The spin-s homogeneous central spin model: exact spectrum and dynamics
Extracted source text · arXiv 1810.03937
\documentclass[amstex,12pt]{article}
\usepackage{amssymb}
\usepackage{amsmath}
\usepackage{IEEEtrantools}
\usepackage{pdflscape}
\usepackage{afterpage}
\usepackage{capt-of}
\usepackage{graphicx}
\usepackage{epstopdf}
\usepackage{subfig}
\usepackage{changepage}
\numberwithin{equation}{section}
% \usepackage{style}
%\draft
\topmargin 0pt \oddsidemargin 0pt
\headheight 0pt \headsep 0pt
\textwidth 6.5in % US paper
\textheight 9in % US paper
% \textwidth 6.1in % A4 paper (old 6.25in)
% \textheight 9.5in % A4 paper
\marginparwidth .875in
\parskip 5pt plus 1pt \jot = 1.5ex
% \makeatletter
% \def\numberbysection{\@addtoreset{equation}{section}
% \def\theequation{\thesection.\arabic{equation}}}
% \makeatother
% \def\baselinestretch{1.2}
% \numberbysection
% GENERAL DEFINITIONS
\newcommand{\be}{\begin{equation}}
\newcommand{\ee}{\end{equation}}
\newcommand{\bea}{\begin{eqnarray}}
\newcommand{\eea}{\end{eqnarray}}
%\newcommand{\ba}{\begin{align}}
%\newcommand{\ea}{\end{align}}
\newcommand{\non}{\nonumber}
\newcommand{\id}{\mathbb{I}}
\newcommand{\C}{\mathbb{C}}
\newcommand{\tr}{\mathop{\rm tr}\nolimits}
\newcommand{\diag}{\mathop{\rm diag}\nolimits}
\newcommand{\uh}{\frac{u}{2}}
\newcommand{\half}{\frac{1}{2}}
\newcommand{\mk}{\mathsf{k}}
\newcommand{\mA}{\mathcal{A}}
\newcommand{\mB}{\mathcal{B}}
\newcommand{\mC}{\mathcal{C}}
\newcommand{\mD}{\mathcal{D}}
\newcommand{\mF}{\mathcal{F}}
\newcommand{\mG}{\mathcal{G}}
\newcommand{\refs}{|0\rangle}
\usepackage[usenames,dvipsnames]{color}
\newcommand{\rodrigo}[1]{$\framebox{\tiny r}$\ \textbf{\texttt{{\color{red}\footnotesize#1}}}}
\newcommand{\nepo}[1]{$\framebox{\tiny n}$\ \textbf{\texttt{{\color{red}\footnotesize#1}}}}
\begin{document}
\begin{titlepage}
\strut\hfill UMTG--297
\vspace{.5in}
\begin{center}
\LARGE The spin-$s$ homogeneous central spin model:\\
exact spectrum and dynamics\\
\vspace{1in}
\large
Rafael I. Nepomechie \footnote{Physics Department,
P.O. Box 248046, University of Miami, Coral Gables, FL 33124 USA, nepomechie@miami.edu}
and
Xi-Wen Guan \footnote{State Key Laboratory of Magnetic Resonance and Atomic and Molecular Physics,
Wuhan Institute of Physics and Mathematics, Chinese Academy of
Sciences, Wuhan 430071, China, xwe105@wipm.ac.cn}
$\! {}^{, }$\footnote{Center for Cold Atom Physics, Chinese Academy of Sciences, Wuhan 430071, China}
$\! {}^{, }$\footnote{Department of Theoretical Physics, Research School of Physics and Engineering,
Australian National University, Canberra ACT 0200, Australia}
\\[0.8in]
\end{center}
\vspace{.5in}
\begin{abstract}
We consider the problem of a central spin with arbitrary spin $s$ that
interacts pairwise and uniformly with a bath of $N$ spins with
$s=1/2$. We present two approaches for determining the exact spectrum
of this model, one based on properties of $SU(2)$, and the other based
on integrability. We also analyze the exact time evolution of a spin
coherent state, and compute the time evolution of various quantities
of physical interest, including the entanglement entropy, spin
polarization and Loschmidt echo.
\end{abstract}
\end{titlepage}
\setcounter{footnote}{0}
\section{Introduction}\label{sec:intro}
The (spin-$\frac{1}{2})$ central spin model is a simple quantum mechanical
model of a central spin $\vec s_{0}$ interacting pairwise with a
``bath'' of $N$ surrounding spins $\vec s_{1}, \ldots, \vec s_{N}$, with
the Hamiltonian
\be
H^{(\frac{1}{2})}_{\rm{inhom}} = B s^{z}_{0} + 2 \sum_{j=1}^{N} A_{j} \vec s_{0} \cdot \vec s_{j}
\,,
\ee
where $B$ and $A_{j}$ are real constants.
This is a special case of the
Richardson-Gaudin model, which was formulated long ago
\cite{Richardson:1963a, Richardson:1964a, Gaudin:1976sv, Gaudin:1983}. Nevertheless,
this model is the focus of renewed attention due to its interesting
new applications, such as quantum dots (see e.g. \cite{Khaetskii2002, Merkulov2002, Khaetskii2003, Erlingsson2004,
Braun2005, Chen2007, Bortz2007, Bortz2010} and references therein).
The homogeneous case $A_{j} = A$ is simple enough to allow for
analytical analysis of spin dynamics, yet exhibits rich phenomena such as quantum collapse and
revival \cite{Dooley2013, Guan:2018a, Guan:2018b}. Moreover, it can be mapped
\cite{Dooley2013, Guan:2018b} to a generalization of the
Jaynes-Cummings model \cite{Jaynes1963, Scully:1997, Bogoliubov2013}
of quantum nonlinear optics.
We consider here a generalization of this model, whereby the central
spin $\vec S_{0}$ has spin $s$. We focus primarily on the homogeneous
case
\be
H^{(s)} = B S^{z}_{0} + 2 A \sum_{j=1}^{N} \vec S_{0} \cdot \vec
s_{j}\,, \qquad s= \tfrac{1}{2}, 1, \ldots \,.
\label{spinShomogH}
\ee
This model, which to our knowledge
has not been considered before, has the potential to be realized
experimentally, and exhibits interesting quantum dynamics. It can
also be mapped to a generalization of the Tavis-Cummings model
\cite{Tavis1968, Scully:1997} of quantum nonlinear
optics, see e.g. \cite{Knap2010, Youssef2010, Agarwal2012, Bogoliubov2013}
for recent work.
The outline of this paper is as follows. In Section
\ref{sec:spectrum}, we discuss two approaches for determining the
exact spectrum of the model, one based on properties of $SU(2)$, and
the other based on integrability. We also briefly describe the
connection with the Tavis-Cummings model. In Section 3, we analyze the
exact time evolution of a spin coherent state. Readers who are
primarily interested in spin dynamics can jump directly to this
section, as it is largely independent of the previous one. We
conclude with a brief summary. Appendix \ref{sec:coeffs} contains the
derivation of a key step in our analysis of spin dynamics, while
Appendix \ref{sec:identities} contains the derivation of some useful
identities.
\section{Exact spectrum}\label{sec:spectrum}
We briefly discuss here two approaches for determining the spectrum of
the spin-$s$ homogeneous central spin model (\ref{spinShomogH}). The
first approach exploits the partial $SU(2)$ symmetry of the problem,
and in principle can give the entire spectrum. The second approach
exploits the integrability of the model; however, it remains to be understood
whether the solution obtained in this way can give the full spectrum.
\subsection{$SU(2)$-based approach}\label{sec:su2}
The Hamiltonian (\ref{spinShomogH}) can evidently be rewritten as
\be
H^{(s)} = B S^{z}_{0} + 2 A \vec S_{0} \cdot \vec J\,,
\label{spinShomogH2}
\ee
where
\be
\vec J = \sum_{j=1}^{N} \vec s_{j}
\ee
is the total spin of the bath. This Hamiltonian clearly has the properties
\be
\left[ H^{(s)}\,, \vec{J}^{2} \right] = 0\,, \qquad \left[ H^{(s)}\,, \vec{S}_{0}^{2} \right] = 0\,,
\label{properties}
\ee
as well as the $U(1)$ symmetry
\be
\left[ H^{(s)}\,, {\cal S}^{z} \right] = 0\,, \qquad {\cal S}^{z} = J^{z} +
S^{z}_{0}\,,
\label{U1symmetry}
\ee
but (for $B \ne 0$) does {\em not} have the full $SU(2)$ symmetry. In view of (\ref{properties}) and (\ref{U1symmetry}), we look for
simultaneous eigenstates of $H$, ${\cal S}^{z}$, $\vec{S}_{0}^{2}$, $\vec{J}^{2}$,
\begin{align}
H^{(s)} |E, m, s, j \rangle &= E |E, m, s, j\rangle \,, \non \\
{\cal S}^{z} |E, m, s, j\rangle &= m |E, m, s, j\rangle \,, \non \\
\vec{S}_{0}^{2} |E, m, s, j\rangle &= s(s+1) |E, m, s, j\rangle \,, \non \\
\vec{J}^{2} |E, m, s, j\rangle &= j(j+1) |E, m, s, j\rangle \,.
\end{align}
We can expand these states in the standard orthonormal spin basis as follows
\be
|E, m, s, j\rangle = \sum_{m_{s}=-s}^{s}\sum_{m_{j}=-j}^{j}
w_{m_{s}, m_{j}}^{(E,m,s,j)} \delta_{m_{s}+m_{j},m}|s,
m_{s}\rangle\otimes |j, m_{j}\rangle\,,
\label{expansion}
\ee
where the coefficients $w_{m_{s}, m_{j}}^{(E,m,s,j)}$ are still to be
determined. Note that the allowed values of $j$ (spin of the bath) are
\be
j= \left\{
\begin{array}{ll}
0, 1, \ldots, \frac{N}{2} & \mbox{ for } N = \mbox{ even } \\ \\
\frac{1}{2}, \frac{3}{2}, \ldots, \frac{N}{2} & \mbox{ for } N = \mbox{ odd }
\end{array} \right. \,,
\label{jvalues}
\ee
while the allowed values of $m$ are given by
\be
-j-s \le m \le j+ s \,.
\label{mvalues}
\ee
We now act on (\ref{expansion}) with the Hamiltonian (\ref{spinShomogH2}) in the form
\be
H^{(s)} = B S^{z}_{0} + A \left( S^{+}_{0} J^{-} + S^{-}_{0} J^{+} + 2 S^{z}_{0}
J^{z}\right) \,,
\label{spinShomogH3}
\ee
where $S^{\pm} = S^{x} \pm i S^{y}$ (and similarly for $J^{\pm}$).
Using the familiar $SU(2)$ raising/lowering formulas
\be
S^{\pm}| s, m\rangle
= \sqrt{(s \mp m)(s \pm m +1)} | s, m\pm 1 \rangle \,,
\label{raisinglowering}
\ee
we obtain
\begin{align}
H^{(s)} |E, m, s, j\rangle &= \sideset{}{'}\sum_{m_{s}, m_{j}} w_{m_{s},
m_{j}}^{(E,m,s,j)} \Bigg\{
\mu_{m_{s}, m_{j}} |s, m_{s}\rangle\otimes |j, m_{j}\rangle \non\\
& + \nu_{m_{s}+1, m_{j}-1}^{(s,j)} |s, m_{s}+1\rangle\otimes |j, m_{j}-1\rangle
+ \nu_{-m_{s}+1, -m_{j}-1}^{(s,j)} |s, m_{s}-1\rangle\otimes |j,
m_{j}+1\rangle \Bigg\}
\,,
\end{align}
where the summation is constrained by $m_{s}+m_{j}=m$, and the
coefficients are defined by
\begin{align}
\mu_{m_{s}, m_{j}} &= B\, m_{s} + 2 A\, m_{s}\, m_{j} \,, \non \\
\nu_{m_{s}, m_{j}}^{(s,j)} &= A \sqrt{(s-m_{s}+1)(s+m_{s})(j+
m_{j}+1)(j-m_{j}) } \,.
\end{align}
Finally, using the orthonormality of the basis, we arrive at an
eigenvalue relation for the energy $E$ and the corresponding
coefficients $w_{m_{s}, m_{j}}^{(E,m,s,j)}$,
\begin{align}
& \mu_{m_{s}, m_{j}}\, w_{m_{s}, m_{j}}^{(E,m,s,j)} +
\nu_{m_{s}, m_{j}}^{(s,j)} \, w_{m_{s}-1, m_{j}+1}^{(E,m,s,j)} +
\nu_{-m_{s}, -m_{j}}^{(s,j)} \, w_{m_{s}+1, m_{j}-1}^{(E,m,s,j)} = E\,
w_{m_{s}, m_{j}}^{(E,m,s,j)} \,, \non\\
& m_{s} = -s, \ldots, s\,, \qquad m_{j} = -j, \ldots, j\,, \qquad
m_{s}+m_{j}=m \,.
\label{eigw}
\end{align}
Letting $d(m, s, j)$ denote the number of allowed values of $(m_{s}, m_{j})$
satisfying $m_{s}+m_{j}=m$, we see that (\ref{eigw}) entails
diagonalizing a $d(m, s, j) \times d(m, s, j)$ matrix.
The eigenvalue relation (\ref{eigw}) for the energy is the main
result of this subsection. Given values of $j$ and $m$ (constrained
by (\ref{jvalues}) and (\ref{mvalues}), respectively), the
eigenvalue problem (\ref{eigw}) can in principle be solved for the
corresponding energies. The advantage of this approach over
brute-force diagonalization of the full Hamiltonian (\ref{spinShomogH2})
is that the matrices to be diagonalized are much smaller. Of course, as
the bath size $N$ becomes large, the number of possible values for $j$ and $m$
also becomes large.
As a simple example, we present in Table \ref{table:s1N2} the energies
that are computed in this way for the case $s=1\,, N=2$. (The final
two columns of the table are explained in the following subsection.)
\begin{table}[htb]
\small
\centering
\begin{tabular}{|c|c|c|c|c|}
\hline
$j$ & $m$ & $E$ & $M$ & $\{ v_{a} \}$\\
\hline
1 & 2 & 1.5 & 0 & -\\
1 & 1 & -0.780776 & 1 & -0.438447\\
1 & 1 & 1.28078 & 1 & -4.56155\\
1 & 0 & -2.14854 & 2 & -0.351465 $\pm 0.262932 i$ \\
1 & 0 & -0.893401 & 2 & -2.71954, -0.493659 \\
1 & 0 & 1.04194 & 2 & -3.54194 $\pm 1.70866 i$ \\
1 & -1 & -1.28078 & 3 & -0.612504, -1.41297 $\pm 0.681796 i$\\
1 & -1 & 0.780776 & 3 & -3.16744, -2.19705 $\pm 2.46224 i$ \\
1 & -2 & 0.5 & 4 & -2.26566 $\pm 0.850941 i$, -0.734342 $\pm
2.43893 i$\\
0 & 1 & 0.5 & & \\
0 & 0 & 0 & & \\
0 & -1 & - 0.5 & & \\
\hline
\end{tabular}
\caption{\small The energies ($E$) and Bethe roots $\{ v_{1},
\ldots, v_{M} \}$ for the Hamiltonian (\ref{spinShomogH}) with
$s=1, N=2, A = B = 0.5$. The total number of levels is $(2s+1) 2^{N} = 12$.}
\label{table:s1N2}
\end{table}
\subsection{Bethe ansatz approach}
The spin-$s$ central spin model (\ref{spinShomogH}) is integrable, as is the
original model with $s=\frac{1}{2}$. We present here its Bethe ansatz
solution, but only sketch the derivation, since it is similar to the one
for the spin-$\frac{1}{2}$ case, see e.g. \cite{Sklyanin:1987ih,
Hikami:1992np, Zhou:2001}. For pedagogical reasons,
we in fact work for the inhomogeneous model (which is also integrable), and consider the
homogeneous limit only at the end.
\subsubsection{Inhomogeneous case}
Our starting point is the $SU(2)$-invariant $(\frac{1}{2},s)$ R-matrix on ${\cal C}^{2}
\otimes {\cal C}^{2s+1}$ (see, e.g.
\cite{Babujian:1983ae}),
\be
R^{(\frac{1}{2},s)}(u)=\frac{1}{u+(s+\frac{1}{2})\eta}
\left((u+\frac{\eta}{2})\id + 2\eta \vec s \overset{\cdot}{\otimes}
\vec S \right) \,, \qquad s = \tfrac{1}{2}\,, 1\,, \ldots\,,
\ee
where (as above) $\vec s$ and $\vec S$ denote the spin operators for
spin-$\frac{1}{2}$ and spin-$s$, respectively. This R-matrix
is a solution of the Yang-Baxter equation on ${\cal C}^{2}
\otimes {\cal C}^{2} \otimes {\cal C}^{2s+1}$
\be
R^{(\frac{1}{2}, \frac{1}{2})}_{12}(u-v)\, R^{(\frac{1}{2}, s)}_{13}(u)\, R^{(\frac{1}{2}, s)}_{23}(v) =
R^{(\frac{1}{2}, s)}_{23}(v)\, R^{(\frac{1}{2}, s)}_{13}(u)\,
R^{(\frac{1}{2}, \frac{1}{2})}_{12}(u-v) \,.
\label{YBE}
\ee
We introduce the following monodromy matrix
\be
T_{a}(u) = G_{a}\, R^{(\frac{1}{2},\frac{1}{2})}_{a
N}(u-\epsilon_{N})\, \ldots R^{(\frac{1}{2},\frac{1}{2})}_{a
1}(u-\epsilon_{1})\, R^{(\frac{1}{2},s)}_{a 0}(u-\epsilon_{0})\,,
\qquad G = e^{\eta s B \sigma^{z}} \,,
\ee
where the ``auxiliary'' space (denoted by $a$) has dimension 2, and there
are $N+1$ ``quantum'' spaces (denoted by $0, 1, \ldots, N$): the
$0^{th}$ quantum space (corresponding to the central spin) has
dimension $2s+1$, while all the others (corresponding to the bath)
have dimension 2. Notice that there are arbitrary inhomogeneities
$\epsilon_{0}, \ldots, \epsilon_{N}$ associated with each of the
quantum spaces. Finally, note that there is a diagonal twist, encoded
by the matrix $G$, which is responsible for breaking $SU(2)$ down to
$U(1)$.
The transfer matrix $t(u)$, obtained by tracing the monodromy matrix
over the auxiliary space
\be
t(u) = \tr_{a} T_{a}(u) \,,
\label{transfer}
\ee
satisfies the important commutativity property
\be
\left[ t(u) \,, t(v) \right] = 0
\ee
owing to the Yang-Baxter equation (\ref{YBE}).
The basic idea, following e.g. \cite{Sklyanin:1987ih, Hikami:1992np, Zhou:2001},
is to evaluate the transfer matrix at $u=\epsilon_{0}$ and expand in terms
of $\eta$. We find that the inhomogeneous spin-$s$ Hamiltonian
\be
H^{(s)}_{\rm{inhom}} = B S^{z}_{0} + \frac{1}{s}\sum_{j=1}^{N}
\frac{1}{\epsilon_{0}-\epsilon_{j}} \vec S_{0} \cdot \vec s_{j}
\label{spinSinhomogH}
\ee
can indeed be obtained in this way
\be
H^{(s)}_{\rm{inhom}} = \frac{(2s+1)}{4s}
\frac{d}{d\eta}t(\epsilon_{0})\Big\vert_{\eta=0} +
\frac{1}{4s}\sum_{j=1}^{N}\frac{1}{\epsilon_{0}-\epsilon_{j}} \id \,.
\label{Hfromtransf}
\ee
Using algebraic Bethe ansatz (ABA), we find that the eigenvalues of the
transfer matrix (\ref{transfer}) are given by
\be
\Lambda(u) = e^{\eta s B} \prod_{a=1}^{M}
\frac{u-v_{a}-\eta}{u-v_{a}} +
e^{-\eta s B}
\left(\frac{u-\epsilon_{0}+(\frac{1}{2}-s)\eta}
{u-\epsilon_{0}+(\frac{1}{2}+s)\eta}\right)
\prod_{j=1}^{N}\frac{u-\epsilon_{j}}{u-\epsilon_{j}+\eta}
\prod_{a=1}^{M}
\frac{u-v_{a}+\eta}{u-v_{a}} \,,
\ee
where $\{v_{1}\,, \ldots\,, v_{M} \}$ are solutions of the Bethe equations
\be
\left(\frac{v_{a}-\epsilon_{0}+(\frac{1}{2}-s)\eta}
{v_{a}-\epsilon_{0}+(\frac{1}{2}+s)\eta}\right)
\prod_{j=1}^{N}\frac{v_{a}-\epsilon_{j}}{v_{a}-\epsilon_{j}+\eta}
= e^{2\eta s B} \prod_{b=1, b\ne a}^{M}
\frac{v_{a}-v_{b}-\eta}{v_{a}-v_{b}+\eta}\,, \quad a = 1, \ldots, M
\,.
\label{BA1}
\ee
It follows from (\ref{Hfromtransf}) that the eigenvalues of
$H^{(s)}_{\rm{inhom}} $ are given by
\begin{align}
E_{\rm{inhom}} &= \frac{(2s+1)}{4s}
\frac{d}{d\eta}\Lambda(\epsilon_{0})\Big\vert_{\eta=0} +
\frac{1}{4s}\sum_{j=1}^{N}\frac{1}{\epsilon_{0}-\epsilon_{j}} \non \\
& = s B +
\frac{1}{2}\sum_{j=1}^{N}\frac{1}{\epsilon_{0}-\epsilon_{j}} +
\sum_{a=1}^{M}\frac{1}{v_{a}-\epsilon_{0}}\,.
\label{energyinhom}
\end{align}
Moreover, expanding the Bethe equations (\ref{BA1}) in $\eta$, we
arrive at the Bethe equations for the inhomogeneous spin-$s$ central spin model
\be
-2 s B - \frac{2s}{v_{a} - \epsilon_{0}}
- \sum_{j=1}^{N} \frac{1}{v_{a} - \epsilon_{j}}
+ 2 \sum_{b=1, b\ne a}^{M} \frac{1}{v_{a}-v_{b}} = 0\,, \quad a = 1, \ldots, M \,.
\label{BAinhom}
\ee
The allowed values of $M$ can be deduced from the formula for the
${\cal S}^{z}$ eigenvalues
\be
m = \frac{N}{2} + s - M
\ee
which also follows from the ABA, together with the ranges
(\ref{jvalues}) and (\ref{mvalues}). We conclude that the number of
Bethe roots can be
\be
M = 0, 1, \ldots, N+2s\,.
\label{Mvalues}
\ee
The energy formula (\ref{energyinhom}) and the corresponding Bethe
equations (\ref{BAinhom}) constitute our main results for the
inhomogeneous spin-$s$ central spin model (\ref{spinSinhomogH}). For
$s=\frac{1}{2}$, the well-known results are recovered.
For distinct values of the inhomogeneities ($\epsilon_{j} \ne
\epsilon_{k}$ for $j \ne k$) and $B \ne 0$, the spectrum is
nondegenerate, and the Bethe ansatz solution appears to be complete.
(We have checked this numerically for small values of $s$ and $N$; a proof
for the case $s=\frac{1}{2}$ can be found in \cite{Links:2017}.) We
conjecture that the number of solutions of these Bethe equations for
given values of $s, N, M$ is given by
\be
{\cal N}(s, N, M) = \sum_{k=0}^{\lfloor s \rfloor} (-1)^{k}
{2s-k\choose k} {N+2s-2k\choose m-k}\,.
\ee
We have also checked this result numerically for small values of $s$ and $N$,
and one can verify that indeed
\be
\sum_{M=0}^{N+2s} {\cal N}(s, N, M) = (2s+1) 2^{N} \,,
\ee
thereby accounting for all the levels of the system.
\subsubsection{Homogeneous case}
Let us finally return to the homogeneous Hamiltonian
(\ref{spinShomogH}). Comparing with the inhomogeneous one
(\ref{spinSinhomogH}), we see that the homogeneous case corresponds
to setting
\be
\epsilon_{1} = \ldots = \epsilon_{N} \equiv \epsilon\,, \qquad
A = \frac{1}{2s(\epsilon_{0} - \epsilon)} \,.
\ee
For these values of parameters, the formulas for the energy (\ref{energyinhom}) and the Bethe
equations (\ref{BAinhom}) reduce to
\be
E = s (B + N A) + \sum_{a=1}^{M}\frac{1}{v_{a}}
\label{EBAhomg}
\ee
and
\be
-2 s B - \frac{2s}{v_{a}}
- \frac{N}{v_{a} + \frac{1}{2 s A}}
+ 2 \sum_{b=1, b\ne a}^{M} \frac{1}{v_{a}-v_{b}} = 0\,, \quad a = 1, \ldots, M \,,
\label{BAhom}
\ee
respectively, after performing a shift $v_{a} \mapsto v_{a}
+ \epsilon_{0}$ of all the Bethe roots. Eqs. (\ref{EBAhomg}) and
(\ref{BAhom}) constitute our main results for the Bethe ansatz
solution of the homogeneous model.
A simple example with $s=1$ and $N=2$ is presented in Table \ref{table:s1N2}. Note that
all the levels with $j = \frac{N}{2}$ are accounted for, but
not those with $j < \frac{N}{2}$. This appears to be a general
feature of the Bethe ansatz solution for the homogeneous
model. Our preliminary investigations indicate that this
difficulty (which is present already for the spin-$\frac{1}{2}$ case)
is due to the necessity of correctly taking into account so-called singular solutions
(e.g., $v=0$).\footnote{For a recent discussion of singular solutions in
the context of the periodic Heisenberg chain, see \cite{Hao:2013jqa}.} We hope
to investigate this matter further in the future.
A nice feature of the Bethe equations (\ref{BAhom}) is that numerical solutions
can be readily found even for relatively large values of $N$ and $M$.
The trick (see e.g. \cite{Gaudin:1971zza, Shastry:2001, Pan:2011, Marquette:2012}
begins with the observation that the q-polynomial
\be
q(u) = \prod_{a=1}^{M}(u-v_{a}) = u^{M} + O(u^{M-1})
\ee
satisfies
\be
\frac{q''(v_{a})}{q'(v_{a})} = 2 \sum_{b=1, b\ne a}^{M}
\frac{1}{v_{a}-v_{b}} \,.
\ee
The Bethe equations (\ref{BAhom}) can therefore be rewritten in the
form
\be
P(v_{a}) = 0\,, \qquad a = 1, \ldots, M \,,
\ee
where
\be
P(u) = u (u + \frac{1}{2 s A}) q''(u)
- 2 s B u (u + \frac{1}{2 s A}) q'(u) - 2 s (u + \frac{1}{2 s A}) q'(u) - N u q'(u) \,.
\ee
One next observes that $P(u)$ is a polynomial in $u$ of degree $M+1$, which
has all $M$ zeros of the polynomial $q(u)$. Therefore, $P(u)/q(u)$ is
a polynomial of degree 1, i.e.
\be
P(u) = (a + b\, u)\, q(u) \,,
\label{Pq}
\ee
where $b= - 2 s B M$ follows from the asymptotic behavior $u
\rightarrow \infty$. Setting
\be
q(u) = \sum_{k=0}^{M} q_{k} u^{k} \,,
\ee
one can obtain from (\ref{Pq}) (by setting the coefficients of
$u^{0}\,, \ldots\,, u^{M+1}$ equal to zero)
a set of $M+2$ equations for the $M+2$ unknowns $a,
q_{0}, \ldots, q_{M}$, which can be readily solved numerically even for
relatively large values of $N$ and $M$; one can then
determine the zeros of $q(u)$, which are the sought-after Bethe
roots.\footnote{Unfortunately, this trick is not nearly as effective
for the inhomogeneous case.}
An example with $N=60, s=1, M=31$ is shown in Figure
\ref{fig:BetherootsN60M31}.
\begin{figure}[htb]
\centering
{\includegraphics[width=5.5cm]{N60M31}}
\caption{A set of solutions of the Bethe equations (\ref{BAhom})
plotted in the complex $v$ plane for
$N=60, s=1, M=31, A=B=0.5$, with corresponding energy $E=30.004$.}
\label{fig:BetherootsN60M31}
\end{figure}
\subsection{Connection with the Tavis-Cummings model}
We briefly note here a mapping of the spin-$s$ homogeneous central
spin model to the Tavis-Cummings model \cite{Tavis1968, Scully:1997}.
This mapping relies, as in the spin-$\frac{1}{2}$
case \cite{Dooley2013, Guan:2018b}, on the Holstein-Primakoff
transformation \cite{Holstein1940}:
\be
J^{+} = \sqrt{N} a^{\dagger} \sqrt{1-\frac{a^{\dagger} a}{N}} \,,
\qquad J^{-} = \left( J^{+} \right)^{\dagger}
=\sqrt{N}\sqrt{1-\frac{a^{\dagger} a}{N}}\, a \,, \qquad
J^{z} = a^{\dagger} a - \frac{N}{2} \,,
\label{HP}
\ee
where $\left[ a \,, a^{\dagger} \right] = 1$. Indeed, let us now consider
an anisotropic generalization of the Hamiltonian (\ref{spinShomogH3}),
\be
H^{(s)} = B S^{z}_{0} + A \left( S^{+}_{0} J^{-} + S^{-}_{0} J^{+}
\right) + 2 \Delta\, S^{z}_{0} J^{z} \,,
\ee
where $\Delta$ is the anisotropy parameter.\footnote{For $\Delta=A$, this
model evidently reduces to the isotropic model (\ref{spinShomogH3}). For simplicity, we focus in this paper
primarily on the isotropic case; however, it is possible to generalize
all of the results to the anisotropic case.} Applying the
transformation (\ref{HP}), and letting $N \rightarrow \infty$, we obtain
\be
H^{(s)} \sim (B- \Delta N) S^{z}_{0} + A \sqrt{N} \left(S^{+}_{0}\, a +
S^{-}_{0}\, a^{\dagger} \right) + 2 \Delta S^{z}_{0}\, a^{\dagger} a
\,.
\ee
The case $\Delta = 0$ reduces to the Tavis-Cummings model, while the
isotropic case $\Delta=A$ corresponds to a generalization of the Tavis-Cummings model.
\section{Exact dynamics}\label{sec:dynamics}
We turn now to the question of how states evolve in time
for the spin-$s$ homogeneous central spin model (\ref{spinShomogH}),
(\ref{spinShomogH2}). For
a generic initial state, this problem appears to pose a formidable challenge,
even for a model as simple as this one. However, if the initial
state is an eigenstate of $\vec{S}_{0}^{2}$ and
$\vec{J}^{2}$, then the symmetries (\ref{properties}) and
(\ref{U1symmetry}) significantly constrain the possible intermediate
states, and the problem becomes tractable yet nevertheless remains nontrivial.
\subsection{Time evolution of a spin coherent state}
Following \cite{Dooley2013, Guan:2018a}, we assume that the bath is
initially in a so-called spin coherent state \cite{Radcliffe1971, Arecchi1972}
\be
|\theta \rangle = \bigotimes_{j=1}^{N}\left[\cos(\tfrac{\theta}{2})
|\tfrac{1}{2}, \tfrac{1}{2}\rangle_{j} + \sin(\tfrac{\theta}{2})
|\tfrac{1}{2}, -\tfrac{1}{2}\rangle_{j} \right]\,,
\label{spincoherentstate}
\ee
which indeed is an eigenstate of $\vec{J}^{2}$ with $j=\frac{N}{2}$.
Moreover, we assume that the central spin is initially ``up'', i.e.
in the state $|s, s \rangle_{0}$. Thus, the initial state of the
system is
\be
|\Psi(0)\rangle = |s, s \rangle \otimes |\theta \rangle \,,
\label{initialstate}
\ee
and our task is to determine its time evolution
\be
|\Psi(t)\rangle = e^{-i H^{(s)} t} |\Psi(0)\rangle \,.
\ee
Expressing the spin coherent state in terms of so-called Dicke states
as in \cite{Dooley2013, Guan:2018b}
\be
|\theta \rangle = \sum_{n=0}^{N} \sqrt{{N \choose n}}
\cos^{N-n}(\tfrac{\theta}{2}) \sin^{n}(\tfrac{\theta}{2})
|n\rangle\,,
\ee
where $|n\rangle \equiv |\frac{N}{2},\frac{N}{2} -n \rangle$, the
problem reduces to computing
\begin{align}
|\Psi(t)\rangle &= \sum_{k=0}^{\infty} \frac{1}{k!} (-i H^{(s)} t)^{k}
|\Psi(0)\rangle \non \\
&= \sum_{n=0}^{N} \sqrt{{N \choose n}}
\cos^{N-n}(\tfrac{\theta}{2}) \sin^{n}(\tfrac{\theta}{2})
\sum_{k=0}^{\infty} \frac{(-i t)^{k}}{k!} (H^{(s)})^{k} \left(|s, s \rangle
\otimes |n\rangle \right) \,.
\label{intermed}
\end{align}
Due to the $U(1)$ symmetry (\ref{U1symmetry}), we know that
\be
(H^{(s)})^{k} \left(|s, s \rangle \otimes |n\rangle \right) =
\sum_{j=0}^{2s} h^{(s, k)}_{j}
|s,s-j \rangle \otimes |n-j\rangle \,,
\label{Hk}
\ee
where the coefficients $h^{(s, k)}_{j}$ are still unknown.
We show in Appendix \ref{sec:coeffs} that these coefficients are given by
\be
h_{j}^{(s,k)} = \sum_{l=0}^{2s} c_{j,l}^{(s)}(n)\,
\left(\omega_{l}^{(s)}(n)\right)^{k} \,, \qquad j = 0, 1, \ldots, 2s
\,,
\label{hjsltnmain}
\ee
see (\ref{hjsltn}), and we provide a straightforward recipe for
numerically computing $c_{j,l}^{(s)}(n)$ and $\omega_{l}^{(s)}(n)$.
Substituting the results (\ref{Hk}) and (\ref{hjsltnmain}) back into
(\ref{intermed}), we obtain
\begin{align}
|\Psi(t)\rangle &= \sum_{n=0}^{N} \sqrt{{N \choose n}}
\cos^{N-n}(\tfrac{\theta}{2}) \sin^{n}(\tfrac{\theta}{2})
\sum_{j=0}^{2s}
\sum_{l=0}^{2s} c_{j,l}^{(s)}(n)\,
\sum_{k=0}^{\infty} \frac{(-i t)^{k}}{k!}
\left(\omega_{l}^{(s)}(n)\right)^{k}
|s,s-j \rangle \otimes |n-j\rangle \non \\
&= \sum_{n=0}^{N} \sqrt{{N \choose n}}
\cos^{N-n}(\tfrac{\theta}{2}) \sin^{n}(\tfrac{\theta}{2})
\sum_{j, l =0}^{2s}
c_{j,l}^{(s)}(n)\,
e^{-i t \omega_{l}^{(s)}(n)}
|s,s-j \rangle \otimes |n-j\rangle \,.
\label{evolvedstate}
\end{align}
For given values of $s$, $N$ and $n$, the
frequencies $\{\omega_{0}^{(s)}(n)\,, \ldots\,,
\omega_{2s}^{(s)}(n) \}$ are the energies in the sector (see Sec. \ref{sec:su2})
with $j=\frac{N}{2}$ and $m = \frac{N}{2} - n + s$, which has (at
most) dimension $2s+1$.
An important check on this result is the verification of unitarity $\langle
\Psi(t) | \Psi(t) \rangle = 1$. Using the
orthonormality of the basis and the fact that
$c_{j,l}^{(s)}(n)$ and $\omega_{l}^{(s)}(n)$ are real, we obtain
\be
\langle \Psi(t) | \Psi(t) \rangle =
\sum_{n=0}^{N} {N \choose n}
\cos^{2(N-n)}(\tfrac{\theta}{2}) \sin^{2n}(\tfrac{\theta}{2})
\sum_{j, l, l' =0}^{2s}
c_{j,l}^{(s)}(n)\, c_{j,l'}^{(s)}(n)\,
e^{-i t \left(\omega_{l}^{(s)}(n) - \omega_{l'}^{(s)}(n) \right)} \,.
\ee
With the help of the identities (see (\ref{id1}) and (\ref{id5}))
\be
\sum_{j=0}^{2s} c_{j,l}^{(s)}(n) \, c_{j,l'}^{(s)}(n) =
c_{0,l}^{(s)}(n)\,
\delta_{l, l'}\,, \qquad \sum_{l=0}^{2s} c_{0,l}^{(s)}(n) = 1\,,
\label{identities}
\ee
we see that unitarity is indeed preserved.
The result (\ref{evolvedstate}) is one of the main results of our
paper. We emphasize that this is an exact result. Note that the infinite sum over $k$ in (\ref{intermed}) has
been effectuated, leaving only finite sums to be performed (assuming
that $N$ and $s$ are finite). For the special case
$s=\frac{1}{2}$, the result (\ref{evolvedstate}) reduces to a corresponding result in
\cite{Guan:2018b}. In the remaining part of this section, we use
(\ref{evolvedstate}) to compute the time evolution of various quantities of physical interest.
\subsection{Reduced density matrix}
The reduced density matrix for the central spin $\rho(t)$ is defined by
\be
\rho(t) = \sum_{\alpha} \langle \alpha| \Psi(t)\rangle \langle \Psi(t) |
\alpha \rangle \,,
\ee
where the trace is performed by summing over an orthonormal basis of
the bath $({\cal C}^{2})^{\otimes N}$. Making use of the result
(\ref{evolvedstate}) and the fact
\be
\sum_{\alpha} \langle \alpha| n-j\rangle \langle n'-j' |
\alpha \rangle = \delta_{n-j,n'-j' } \,,
\ee
we obtain the $(2s+1) \times (2s+1)$ matrix
\be
\rho(t) = \sum_{j, j' =0}^{2s} \rho_{j j'}(t) |s-j\rangle \langle s-j'|
\,,
\ee
whose matrix elements are given by
\begin{align}
\rho_{j j'}(t) &= \sum_{n, n'=0}^{N}
\sum_{l, l' =0}^{2s}
\sqrt{{N \choose n} {N \choose n'}}
\cos^{2N-n-n'}(\tfrac{\theta}{2}) \sin^{n+n'}(\tfrac{\theta}{2})\,
\delta_{n-j,n'-j'} \non \\
& \qquad \qquad
\times c_{j,l}^{(s)}(n)\, c_{j',l'}^{(s)}(n')\, e^{-i t
\left(\omega_{l}^{(s)}(n) - \omega_{l'}^{(s)}(n') \right)} \,.
\label{rhomat}
\end{align}
The matrix (\ref{rhomat}) is manifestly Hermitian, $\rho^{\dagger}=\rho$.
Knowing the reduced density matrix, we can directly compute
the von Neumann entanglement entropy $S(t)$
\be
S(t) = - \tr \left[ \rho(t) \ln \rho(t) \right] = - \sum_{l=0}^{2s}
\lambda_{l}(t) \ln \lambda_{l}(t) \,,
\label{entropy}
\ee
and the quantum purity $\gamma(t)$
\be
\gamma(t) = \tr \left[ \rho^{2}(t) \right] = \sum_{l=0}^{2s}
\lambda_{l}^{2}(t) \,,
\ee
where $\{ \lambda_{l}(t) \}$ are the eigenvalues of $\rho(t)$.
An example of the von Neumann entanglement entropy $S(t)$ for $s=1$ is presented in
Fig. \ref{fig:entanglement_entropy}. In contrast with the
$s=\frac{1}{2}$ case \cite{Guan:2018b}, here $S(t)$ displays rapid
irregular oscillations except at the collapsed regions. (See also
Fig. \ref{fig:spin_polarization}.)
\begin{figure}[htb]
\centering
{\includegraphics[width=5.5cm]{entanglement_entropy}}
\caption{The von Neumann entanglement entropy (\ref{entropy})
as a function of time for $N=15, s=1, \theta=0.5 \pi, A=B=1.0$.}
\label{fig:entanglement_entropy}
\end{figure}
\subsection{Spin expectation value}
The expectation value of the central spin can be computed using the
reduced density matrix
\be
\vec S(t) = \langle \Psi(t) | \vec S | \Psi(t) \rangle = \tr \left[
\vec S\, \rho(t) \right] \,.
\ee
In particular, the so-called spin polarization is given in terms of the
diagonal elements of (\ref{rhomat})
\be
S^{z}(t) = \sum_{n=0}^{N}
\sum_{j, l, l' =0}^{2s}
{N \choose n}
\cos^{2(N-n)}(\tfrac{\theta}{2}) \sin^{2n}(\tfrac{\theta}{2})\,
(s-j)\, c_{j,l}^{(s)}(n)\, c_{j,l'}^{(s)}(n)\, e^{-i t
\left(\omega_{l}^{(s)}(n) - \omega_{l'}^{(s)}(n) \right)} \,.
\label{spinpolarization}
\ee
An example of the spin polarization for $s=1$ is presented in
Fig. \ref{fig:spin_polarization}. In contrast with the
$s=\frac{1}{2}$ case \cite{Guan:2018b}, the duration of collapses
quickly tend to zero as $t$ increases. This indicates
that the bath is mixing all the states of the central spin. Note also
that, for spin $s$, the revival regions consist of up to $2s+1$
revival peaks.
\begin{figure}[htb]
\centering
{\includegraphics[width=5.5cm]{spin_polarization}}
\caption{The spin polarization of the central spin (\ref{spinpolarization})
as a function of time for $N=15, s=1, \theta=0.5 \pi, A=B=1.0$.}
\label{fig:spin_polarization}
\end{figure}
On the other hand, the so-called coherent factor $S^{-}(t) =
\langle \Psi(t) | S^{-} | \Psi(t) \rangle$ involves
off-diagonal elements of the reduced density matrix, an example of
which is presented in Fig. \ref{fig:coherentfactor}.
\begin{figure}[htb]
\centering
{\includegraphics[width=5.5cm]{coherent_factor}}
\caption{The squared norm coherent factor
of the central spin $|S^{-}(t)|^{2}$ as a function of time for $N=15,
s=1, \theta=0.5 \pi, A=B=1.0$.}
\label{fig:coherentfactor}
\end{figure}
\subsection{Loschmidt echo}
The Loschmidt echo can also be readily computed using the result
(\ref{evolvedstate})
\be
L(t) \equiv |\langle \Psi(0)| \Psi(t) \rangle |^{2}
= \left\vert \sum_{n=0}^{N} {N \choose n}
\cos^{2(N-n)}(\tfrac{\theta}{2}) \sin^{2n}(\tfrac{\theta}{2})
\sum_{l =0}^{2s}
c_{0,l}^{(s)}(n)\,
e^{-i t \omega_{l}^{(s)}(n) } \right\vert^{2}\,.
\label{Loschmidt}
\ee
The fact that $L(0)=1$ is ensured by the second identity
in (\ref{identities}).
An example of the Loschmidt echo $L(t)$ for $s=1$ is presented in
Fig. \ref{fig:Loschmidt}. In contrast with the
$s=\frac{1}{2}$ case \cite{Guan:2018b}, here $L(t)$ displays rapid
irregular oscillations. Notice the appearance of points when
$L(t)=0$, at which times the states are completely orthogonal to the
initial state.
\begin{figure}[htb]
\centering
{\includegraphics[width=5.5cm]{Loschmidt}}
\caption{The Loschmidt echo (\ref{Loschmidt})
as a function of time for $N=15, s=1, \theta=0.5 \pi, A=B=1.0$.}
\label{fig:Loschmidt}
\end{figure}
\section{Conclusions}\label{sec:conclusion}
We have presented two approaches for determining the exact spectrum
of the spin-$s$ homogeneous central spin model. The first approach,
based on properties of $SU(2)$, leads to the reduced eigenvalue problem
(\ref{eigw}). The second approach, based on integrability, leads to
the Bethe ansatz solution (\ref{EBAhomg}) and (\ref{BAhom}); however,
it remains to be understood whether this solution can give the full spectrum.
We have also performed an exact analysis of the time evolution of a
spin coherent state, leading to the result (\ref{evolvedstate}).
This allows us to compute the time evolution of various quantities of physical
interest, including the entanglement entropy (\ref{entropy}), spin
polarization (\ref{spinpolarization}) and Loschmidt echo
(\ref{Loschmidt}). For $s>\frac{1}{2}$, we have found interesting differences in
comparison with the $s=\frac{1}{2}$ case \cite{Guan:2018b}.
While we have restricted for simplicity to the case
that the central spin is initially ``up'', it should be
possible to generalize this analysis to other cases, such as the
GHZ \cite{Greenberger:1990}-like state $(|s, s \rangle + |s, -s
\rangle)/\sqrt{2}$, or states corresponding to entangled states of
lower spin. We expect that this model can help usher in the era of tunable quantum
metrology.
\section*{Acknowledgments}
RN was supported by the Chinese Academy of Sciences President's International Fellowship
Initiative Grant No. 2018VMA0017, and he is grateful for the hospitality extended to
him at the Wuhan Institute of Physics and Mathematics, where this
work was performed. This work was supported by
Key NNSFC grant number 11534014 and by MOST grant number 2017YFA0304500.
\appendix
\section{Computation of the coefficients $h^{(s,
k)}_{j}$}\label{sec:coeffs}
Here we compute the important coefficients $h^{(s, k)}_{j}$ appearing
in (\ref{Hk}). Our strategy is to determine these coefficients by means of
recurrence relations. Acting on both sides of (\ref{Hk}) with the
Hamiltonian $H^{(s)}$, we
obtain
\begin{align}
(H^{(s)})^{k+1}\left(|s, s \rangle \otimes |n\rangle \right) &=
\sum_{j=0}^{2s} h^{(s,k)}_{j}
H^{(s)} \left(|s,s-j \rangle \otimes |n-j\rangle \right) \non\\
&= \sum_{j=0}^{2s} h^{(s, k+1)}_{j}
|s,s-j \rangle \otimes |n-j\rangle \,.
\end{align}
Writing the Hamiltonian in terms of spin raising and lowering
operators (\ref{spinShomogH3}) and making use of
(\ref{raisinglowering}), we arrive at a system of recurrence
relations
\be
h^{(s,k+1)}_{j} = \alpha_{j}^{(s)}\, h^{(s,k)}_{j} +
\beta_{j}^{(s)}\, h^{(s,k)}_{j+1} +
\beta_{j-1}^{(s)}\, h^{(s,k)}_{j-1}\,, \qquad h^{(s,0)}_{j} = \delta_{j,0} \,,
\qquad j = 0, 1, \ldots, 2s\,,
\label{recurrence}
\ee
whose coefficients are given by
\begin{align}
\alpha_{j}^{(s)} &=B (s-j) + 2 A(s-j)(\frac{N}{2} + j-n) \,, \non\\
\beta_{j}^{(s)} &= A \sqrt{(j+1)(2s-j)(n-j)(N+j+1-n)} \,.
\label{alphabeta}
\end{align}
In order to solve these recurrence relations, we
define the generating functions
\be
h_{j}^{(s)}(z) = \sum_{k=0}^{\infty} h^{(s,k)}_{j} z^{k} \,.
\label{genfuncs}
\ee
By virtue of (\ref{recurrence}), these generating functions satisfy a system of linear relations
\be
(\alpha_{j}^{(s)} - \frac{1}{z})\, h_{j}^{(s)}(z) + \beta_{j}^{(s)}\, h_{j+1}^{(s)}(z) +
\beta_{j-1}^{(s)}\, h_{j-1}^{(s)}(z) = - \frac{1}{z} \delta_{j,0} \,,
\qquad j = 0, 1, \ldots, 2s \,,
\label{genfuncrltns}
\ee
which we can also write in matrix form
\be
M^{(s)}(z)\, h^{(s)}(z) = C^{(s)}(z)\,,
\label{matform}
\ee
where we have introduced the column vectors $h^{(s)}(z) =
(h_{0}^{(s)}(z), \ldots, h_{2s}^{(s)}(z))^{T}$ and $C^{(s)}(z) = (-
\frac{1}{z}, 0, \ldots , 0)^{T}$, and $M^{(s)}(z)$ is the tridiagonal
matrix\footnote{We suppress here the superscripts ${}^{(s)}$ on the
$\alpha$'s and $\beta$'s.}
\be
M^{(s)}(z) = \left(\begin{array}{ccccccc}
\alpha_{0}-\frac{1}{z} & \beta_{0} & 0 & 0 &\ldots & 0 & 0 \\
\beta_{0} & \alpha_{1}- \frac{1}{z} & \beta_{1} & 0 & \ldots & 0 & 0 \\
0 & \beta_{1} & \alpha_{2}- \frac{1}{z} & \beta_{2} & \ldots & 0 & 0 \\
\vdots & \vdots & \vdots & \vdots & & \vdots & \vdots \\
0 & 0 & 0 & 0 & \ldots & \beta_{2s-1} & \alpha_{2s}-\frac{1}{z}
\end{array}
\right)_{(2s+1) \times (2s+1)} \,.
\label{Mmat}
\ee
The solutions of the relations (\ref{genfuncrltns}) are given by the
following rational functions of $z$
\be
h_{j}^{(s)}(z) = \frac{\sum_{l=0}^{2s} n_{l}^{(s, j)}
z^{l}}{\sum_{l=0}^{2s+1} d_{l}^{(s)} z^{l}} \,, \qquad j = 0, 1, \ldots, 2s
\,.
\label{hjzsltn}
\ee
The coefficients in the denominator $d_{l}^{(s)}$ can be read
off from an expansion of the determinant of the matrix $M^{(s)}(z)$
(\ref{Mmat}) in inverse powers of $z$,
\be
\det \left( M^{(s)}(z) \right) =\sum_{l=0}^{2s+1} \frac{ (-1)^{2s+1} }{z^{2s+1-l}}
d^{(s)}_{l} \,, \qquad d_{0}^{(s)} = 1 \,.
\label{denomcoeffs}
\ee
Note that $d^{(s)}_{l}$ is independent of the value of $j$ in
(\ref{hjzsltn}). The coefficients in the numerator $n_{l}^{(s, j)}$,
which do depend on the value of $j$, can be read off from a similar
expansion of the {\em minors} of the first row of the matrix
$M^{(s)}(z)$,
\be
{\rm minor}_{(1,1+j)} \left( M^{(s)}(z) \right) =\sum_{l=0}^{2s}
\frac{ (-1)^{2s+j} }{z^{2s-l}}
n^{(s,j)}_{l} \,, \qquad n^{(s,j)}_{0} = \delta_{j,0}\,, \qquad j = 0, 1, \ldots, 2s
\label{numcoeffs}
\,.
\ee
Clearly, both $d^{(s)}_{l}$ and $n_{l}^{(s, j)}$ are expressed in
terms of the $\alpha$'s and $\beta$'s (\ref{alphabeta}).
Performing a partial fraction decomposition of the solutions
$h_{j}^{(s)}(z)$ (\ref{hjzsltn}), we obtain
\be
h_{j}^{(s)}(z) = \sum_{l=0}^{2s}\frac{c_{j,l}^{(s)}}{1-\omega_{l}^{(s)} z}\,,
\label{hpf}
\ee
where $\omega_{l}^{(s)}$ is the $(l+1)^{th}$ root of the following polynomial
equation of degree $2s+1$
\be
\sum_{i=0}^{2s+1} d_{i}^{(s)}\, z^{2s+1-i} = 0 \,.
\label{gpoly}
\ee
Moreover, we obtain formulas for
$c_{j,l}^{(s)}$ by evaluating the residues of $h_{j}^{(s)}(z)$
(\ref{hjzsltn}) at $z=1/\omega_{l}^{(s)}$,
\be
c_{j,l}^{(s)} = \frac{\sum_{i=0}^{2s} n_{i}^{(s, j)}
\left(\omega_{l}^{(s)}\right)^{2s-i} }{\prod_{i=1, i\ne
l}^{2s+1}(\omega_{l}^{(s)} - \omega_{i}^{(s)})} \,.
\label{cjl}
\ee
In view of the definition (\ref{genfuncs}) of the generating
functions, the sought-after
coefficients $h_{j}^{(s,k)}$ can be obtained from
\be
h_{j}^{(s,k)} = \frac{1}{k!} \frac{d^{k}}{dz^{k}} h_{j}^{(s)}(z)
\Big\vert_{z=0} \,.
\ee
Performing this computation using the solution (\ref{hpf}), we
conclude that
\be
h_{j}^{(s,k)} = \sum_{l=0}^{2s} c_{j,l}^{(s)}\,
\left(\omega_{l}^{(s)}\right)^{k} \,, \qquad j = 0, 1, \ldots, 2s
\,.
\label{hjsltn}
\ee
As a consistency check, note that substituting the result
(\ref{hjsltn}) back into (\ref{genfuncs}) and interchanging the order of
summations, one obtains
\be
h_{j}^{(s)}(z) = \sum_{l=0}^{2s} c_{j,l}^{(s)} \left[
\sum_{k=0}^{\infty} \left(\omega_{l}^{(s)} z \right)^{k} \right] \,.
\label{geometric}
\ee
Summing the geometric series in (\ref{geometric}), one recovers the result (\ref{hpf}).
In short, we have the following recipe for computing the coefficients
$h_{j}^{(s,k)}$ appearing in (\ref{Hk}):
\begin{enumerate}
\item Construct the matrix $M^{(s)}(z)$ (\ref{Mmat}), where the
$\alpha$'s and $\beta$'s are defined in (\ref{alphabeta}).
\item Read off $d_{l}^{(s)}$ and $n_{l}^{(s, j)}$ using
(\ref{denomcoeffs}) and (\ref{numcoeffs}), respectively.
\item Solve the polynomial equation (\ref{gpoly}) to obtain
$\omega_{l}^{(s)}$.
\item Obtain $c_{j,l}^{(s)}$ using (\ref{cjl}).
\item Obtain $h_{j}^{(s,k)}$ using (\ref{hjsltn}).
\end{enumerate}
It is straightforward to implement this recipe numerically on a
computer.
Note that $\omega_{l}^{(s)}$ and $c_{j,l}^{(s)}$ also depend on
$n$, $N$, $A$, and $B$ (through the $\alpha$'s and $\beta$'s), but we
have not explicitly displayed these dependences here in order to lighten the
notation. However, we {\em do} display the dependence on $n$ in the
corresponding formula in the body of the paper (\ref{hjsltnmain}),
due to the presence there of a summation over $n$.
\section{Identities for $c_{j,l}^{(s)}$}\label{sec:identities}
We obtain here some useful identities for the coefficients
$c_{j,l}^{(s)}$ appearing in $h_{j}^{(s,k)} $ (\ref{hjsltn}).
We begin by evaluating the generating functions $h_{j}^{(s)}(z)$ at
$z=0$ in two different ways: using (\ref{hjzsltn}) we obtain
\be
h_{j}^{(s)}(0) = \frac{n_{0}^{(s, j)}}{d_{0}^{(s)}} = \delta_{j,0} \,,
\label{id1a}
\ee
while (\ref{hpf}) gives
\be
h_{j}^{(s)}(0) = \sum_{l=0}^{2s} c_{j,l}^{(s)}\,.
\label{id1b}
\ee
We conclude from (\ref{id1a}) and (\ref{id1b}) that
\be
\sum_{l=0}^{2s} c_{j,l}^{(s)} = \delta_{j,0} \,.
\label{id1}
\ee
Multiplying both sides of (\ref{id1}) by $c_{j,l'}^{(s)}$ and summing
over $j$, we obtain
\be
\sum_{j=0}^{2s} \sum_{l=0}^{2s} c_{j,l}^{(s)}\, c_{j,l'}^{(s)} = c_{0,l'}^{(s)} \,.
\label{id2}
\ee
Let us now evaluate the expression $\sum_{j=0}^{2s} h_{j}^{(s)}(z)\,
h_{j}^{(s)}(z')$ in two different ways. Using (\ref{hpf}) we obtain
\be
\sum_{j=0}^{2s} h_{j}^{(s)}(z)\, h_{j}^{(s)}(z') = \sum_{j=0}^{2s}
\sum_{l, l'=0}^{2s}\frac{c_{j,l}^{(s)}\, c_{j,l'}^{(s)}}{(1-\omega_{l}^{(s)}
z)(1-\omega_{l'}^{(s)} z')} \,.
\label{id3}
\ee
On the other hand, we know from (\ref{matform}) that $h^{(s)}(z) =
(M^{(s)}(z))^{-1}\, C^{(s)}(z)$. Therefore,
\be
\sum_{j=0}^{2s} h_{j}^{(s)}(z)\, h_{j}^{(s)}(z') = \frac{1}{z z'}\left( M(z) M(z')
\right)^{-1}_{00} \,.
\label{other}
\ee
Since $\det(M(z))$ does not vanish at $z=\omega_{l}^{(s)}$ (it
vanishes instead at $z=1/\omega_{l}^{(s)}$), and similarly for
$\det(M(z'))$, we see from (\ref{other}) that $\sum_{j=0}^{2s} h_{j}^{(s)}(z)\,
h_{j}^{(s)}(z')$ is regular at $z=\omega_{l}^{(s)}\,,
z'=\omega_{l'}^{(s)}$. Evaluating the residues of both sides of (\ref{id3})
at $z=\omega_{l}^{(s)}\,, z'=\omega_{l'}^{(s)}$ with $l \ne
l'$, we obtain
\be
\sum_{j=0}^{2s} c_{j,l}^{(s)}\, c_{j,l'}^{(s)} = 0 \,, \qquad l \ne
l' \,.
\label{id4}
\ee
Combining this result with (\ref{id2}), we conclude that
\be
\sum_{j=0}^{2s} c_{j,l}^{(s)}\, c_{j,l'}^{(s)} = c_{0,l}^{(s)} \,
\delta_{l,l'} \,.
\label{id5}
\ee
The main results of this appendix are the identities (\ref{id1}) and (\ref{id5}).
% \newpage
% \clearpage
% \bibliographystyle{utphys}
% \bibliography{refs}
\providecommand{\href}[2]{#2}\begingroup\raggedright\begin{thebibliography}{10}
\bibitem{Richardson:1963a}
R.~W. Richardson, ``{A restricted class of exact eigenstates of the
pairing-force Hamiltonian},'' {\em Phys. Lett} {\bfseries 3} (1963) 277--279.
\bibitem{Richardson:1964a}
R.~W. Richardson and N.~Sherman, ``{Exact eigenstates of the pairing-force
Hamiltonian},'' {\em Nucl. Phys.} {\bfseries 52} (1964) 221--238.
\bibitem{Gaudin:1976sv}
M.~Gaudin, ``{Diagonalization of a class of spin Hamiltonians},''
{\em J. Physique} {\bfseries 37} (1976) 1087--1098.
%%CITATION = SACLAY-DPH-T/76-38;%%.
\bibitem{Gaudin:1983}
M.~Gaudin, {\em La fonction d'onde de Bethe}.
\newblock Masson, 1983.
\newblock English translation by J.-S. Caux, The Bethe Wavefunction, Cambridge
University Press, 2014.
\bibitem{Khaetskii2002}
A.~V. Khaetskii, D.~Loss, and L.~Glazman, ``{Electron spin decoherence in
quantum dots due to interaction with nuclei},''
\href{http://dx.doi.org/10.1103/PhysRevLett.88.186802}{{\em Phys. Rev. Lett.}
{\bfseries 88} (2002) 186802},
\href{http://arxiv.org/abs/cond-mat/0201303}{{\ttfamily cond-mat/0201303}}.
\bibitem{Merkulov2002}
I.~A. Merkulov, A.~L. Efros, and M.~Rosen, ``{Electron spin relaxation by
nuclei in semiconductor quantum dots},''
\href{http://dx.doi.org/10.1103/PhysRevB.65.205309}{{\em Phys. Rev. B}
{\bfseries 65} (2002) 205309},
\href{http://arxiv.org/abs/cond-mat/0202271}{{\ttfamily cond-mat/0202271}}.
\bibitem{Khaetskii2003}
A.~V. Khaetskii, D.~Loss, and L.~Glazman, ``{Electron spin evolution induced by
interaction with nuclei in a quantum dot},''
\href{http://dx.doi.org/10.1103/PhysRevB.67.195329}{{\em Phys. Rev. B}
{\bfseries 67} (2003) 195329},
\href{http://arxiv.org/abs/cond-mat/0211678}{{\ttfamily cond-mat/0211678}}.
\bibitem{Erlingsson2004}
S.~I. Erlingsson and Y.~V. Nazarov, ``{Evolution of localized electron spin in
a nuclear spin environment},''
\href{http://dx.doi.org/10.1103/PhysRevB.70.205327}{{\em Phys. Rev. B}
{\bfseries 70} (2004) 205327}.
\bibitem{Braun2005}
P.-F. Braun, X.~Marie, L.~Lombez, B.~Urbaszek, T.~Amand, P.~Renucci, V.~K.
Kalevich, K.~V. Kavokin, O.~Krebs, P.~Voisin, and Y.~Masumoto, ``{Direct
observation of the electron spin relaxation induced by nuclei in quantum
dots},'' \href{http://dx.doi.org/10.1103/PhysRevLett.94.116601}{{\em Phys.
Rev. Lett.} {\bfseries 94} (2005) 116601}.
\bibitem{Chen2007}
G.~Chen, D.~L. Bergman, and L.~Balents, ``{Semiclassical dynamics and long-time
asymptotics of the central-spin problem in a quantum dot},''
\href{http://dx.doi.org/10.1103/PhysRevB.76.045312}{{\em Phys. Rev. B}
{\bfseries 76} (2007) 045312},
\href{http://arxiv.org/abs/cond-mat/0703631}{{\ttfamily cond-mat/0703631}}.
\bibitem{Bortz2007}
M.~Bortz and J.~Stolze, ``{Exact dynamics in the inhomogeneous central-spin
model},'' \href{http://dx.doi.org/10.1103/PhysRevB.76.014304}{{\em Phys. Rev.
B} {\bfseries 76} (2007) 014304},
\href{http://arxiv.org/abs/cond-mat/0612382}{{\ttfamily cond-mat/0612382}}.
\bibitem{Bortz2010}
M.~Bortz, S.~Eggert, C.~Schneider, R.~Stubner, and J.~Stolze, ``{Dynamics and
decoherence in the central spin model using exact methods},''
\href{http://dx.doi.org/10.1103/PhysRevB.82.161308}{{\em Phys. Rev. B}
{\bfseries 82} (2010) 161308},
\href{http://arxiv.org/abs/1005.0001}{{\ttfamily 1005.0001}}.
\bibitem{Dooley2013}
S.~Dooley, F.~McCrossan, D.~Harland, M.~J. Everitt, and T.~P. Spiller,
``{Collapse and revival and cat states with an $N$-spin system},''
\href{http://dx.doi.org/10.1103/PhysRevA.87.052323}{{\em Phys. Rev. A}
{\bfseries 87} (2013) 052323},
\href{http://arxiv.org/abs/arXiv:1302.2806}{{\ttfamily arXiv:1302.2806
[quant-ph]}}.
\bibitem{Guan:2018a}
W.-B. He and X.-W. Guan, ``{Exact results of quantum dynamics of Gaudin
models},'' in preparation.
\bibitem{Guan:2018b}
W.-B. He, S. Chesi, H.-Q. Lin and X.-W. Guan, ``{Exact quantum dynamics of
XXZ central spin problems},'' {{\ttfamily
arXiv:1810.03012 [quant-ph]}}.
\bibitem{Jaynes1963}
E.~T. Jaynes and F.~W. Cummings, ``{Comparison of quantum and semiclassical
radiation theories with application to the beam maser},''
\href{http://dx.doi.org/10.1109/PROC.1963.1664}{{\em Proc. IEEE} {\bfseries
51} (1963) 89--109}.
\bibitem{Scully:1997}
M.~O. Scully and M.~S. Zubairy, {\em Quantum Optics}.
\newblock Cambridge University Press, 1997.
\bibitem{Bogoliubov2013}
N.~M. Bogoliubov and P.~P. Kulish, ``{Exactly solvable models of nonlinear
quantum optics},'' {\em J. Math. Sciences} {\bfseries 192} (2013)
14--30.
\bibitem{Tavis1968}
M.~Tavis and F.~W. Cummings, ``{Exact solution for an
N-moleculeÑradiation-field Hamiltonian},''
\href{http://dx.doi.org/10.1103/PhysRev.170.379}{{\em Phys. Rev.} {\bfseries
170} (1968) 379--384}.
\bibitem{Knap2010}
M.~Knap, E.~Arrigoni, and W.~von~der Linden, ``{Quantum phase transition and
excitations of the Tavis-Cummings lattice model},''
\href{http://dx.doi.org/10.1103/PhysRevB.82.045126}{{\em Phys. Rev. B}
{\bfseries 82} (2010) 045126}.
\bibitem{Youssef2010}
M.~Youssef, N.~Metwally, and A.-S.~F. Obada, ``{Some entanglement features of a
three-atom TavisÐCummings model: a cooperative case},''
\href{http://dx.doi.org/10.1088/0953-4075/43/9/095501}{{\em J. Phys. B: At.
Mol. Opt. Phys} {\bfseries 43} (2010) 095501}.
\bibitem{Agarwal2012}
S.~Agarwal, S.~M. Hashemi~Rafsanjani, and J.~H. Eberly, ``{Tavis-Cummings model
beyond the rotating wave approximation: Quasidegenerate qubits},''
\href{http://dx.doi.org/10.1103/PhysRevA.85.043815}{{\em Phys. Rev. A}
{\bfseries 85} (2012) 043815}.
\bibitem{Sklyanin:1987ih}
E.~K. Sklyanin, ``{Separation of variables in the Gaudin model},''
\href{http://dx.doi.org/10.1007/BF01840429}{{\em J. Sov. Math.} {\bfseries
47} (1989) 2473--2488}.
[Zap. Nauchn. Semin.164, 151 (1987)].
%%CITATION = JOSMA,47,2473;%%.
\bibitem{Hikami:1992np}
K.~Hikami, P.~P. Kulish, and M.~Wadati, ``{Integrable spin systems with long
range interactions},''
\href{http://dx.doi.org/10.1143/JPSJ.61.3071}{{\em J. Phys. Soc. Jap.}
{\bfseries 61} (1992) 3071--3076}.
%%CITATION = JUPSA,61,3071;%%.
\bibitem{Zhou:2001}
H.-Q. Zhou, J.~Links, R.~H. McKenzie, and M.~D. Gould, ``{Superconducting
correlations in metallic nanoparticles: Exact solution of the BCS model by
the algebraic Bethe ansatz},''
\href{http://dx.doi.org/10.1103/PhysRevB.65.060502}{{\em Phys. Rev. B}
{\bfseries 65} (2002) 060502}, \href{http://arxiv.org/abs/0106390}{{\ttfamily
arXiv:0106390 [cond-mat.supr-con]}}.
\bibitem{Babujian:1983ae}
H.~M. Babujian, ``{Exact solution of the isotropic Heisenberg chain with
arbitrary spins: thermodynamics of the model},''
\href{http://dx.doi.org/10.1016/0550-3213(83)90668-5}{{\em Nucl. Phys.}
{\bfseries B215} (1983) 317--336}.
%%CITATION = NUPHA,B215,317;%%.
\bibitem{Links:2017}
J.~Links, ``{Completeness of the Bethe states for the rational, spin-1/2
Richardson--Gaudin system},''
\href{http://dx.doi.org/10.21468/SciPostPhys.3.1.007}{{\em SciPost Phys.}
{\bfseries 3} (2017) 007}, \href{http://arxiv.org/abs/1603.03542}{{\ttfamily
arXiv:1603.03542 [nlin.SI]}}.
\bibitem{Hao:2013jqa}
W.~Hao, R.~I. Nepomechie, and A.~J. Sommese, ``{Completeness of solutions of
Bethe's equations},''
\href{http://dx.doi.org/10.1103/PhysRevE.88.052113}{{\em Phys. Rev.}
{\bfseries E88} no.~5, (2013) 052113},
\href{http://arxiv.org/abs/1308.4645}{{\ttfamily arXiv:1308.4645 [math-ph]}}.
%%CITATION = ARXIV:1308.4645;%%.
\bibitem{Gaudin:1971zza}
M.~Gaudin, ``{Boundary Energy of a Bose Gas in One Dimension},''
\href{http://dx.doi.org/10.1103/PhysRevA.4.386}{{\em Phys. Rev.} {\bfseries A4}
(1971) 386--394}.
%%CITATION = PHRVA,A4,386;%%.
\bibitem{Shastry:2001}
S.~Shastry and A.~Dhar, ``{Solution of a generalized Stieltjes problem},'' {\em
J.Phys.} {\bfseries A34} (2001) 6197,
\href{http://arxiv.org/abs/cond-mat/0101464}{{\ttfamily
arXiv:cond-mat/0101464 [cond-mat]}}.
\bibitem{Pan:2011}
X.~Guan, K.~D. Launey, M.~Xie, L.~Bao, F.~Pan, and J.~P. Draayer,
``{Heine-Stieltjes correspondence and the polynomial approach to the standard
pairing problem},'' {\em Phys.Rev.} {\bfseries C86} (2012) 024313,
\href{http://arxiv.org/abs/1106.5237}{{\ttfamily arXiv:1106.5237 [nucl-th]}}.
\bibitem{Marquette:2012}
I.~Marquette and J.~Links, ``{Generalised Heine-Stieltjes and Van Vleck
polynomials associated with degenerate, integrable BCS models},'' {\em
J.Stat.Mech.} {\bfseries P08019} (2012) ,
\href{http://arxiv.org/abs/1206.2988}{{\ttfamily arXiv:1206.2988 [nlin.SI]}}.
\bibitem{Holstein1940}
T.~Holstein and H.~Primakoff, ``{Field dependence of the intrinsic domain
magnetization of a ferromagnet},''
\href{http://dx.doi.org/10.1103/PhysRev.58.1098}{{\em Phys. Rev.} {\bfseries
58} (1940) 1098--1113}.
\bibitem{Radcliffe1971}
J.~M. Radcliffe, ``{Some properties of coherent spin states},'' {\em J. Phys.
A} {\bfseries 4} (1971) 313.
\bibitem{Arecchi1972}
F.~T. Arecchi, E.~Courtens, R.~Gilmore, and H.~Thomas, ``{Atomic coherent
states in quantum optics},''
\href{http://dx.doi.org/10.1103/PhysRevA.6.2211}{{\em Phys. Rev. A}
{\bfseries 6} (1972) 2211--2237}.
\bibitem{Greenberger:1990}
D.~M. Greenberger, M.~Horne, and A.~Zeilinger, ``BellÕs theorem without
inequalities,'' {\em Am. J. Phys} {\bfseries 58} (1990) 1131.
\end{thebibliography}\endgroup
\end{document}