Homotopy Non-Invariance of the String Cobracket and the Failure of the Lie Bialgebra Structure
Extracted source text · arXiv 2605.10273
\documentclass[10pt,leqno]{amsart}
\usepackage{graphicx}
\baselineskip=16pt
\usepackage{indentfirst,csquotes}
\topmargin= .5cm
\textheight= 20cm
\textwidth= 32cc
\baselineskip=16pt
\evensidemargin= .9cm
\oddsidemargin= .9cm
\usepackage{amssymb,amsthm,amsmath,amsfonts,ascmac,mathrsfs,mathtools}
\usepackage{paralist,hyperref,fancyhdr,etoolbox}
\usepackage{tikz}
\usetikzlibrary{cd}
\theoremstyle{plain}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{example}[theorem]{Example}
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{problem}[theorem]{Problem}
\theoremstyle{remark}
\newtheorem{remark}[theorem]{Remark}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\newcommand{\Z}{\mathbb{Z}}
\newcommand{\mr}{\mathrm}
\newcommand{\ra}{\rightarrow}
\hypersetup{
colorlinks = true,
linkcolor = [rgb]{0.3,0.55,0.25},
citecolor = [rgb]{0.164,0.39,0.77},
urlcolor = [rgb]{0.7,0.2,0.4}
}
\def\proof{\noindent {\it Proof. $\, $}}
\def\endproof{\hfill $\Box$ \vskip 5 pt }
\begin{document}
\title[String Cobracket and Lie Bialgebra Structure]{Homotopy Non-Invariance of the String Cobracket and the Failure of the Lie Bialgebra Structure}
\author{Isana SUMOTO}
\address{Graduate School of Mathematical Sciences, The University of Tokyo, 3-8-1 Komaba, Meguro-ku, Tokyo, 153-8914, Japan}
\email{isana(at)ms.u-tokyo.ac.jp}
\begin{abstract}
We prove that the string cobracket is \emph{not} a homotopy invariant. Adapting Naef's method \cite{Naef24} for computing the string coproduct, we show that the string cobrackets on the three-dimensional lens spaces $L(9;1)$ and $L(9;4)$ differ. We further relate the string cobracket to the Whitehead torsion, analogously to the case of the string coproduct. In addition, we show that the string bracket and the string cobracket do \emph{not} endow the $S^1$-equivariant homology of the free loop space with a Lie bialgebra structure. These findings indicate that the analogy with the Turaev cobracket breaks down in a higher-dimensional string topology.
\end{abstract}
\maketitle
\begingroup
\let\thefootnote\relax
\footnotetext{\emph{2020 Mathematics Subject Classification}. Primary 55P50; Secondary 57Q10, 57K31.\\
\indent \emph{Key words and phrases}. String topology, Lens spaces, Lie bialgebra.}
\endgroup
%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
String topology, introduced in \cite{ChasSullivan}, investigates algebraic structures on the homology of free loop spaces and was developed as a higher‑dimensional analogue of Goldman's work on closed curves on surfaces \cite{Goldman}. Among its fundamental operations are the string product and the string coproduct\footnote{While the string product is referred to as the ``loop product'' in \cite{ChasSullivan}, the terms ``loop coproduct'' and ``string coproduct'' often denote different operations. In this paper, following the convention in \cite{Naef24}, we refer to the product defined by Chas and Sullivan as the string product, and the coproduct investigated by Goresky and Hingston as the string coproduct.}. They induce operations on the $S^1$-equivariant homology of the free loop space, called the string bracket and the string cobracket. The string bracket generalizes the Goldman bracket, and the string cobracket generalizes the Turaev cobracket with rational coefficients \cite{HartensteinStegemeyer}.\\
The string product is known to be a \emph{homotopy invariant} \cite{Cohen08,Crabb08,GruherSalvatore08} (or could be deduced from \cite{CohenJones}). The string coproduct is shown to be a simple-homotopy invariant in \cite{NaefSafronov}. Moreover, for a simply connected manifold, it is a homotopy invariant over $\mathbb{Q}$ \cite{RiveraWang}. However, Naef \cite{Naef24} computed the string coproduct on the three-dimensional lens spaces $L(7;1)$ and $L(7;2)$---which are homotopy equivalent but not homeomorphic---and obtained different values. Thus, the string coproduct is \emph{not} a homotopy invariant in general. Moreover, the string coproduct distinguishes non-homeomorphic three-dimensional lens spaces \cite{NaefManuelNathalie}.\\
This raises the question of whether the \emph{string cobracket} behaves analogously to the Turaev cobracket. In this paper, we focus on the \emph{string cobracket} of $L(9;1)$ and $L(9;4)$, a pair of manifolds that are homotopy-equivalent but non-homeomorphic and admit explicit computations. The choice of $9$ (rather than $7$) is essential: the nontrivial divisor $3$ plays a crucial role in our spectral sequence and torsion calculations.\\
Furthermore, we investigate the algebraic structure of the homology of free loop spaces. Turaev \cite{Turaev} proved that the Goldman bracket and the Turaev cobracket form a Lie bialgebra. Similarly, Chas and Sullivan \cite{Chas2004} discussed that the string bracket and the string cobracket on the rational homology of the free loop space of smooth, closed, oriented, and simply connected manifolds form a Lie bialgebra. It is natural to ask whether a similar statement holds for higher-dimensional manifolds with integer coefficients.\\
The main contributions of this paper are as follows:
\begin{enumerate}
\item Like the string coproduct, the string cobracket is also \emph{not} a homotopy invariant. More precisely, we prove that the number of components on which the string cobracket is nonzero differs between $L(9;1)$ and $L(9;4)$.
\item The string cobracket can be expressed in terms of Whitehead torsion, extending the relationship previously observed for the string coproduct by Naef \cite{Naef24}.
\item The pair consisting of the string bracket and the string cobracket does \emph{not} form a Lie bialgebra. This contrasts with the classical surface case, where the Goldman bracket and the Turaev cobracket form a Lie bialgebra \cite{Turaev}, as well as with the rational string topology of simply connected manifolds \cite{Chas2004}.
\end{enumerate}
These results reveal that the string cobracket is not merely a higher-dimensional analogue of the Turaev cobracket or the rational string cobracket; rather, it exhibits a richer and more subtle algebraic structure.\\
The paper is organized as follows. In Section \ref{section2}, we review the necessary background on string topology and the results by Naef \cite{Naef24}. In Section \ref{section3}, we compute the string cobracket on $L(9;1)$ and $L(9;4)$ using the Leray--Serre spectral sequence, and demonstrate that the string cobracket is not a homotopy invariant. Section \ref{section4} is devoted to discussing the relationship between our results and Whitehead torsion. Finally, in Section \ref{section5}, we apply the computations from Section \ref{section3} to prove that the string bracket and the string cobracket do not form a Lie bialgebra.
\section*{Acknowledgments}
I would like to express my gratitude to my supervisor, Takuya Sakasai, for his invaluable advice and tremendous support. I also thank Katsuhiko Kuribayashi, Takahito Naito, and Toyo Taniguchi for carefully reading the manuscript, and providing constructive suggestions and insightful comments.
%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Preparation}\label{section2}
\subsection{Definitions}
Throughout this paper, all homology and cohomology are taken with $\Z$-coefficients. Let $M$ be an oriented compact $d$-dimensional manifold, and let $LM$ be its free loop space $\mathrm{Map}(S^1,M)$. Throughout this paper, we adopt the Sobolev $H^1$-loop space as our model for $LM$ (see \cite{HingstonWahl}); consequently, $LM$ is regarded as a Hilbert manifold.\\
The \emph{string product}
\begin{align*}
\wedge\colon H_*(LM\times LM)\ra H_{*-d}(LM)
\end{align*}
is an operation that equips the ($d$-shifted) homology of $LM$ with a ring structure. Dually, the \emph{string coproduct}
\begin{align*}
\vee\colon H_*(LM,M)\ra H_{*+1-d}(LM\times LM,LM\times M\cup M\times LM)
\end{align*}
is defined in \cite{Sullivan_2004} (also see \cite{GoreskyHingston}). Here, the subspace $M\subset LM$ is obtained by identifying each point $x\in M$ with the constant loop at $x$ via the natural inclusion $M\hookrightarrow LM$. The string bracket (resp. string cobracket) is defined by a combination of string product (resp. string coproduct) with the BV-operator introduced below.\\
First, we see the property of the string coproduct following Proposition 3.7 in \cite{HingstonWahl}. Let $N$ be an oriented compact $k$-dimensional manifold, and let $\alpha\colon N\ra LM$ be a smooth map. Assume transversality in the following sense:
\begin{enumerate}
\item For all $n\in N$, $\partial\alpha(n)(t)/\partial t$ is nonzero at $t=0$.
\item The map $\bar{\alpha}\colon N\times (0,1)\ra M\times M$ given by $(n,t)\mapsto (\alpha(n)(0),\alpha(n)(t))$, $\bar{\alpha}$ intersects $V$ the diagonal transversely in a compact submanifold of $N\times (0,1)$. Here,
\begin{align*}
V\coloneqq \{(n,t)\in N\times S^1 \mid \alpha(n)(t)=\alpha(n)(0),t\neq 0\}
\end{align*}
is the self-intersection locus of $\alpha$.
\end{enumerate}
Let $x$ be a homology class given by $\alpha_*[N]\in H_*(LM,M)$. The map $\alpha$ induces
\begin{align*}
\vee(\bar{\alpha})\colon V\ra& LM\times LM\\
(n,t)\mapsto& \left(s\mapsto \alpha(n)(st),s\mapsto \alpha(n)((1-t)s+t)\right).
\end{align*}
The string coproduct $\vee x$ coincides with
\begin{align*}
\vee x= \vee(\alpha_*[N])=(\vee\bar{\alpha})_*([V])\in H_{k+1-d}(LM\times LM,LM\times M\cup M\times LM).
\end{align*}
Second, we define the BV-operator, which plays a central role in the homology of free loop spaces.
\begin{definition}[BV-operator]
The free loop space $LM$ carries an $S^1$-action $\rho:S^1\times LM\ra LM;~(\theta,\gamma)\mapsto (t\mapsto \gamma(t+\theta))$. The \emph{Batalin--Vilkovisky(BV) operator} is the map
\begin{align*}
H_*(LM)&\ra H_{*+1}(LM)\\
\alpha &\mapsto \rho_*([S^1]\otimes\alpha).
\end{align*}
\end{definition}
For a space $X$ with an $S^1$-action, we denote its $S^1$-equivariant homology by $H_*^{S^1}(X)=H_*(ES^1\times_{S^1}X)$, where $ES^1 \times_{S^1} X$ is the Borel construction.\\
Finally, following \cite{NaefWillwacher}, we will define the string cobracket on the $S^1$-equivariant homology of $LM$ as follows. Since we work with $\Z$-coefficients, torsion terms appear. Also, one must be careful about whether the string cobracket can be defined in the same manner as with $\mathbb{Q}$-coefficients. However, a calculation analogous to that in the case of the string bracket confirms that this is not an issue.
\begin{definition}[string cobracket]
The string cobracket $\vee_{S^1}$ is defined by the composition
\begin{align*}
\vee_{S^1}:&H_*^{S^1}(LM,M) \xrightarrow{\tau_*} H_{*+1}(LM,M)\\
&\xrightarrow[\substack{\text{string}\\ \text{coproduct}}]{\vee} H_{*+2-d}(LM\times LM,LM\times M\cup M\times LM) \\
&\xrightarrow{\text{Künneth}}\left(H_*(LM,M)\otimes H_*(LM,M)\right)[2-d]\oplus \mr{Tor}\left(H_*(LM,M), H_*(LM,M)\right)[1-d]\\
&\xrightarrow{(\pi_*\otimes\pi_*)\oplus \mr{Tor}(\pi_*,\pi_*)} \left(H_*^{S^1}(LM,M)\otimes H_*^{S^1}(LM,M)\right)[2-d]\\
&\hspace{83pt}\oplus \mr{Tor}\left(H_*^{S^1}(LM,M), H_*^{S^1}(LM,M)\right)[1-d].
\end{align*}
Here, $\tau$ is the relative $S^1$-equivariant transfer map. The natural inclusion $LM\hookrightarrow ES^1\times LM$ is a homotopy equivalence and since the diagram
\[
\begin{tikzcd}
ES^1\times LM \arrow[d,"\simeq"]\arrow[rd,"\pi"]\\
LM\arrow[r,"i"] &ES^1\times_{S^1} LM
\end{tikzcd}
\]
commutes, we identify $i$ with $\pi$ by abuse of notation.
\end{definition}
In section \ref{section3}, we study the maps $\tau_*$ and $\pi_*$ in more detail. The string bracket $\wedge_{S^1}$ is defined analogously.
\subsection{Application of the results by Naef}
In the following, let $M$ be the lens space $L(9;k)$ with $k=1,4$. The lens spaces $L(9;1)$ and $L(9;4)$ are homotopy equivalent but not homeomorphic. Since $\pi_1(M)\cong \Z/9\Z$, the free loop space $LM$ decomposes into $9$ path-connected components, which we denote by $L_l M$ indexed by $l \in \Z/9\Z$.\\
Following Naef \cite{Naef24}, the homology of $L_lM$ can be computed as follows:
\begin{align*}
H_0(L_lM)\cong \Z,\quad H_1(L_lM)\cong \Z/9\Z,\quad H_2(L_lM)\cong \Z,\\
H_3(L_lM)\cong \Z\oplus(\Z/9\Z),\quad H_4(L_lM)\cong \Z,~\dots.
\end{align*}
\begin{remark}
In \cite{Naef24}, Naef computed the homology of the free loop space of the lens spaces $L(7;1)$ and $L(7;2)$. Since the argument does not rely on the fact that $7$ is prime, the same method applies to $L(9;k)$ without modification. On the other hand, in the latter part of our discussion, the fact that $9$ is not prime plays an essential role.
\end{remark}
\begin{proposition}\label{rho_0}
Let $x_l \in H_1(L_lM)$ denote the positive generator. Then, the equality $\rho_*([S^1]\otimes 1)=l\cdot x_l$ holds.
\end{proposition}
\begin{proof}
Following \cite{Naef24}, we identify $H_0(LM)$ with $\Z[\Z/9\Z]$, and consider the module of formal de-Rham 1-forms $\Omega^1=\Omega^1(\Z[\Z/9\Z])=((\Z/9\Z)[t]/(t^9-1))~dt/t$. The correspondence
\begin{align*}
(\Z/9\Z)[t]/(t^9-1)~dt/t&\ra H_1(LM)=\bigoplus_{l\in\Z/9\Z} H_1(L_lM)
\end{align*}
given by
\begin{align*}
(c_0+c_1t+\dots+c_8t^8)\frac{dt}{t}&\mapsto (c_0,\dots,c_8)
\end{align*}
identifies coefficients with the generators of each $H_1(L_lM)$. Under this identification, the $S^1$ action induces the map $H_0(LM)\ra H_1(LM)$ corresponding to the formal de Rham differential
\begin{align*}
\Z[\Z/9\Z]&\ra (\Z/9\Z)[t]/(t^9-1)~dt/t\\
(c_0+c_1t+\dots+c_8t^8)&\mapsto (c_1t+2c_2t^2+\dots+8c_8t^8)\frac{dt}{t}.
\end{align*}
\end{proof}
Now, we consider the following component:
\begin{align*}
\vee_{S^1}\colon H_2^{S^1}(LM)\ra\left(H_1^{S^1}(LM)\otimes H_0^{S^1}(LM)\right)\oplus\left(H_0^{S^1}(LM)\otimes H_1^{S^1}(LM)\right).
\end{align*}
Since $\mr{Tor}\left(H_0(LM), H_0(LM)\right)=0$, torsion terms do not appear. Let $\mr{pr}_1$ denote the projection onto the first summand. Since $\mr{pr}_1$ commutes with $\pi_*\otimes \pi_*$, we may rearrange the order of composition as convenient. (Since both the string coproduct and the string cobracket satisfy coassociativity, composing with the projection does not affect the componentwise structure and hence causes no difficulty.)\\
Let $t^pt_2^q~dt/t\in H_1(L_pM)\otimes H_0(L_qM)$ be the homology class whose projection to
\begin{align*}
H_1(M)\otimes H_0(L_qM)\cong\Z/9\Z
\end{align*}
is the positive generator. For integers $l,m$, let $[\rho_{l,m}] \in H_3(L_lM)$ be the homology class obtained by pushing forward the fundamental class $[M]$ of $M$ along $\rho_{l,m}$. Here, $\rho_{l,m}\colon M\rightarrow LM$ is the map induced by
\begin{align*}
\tilde\rho_{l,m}\colon S^3&\ra LM\\
(z_1,z_2)&\mapsto (t\mapsto p\left(e^{2\pi \sqrt{-1}lt}z_1,e^{2\pi \sqrt{-1}(kl+9m)t}z_2\right)
\end{align*}
where $p\colon S^3\ra M$ is the canonical quotient map. Using the identification $H_1(LM)\otimes H_0(LM)\cong R/I$, where $R=(\Omega^1/(\Z/9\Z)~dt/t)\otimes (\Z[\Z/9\Z]/\Z\cdot 1)$ and $I$ is the ideal generated by $t^9-1$ and $t_2^9-1$. The same argument as in \cite[Proposition 3.1]{Naef24}, shows that for $l\neq 0$, the equality
\begin{align}
\begin{aligned}\label{string_coproduct_calc}
\mr{pr}_1\circ\vee(j_*[\rho_{l,m}])&=\left(tt_2^{l-1}+t^2t_2^{l-2}+\cdots+t^{l-1}t_2\right)\frac{dt}{t}\\
&+r\left(t^rt_2^{(kl+9m-1)r}+t^{2r}t_2^{(kl+9m-2)r}+\cdots+t^{(kl+9m-1)r}t_2^r\right)\frac{dt}{t} \mod{I}
\end{aligned}
\end{align}
holds. Here, $j\colon LM\ra (LM,M)$ is natural inclusion, and $r$ is the inverse of $k$ modulo $9$.
Since the string coproducts of $\{[\rho_{l,1+nl}]\}_{n=1,2,\dots,9}$ are all distinct, they span $H_3(L_lM)$.
\section{Computation of the string cobracket on the lens space}\label{section3}
In this section, we compute the string cobracket on the lens space $L(9;k)$ for $k=1,4$. More precisely, we show that the number of components on which the string cobracket is nonzero differs between two lens spaces. We calculate $H_r(L_lM)$ for $r=0,1,2,3$, then we investigate $\tau_*$ and $\pi_*$. We denote a nonzero element of $\Z/9\Z$ by $l$.
\subsection{Computation of $H_r^{S^1}(L_lM)$ for $r\leq 3$}
We analyze the Leray--Serre spectral sequence associated with the fibration
\begin{align*}
L_lM\ra ES^1\times_{S^1}L_lM\xrightarrow{\varpi} BS^1.
\end{align*}
\begin{proposition}
Let $p$ be even. Then the differential of the $E^2$-term $d_{p,q}^2\colon E_{p,q}^2\ra E_{p-2,q+1}^2$ coincides with $u_p\otimes x\mapsto u_{p-2}\otimes \rho_*([S^1]\otimes x)$ where $u_p\in H_p(BS^1)$ denotes the generator.
\end{proposition}
\begin{proof}
Let $p$ be even and $(BS^1)^{(p)}$ denote the $p$–skeleton of $BS^1$ as $\mathbb{C}P^\infty$, and set $Y^p=\varpi^{-1}((BS^1)^{(p)})$. By the definition of the Leray--Serre spectral sequence, $E_{p,q}^1=H_{p+q}(Y^p,Y^{p-1})$. If $p$ is odd, then $Y^p$ is $Y^{p-1}$, and hence $E_{p,q}^1=0$. Therefore
\begin{align*}
d_{p,q}^1\colon E_{p,q}^1\ra E_{p-1,q}^1
\end{align*}
is the zero map. Consequently,
\begin{align*}
E_{p,q}^2=H_{p+q}(Y^p,Y^{p-1})=H_{p+q}(Y^p,Y^{p-2}).
\end{align*}
From now on, $p$ will be even.\\
\underline{Step 1: The case $p=2$}\\
We follow the proof of Theorem 5.3 in \cite{Hatcher}. Let $\Phi_2\colon D^2\ra (BS^1)^{(2)}$ be the characteristic map for the $2$-cell of $BS^1$. Let $e_0$ be a point in $(BS^1)^{(2)}$. Then we obtain a map $\tilde\Phi_2\colon (D^2,S^1)\times\varpi^{-1}(\Phi_2(e_0))\ra (Y^p,Y^{p-1})$. As in \cite{Hatcher}, the map induces an isomorphism
\begin{align*}
\varphi\colon H_2(S^{\Delta}((BS^1)^{(2)},(BS^1)^{(1)};\mathcal{H}_q))\ra& H_{q+2}(Y^2,Y^1)\\
[\Phi_2]\otimes x \mapsto&(\tilde\Phi_2)_*(\mu\otimes x).
\end{align*}
where $\mu$ is the fundamental class of $(D^2,S^1)$, $S^{\Delta}(-,-;-)$ is the relative cellular singular chain complex and $\mathcal{H}_q$ is the local system on $BS^1$. Since $L_lM$ is path-connected, the local system $\mathcal{H}_q$ is trivial. Thus,
\begin{align*}
H_2(S^{\Delta}((BS^1)^{(2)},(BS^1)^{(1)};\mathcal{H}_q))\cong H_2((BS^1)^{(2)},(BS^1)^{(1)})\otimes H_q(L_lM).
\end{align*}
The differential $d_{2,q}^2\colon H_q(Y^2,Y^0)\ra H_{q+1}(Y^0)$ is the boundary map in the long exact sequence of the pair $(Y^2,Y^0)$. Therefore,
\begin{align*}
d_{2,q}^2(\varphi ([\Phi_2]\otimes x))=\partial_*((\tilde\Phi_2)_*(\mu\otimes x)).
\end{align*}
By construction,
\begin{align*}
\tilde\Phi_2\colon (D^2,S^1)\times L_lM\ra& (Y_2,Y_0)
\end{align*}
is described as $(z,\gamma)\mapsto [[z:1],\gamma]$ and its restriction to $S^1\times L_lM$ is
\begin{align*}
(\tilde\Phi_2)|_{S^1\times L_lM}\colon S^1\times L_lM\ra& Y_0\\
(e^{2\pi \sqrt{-1}t},\gamma)\mapsto& [e^{2\pi \sqrt{-1}t},\gamma]=\rho(t,\gamma).
\end{align*}
Thus, $(\tilde\Phi_2)|_{S^1\times L_lM}=\rho$. By naturality of the boundary map,
$\partial_*((\tilde\Phi_2)_*(\mu\otimes x))=\rho_*(\partial_*(\mu\otimes x))=\rho_*([S^1]\otimes x)$ holds.\\
\noindent\underline{Step 2: The case of general even $p$}\\
Let $e\in H^2(BS^1)$ be the Euler class of the tautological complex line bundle. Using the cap product and the naturality of the differential of the spectral sequence, we compute:
\begin{align*}
&(e\otimes 1)\cap (u_{p-2}\otimes \rho_*([S^1]\otimes x)-d_{p,q}^2(u_p\otimes x))\\
&=u_{p-4}\otimes \rho_*([S^1]\otimes x)- d_{p,q}^2(e\otimes 1 \cap u_p\otimes x)\\
&=u_{p-4}\otimes \rho_*([S^1]\otimes x)-d_{p-2,q}^2(u_{p-2}\otimes x)\\
&=0.
\end{align*}
Since the cap product
\begin{align*}
e\otimes 1\cap ~\colon H_p(BS^1)\otimes H_q(L_lM)\ra H_{p-2}(BS^1)\otimes H_q(L_lM)
\end{align*}
is an isomorphism, we conclude that $u_p\otimes \rho_*([S^1]\otimes x)=d_{p,q}^2(u_{p-2}\otimes x)$. This completes the proof.
\end{proof}
\begin{proposition}\label{rho_2}
Let $y_l\in H_2(L_lM)$ be a generator. Then, the equality $\mr{ev}_*\circ\rho_*([S^1]\otimes y_l)=[M]$ holds.
\end{proposition}
\begin{proof}
Let $p\colon S^3\ra M$ be the canonical quotient map, viewing $S^3$ as a subspace of $\mathbb{C}^2$. Let $(f_1,f_2)$ be an element in $\Omega S^3$. For a loop $(f_1,f_2)\in \Omega S^3$, the composition $p\circ (f_1,f_2)$ lies in $\Omega M$. Thus, we obtain a homeomorphism $\Omega_0 M\ra\Omega_l M$ induced by the self-homomorphism
\begin{align*}
\phi\colon (x_1,x_2)\mapsto(e^{2\pi\sqrt{-1}lt/9}x_1,e^{2\pi\sqrt{-1}klt/9}x_2).
\end{align*}
The map $p\circ (e^{2\pi\sqrt{-1}lt/9}f_1,e^{2\pi\sqrt{-1}klt/9}f_2)$ represents the generator of $\pi_1(M)$. Indeed, for the generator $\alpha_l=p\circ (e^{2\pi\sqrt{-1}lt/9},e^{2\pi\sqrt{-1}klt/9})$, since $(f_1,f_2)$ is contractible, a homotopy $H$ from the constant loop at $(1,1)$ to $(f_1,f_2)$ induces a homotopy $p\circ\phi\circ H$ from $\alpha_l$ to $p\circ (e^{2\pi\sqrt{-1}l/9}f_1,e^{2\pi\sqrt{-1}kl/9}f_2)$.\\
Let $[\mr{id}]\in\pi_3(S^3)$ denote the identity class. Under the correspondence $\pi_3(S^3)\cong \pi_2(\Omega S^3)$, this class is represented by the map
\begin{align*}
(x_1,x_2,x_3)\mapsto \left[t\mapsto
\begin{pmatrix}
(x_1+\sqrt{-1}x_2)\sin (2\pi t)\\
x_3\sin(2\pi t)+\sqrt{-1}\cos(2\pi t)
\end{pmatrix}\right]
\end{align*}
which we denote by $F$.\\
Now we compute the degree of $F$ by treating it as a smooth map $F\colon S^2\times S^1\ra S^3$. Its degree is computed as the sum of the signs of the points in the preimage of a regular value (cf. \cite{Hirsch}). We choose $y=((1,0,0),1/4)\in S^2\times S^1$ and show that it is a regular point. Define $\tilde{F}\colon \mathbb{R}^3\times\mathbb{C}\ra \mathbb{C}^2$ which is an extension of $F$ and $g\colon \mathbb{R}^3\times\mathbb{C}\ra \mathbb{R}^2$ by
\begin{align*}
\tilde{F}\left(x_1,x_2,x_3,Re^{2\pi\sqrt{-1}t}\right)&=R
\begin{pmatrix}
(x_1+\sqrt{-1}x_2)\sin (2\pi t)\\
x_3\sin(2\pi t)+\sqrt{-1}\cos(2\pi t)
\end{pmatrix}\\
g\left(x_1,x_2,x_3,Re^{2\pi\sqrt{-1}t}\right)&=\left(\sqrt{x_1^2+x_2^2+x_3^2},R\right).
\end{align*}
Note that $g^{-1}((-1,1))=S^2\times S^1$ and $(1,1)$ is a regular value of $g$. Since
\begin{align*}
\mr{rank}(dF)_y=3-\dim\mr{Ker}(dF)_y=3-(5-\dim \mr{Ker}(d\tilde{F})_y\cap T_y(S^2\times S^1)),
\end{align*}
it suffices to compute $\mr{Ker}(d\tilde{F})_y\cap T_y(S^2\times S^1)$. Because
\begin{align*}
\mr{Ker}(dF)_y&=\mr{Ker}(d\tilde{F})_y\cap T_y(S^2\times S^1)\\
&=\mr{Ker}(d\tilde{F})_y\cap \mr{Ker}(dg)_y\\
&=\mr{Ker}
\begin{pmatrix}
(d\tilde{F})_y\\
(dg)_y
\end{pmatrix},
\end{align*}
the Jacobian matrix at $y$ is
\begin{align*}
\begin{pmatrix}
(d\tilde{F})_y\\
(dg)_y
\end{pmatrix}=
\begin{pmatrix}
1 & 0 & 0 & 0 & 0\\
0 & 1 & 0 & 0 & 0\\
0 & 0 & 1 & 0 & 0\\
0 & 0 & 0 & 0 & -2\pi\\
1 & 0 & 0 & 0 & 0\\
0 & 0 & 0 & 1 & 0
\end{pmatrix}.
\end{align*}
Thus, $\dim \mr{Ker}(d\tilde{F})_y\cap T_y(S^2\times S^1)=0$, and hence $\mr{rank}(dF)_y=3$. Since both $p$ and $\phi$ are local diffeomorphisms and $[e^{\pi\sqrt{-1}l/9},0]$ is a regular value of $p\circ\phi\circ F$, $(p\circ\phi\circ F)^{-1}([e^{\pi\sqrt{-1}l/9},0])$ consists of exactly one point. Hence, the degree of $p\circ \phi\circ F\colon S^2\times S^1\ra M$ is $1$.\\
Finally, we consider the homology of $L_lM$. Since $\Omega_lM$ is simply connected, the Hurewicz theorem yields $\pi_2(\Omega_l M)\cong H_2(\Omega_lM)$. Moreover, the map $H_2(\Omega_l M)\xrightarrow{\iota_*}H_2(L_lM)$ is an isomorphism, so a generator of $H_2(L_lM)$ is $(\iota\circ p\circ\phi\circ F)_*[S^2]$. Using $\mr{ev}\circ\rho(\mr{id}\times p\circ\phi\circ F)= p\circ\phi\circ F$, we obtain
\begin{align*}
\mr{ev}_*\rho_*([S^1]\otimes(\iota\circ p\circ\phi\circ F)_*[S^2])=(p\circ\phi\circ F)_*([S^1\times S^2])=[M]
\end{align*}
as required.
\end{proof}
Combining these two propositions and Proposition \ref{rho_0}, we compute $H_r^{S^1}(L_lM)$ for $r=0,1,2,3$. Under the natural identification $E_{p,q}^2=H_p(BS^1)\otimes H_q(L_lM)$. We see that the low-degree $E^2$-page is:
\begin{align*}
\begin{array}{c|ccccc}
3 & \Z\oplus\Z/9\Z & 0 & \Z\oplus\Z/9\Z & 0 & \Z\oplus\Z/9\Z\\
2 & \Z & 0 & \Z & 0 & \Z\\
1 & \Z/9\Z & 0 & \Z/9\Z & 0 & \Z/9\Z\\
0 & \Z & 0 & \Z & 0 & \Z\\
\hline
& 0 & 1 & 2 & 3 & 4
\end{array}
\end{align*}
Because $L_lM$ is a path-connected component, $H_0^{S^1}(L_lM)\cong \Z$. Moreover, Proposition \ref{rho_0} gives $\mr{Im}d_{2,0}^2\cong l~\Z/9\Z$, hence $H_1^{S^1}(L_lM)=0$ for $l\neq 3,6$ and $H_1^{S^1}(L_lM)\cong \Z/3\Z$ for $l=3,6$\footnotemark. $H_2^{S^1}(L_lM)$ is isomorphic to $\Z^2$. From proposition \ref{rho_2}, $E_{0,3}^3\cong \Z/9\Z$ and $E_{0,3}^5\cong (\Z/9\Z)/\mr{Im}d_{4,0}^4$ satisfy. In addition, $E_{1,2}^\infty=E_{3,0}^\infty=0$ holds and $E_{2,1}^3=0$ for $l\neq 3,6$ and $E_{2,1}^3\cong \Z/3\Z$ for $l=3,6$. Therefore, $H_3^{S^1}(L_lM)$ satisfies the exact sequence
\begin{align*}
0\ra E_{0,3}^5 \ra H_3^{S^1}(L_lM)\ra E_{2,1}^3\ra 0.
\end{align*}
\footnotetext{This phenomenon arises from the fact that $9$ is not prime, which allows nontrivial $S^1$-equivariant homology in the components $L_3M$and $L_6M$, and it does not occur for prime lens spaces such as $L(7;k)$.}
\subsection{Computation of $\tau_*$}
We determine on the image of $\tau_*$ by considering the Gysin exact sequence for $S^1$-bundle $ES^1\times L_lM\ra ES^1\times_{S^1} L_lM$. We consider $l\neq 3,6$. Since the diagram
\[
\begin{tikzcd}
H_2^{S^1}(L_lM)\arrow[r,"\tau_*"]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& H_3(L_lM) \arrow[r,"\pi_*"]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& H_3^{S^1}(L_lM)\arrow[r] \arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]&H_1^{S^1}(L_lM)\arrow[d,"\cong"]\\
\Z^2\arrow[r]& \Z\oplus\Z/9\Z\arrow[r,->>]& (\Z/9\Z)/\mr{Im}d_{4,0}^4\arrow[r]& 0
\end{tikzcd}
\]
commutes and the row is exact, $\tau_*$ is the surjection.\\
For $l=3,6$, the image of $\pi_*$ coincides with $H_3(L_lM)\twoheadrightarrow E_{0,3}^\infty\hookrightarrow H_3^{S^1}(L_lM)$ (cf. \cite[Theorem 5.9]{Hatcher}). Hence, $\mr{Im}\pi_*\cong (\Z/9\Z)/\mr{Im}d_{4,0}^4$ holds. Thus, we obtain
\[
\begin{tikzcd}
H_2^{S^1}(L_lM)\arrow[r,"\tau_*"]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& H_3(L_lM) \arrow[r,->>,"\pi_*"]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& E_{0,3}^5\arrow[r] \arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]&0\arrow[d,"\cong"]\\
\Z^2\arrow[r]& \Z\oplus\Z/9\Z\arrow[r,->>]& (\Z/9\Z)/\mr{Im}d_{4,0}^4\arrow[r]&0
\end{tikzcd}
.
\]
Therefore, for $l\neq 0$ and a generator $y_l\in H_2(L_lM)$, there exists $m\in\Z$ such that $\tau_*\circ\pi_*(y_l)=[\rho_{l,m}]$ holds.
\subsection{Computation of $\pi_*$}\label{pi_*}
For the $S^1$-bundle $\pi\colon ES^1\times L_lM\rightarrow ES^1\times_{S^1}L_lM$, we compute $\pi_*\colon H_r(L_lM)\ra H_r^{S^1}(L_lM)$ for $r=0,1,3$.\\
For $r=0$, the map $\pi_*$ is an isomorphism by the Gysin exact sequence of the $S^1$-bundle $ES^1\times L_lM\ra ES^1\times_{S^1} L_lM$.\\
Next, we consider $r=1$. If $l\neq 3,6$, then $H_1^{S^1}(L_lM)=0$ and hence $\pi_*\otimes\pi_*\circ\vee(j_*[\rho_{l,m}])=0$. By contrast, for $l=3,6$, the Gysin exact sequence of the $S^1$-bundle $ES^1\times L_lM\ra ES^1\times_{S^1} L_lM$ near degrees $0,1$ yields:
\[
\begin{tikzcd}
H_1(L_lM)\arrow[r,->>,"\pi_*"]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& H_1^{S^1}(L_lM) \arrow[r]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& 0\arrow[d,"\cong"]\\
\Z/9\Z\arrow[r,->>]& \Z/3\Z\arrow[r]&0
\end{tikzcd}
\]
and
\[
\begin{tikzcd}
H_0(L_lM)\arrow[r,->>,"\pi_*"]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& H_0^{S^1}(L_lM) \arrow[r]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& 0\arrow[d,"\cong"]\\
\Z\arrow[r,->>]& \Z\arrow[r]&0
\end{tikzcd}
.
\]
From these commutative diagrams, the image of $ct^pt_2^q~(c\in\Z/9\Z)$ by $\pi_*$ is $0$ for $p\neq 3,6$ and $c'\alpha_p\otimes \beta_q$ for $p=3,6$. Here, $c'$ is the value of $c$ in modulo $3$, and $\alpha_p,\beta_q$ are the generator of $H_1^{S^1}(L_lM),~H_0^{S^1}(L_lM)$.\\
For $r=3$, we will use the following proposition. As a consequence, when $l\neq 3,6$, there exists $y_l'\in H_2^{S^1}(L_lM)$ such that $\tau_*(y_l')=[\rho_{l,1+n_ll}]-[\rho_{l,1+l}]$.
\begin{proposition}
$\mr{Im}d_{4,0}^4$ is isomorphic to $l\Z/9\Z$.
\end{proposition}
\begin{proof}
Let $c\in H^2(\Omega_lM)$ be a generator. Using the identification $H^*(L_lM)\cong H^*(M)\otimes H^*(\Omega_lM)$ and the cohomology ring of $\Omega_lM$, define $1\otimes c\cap\colon H_*(L_lM)\ra H_{*-2}(L_lM)$. Define $\tilde{c}\in H^2(ES^1\times_{S^1}L_lM,L_lM)$ by $\delta^*c+\pi^*e$, where $\delta$ is the connecting morphism in the long exact sequence in cohomology for the pair $(ES^1\times_{S^1} L_lM, L_lM)$, and $e\in H_2(BS^1)$ is the Euler class. Then the diagram
\[
\begin{tikzcd}
H_4(BS^1)\arrow[d,"e\cap"]\arrow[rd, phantom, "\circlearrowright"]&H_4(ES^1\times_{S^1}L_lM)\arrow[l,"\pi_*"']\arrow[r,"\partial_*"]\arrow[d,"\tilde{c}\cap"]\arrow[rd, phantom, "\circlearrowright"]&H_3(L_lM)\arrow[d,"c\cap"]\\
H_2(BS^1)&H_2(ES^1\times_{S^1}L_lM)\arrow[l,"\pi_*"']\arrow[r,"\partial_*"]&H_1(L_lM)
\end{tikzcd}
\]
commutes.
By \cite[Proposition 5.14]{Hatcher}, the transgression $d_{p,0}^p$ corresponds to $\partial_*\circ\pi_*^{-1}\colon \mr{Im}\pi_*\ra H_{p-1}(L_lM)/\partial_*(\mr{Ker}\pi_*)$. Using naturality of the cap product, we obtain
\begin{align*}
c\cap d_4(u_4)=d_2(e\cap u_4)=d_2(u_2)=lx_l.
\end{align*}
Let $\omega$ be a generator of $H_2(\Omega_lM)$. Since $\iota_*^{-1}(x_l)\otimes\omega\in H_1(M)\otimes H_2(\Omega_lM)\cong H_3(L_lM)$ is a generator of $E_{0,3}^4$, $c\cap (\iota_*^{-1}(x_l)\otimes\omega)=\iota_*^{-1}(x_l)\otimes 1=x_l$ holds. Therefore $c\cap$ is an isomorphism. Therefore,
\begin{align*}
d_4(u_4)=(c\cap)^{-1}c\cap d_4(u_4)=(c\cap)^{-1}(lx_l)=l\iota_*^{-1}(x_l)\otimes\omega \in E_{0,3}^4.
\end{align*}
This proves the claim.
\end{proof}
%%%%%%%%%%%%%
\subsection{Computation of the string cobracket on $L(9;1)$ and $L(9;4)$}
Let $K=\mr{Ker}(H_3(LM)\ra \bigoplus_{l\in\Z/9\Z}H_3(M))$. Using the formula in \ref{string_coproduct_calc}, we compute $\mr{pr}_1\circ\vee(j_*[\rho_{l,m}])\mod {\mr{pr}_1\circ\vee(K)}$ for $l\neq 0$. The resulting expressions, organized by $l$, are listed in Table \ref{string_coproduct_1} for $k=1$ and Table \ref{string_coproduct_4} for $k=4$.
\begin{table}[htbp]
\centering
\begin{align*}
\begin{array}{c|l}
l&\mr{pr}_1\circ\vee(j_*[\rho_{l,m}])\\
\hline
1&0\\
2&2t^{1}t_2^{1}\frac{dt}{t}\\
3&2t^{1}t_2^{2}+2t^{2}t_2^{1}\frac{dt}{t}\\
4&2t^{1}t_2^{3}+2t^{2}t_2^{2}+2t^{3}t_2^{1}\frac{dt}{t}\\
5&2t^{1}t_2^{4}+2t^{2}t_2^{3}+2t^{3}t_2^{2}+2t^{4}t_2^{1}\frac{dt}{t}\\
6&2t^{1}t_2^{5}+2t^{2}t_2^{4}+2t^{3}t_2^{3}+2t^{4}t_2^{2}+2t^{5}t_2^{1}\frac{dt}{t}\\
7&2t^{1}t_2^{6}+2t^{2}t_2^{5}+2t^{3}t_2^{4}+2t^{4}t_2^{3}+2t^{5}t_2^{2}+2t^{6}t_2^{1}\frac{dt}{t}\\
8&2t^{1}t_2^{7}+2t^{2}t_2^{6}+2t^{3}t_2^{5}+2t^{4}t_2^{4}+2t^{5}t_2^{3}+2t^{6}t_2^{2}+2t^{7}t_2^{1}\frac{dt}{t}
\end{array}
\end{align*}
\caption{String coproduct for $k=1$}
\label{string_coproduct_1}
\end{table}
\begin{table}[htbp]
\centering
\begin{align*}
\begin{array}{c|l}
l&\mr{pr}_1\circ\vee(j_*[\rho_{l,m}])\\
\hline
1&7t^{3}t_2^{7}+7t^{5}t_2^{5}+7t^{7}t_2^{3}\frac{dt}{t}\\
2&8t^{1}t_2^{1}+7t^{3}t_2^{8}+7t^{4}t_2^{7}+7t^{5}t_2^{6}+7t^{6}t_2^{5}+7t^{7}t_2^{4}+7t^{8}t_2^{3}\frac{dt}{t}\\
3&8t^{1}t_2^{2}+8t^{2}t_2^{1}+7t^{4}t_2^{8}+5t^{5}t_2^{7}+7t^{6}t_2^{6}+5t^{7}t_2^{5}+7t^{8}t_2^{4}\frac{dt}{t}\\
4&6t^{1}t_2^{3}+8t^{2}t_2^{2}+6t^{3}t_2^{1}+5t^{5}t_2^{8}+5t^{6}t_2^{7}+5t^{7}t_2^{6}+5t^{8}t_2^{5}\frac{dt}{t}\\
5&6t^{1}t_2^{4}+6t^{2}t_2^{3}+6t^{3}t_2^{2}+6t^{4}t_2^{1}+5t^{6}t_2^{8}+3t^{7}t_2^{7}+5t^{8}t_2^{6}\frac{dt}{t}\\
6&4t^{1}t_2^{5}+6t^{2}t_2^{4}+4t^{3}t_2^{3}+6t^{4}t_2^{2}+4t^{5}t_2^{1}+3t^{7}t_2^{8}+3t^{8}t_2^{7}\frac{dt}{t}\\
7&4t^{1}t_2^{6}+4t^{2}t_2^{5}+4t^{3}t_2^{4}+4t^{4}t_2^{3}+4t^{5}t_2^{2}+4t^{6}t_2^{1}+3t^{8}t_2^{8}\frac{dt}{t}\\
8&2t^{1}t_2^{7}+4t^{2}t_2^{6}+2t^{3}t_2^{5}+4t^{4}t_2^{4}+2t^{5}t_2^{3}+4t^{6}t_2^{2}+2t^{7}t_2^{1}\frac{dt}{t}
\end{array}
\end{align*}
\caption{String coproduct for $k=4$}
\label{string_coproduct_4}
\end{table}
\newpage
Next, using Proposition~\ref{pi_*}, we compute $\mr{pr}_1\circ\vee_{S^1}(j_*\pi_*y_l) \mod {\mr{pr}_1\circ\vee_{S^1}(K)}$ as \ref{string_cobracket_1} and \ref{string_cobracket_4}. Note that the string coproduct on $L_0M$ is the zero map by definition and the following diagram commutes
\[
\begin{tikzcd}
H_2^{S^1}(LM)\arrow[r,"j_*"]\arrow[d,"\tau_*"]\arrow[rd, phantom, "\circlearrowright"]&H_2^{S^1}(LM,M)\arrow[d,"\tau_*"]\\
H_3(LM)\arrow[r,"j_*"] &H_3(LM,M).
\end{tikzcd}
\]
From the following long exact sequences of a pair
\[
\begin{tikzcd}
H_2^{S^1}(M)\arrow[r]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& H_2^{S^1}(L_0M)\arrow[r]\arrow[d,"\cong"]& H_2^{S^1}(L_0M,M)\arrow[r]& H_1^{S^1}(M)\arrow[r,"\cong"]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]&H_1^{S^1}(L_0M)\arrow[d,"\cong"]\\
\Z \arrow[r]& \Z^2&& \Z/9\Z\arrow[r,"\cong"]&\Z/9\Z,
\end{tikzcd}
\]
the generator of $H_2^{S^1}(LM,M)$ are $j_*\pi_*y_l$ and $j_*y_l'$ for $l\neq 0$ and $j_*y_0'$. Therefore, we consider only $\vee_{S^1}(j_*\pi_*y_l)$ and $\vee_{S^1}(j_*y_l')$.
\newpage
\begin{table}[htbp]
\centering
\begin{minipage}[t]{0.48\textwidth}
\begin{align*}
\begin{array}{c|l}
l & \mr{pr}_1\circ\vee_{S^1}(\pi_*y_l) \mod {\mr{pr}_1\circ\vee_{S^1}(K)}\\
\hline
1 & 0\\
2 & 0\\
3 & 0\\
4 & 2\alpha_3\otimes\beta_1\\
5 & 2\alpha_3\otimes\beta_2\\
6 & 2\alpha_3\otimes\beta_3\\
7 & 2\alpha_3\otimes\beta_4+2\alpha_6\otimes\beta_1\\
8 & 2\alpha_3\otimes\beta_5+2\alpha_6\otimes\beta_2
\end{array}
\end{align*}
\caption{String cobracket for $k=1$}
\label{string_cobracket_1}
\centering
\end{minipage}
\begin{minipage}[t]{0.48\textwidth}
\begin{align*}
\begin{array}{c|l}
l & \mr{pr}_1\circ\vee_{S^1}(\pi_*y_l) \mod {\mr{pr}_1\circ\vee_{S^1}(K)}\\
\hline
1 & \alpha_3\otimes\beta_7\\
2 & \alpha_3\otimes\beta_8+\alpha_6\otimes\beta_5\\
3 & \alpha_6\otimes\beta_6\\
4 & 2\alpha_6\otimes\beta_7\\
5 & 2\alpha_6\otimes\beta_8\\
6 & \alpha_3\otimes\beta_3\\
7 & \alpha_3\otimes\beta_4+\alpha_6\otimes\beta_1\\
8 & 2\alpha_3\otimes\beta_5+\alpha_6\otimes\beta_2
\end{array}
\end{align*}
\caption{String cobracket for $k=4$}
\label{string_cobracket_4}
\end{minipage}
\end{table}
From Tables \ref{string_cobracket_1} and \ref{string_cobracket_4}, the number of indices for which $\mr{pr}_1\circ\vee_{S^1}(\pi_*y_l) \mod {\mr{pr}_1\circ\vee_{S^1}(K)}$ is nonzero differs between $k=1$ and $k=4$.\\
We now compute $\mr{pr}_1\circ\vee_{S^1}(K)$. For $l\neq 0,3,6$,
\begin{align*}
\mr{pr}_1\circ\vee_{S^1}(y_l')=r(n_l-1)l(\alpha_3\otimes\beta_{6-kl}+\alpha_6\otimes\beta_{9-kl})
\end{align*}
and for $l=0,3,6$, the image is zero. All indices are taken modulo $9$.
\begin{table}[htbp]
\begin{minipage}[t]{0.48\textwidth}
\centering
\begin{align*}
\begin{array}{c|l}
l & \mr{pr}_1\circ\vee_{S^1}(y_l')\\
\hline
1 & \alpha_3\otimes\beta_5+\alpha_6\otimes\beta_8\\
2 & \alpha_3\otimes\beta_4+\alpha_6\otimes\beta_7\\
3 & 0\\
4 & \alpha_3\otimes\beta_2+\alpha_6\otimes\beta_5\\
5 & \alpha_3\otimes\beta_1+\alpha_6\otimes\beta_4\\
6 & 0\\
7 & \alpha_3\otimes\beta_8+\alpha_6\otimes\beta_2\\
8 & \alpha_3\otimes\beta_7+\alpha_6\otimes\beta_1
\end{array}
\end{align*}
\caption{String cobracket of $K$ for $k=1$}
\end{minipage}
\begin{minipage}[t]{0.48\textwidth}
\centering
\begin{align*}
\begin{array}{c|l}
l & \mr{pr}_1\circ\vee_{S^1}(y_l')\\
\hline
1 & \alpha_3\otimes\beta_2+\alpha_6\otimes\beta_5\\
2 & \alpha_3\otimes\beta_7+\alpha_6\otimes\beta_1\\
3 & 0\\
4 & \alpha_3\otimes\beta_8+\alpha_6\otimes\beta_2\\
5 & \alpha_3\otimes\beta_4+\alpha_6\otimes\beta_7\\
6 & 0\\
7 & \alpha_3\otimes\beta_5+\alpha_6\otimes\beta_8\\
8 & \alpha_3\otimes\beta_1+\alpha_6\otimes\beta_4
\end{array}
\end{align*}
\caption{String cobracket of $K$ for $k=4$}
\end{minipage}
\end{table}
Consequently, one can see that the string cobrackets for $k=1,4$ differ. We summarize our findings in the following.
\begin{theorem}
The string cobracket is not a homotopy invariant. More precisely, the number of components on which the string cobracket is nonzero differs between $L(9;1)$ and $L(9;4)$.
\end{theorem}
\begin{proof}
For $k=1$, the components on which $\mr{pr}_1\circ\vee_{S^1}(\pi_*y_l)$ is nonzero are $l=4,5,6,7,8$ and those on which $\mr{pr}_1\circ\vee_{S^1}(y_l')$ is nonzero are $l=1,2,4,5,7,8$. These lie in distinct components, so the total number of nonzero components is $11$.\\
For $k=4$, the components on which $\mr{pr}_1\circ\vee_{S^1}(\pi_*y_l)$ is nonzero are $l=1,2,3,4,5,6,7,8$ and those on which $\mr{pr}_1\circ\vee_{S^1}(y_l')$ is nonzero are $l=1,2,4,5,7,8$. Again, these lie in distinct components, giving a total of $14$.\\
Since the string cobracket is coassociative, composing with the projection $\mathrm{pr}_1$ does not change which components are nonzero. Therefore, the number of nonzero components of the string cobracket is $11$ for $k=1$ and $14$ for $k=4$.
\end{proof}
\begin{remark}
There exist pairs of lens spaces pairs for which the numbers of components in which the string coproduct are nonzero agree, while those of the string cobracket differ. For example, for lens spaces $L(21;2)$ and $L(21;8)$, the number of nonzero components for the string coproduct is $20$ in both cases, while the number of nonzero components for the string cobracket is $19$ and $20$, respectively. Thus, counting nonzero components is insufficient to capture the full algebraic structure of either operation. Note that a direct calculation for each natural number shows that $21$ is the smallest number for which such a pair exists.
\end{remark}
\section{Relationship between string cobracket and Whitehead torsion}\label{section4}
We now investigate how the string cobracket transforms under homotopy equivalence and how this behavior reflects Whitehead torsion. Fix a homotopy equivalence $f\colon L(9;1)\ra L(9;4)$ that sends the $l$-component of $L(9;1)$ to the $2l$-component of $L(9;4)$. As in \cite[Section 4.2]{Naef24}, we have the transformation formula for the string coproduct
\begin{align*}
\vee\circ f_*(x)=f_*\circ \vee(x)+f_*(x\wedge d\log \tau(f))
\end{align*}
where $\wedge$ is the string product, $\tau(f)$ is the Whitehead torsion of $f$, and $d\log \tau(f)$ is the image of the Dennis trace map of $f$. Using the commutativity of $f$ and $\tau_*$ and $\pi_*$, we obtain the transformation formula for the string cobracket
\begin{align*}
\vee_{S^1}\circ f_*(x)=f_*\circ \vee_{S^1}(x)+\pi_*\circ f_*(\tau_*(x)\wedge d\log\tau(f)).
\end{align*}
We now compute
\begin{align*}
\pi_* f(\tau_*(\pi_*y_l)\wedge d\log\tau(f)), \quad \pi_*y_l\in H_2^{S^1}(L_l(L(9;1))).
\end{align*}
Since
\begin{align*}
\tau(f)&=\frac{(t^7-1)(t-1)}{(t-1)(t-1)}= 1+t+\cdots +t^6\\
\tau(f)^{-1}&=\frac{t-1}{t^7-1}=\frac{t^{9\cdot 3+1}-1}{t^7-1}=1+t^3+t^5+t^7,
\end{align*}
we obtain
\begin{align*}
d\log\tau(f)&=(1+t^3+t^5+t^7)(1+2t+3t^2+\cdots +6t^5)\\
\equiv &2t^2+2t^3+t^4+t^5dt \in \Z/3\Z[t]/(t^9-1)dt.
\end{align*}
Since the operations on the coefficients are invariant, as well as $\pi_*$ and $f_*$, we may first work modulo $3$. After homogenizing, we compute $\pi_*\circ f_*(\tau_*(\pi_*y_l)\wedge d\log\tau(f))$. Here, “homogenizing’’ means rewriting the expression in terms of the homogeneous basis $t^it_2^j\frac{dt}{t}$ (with $i,j$ taken modulo $9$), so that each term lies in a well-defined component $H_1(L_iM)\otimes H_0(L_jM)$, and the projection $\pi _*$ can be applied. (We use the results of \cite[Appendix A]{Naef24}.)
\begin{align*}
\pi_*\circ f_*(\tau_*(\pi_*y_l)\wedge d\log\tau(f))&= \pi_*\circ f_*((t^{l}-t_2^{l})(2t^2t_2^6+2t^3t_2^5+t^4t_2^4+t^5t_2^3)dt)\\
&=\pi_*\left((t^{2l}-t_2^{2l})(2t^4t_2^{12}+2t^6t_2^{10}+t^8t_2^8+t^{10}t_2^6)2t^2 \frac{dt}{t}\right).
\end{align*}
For example, when $l=1$ we obtain
\begin{align*}
\pi_* f(\tau_*(\pi_*y_l)\wedge d\log\tau(f))&=2\alpha_3\otimes\beta_8-2\alpha_3\otimes\beta_8-2\alpha_6\otimes \beta_{14}+\alpha_{12}\otimes\beta_8\\
&=\alpha_6\otimes \beta_5+\alpha_3\otimes\beta_8.
\end{align*}
This agrees with the computation as Section \ref{section3}. This confirms that the string cobracket, like the string coproduct, detects Whitehead torsion via the Dennis trace map.\\
Finally, since the string coproduct is a simple-homotopy invariant \cite{NaefSafronov}, and the maps $\tau_*$ and $\pi_*$ are natural, it follows that the string cobracket is also invariant under simple–homotopy equivalence.
\section{Failure of Lie bialgebra structure}\label{section5}
We now show that the pair consisting of the string bracket and the string cobracket does not satisfy the compatibility condition of a Lie bialgebra in general. Recall that a Lie bialgebra consists of a Lie algebra structure and a Lie coalgebra structure on a vector space satisfying the following compatibility condition in the sense of Drinfeld:
\begin{align*}
\vee_{S^1}\circ\wedge_{S^1}(X,Y)=(\mr{ad}_X\otimes 1+1\otimes \mr{ad}_X)\vee_{S^1}(Y)-(-1)^{|X||Y|}(\mr{ad}_Y\otimes 1+1\otimes \mr{ad}_Y)\vee_{S^1}(X)
\end{align*}
for all $X,Y$. Here, $\mr{ad}_X Y=\wedge_{S^1}(X,Y)$. We show that this compatibility fails, in general.
\begin{theorem}
There exists an oriented compact finite-dimensional manifold $M$ such that $(H_*^{S^1}(LM,M),\wedge_{S^1},\vee_{S^1})$ does not form a Lie bialgebra.
\end{theorem}
\begin{proof}
Take $M=L(9;4)$, and set $X=\pi_*y_1$, $Y=\pi_*y_8$. Since
\[
\begin{tikzcd}
H_3^{S^1}(L_lM)\arrow[r,"\tau_*"]\arrow[d,"\cong"]\arrow[rd, phantom, "\circlearrowright"]& H_4(L_lM)\arrow[d,"\cong"]\\
(\Z/9\Z)/\mr{Im}d_{4,0}^4 \arrow[r] & \Z
\end{tikzcd}
\]
is the zero map, the left-hand side is
\begin{align*}
\vee_{S^1}\circ\wedge_{S^1}(X,Y)=\pi_*\circ\vee\circ\tau_*\circ \wedge_{S^1}(X,Y)=0
\end{align*}
We consider the right-hand side. Using Table \ref{string_cobracket_4}, we obtain
\begin{align*}
\mr{pr}_1\circ\vee_{S^1}(\pi_*(y_8))=\alpha_3\otimes\beta_1+\alpha_6\otimes\beta_4.
\end{align*}
The first term of the above formula is
\begin{align*}
(\mr{ad}_X\otimes 1+1\otimes \mr{ad}_X)(\alpha_3\otimes\beta_1)=\wedge_{S^1}(\pi_*y_1,\alpha_3)\otimes\beta_1+\alpha_3\otimes\wedge_{S^1}(\pi_*y_1,\beta_1).
\end{align*}
Since $\alpha_3$ is torsion, while $\tau_*\alpha_3\in H_2(L_3M)\cong\Z$ is satisfied, thus the map $\tau_*$ must be zero on $\alpha_3$, hence $\tau_*\alpha_3=0$. Moreover, since $\tau_*\beta_1=1t^1dt/t$, using \cite[Appendix A]{Naef24} again, we obtain
\begin{align*}
\alpha_3\otimes\wedge_{S^1}(\pi_*y_1,\beta_1)&=\pi_*\wedge([\rho_{1,m}],tdt/t)\\
&=\alpha_3\otimes (\pi_*t^2dt/t)\\
&=0.
\end{align*}
Since $H_1^{S^1}(L_lM)=0$ for $l\neq 3,6$, the remaining terms are computed similarly. The right-hand side of the compatibility condition becomes
\begin{align*}
&(\mr{ad}_X\otimes 1+1\otimes \mr{ad}_X)(\alpha_3\otimes\beta_1+\alpha_6\otimes\beta_4-\beta_1\otimes\alpha_3-\beta_4\otimes\alpha_6)\\
&-(\mr{ad}_Y\otimes 1+1\otimes \mr{ad}_Y)(\alpha_3\otimes\beta_7-\beta_7\otimes\alpha_3)\\
&=\alpha_3\otimes\pi_*(t^2dt/t)+4\alpha_6\otimes\pi_*(t^5dt/t)-\pi_*(t^2dt/t)\otimes\alpha_3-4\pi_*(t^5dt/t)\otimes\alpha_6\\
&-7\alpha_3\otimes\pi_*(t^{15}dt/t)+7\pi_*(t^{15}dt/t)\otimes\alpha_3\\
&=-\alpha_3\otimes\alpha_6+\alpha_6\otimes\alpha_3.
\end{align*}
Note that the degree of $X$ and $Y$ is one, and the sign of the second term is negative. Since the left-hand side and right-hand side differ, the compatibility condition fails. Therefore, the triple $(H_*^{S^1}(LL(9;4),L(9;4)),\wedge_{S^1},\vee_{S^1})$ does not form a Lie bialgebra.
\end{proof}
\bibliographystyle{amsalpha}
\bibliography{Isana_Sumoto_1_rev}
\end{document}