G-birationally rigid cubic threefolds

Extracted source text · arXiv 2604.20426

\documentclass[12pt]{amsart}
\usepackage{amscd,amssymb,amsthm,amsmath,amssymb,mathrsfs,enumerate,url,hyperref}
\usepackage{graphicx}
\usepackage[matrix,arrow,curve]{xy}
\usepackage[margin=2cm]{geometry}
\usepackage[normalem]{ulem}
\usepackage{verbatim}

\newtheorem{theorem}{Theorem}

\newtheorem*{theorem*}{Theorem}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem*{proposition*}{Proposition}
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem*{lemma*}{Lemma}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem*{corollary*}{Corollary}
\newtheorem{conjecture}[theorem]{Conjecture}
\newtheorem{question}[theorem]{Question}
\newtheorem{problem}[theorem]{Problem}
\newtheorem*{problem*}{Problem}

\theoremstyle{definition}
\newtheorem{example}[theorem]{Example}
\newtheorem*{construction*}{Construction}
\newtheorem*{example*}{Example}
\newtheorem{definition}[theorem]{Definition}
\newtheorem*{observation*}{Observation}

\theoremstyle{remark}
\newtheorem{remark}[theorem]{Remark}
\newtheorem*{remark*}{Remark}

\sloppy\pagestyle{plain}

\makeatletter\@addtoreset{equation}{section} \makeatother
\renewcommand{\theequation}{\thesection.\arabic{equation}}

\newcommand\mc{\mathcal}


\newcommand{\CM}{\mc M}

\author{Ivan Cheltsov, Igor Krylov, Sione Ma'u}

\thanks{Throughout this paper, all varieties are assumed to be complex and projective.}

\title{$G$-birationally rigid cubic threefolds}

\begin{document}

\begin{abstract}
We classify pairs $(X,G)$ consisting of a (possibly singular) cubic threefold $X\subset\mathbb{P}^4$ and a finite subgroup $G\subset\mathrm{Aut}(X)$ such that $X$ is $G$-birationally rigid, i.e., $X$ is a $G$-Mori fiber space (over a point), and $X$ is not $G$-birational to any $G$-Mori fibre space that is not $G$-biregular to $X$.
\end{abstract}


\address{ \emph{Ivan Cheltsov}\newline
\textnormal{University of Edinburgh, Edinburgh, Scotland}
\newline
\textnormal{\texttt{i.cheltsov@ed.ac.uk}}}


\address{\emph{Igor Krylov}
\newline
\textnormal{Institute for Basic Science, Pohang, Korea}
\newline
\textnormal{\texttt{krylov.igor.o@gmail.com}}}

\address{\emph{Sione Ma'u}
\newline
\textnormal{University of Auckland, Auckland, New Zealand}
\newline
\textnormal{\texttt{s.mau@auckland.ac.nz}}}


\maketitle

\tableofcontents

\section{Introduction}
\label{section:introduction}

The notion of birational rigidity originated in the~work of Iskovskikh and Manin \cite{IskovskikhManin}, where they implicitly proved that every smooth hypersurface of degree $4$ in $\mathbb{P}^4$ is birationally superrigid and, in particular, non-rational. Recall that a Fano variety $X$ is \emph{birationally rigid} if $X$ is a Mori fiber space over a point (that is, $X$ has terminal singularities and the~class group $\mathrm{Cl}(X)$ has rank~$1$), and $X$ is the~only Mori fiber space in its birational equivalence class. If, in addition, 
$$
\mathrm{Bir}(X)=\mathrm{Aut}(X),
$$
then $X$ is said to be \emph{birationally superrigid}. Birational rigidity is an obstruction to rationality.

Problems concerning the~irrationality of Fano varieties defined over non-algebraically closed fields lead naturally to an equivariant generalization of the~notion of birational rigidity. To formulate this notion, we fix a Fano variety $X$ that have at most terminal singularities, and we fix a finite subgroup $G\subset \mathrm{Aut}(X)$. Then the~Fano variety $X$ is said to be $G$-birationally rigid if the~following two conditions are satisfied:
\begin{enumerate}
\item $X$ is a~$G$-Mori fiber space over a~point. i.e., $X$ has terminal singularities and $\mathrm{rk}\,\mathrm{Cl}(X)^G=1$,
\item if $Y$ is a $G$-Mori fiber space that is $G$-birational to $X$, then $Y$ is $G$-biregular to $X$.
\end{enumerate}
If, in addition, we have 
$$
\mathrm{Bir}^G(X)=\mathrm{Aut}^G(X),
$$
then $X$ is said to be $G$-birationally superrigid. The~first result on $G$-birational rigidity of Fano varieties has been implicitly proved by Segre and Manin in  \cite{Segre,Manin1968}, where they studied irrationality of smooth cubic surfaces defined over an algebraically non-closed field.

\begin{theorem}[{Segre--Manin Theorem}]
\label{theorem:Manin-Segree}
Let $X$ be a~smooth cubic surface in $\mathbb{P}^3$, and let $G$ be a finite subgroup in $\mathrm{Aut}(X)$ such that $\mathrm{rk}\,\mathrm{Cl}(X)^G=1$. Then $X$ is $G$-birationally rigid.
\end{theorem}

\begin{remark}
\label{remark:del-Pezzo}
Cubic surfaces are del Pezzo surfaces of degree $3$. Segre--Manin Theorem has a natural generalization for del Pezzo surfaces of degree $\leqslant 2$. For the~classification of all $G$-birationally rigid del Pezzo surfaces, see \cite{Cheltsov2008,DolgachevIskovskikh2009,Cheltsov2014,Sakovich,Pinardin2024,Yasinsky2025,PSY,CheltsovTschinkelZhang2026}.
\end{remark}

In this paper, we generalize Segree--Manin theorem for three-dimensional cubics.

\begin{theorem}[Main Theorem]
\label{theorem:main}
Let $X$ be a smooth cubic threefold in $\mathbb{P}^4$, and let $G$ be a finite subgroup in $\mathrm{Aut}(X)$. Then the~following conditions are equivalent:
\begin{itemize}
\item[$(\mathrm{a})$] $\mathbb{P}^4$ contains no $G$-invariant planes, and $G\not\simeq C_5\rtimes C_4, C_3\times(C_5\rtimes C_4), C_{11}\rtimes C_5$;
\item[$(\mathrm{b})$] $X$ is $G$-birationally superrigid;
\item[$(\mathrm{c})$] $X$ is $G$-birationally rigid.
\end{itemize}
\end{theorem}

In this theorem, we consider $G$ as a subgroup of $\mathrm{PGL}_5(\mathbb{C}$), since the~$G$-action lifts to $\mathbb{P}^4$.

All possibilities for the~group $G$ in Theorem~\ref{theorem:main} are known \cite{WeiYu2020}. Namely, if $G$ is a finite group that faithfully  acts on a smooth cubic threefold, it follows from \cite{WeiYu2020} that $G$ is isomorphic to a~subgroup of one of the~following six finite groups:
\begin{enumerate}
\item $C_4^3\rtimes\mathfrak{S}_5$, which is the~automorphism group of the~Fermat cubic threefold
\begin{equation}
\label{equation:Fermat}\tag{$\clubsuit$}
\big\{x_0^3+x_1^3+x_2^3+x_3^3+x_4^3=0\big\}\subset\mathbb{P}^4;
\end{equation}
\item $\mathrm{PSL}_2(\mathbf{F}_{11})$, which is the~automorphism group of the~Klein cubic threefold
\begin{equation}
\label{equation:Klein}\tag{$\spadesuit$}
\big\{x_0x_1^2+x_1x_2^2+x_2x_3^2+x_3x_4^2+x_4x_0^2=0\big\}\subset\mathbb{P}^4;
\end{equation}
\item $((C_3^2\rtimes C_3)\rtimes C_4)\times\mathfrak{S}_3$, which is the~finite group with GAP ID [648,541];
\item $C_3\times\mathfrak{S}_5$, which is the~automorphism group of the~cubic threefold
\begin{equation}
\label{equation:360-119}\tag{$\maltese$}
\big\{x_0+x_1+x_2+x_3+x_4=0,x_0^3+x_1^3+x_2^3+x_3^3+x_4^3+x_5^3=0\big\}\subset\mathbb{P}^5;
\end{equation}
\item $C_{24}$;
\item $C_{16}$.
\end{enumerate}
Using this classification and the proof of Theorem~\ref{theorem:main}, we obtain the following classification.

\begin{corollary}
\label{corollary:main}    
Let $X$ be a smooth cubic threefold in $\mathbb{P}^4$, and let $G$ be a finite subgroup in $\mathrm{Aut}(X)$. Then $X$ is $G$-birationally rigid if and only if one of the following cases holds:
\begin{enumerate}
\item[$(\mathrm{i})$] $X$ is the Fermat cubic \eqref{equation:Fermat}, and $G$ is a subgroup listed in Figure~\ref{figure:26} such that \mbox{$G\not\simeq C_5\rtimes C_4$}, and $\mathbb{P}^4$ contains no $G$-invariant planes (generators of a subgroup conjugated to $G$ can be found using Magma code provided in Appendix~\ref{subsection:plane-free-groups});
\item[$(\mathrm{ii})$] $X$ is the Klein cubic \eqref{equation:Klein}, and $G\simeq\mathrm{PSL}_2(\mathbf{F}_{11}),\mathfrak{A}_5$;
\item[$(\mathrm{iii})$] $X$ is the cubic \eqref{equation:360-119}, and $G\simeq C_3\times\mathfrak{S}_5, C_3\times\mathfrak{A}_5, \mathfrak{S}_5, \mathfrak{A}_5$,
\item[$(\mathrm{iv})$] $X$ is the cubic
$$
\Big\{x_0^3+x_1^3+x_2^3+x_3^3+x_4^3+ax_3(x_0x_2+x_1x_4)=0\Big\}\subset\mathbb{P}^4
$$
for some $a\in \mathbb{C}$, and $G\simeq C_3^2\rtimes C_4, \mathfrak{S}_3^2,\mathfrak{S}_3\wr C_2$ such that $\mathbb{P}^4$ does not contain $G$-invariant planes (generators of a subgroup conjugated to $G$ can be found in Section~\ref{section:36-9-36-10}).  
\end{enumerate}
\end{corollary}

\begin{remark}
\label{remark:Costya}    
If $X$ is the~Klein cubic threefold \eqref{equation:Klein} and $G\simeq\mathrm{PSL}_2(\mathbf{F}_{11})$, then $X$ is $G$-birationally superrigid by \cite[Theorem~A.5]{CheltsovShramov2014}. But we will not use this result in the proof of Theorem~\ref{theorem:main}.
\end{remark}

As a by-product of the proof of Theorem~\ref{theorem:main} (Main Theorem), we obtain the~following result, which completes the investigations started in \cite{CheltsovShramov2014,Avilov2016b,Avilov2018,CheltsovTschinkelZhang2025}. 

\begin{theorem}
\label{theorem:singular}
Let $X$ be a singular cubic threefold in $\mathbb{P}^4$ that has at most terminal singularities, and let $G$ be a finite subgroup in $\mathrm{Aut}(X)$ such that $\mathrm{rk}\,\mathrm{Cl}(X)^G=1$. Then $X$ is $G$-birationally rigid if and only if $X$ is $G$-birationally superrigid, which happens if and only if one of the~following holds:
\begin{itemize}
\item[$(\mathrm{v})$] $X$ is the~(ten nodal) Segre cubic threefold
\begin{equation}
\label{equation:cubic-Segre}\tag{$\diamondsuit$}
\big\{x_0+x_1+x_2+x_3+x_4+x_5=0,x_0^3+x_1^3+x_2^3+x_3^3+x_4^3+x_5^3=0\big\}\subset\mathbb{P}^5,
\end{equation}
and $G\simeq\mathfrak{S}_6,\mathfrak{A}_6$, or $G\simeq\mathfrak{S}_5,\mathfrak{A}_5$ and $G$ leaves invariant a hyperplane in $\mathbb{P}^4$;

\item[$(\mathrm{vi})$] $X$ is the~(nine-nodal) cubic threefold
\begin{equation}\tag{$\heartsuit$}
\label{equation:cubic-9-nodes}
\big\{x_0+x_1+x_2=x_3+x_4+x_5, x_0x_1x_2=x_3x_4x_5\big\}\subset\mathbb{P}^5,
\end{equation}
and $G\simeq\mathfrak{S}_3^2\rtimes C_3, C_3^2\rtimes C_4$, or $G\simeq\mathfrak{S}_3^2$ and $G$ acts transitively on $\mathrm{Sing}(X)$.
\end{itemize}
\end{theorem}

\begin{proof}
If $X$ is the~cubic \eqref{equation:cubic-9-nodes}, and $G\simeq C_3^2\rtimes C_4$, or $G\simeq\mathfrak{S}_3^2$ and $G$ acts transitively on $\mathrm{Sing}(X)$, then the cubic threefold $X$ is $G$-birationally superrigid by Theorem~\ref{theorem:36-9}, which is proved in Section~\ref{section:36-9-36-10}. The remaining assertions are proved in \cite{CheltsovShramov2014,Avilov2016b,Avilov2018,CheltsovTschinkelZhang2025}.
\end{proof}

It would be natural to generalize both Theorems~\ref{theorem:main} and \ref{theorem:singular} for del Pezzo threefolds. Let us recall their classification. Let $X$ a Fano threefold with terminal Gorenstein singularities such that 
$$
-K_X\sim 2H,
$$
where $H$ is a Cartier divisor on $X$. Suppose that $\mathrm{rk}\,\mathrm{Cl}(X)^G=1$ for a finite subgroup $G\subset\mathrm{Aut}(X)$. Set $d=H^3$. Then $d\in\{1,2,3,4,5,6,8\}$, the threefold $X$ is a del Pezzo threefold of degree $d$, and it follows from \cite{Prokhorov2013a,Prokhorov2013b,Prokhorov2015,KuznetsovProkhorov2023} that one of the following cases holds:
\begin{itemize}
\item[$(\mathrm{I})$] $d=1$ and $X$ is a sextic hypersurface in $\mathbb{P}(1,1,1,2,3)$;
\item[$(\mathrm{II})$] $d=2$ and $X$ is a quartic hypersurface in $\mathbb{P}(1,1,1,1,2)$; 
\item[$(\mathrm{III})$] $d=3$ and $X$ is a cubic hypersurface in $\mathbb{P}^4$;
\item[$(\mathrm{IV})$] $d=4$ and $X$ is a complete intersection of two quadrics in $\mathbb{P}^5$;
\item[$(\mathrm{V})$] $d=5$ and $X$ is a smooth intersection of the Grassmannian
$\mathrm{Gr}(2,5)\hookrightarrow\mathbb{P}^9$  with a linear subspace of dimension $6$ ($X$ is unique up to an isomorphism, and $\mathrm{Aut}(X)\simeq\mathrm{PGL}_2(\mathbb{C})$);
\item[$(\mathrm{VI})$] $d=6$ and $X=\mathbb{P}^1\times\mathbb{P}^1\times \mathbb{P}^1$;
\item[$(\mathrm{VI}^\prime)$] $d=6$ and $X=\{x_0y_0+x_1y_1+x_2y_2=0\}\subset\mathbb{P}^2_{x_0,x_1,x_2}\times\mathbb{P}^2_{y_0,y_1,y_2}$; 
\item[$(\mathrm{VIII})$] $d=8$ and $X=\mathbb{P}^3$ (this case is usually excluded).
\end{itemize}
It seems feasible to find all possibilities for $X$ and $G$ such that $X$ is $G$-birationally rigid for $d\ne 3$. For $X=\mathbb{P}^3$, this is done in \cite{CheltsovShramov2012,CheltsovShramov2014,CheltsovShramov2019}. For $X=\mathbb{P}^1\times\mathbb{P}^1\times \mathbb{P}^1$, this is done in \cite{CheltsovDuboulozKishimoto2023} under an assumption that $|G|\geqslant 32\times 24^4$. If $d=5$, $X$ is $G$-birationally rigid if and only if $G\simeq\mathfrak{A}_5$ \cite{CheltsovShramov2016}. For $d\in\{1,2,4\}$, partial results are obtained in \cite{CheltsovShramovV5,Avilov2016a,Avilov2019,Avilov2023,Cheltsov2023,AbbanCheltsovParkShramov2025}.

The structure of this paper is simple. In Section~\ref{section:proof}, we prove Theorem~\ref{theorem:main}  modulo $5$ results, which are proved in Sections~\ref{section:20-3}, \ref{section:55-1}, \ref{section:405-15}, \ref{section:A5},~\ref{section:36-9-36-10}. In~Appendix~\ref{section:Magma}, we present Magma codes used in the~proof.

\medskip
\noindent
\textbf{Notation.}
Throughout the~paper, we denote by $C_n$ the~cyclic group of order $n$, and by  $\mathfrak{D}_n$ the~dihedral group of order~$2n\geqslant 6$.
Similarly, we denote by $\mathfrak{S}_n$ the~symmetric group of degree $n$, and we denote by $\mathfrak{A}_n$ its alternating subgroup. We denote by $\zeta_n$ a primitive $n$-th root of unity.


\medskip
\noindent
\textbf{Acknowledgements.}
Most of the~work presented in this paper was carried out during a visit by the~first two authors to the~University of Auckland (Tamaki Makaurau, Aotearoa) in March~2026. We thank the~university for providing an excellent environment for conducting research. 

We are very grateful to Paolo Cascini and Zhijia Zhang for many useful discussions. 

The~first author has been supported by Simons Collaboration grant \emph{Moduli of varieties}.
The~second author was supported by \emph{IBS-R003-D1} grant.
\begin{center}
\begin{figure}
\caption{Plane free groups acting on Fermat cubic}
  \label{figure:26}
\includegraphics[height=\textheight]{graphfigure.pdf}
\end{figure}
\end{center}



\section{The proof of Main Theorem}
\label{section:proof}

Let $X$ be a smooth cubic threefold in $\mathbb{P}^4$, and let $G$ be a subgroup in $\mathrm{Aut}(X)$. Then $G$ is finite, and its action on $X$ lifts to $\mathbb{P}^4$, so we consider $G$ as a subgroup in $\mathrm{PGL}_5(\mathbb{C})$. To prove Theorem~\ref{theorem:main}, we must show that the~following three conditions are equivalent:
\begin{itemize}
\item[$(\mathrm{a})$] $\mathbb{P}^4$ does not contain $G$-invariant planes, $G\not\simeq C_5\rtimes C_4, C_3\times(C_5\rtimes C_4), C_{11}\rtimes C_5$;
\item[$(\mathrm{b})$] $X$ is $G$-birationally superrigid;
\item[$(\mathrm{c})$] $X$ is $G$-birationally rigid.
\end{itemize}
The implication $(\mathrm{b})\Rightarrow (\mathrm{c})$ follows from the~definitions. If $G\simeq C_5\rtimes C_4$ or $G\simeq C_3\times(C_5\rtimes C_4)$, then the~cubic threefold $X$ is not $G$-birationally rigid by Theorem~\ref{theorem:20-3}, which will be proved in Section~\ref{section:20-3}. Similarly, if $G\simeq C_{11}\rtimes C_5$, then the~cubic threefold $X$ is not $G$-birationally rigid by Theorem~\ref{theorem:Klein-cubic}, which will be proved in Section~\ref{section:55-1}. Therefore, the~implication $(\mathrm{c})\Rightarrow (\mathrm{a})$ follows from

\begin{lemma}
\label{lemma:Pn-pencil}
Suppose that $\mathbb{P}^4$ contains a $G$-invariant plane. Then~$X$ is not $G$-birationally rigid.
\end{lemma}

\begin{proof}
Let $\mathcal{P}$ be the~pencil in $|\mathcal{O}_X(1)|$ consisting of all surfaces that are cut out by hyperplanes that pass through $\Pi$.
Then $\mathcal{P}$ is mobile, and all its members are irreducible cubic surfaces.
Let $\psi\colon X\dasharrow\mathbb{P}^1$ be the~map given by the pencil~$\mathcal{P}$. Then $\psi$ is $G$-equivariant and dominant.
Moreover, equivariantly resolving the~indeterminacy of this map, we get the~following $G$-equivariant commutative diagram:
$$
\xymatrix{
& Y\ar@{->}[ld]_\pi\ar@{->}[rd]^{\phi} & \\
X\ar@{-->}[rr]^\psi  && \mathbb{P}^1}
$$
where $Y$ is a smooth threefold, $\pi$ is a birational morphism, and $\phi$ is a surjective morphism, whose general fiber is birational to a cubic surface. Now, applying relative $G$-Minimal Model Program to~$Y$ over~$\mathbb{P}^1$, we obtain a $G$-birational map from $Y$ to a $G$-Mori fiber space with a positive-dimensional base, which implies that~$X$ is not $G$-birationally rigid.
\end{proof}

Thus, to complete the~proof of Theorem~\ref{theorem:main}, it is enough to prove the~implication $(\mathrm{a})\Rightarrow (\mathrm{b})$. Hence, we suppose that $\mathbb{P}^4$ does not contain $G$-invariant planes, $G\not\simeq C_5\rtimes C_4$, $G\not\simeq C_3\times(C_5\rtimes C_4)$, and $G\not\simeq C_{11}\rtimes C_5$. Let us show that $X$ is $G$-birationally superrigid.

By Duncan's lemma \cite[Lemma 2.4]{AbbanCheltsovKishimotoMangolte2025}, the~subgroup $G$ lifts isomorphically to $\mathrm{GL}_5(\mathbb{C})$, so the~action of the~group $G$ on $\mathbb{P}^4$ is given by its faithful $5$-dimensional representations $\mathbb{V}_5$. By assumption, $\mathbb{V}_5$ has no three-dimensional subrepresentations. Hence, twisting $\mathbb{V}_5$ by an appropriate $1$-dimensional representation of the~group $G$, we may assume that either $\mathbb{V}_5$ is irreducible, or
$$
\mathbb{V}_5=\mathbb{I}\oplus\mathbb{V}_4,
$$
where $\mathbb{V}_4$ is an irreducible faithful $4$-dimensional representation of the~group $G$, and $\mathbb{I}$ is the~trivial representation. In particular, we see that $G$ is neither cyclic nor dyhedral. Hence, it follows from the~classification \cite{WeiYu2020} that $G$ is isomorphic to a subgroup of one of the~following four groups:
\begin{enumerate}
\item $C_3^4\rtimes\mathfrak{S}_5$ --- the~automorphism group of the~Fermat cubic \eqref{equation:Fermat};
\item $\mathrm{PSL}_2(\mathbf{F}_{11})$ --- the~automorphism group of the~Klein cubic \eqref{equation:Klein};
\item $((C_3^2\rtimes C_3)\rtimes C_4)\times\mathfrak{S}_3$ --- the~group with GAP ID [648,541];
\item $C_3\times\mathfrak{S}_5$.
\end{enumerate}

\begin{lemma}
\label{lemma:Fermat}
Let $H$ be a subgroup of the~group $C_4^3\rtimes\mathfrak{S}_5$ such that $H$ has a faithful irreducible representation of dimension $4$ or $5$. Then $H$ is isomorphic to one of the~groups listed in Figure~\ref{figure:26}.
\end{lemma}

\begin{proof}
Recall that $C_4^3\rtimes\mathfrak{S}_5$ is the~automorphism group of the~Fermat cubic \eqref{equation:Fermat}.
Using the~Magma code presented in Appendix~\ref{subsection:plane-free-groups}, we find all subgroups of the~automorphism group of this cubic (up to conjugation) that does not leave invariant planes in $\mathbb{P}^4$. The~graph of these subgroups is presented in Figure~\ref{figure:26} (where edges corresponds to inclusions). Up to isomorphism, this graph contains $24$ groups, and all of them have a faithful irreducible representation of dimension $4$ or $5$.

Now, using the~Magma code presented in Appendix~\ref{subsection:qualifying-subgroups}, we obtain the~list of all subgroups of the~group $C_4^3\rtimes\mathfrak{S}_5$ that have an irreducible representation of dimension $4$ or~$5$, which contains the~same subgroups that are listed in Figure~\ref{figure:26}.
\end{proof}

\begin{lemma}
\label{lemma:648-541}
Let $H$ be a subgroup of the~group $((C_3^2\rtimes C_3)\rtimes C_4)\times\mathfrak{S}_3$ such that $H$ has a faithful irreducible representation of dimension $4$ or $5$. Then one of the~following cases holds:
\begin{itemize}
\item $H\simeq\mathfrak{S}_3^2$ and its GAP ID is [36,10];
\item $H\simeq C_3\times (C_3^2\rtimes C_4)$ and its GAP ID is [108,36];
\item $H\simeq C_3\times\mathfrak{S}_3^2$ and its GAP ID is [108,38].
\end{itemize}
\end{lemma}

\begin{proof}
As in the~proof of Lemma~\ref{lemma:Fermat}, we use the~Magma code presented in Appendix~\ref{subsection:weird-threefold-qualifying-subgroups} to obtain the~list of all subgroups of the~group $((C_3^2\rtimes C_3)\rtimes C_4)\times\mathfrak{S}_3$ that have an irreducible representation of dimension $4$ or~$5$. This list consists of $9$ subgroups (up to isomorphism). Then, using GAP, we check that only three of them has a faithful irreducible representation of dimension $4$ or $5$. These are the~groups $\mathfrak{S}_3^2$, $C_3\times (C_3^2\rtimes C_4)$ and $C_3\times\mathfrak{S}_3^2$. For instance, the~following GAP code verifies that the~group $((C_3^2\rtimes C_3)\rtimes C_4)\times\mathfrak{S}_3$ does not have faithful $4$-dimensional irreducible representations.
\begin{verbatim}
    G := SmallGroup(648,541);
    tbl := CharacterTable(G);
    irr := Irr(tbl);
    deg4 := Filtered(irr, chi -> Degree(chi) = 4);
    Print("Number of irreducible representations: ", Length(deg4), "\n");
    faithful := [];
    for chi in deg4 do
        ker := KernelOfCharacter(chi);
        if Size(ker) = 1 then
            Add(faithful, chi);
        fi;
    od;
    if Length(faithful) > 0 then
        Print("Faithful representation exists.\n");
        Print("Number of representations: ", Length(faithful),"\n");
    else
        Print("No faithful representation found.\n");
    fi;
\end{verbatim}
Hence, we conclude that $H\simeq\mathfrak{S}_3^2$ or $H\simeq C_3\times (C_3^2\rtimes C_4)$ or $H\simeq C_3\times\mathfrak{S}_3^2$, as required.
\end{proof}

If $G$ is a isomorphic to a subgroups of the~group $\mathrm{PSL}_2(\mathbf{F}_{11})$, then  
$G\simeq\mathrm{PSL}_2(\mathbf{F}_{11})$ or $G\simeq\mathfrak{A}_5$, because $G\not\simeq C_{11}\rtimes C_5$ by assumption,
and all other subgroups of the~group $\mathrm{PSL}_2(\mathbf{F}_{11})$ do not have irreducible representations of dimension $4$ or $5$. Similarly, up to isomorphism,  $C_3\times\mathfrak{S}_5$ contains three subgroups that have a faithful irreducible representation of dimension $4$ or $5$ that are not listed in Figure~\ref{figure:26}. These are the~groups $C_3\times\mathfrak{S}_5$, $C_3\times\mathfrak{A}_5$, $C_3\rtimes(C_5\rtimes C_4)$.
Since $G\not\simeq C_3\rtimes(C_5\rtimes C_4)$, we see that one of the~following cases holds:
\begin{enumerate}
\item $G\simeq\mathrm{PSL}_2(\mathbf{F}_{11})$,
\item $G\simeq C_3\times\mathfrak{S}_5$,
\item $G\simeq C_3\times\mathfrak{A}_5$,
\item $G$ is isomorphic to one of the~groups listed in Figure~\ref{figure:26} except $C_5\rtimes C_4$.
\end{enumerate}
Now, using Lemmas~\ref{lemma:Fermat} and \ref{lemma:648-541} (and looking at the~graph presented in Figure~\ref{figure:26}), we conclude that the~group $G$ contains a subgroup $H$ such that the~subgroup $H$ isomorphic to one of the~following four \textit{minimal} groups:
\begin{center}
$C_3^4\rtimes C_5$ (GAP ID is [405,15]), $\mathfrak{A}_5$, $C_3^2\rtimes C_4$ (GAP ID is [36,9]), $\mathfrak{S}_3^2$ (GAP ID is [36,10]).
\end{center}
In Sections~\ref{section:405-15}, \ref{section:A5}, \ref{section:36-9-36-10} we will show that $X$ is $G$-birationally superrigid if $G$ is one of these four \textit{minimal} groups. This give the~following result.

\begin{theorem}[{cf. \cite{Kollar2009}}]
\label{theorem:minimal}
Suppose that the~restriction of the~$G$-representation $\mathbb{V}_5$ to $H$ is
\begin{itemize}
\item either an irreducible $5$-dimensional representation of the~group $H$,
\item or a sum of a $1$-dimensional representation and an irreducible $4$-dimensional representation.
\end{itemize}
Then $X$ is $H$-birationally superrigid, and $X$ is $G$-birationally superrigid.
\end{theorem}

\begin{proof}
If $H\simeq C_3^4\rtimes C_5$, then $X$ is $H$-birationally superrigid by Theorem~\ref{theorem:405-15}. Similarly, if $H\simeq \mathfrak{A}_5$, then $X$ is $H$-birationally superrigid by Theorem~\ref{theorem:A5}. If $H\simeq C_3^2\rtimes C_4,\mathfrak{S}_3^2$, then $X$ is $H$-birationally superrigid by Theorem~\ref{theorem:36-9}. Using \cite[Corollary 3.3.3]{CheltsovShramov2016} and \cite[Corollary~1.4.3]{Cheltsov2005}, we see that the~following two conditions are equivalent:
\begin{enumerate}
\item $X$ is not $G$-birationally superrigid;
\item there is a non-empty $G$-invariant mobile linear subsystem $\mathcal{M}\subset |\mathcal{O}_X(n)|$, for some  $n\in\mathbb{Z}_{>0}$, such that the~singularities of the~log pair $(X,\frac{2}{n}\mathcal{M})$ are not canonical.
\end{enumerate}
Thus, if $X$ is not $G$-birationally superrigid, then $X$ is not $H$-birationally rigid. Hence, we conclude that $X$ is  $G$-birationally superrigid.
\end{proof}

In the~proof of Theorem~\ref{theorem:minimal}, we used Theorems~\ref{theorem:405-15}, \ref{theorem:A5}, \ref{theorem:36-9}. We will prove these theorems later in Section~\ref{section:405-15}, \ref{section:A5}, \ref{section:36-9-36-10}, respectively. Their proofs are independent of each other.

Now, we are ready to conclude the~proof of Theorem~\ref{theorem:main}. Suppose that $X$ is not $G$-birationally superrigid.  Then, by Theorem~\ref{theorem:minimal}, the~restriction of the~$G$-representation $\mathbb{V}_5$ to $H$ is a faithful $5$-dimensional representation of the~subgroup $H$ that  has a $3$-dimensional subrepresentation $\mathbb{V}_3$. This suprepresentation $\mathbb{V}_3$ corresponds to a $H$-invariant plane
$$
\Pi\subset\mathbb{P}^4.
$$
But the~groups $C_3^4\rtimes C_5$ and $C_3^2\rtimes C_4$ do not have a faithful $5$-dimensional representation that have a $3$-dimensional subrepresentation.
This shows that $H\not\simeq C_3^4\rtimes C_5$ and $H\not\simeq C_3^2\rtimes C_4$.

If $H\simeq \mathfrak{A}_5$, then $\mathbb{V}_3$ must be an irreducible representation of the~group $H$, which implies that the~plane $\Pi$ does not contain $H$-invarant cubic curves, which is a contradiction, because $\Pi\not\subset X$, so $X\vert_{\Pi}$ is an $H$-invariant cubic curve. This shows that $H\not\simeq \mathfrak{A}_5$.

Hence, we have $H\simeq \mathfrak{S}_3^2$. Applying previous arguments to other subgroups of $G$, we conclude that $G$ does not contain subgroups isomorphic to $C_3^4\rtimes C_5$, $\mathfrak{A}_5$, $C_3^2\rtimes C_4$. Now, looking at the~graph presented in Figure~\ref{figure:26}, we conclude that the~GAP ID of the~group $G$ is one of the~following:
\begin{center}
[108,38], [108,40], [324,160], [324,167], [972,877].
\end{center}
These groups do not have irreducible $5$-dimensional representations. Thus, we have
$$
\mathbb{V}_5=\mathbb{I}\oplus\mathbb{V}_4,
$$
where $\mathbb{V}_4$ is an irreducible faithful $4$-dimensional representation of the~group $G$ such that its restriction to the~subgroup $H$ is a faithful non-irreducible representation of the~group $H\simeq \mathfrak{S}_3^2$. This is purely a group-theoretic condition. Let us verify it using Magma.

If the~GAP ID of $G$ is [108,38], then $G$ has a unique subgroup isomorphic to $\mathfrak{S}_3^2$, which is normal, so $H$ must be this subgroup. This can be checked using the following Magma code:
\begin{verbatim}
    G:=SmallGroup(108,38);
    S:=Subgroups(G);
    for s in S do
        H:=s`subgroup;
        print(IdentifyGroup(H));
        if (IsNormal(G,H)) then 
        print ("Normal\n" );
        end if;
    end for;
\end{verbatim}
In this case, $G$ has two faithful $4$-dimensional representations. Moreover, restricting them to $H$, we obtain irreducible representations (this also follows from the~Clifford theory), which implies that the~GAP ID of $G$ is not [108,38].
Here, we used the following Magma code:
\begin{verbatim}
    G:=SmallGroup(108,38);
    K:=CyclotomicField(108);
    L:=IrreducibleModules(G,K);
    //print(L);
    R:=L[27];
    print(R);
    print(IdentifyGroup(Kernel(R)));
    S:=Subgroups(G);
    H:=S[48]`subgroup;
    print(IdentifyGroup(H));
    W:=Restriction(R, H);
    print(W);
    IsIrreducible(W);
\end{verbatim}

If  the~GAP ID of $G$ is [108,40], then $G$ has three subgroups isomorphic to $\mathfrak{S}_3^2$ (up to conjugation), and $G$ has two (non-isomorphic) faithful $4$-dimensional representations. Restricting each of these representations to each of these subgroups, we obtain irreducible representations, which shows that the~GAP ID of $G$ is not [108,40].

If the~GAP ID of $G$ is [324,160], it has a unique subgroup isomorphic to $\mathfrak{S}_3^2$ (up to conjugation), and $G$ has $6$ (non-isomorphic) $4$-dimensional representations, which are all faithful, and their restrictions to $H$ are irreducible. Hence, the~GAP ID of $G$ is not [324,160].

Similarly, if  the~GAP ID of $G$ is [324,167], then $G$ contains three subgroups isomorphic to $\mathfrak{S}_3^2$ (up to conjugation), and $G$ has a unique faithful $4$-dimensional representations. Restricting this representation to these three subgroups, we obtain irreducible representations, which shows that the~GAP ID of $G$ is not [324,167].

Finally, if the~GAP ID of the~group $G$ is [972,877], then $G$ contains a unique subgroup isomorphic to $\mathfrak{S}_3^2$ (up to conjugation), and $G$ has $12$ (non-isomorphic) faithful $4$-dimensional representations. Restricting each of these representation to this subgroup, we obtain an irreducible representation. Thus, the~GAP ID of $G$ is not [972,877].

The obtained contradiction completes the~proof of Theorem~\ref{theorem:main} (Main Theorem).

\section{Groups $C_5\rtimes C_4$ and $C_3\times(C_5\rtimes C_4)$}
\label{section:20-3}

Let $X$ be a smooth cubic threefold in $\mathbb{P}^4$. Suppose that $\mathrm{Aut}(X)$ contains a subgroup such that either $G\simeq C_5\rtimes C_4$ or $G\simeq C_3\times(C_5\rtimes C_4)$. The~goal of this section is to prove the~following result.

\begin{theorem}
\label{theorem:20-3}
The cubic threefold $X$ is not $G$-birationally rigid.
\end{theorem}

\begin{proof}
The $G$-action on $X$ lifts to $\mathbb{P}^4$. Moreover, it follows from \cite[Lemma 2.4]{AbbanCheltsovKishimotoMangolte2025} that the~action of the~group $G$ on $\mathbb{P}^4$ is given by its faithful $5$-dimensional representation. Looking at representations of the~group $G$, we see that this representation splits as a sum of a $1$-dimensional representation and an irreducible $4$-dimensional representation.

Let $S$ be the $G$-invariant hyperplane section of $X$, and let $\overline{G}\subset\mathrm{Aut}(S)$ be the image of $G$ via the  natural homomorphism $G\to \mathrm{Aut}(S)$. Then $\overline{G}\simeq C_5\rtimes C_4$, and $S$ has isolated singularities, since $X$ is smooth. In fact, the~surface $S$ is smooth. Indeed, if $S$ is singular, then it has exactly four singular points, because $S$ does not have $G$-orbits of length less than $4$, and there are no cubic surfaces with isolated singularities that have more than four singular points. But cubic surface with four singular points is unique, and it does not admit a faithful $\overline{G}$-action. Hence, $S$ is smooth.

Since $S$ is smooth, it follows from \cite{DolgachevIskovskikh2009} that $S$ is isomorphic to the~Clebsch cubic surface, so we have $\mathrm{Aut}(S)\simeq\mathfrak{S}_5$.
Moreover, by \cite{Iskovskikh1996,Wolter},  we have the~following $G$-Sarkisov link:
$$
\xymatrix{
&S\ar[dl]_{\pi}\ar[dr]^{\phi}&\\
\overline{S}&&\mathbb{P}^1\times\mathbb{P}^1}
$$
where $\overline{S}$ is a smooth quintic del Pezzo surface, $\pi$ and $\phi$ are blowups of $2$ and $5$ points, respectively. Let $L_1$ and $L_2$ be the~$\pi$-exceptional curves, and let $\ell_1$, $\ell_2$, $\ell_3$, $\ell_4$, $\ell_5$ be the~$\phi$-exceptional curves. Then $L_1$, $L_2$, $\ell_1$, $\ell_2$, $\ell_3$, $\ell_4$, $\ell_5$ are lines, and it follows from \cite{CheltsovKrylovMartinezShinder2024,SashaYura2025} that this $G$-Sarkisov link can be extended to the~following three-dimensional $G$-Sarkisov link:
$$
\xymatrix{
&\widetilde{X}\ar[dl]_{\alpha}\ar[dr]^{\beta}\ar@{-->}[rr]^{\chi}&&\widehat{X}\ar[dl]_{\gamma}\ar[dr]^{\delta}&\\
X&&V&&Y}
$$
where $\alpha$ is the~blow up of the~lines $L_1$ and $L_2$, $V$ is a divisor of degree $(2,2)$ in $\mathbb{P}^2\times\mathbb{P}^2$ that has five isolated ordinary double points, $\beta$ is a small birational morphism that contracts strict transforms of the~lines $\ell_1$, $\ell_2$, $\ell_3$, $\ell_4$, $\ell_5$ to the~five singular points of $V$, $\chi$ is a composition of Atiyah flops of the~curves contracted by $\beta$, $Y$ is a Fano threefold with $\mathrm{Pic}(Y)=\mathbb{Z}$ and $(-K_Y)^3=14$ that have one isolated ordinary double point, $\gamma$ is a smooth birational morphism, and $\delta$ is a birational morphism that contracts the~strict transform of the~surface $S$ to the~singular point of the~threefold $Y$. 

By construction, $Y$ is a $G$-Mori fibre space (over a point) and $Y\not\simeq X$. Thus, we conclude that the cubic threefold $X$ is not $G$-birationally rigid.
\end{proof}

\section{Group $C_{11}\rtimes C_5$}
\label{section:55-1}

Let $X$ be a smooth cubic threefold in $\mathbb{P}^4$ such that $\mathrm{Aut}(X)$ contains a subgroup $G\simeq C_{11}\rtimes C_5$. Then the~$G$-action on $X$ lifts to $\mathbb{P}^4$, and it follows from \cite[Lemma 2.4]{AbbanCheltsovKishimotoMangolte2025} that the~action of the~group $G$ on the~projective space $\mathbb{P}^4$ is given by its faithful $5$-dimensional representation. This representation must be irreducible, and it is isomorphic to one of the~two distinct irreducible $5$-dimensional representations of the~group $G$. Since these $G$-representations differ by an outer automorphism of the~group $G$, we see that $G$ is uniquely determined as a subgroup of $\mathrm{PGL}_{5}(\mathbb{C})$ up to a conjugation. Hence, changing coordinates on $\mathbb{P}^4$ if neccesary, we may assume that $G$ is generated by the~following two projective transformations:
\begin{align*}
[x_0:x_1:x_2:x_3:x_4]&\mapsto [x_1:x_2:x_3:x_4:x_0],\\
[x_0:x_1:x_2:x_3:x_4]&\mapsto [\zeta_{11}x_0:\zeta_{11}^5x_1:\zeta_{11}^3x_2:\zeta_{11}^4x_3:\zeta_{11}^9x_4].
\end{align*}
Then explicit computations show that $G$ leaves invariant $5$ smooth cubic threefolds:
$$
X_a=\big\{x_0x_1^2+\zeta_5^ax_1x_2^2+\zeta_5^{2a}x_2x_3^2+\zeta_5^{3a}x_3x_4^2+\zeta_5^{4a}x_4x_0^2=0\},
$$
where $a\in\{0,1,2,3,4\}$. Since all these hypersurfaces are $G$-biregular, we may assume that $X=X_0$, which also follows from \cite{Roulleau}. Moreover, it follows from \cite{Adler} that $\mathrm{Aut}(X)\simeq\mathrm{PSL}_2(\mathbf{F}_{11})$, and the cubic $X$ is known as the~Klein cubic threefold.

\begin{theorem}
\label{theorem:Klein-cubic}
The cubic threefold $X$ is not $G$-birationally rigid.
\end{theorem}

To prove this theorem, let $\chi\colon X\dasharrow\mathbb{P}^9$ be the~rational map given by
$$
[x_0:x_1:x_2:x_3:x_4]\mapsto [f_0:f_1:f_2:f_3:f_4:f_5:f_6:f_7:f_8:f_9],
$$
where
\begin{multline*}
f_0=x_0x_1(x_0x_1x_2x_3x_4),
f_1=x_1x_2(x_0x_1x_2x_3x_4),
f_2=x_2x_3(x_0x_1x_2x_3x_4),
f_3=x_3x_4(x_0x_1x_2x_3x_4),\\
f_4=x_4x_0(x_0x_1x_2x_3x_4),
f_5=x_3(x_1x_2^2+x_3x_4^2)x_2x_3x_4,
f_6=x_4(x_0^2x_4+x_2x_3^2)x_3x_4x_0,\\
f_7=x_0(x_0x_1^2+x_3x_4^2)x_4x_0x_1,
f_8=x_1(x_0^2x_4+x_1x_2^2)x_0x_1x_2,
f_9=x_2(x_0x_1^2+x_2x_3^2)x_1x_2x_3,
\end{multline*}
and let $Y$ be the~subscheme in $\mathbb{P}^9$ that is given by the~following equations:
\begin{multline*}
\quad \quad \quad \quad \quad y_0y_9+y_1y_3+y_2y_8=0, y_2y_7-y_0^2-y_3y_4=0,y_0y_6+y_1y_4+y_3y_7=0,\\
y_0y_4+y_1^2-y_3y_8=0, y_1y_2-y_0y_5+y_3^2=0, y_1y_5+y_2y_4+y_3y_9=0,\\
y_0y_3+y_2y_6+y_4y_5=0, y_0y_2+y_1y_7+y_4y_8=0, y_0y_1+y_2^2-y_4y_9=0,\\
y_0y_1-y_3y_6-y_5y_7=0, y_1y_6-y_2y_3-y_4^2=0, y_4y_7-y_1y_2+y_6y_8=0,\\
y_0y_4-y_2y_5-y_6y_9=0, y_2y_3-y_0y_8-y_7y_9=0, y_3y_4-y_1y_9-y_5y_8=0.\quad \quad \quad \quad \quad
\end{multline*}
Then, using Magma, one can check that $Y$ is an irreducible smooth three-dimensional subvariety.
Moreover, $Y$ is $G$-invariant, $\chi$ is $G$-equivariant, and $\chi(X)\subset\mathrm{Supp}(Y)$. 

\begin{lemma}
\label{lemma:55-1-Gr-2-6}
The threefold $Y$ is an intersection of the~Grassmannian $\mathrm{Gr}(2,6)\subset \mathbb{P}^{14}$ with a linear subspace of dimension nine, where $\mathrm{Gr}(2,6)\hookrightarrow\mathbb{P}^{14}$ is the~Pl\"ucker embedding.
\end{lemma}

\begin{proof}
We may assume that $\mathrm{Gr}(2,6)$ is given in $\mathbb{P}^{14}$ by the~following system of equations:
\begin{multline*}
\quad \quad p_{01}p_{23}-p_{02}p_{13}+p_{03}p_{12}=0, p_{01}p_{24}-p_{02}p_{14}+p_{04}p_{12}=0, p_{01}p_{25}-p_{02}p_{15}+p_{05}p_{12}=0,\\
p_{01}p_{34}-p_{03}p_{14}+p_{04}p_{13}=0, p_{01}p_{35}-p_{03}p_{15}+p_{05}p_{13}=0, p_{01}p_{45}-p_{04}p_{15}+p_{05}p_{14}=0,\\
p_{02}p_{34}-p_{03}p_{24}+p_{04}p_{23}=0, p_{02}p_{35}-p_{03}p_{25}+p_{05}p_{23}=0, p_{02}p_{45}-p_{04}p_{25}+p_{05}p_{24}=0,\\
p_{03}p_{45}-p_{04}p_{35}+p_{05}p_{34}=0, p_{12}p_{34}-p_{13}p_{24}+p_{14}p_{23}=0, p_{12}p_{35}-p_{13}p_{25}+p_{15}p_{23}=0,\\
p_{12}p_{45}-p_{14}p_{25}+p_{15}p_{24}=0, p_{13}p_{45}-p_{14}p_{35}+p_{15}p_{34}=0, p_{23}p_{45}-p_{24}p_{35}+p_{25}p_{34}=0,
\quad \quad
\end{multline*}
where $p_{ij}$ are Pl\"ucker coordinates on $\mathbb{P}^{14}$. Let us rename these coordinates as follows: $t_0=p_{01}$, $t_1=p_{12}$, $t_2=p_{23}$, $t_3=p_{34}$, $t_4=p_{04}$, $t_5=p_{13}$, $t_6=p_{24}$, $t_7=p_{03}$, $t_8=p_{14}$, $t_9=p_{02}$,
$t_{10}=p_{45}$, $t_{11}=p_{05}$, $t_{12}=p_{15}$, $t_{13}=p_{25}$, $t_{14}=p_{35}$. Set
$$
\Lambda=\big\{t_{10}=t_5, t_{11}=t_6, t_{12}=t_7, t_{13}=t_8, t_{14}=t_9\big\}\subset\mathbb{P}^{14},
$$
and identify $\Lambda=\mathbb{P}^9$ with coordinates $t_{0}$, $t_1$, $t_2$, $t_3$, $t_4$, $t_5$, $t_6$, $t_7$, $t_8$, $t_9$. Finally, we change coordinates on $\Lambda$ as follows:  $y_0=-t_8$, $y_1=t_9$,  $y_2=t_5$,
$y_3=-t_6$, $y_4=t_7$, $y_5=-t_4$, $y_6=t_0$, $y_7=t_1$, $y_8=t_2$, $y_9=-t_3$. Then $\Lambda\cap \mathrm{Gr}(2,6)=Y$.
\end{proof}

Now, it follows from \cite{Gushel,IskovskikhProkhorov} that $Y$ is an anticanonically embedded smooth Fano threefold such that $\mathrm{Pic}(Y)=\mathbb{Z}[-K_Y]$ and $(-K_Y)^3=14$. Moreover, one can show that the~$G$-equivariant rational map $X\dasharrow Y$ induced by $\chi$ is birational. Hence, the cubic $X$ is not $G$-birationally rigid, which proves Theorem~\ref{theorem:Klein-cubic}. We~expect that $X$ is $G$-birationally solid \cite{CheltsovDuboulozKishimoto2023,CheltsovSarikyan2023,Pinardin2024,PinardinZhang2025}.

Let us explain how did we find $\chi$. Consider $G$-invariant singular quartic threefold:
$$
V_4=\{x_0^2x_1x_3+x_0x_2^2x_3+x_0x_2x_4^2+x_1^2x_2x_4+x_1x_3^2x_4=0\}\subset\mathbb{P}^4.
$$
Observe that the~singular locus of $V_4$ consists of $5$ terminal Gorenstein singular points, which form the~$G$-orbit of the~point $[1:0:0:0:0]$. Note also that the~threefolds $X$ and $V_4$ are $G$-birational. Indeed, let $\psi\colon X\dasharrow V_4$ be the~rational map given by
$$
[x_0:x_1:x_2:x_3:x_4]\mapsto [x_0x_1:x_1x_2:x_2x_3:x_3x_4:x_0x_4],
$$
and let $C$ be the~$G$-invariant curve in $X$ that consists of the~following $5$ lines:
\begin{align*}
\ell_1&=\{x_0=0, x_1=0, x_3=0\},\\
\ell_2&=\{x_1=0, x_2=0, x_4=0\},\\
\ell_3&=\{x_2=0, x_3=0, x_0=0\},\\
\ell_4&=\{x_3=0, x_4=0, x_1=0\},\\
\ell_5&=\{x_4=0, x_0=0, x_2=0\}.
\end{align*}
Then $\psi$ is birational, its indeterminacy locus consists of the~curve $C$, and we have the~following $G$-equivariant commutative diagram:
\begin{equation}
\label{equation:55-1}
\xymatrix{
&\widetilde{X}\ar[dl]_{\alpha}\ar[dr]^{\beta}&\\
X\ar@{-->}[rr]_{\psi}&&V_4}
\end{equation}
where $\alpha$ is the~blow up of the~curve $C$, the~divisor $-K_{\widetilde{X}}$ is nef and big, and $\beta$ is a small birational morphism, which is given by the~linear system $|-K_{\widetilde{X}}|$. It follows from \eqref{equation:55-1} that $\mathrm{rk}\,\mathrm{Cl}(V_4)^G=2$, which implies that $V_4$ is not a $G$-Mori fiber space.

\begin{remark}
\label{remark:55-1-lines}
The lines $\ell_1$, $\ell_2$, $\ell_3$, $\ell_4$, $\ell_5$ form a pentagon configuration, so $C$ has $5$ singular points, which form the~$G$-orbit of the~point $[1:0:0:0:0]$. Thus, the~threefold $\widetilde{X}$ has $5$ isolated ordinary double points. Hence, $C$ is a (reducible) curve of degree $5$ and arithmetic genus $1$.
\end{remark}

Arguing as in \cite{CutroneMarshburn2013,ArapCutroneMarshburn2017,CutroneLimarziMarshburn2019,CutroneMarshburn2025,Kaloghiros2012,BlancLamy}, we see that  \eqref{equation:55-1} can be uniquely extended to a $G$-Sarkisov link that starts at $X$. Let us describe this $G$-Sarkisov link.

\begin{remark}
\label{remark:55-1-link}
If $C$ were a smooth elliptic curve, the~$G$-Sarkisov link induced by \eqref{equation:55-1} would be a~usual Sarkisov link that starts at $X$ and ends up at a smooth Fano threefold of Picard rank $1$ and anticanonical degree $14$. This Sarkisov link is classically known \cite{Fano,Iskovskikh1980,Puts}, and it appeared in \cite{Beauville2002,Kuznetsov2004,SashaYuraCostya,AtanasDima,AtanasLaurent,IskovskikhProkhorov,YuraZhijia}. Another Sarkisov link that starts at a smooth cubic threefold and ends at a smooth Fano threefold of Picard rank $1$ and anticanonical degree $14$ is described in \cite{Takeuchi1989,Tregub}.
\end{remark}

Thus, to construct the~required $G$-Sarkisov link, we can follow the~proof of \cite[Theorem~1.4]{Iskovskikh1980}. Namely, let $\mathcal{M}$ be the~linear system on $X$ that consists of the~following surfaces:
$$
\big\{a_0f_0+a_1f_1+a_2f_2+a_3f_3+a_4f_4+a_5h_0+a_6h_1+a_7h_2+a_8h_3+a_9h_4=0\big\}\cap X,
$$
where $[a_0:a_1:a_2:a_3:a_4:a_5:a_6:a_7:a_8:a_9]\in\mathbb{P}^9$. Then $\mathcal{M}$ is $G$-invariant, and the~constructed rational map $\chi$ is given by the~linear system $\mathcal{M}$. Note that the~base locus of this linear system consists of the~curve $C$. Moreover, we have
$$
\mathrm{mult}_{C}\big(\mathcal{M}\big)=4.
$$
Now, we set $S_i=X\cap\{x_i=0\}$ for $i\in\{0,1,2,3,4\}$. Then each $S_i$ is a singular cubic surface that have a Du Val singular point of type $\mathbb{D}_4$, each $\psi(S_i)$ is a plane in $V_4$, and each $\chi(S_i)$ is a line in $Y$. Set $L_1=\chi(S_1)$, $L_2=\chi(S_2)$, $L_3=\chi(S_3)$, $L_4=\chi(S_4)$, $L_5=\chi(S_5)$, and set
$$
Z=L_1+L_2+L_3+L_4+L_5.
$$
Then $L_1$, $L_2$, $L_3$, $L_4$, $L_5$ span a $4$-dimensional linear subspace $\Pi\subset\mathbb{P}^9$ such that
$$
Y\big\vert_{\Pi}=Z,
$$
and  $L_1$, $L_2$, $L_3$, $L_4$, $L_5$ form a pentagon configuration. So, the~arithmetic genus of $Z$ is $1$.

Let $\delta\colon\widehat{X}\to Y$ be the~blow up of the~curve $Z$. Then $\widehat{X}$ has $5$ isolated ordinary double points, the~divisor $-K_{\widehat{X}}$ is nef and big, and we have the~following $G$-equivariant commutative diagram:
\begin{equation}
\label{equation:55-1-half-link}
\xymatrix{
&\widehat{X}\ar[dl]_{\gamma}\ar[dr]^{\delta}&\\
V_4&&Y\ar@{-->}[ll]}
\end{equation}
where $\gamma$ is a small birational morphism given by $|-K_{\widehat{X}}|$, and $Y\dasharrow V_4$ is the~rational map induced by the~linear projection from $\Pi$. Now, combining \eqref{equation:55-1} and \eqref{equation:55-1-half-link}, we obtain the~following $G$-equivariant commutative diagram:
\begin{equation}
\label{equation:55-1-link}
\xymatrix{
\widetilde{X}\ar[d]_{\alpha}\ar[rr]^{\beta}&& V_4&&\widehat{X}\ar[ll]_{\gamma}\ar[d]^{\delta}\\
X\ar@{-->}[rrrr]\ar@{-->}[rru]_{\psi}&&&&Y\ar@{-->}[ull]}
\end{equation}
where $X\dasharrow Y$ is the~birational map induced by $\chi$. Note that  \eqref{equation:55-1-link} is a $G$-Sarkisov link, and $\delta$-exceptional surfaces are the~strict transforms of the~cubic surfaces $S_1$, $S_2$, $S_3$, $S_4$, $S_5$.

\section{Group $C_3^4\rtimes C_5$}
\label{section:405-15}

Let $X$ be a smooth cubic threefold in $\mathbb{P}^4$. Suppose that $\mathrm{Aut}(X)$ has a subgroup $G\simeq C_3^4\rtimes C_5$. Since the~$G$-action lifts to $\mathbb{P}^4$, we can consider $G$ as a subgroup in $\mathrm{PGL}_5(\mathbb{C})$. Moreover, it follows from Duncan's lemma \cite[Lemma 2.4]{AbbanCheltsovKishimotoMangolte2025} that the group $G$ lifts isomorphically to $\mathrm{GL}_5(\mathbb{C})$. Looking at representations of the group $G$, we see that the~embedding $G\hookrightarrow\mathrm{GL}_5(\mathbb{C})$ is given~by an~irreducible representation. 

The~group $G$ has $16$ irreducible $5$-dimensional representations, but all of them define the same subgroup in $\mathrm{PGL}_5(\mathbb{C})$ up to conjugation. Thus, we may assume that $G$ is generated by 
\begin{align*}
[x_0:x_1:x_2:x_3:x_4]&\mapsto [x_0:\xi_3x_1:x_2:x_3:\xi_3^2x_4],\\
[x_0:x_1:x_2:x_3:x_4]&\mapsto [x_0:\xi_3x_1:x_2:\xi_3^2x_3:x_4],\\
[x_0:x_1:x_2:x_3:x_4]&\mapsto [x_0:\xi_3x_1:\xi_3^2x_2:x_3:x_4],\\
[x_0:x_1:x_2:x_3:x_4]&\mapsto [\xi_3x_0:\xi_3^2x_1:x_2:x_3:x_4],\\
[x_0:x_1:x_2:x_3:x_4]&\mapsto [x_4:x_3:x_1:x_0:x_2].
\end{align*}
Then $X=\{x_0^3+\zeta_5^ax_1^3+\zeta_5^{2a}x_2^3+\zeta_5^{3a}x_3^3+\zeta_5^{4a}x_4^3=0\big\}\subset\mathbb{P}^4$ for some $a\in\{0,1,2,3,4\}$. Since all these cubic threefolds are $G$-biregular, we may assume that $a=0$, so $X$ is the~Fermat cubic \eqref{equation:Fermat}. 

\begin{theorem}
\label{theorem:405-15}
The cubic theeefold $X$ is $G$-birationally superrigid.
\end{theorem}

Before proving this result, let us state without proof $10$ facts.
\begin{enumerate}
\item The group $G$ cannot act faithfully on a smooth curve  of genus $\leqslant 15$ \cite{Breuer,Jen,lmfdb}.
\item Every proper subgroup of $G$ is isomorphic to $C_3$, $C_5$, $C_3^2$, $C_3^3$ or $C_3^4$.
\item The group $G$ contains a unique subgroup of index $5$, which is isomorphic to $C_3^4$.
\item If $H$ is the~subgroup of $G$ of index $5$, then $X$ does not contain $H$-invariant lines.
\item The cubic $X$ does not contain $G$-orbits of length $<15$;
\item The cubic $X$ contains two $G$-orbits of length $15$, which can be described as follows:
\begin{align*}
\Sigma_{15}&=\mathrm{Orb}_G\big([0:0:0:1:-1]\big),\\
\Sigma_{15}^\prime&=\mathrm{Orb}_G\big([0:0:1:0:-1]\big).
\end{align*}
\item As a $G$-representation, $H^0(X,\mathcal{O}_{X}(-K_X))$ splits as a sum of three pairwise non-isomorphic $5$-dimensional irreducible representations, which correspond to the~$4$-dimensional linear subsystems in $|\mathcal{O}_{X}(2)|$ that can be described as follows:
\begin{itemize}
\item $\mathcal{H}$ is the~linear system that is cut out by quadrics
$$
a_0x_0x_1+a_1x_1x_2+a_2x_2x_3+a_3x_3x_4+a_4x_0x_4=0,
$$
\item $\mathcal{H}^\prime$ is the~linear system that is cut out by quadrics
$$
a_0x_0x_2+a_1x_1x_3+a_2x_2x_4+a_3x_3x_5+a_4x_1x_4x=0,
$$
\item $\mathcal{H}^{\prime\prime}$ is the~linear system that is cut out by quadrics
$$
a_0x_0^2+a_1x_1^2+a_2x_2^2+a_3x_3^2+a_4x_4^2=0,
$$
\end{itemize}
where $[a_0:a_1:a_2:a_3:a_4]\in\mathbb{P}^4$.
\item The $G$-orbit $\Sigma_{15}$ is the~base locus of the linear system $\mathcal{H}$.
\item The $G$-orbit $\Sigma_{15}^\prime$ is the~base locus of the linear system $\mathcal{H}^\prime$.
\item The linear system $\mathcal{H}^{\prime\prime}$ is base point free.
\end{enumerate}

\begin{proof}[Proof of Theorem~\ref{theorem:405-15}]
Suppose $X$ is not $G$-birationally superrigid. Let us seek for a contradiction. 

It follows from \cite[Corollary 3.3.3]{CheltsovShramov2016} that there exists a non-empty $G$-invariant mobile linear subsystem $\mathcal{M}\subset |\mathcal{O}_X(n)|$, for some  $n\in\mathbb{Z}_{>0}$, such that the~log pair $(X,\frac{2}{n}\mathcal{M})$ is not canonical. Then, by \cite[Corollary~2.3]{CheltsovShramov2014}, the~log pair $(X,\frac{4}{n}\mathcal{M})$ is not log canonical. 

Choose a positive rational number $\mu\leqslant \frac{4}{n}$ such that the~singularities of  $(X,\mu\mathcal{M})$ are strictly log canonical. Let $C$ be an irreducible subvariety in $X$ that is a minimal center of log canonical singularities of the~log pair $(X,\mu\mathcal{M})$, and let $Z$ be the $G$-irreducible curve in $X$ whose irreducible component is the curve $C$. Then every irreducible component of the curve $Z$ is a minimal center of log canonical singularities of the log pair $(X,\mu\mathcal{M})$, which implies that these irreducible components are disjoint \cite[Proposition 1.5]{Kawamata-1}. Moreover, we know that $Z$ is not a surface, since $\mathcal{M}$ is mobile. Therefore, either $Z$ is a $G$-irreducible curve or $Z$ is the~$G$-orbit of a point. 

Now, we are going to use \cite[Lemma 2.8]{cheltsov2011exceptional}, \cite[Lemma 2.4.10]{CheltsovShramov2016} or \cite[Lemma A.28]{CalabiBook}, which are equivariant versions of \emph{Tie Breaking} trick used in the~proof of \cite[Theorem 1.10]{Kawamata-1}. Namely, using this \textit{equivariant trick} together with \cite[Theorem 4.8]{Kollar1997}, we can replace the~rational number $\mu$ and the~mobile linear system $\mathcal{M}$ by a positive rational number $\lambda<2$ and an effective $G$-invariant $\mathbb{Q}$-divisor $D\sim_{\mathbb{Q}}-K_{X}$ such that the log pair $(X,\lambda D)$ is log canonical, $C$ is a minimal center of the~non-Kawamata log terminal singularities of this pair, and $\mathrm{Nklt}(X,\lambda D)=Z$,
where 
$$
\mathrm{Nklt}\big(X,\lambda D\big)=\big\{P\in X\ \text{such that $(X,\lambda D)$ is not Kawamata log terminal at $P$}\big\}.
$$

Let $\mathcal{I}_Z$ be the~ideal sheaf of the~subvariety $Z\subset X$. Then it follows from the Nadel vanishing theorem \cite{Nadel} in the~form of \cite[Theorem 9.4.8]{Lazarsfeld2004} that
$$
h^1\big(X,\mathcal{O}_{X}(-K_X)\otimes\mathcal{I}_Z\big)=0.
$$
Hence, we have the~following exact sequence of $G$-representations:
$$
0\longrightarrow H^0\big(X,\mathcal{I}_Z\otimes\mathcal{O}_{X}(-K_X)\big)\longrightarrow H^0\big(X,\mathcal{O}_{X}(-K_X)\big)\longrightarrow H^0\big(Z,\mathcal{O}_{Z}(-K_X\vert_{Z})\big)\longrightarrow 0,
$$
which gives
\begin{equation}
\label{equation:405-15}
h^0\big(Z,\mathcal{O}_{Z}(-K_X\vert_{Z})\big)=15-h^0\big(X,\mathcal{I}_Z\otimes\mathcal{O}_{X}(-K_X)\big),
\end{equation}
because $h^0(X,\mathcal{O}_{X}(-K_X))=15$.  Thus, if $Z$ is a $G$-orbit of a point, \eqref{equation:405-15} gives
$$
|Z|=h^0(Z,\mathcal{O}_{Z})=15-h^0\big(X,\mathcal{I}_Z\otimes\mathcal{O}_{X}(-K_X)\big)\leqslant 15,
$$
so $Z$ is a $G$-orbit of length $15$ and $h^0(X,\mathcal{I}_Z\otimes\mathcal{O}_{X}(-K_X))=0$, which implies that $Z$ is not contained in any surface in $|-K_X|$.
On the~other hand, the~only $G$-orbits of length $15$ in $X$ are $\Sigma_{15}$ and~$\Sigma_{15}^\prime$, and each of them is contained in many surfaces in $|-K_X|$, since $\Sigma_{15}$ is the~base locus of the~linear system $\mathcal{H}$, and $\Sigma_{15}^\prime$ is the~base locus of the~linear system $\mathcal{H}^\prime$.
Thus, $Z$ is not a $G$-orbit of a point.

We see that $Z$ is a curve. Then it follows from \cite{Kawamata-2} that $C$ is smooth, so $Z$ is also smooth. Moreover, we have
$$
h^0\big(X,\mathcal{I}_Z\otimes\mathcal{O}_{X}(-K_X)\big)=0,
$$
because the~only $G$-invariant linear subsystems in $|-K_X|$ are $\mathcal{H}$, $\mathcal{H}^\prime$, $\mathcal{H}^{\prime\prime}$, and their base loci do not contain curves.
This and \eqref{equation:405-15} give:
\begin{equation}
\label{equation:405-15-easy}
h^0\big(Z,\mathcal{O}_{Z}(-K_X\vert_{Z})\big)=15.
\end{equation}

Let $r$ be the~number of irreducible components of the curve $Z$, let $d$ be the~degree of the~curve~$C$, let $g$ be the~genus of the curve $C$, and let $A=\epsilon (-K_X)$ for any  $\epsilon\in\mathbb{Q}_{>0}$. Then it follows from the~Kawamata subadjunction theorem \cite[Theorem 1]{Kawamata-2} that
$$
\big(K_{X}+\lambda D+A\big)\big\vert_{C}\sim_{\mathbb{Q}} K_C+\Delta_C
$$
for some effective $\mathbb{Q}$-divisor $\Delta_C$ on the~curve $C$. This gives
$$
(\lambda-1+\epsilon)2d\geqslant 2g-2.
$$
Since this inequality holds for any $\epsilon>0$, we get $2d>2g-2$, because $\lambda<2$. Then $g\leqslant d$, which implies that the~divisor $-K_X\vert_{C}$ is not special. Using \eqref{equation:405-15-easy} and the~Riemann--Roch formula, we get
$$
15=h^0\big(Z,\mathcal{O}_{Z}(-K_X\vert_{Z})\big)=r(2d-g+1).
$$
On the~other hand, going through the~subgroups of $G$, we see that $r\in\{1,5,15,45,81,135,405\}$.

If $g=0$, then $15=r(2d+1)$, which implies that $r=5$ and $d=1$, so $C$ is a line, and its stabilizer in $G$ is the~unique subgroup of index $5$. As we mentioned earlier, $X$ does not contain such lines. Hence,  $g\ne 0$. Then $d\geqslant 3$ and
$$
15=r(2d-g+1)\geqslant r(d+1)\geqslant 4r,
$$
which implies that $r=1$, so $C=Z$ and $G$ acts faithfully on $C$. Now, using $g\leqslant d$, we get
$$
15=2d-g+1\geqslant d+1,
$$
so $d\leqslant 14$. Then $g\leqslant d\leqslant 14$. But $G$ cannot act faithfully on a smooth curve  of genus $\leqslant 15$. 
\end{proof}

\section{Group $\mathfrak{A}_5$}
\label{section:A5}

Let $X$ be a smooth cubic hypersurface in $\mathbb{P}^4$ such that $\mathrm{Aut}(X)$ contains a subgroup $G\simeq\mathfrak{A}_5$. The goal of this section is to prove the following result.

\begin{theorem}
\label{theorem:A5}
The cubic $X$ is $G$-birationally superrigid.
\end{theorem}

First, we recall the following result from \cite{CheltsovShramov2016}.

\begin{lemma}
\label{lemma:A5-curves}
Let $C$ be a smooth irreducible curve that admits a non-trivial action of the group $G$. Then the length of a $G$-orbit in $C$ is $12$, $20$, $30$ or $60$. Moreover, if $g\leqslant 16$, then 
$$
g\in\big\{0,4,5,6,9,10,11,13,15,16\big\},
$$
and the possible numbers of $G$-orbits in $C$ consisting of $12$, $20$, $30$ points are contained in the table below, where where $a_k$ is the~number of $G$-orbits in $C$.
\begin{center}\renewcommand\arraystretch{1.1}
\begin{tabular}{|c||c|c|c|c|c|c|c|c|c|c|}
\hline
$g$ &      $0$ & $4$ & $5$ & $6$ & $9$ & $10$ & $11$ & $13$ & $15$ & $16$ \\
\hline
$a_{12}$ & $1$ & $2$ & $1$ & $0$ & $2$ & $1$  & $0$  & $3$  & $1$  & $0$  \\
\hline
$a_{20}$ & $1$ & $0$ & $2$ & $1$ & $1$ & $0$  & $2$  & $0$  & $1$  & $3$  \\
\hline
$a_{30}$ & $1$ & $1$ & $0$ & $3$ & $0$ & $3$  & $2$  & $0$  & $2$  & $1$  \\
\hline
\end{tabular}
\end{center}
\end{lemma}

\begin{proof}
This is \cite[Lemma 5.1.4]{CheltsovShramov2016} and \cite[Lemma 5.1.5]{CheltsovShramov2016}, see also \cite{lmfdb,Jen,Breuer}.
\end{proof}

Since the~$G$-action on the cubic $X$ lifts to $\mathbb{P}^4$, we can also consider $G$ as a subgroup in $\mathrm{PGL}_5(\mathbb{C})$. Moreover, the subgroup $G\subset\mathrm{PGL}_5(\mathbb{C})$ lifts isomorphically to $\mathrm{GL}_5(\mathbb{C})$ by \cite[Lemma 2.4]{AbbanCheltsovKishimotoMangolte2025}. Note that the~embedding $G\hookrightarrow\mathrm{GL}_5(\mathbb{C})$ is given by a faithful $5$-dimensional $G$-representation $\mathbb{V}_5$.
Then either $\mathbb{V}_5$ is the unique irreducible $5$-dimensional representation of the group $G$, or
$$
\mathbb{V}_5\simeq \mathbb{I}\oplus\mathbb{V}_4,
$$
where $\mathbb{I}$ is the trivial representation, and $\mathbb{V}_4$ is the unique (standard) $4$-dimensional irreducible representation. If $\mathbb{V}_5$ is reducible, we say that $G$ is a \textit{standard} subgroup of the~group $\mathrm{PGL}_5(\mathbb{C})$.
If~$\mathbb{V}_5$ is irreducible, we say that $G$ is a \textit{non-standard} subgroup of the group~$\mathrm{PGL}_5(\mathbb{C})$. Set
$$
M_1=\begin{pmatrix}%(1,2,4)
0 & 0 & 0 & 1 & 0 \\
1 & 0 & 0 & 0 & 0 \\
0 & 0 & 1 & 0 & 0 \\
0 & 1 & 0 & 0 & 0 \\
0 & 0 & 0 & 0 & 1
\end{pmatrix}, M_2=\begin{pmatrix}%(1,2,4)
0 & 0 & 0 & 1 & 0 \\
1 & 0 & 0 & 0 & 0 \\
0 & 0 & \xi_3& 0 & 0 \\
0 & 1 & 0 & 0 & 0 \\
0 & 0 & 0 & 0 & \xi_3^2
\end{pmatrix},
M_3=\begin{pmatrix}%(1,5)(3,4)
0 & 0 & 0 & 0 & 1 \\
0 & 1 & 0 & 0 & 0 \\
0 & 0 & 0 & 1 & 0 \\
0 & 0 & 1 & 0 & 0 \\
1 & 0 & 0 & 0 & 0
\end{pmatrix}. %(1,2,4)(1,5)(3,4)=(1,5,2,4,3)
$$
Choosing appropriate coordinates on $\mathbb{P}^4$, we may assume that one of the following cases holds:
\begin{enumerate}
\item $G=\langle M_1,M_3\rangle$, and $G$ is a standard subgroup of $\mathrm{PGL}_5(\mathbb{C})$;
\item $G=\langle M_2,M_3\rangle$, and $G$ is a non-standard subgroup of $\mathrm{PGL}_5(\mathbb{C})$.
\end{enumerate}

\begin{lemma}
\label{lemma:A5-standard-equation}
Suppose that $G=\langle M_1,M_3\rangle$. Then $X$ can be given by
\begin{equation}
\label{equation:A5-standard}
x_0^3+x_1^3+x_2^3+x_3^3+x_4^3+a(x_0+x_1+x_2+x_3+x_4)(x_0^2+x_1^2+x_2^2+x_3^2+x_4^2)=0
\end{equation}
for some $a\in\mathbb{C}$ such that $a\ne -\frac{1}{5}$, $80a^3+144a^2+135a+27\ne 0$ and $40a^3+72a^2+45a+9\ne 0$.
\end{lemma}

\begin{proof}
It is well-known that the cubic threefold $X$ is given by
$$
\sum_{i=0}^4x_i^3+a\Big(\sum_{i=0}^4x_i\Big)\Big(\sum_{i=0}^4x_i^2\Big)+b\Big(\sum_{i=0}^4x_i\Big)^3=0
$$
for some $a,b\in\mathbb{C}$, where $(a,b)\ne (-\frac{3}{5},\frac{2}{25})$, since otherwise $X$ would be singular. Take some $s\in\mathbb{C}$. Then, applying the projective transformation
$$
\begin{pmatrix}
4+s&s-1&s-1&s-1&s-1\\
s-1&4+s&s-1&s-1&s-1\\
s-1&s-1&4+s&s-1&s-1\\
s-1&s-1&s-1&4+s&s-1\\
s-1&s-1&s-1&s-1&4+s
\end{pmatrix},
$$
we transform the equation of $X$ into
$$
\sum_{i=0}^4x_i^3+\frac{(5a+3)s-3}{5}\Big(\sum_{i=0}^4x_i\Big)\Big(\sum_{i=0}^4x_i^2\Big)+\frac{2+(5a+25b+1)s^3-(5a-3)s}{25}\Big(\sum_{i=0}^4x_i\Big)^3=0.
$$
Note that $2+(5a+25b+1)s^3-(5a-3)s$ is a non-constant polynomial in $s$, because $(a,b)\ne (-\frac{3}{5},\frac{2}{25})$. Therefore,  we can choose $s\in\mathbb{C}$ such that $2+(5a+25b+1)s^3-(5a-3)s=0$, which implies that may assume that $b=0$, so the cubic $X$ is given by \eqref{equation:A5-standard} for some $a\in\mathbb{C}$ as claimed.

If $a=-\frac{1}{5}$, then $X$ has a node at $[1:1:1:1:1]$. Similarly, if  $80a^3+144a^2+135a+27=0$, then $X$ has five nodes. Finally, if $40a^3+72a^2+45a+9=0$, then $X$ has ten nodes.
\end{proof}

The proof of the following result is very similar to the proof of \cite[Proposition 5.1]{PinardinZhang2025}.

\begin{proposition}
\label{proposition:A5-standard}
Suppose that $G=\langle M_1,M_3\rangle$. Then $X$ is $G$-birationally superrigid.
\end{proposition}

\begin{proof}
By Lemma~\ref{lemma:A5-standard-equation}, we may assume that $X$ is given by \eqref{equation:A5-standard}.
Let's describe small $G$-orbits in~$X$. First, we observe that the cubic threefold $X$ does not contain $G$-orbits of length $6$,
which implies that all $G$-orbits in $X$ of length $<12$ have length $5$ or $10$.

To describe $G$-orbits in $X$ of length $5$, take  $[s:t]\in\mathbb{P}^1$ such that
$$
(16a+4)s^3+4as^2t+4ast^2+(a+1)t^3=0.
$$
Then $[s:s:s:s:t]\in X$, and the stabilizer in $G$ of this point is isomorphic to $\mathfrak{A}_4$, which implies that the $G$-orbit of the point $[s:s:s:t:t]$ has length $5$, and every $G$-orbit of length $5$ can be obtained in this way. Note that there are exactly three possibilities for $[s:s:s:s:t]$, because the discriminant of the polynomial $(16a+4)s^3+4as^2t+4ast^2+(a+1)t^3$ is not zero, since $a\ne -\frac{1}{5}$ and $80a^3+144a^2+135a+27\ne 0$ by Lemma~\ref{lemma:A5-standard-equation}.
Hence, $X$ has three $G$-orbits of length $5$.

To describe $G$-orbits of length $10$ in $X$, we let $O=[0:0:0:1:-1]$, let 
$$
\Sigma_{10}=\mathrm{Orb}_G(O),
$$
and let $G_O$ be the stabilizer in $G$ of the point $O$. Then $\Sigma_{10}\subset X$, $|\Sigma_{10}|=10$, $G_O\simeq\mathfrak{S}_3$, and the~group $G_O$ fixes $3$ extra points in $X$, which can be described as follows. Take $[s:t]\in\mathbb{P}^1$~such~that
$$
(3+9a)s^3+6ats^2+6at^2s+(4a+2)t^3=0.
$$
Then $[s:s:s:t:t]\in X$, and this point is fixed by $G_O$. This implies that the $G$-orbit of the point $[s:s:s:t:t]$ has length $10$, and every $G$-orbit of length $10$ in $X$ different from $\Sigma_{10}$ can be obtained in this way. Note that there are exactly three possibilities for $[s:s:s:t:t]$, because the~discriminant of the polynomial $(3+9a)s^3+6ats^2+6at^2s+(4a+2)t^3$ is not zero, since $a\ne -\frac{1}{5}$ and $80a^3+144a^2+135a+27\ne 0$ by Lemma~\ref{lemma:A5-standard-equation}. Thus, $X$ has four $G$-orbits of length $10$.

We have described all $G$-orbits in $X$ of length $<12$. Now, we set
$$
S=X\cap\big\{x_0+x_1+x_2+x_3+x_4=0\big\}.
$$
Then $S$ is a smooth cubic surface, the group $G$ acts faithfully on $S$, and $S$ is the so-called Clebsch cubic surface. Observe that $\Sigma_{10}\subset S$, and $\Sigma_{10}$ is the unique $G$-orbit in $S$ of length $<12$, which also follows from \cite[Lemma 6.3.12]{CheltsovShramov2016}. Moreover, it follows from \cite[Lemma 5.8]{PinardinZhang2025} that $\alpha_G(S)=2$. Furthermore, $S$ does not contain $G$-invariant curves of degree $<6$ by \cite[Theorem 6.3.18]{CheltsovShramov2016}.

We claim that $X$ contains no $G$-invariant curves of degree $<11$ that are not contained in~$S$. Indeed, if $Z$ is a $G$-irreducible curve  in $X$ of degree $d<11$ and $Z\not\subset S$, then $S\cap Z$ is a $G$-invariant subset consisting of at most $S\cdot Z=d$ points, which must be a union of $G$-orbits. Thus, we have
$$
S\cap Z=\Sigma_{10},
$$
so $d=S\cdot Z=10$, which implies that $X$ does not contain curves of degree $<6$. Moreover, since 
$$
|S\cap Z|=|\Sigma_{10}|=10=S\cdot Z,
$$
the surface $S$ intersects $Z$ transversally at every point of $\Sigma_{10}$, and, therefore $C$ must be smooth at every point of $\Sigma_{10}$. This implies that $Z$ is reducible, since otherwise $G$ would act faithfully on $Z$, and the stabilizer of any point in $\Sigma_{10}\subset Z\setminus\mathrm{Sing}(Z)$ is cyclic, which is not the case. Then
\begin{enumerate}
\item either $Z$ is a union of $10$ lines;
\item or $Z$ is a union of $5$ smooth conics.
\end{enumerate}
Let $C$ be an irreducible component of $Z$, and let $G_C$ be its stabilizer in $G$. If $Z$ consists of $10$ lines, then $G_C\simeq\mathfrak{S}_3$, so we may assume that $G_C=G_O$. Then, since $S\cap C$ is a  $G_O$-fixed point, we see that $S\cap C=[0:0:0:1:-1]$, and $G_O$ fixes another point in $C$, which is not contained in $S$. Hence, $C$ contains $[s:s:s:t:t]$ for $[s:t]\in\mathbb{P}^1$ such that $(3+9a)s^3+6ats^2+6at^2s+(4a+2)t^3=0$,
which gives $80a^3+144a^2+135a+27=0$, which is impossible by Lemma~\ref{lemma:A5-standard-equation}. Hence, we conclude that $C$ is a smooth conic. Let $\Pi_C$ be the plane in $\mathbb{P}^4$ that contains $C$. Then
$$
\Pi_C\cap X=C+L,
$$
for a line $L$. Now, applying $G$ to the line $L$, we obtain a $G$-irreducible curve in $X$ of degree $\leqslant 5$, which does not exists. So, $X$ contains no $G$-irreducible curves of degree $<12$ not contained in $S$.

Now, we suppose that $X$ is not $G$-birationally superrigid.
Then, by \cite[Corollary 3.3.3]{CheltsovShramov2016}, there exists a non-empty $G$-invariant mobile linear subsystem
$\mathcal{M}\subset |nS|$, for a positive integer~$n$, such that $(X,\frac{2}{n}\mathcal{M})$ is not canonical.
Let 
$$
\Sigma=\big\{P\in X\ \text{such that $(X,\frac{2}{n}\mathcal{M})$ is not canonical at~$P$}\big\}.
$$
We claim that $\Sigma$ is a finite subset.

Indeed, suppose that $\Sigma$ is not finite subset. Then $\Sigma$ contains a $G$-irreducible curve $Z$, and it follows from \cite[Excercise 6.18]{KollarSmithCorti} that
$$
\mathrm{mult}_Z\big(\mathcal{M}\big)>\frac{n}{2}.
$$
Let $M$ and $M^\prime$ be general surfaces in $\mathcal{M}$. Then
$$
\mathrm{mult}_Z\big(M\big)=\mathrm{mult}_Z\big(M^\prime\big)=\mathrm{mult}_Z\big(\mathcal{M}\big)>\frac{n}{2}.
$$
Then, in particular, we have
$$
M\cdot M^\prime=mZ+\Delta,
$$
where $m$ is a positive integer such that $m\geqslant\mathrm{mult}^2_Z(\mathcal{M})$, and $\Delta$ is an effective one-cycle on $X$ whose does not contain $Z$. Then
$$
3n^2=S\cdot M\cdot M^\prime=m\mathrm{deg}(Z)+S\cdot\Delta\geqslant\mathrm{mult}^2_Z\big(\mathcal{M}\big)\mathrm{deg}(Z)>\frac{n^2}{4}\mathrm{deg}(Z),
$$
which gives $\mathrm{deg}(Z)<12$. Thus, we conclude that $Z\subset S$ and $\mathrm{deg}(Z)\geqslant 6$. Then 
$$
n(-K_S)\sim M\big\vert_{S}=m_SZ+\Delta_S
$$
for some $m_S\geqslant \mathrm{mult}_Z(\mathcal{M})>\frac{n}{2}$ and some effective divisor $\Delta_S$ on the~surface $S$ whose support does not contain the curve $Z$. Then 
$$
3n=-K_S\cdot (mZ+\Delta_S)=m\mathrm{deg}(Z)-K_S\cdot\Delta_S\geqslant m\mathrm{deg}(Z)\geqslant \mathrm{mult}_Z(\mathcal{M})\mathrm{deg}(Z)>\frac{n\mathrm{deg}(Z)}{2},
$$
so $\mathrm{deg}(Z)<6$, which is impossible, since $S$ contains ni $G$-invariant curves of degree $<6$.

Therefore, we conclude that $\Sigma$ is a finite subset. Then $\Sigma\cap S=\varnothing$. Indeed, if $\Sigma\cap S\ne\varnothing$, then the log pair $(S,\frac{2}{n}\mathcal{M}\vert_{S})$ is not log canonical by the~inversion of adjunction \cite[Theorem 5.50]{KollarMori}, so~$\alpha_G(S)<2$, which is impossible, since $\alpha_G(S)=2$ by \cite[Lemma 5.8]{PinardinZhang2025}.

Arguing as in the proof of \cite[Lemma 4.6]{Avilov2018}, we see that $\Sigma$ contains no $G$-orbits of length~$5$. Indeed, suppose that $\Sigma$ contains a $G$-orbit $\Sigma_5$ that have length $5$. Since $\Sigma_5$ is not contained in $S$, the points of $\Sigma_5$ are in linearly general position. Further, it follows from \cite{Pukhlikov1998,Corti2000} that 
$$
\big(M\cdot M^\prime)_P>n^2,
$$
for every point $P\in\Sigma$, where as above, $M$ and $M^\prime$ are general surfaces in $\mathcal{M}$.  Let $\mathcal{P}$ be the linear subsystem in $|\mathcal{O}_{X}(3)|$ consisting of all surfaces that are singular at every point of the $G$-orbit $\Sigma_5$. Then $\mathcal{P}$ is non-empty, mobile, and its base locus consists of the $G$-orbit $\Sigma_5$, because $X$ does not contain lines in $\mathbb{P}^4$ that pass through pair of points of $\Sigma_5$. Hence, if $\mathscr{S}$ is a general surface in $\mathcal{P}$, then
$$
9n^2=\mathscr{S}\cdot M\cdot M^\prime\geqslant\sum_{P\in\Sigma_5}\mathrm{mult}_P(\mathscr{S})\big(M\cdot M^\prime)_M>2n^2|\Sigma_5|=10n^2,
$$
which is absurd. Therefore, we conclude that $\Sigma$ does not contain $G$-orbits of length $5$.

Now, we fix a point $P\in\Sigma$, and let $\Sigma_P$ is the $G$-orbit of the point $P$. Then $\Sigma_P\not\subset S$ and $|\Sigma_P|>5$. Choose $\mu\in\mathbb{Q}_{>0}$ such that the~log pair  $(X,\mu\mathcal{M})$ is strictly log canonical at~$P$. Then 
$$
\mu<\frac{3}{n}
$$ 
by \cite[Remark 3.6]{CheltsovSarikyanZhuang2024}. Recall from \cite{CheltsovShramov2016} that
$$
\mathrm{Nklt}\big(X,\mu\mathcal{M}\big)=\big\{P\in X\ \text{such that $(X,\mu\mathcal{M})$ is not Kawamata log terminal at $P$}\big\}.
$$
We claim that the locus $\mathrm{Nklt}(X,\mu\mathcal{M})$ does not contain curves that are not disjoint from $\Sigma_P$. Indeed, let $M$ and $M^\prime$ be general surfaces in $\mathcal{M}$.
If $Z$ is a $G$-irreducible curve in $\mathrm{Nklt}(X,\mu\mathcal{M})$, then it follows from \cite[Lemma 1.8]{Corti2000} that
$$
\big(M\cdot M^\prime\big)_Z\geqslant\frac{4}{\mu^2}>\frac{4n^2}{9}
$$
which gives $\mathrm{deg}(Z)\leqslant 6$, so $Z\subset S$, and, therefore $Z\cap\Sigma_P=\varnothing$. In fact, using \cite[Theorem~1.2]{DemaillyPham2014}, we get $\mathrm{deg}(Z)<6$, which would imply that $\mathrm{Nklt}(X,\mu\mathcal{M})$ consists of finitely many points.

We see that the points of the $G$-orbit $\Sigma_P$ are isolated components of the locus $\mathrm{Nklt}(X,\mu\mathcal{M})$. Now, using the~Nadel vanishing theorem as in the~proof of Theorem~\ref{theorem:405-15}, we see that
$$
5<\big|\Sigma_P\big|\leqslant h^0\big(X,\mathcal{O}_{X}(1)\big)=5,
$$
which is absurd. The~obtained contradiction completes the~proof of Proposition~\ref{proposition:A5-standard}.
\end{proof}

By Proposition~\ref{proposition:A5-standard}, to prove Theorem~\ref{theorem:A5}, we may assume that $G=\langle M_2,M_3\rangle$. Then, arguing as in the proof of Lemma~\ref{lemma:A5-standard-equation}, we see that $X$ is given by
\begin{multline*}
\quad \quad \quad \quad x_0^3+x_1^3+x_2^3+x_3^3+x_4^3+b(x_0x_1x_3+\zeta_3^2x_0x_1x_2+\zeta_3x_0x_1x_4+x_2x_3x_4+\\
+x_1x_2x_4+\zeta_3^2x_1x_3x_4+x_0x_3x_4+\zeta_3x_1x_2x_3+x_3x_2x_0+x_0x_2x_4)=0\quad \quad \quad \quad
\end{multline*}
for some $b\in\mathbb{C}$. Furthermore, we have $4b^2-3b+9\ne 0$, and %r := -3/4*w^7 + 3/4*w^5 - 3/2*w^4 + 3/4*w^3 - 3/4*w^2 - 3/2*w + 3/2
$$
b\not\in\{-1,1,-3\},
$$
because the cubic threefold $X$ would be singular otherwise. In the remaining part of this section, we will prove that $X$ is $G$-birationally superrigid arguing as in the proof of \cite[Proposition~7.1]{PinardinZhang2025}.

As in the proof of Proposition~\ref{proposition:A5-standard}, we start with the classification of orbits of small length in $X$.
First, we present some $G$-orbits in $\mathbb{P}^4$. Let
\begin{align*}
\Sigma_5&=\mathrm{Orb}_G\big([0:0:0:0:1]\big),\\
\Sigma_5^\prime&=\mathrm{Orb}_G\big([1:\zeta_3^2:\zeta_3^2:\zeta_3:0]\big).
\end{align*}
Then $\Sigma_5$ and $\Sigma_5^\prime$ are $G$-orbits of length $5$, which are not contained in $X$.
Likewise, let
$$
\Sigma_6=\mathrm{Orb}_G\big([1:\zeta_3:\zeta_3:\zeta_3:1]\big).
$$
Then $\Sigma_6$ is a $G$-orbit of length $6$, which is not contained in $X$.
Similarly, let
$$
\Sigma_{10}=\mathrm{Orb}_G\big([1:\zeta_3:1:0:0]\big).
$$
Then $\Sigma_{10}$ is a $G$-orbit of length $10$, which is not contained in $X$.
Let
\begin{align*}
\Sigma_{12}&=\mathrm{Orb}_G\big([\zeta_{15}:\zeta_{15}^9:1:\zeta_{15}^3:\zeta_{15}^7]\big),\\
\Sigma_{12}^\prime&=\mathrm{Orb}_G\big([\zeta_{15}^4:1:\zeta_{15}^{12}:\zeta_{15}^3:\zeta_{15}]\big).
\end{align*}
Then $\Sigma_{12}$ and $\Sigma_{12}^\prime$ are $G$-orbits of length $12$, which are contained in $X$. Let
$$
\Sigma_{15}=\mathrm{Orb}_G\big([-1:0:1:-1:1]\big),
$$
and let $\mathcal{L}_5$ be the $G$-irreducible curve  in $\mathbb{P}^4$ whose irreducible component is the line
$$
\big\{x_0=x_3, x_1=x_2, x_0=x_4\big\}\subset\mathbb{P}^4.
$$
Then $\Sigma_{15}$ is a $G$-orbit of length~$15$, $\Sigma_{15}\subset X$,  the curve $\mathcal{L}_{5}$ consists of $5$ disjoint lines, $\Sigma_5\cup\Sigma^\prime_5\subset\mathcal{L}_{5}$, the $G$-orbit of every point in $\mathcal{L}_{5}\setminus(\Sigma_5\cup\Sigma^\prime_5)$ has length $15$, and $\mathcal{L}_{5}\cap X$ is a union of three $G$-orbits  
$$
\mathrm{Orb}_G\big(\big[1:t:1:1:1]\big),
$$
where $t\in\mathbb{C}$ such that $t^3+4b+4=0$. Let $\mathcal{L}_{20}$ be the $G$-irreducible curve in $\mathbb{P}^4$ whose irreducible component is the line
$$
\big\{x_0=\zeta_3x_1, x_1=x_2, x_4=0\big\}\subset\mathbb{P}^4.
$$
Note that $\Sigma_5\cup\Sigma^\prime_5\subset\mathcal{L}_{20}$.
Furthermore, the $G$-orbit of every point in $\mathcal{L}_{20}\setminus(\Sigma_5\cup\Sigma^\prime_5)$ has length~$20$,
the curve $\mathcal{L}_{20}$ consists of $20$ disjoint lines, and $\mathcal{L}_{20}\cap X$ is a union of three $G$-orbits 
$$
\mathrm{Orb}_G\big(\big[q:1:1:t:0]\big),
$$
where $t\in\mathbb{C}$ such that $t^3+3b\zeta_3t+b+3=0$. 

\begin{lemma}
\label{lemma:A5-orbits}
Let $\Sigma$ be a $G$-orbit in $X$ of length $|\Sigma|\leqslant 20$. Then $|\Sigma|\in\{12,15,20\}$ and
$$
\Sigma\in\Sigma_{12}\cup \Sigma_{12}^\prime\cup \Sigma_{15}\cup\big(\mathcal{L}_{5}\cap X\big)\cup\big(\mathcal{L}_{20}\cap X\big).
$$
\end{lemma}

\begin{proof}
Let $P$ be a point in $\Sigma$, and let $G_P$ be its stabilizer in $G$. Then $G_P$ is a proper subgroup, because $G$ does not fix points in the cubic threefold $X$. Thus, $|\Sigma|=|G/G_P|\in\{5,6,10,12,15,20\}$, and $G_P$ is isomorphic to one of the following groups:
\begin{center}
$\mathfrak{A}_4$, $\mathfrak{D}_5$, $\mathfrak{S}_3$, $C_5$, $C_2^2$, $C_3$.
\end{center}
Note also that $G$ contains exactly one subgroup isomorphic to $G_P$ up to conjugation.

If $G_P\simeq \mathfrak{A}_4$, then $G_P$ fixes two points in $\mathbb{P}^4$, but none of them is contained in $X$, so that $|\Sigma|\ne 5$. Similarly, if $G_P\simeq \mathfrak{D}_5$ or $G_P\simeq \mathfrak{S}_3$, then $G_P$ fixes one point in $\mathbb{P}^4$, but it is not contained in $X$, which implies that $|\Sigma|\ne 6$ and $|\Sigma|\ne 10$ either.

If $|\Sigma|=12$, then $G_P\simeq C_5$ fixes $5$ points in $\mathbb{P}^5$, but only $4$ of them are contained in $X$, and these points are contained in $\Sigma_{12}\cup\Sigma_{12}^\prime$, which implies that $\Sigma\subset\Sigma_{12}\cup\Sigma_{12}^\prime$,

If $|\Sigma|=15$, then $G_P\simeq C_2^2$ fixes three points in $\Sigma_{15}$, and $G_P$ pointwise fixes an irreducible component of the line $\mathcal{L}_5$, which implies that $\Sigma\subset\Sigma_{15}\cup(\mathcal{L}_{5}\cap X)$.

Finally, if $|\Sigma|=20$, then $G_P\simeq C_3$ fixes fixes a point in $\Sigma_{10}$, and $G_P$ pointwise fixes two irreducible components of the curve $\mathcal{L}_{20}$, which implies that $\Sigma\subset\mathcal{L}_{20}$.
\end{proof}

Now, we describe some reducible $G$-irreducible curves in $X$. Let $\mathcal{L}_6$ be the $G$-irreducible curve in $\mathbb{P}^4$ such that its irreducible component is the line
$$
\left\{\aligned
&x_0-(\zeta_{15}^7+\zeta_{15}^{13})x_1+x_4=0,\\
&\zeta_{15}^5x_0+x_3+(\zeta_{15}+\zeta_{15}^6-\zeta_{15}^{14})x_4=0,\\
&(\zeta_{15}+\zeta_{15}^6-\zeta_{15}^{14})x_0+x_2+\zeta_{15}^5x_4=0.
\endaligned
\right.
$$
and let $\mathcal{L}_6^\prime$ be the $G$-irreducible curve in $\mathbb{P}^4$  such that its irreducible component is the line
$$
\left\{\aligned
&(\zeta_{15}^7+\zeta_{15}^4-\zeta_{15}^3+\zeta_{15}^2+\zeta_{15}-1)x_0+x_1+x_3=0,\\
&(\zeta_{15}^7-\zeta_{15}^3+\zeta_{15}^2)x_1+x_2+x_3=0,\\
&x_0-(\zeta_{15}^4+\zeta_{15})x_1+x_4=0.
\endaligned
\right.
$$
Then $\mathcal{L}_6\cup\mathcal{L}_6^\prime\subset X$. Moreover, each curve $\mathcal{L}_6$ and $\mathcal{L}_6^\prime$ consists of $6$ disjoint lines.

\begin{lemma}
\label{lemma:A5-L6}
Let $Z$ be one of the curves $\mathcal{L}_6$ or $\mathcal{L}_6^\prime$, and let $\mathcal{I}_Z$ be its ideal sheaf in $\mathbb{P}^4$. Then 
$$
h^0\big(\mathbb{P}^4,\mathcal{I}_Z\otimes\mathcal{O}_{\mathbb{P}^4}(3)\big)=11,
$$
and $\mathcal{I}_Z\otimes\mathcal{O}_{\mathbb{P}^4}(3)$ is generated by global sections.
\end{lemma}

\begin{proof}
First, we found a basis of $H^0(\mathbb{P}^4,\mathcal{I}_Z\otimes\mathcal{O}_{\mathbb{P}^4}(3))$ using Maple, and then we used Magma to evrify that the polynomials of this basis determine $Z$ as a subscheme in $\mathbb{P}^4$. 
\end{proof}

For every $t\in\mathbb{C}$, let $\mathcal{L}_{10}^t$ be the $G$-irreducible curve in $\mathbb{P}^4$ whose irreducible component is the line
\begin{equation}
\label{equation:A5-10-lines}
\left\{\aligned
&t(x_0-x_1)+(1-\zeta_3)x_3=0,\\
&t(\zeta_3x_1-x_2)+(1-\zeta_3)x_3=0,\\
&tx_0-x_1+(1+\zeta_3)x_4=0.
\endaligned
\right.
\end{equation}
Then $\mathcal{L}^t_{10}$ consists of $10$ lines, and
$$
\mathcal{L}^t_{10}\subset X\iff t^3+3b\zeta_3t+b+3=0.
$$
Observe that the polynomial $(b+3)t^3+3b\zeta_3t^2+1$ has three distinct roots, so we have exactly three
possibilities for  $\mathcal{L}^t_{10}$ such that $\mathcal{L}^t_{10}\subset X$. Moreover, the following holds.
\begin{itemize}
\item If  If $(b+3)t^3+3b\zeta_3t^2+1=0$ and $t\ne\pm\frac{\zeta_3+2}{3}$, then $\mathcal{L}^t_{10}$ is a union of $10$ disjoint lines. 
\item If $t=\pm\frac{\zeta_3+2}{3}$, then $\mathcal{L}^t_{10}$ is a nodal curve of arithmetic genus $5$, and $\mathrm{Sing}(\mathcal{L}^t_{10})=\Sigma_{15}$.
\end{itemize}

\begin{lemma}
\label{lemma:A5-L10}
Let $Z=\mathcal{L}^t_{10}$ for $t=\pm\frac{\zeta_3+2}{3}$, and let $\mathcal{I}_Z$ be the ideal sheaf of the curve $Z$ in $\mathbb{P}^4$. Then 
$$
h^0\big(\mathbb{P}^4,\mathcal{I}_Z\otimes\mathcal{O}_{\mathbb{P}^4}(3)\big)=10,
$$
and $\mathcal{I}_Z\otimes\mathcal{O}_{\mathbb{P}^4}(3)$ is generated by global sections.
\end{lemma}

\begin{proof}
Explicit computations. For instance, if $t=\frac{\zeta_3+2}{3}$, a basis of  $H^0(\mathbb{P}^4,\mathcal{I}_Z\otimes\mathcal{O}_{\mathbb{P}^4}(3))$ consists of
\begin{multline*}
f_1=-\zeta_3x_0^3+(-1-2\zeta_3)x_0^2x_1+(-\zeta_3-2)x_2x_0^2+\zeta_3x_0x_1^2-\\
4\zeta_3x_0x_1x_3+(2\zeta_3+2)x_4x_0x_1+(1+\zeta_3)x_0x_2^2+4x_3x_2x_0+\\
(-2\zeta_3-2)x_4x_0x_2-2x_3^2x_0+(-1-2\zeta_3)x_4^2x_0+(-1-2\zeta_3)x_1^3+\zeta_3x_1^2x_2+\\
2\zeta_3x_1^2x_4+(\zeta_3-2)x_2^2x_1+(-4\zeta_3-2)x_4x_1x_2+(-1-\zeta_3)x_4^2x_1+\\
(-1-\zeta_3)x_2^3-2x_2^2x_4+2x_3^2x_2-x_4^2x_2-2x_3^2x_4+(-2\zeta_3-2)x_4^3,
\end{multline*}
\begin{multline*}
f_2=(-1-2\zeta_3)x_0^3-3\zeta_3x_0^2x_1+(-3\zeta_3-3)x_2x_0^2+6x_0^2x_4+(-6\zeta_3+6)x_3x_0x_1+\\
(\zeta_3-1)x_0x_2^2+(-6\zeta_3-6)x_4x_0x_2+(-4\zeta_3-2)x_3^2x_0-6\zeta_3x_0x_3x_4-3\zeta_3x_0x_4^2+\\
(-1-2\zeta_3)x_1^3+(3\zeta_3-3)x_1^2x_2+(3\zeta_3-3)x_4x_1^2+(\zeta_3-4)x_2^2x_1+(-12\zeta_3-6)x_4x_1x_2+\\
(-4\zeta_3-5)x_3^2x_1-3x_4^2x_1+(-4\zeta_3-5)x_2^3+(-2\zeta_3+2)x_2^2x_3+(-2\zeta_3-4)x_4x_2^2+\\
(2\zeta_3+4)x_3^2x_2+3\zeta_3x_2x_4^2+(-\zeta_3-2)x_4x_3^2+(-5\zeta_3-4)x_4^3,
\end{multline*}
\begin{multline*}
f_3=(\zeta_3-1)x_0^3+6x_0^2x_2+12x_0^2x_3+(-6\zeta_3-3)x_1^2x_0+(12\zeta_3+6)x_0x_1x_3+\\
(-6\zeta_3-12)x_4x_0x_1+(\zeta_3+5)x_0x_2^2+(-6\zeta_3-6)x_4x_0x_2+(-4\zeta_3-2)x_3^2x_0+\\
(-6\zeta_3-18)x_4x_0x_3+6x_4^2x_0+(10\zeta_3+5)x_1^3+(-6\zeta_3-3)x_4x_1^2+(\zeta_3+2)x_2^2x_1+\\
(12\zeta_3+6)x_4x_1x_2+(5\zeta_3+4)x_3^2x_1+6x_4^2x_1+(2\zeta_3+4)x_2^3+(4\zeta_3+2)x_2^2x_3+\\
(-2\zeta_3+2)x_4x_2^2+(2\zeta_3+4)x_3^2x_2+12x_4^2x_2+(-\zeta_3+7)x_4x_3^2+(\zeta_3-1)x_4^3,
\end{multline*}
\begin{multline*}
f_4=-6\zeta_3x_0x_1x_4-6x_0x_2x_4+12x_1x_2x_3+(3\zeta_3+6)x_1^2x_2+(-6-3\zeta_3)x_2x_0^2+\\
(-3\zeta_3-3)x_4^2x_1+(2\zeta_3-2)x_4x_2^2+(-2\zeta_3-4)x_3^2x_2+(-5\zeta_3-4)x_4x_3^2+(1-\zeta_3)x_2^2x_1+\\
(-1-2\zeta_3)x_3^2x_1+(2\zeta_3-2)x_2^2x_3-3x_0^2x_1-3x_4^2x_0+3x_4^2x_2+6\zeta_3x_1x_2x_4+(-6\zeta_3-6)x_0x_1x_3+\\
6\zeta_3x_0x_3x_4+(-4\zeta_3+1)x_0x_2^2+(4\zeta_3-4)x_3^2x_0+(-3\zeta_3+3)x_4x_1^2+(5\zeta_3-2)x_4^3+\\
(-\zeta_3-2)x_1^3+(\zeta_3-1)x_2^3+(2\zeta_3-2)x_0^3+(-6\zeta_3-3)x_1^2x_0,
\end{multline*}
\begin{multline*}
f_5=12x_1x_3x_4+(-4\zeta_3-2)x_3^2x_0+(-7\zeta_3-5)x_3^2x_1+3\zeta_3x_1^2x_2+(6\zeta_3+9)x_4^2x_0+\\
(3\zeta_3-6)x_4x_1^2+(-3+6\zeta_3)x_4^2x_2+(2\zeta_3+7)x_4x_3^2+(8\zeta_3+10)x_3^2x_2+(-6\zeta_3-3)x_0^2x_1+\\
(\zeta_3-4)x_0x_2^2+(-1-2\zeta_3)x_2^2x_1+(-2\zeta_3-4)x_2^2x_3+(3\zeta_3+6)x_2x_0^2+(3+3\zeta_3)x_4^2x_1+\\
(4\zeta_3+2)x_4x_2^2+6x_1^2x_0+(\zeta_3-1)x_1^3+(1-\zeta_3)x_2^3+(-1-2\zeta_3)x_4^3+(\zeta_3+5)x_0^3+\\
(6\zeta_3+12)x_3x_0x_1+(6\zeta_3+12)x_4x_0x_2+(-12\zeta_3-6)x_4x_0x_3+(-6\zeta_3-6)x_4x_1x_2,
\end{multline*}
\begin{multline*}
f_6=(-5\zeta_3-4)x_0^3+3x_0^2x_1-3\zeta_3x_0^2x_2+3\zeta_3x_0x_1^2+(6\zeta_3+6)x_4x_0x_1+\\
(\zeta_3-1)x_0x_2^2+(-6\zeta_3-6)x_4x_0x_2+(2\zeta_3-2)x_3^2x_0-3x_4^2x_0+(-1-2\zeta_3)x_1^3-\\
3\zeta_3x_1^2x_2+(\zeta_3+2)x_2^2x_1+6x_1x_2x_4+(2\zeta_3-2)x_3^2x_1+(-9\zeta_3-3)x_4^2x_1+\\
(1-\zeta_3)x_2^3+(4\zeta_3+8)x_2^2x_3+(-2\zeta_3-4)x_4x_2^2+(2\zeta_3-2)x_3^2x_2+\\
12x_2x_3x_4+3x_4^2x_2+(2\zeta_3+4)x_4x_3^2+(-2\zeta_3+2)x_4^3,
\end{multline*}
\begin{multline*}
f_7=(4\zeta_3+3)x_0^3+6\zeta_3x_0^2x_2-3\zeta_3x_0x_1^2+6\zeta_3x_0x_1x_3+(-6\zeta_3-6)x_4x_0x_1+\\
(-1-2\zeta_3)x_0x_2^2+(12\zeta_3+6)x_4x_0x_2+(2\zeta_3+4)x_3^2x_0+6x_0x_3x_4+6\zeta_3x_0x_4^2+\\
\zeta_3x_1^3-3\zeta_3x_1^2x_4+(\zeta_3+5)x_2^2x_1+6\zeta_3x_1x_2x_4+(1-\zeta_3)x_3^2x_1+\\
6\zeta_3x_1x_4^2+(2\zeta_3+2)x_2^3+(-2\zeta_3-4)x_2^2x_3+(4\zeta_3+2)x_4x_2^2+\\
(2\zeta_3-2)x_3^2x_2+4x_3^3+(2\zeta_3+1)x_4x_3^2+(4\zeta_3+3)x_4^3,
\end{multline*}
\begin{multline*}
f_8=-x_0^3+(-\zeta_3-2)x_0^2x_1+(1-\zeta_3)x_2x_0^2-2x_1^2x_0+4x_0x_1x_2+(-2\zeta_3-2)x_3x_0x_1+\\
(\zeta_3+3)x_0x_2^2+(-2\zeta_3-2)x_4x_0x_2-2x_3^2x_0+(2\zeta_3-4)x_4x_0x_3-\zeta_3x_0x_4^2+\\
(2\zeta_3+1)x_1^3+(1+\zeta_3)x_1^2x_2+(\zeta_3+3)x_4x_1^2-\zeta_3x_1x_2^2+(4\zeta_3+2)x_4x_1x_2+\\
(2\zeta_3+1)x_3^2x_1+(-2\zeta_3+1)x_4^2x_1+(2\zeta_3+1)x_2^3+(2\zeta_3+2)x_3x_2^2-\\
2\zeta_3x_2x_3^2+(-\zeta_3+2)x_4^2x_2-\zeta_3x_3^2x_4+\zeta_3x_4^3,
\end{multline*}
\begin{multline*}
f_9=(\zeta_3+2)x_0^3+(-1-\zeta_3)x_1^3+(-2\zeta_3-2)x_2^3+(-1-2\zeta_3)x_3^2x_1+\\
2\zeta_3x_2^2x_4-2x_1^2x_2-x_2^2x_1-3\zeta_3x_4^3+(\zeta_3-3)x_4x_1^2+(\zeta_3-2)x_0x_2^2+\\
(2\zeta_3+4)x_2x_0^2+(4\zeta_3+2)x_4^2x_1+(2\zeta_3-2)x_4^2x_2+(\zeta_3+2)x_4x_3^2+\\
(-2\zeta_3-2)x_2^2x_3+(4\zeta_3+2)x_3^2x_2+(-2\zeta_3+2)x_0^2x_1+(\zeta_3+3)x_1^2x_0+\\
(-2\zeta_3+2)x_3^2x_0-6\zeta_3x_0x_3x_4+(-6\zeta_3-6)x_4x_1x_2+(6\zeta_3+6)x_0x_1x_3+\\
(4\zeta_3+2)x_4x_0x_1+(2\zeta_3+4)x_4x_0x_2+4x_1^2x_3,
\end{multline*}
\begin{multline*}
f_{10}=-\zeta_3x_0^3+(2\zeta_3+4)x_1x_0^2+(\zeta_3-1)x_1^2x_0+(2\zeta_3-2)x_0x_1x_3+\\
(4\zeta_3+2)x_4x_0x_1+(-\zeta_3+2)x_0x_2^2+(-2\zeta_3-4)x_4x_0x_2+(2\zeta_3+2)x_3^2x_0+\\
(2\zeta_3+4)x_4x_0x_3-2x_4^2x_0+(1+\zeta_3)x_1^3+(-2\zeta_3+2)x_1^2x_2+(1-\zeta_3)x_4x_1^2+\\
x_2^2x_1+(6\zeta_3+6)x_4x_1x_2+x_3^2x_1-2\zeta_3x_1x_4^2+(2\zeta_3+2)x_3x_2^2-\\
2\zeta_3x_2^2x_4-2x_3^2x_2-2\zeta_3x_2x_4^2+\zeta_3x_3^2x_4+4x_3x_4^2+(\zeta_3+2)x_4^3.
\end{multline*}
Using Magma one can check that the subscheme defined by these polynomials is indeed $Z$. 
\end{proof}

Now, we are ready to prove the following result. 

\begin{lemma}
\label{lemma:A5-reducible-curves}
Let $Z$ be a $G$-invariant reducible curve of degree $<12$. Then
\begin{itemize}
\item either $Z=\mathcal{L}_6$ or $Z=\mathcal{L}_6^\prime$,
\item or $Z=\mathcal{L}^t_{10}$ for some $t\in\mathbb{C}$ such that $(b+3)t^3+3b\zeta_3t^2+1=0$.
\end{itemize}
\end{lemma}

\begin{proof}
Let $C$ be an irreducible component of the curve $Z$, and let $G_C$ be its stabilizer in $G\simeq\mathfrak{A}_5$.
Then one of the following cases hold:
\begin{itemize}
\item $Z$ is a union of $5$ lines, and $G_C\simeq \mathfrak{A}_4$,
\item $Z$ is a union of $5$ conics, and $G_C\simeq \mathfrak{A}_4$,
\item $Z$ is a union of $6$ lines, and $G_C\simeq \mathfrak{D}_5$,
\item $Z$ is a union of $10$ lines, and $G_C\simeq \mathfrak{S}_3$.
\end{itemize}
Let us deal with these cases one by one.

Suppose that $Z$ is a union of $5$ lines. Then the restriction of the representation $\mathbb{V}_5$ to $G_C\simeq \mathfrak{A}_4$ splits as a sum of an irreducible three-dimensional representation, trivial representation, and sign representation. Hence, there is a unique choice for $C$, so $C$ must be an irreducible component of the curve $\mathcal{L}_5$, which is not contained in $X$, which is a contradiction.

Suppose that $Z$ is a union of $5$ conics. Let $\Pi$ be the plane in $\mathbb{P}^4$ that contains $C$. Then
$$
X\cap \Pi=C+L
$$
for some line $L$, which must be an irreducible component of a $G$-irreducible curve consisting of five lines, because $G$ does not leave any line in $\mathbb{P}^4$ invariant. However, we already proved that $X$ does not contain $G$-irreducible curve consisting of five lines, so $Z$ is not a union of $5$ conics.

Suppose that $Z$ is a union of $6$ lines. Then the restriction of the representation $\mathbb{V}_5$ to $G_C\simeq \mathfrak{D}_5$ splits as a sum of two non-isomorphic irreducible two-dimensional representations, and trivial representation. This implies that $C$ is an irreducible component of one of the curves $\mathcal{L}_6$ or $\mathcal{L}_6^\prime$, and, therefore either $Z=\mathcal{L}_6$ or $Z=\mathcal{L}_6^\prime$ as claimed.

Suppose that $Z$ is a union of $10$ lines. The restriction of the representation $\mathbb{V}_5$ to~\mbox{$G_C\simeq \mathfrak{S}_3$} splits as a sum of two isomorphic irreducible two-dimensional representations, and trivial representation, so the line $C$ corresponds to a two-dimensional irreducible representation. Then $C$ is an irreducible component of the curve $\mathcal{L}^t_{10}$ for some $t\in\mathbb{C}$. Indeed, since $G$ contains a unique subgroup isomorphic to $\mathfrak{S}_3$ up to conjugation, we may assume that $G_C$ is generated by the following transformations:
\begin{align*}
T_1\colon [x_0:x_1:x_2:x_3:x_4]&\mapsto[x_2:x_1:x_0:x_3:x_4],\\
T_2\colon [x_0:x_1:x_2:x_3:x_4]&\mapsto[\zeta_3^2x_1:\zeta_3x_2:x_0:\zeta_3x_3:\zeta_3^2x_4].
\end{align*}
But $C$ has a unique $G_C$-orbits of length $2$, and the points of this orbit must be fixed by $\langle T_2\rangle\simeq C_3$. Computing the eigenvalues and eigenvectors of the matrix that corresponds to $T_2$, we see that this $G_P$-orbit of length $2$ consists of the points $[q:1:1:t:0]$ and $[1:1:1:0:t]$ for some $t\in\mathbb{C}$, which implies that $C$ is given by \eqref{equation:A5-10-lines}. Therefore, $C$ is  component of the curve $\mathcal{L}^t_{10}$ for some $t\in\mathbb{C}$. Now, using $C\subset X$, we get $(b+3)t^3+3b\zeta_3t^2+1=0$.
\end{proof}

Now, let us introduce $G$-irreducible surfaces in $X$ that has small degree. To do this, we set
\begin{multline*}
\quad \quad \quad \quad \quad \quad 
f_2=x_1x_2+x_1x_3+x_1x_4+\zeta_3^2x_2x_3+x_2x_4+\\
+\zeta_3x_3x_4+x_0x_1+\zeta_3x_0x_2+x_0x_3+\zeta_3^2x_0x_4,
\quad \quad \quad \quad \quad \quad
\end{multline*}
\begin{multline*}
\quad \quad \quad \quad \quad \quad f_3=x_0x_1x_3+\zeta_3^2x_0x_1x_2+\zeta_3x_0x_1x_4+x_2x_3x_4+x_1x_2x_4+\\
+\zeta_3^2x_1x_3x_4+x_0x_3x_4+\zeta_3x_1x_2x_3+x_3x_2x_0+x_0x_2x_4,\quad \quad \quad \quad \quad \quad
\end{multline*}
\begin{multline*}
\quad \quad \quad \quad \quad \quad f_4=\zeta_3^2x_0^2x_4^2+\zeta_3^2x_2^2x_3^2+x_3^2x_4^2+x_1^2x_4^2\zeta_3+\\
+\zeta_3x_0^2x_1^2+\zeta_3x_0^2x_3^2+\zeta_3x_1^2x_2^2+\zeta_3x_1^2x_3^2+\zeta_3x_2^2x_4^2+x_2^2x_0^2.\quad \quad \quad \quad \quad \quad
\end{multline*}
Let $Q=\{f_2=0\}\cap X$, let $S=\{f_3=0\}\cap X$, and let $S_\lambda$ be the surface in $X$ that is cut out by
$$
f_4+\lambda f_2^2=0,
$$
where $\lambda\in\mathbb{C}$. Then $Q$, $S$, $S_\lambda$ are $G$-invariant surfaces.

\begin{lemma}
\label{lemma:A5-Q-S-orbits}
The following assertions hold:
\begin{itemize}
\item[($\mathrm{i})$] $Q$ contains $\Sigma_{12}$ and $\Sigma_{12}^\prime$;
\item[($\mathrm{ii})$] $Q$ does not contain $G$-orbits of length $15$;
\item[($\mathrm{iii})$] $Q$ contains all $G$-orbits in $X$ of length $20$;
\item[($\mathrm{iv})$] $Q$ contains neither $\mathcal{L}_6$ nor $\mathcal{L}_6^\prime$;
\item[($\mathrm{v})$] $Q$ contains $\mathcal{L}_{10}^t$ if and only if $(b+3)t^3+3b\zeta_3t^2+1=0$, $11b^2+174b+207=0$ 
and 
$$
t=\frac{(11b + 150)\zeta_3}{63};
$$
\item[($\mathrm{vi})$] $S$ contains $\Sigma_{12}$ and $\Sigma_{12}^\prime$;
\item[($\mathrm{vii})$] $S$ contains $\Sigma_{15}$, but $S\cap(\mathcal{L}_5\cap X)=\varnothing$;
\item[($\mathrm{viii})$] $S$ does not contain $G$-orbits of length $20$;
\item[($\mathrm{ix})$] $S$  contains $\mathcal{L}_6$ and $\mathcal{L}_6^\prime$; 
\item[($\mathrm{x})$] $S$ does not contains $\mathcal{L}_{10}^t$ for any $t\in\mathbb{C}$ such that $(b+3)t^3+3b\zeta_3t^2+1=0$;
\item[($\mathrm{xi})$] $S_\lambda$ contains $\Sigma_{12}$ and $\Sigma_{12}^\prime$;
\item[($\mathrm{xii})$] $S_\lambda$ contains $\Sigma_{15}$ if and only if $\lambda=0$, and
$$
\big(\mathcal{L}_5\cap X\big)\cap S_\lambda\ne\varnothing \iff \big(\mathcal{L}_5\cap X\big)\subset S_\lambda \iff c=-\frac{\zeta_3}{4};
$$
\item[($\mathrm{xiii})$] $S_\lambda$ contains all $G$-orbits in $X$ of length $20$;
\item[($\mathrm{xiv})$] $S_\lambda$ contains $\mathcal{L}_6$ or $\mathcal{L}_6^\prime$ if and only if $\lambda=-\frac{\zeta_3}{5}$, 
\item[($\mathrm{xv})$] $S_\lambda$ contains $\mathcal{L}_{10}^t$ if and only if $(b+3)t^3+3b\zeta_3t^2+1=0$, $6\zeta_3t-3t^2-\zeta_3-1\ne 0$ and
$$
\lambda=\frac{1+\zeta_3-6t^2-9\zeta_3t^4}{(6\zeta_3t-3t^2-\zeta_3-1)^2}.
$$
\end{itemize}
\end{lemma}

\begin{proof}
Explicit computations.
\end{proof}

Set $\mathcal{C}_{18}=Q\cap S$. Then $\mathcal{C}_{18}$ is a $G$-invariant curve of degree $18$.

\begin{lemma}
\label{lemma:A5-Q-S-smooth}
Surfaces $Q$, $S$, $S_\lambda$ are irreducible,
$\mathcal{C}_{18}$ is irreducible, $Q$ is smooth, 
\begin{equation}
\label{equation:S-C-singularities}
\mathrm{Sing}(\mathcal{C}_{18})=\mathrm{Sing}(S)=\Sigma_{12}\cup\Sigma_{12}^\prime,
\end{equation}
and $S$ has isolated ordinary double points (nodes) at $\Sigma_{12}\cup\Sigma_{12}^\prime$. If $b\in\mathbb{C}$ is general, $S_\lambda$ is normal. The surface $S_\lambda$ is smooth at every point of $\Sigma_{12}\cup\Sigma_{12}^\prime$, and $S_\lambda$ is smooth at every point of  $\mathcal{L}_{20}\cap X$. 
One has $\Sigma_{15}\subseteq \mathrm{Sing}(S_0)$, and 
\begin{itemize}
\item either $5b^2-6b+9\ne 0$, and $S_0$ has isolated ordinary double points at $\Sigma_{15}$;
\item or $5b^2-6b+9=0$, and $S_0$ has Du Val singular points of type $\mathbb{A}_3$ at $\Sigma_{15}$.
\end{itemize}
If $c=-\frac{\zeta_3}{4}$, then $\mathcal{L}_5\cap X\subseteq\mathrm{Sing}(S_\lambda)$, and $S_\lambda$ has isolated ordinary double points at $\mathcal{L}_5\cap X$.
\end{lemma}

\begin{proof}
Note that $S$ and $\mathcal{C}_{18}$ do not depend on $b$, so we can use Magma to verify \eqref{equation:S-C-singularities}, and to check that $S$ has an isolated ordinary double singularity (a node) at every point of $\Sigma_{12}\cup\Sigma_{12}^\prime$.

Explicit computations show that $Q$ is smooth at every point of $\Sigma_{12}\cup\Sigma_{12}^\prime$, so $Q$ has isolated singularities, because $\mathcal{C}_{18}=Q\cap S$ is smooth away from $\Sigma_{12}\cup\Sigma_{12}^\prime$. Hence, $Q$ is normal and $K_Q\sim 0$, so it follows from \cite{Umezu} that $Q$ can have at most two non-Du Val singular points, which also follows from \cite[Theorem 6.9]{Shokurov92} and \cite{Fujino}.
Thus, by Lemma~\ref{lemma:A5-orbits}, $Q$ has  Du Val singularities.
Let $\widetilde{Q}$ be the minimal resolution of singularities of the surface $Q$. Then $\widetilde{Q}$ is a smooth K3 surface, so that $\mathrm{rk}\,\mathrm{Pic}(\widetilde{Q})\leqslant 19$, which gives $|\mathrm{Sing}(Q)|<19$. Thus, the surface $Q$ is smooth by Lemma~\ref{lemma:A5-orbits}, because $Q$ does not contains $G$-orbits $\Sigma_{15}$ and $\Sigma_{15}^\prime$.

If $b\in\mathbb{C}$ is general, then $S_\lambda\cap Q=\{f_4=0,f_2=0\}\cap X$ is a smooth irreducible curve, which implies that $S_\lambda$ has isolated singularities, so $S_\lambda$ is normal.

The remaining assertions can be checked using Magma and Maple.
\end{proof}

\begin{remark}
\label{remark:A5-quartics-normal}
We believe that $S_\lambda$ is normal for every $\lambda,b\in\mathbb{C}$. However, we were unable to confirm this computationally using Magma or Macaulay2, as the problem is too complex.
\end{remark}

Now, we study irreducible $G$-invariant curves in surfaces $Q$ and $S$.

\begin{lemma}
\label{lemma:A5-irreducible-curves-in-Q}
Let $C$ be an irreducible $G$-invariant curve contained in $Q$, and let $d$ be its degree. Suppose that $d<12$. Then $C$ is smooth, and one of the following two cases holds:
\begin{itemize}
\item either $d=8$ and $C\simeq\mathbb{P}^1$,
\item or $d=10$, and $C$ is a curve of genus $6$.
\end{itemize}
\end{lemma}

\begin{proof}
Let $\overline{C}$ be the normalization of the curve $C$, and let $g$ be the genus of the smooth curve $\overline{C}$. Since $\mathbb{V}_5$ is irreducible representation, $G$ acts faithfully on the curve $C$, and its action lifts to $\overline{C}$. Moreover, if $g\leqslant 16$, it follows from Lemma~\ref{lemma:A5-curves} that
$$
g\in\{0,4,5,6,9,10,11,13,15\},
$$
and the length of a $G$-orbits in $\overline{C}$ is $12$, $20$ or $30$.

Since $\mathcal{C}_{18}=Q\cap S$ is an irreducible curve of degree $18$, we see that $C\ne \mathcal{C}_{18}$. Thus, it follows from Lemma~\ref{lemma:A5-orbits} and the description of the $G$-orbits in $Q$ and $S$ that
$$
33\geqslant 3d=\mathcal{C}_{18}\cdot C=12\times 2\times a+30\times b,
$$
for some non-negative integers $a$ and $b$, because $\mathcal{C}_{18}$ is singular at $\Sigma_{12}\cup\Sigma_{12}^\prime$,
and $\mathcal{C}_{18}$ does not contain $G$-orbits of length $20$. Hence, we see that either $(d,a,b)=(8,1,0)$ or $(d,a,b)=(10,0,1)$.
Moreover, since $\Sigma_{12}$ or $\Sigma_{12}^\prime$ are the only $G$-orbits of length $12$ in $X$ by Lemma~\ref{lemma:A5-orbits}, we conclude that one of the following two cases holds:
\begin{enumerate}
\item $d=8$, the curve $C$ contains one orbit among $\Sigma_{12}$ or $\Sigma_{12}^\prime$, and $C$ is smooth at this orbit;
\item $d=10$, and the curve $C$ contains neither $\Sigma_{12}$ nor $\Sigma_{12}^\prime$.
\end{enumerate}

Recall from Lemma~\ref{lemma:A5-Q-S-smooth} that $Q$ is a smooth K3 surface.
Let $H_Q$ be its general hyperplane section. Then it follows from the Hodge index theorem that
$$
6C^2-d^2=\mathrm{det}\begin{pmatrix}
C^2 &d \\
d & 6
\end{pmatrix}=\mathrm{det}\begin{pmatrix}
C^2 &H_{Q}\cdot C \\
H_{Q}\cdot C & H_{Q}^2
\end{pmatrix}\leqslant 0.
$$
Therefore, it follows from the adjunction formula that
$$
2g-2\leqslant 2\mathrm{p}_a(C)-2=C^2\leqslant\Big\lfloor\frac{d^2}{6}\Big\rfloor=
\left\{\aligned
&10\ \text{if $d=8$},\\
&16\ \text{if $d=10$},
\endaligned
\right.
$$
where $\mathrm{p}_a(C)$ is the arithmetic genus of the curve $C$.
Therefore, we have
$$
g\leqslant \mathrm{p}_a(C)\leqslant \left\{\aligned
&6\ \text{if $d=8$},\\
&9\ \text{if $d=10$}.
\endaligned
\right.
$$
Hence, if $d=8$, then $g\in\{0,4,5,6\}$. Similarly, if $d=10$, then $g\in\{0,4,5,6,9\}$

We claim that $C$ is smooth. Indeed, if $C$ is singular, $\mathrm{Sing}(C)$ is a union of orbits $\Sigma_1,\ldots\Sigma_r$, so
$$
0\leqslant g=\mathrm{p}_a(C)-\sum_{i=1}^rn_i|\Sigma_i|\leqslant 9-\sum_{i=1}^rn_i|\Sigma_i|
$$
for some positive integers $n_1,\ldots,n_r$, which contradicts Lemmas~\ref{lemma:A5-orbits}. Thus, $C$ is a smooth curve.

If $d=10$, then $C$ does not have $G$-orbits of length $12$, so it follows from Lemma~\ref{lemma:A5-curves} that $g=6$. Similarly, if $d=8$, the curve $C$ has one $G$-orbit of length $12$, so $g\ne 6$ and $g\ne 4$ by Lemma~\ref{lemma:A5-curves}. Thus, one of the following holds:
\begin{itemize}
\item $d=8$ and $g=0$,
\item $d=8$ and $g=5$,
\item $d=10$ and $g=6$.
\end{itemize}

Suppose that $d=8$ and $g=5$. To complete the proof, we must show that this case is impossible.
In this case, we have $C^2=8$ on the surface $Q$, which gives
$$
\big(2H_{Q}-C\big)^2=0.
$$
Thus, it follows from the~Riemann--Roch formula and Serre duality that
$$
h^0\big(Q,\mathcal{O}_{Q}(2H_Q-C)\big)\geqslant 2.
$$
In particular, the~linear system $|2H_Q-C|$ is not empty. On the other hand, this linear system does not have base curves, because curves in $|2H_Q-C|$ have degree $4$. Thus, since $(2H_Q-C)^2=0$, the linear system $|2H_Q-C|$ is basepoint free. Moreover, general curve in $|2H_Q-C|$ is irreducible, since otherwise $S$ would be covered by rational curves, which is impossible, since $S$ is not uniruled. Hence, $h^0(Q,\mathcal{O}_{Q}(2H_Q-C))=2$ \cite{SD}.
We have the following embedding of $G$-representations:
$$
H^0(Q,\mathcal{O}_{Q}(2H_Q-C))\hookrightarrow H^0\big(Q,\mathcal{O}_{Q}(2H_Q)\big)\simeq H^0\big(X,\mathcal{O}_X(2)\big).
$$
On the other hand, GAP computations show that $H^0(X,\mathcal{O}_X(2))$ dooes not have two-dimensional subrepresentations. The obtained contradiction completes the proof.
\end{proof}

\begin{lemma}
\label{lemma:A5-irreducible-curves-in-S}
The surface $R$ does not contain $G$-invariant curves of degree $<12$.
\end{lemma}

\begin{proof}
Let $C$ be an irreducible $G$-invariant curve contained in $S$, let $\overline{C}\to C$ be its normalization, and let $g$ be the genus of $\overline{C}$. As in the proof of Lemma~\ref{lemma:A5-irreducible-curves-in-Q}, we see that $G$ acts faithfully on $C$, and this action lifts to $\overline{C}$. Moreover, if $g\leqslant 15$, then  Lemma~\ref{lemma:A5-curves} gives
$$
g\in\{0,4,5,6,9,10,11,13,15\}.
$$
Similarly, it follows from  Lemma~\ref{lemma:A5-curves} that the $G$-orbits in $\overline{C}$ are $12$, $20$, $30$.

Let $d$ be the degree of the curve $C$. Suppose that $d<12$. Let us seek for a contradiction.

Observe that $C\ne \mathcal{C}_{18}=Q\cap S$, because $\mathcal{C}_{18}$ is an irreducible curve of degree $18$. Thus, $C\not\subset Q$. On the other hand, it follows from Lemma~\ref{lemma:A5-orbits} that $G$-orbits in $Q$ have length $12$, $20$, $30$ or $60$, which also follows from \cite[Lemma 6.7.1]{CheltsovShramov2016}. Therefore, we have
$$
2d=Q\cdot C=12\times a+20\times b+30\times c
$$
for some non-negative integers $a$, $b$, $c$, and, therefore $11\geqslant d=6a+10b+15c$, which gives $c=0$, and either $(d,a,b)=(6,1,0)$ or $(d,a,b)=(10,0,1)$. Thus, we see that either $d=6$ or $d=10$. Moreover, our computations and Lemma~\ref{lemma:A5-orbits} also imply that one of the following cases holds:
\begin{enumerate}
\item $d=6$, the curve $C$ contains one orbit of length $12$, and $C$ is smooth at this orbit;
\item $d=10$, and $C$ does not contain $G$-orbits of length $12$.
\end{enumerate}

Let $H_S$ be a general hyperplane section of the surface $S$. Then it follows from that Hodge index theorem that
$$
9C^2-d^2=\mathrm{det}\begin{pmatrix}
C^2 &d \\
d & 9
\end{pmatrix}=\mathrm{det}\begin{pmatrix}
C^2 &H_{S}\cdot C \\
H_{S}\cdot C & H_S^2
\end{pmatrix}\leqslant 0,
$$
which gives
$$
C^2\leqslant \Big\lfloor\frac{d^2}{9}\Big\rfloor=
\left\{\aligned
&4\ \text{if $d=6$},\\
&11\ \text{if $d=10$}
\endaligned
\right.
$$

Suppose that $d=10$. Then $C$ is contained in the smooth locus of the surface $S$, because the surface $S$ is smooth away from the locus $\Sigma_{12}\cup\Sigma_{12}^\prime$ by Lemma~\ref{lemma:A5-Q-S-smooth}, and $C$ does not contain $G$-orbits of length $12$. Thus, if $d=10$, then it follows from the adjunction formula that
$$
2g-2\leqslant 2\mathrm{p}_a(C)-2=(K_S+C)\cdot C=d+C^2=10+C^2\leqslant 21,
$$
where $\mathrm{p}_a(C)$ is the arithmetic genus of the curve $C$. Hence, we have $g\leqslant  \mathrm{p}_a(C)\leqslant 11$, so arguing as in the proof of Lemma~\ref{lemma:A5-irreducible-curves-in-Q}, we see that $C$ is smooth.
Then $g\in\{0,4,5,6,9,10,11\}$, and it follows from  Lemma~\ref{lemma:A5-curves} that $C$ has a $G$-orbit of length $12$ or $20$. But $C$ has no $G$-orbits of length~$12$, and it does not contain $G$-orbits of length $20$ by Lemma~\ref{lemma:A5-Q-S-orbits}.

Thus, $d=6$, and $C$ has exactly one $G$-orbit of length $12$, which must be $\Sigma_{12}$ or $\Sigma_{12}^\prime$ by Lemma~\ref{lemma:A5-orbits}. Without loss of generality, we may assume that $\Sigma_{12}\subset C$, and $\Sigma_{12}^\prime\not\subset C$. Recall that $C$ is smooth at the points of $\Sigma_{12}$, and it follows from  Lemma~\ref{lemma:A5-Q-S-smooth} that $S$ has nodes at points of $\Sigma_{12}$.

Let $\pi\colon\widetilde{S}\rightarrow S$ be the blowup of  $\Sigma_{12}$, let $E$ be the $\pi$-exceptional divisor, and $\widetilde{C}$ be the strict transform on $\widetilde{S}$ of the curve $C$. Then, since $C$ is smooth at the points of $\Sigma_{12}$, we have
$$
\widetilde{C}\sim_{\mathbb{Q}} \pi^*(C)-\frac{1}{2}E.
$$
Then, it follows from the adjunction formula that
$$
2g-2\leqslant 2\mathrm{p}_a\big(\widetilde{C}\big)-2=\big(K_{\widetilde{S}}+\widetilde{C}\big)\cdot\widetilde{C}=\Big(\pi^*\big(\mathcal{O}_{X}(1)\vert_{S}\big)+\pi^*(C)-\frac{1}{2}E\Big)\cdot\widetilde{C}=C^2\leqslant 4,
$$
where $\mathrm{p}_a(\widetilde{C})$ is the arithmetic genus of the curve $\widetilde{C}$.
Thus, $0\leqslant g\leqslant\mathrm{p}_a(\widetilde{C})\leqslant 3$, which gives $g=0$.

Now, arguing as in the proof of Lemma~\ref{lemma:A5-irreducible-curves-in-Q}, we see that the curve $\widetilde{C}$ is smooth, so $C$ is also smooth. Thus, $C$ contains a $G$-orbit of length $20$ by \cite[Lemma 5.1.5]{CheltsovShramov2016}, but $S$ does not contain $G$-orbits of length $20$, which is a contradiction.
 \end{proof}

\begin{corollary}
\label{corollary:A5-irreducible-curves}
Let $C$ be an irreducible $G$-invariant curve in $X$ of degree $d<12$ such that $C\not\subset Q$.
Then $d=10$, and $C$ does not contain $G$-orbits of length $12$ or $15$.
\end{corollary}

\begin{proof}
Recall that all $G$-orbits in the smooth K3 surface $Q$ are of length $12$, $20$, $30$ and $60$, which also follows from \cite[Lemma 6.7.1]{CheltsovShramov2016}. Thus, since $C\not\subset Q$, we have
$$
2d=Q\cdot C=12\times a+20\times b+30\times c
$$
for some non-negative integers $a$, $b$, $c$. So, either $(d,a,b,c)=(6,1,0,0)$ or $(d,a,b,c)=(10,0,1,0)$. 

Recall that $C\not\subset S$ by Lemma~\ref{lemma:A5-irreducible-curves-in-S}, 
and $S$ is singular at the points of $\Sigma_{12}\cup\Sigma_{12}^\prime$ by Lemma~\ref{lemma:A5-Q-S-smooth}.
Moreover, it follows from Lemmas~\ref{lemma:A5-orbits} and \ref{lemma:A5-Q-S-orbits} that $\Sigma_{15}$ is the only $G$-orbit of length $15$ that is contained in $S$, and it follows from Lemma~\ref{lemma:A5-curves} that 
$$
\Sigma_{15}\cap\big(C\setminus\mathrm{Sing}(C)\big)=\varnothing.
$$
Thus, it follows from Lemma~\ref{lemma:A5-Q-S-orbits} that 
$$
3d=S\cdot C=12\times 2\times a^\prime+15\times 2\times b^\prime,
$$
for some non-negative integers $a^\prime$ and $b^\prime$. Hence, either $(d,a^\prime,b^\prime)=(8,1,0)$ or $(d,a^\prime,b^\prime)=(10,0,1)$. Therefore, $d=10$, and $C$ does not contain $G$-orbits of length $12$ by Lemmas~\ref{lemma:A5-orbits} and \ref{lemma:A5-Q-S-orbits}.

To complete the proof, we must show that the curve $C$ does not contain $G$-orbits of length $15$.
Suppose that this is not the case. Then $C$ contains a $G$-orbit $\Sigma$ such that $|\Sigma|=15$. 
Then it follows from Lemma~\ref{lemma:A5-orbits} that either $\Sigma=\Sigma_{15}$ or $\Sigma\subset\mathcal{L}_5\cap X$. Moreover, by Lemma~\ref{lemma:A5-curves}, we have 
$$
\Sigma\subset\mathrm{Sing}(C).
$$
Fix a point $O\in \Sigma$. Let $\mathcal{M}$ be the linear system in $|\mathcal{O}_Q(3)|$ consisting of all surfaces that contains the $G$-orbit $\Sigma$ and singular at $O$. Then explicit computations shows that the base locus of $\mathcal{M}$ does not contain curves of degree $\ne 1$, so, in particular, the curve $C$ is not contained in this base locus. Let $M$ be a general surface in $\mathcal{M}$. Then $C\not\subset M$, so that
\begin{multline*}
30=M\cdot C=\sum_{P\in C\cap M}\big(M\cdot C\big)_P\geqslant \sum_{P\in \Sigma}\big(M\cdot C\big)_P\geqslant \sum_{P\in \Sigma}\mathrm{mult}_P(M)\mathrm{mult}_P(C)\geqslant\\ 
\geqslant 2\sum_{P\in\Sigma}\mathrm{mult}_P(M)\geqslant 
4+2\sum_{P\in\Sigma\setminus O}\mathrm{mult}_P(M)\geqslant 4+\sum_{P\in\Sigma\setminus O}2=32,
\end{multline*}
which is absurd.
\end{proof}

\begin{lemma}
\label{lemma:A5-irreducible-curves-degree-10}
Let $C$ be an irreducible $G$-invariant curve in $X$ of degree $d<12$ such that $C\not\subset Q$.
Then $d=10$, $\Sigma_{12}\not\subset C$, $\Sigma_{12}^\prime\not\subset C$, and $C\subset S_\lambda$ for some $\lambda\in\mathbb{C}$. Further, if the surface $S_\lambda$~is~normal, then $C$ is a smooth curve of degree $10$ and genus $6$.
\end{lemma}

\begin{proof}
Let $\overline{C}\to C$ be be the normalization of the curve $C$, and let $g$ be the genus of the curve~$\overline{C}$. Arguing, as in the proof of Lemma~\ref{lemma:A5-irreducible-curves-in-Q}, we see that $G$ acts faithfully on $C$, and the action lifts to~$\overline{C}$. If $g\leqslant 15$, then it follows from Lemma~\ref{lemma:A5-curves} that
$$
g\in\{0,4,5,6,9,10,11,13,15\}.
$$
By  Lemma~\ref{lemma:A5-curves}, the length of the $G$-orbits in $\overline{C}$ are $12$, $20$, $30$. So, if $C$ has a $G$-orbit of length $15$, then $C$ is singular at every point of this orbit. By Corollary~\ref{corollary:A5-irreducible-curves}, $d=10$, $\Sigma_{12}\not\subset C$ and $\Sigma_{12}^\prime\not\subset C$.

Let $P$ be a general point in $C$, and let $\Sigma_P$ be its $G$-orbit. Then $|\Sigma|=60$, and there is $\lambda\in\mathbb{C}$ such that $P\in S_\lambda$. Moreover, if $C\not\subset S_\lambda$, then 
$$
40\geqslant 4d=S_\lambda\cdot C\geqslant |S_\lambda\cap C|\geqslant |\Sigma_P|=60.
$$
Hence, we conclude that $C\subset S_\lambda$. 

Now, we suppose that $S_\lambda$ is normal, which is always the case if $b\in\mathbb{C}$ is general, cf. Remark~\ref{remark:A5-quartics-normal}.

Set $H_{S_{\lambda}}=H\vert_{S_{\lambda}}$. Then $K_{S_{\lambda}}\sim 2H_{S_{\lambda}}$, and it follows from the Hodge index theorem that
$$
12C^2-100=12C^2-d^2=\mathrm{det}\begin{pmatrix}
C^2 &d \\
d & 12
\end{pmatrix}=\mathrm{det}\begin{pmatrix}
C^2 &H_{S_{\lambda}}\cdot C \\
H_{S_{\lambda}}\cdot C & H_{S_{\lambda}}^2
\end{pmatrix}\leqslant 0.
$$
This gives $C^2\leqslant 8$. Here, we consider intersection of divisors on $S_\lambda$ as in \cite{Sakai1984}.

Suppose that $C$ is contained in the~smooth locus of the surface $S_\lambda$.
Then, by adjunction formula, we have $C^2=2\mathrm{p}_a(C)-2$, where $\mathrm{p}_a(C)$ is the~arithmetic genus of the~curve $C$.
Thus, we have
$$
2g-2\leqslant 2\mathrm{p}_a(C)-2=\big(K_{S_\lambda}+C\big)\cdot C^2=\big(2H_{S_\lambda}+C\big)\cdot C^2=2d+C^2=20+C^2\leqslant 28,
$$
so $g\leqslant 15$. Thus, $g\in\{0,4,5,6,9,10,11,13,15\}$. Since $C$ does not contain $G$-orbits of length $12$, it follows from Lemmas~\ref{lemma:A5-orbits} and \ref{lemma:A5-Q-S-orbits} that the $G$-orbits in $C$ have length $20$, $30$,~$60$. Hence, arguing as in the proof of Lemma~\ref{lemma:A5-irreducible-curves-in-Q}, we see that $C$ is smooth. Then, by Lemma~\ref{lemma:A5-curves}, either $g=6$ or $g=11$. But $g=11$ is impossible by the Castelnuovo bound \cite{ACGH}. Hence, $g=6$ as claimed.

Thus, we may assume that $S_\lambda$ is singular, and $C$ contains some of its singular points. Then 
$$
\mathrm{Sing}(\Sigma_\lambda)\cap C=\Sigma_1\cup\cdots\cup\Sigma_r,
$$ 
where $\Sigma_1,\ldots,\Sigma_r$ are $G$-orbits.  Then it follows from Lemma~\ref{lemma:A5-Q-S-orbits} that
$$
|\Sigma_i|\in\{15,30,60\big\}
$$
for every $i\in\{1,\ldots,r\}$. Without loss of generality, we may assume that $|\Sigma_i|\leqslant |\Sigma_j|$ if $i<j$.

Let $\pi\colon\widetilde{S}_\lambda\to S_\lambda$ to be the minimal resolution of singularities of the points in $\Sigma_1\cup\cdots\cup\Sigma_r$, let $\widetilde{C}$ be the strict transform on $\widetilde{S}_\lambda$ of the curve $C$, and let $E_1,\ldots,E_r$ be the exceptional $G$-irreducible curves that are mapped to $\Sigma_1,\ldots,\Sigma_r$, respectively. Then $$
K_{\widetilde{S}_\lambda}\sim \pi^*\big(K_{S_\lambda}\big)-\sum_{i=1}^{r}a_iE_i\sim \pi^*\big(2H_{S_\lambda}\big)-\sum_{i=1}^{r}a_iE_i
$$
for some non-negative integers $a_1,\ldots,a_r$, and
$$
\widetilde{C}\sim_{\mathbb{Q}} \pi^*(C)-\sum_{i=1}^{r}m_iE_i
$$
for some positive rational numbers $m_1,\ldots,m_r$. Then 
\begin{multline*}
-2\leqslant 2g-2\leqslant 2\mathrm{p}_a\big(\widetilde{C}\big)-2=\big(K_{\widetilde{S}_\lambda}+\widetilde{C}\big)\cdot\widetilde{C}=\\
=\Big(\pi^*\big(2H_{S_\lambda}+C\big)-\sum_{i=1}^{r}(a_i+m_i)E_i\Big)\cdot\widetilde{C}=20+C^2-\sum_{i=1}^{r}(a_i+m_i)E_i\cdot\widetilde{C}\leqslant \\
\leqslant 20+C^2-(a_k+m_k)E_k\cdot\widetilde{C}\leqslant 20+C^2-(a_k+m_k)|\Sigma_k|\leqslant 28-(a_k+m_k)|\Sigma_1|< 28-a_k|\Sigma_k|
\end{multline*}
for every $k$, where $\mathrm{p}_a(\widetilde{C})$ is the arithmetic genus of the curve $\widetilde{C}$. If $|\Sigma_k|\ne 15$ and $a_k>0$, then
$$
2g-2<28-a_k|\Sigma_k|\leqslant 28-|\Sigma_k|\leqslant -2,
$$
which is absurd. If $|\Sigma_k|=15$, then $S_\lambda$ has Du Val singularities at the points of $\Sigma_k$ by Lemma~\ref{lemma:A5-Q-S-smooth}, so in this case we have $a_k=0$. Thus, we conclude that $a_k=0$ for every possible $k$.
This means that $S_\lambda$ has Du Val singularities at every point in $\Sigma_1\cup\cdots\cup\Sigma_r$. Then  Lemma~\ref{lemma:DuVal} gives
\begin{multline*}
\quad \quad -2\leqslant 2g-2\leqslant 2\mathrm{p}_a\big(\widetilde{C}\big)-2=\big(K_{\widetilde{S}_\lambda}+\widetilde{C}\big)\cdot\widetilde{C}=\\
=\big(\pi^*\big(2H_{S_\lambda}\big)+\widetilde{C}\big)\cdot\widetilde{C}=20+\widetilde{C}^2\leqslant 20+C^2-\sum_{i=1}^{r}\frac{1}{2}|\Sigma_i|\leqslant 28-\sum_{i=1}^{r}\frac{1}{2}|\Sigma_i|.
\end{multline*}
Thus, $g\leqslant \mathrm{p}_a\leqslant 8$, because $|\Sigma_i|\in\{15,30,60\}$ for each $i$. So, arguing as in the proof of Lemma~\ref{lemma:A5-irreducible-curves-in-Q}, we see that $\widetilde{C}$ is smooth. 
Then it follows from \cite{lmfdb} that $g=\mathrm{p}_a=6$, because $\widetilde{C}$ does not have $G$-orbits of length $12$, since $C$ does not contain $G$-orbits of length $12$. Then
$$
10=2g-2=20+\widetilde{C}^2\leqslant 20+C^2-\sum_{i=1}^{r}\frac{1}{2}|\Sigma_i|\leqslant 28-\sum_{i=1}^{r}\frac{1}{2}|\Sigma_i|
$$
Thus, it follows from Lemma~\ref{lemma:A5-Q-S-smooth}, that one of the following cases cases holds:
\begin{enumerate}
\item $r=1$ and $|\Sigma_1|=30$;
\item $r=1$, $\lambda=0$, $\Sigma_1=\Sigma_{15}$;
\item $r=2$, $\lambda=-\frac{\zeta_3}{4}$, $|\Sigma_1|=|\Sigma_2|=15$, $\Sigma_1\cup\Sigma_2\subset\mathcal{L}_5\cap X$.
\end{enumerate}

Suppose that $r=1$ and $|\Sigma_1|=30$. If the singularities of $S_\lambda$ at the points of $\Sigma_1$ are not isolated ordinary double points, then it follows from Lemma~\ref{lemma:DuVal} that
$$
10=2g-2=20+\widetilde{C}^2\leqslant 20+C^2-\frac{2}{3}|\Sigma_1|\leqslant 8,
$$
which is absurd. So, we conclude that $S_\lambda$ has isolated ordinary double points at the points of $\Sigma_1$. If~the curve $C$ is singular at these points, then Lemma~\ref{lemma:DuVal} gives
$$
10=2g-2=20+\widetilde{C}^2\leqslant 20+C^2-2|\Sigma_1|=C^2-40<-32,
$$
$$
10=2g-2=20+C^2-m_1E_1\cdot\widetilde{C}=20+C^2-2m_1^2|\Sigma_1|=20+C^2-60m_1^2\leqslant C^2-40<-32.
$$
Thus, $C$ is smooth, and, therefore, $C\simeq\widetilde{C}$ is a curve of degree $10$ and genus $6$ as claimed.

Suppose that $r=1$, $\lambda=0$, $\Sigma_1=\Sigma_{15}$. Recall from Lemma~\ref{lemma:A5-Q-S-smooth} that either $S_\lambda$ has isolated ordinary double points at the points of $\Sigma_{15}$, or $S_\lambda$ has Du Val singularities of type $\mathbb{A}_3$ at the points of $\Sigma_{15}$. In both cases, if $C$ is singular at $\Sigma_1$, then it follows from Lemma~\ref{lemma:DuVal} that
$$
10=2g-2=20+\widetilde{C}^2\leqslant 20+C^2-2|\Sigma_1|=C^2-10\leqslant -2,
$$
which is absurd. Thus, $C\simeq\widetilde{C}$ is a curve of degree $10$ and genus $6$ as claimed.

Finally, we suppose that $r=2$, $\lambda=-\frac{\zeta_3}{4}$, $|\Sigma_1|=|\Sigma_2|=15$, $\Sigma_1\cup\Sigma_2\subset\mathcal{L}_5\cap X$. If $C$ is singular at the points of the $G$-orbit $\Sigma_1$, then, as above, we have 
$$
10=2g-2=20+\widetilde{C}^2\leqslant 20+C^2-2|\Sigma_1|-\frac{1}{2}|\Sigma_1|\leqslant -2,
$$
which is a contradiction. Thus, we see that $C$ is smooth at $\Sigma_1$.
Similarly, we see that $C$ is smooth at the points of the $G$-orbit $\Sigma_2$.
Hence, $C$ is smooth, and $C\simeq\widetilde{C}$ is a curve of genus $6$.
\end{proof}

In the proof of Corollary~\ref{corollary:A5-irreducible-curves}, we used the following result.

\begin{lemma}
\label{lemma:DuVal} Let $Y$ be a normal singular surface, let $P$ be its singular point, let $C\subset Y$ be a curve such that $P\in C$, let $f\colon\widetilde{Y}\to Y$ be the~minimal resolution of the~point~$P$, let $\widetilde{C}$ be the~strict transform on $\widetilde{Y}$ of the~curve $C$. Suppose that $(P\in Y)$ is a Du Val singularity. Then 
\begin{equation}
\label{equation:DuVal}
C^2-\widetilde{C}^2\geqslant \left\{\aligned
&\frac{n}{n+1}\ \text{if $(P\in Y)$ is a singular point of type $\mathbb{A}_n$},\\
&1\ \text{if $(P\in Y)$ is a singular point of type $\mathbb{D}_n$},\\
&\frac{4}{3}\ \text{if $(P\in Y)$ is a singular point of type $\mathbb{E}_6$},\\
&\frac{3}{2}\ \text{if $(P\in Y)$ is a singular point of type $\mathbb{E}_7$},\\
&2\ \text{if $(P\in Y)$ is a singular point of type $\mathbb{E}_8$}.
\endaligned
\right.
\end{equation}
Moreover, if the curve $C$ is singular at $P$, then $C^2-\widetilde{C}^2\geqslant 2$.
\end{lemma}

\begin{proof}
If $(P\in Y)$ is a singular point of type $\mathbb{E}_8$, then $C$ is singular at $P$ by \cite[Lemma~A.4.3]{CheltsovPrzyjalkowski2025}. Moreover, if $C$ is smooth at $P$, then the required assertion follows from \cite[Proposition A.1.3]{CheltsovPrzyjalkowski2025}. Thus, we may assume that the curve $C$ is singular at the point $P$. Hence,  $\mathrm{mult}_P(C)\geqslant 2$. 

Let $E_1,\ldots,E_n$ be $f$-exceptional curves in $\widetilde{Y}$. Then
$$
\widetilde{C}\sim_{\mathbb{Q}}f^*(C)-\sum_{i=1}^{n}a_iE_i,
$$
where each $a_i$ is a positive rational number. Define the function $\Delta\colon\mathbb{R}^n\to\mathbb{R}$ as follows:
$$
\Delta(x_1,\ldots,x_n)=-\Big(\sum_{i=1}^{n}x_iE_i\Big)^2.
$$
Recall from \cite{Artin1962} that $\Delta(x_1,\ldots,x_n)\geqslant 0$ for every $(x_1,\ldots,x_n)\in\mathbb{R}^n$, because the intersection form of the curves $E_1,\ldots,E_n$ is negative definite. On the other hand, we have 
$$
C^2-\widetilde{C}^2=\Delta(a_1,\ldots,a_n).
$$
Thus, we have to show that $\Delta(a_1,\ldots,a_n)\geqslant 2$. 

Suppose that $(P\in Y)$ is a Du Val singular point of type $\mathbb{A}_n$. We may assume that 
$$
E_i\cdot E_j=\left\{\aligned
&0\ \text{if $|i-j|\geqslant 2$},\\
&1\ \text{if $|i-j|=1$},\\
&-2\ \text{if $i=j$}.
\endaligned
\right.
$$
If $n=1$, then $\Delta(a_1)=2a_1^2$ and $2a_1=E_1\cdot \widetilde{C}=\mathrm{mult}_P(C)$, so
$$
\Delta(a_1)=\frac{\mathrm{mult}^2_P(C)}{2}\geqslant 2.
$$
Hence, we may further assume $n\geqslant 2$. Then 
$$
\Delta(a_1,\ldots,a_n)=2\sum_{i=1}^na_i^2-2\sum_{i=1}^{n-1}a_ia_{i+1}=a_1^2+a_n^2+\sum_{i=1}^{n-1}(a_i-a_{i+1})^2,
$$
and it follows from \cite{Artin1962,Artin1966} that 
$$
a_1+a_n=\widetilde{C}\cdot\Big(\sum_{i=1}^{n}E_i\Big)=\mathrm{mult}_P(C)\geqslant 2.
$$
On the other hand, we have 
$$
\Delta(a_1,\ldots,a_n)=\Delta(a_1-1,a_2-1,\ldots,a_n-1)+2(a_1+a_{n}-2)+2\geqslant 2(a_1+a_{n}-2)+2\geqslant 2
$$
as claimed. 

Suppose that $(P\in Y)$ is a Du Val singular point of type $\mathbb{D}_4$. Then $n=4$, and we may assume that the intersection form of the curves $E_1$, $E_2$, $E_3$, $E_4$ is given~in the following table:
\begin{center}\renewcommand\arraystretch{1.42}
\begin{tabular}{|c||c|c|c|c|}
\hline
 $\bullet$  & $E_1$ & $E_2$& $E_3$ & $E_4$\\
\hline\hline
 $E_1$ &  $-2$ & $1$ & $1$ & $1$ \\
\hline
 $E_2$ &  $1$ & $-2$ & $0$ & $0$\\
\hline
 $E_3$ &  $1$ & $0$ & $-2$ & $0$\\
\hline
 $E_4$ &  $1$ & $0$ & $0$ & $-2$\\
\hline
\end{tabular}
\end{center}
Then 
$$
\Delta(a_1,a_2,a_3,a_4)=2a_1^2+2a_2^2+2a_3^2+2a_4^2-2a_1a_2-2a_1a_3-2a_1a_4
$$
Moreover, since $2E_1+E_2+E_3+E_4$ is the fundamental cycle of the singular point $O$, it follows from \cite{Artin1962,Artin1966} that 
$$
a_1=\widetilde{C}\cdot\big(2E_1+E_2+E_3+E_4\big)=\mathrm{mult}_{O}\big(C\big)\geqslant 2.
$$
As above, we see that 
$$
\Delta(a_1,a_2,a_3,a_4)=\Delta(a_1-2,a_2-1,a_3-1,a_4-1)+2(a_{1}-2)+2\geqslant 2(a_{1}-2)+2\geqslant 2.
$$

Suppose that $(P\in Y)$ is a Du Val singular point of type $\mathbb{D}_n$ with $n\geqslant 5$. In this case, we may assume that the intersection form of the curves $E_1,\ldots,E_n$ is given~in the following table:
\begin{center}\renewcommand\arraystretch{1.42}
\begin{tabular}{|c||c|c|c|c|c|c|c|c|}
\hline
 $\bullet$  & $E_1$ & $E_2$& $E_3$ & $E_4$ & $E_5$ & $\ldots$ & $E_{n-1}$ & $E_n$\\
\hline\hline
 $E_1$ &  $-2$ & $1$ & $1$ & $1$ & $0$ & $\ldots$ & $0$ & $0$\\
\hline
 $E_2$ &  $1$ & $-2$ & $0$ & $0$ & $0$ & $\ldots$ & $0$ & $0$\\
\hline
 $E_3$ &  $1$ & $0$ & $-2$ & $0$ & $0$ & $\ldots$ & $0$ & $0$\\
\hline
 $E_4$ &  $1$ & $0$ & $0$ & $-2$ & $1$ & $\ldots$ & $0$ & $0$\\
\hline
 $E_5$ &  $0$ & $0$ & $0$ & $1$ & $-2$ & $\ldots$ & $0$ & $0$ \\
\hline
 $\ldots$ & $\ldots$ & $\ldots$ & $\ldots$ & $\ldots$ & $\ldots$ & $\ddots$ & $\ldots$ & $\ldots$\\
\hline
 $E_{n-1}$ &  $0$ & $0$ & $0$ & $0$ & $0$ & $\ldots$ & $-2$ & $1$\\
\hline
 $E_n$ &  $0$ & $0$ & $0$ & $0$ & $0$ & $\ldots$ & $1$ & $-2$\\
\hline
\end{tabular}
\end{center}
Then 
$$
\Delta(a_1,\ldots,a_n)=2\sum_{i=1}^na_i^2-2a_1(a_2+a_3+a_4)-\sum_{i=4}^{n-1}a_ia_{i+1}.
$$
Since  $2E_1+E_2+E_3+2E_4+\cdots+2E_{n-1}+E_n$ is the fundamental cycle of the singular point $O$,  it follows from \cite{Artin1962,Artin1966} that 
$$
a_{n-1}=\widetilde{C}\cdot\big(2E_1+E_2+E_3+2E_4+\cdots+2E_{n-1}+E_n\big)=\mathrm{mult}_{O}\big(C\big)\geqslant 2.
$$
Then 
$$
\Delta(a_1,\ldots,a_n)=\Delta(a_1-2,a_2-1,a_3-1,a_4-2,\ldots,a_{n-1}-2,a_n-1)+2(a_{n-1}-2)+2\geqslant 2(a_{n-1}-2)+2\geqslant 2.
$$

Suppose that $(P\in Y)$ is a Du Val singular point of type $\mathbb{E}_6$. Then $n=6$, and we may assume that the intersection form of the curves  $E_1$, $E_2$, $E_3$, $E_4$, $E_5$, $E_6$  is given~in the following table:
\begin{center}\renewcommand\arraystretch{1.42}
\begin{tabular}{|c||c|c|c|c|c|c|}
\hline
 $\bullet$  & $E_1$ & $E_2$& $E_3$ & $E_4$ & $E_5$ & $E_6$\\
\hline\hline
 $E_1$ &  $-2$ & $1$ & $0$ & $0$ & $0$ & $0$\\
\hline
 $E_2$ &  $1$ & $-2$ & $1$ & $0$ & $0$ & $0$\\
\hline
 $E_3$ &  $0$ & $1$ & $-2$ & $1$ & $1$ & $0$\\
\hline
 $E_4$ &  $0$ & $0$ & $1$ & $-2$ & $0$ & $0$\\
\hline
 $E_5$ &  $0$ & $0$ & $1$ & $0$ & $-2$ & $1$\\
\hline
 $E_6$ &  $0$ & $0$ & $0$ & $0$ & $1$ & $-2$\\
\hline
\end{tabular}
\end{center}
Then 
$$
\Delta(a_1,a_2,a_3,a_4,a_5,a_6)=2a_1^2+2a_2^2+2a_3^2+2a_4^2+2a_5^2+2a_6^2-2a_1a_2-2a_2a_3-2a_3a_4-2a_3a_5-2a_5a_6.
$$
Since $E_1+2E_2+3E_3+2E_4+2E_5+E_6$ is the fundamental cycle of the singular point $O$, it follows from \cite{Artin1962,Artin1966} that 
$$
a_{4}=\widetilde{C}\cdot\big(E_1+2E_2+3E_3+2E_4+2E_5+E_6\big)=\mathrm{mult}_{O}\big(C\big)\geqslant 2.
$$
Then 
$$
\Delta(a_1,a_2,a_3,a_4,a_5,a_6)=\Delta(a_1-1,a_2-2,a_3-3,a_4-2,a_5-2,a_6-1)+2(a_{4}-2)+2\geqslant 2(a_{4}-2)+2\geqslant 2.
$$

Suppose that $(P\in Y)$ is a Du Val singular point of type $\mathbb{E}_7$. Then $n=7$, and we may assume that the intersection form of the curves $E_1$, $E_2$, $E_3$, $E_4$, $E_5$, $E_6$, $E_7$ is given~in the following table:
\begin{center}\renewcommand\arraystretch{1.42}
\begin{tabular}{|c||c|c|c|c|c|c|c|}
\hline
 $\bullet$  & $E_1$ & $E_2$& $E_3$ & $E_4$ & $E_5$ & $E_6$ & $E_7$\\
\hline\hline
 $E_1$ &  $-2$ & $1$ & $0$ & $0$ & $0$ & $0$ & $0$ \\
\hline
 $E_2$ &  $1$ & $-2$ & $1$ & $0$ & $0$ & $0$& $0$ \\
\hline
 $E_3$ &  $0$ & $1$ & $-2$ & $1$ & $1$ & $0$ & $0$ \\
\hline
 $E_4$ &  $0$ & $0$ & $1$ & $-2$ & $0$ & $0$& $0$ \\
\hline
 $E_5$ &  $0$ & $0$ & $1$ & $0$ & $-2$ & $1$& $0$ \\
\hline
 $E_6$ &  $0$ & $0$ & $0$ & $0$ & $1$ & $-2$ & $1$ \\
\hline
 $E_7$ &  $0$ & $0$ & $0$ & $0$ & $0$ & $1$ & $-2$ \\
\hline
\end{tabular}
\end{center}
Then 
$$
\Delta(a_1,a_2,a_3,a_4,a_5,a_6,a_7)=2(a_1^2+a_2^2+a_3^2+a_4^2+a_5^2+a_6^2+a_7^2-a_1a_2-a_2a_3-a_3a_4-a_3a_5-a_5a_6-a_6a_7).
$$
Since $2E_1+3E_2+4E_3+2E_4+3E_5+2E_6+E_7$ is the fundamental cycle of  $(Y\ni O)$, we get 
$$
a_{1}=\widetilde{C}\cdot\big(2E_1+3E_2+4E_3+2E_4+3E_5+2E_6+E_7\big)=\mathrm{mult}_{O}\big(C\big)\geqslant 2.
$$
As above, we have 
$$
\Delta(a_1,a_2,a_3,a_4,a_5,a_6,a_7)=\Delta(a_1-2,a_2-3,a_3-4,a_4-2,a_5-3,a_6-2,a_7-1)+2(a_{1}-2)+2,
$$
which gives $\Delta(a_1,a_2,a_3,a_4,a_5,a_6,a_7)\geqslant 2(a_{1}-2)+2\geqslant 2$.

Suppose that $(P\in Y)$ is a singular point of type $\mathbb{E}_8$. Then $n=8$, and we may assume that the~intersection form of the curves $E_1$, $E_2$, $E_3$, $E_4$, $E_5$, $E_6$, $E_7$, $E_8$ is given~in the following table:
\begin{center}\renewcommand\arraystretch{1.42}
\begin{tabular}{|c||c|c|c|c|c|c|c|c|}
\hline
 $\bullet$  & $E_1$ & $E_2$& $E_3$ & $E_4$ & $E_5$ & $E_6$ & $E_7$ & $E_8$\\
\hline\hline
 $E_1$ &  $-2$ & $1$ & $0$ & $0$ & $0$ & $0$ & $0$ & $0$\\
\hline
 $E_2$ &  $1$ & $-2$ & $1$ & $0$ & $0$ & $0$& $0$ & $0$\\
\hline
 $E_3$ &  $0$ & $1$ & $-2$ & $1$ & $1$ & $0$ & $0$ & $0$\\
\hline
 $E_4$ &  $0$ & $0$ & $1$ & $-2$ & $0$ & $0$& $0$ & $0$\\
\hline
 $E_5$ &  $0$ & $0$ & $1$ & $0$ & $-2$ & $1$& $0$ & $0$\\
\hline
 $E_6$ &  $0$ & $0$ & $0$ & $0$ & $1$ & $-2$ & $1$ & $0$\\
\hline
 $E_7$ &  $0$ & $0$ & $0$ & $0$ & $0$ & $1$ & $-2$ & $1$\\
\hline
 $E_8$ &  $0$ & $0$ & $0$ & $0$ & $0$ & $0$ & $1$ & $-2$\\
\hline
\end{tabular}
\end{center}
Then 
\begin{multline*}
\Delta(a_1,a_2,a_3,a_4,a_5,a_6,a_7,a_8)=2a_1^2+2a_2^2+2a_3^2+2a_4^2+2a_5^2+2a_6^2+2a_7^2+2a_8^2-\\
-2a_1a_2-2a_2a_3-2a_3a_4-2a_3a_5-2a_5a_6-2a_6a_7-2a_7a_8.\quad \quad 
\end{multline*}
Since $2E_1+4E_2+6E_3+3E_4+5E_5+4E_6+3E_7+2E_8$ is the fundamental cycle of the singular point $O$, it follows from \cite{Artin1962,Artin1966} that 
$$
a_{8}=\widetilde{C}\cdot\big(2E_1+4E_2+6E_3+3E_4+5E_5+4E_6+3E_7+2E_8\big)=\mathrm{mult}_{O}\big(C\big)\geqslant 2.
$$
Since $\Delta$ is a non-negative function, we have 
$$
\Delta(a_1,a_2,a_3,a_4,a_5,a_6,a_7,a_8)=\Delta(a_1-2,a_2-4,a_3-6,a_4-3,a_5-5,a_6-4,a_7-3,a_8-2)+2(a_{8}-2)+2,
$$
which implies that $\Delta(a_1,a_2,a_3,a_4,a_5,a_6,a_7,a_8)\geqslant 2(a_{8}-2)+2\geqslant 2$ as claimed.
\end{proof}

To proceed, we need the following lemma.

\begin{lemma}
\label{lemma:A5-3-regular}
Let $C$ be a smooth irreducible $G$-invariant curve of degree $10$ and genus $6$ in $\mathbb{P}^4$, and let  $\mathcal{I}_C$ be the ideal sheaf of the curve $C$. Suppose that $C\subset X$ and $C\not\subset Q$. Then 
\begin{equation}
\label{equation:3-regular}
H^i\big(\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(3-i)\big)=0
\end{equation}
for every $i>0$, i.e., the curve $C$ is $3$-regular \cite{Mumford}.
\end{lemma}

\begin{proof}
We claim that $C$ is not contained in any surface in $|-K_X|$. Indeed, let $\mathcal{M}$ be the linear subsystem in $|-K_X|$ consisting of surfaces that pass through $C$. Suppose that $\mathcal{M}$ is not empty. Then $\mathcal{M}$ does not have fixed components, since $C\not\subset Q$. Moreover, the base locus of the linear system $\mathcal{M}$ does not contain curves except $C$. Indeed, if $C^\prime$ is another $G$-irreducible curve contained in the base locus of the linear system $\mathcal{M}$, then 
$$
M\cdot M^\prime=mC+m^\prime C^\prime+\Delta,
$$
for general surfaces $M$ and $M^\prime$ in $\mathcal{M}$, an effective one-cycle $\Delta$, and $m,m^\prime\in\mathbb{Z}_{>0}$, which gives 
$$
\mathrm{deg}(C^\prime)\leqslant\mathrm{deg}\big(M\cdot M^\prime\big)-\mathrm{deg}(C)=12-\mathrm{deg}(C)=2,
$$
which is impossible by Lemmas~\ref{lemma:A5-reducible-curves}, \ref{lemma:A5-irreducible-curves-in-Q}, \ref{lemma:A5-irreducible-curves-in-S} and Corollary~\ref{corollary:A5-irreducible-curves}. Thus, $C$ is the only curve in the base locus of the linear system $\mathcal{M}$. Furthermore, the same arguments imply that two general surfaces in the linear system $\mathcal{M}$ are smooth at general point of the curve $C$, and they are not tangent to each other along $C$. Now, we let $\pi\colon\widetilde{X}\to X$ be the blow up of the curve $C$, let $E$ be the exceptional surface, and let $\widetilde{\mathcal{M}}$ be the strict transform of the linear system $\mathcal{M}$ on the threefold $\widetilde{X}$. Then
$$
\widetilde{\mathcal{M}}\sim\pi^*(-K_X)-E,
$$ 
and $\widetilde{\mathcal{M}}$ does not have base curves except possibly fibers of the~projection $E\to C$, which implies that the divisor $\pi^*(-K_X)-E$ is nef. But $\pi^*(-K_X)-E$  cannot be nef, since 
$$
\big(\pi^*(-K_X)-E\big)^3=-6.
$$
Hence, $C$ is not contained in any surface in $|-K_X|$, which implies that $C$ is not contained in any quadric hypersurface in $\mathbb{P}^4$. Let us use this to prove \eqref{equation:3-regular} for $i=1$.

Recall that $C$ is not contain in any hyperplane in $\mathbb{P}^4$, because the representation $\mathbb{V}_5$ is irreducible.
Let $H$ be a hyperplane section of the curve $C$. Consider the following exact sequence: 
\begin{multline*}
\quad \quad \quad 0=H^0\big(\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(2)\big)\longrightarrow H^0\big(\mathcal{O}_{\mathbb{P}^4}(2)\big)\longrightarrow H^0\big(C,\mathcal{O}_{C}(2H)\big)\longrightarrow \\
\longrightarrow H^1\big(\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(2)\big)\longrightarrow H^1\big(\mathcal{O}_{\mathbb{P}^4}(2)\big)=0.\quad \quad \quad \quad 
\end{multline*}
Now, using the~Riemann--Roch formula and Serre duality, we get 
$$
h^1\big(\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(2)\big)=h^0\big(C,\mathcal{O}_{C}(2H)\vert_{C}\big)-15=h^0\big(C,\mathcal{O}_{C}(K_C-2H)\big)=0,
$$
because $\mathrm{deg}(K_C-2H)=-10$. This gives \eqref{equation:3-regular} for $i=1$.

Similarly, to prove \eqref{equation:3-regular} for $i=2$, we consider the following exact sequence: 
$$
0=H^1\big(\mathcal{O}_{\mathbb{P}^4}(1)\big)\longrightarrow H^1\big(C,\mathcal{O}_{C}(H)\big)\longrightarrow H^2\big(\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(1)\big)\longrightarrow H^2\big(\mathcal{O}_{\mathbb{P}^4}(1)\big)=0.
$$
Thus, it follows from the~Serre duality that
$$
h^2\big(\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(1)\big)=h^1\big(C,\mathcal{O}_{C}(H)\big)=h^0\big(C,\mathcal{O}_{C}(K_C-H)\big)=\left\{\aligned
&1\ \text{if $K_C\sim H$},\\
&0\ \text{if $K_C\not\sim H$}.
\endaligned
\right.
$$
On the other hand, if $K_C\sim H$, then the embedding $C\hookrightarrow\mathbb{P}^4$ is factored through the $G$-equivariant canonical embedding $\phi\colon C\hookrightarrow\mathbb{P}^5$ followed by the linear projection $\mathbb{P}^5\dasharrow \mathbb{P}^4$ from a $G$-fixed point. However, if $\mathbb{P}^5$ contains a $G$-fixed point, then $\mathbb{P}^5$ also contain a $G$-invariant hyperplane, which must intersect $C$ by $\leqslant 10$ points, which contradicts Lemma~\ref{lemma:A5-curves}. So, $K_C\not\sim H$, which gives \eqref{equation:3-regular} for $i=2$. 

For $i=3$, \eqref{equation:3-regular} follows from the exact sequence
$$
0=H^2\big(C,\mathcal{O}_{C}(H)\big)\longrightarrow H^3\big(\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(1)\big)\longrightarrow H^3\big(\mathcal{O}_{\mathbb{P}^4}(1)\big)=0.
$$
For $i=4$, \eqref{equation:3-regular} follows from the exact sequence
$$
0=H^3\big(C,\mathcal{O}_{C}(H)\big)\longrightarrow H^4\big(\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(1)\big)\longrightarrow H^4\big(\mathcal{O}_{\mathbb{P}^4}(1)\big)=0.
$$
This shows that the curve $C$ is $3$-regular.
\end{proof}

\begin{corollary}
\label{corollary:A5-3-regular}
In the assumptions of Lemma~\ref{lemma:A5-3-regular}, $\mathcal{I}_C\otimes\mathcal{O}_{\mathbb{P}^4}(3)$ is generated by global sections.
\end{corollary}

\begin{proof}
The required assertion follows from Lemma~\ref{lemma:A5-3-regular} and Proposition in  \cite[Lecture 14]{Mumford}.
\end{proof}

Let $H$ be a general hyperplane section of the cubic threefold $X$.

\begin{proposition}
\label{proposition:A5-NFI-curves}
Let $Z$ be a curve in $X$ such that one of the following cases holds:
\begin{enumerate}
\item[$(\mathrm{i})$] $Z=\mathcal{L}_6$ or $Z=\mathcal{L}_6^\prime$;
\item[$(\mathrm{ii})$] $Z=\mathcal{L}^t_{10}$ for some $t\in\mathbb{C}$ such that $(b+3)t^3+3b\zeta_3t^2+1=0$;
\item[$(\mathrm{iii})$] $Z$ is a smooth rational curve of degree $8$, and $Z\subset Q$;
\item[$(\mathrm{iv})$] $Z$ is a smooth curve of degree $10$ and genus $6$.
\end{enumerate}
Let $\mathcal{M}$ be a non-empty $G$-invariant mobile linear subsystem in $|nH|$, where $n\in\mathbb{Z}_{>0}$. Then
$$
\mathrm{mult}_Z\big(\mathcal{M}\big)\leqslant\frac{n}{2}.
$$
\end{proposition}

\begin{proof}
Let $\pi\colon \widetilde{X}\to X$ be the~blow up~of~$Z$, let $F$ be the $\pi$-exceptional divisor, 
let $\widetilde{\mathcal{M}}$ be the strict transform on $\widetilde{X}$ of the linear system $\mathcal{M}$,
let~$\widetilde{M}$ and $\widetilde{M}^\prime$ be general surfaces in $\widetilde{\mathcal{M}}$, let  $\widetilde{H}=\pi^*(H)$. Then
$$
\widetilde{M}\sim \widetilde{M}^\prime\sim n\widetilde{H}-\mathrm{mult}_Z\big(\mathcal{M}\big)F.
$$
Note that $\widetilde{M}\cdot\widetilde{M}^\prime$ is an effective one-cycle.
Take $k\in\mathbb{Z}_{>0}$ such that the divisor $k\widetilde{H}-F$ is nef. Then
\begin{equation}
\label{equation:A5-test-class}
0\leqslant \widetilde{M}\cdot \widetilde{M}^\prime\cdot \big(k\widetilde{H}-F\big)=3kn^2-2n\mathrm{deg}(Z)\mathrm{mult}_Z\big(\mathcal{M}\big)-(k\mathrm{deg}(Z)+F^3)\mathrm{mult}^2_Z\big(\mathcal{M}\big).
\end{equation}
Note that $Z$ is smooth unless $Z=\mathcal{L}^t_{10}$ and $t=\pm\frac{\zeta_3+2}{3}$. Moreover, we have
$$
F^3=\left\{\aligned
&0\ \text{if $Z=\mathcal{L}_6$ or $Z=\mathcal{L}_6^\prime$},\\
&0\ \text{if $Z=\mathcal{L}^t_{10}$ for $t\in\mathbb{C}$ such that $(b+3)t^3+3b\zeta_3t^2+1=0$ and $t\ne\pm\frac{\zeta_3+2}{3}$},\\
&-30\ \text{if $Z=\mathcal{L}^t_{10}$ for $t=\pm\frac{\zeta_3+2}{3}$},\\
&-14\ \text{if $Z$ is a smooth rational curve of degree $8$},\\
&-30\ \text{if $Z$ is a smooth curve of degree $10$ and genus $9$}.
\endaligned
\right.
$$

If $Z=\mathcal{L}_6$ or $Z=\mathcal{L}_6^\prime$, then $3\widetilde{H}-F$ is nef by Lemma~\ref{lemma:A5-L6}, and we can let $k=3$ in \eqref{equation:A5-test-class} to get 
$$
0\leqslant \widetilde{M}\cdot \widetilde{M}^\prime\cdot \big(3\widetilde{H}-F\big)=
9n^2-12n\mathrm{mult}_Z\big(\mathcal{M}\big)-18\mathrm{mult}^2_Z\big(\mathcal{M}\big),
$$
which gives $\mathrm{mult}_Z(\mathcal{M})\leqslant\frac{n}{2}$. 

If $Z=\mathcal{L}^t_{10}$ for  $t\in\mathbb{C}$ such that $(b+3)t^3+3b\zeta_3t^2+1=0$ and $t\ne\pm\frac{\zeta_3+2}{3}$, then $Z$ is a disjoint union of $10$ lines, so the divisor $10\widetilde{H}-F$ is nef, and, therefore, we can let $k=10$ in \eqref{equation:A5-test-class} to get 
$$
0\leqslant \widetilde{M}\cdot \widetilde{M}^\prime\cdot \big(10\widetilde{H}-F\big)=
30n^2-20n\mathrm{mult}_Z\big(\mathcal{M}\big)-100\mathrm{mult}^2_Z\big(\mathcal{M}\big),
$$
which gives $\mathrm{mult}_Z(\mathcal{M})\leqslant\frac{n}{2}$. 

If $Z=\mathcal{L}^t_{10}$ for  $t=\pm\frac{\zeta_3+2}{3}$, then $3\widetilde{H}-F$ is nef by Lemma~\ref{lemma:A5-L10}. If $Z$ is a curve of degree $10$ and genus $9$ such that $C\not\subset Q$, then $3\widetilde{H}-F$ is nef by Corollary~\ref{corollary:A5-3-regular}. In both cases,  \eqref{equation:A5-test-class} gives
$$
0\leqslant \widetilde{M}\cdot \widetilde{M}^\prime\cdot \big(3\widetilde{H}-F\big)=
9n^2-20n\mathrm{mult}_Z\big(\mathcal{M}\big),
$$
which implies $\mathrm{mult}_Z(\mathcal{M})\leqslant\frac{n}{2}$. 

Hence, we may further assume that $Z\subset Q$, and either $Z$ is a smooth rational curve of degree~$8$, or~$Z$ is a smooth curve of degree $10$ and genus $9$. Let us show that $\mathrm{mult}_Z(\mathcal{M})\leqslant\frac{n}{2}$.

Set $H_{Q}=H\vert_{Q}$. We claim that $|3H_Q-Z|$ does not have fixed curves. Indeed, write
$$
|3H_Q-Z|=\mathscr{F}+|\mathscr{M}|,
$$
where $\mathscr{F}$ is the~fixed part of this linear system, and $\mathscr{M}$ is a  divisor  such that $|\mathscr{M}|$ is the mobile part of the linear system $|3H_Q-Z-E|$. We may assume that $\mathscr{M}$ is a general surface in $|\mathscr{M}|$. Then
$$
H_Q\cdot \mathscr{M}=\mathrm{deg}\big(\mathscr{M}\big)\geqslant 3,
$$
because $Q$ is a smooth K3 surface, so it is not uniruled. On the other hand, $\mathscr{F}$ is $G$-invariant, and
$$
18-\mathrm{deg}(Z)=H_Q\cdot\big(3H_Q-Z\big)=H_Q\cdot\big(\mathscr{F}+\mathscr{M}\big)\geqslant H_Q\cdot\mathscr{F}+3=\mathrm{deg}\big(\mathscr{F}\big)+3.
$$
Thus, $\mathrm{deg}(\mathscr{F})\leqslant 7$, which gives $\mathscr{F}=0$ by Lemmas~\ref{lemma:A5-reducible-curves}, \ref{lemma:A5-irreducible-curves-in-Q}, \ref{lemma:A5-irreducible-curves-in-S} and Corollary~\ref{corollary:A5-irreducible-curves}, because $Q$ contains neither $\mathcal{L}_{6}$ nor $\mathcal{L}_{6}^\prime$ by Lemma~\ref{lemma:A5-Q-S-orbits}. Hence, $|3H_Q-Z|$ does not have fixed curves. 

In particular, the divisor $3H_Q-Z$ is nef.  Let $M$ be a general surface in $\mathcal{M}$. Then
$$
M\big\vert_{Q}=mZ+D
$$
for an integer $m\geqslant \mathrm{mult}_Z(\mathcal{M})$ and an
effective divisor $D$ on the~surface $Q$. If $Z$ is a rational curve of degree $8$, then $H_Q\cdot Z=8$ and $Z^2=-2$, so 
$$
0\leqslant \big(3H_Q-Z\big)\cdot D=\big(3H_Q-Z\big)\cdot\big(nH_Q-mZ\big)=10n-26m,
$$
which gives $\mathrm{mult}_Z(\mathcal{M})\leqslant m\leqslant\frac{n}{2}$. If $Z$ is a curve of
degree $10$, then $H_Q\cdot Z=10$ and $Z^2=10$, so 
$$
0\leqslant \big(3H_Q-Z\big)\cdot D=\big(3H_Q-Z\big)\cdot\big(nH_Q-mZ\big)=8n-20m,
$$
which gives $\mathrm{mult}_Z(\mathcal{M})\leqslant m\leqslant\frac{n}{2}$. 
\end{proof}

Now, we are ready to prove the following result, which implies Theorem~\ref{theorem:A5}. 

\begin{proposition}
\label{proposition:A5-non-standard}
Suppose that $G=\langle M_2,M_3\rangle$. Then $X$ is $G$-birationally superrigid.
\end{proposition}

Suppose that $X$ is not $G$-birationally superrigid. Let us seek for a contradiction.

By \cite[Corollary 3.3.3]{CheltsovShramov2016}, there is a non-empty $G$-invariant mobile linear subsystem
$\mathcal{M}\subset |nH|$, for a positive integer~$n$, such that the singularities of the log pair $(X,\frac{2}{n}\mathcal{M})$ are not canonical.
Let 
$$
\Sigma=\Big\{P\in X\ \text{such that the~log pair $\Big(X,\frac{2}{n}\mathcal{M}\Big)$ is not canonical at~$P$}\Big\}. 
$$
Then $\Sigma$ is a union of $G$-orbits and $G$-irreducible curves, since $\mathcal{M}$ is mobile.

\begin{lemma}
\label{lemma:A5-Sigma-not-finite}
The subset $\Sigma$ contains a $G$-irreducible curve.
\end{lemma}

\begin{proof}
Suppose that $\Sigma$ does not contain $G$-irreducible curves. Then $\Sigma$ is a finite subset, so it follows from \cite[Remark 3.6]{CheltsovSarikyanZhuang2024} that the~singularities of the~log pair $(X,\frac{3}{n}\mathcal{M})$ are not log canonical at every point of $\Sigma$. Hence, there is $\mu\in\mathbb{Q}_{>0}$ such that $\mu\leqslant \frac{3}{n}$ and  $(X,\mu\mathcal{M})$ is strictly log canonical. 

Let $C$ be an irreducible subvariety in $X$ that is a minimal center of log canonical singularities of the~log pair $(X,\mu\mathcal{M})$, and let $Z$ be a $G$-irreducible subvariety in $X$ that has $C$ as an irreducible component. Then all component of $Z$ are minimal centers of log canonical singularities of  $(X,\mu\mathcal{M})$, which implies that these irreducible components are disjoint \cite[Proposition 1.5]{Kawamata-1}. Note that 
$$
\mathrm{dim}(Z)\ne 2,
$$
because $\mathcal{M}$ is mobile. Hence, either $Z$ is a $G$-irreducible curve or $Z$ is the~$G$-orbit of a point. 

Now, arguing as in the proof of Theorem~\ref{theorem:405-15}, we can replace the~rational number $\mu$ and the~mobile linear system $\mathcal{M}$ by a rational number $\lambda<3$ and an effective $G$-invariant $\mathbb{Q}$-divisor $D\sim_{\mathbb{Q}} H$
such that the~singularities of the~log pair $(X,\lambda D)$ are strictly log canonical, $C$ is a minimal center of the~non-Kawamata log terminal singularities of this log pair, and
$$
\mathrm{Nklt}\big(X,\lambda D\big)=Z,
$$
where $\mathrm{Nklt}(X,\lambda D)=\big\{P\in X\ \text{such that $(X,\lambda D)$ is not Kawamata log terminal at $P$}\big\}$.

Let $\mathcal{I}_Z$ be the~ideal sheaf of the~subvariety $Z\subset X$. Then $\mathcal{I}_Z$ is the~multiplier ideal sheaf of the~log pair $(X,\lambda D)$, and it follows from the Nadel vanishing theorem \cite[Theorem 9.4.8]{Lazarsfeld2004} that 
$$
h^1\big(X,\mathcal{I}_Z\otimes\mathcal{O}_{X}(H)\big)=0.
$$
Hence, we have the~following exact sequence of $G$-representations:
$$
0\longrightarrow H^0\big(X,\mathcal{I}_Z\otimes\mathcal{O}_{X}(H)\big)\longrightarrow H^0\big(X,\mathcal{O}_{X}(H)\big)\longrightarrow H^0\big(Z,\mathcal{O}_{Z}(H\vert_{Z})\big)\longrightarrow 0.
$$
On the other hand, we have 
$$
H^0\big(X,\mathcal{I}_Z\otimes\mathcal{O}_{X}(H)\big)=0,
$$
because the representation $\mathbb{V}_5$ is irreducible. Thus, we get 
\begin{equation}
\label{equation:A5-final}
h^0\big(Z,\mathcal{O}_{Z}(H\vert_{Z})\big)=5.
\end{equation}
Thus, if $Z$ is a $G$-orbit of a point, then $|Z|=5$, which is impossible by Lemma~\ref{lemma:A5-orbits}.

Thus, we conclude that $Z$ is a curve. Then it follows from \cite[Lemma 1.8]{Corti2000} that $\mathrm{deg}(Z)\leqslant 6$. Indeed, let $M$ and $M^\prime$ be general surfaces in $\mathcal{M}$. Then it follows from \cite[Lemma 1.8]{Corti2000} that
$$
\big(M\cdot M^\prime\big)_Z\geqslant\frac{4}{\mu^2}>\frac{4n^2}{9}
$$
which gives $\mathrm{deg}(Z)\leqslant 6$. Then  $Z=\mathcal{L}_6$ or $Z=\mathcal{L}_6^\prime$ by  
Lemmas~\ref{lemma:A5-reducible-curves}, \ref{lemma:A5-irreducible-curves-in-Q}, \ref{lemma:A5-irreducible-curves-in-S} and Corollary~\ref{corollary:A5-irreducible-curves}. Then 
$$
h^0\Big(Z,\mathcal{O}_{Z}\big(H\vert_{Z}\big)\Big)=12,
$$
because $Z$ is a dijoint union of $6$ lines. This contradicts to \eqref{equation:A5-final}. 
\end{proof}

Thus, $\Sigma$ contains a $G$-irreducible curve $\mathscr{C}$. Then, see \cite[Excercise 6.18]{KollarSmithCorti}, we have 
\begin{equation}
\label{equation:A5-mult}
\mathrm{mult}_{\mathscr{C}}\big(\mathcal{M}\big)>\frac{n}{2}.
\end{equation}
This gives $\mathrm{deg}(\mathscr{C})<12$. Indeed, let $M$ and $M^\prime$ be general surfaces in $\mathcal{M}$. Then
$$
M\cdot M^\prime=m\mathscr{C}+\Delta,
$$
where $\Delta$ is an effective one-cycle whose support does not contain $\mathscr{C}$, and $m$ is a positive integer such that
$m\geqslant \mathrm{mult}_\mathscr{C}^2(\mathcal{M})>\frac{n^2}{4}$. Thus, we have
$$
3n^2=H\cdot M\cdot M^\prime=m\mathrm{deg}(Z)+H\cdot\Delta\geqslant\mathrm{mult}^2_\mathscr{C}\big(\mathcal{M}\big)\mathrm{deg}\big(\mathscr{C}\big)>\frac{n^2}{4}\mathrm{deg}(\mathscr{C}),
$$
so $\mathrm{deg}(Z)<12$. Then, by Lemmas~\ref{lemma:A5-reducible-curves}, \ref{lemma:A5-irreducible-curves-in-Q}, \ref{lemma:A5-irreducible-curves-in-S}, \ref{lemma:A5-irreducible-curves-degree-10}, Corollary~\ref{corollary:A5-irreducible-curves} and Proposition~\ref{proposition:A5-NFI-curves}, we see that 
\begin{itemize}
\item $\mathscr{C}$ is an irreducible singular curve of degree $10$, 
\item $\mathscr{C}$ does not contain $G$-orbits of length $12$ or $15$,
\item $\mathscr{C}$ is contained in $S_\lambda$ for some $\lambda\in\mathbb{C}$, and $S_\lambda$ is not normal.
\end{itemize}

\begin{remark}
\label{remark:A5-simplified-proof}
By Lemma~\ref{lemma:A5-Q-S-smooth}, $S_\lambda$ is normal if $b\in\mathbb{C}$ is general. Hence, Proposition~\ref{proposition:A5-non-standard} is proved for general $b\in\mathbb{C}$. As we mentioned in Remark~\ref{remark:A5-quartics-normal}, we think that $S_\lambda$ is normal for every $b\in\mathbb{C}$. Unfortunately, we were unable to show this using the tools we have.
\end{remark}

It follows from \eqref{equation:A5-mult} that the~singularities of the mobile log pair $(X,\frac{4}{n}\mathcal{M})$ are not log canonical along the curve $\mathscr{C}$. Let 
$$
\mu=\mathrm{lct}_{\mathscr{C}}\big(X,\mathcal{M}\big)=\mathrm{sup}\big\{\lambda\in\mathbb{Q}_{>0}\ \big\vert\ \text{$(X,\lambda\mathcal{M})$ is log canonical at every point of the curve $\mathscr{C}$}\big\}.
$$
Then $\mu\leqslant \frac{4}{n}$, and $\mathrm{Nklt}(X,\mu\mathcal{M})$ contains an irreducible subvariety $C$ such that $C\cap\mathscr{C}\ne \varnothing$, where 
$$
\mathrm{Nklt}\big(X,\mu\mathcal{M}\big)=\big\{P\in X\ \text{such that $(X,\mu\mathcal{M})$ is not Kawamata log terminat at $P$}\big\}.
$$
Moreover, replacing $C$ if necessarily, may further assume that $C$ has minimal dimension among all subvarieties in $\mathrm{Nklt}(X,\mu\mathcal{M})$ with this property.

Let $Z$ be the $G$-irreducible subvariety in the hypersurface $X$ whose irreducible component is~$C$. Then we have the following three possibilities:
\begin{enumerate}
\item $Z$ is a $G$-irreducible curve such that $C\ne \mathscr{C}$ and $Z\cap\mathscr{C}\ne\varnothing$; 
\item $Z$ is a $G$-orbit of a point in $\mathscr{C}$;
\item $Z=\mathscr{C}$.
\end{enumerate}
Note that the locus $\mathrm{Nklt}(X,\mu\mathcal{M})$ is $G$-invariant by construction.

\begin{lemma}
\label{lemma:A5-end-Corti}
The locus $\mathrm{Nklt}(X,\mu\mathcal{M})$ does not contain curves different from $\mathscr{C}$.
\end{lemma}

\begin{proof}
Suppose that the locus $\mathrm{Nklt}(X,\mu\mathcal{M})$ contains a $G$-irreducible curve $\mathscr{Z}$ such that $\mathscr{Z}\ne \mathscr{C}$. Let $M$ and $M^\prime$ be general surfaces in $\mathcal{M}$. 
Write 
$$
M\cdot M^\prime=m_\mathscr{Z}\mathscr{Z}+m_\mathscr{C}\mathscr{C}+\mathscr{D},
$$
where $m_\mathscr{Z}$ and $m_\mathscr{C}$ are non-negative integers, and $\mathscr{D}$ is an effective one-cycle on $X$ whose support does not contain curves $\mathscr{Z}$ and $\mathscr{C}$. Recall from \eqref{equation:A5-mult} that
$$
m_\mathscr{C}=\big(M\cdot M^\prime\big)_\mathscr{C}\geqslant\mathrm{mult}_{\mathscr{C}}\big(M\big)\mathrm{mult}_{\mathscr{C}}\big(M^\prime\big)=\mathrm{mult}_{\mathscr{C}}^2\big(\mathcal{M}\big)>\frac{n^2}{4}.
$$
Moreover, it follows from \cite[Lemma 1.8]{Corti2000} that
$$
m_\mathscr{Z}=\big(M\cdot M^\prime\big)_\mathscr{Z}\geqslant\frac{4}{\mu^2}>\frac{n^2}{4}.
$$
Thus, we have
$$
3n^2=H\cdot M\cdot M^\prime=m_\mathscr{Z}\mathrm{deg}(\mathscr{Z})+m_\mathscr{C}\mathrm{deg}\big(\mathscr{C}\big)+H\cdot \mathscr{D}>\frac{n^2}{4}\Big(\mathrm{deg}(\mathscr{Z})+\mathrm{deg}\big(\mathscr{C}\big)\Big),
$$
so $\mathrm{deg}(\mathscr{Z})<12-\mathrm{deg}(\mathscr{C})=2$, which is impossible by
Lemmas~\ref{lemma:A5-reducible-curves}, \ref{lemma:A5-irreducible-curves-in-Q}, \ref{lemma:A5-irreducible-curves-in-S}, \ref{lemma:A5-irreducible-curves-degree-10} and Corollary~\ref{corollary:A5-irreducible-curves}.
\end{proof}

Hence, $\mathrm{Nklt}(X,\mu\mathcal{M})\setminus\mathscr{C}$ is a finite (possibly empty) set. This, we see that
\begin{itemize}
\item either $Z$ is a $G$-orbit of a point in $\mathscr{C}$. 
\item or $Z=\mathscr{C}$.
\end{itemize}
Moreover, since $(X,\mu\mathcal{M})$ is log canonical along $\mathscr{C}$, it follows from \cite{Kawamata-2} that $Z$ is smooth, which gives $Z\ne\mathscr{C}$, because the curve $\mathscr{C}$ is singular. Thus, $Z$ is a $G$-orbit of a point in $\mathscr{C}$.

Now, arguing as in the proofs of \cite[Lemma 2.4.10]{CheltsovShramov2016} or \cite[Lemma A.28]{CalabiBook}, we can replace $\mu$ and  $\mathcal{M}$ by a positive rational number $\lambda<2$ and an effective $G$-invariant $\mathbb{Q}$-divisor
$$
D\sim_{\mathbb{Q}}-K_{X}
$$
such that the log pair $(X,\lambda D)$ has strictly log canonical singularities at every point of the orbit~$Z$, the locus $\mathrm{Nklt}\big(X,\lambda D\big)$ consists of finitely many points, and
$$
Z\subset\mathrm{Nklt}\big(X,\lambda D\big),
$$
 where $\mathrm{Nklt}(X,\lambda D)=\big\{P\in X\ \text{such that $(X,\lambda D)$ is not Kawamata log terminal at $P$}\big\}$.

Let $\mathcal{I}\subset\mathcal{O}_X$ be the~multiplier ideal sheaf of the~pair $(X,\lambda D)$, and let $\mathcal{L}$ be the zero-dimensional subscheme in $X$ that is given by $\mathcal{I}$. Then 
$$
Z\subset\mathrm{Supp}\big(\mathcal{L}\big)=\mathrm{Nklt}\big(X,\lambda D\big).
$$
Moreover, it follows from the Nadel vanishing theorem \cite[Theorem 9.4.8]{Lazarsfeld2004} that
$$
h^1\big(X,\mathcal{I}\otimes \mathcal{O}_{X}(2)\big)=0.
$$
Hence, we have the~following exact sequence of $G$-representations:
$$
0\longrightarrow H^0\big(X,\mathcal{I}\otimes \mathcal{O}_{X}(2)\big)\longrightarrow H^0\big(X,\mathcal{O}_{X}(2)\big)\longrightarrow H^0\big(\mathcal{L},\mathcal{O}_{\mathcal{L}}\big)\longrightarrow 0,
$$
because $\mathcal{L}$ is zero-dimensional. This gives 
$$
|Z|\leqslant h^0\big(\mathcal{L},\mathcal{O}_{\mathcal{L}}\big)=h^0\big(X,\mathcal{O}_{X}(2)\big)-h^0\big(X,\mathcal{O}_{X}(2)\otimes\mathcal{I}\big)=15-h^0\big(X,\mathcal{O}_{X}(2)\otimes\mathcal{I}\big)\leqslant 15,
$$
Hence, $|Z|\in\{12,15\}$ by Lemma~\ref{lemma:A5-orbits}. But $Z\subset\mathscr{C}$, and $\mathscr{C}$ contains no $G$-orbits of length $12$ or~$15$. The obtained contradiction completes the proof of Proposition~\ref{proposition:A5-non-standard}. 

By Propositions~\ref{proposition:A5-standard} and \ref{proposition:A5-non-standard},  Theorem~\ref{theorem:A5} is proved.

\section{Groups $C_3^2\rtimes C_4$ and $\mathfrak{S}_3\times\mathfrak{S}_3$}
\label{section:36-9-36-10}

Let $X$ be a smooth cubic threefold in $\mathbb{P}^4$. Suppose that $\mathrm{Aut}(X)$ has a subgroup $G$ such that
\begin{itemize}
\item either $G\simeq C_3^2\rtimes C_4$ with GAP ID [36,9],
\item or $G\simeq\mathfrak{S}_3\times\mathfrak{S}_3$ with GAP ID [36,10].
\end{itemize}
Then the~$G$-action on $X$ lifts to $\mathbb{P}^4$, so we can consider $G$ as a subgroup in $\mathrm{PGL}_5(\mathbb{C})$.
We suppose, in~addition, that $\mathbb{P}^4$ contains no $G$-invariant planes. In this section, we prove the following result:

\begin{theorem}
\label{theorem:36-9-smooth}
The cubic threefold $X$ is $G$-birationally superrigid.
\end{theorem}

First, we state few auxiliary results about the groups $C_3^2\rtimes C_4$ and $\mathfrak{S}_3\times\mathfrak{S}_3$.

\begin{lemma}
\label{lemma:36-9-subgroups}
Conjugacy classes of proper subgroups of $C_3^2\rtimes C_4$ are described as follows:
\begin{center}\renewcommand{\arraystretch}{1.5}
\begin{tabular}{|c||c|c|c|c|c|c|c|c|}
\hline
$\mathcal{C}$&$\mathcal{C}_2$&$\mathcal{C}_3$&$\mathcal{C}_3^\prime$&$\mathcal{C}_4$&$\mathcal{C}_6$&$\mathcal{C}_6^\prime$&$\mathcal{C}_9$&$\mathcal{C}_{18}$\\\hline
$H\in \mathcal{C}$ &$C_2$& $C_3$&$C_3$& $C_4$&$\mathfrak{S}_3$&$\mathfrak{S}_3$&$C_3^2$&$C_3\rtimes\mathfrak{S}_3$\\
\hline
$|\mathcal{C}|$ &$9$& $2$&$2$& $9$&$6$&$6$&$1$&$1$\\
\hline
$[C_3^2\rtimes C_4:H]$ & $18$ & $12$ & $12$ & $9$ & $6$ & $6$ & $4$ & $2$\\
\hline
\end{tabular}
\end{center}
\end{lemma}

\begin{corollary}
\label{corollary:36-9-act}
Let $\Sigma$ be a finite set such that $|\Sigma|>1$ and $C_3^2\rtimes C_4$ acts transitively on $\Sigma$. Then
$$
|\Sigma|\in \{2,4,6,9,12,18,36\}.
$$
\end{corollary}

\begin{lemma}
\label{lemma:36-10-subgroups}
Conjugacy classes of proper subgroups of $\mathfrak{S}_3\times\mathfrak{S}_3$ are described as follows:
\begin{center}\renewcommand{\arraystretch}{1.5}
\begin{tabular}{|c||c|c|c|c|c|c|c|c|c|}
\hline
$\mathcal{C}$&$\mathcal{C}_2$&$\mathcal{C}_2^\prime$&$\mathcal{C}_2^{\prime\prime}$&$\mathcal{C}_3$&$\mathcal{C}_3^\prime$&$\mathcal{C}_3^{\prime\prime}$&$\mathcal{C}_4$&$\mathcal{C}_6$&$\mathcal{C}_{6}^\prime$\\
\hline
$H\in \mathcal{C}$ &$C_2$& $C_2$&$C_2$& $C_3$&$C_3$&$C_3$&$C_2^2$&$C_6$&$C_6$\\
\hline
$|\mathcal{C}|$ &$3$& $3$&$9$& $1$&$1$&$2$&$9$&$3$&$3$\\
\hline
$[\mathfrak{S}_3\times\mathfrak{S}_3:H]$ & $18$ & $18$ & $18$ & $12$ & $12$ & $12$ & $9$& $6$ & $6$\\
\hline
\end{tabular}
\end{center}
\begin{center}\renewcommand{\arraystretch}{1.5}
\begin{tabular}{|c||c|c|c|c|c|c|c|c|c|c|c|}
\hline
$\mathcal{C}$&$\mathcal{C}_6^{\prime\prime}$&$\mathcal{C}_6^{\prime\prime\prime}$&$\mathcal{C}_6^{\prime\prime\prime\prime}$&$\mathcal{C}_6^{\prime\prime\prime\prime\prime}$&$\mathcal{C}_6^{\prime\prime\prime\prime\prime\prime}$&$\mathcal{C}_9$&$\mathcal{C}_{12}$&$\mathcal{C}_{12}^\prime$&$\mathcal{C}_{18}$ & $\mathcal{C}_{18}^\prime$& $\mathcal{C}_{18}^{\prime\prime}$\\
\hline
$H\in \mathcal{C}$ &$\mathfrak{S}_3$& $\mathfrak{S}_3$&$\mathfrak{S}_3$& $\mathfrak{S}_3$&$\mathfrak{S}_3$&$C_3^2$&$\mathfrak{D}_6$&$\mathfrak{D}_6$&$C_3\times \mathfrak{S}_3$ & $C_3\times \mathfrak{S}_3$ &$C_3\rtimes \mathfrak{S}_3$ \\
\hline
$|\mathcal{C}|$ &$1$& $1$&$3$& $3$&$6$&$1$&$3$&$3$&$1$&$1$ &$1$ \\
\hline
$[\mathfrak{S}_3\times\mathfrak{S}_3:H]$ & $6$ & $6$ & $6$ & $6$ & $6$ & $4$ & $3$& $3$ & $2$ & $2$& $2$\\
\hline
\end{tabular}
\end{center}
\end{lemma}

\begin{corollary}
\label{corollary:36-10-act}
Let $\Sigma$ be a finite set such that $|\Sigma|>1$ and $\mathfrak{S}_3\times\mathfrak{S}_3$ acts transitively on $\Sigma$. Then
$$
|\Sigma|\in \{2,3,4,6,9,12,18,36\}.
$$
\end{corollary}

\begin{lemma}
\label{lemma:36-9-reps}
Irreducible complex representations of the~group $C_3^2\rtimes C_4$ can be described as follows:
\begin{enumerate}
\item[$(\mathrm{1})$] $\mathbb{I}$ is the~trivial $1$-dimensional representation,
\item[$(\mathrm{2})$] $\mathbb{V}_1$ is the~$1$-dimensional representation of order $2$,
\item[$(\mathrm{3})$] $\mathbb{V}_1^\prime$ is a $1$-dimensional representation of order $4$,
\item[$(\mathrm{4})$] $\mathbb{V}_1^{\prime\prime}$ is a $1$-dimensional representation of order $4$,
\item[$(\mathrm{5})$] $\mathbb{V}_4$ is a faithful $4$-dimensional  representation,
\item[$(\mathrm{6})$] $\mathbb{V}_4^\prime$ is a faithful $4$-dimensional  representation.
\end{enumerate}
The representations $\mathbb{V}_4$ and $\mathbb{V}_4^\prime$ differ by an automorphism of the~group $G$.
\end{lemma}

\begin{lemma}
\label{lemma:36-10-reps}
Irreducible complex representations of the~group $\mathfrak{S}_3\times\mathfrak{S}_3$ can be described as follows:
\begin{enumerate}
\item[$(\mathrm{1})$] $\mathbb{I}$ is the~trivial $1$-dimensional representation,
\item[$(\mathrm{2})$] $\mathbb{V}_1$ is a~$1$-dimensional representation of order $2$,
\item[$(\mathrm{3})$] $\mathbb{V}_1^\prime$ is a $1$-dimensional representation of order $2$,
\item[$(\mathrm{4})$] $\mathbb{V}_1^{\prime\prime}$ is a $1$-dimensional representation of order $2$,
\item[$(\mathrm{5})$] $\mathbb{V}_2$ is a $2$-dimensional  representation lifted from $\mathfrak{S}_3$,
\item[$(\mathrm{6})$] $\mathbb{V}_2^\prime$ is a $2$-dimensional representation lifted from $\mathfrak{D}_6$,
\item[$(\mathrm{7})$] $\mathbb{V}_2^{\prime\prime}$ is a $4$-dimensional  representation lifted from $\mathfrak{S}_3$,
\item[$(\mathrm{8})$] $\mathbb{V}_2^{\prime\prime\prime}$ is a $4$-dimensional  representation lifted from $\mathfrak{D}_6$,
\item[$(\mathrm{9})$] $\mathbb{V}_4$ is a faithful $4$-dimensional  representation.
\end{enumerate}
\end{lemma}

We proceed to actions of the~groups $C_3^2\rtimes C_4$ and $\mathfrak{S}_3\times\mathfrak{S}_3$ on curves.

\begin{lemma}
\label{lemma:36-9-curves}
Let $C$ be a smooth irreducible curve that admits a faithful action of the~group $C_3^2\rtimes C_4$, and let $g$ be the genus of the curve $C$. Then the~following assertions holds:
\begin{itemize}
\item orbits in $C$ are of length $9$, $12$, $18$ or $36$;
\item the~curve $C$ is not hyperelliptic;
\item if $g<10$, then either $g=1$ or $g=4$;
\item if $g=4$, then orbits of length $<36$ in $C$ can be described as follows:
\begin{itemize}
\item one orbit of length $12$;
\item two orbits of length $9$;
\end{itemize}
\item if $g=1$ and $C$ has at least one orbit of length $9$, then $C$ has an orbit of length $18$, it has two orbits of length $9$, and all other orbits in $C$ are of lenth $36$.
\end{itemize}
\end{lemma}

\begin{proof}
If $P$ is a point in $C$, its stablizer in $C_3^2\rtimes C_4$ acts faithfully on the~tangent space $T_{P}(C)$, so it must be cyclic, which implies that orbits in $C$ have length $9$, $12$, $18$ or $36$ by Lemma~\ref{lemma:36-9-subgroups}.

The curve $C$ is not hyperelliptic, since $C_3^2\rtimes C_4$ is not contained in $\mathrm{PGL}_2(\mathbb{C})$ and it does not have normal subgroups isomorphic to $C_2$ by Lemma~\ref{lemma:36-9-subgroups}.

The genus bound follows from \cite{Breuer,Jen,lmfdb}.

If $g=4$, then it follows from \cite{lmfdb} that the~curve $C$ has one orbit of length $12$, and $C$ has two orbits of length $9$.

Finally, we suppose that $g=1$, and the curve $C$  contains at least one $C_3^2\rtimes C_4$-orbit of length~$9$. Then the quotient of the curve $C$ by the group $C_3^2\rtimes C_4$
is isomorphic to $\mathbb{P}^1$. Moreover, applying the~Hurwitz formula to the~quotient map $C\to\mathbb{P}^1$, we get
$$
72=1\times 18a_{18}+2\times 12a_{12}+3\times 9a_{9}.
$$
where $a_{18}$, $a_{12}$ and $a_{9}$ are the~numbers of the~orbits in $C$ of length $18$, $12$ and $9$, respectively.
Since $a_{9}\geqslant 1$, we get $a_{18}=1$, $a_{12}=0$ and $a_9=0$ as required.
\end{proof}

\begin{lemma}
\label{lemma:36-10-curves}
Let $C$ be a smooth irreducible curve that admits a faithful action of the~group $\mathfrak{S}_3\times\mathfrak{S}_3$, and let $g$ be the genus of the curve $C$. Then the~following assertions holds:
\begin{itemize}
\item orbits in $C$ are of length $6$, $12$, $18$ or $36$;
\item the~curve $C$ is not hyperelliptic;
\item if $g<10$, then either $g=4$ or $g=7$;
\item if $g=7$, then orbits of length $<36$ in $C$ can be described as follows:
\begin{itemize}
\item three orbits of length $18$;
\item one orbit of length $6$;
\end{itemize}
\item if $g=4$, then orbits of length $<36$ in $C$ can be described as follows:
\begin{itemize}
\item either $C$ has one orbit of length $18$ and two orbits of length $6$;
\item or $C$ has  three orbits of length $18$ and one orbit of length $12$.
\end{itemize}
\end{itemize}
\end{lemma}

\begin{proof}
If $g=1$, $\mathfrak{S}_3\times\mathfrak{S}_3$ has an abelian subgroup $H$ such that $\mathfrak{S}_3\times\mathfrak{S}_3/H\simeq C_n$ for $n\in\{2,3,4,6\}$, which is impossible by Lemma~\ref{lemma:36-10-subgroups}.
This gives  $g\ne 1$. Now, arguing as in the proof of Lemma~\ref{lemma:36-9-curves} and using Lemma~\ref{lemma:36-10-subgroups}, we obtain all remaining assertions.
\end{proof}

Let us classify $G$ up to conjugation in  $\mathrm{GL}_5(\mathbb{C})$. To do this, consider the following projective transformations:
$$
M_1=\begin{pmatrix}
\xi_3 &  0  & 0 & 0 & 0 \\
 0  & 1 & 0 & 0 & 0 \\
 0  &   0   & \xi_3^2  & 0 & 0 \\
 0  &   0   & 0 & 1 & 0 \\
 0  &   0   & 0 & 0 & 1
\end{pmatrix}, \quad
M_2=\begin{pmatrix}
1 & 0 & 0 & 0 & 0 \\
0 & \xi_3 & 0 & 0 & 0 \\
0 & 0 & 1 & 0 & 0 \\
0 & 0 & 0 & 1 & 0 \\
0 & 0 & 0 & 0 & \xi_3^2
\end{pmatrix},
$$
$$
M_3=\begin{pmatrix}
0 & 1 & 0 & 0 & 0 \\
0 & 0 & 1 & 0 & 0 \\
0 & 0 & 0 & 0 & 1 \\
0 & 0 & 0 & 1 & 0 \\
1 & 0 & 0 & 0 & 0
\end{pmatrix},\quad
M_4=\begin{pmatrix}
0 & 0 & 1 & 0 & 0 \\
0 & 0 & 0 & 0 & 1 \\
1 & 0 & 0 & 0 & 0 \\
0 & 0 & 0 & 1 & 0 \\
0 & 1 & 0 & 0 & 0
\end{pmatrix},\quad %(0,2)(1,4)
M_5=\begin{pmatrix}
0 & 1 & 0 & 0 & 0 \\
1 & 0 & 0 & 0 & 0 \\
0 & 0 & 0 & 0 & 1 \\
0 & 0 & 0 & 1 & 0 \\
0 & 0 & 1 & 0 & 0 %(0,1)(2,4)
\end{pmatrix}.
$$
Let $G_{36,9}$, $G_{36,10}$ and $G_{72,40}$ be subgroups in $\mathrm{PGL}_5(\mathbb{C})$ defined as follows:
\begin{align*}
G_{36,9}&=\langle M_1,M_2,M_3\rangle,\\
G_{36,10}&=\langle M_1,M_2,M_4,M_5\rangle,\\
G_{72,40}&=\langle M_1,M_2,M_3,M_4,M_5\rangle.
\end{align*}
Then $G_{36,9}\simeq C_3^2\rtimes C_4$, $G_{36,10}\simeq \mathfrak{S}_3\times\mathfrak{S}_3$, $G_{72,40}\simeq \mathfrak{S}_3\wr C_2$,
and the subgroup $G_{72,40}$ contains both subgroups $G_{36,9}$ and $G_{36,10}$ as normal subgroups.
We claim that $G$ is conjugated to $G_{39,9}$ or $G_{39,10}$.

Indeed, the subgroup $G\subset\mathrm{PGL}_5(\mathbb{C})$ lifts isomorphically to $\mathrm{GL}_5(\mathbb{C})$ by \cite[Lemma 2.4]{AbbanCheltsovKishimotoMangolte2025}, and the~embedding $G\hookrightarrow\mathrm{GL}_5(\mathbb{C})$ is given by a $5$-dimensional $G$-representation $\mathbb{U}$,
which does not contain $3$-dimensional subrepresentations.
Thus, it follows from Lemmas~\ref{lemma:36-9-reps} and \ref{lemma:36-10-reps} that
$$
\mathbb{U}\simeq \mathbb{U}_1\oplus\mathbb{U}_4,
$$
where $\mathbb{U}_1$ is a one-dimensional representation of the group $G$, and $\mathbb{U}_4$ is its $4$-dimensional irreducible representation.
Now, tensoring $\mathbb{U}$ by the representation $\mathbb{U}_1^\vee$, we may  assume that $\mathbb{U}_1$ is the trivial representation.
Hence, using Lemmas~\ref{lemma:36-9-reps} and \ref{lemma:36-10-reps} and the notations introduced in these lemmas, we may assume that
$$
\mathbb{V}=\mathbb{I}\oplus\mathbb{V}_4,
$$
Thus, $G$ is conjugated to $G_{36,9}$ or $G_{36,10}$.
Hence, to prove Theorem~\ref{theorem:36-9}, we may assume that
\begin{itemize}
\item either $G=G_{36,9}\simeq C_3^2\rtimes C_4$,
\item or $G=G_{36,10}\simeq\mathfrak{S}_3\times\mathfrak{S}_3$.
\end{itemize}
Now, we can find the defining equation of $X$. Namely, for every $a\in \mathbb{C}$, let
$$
X_a=\Big\{x_0^3+x_1^3+x_2^3+x_3^3+x_4^3+ax_3(x_0x_2+x_1x_4)=0\Big\}\subset\mathbb{P}^4.
$$
Then $X_a$ is an irreducible normal cubic hypersurface in $\mathbb{P}^4$, which has at most isolated singularities. Note that $X_a$ is $G_{72,40}$-invariant, so it is $G_{36,9}$-invariant and $G_{36,10}$-invariant.

\begin{lemma}
\label{lemma:36-9-equation}
One has $X=X_a$ for some $a\in\mathbb{C}$.
\end{lemma}

\begin{proof}
Let us only consider the case $G=G_{36,9}$, since the proof is similar in the case $G=G_{36,10}$.
In the notations of Lemma~\ref{lemma:36-9-reps}, we have
$$
\mathrm{Sym}^3\big(\mathbb{I}\oplus\mathbb{V}_4\big)\simeq \mathbb{I}\oplus \mathbb{I}\oplus \mathbb{I}\oplus \mathbb{V}_1\oplus \mathbb{V}_1\oplus \mathbb{V}_1^\prime \oplus \mathbb{V}_1^{\prime\prime}\oplus \mathbb{V}_4\oplus \mathbb{V}_4\oplus \mathbb{V}_4\oplus \mathbb{V}_4\oplus \mathbb{V}_4^\prime\oplus \mathbb{V}_4^\prime\oplus\mathbb{V}_4^\prime.
$$
In this decomposition, the summand $\mathbb{V}_1\oplus \mathbb{V}_1$ corresponds to the singular cubics given by
$$
\lambda x_3(x_0x_2-x_1x_4)+\mu (x_0^3-x_1^3+x_2^3-x_4^3)=0,
$$
where $[\lambda:\mu]\in\mathbb{P}^1$. Similarly, the summands $\mathbb{V}_1^\prime$ and $\mathbb{V}_1^{\prime\prime}$ correspond to the cubic cones
$$
\{x_0^3\pm ix_1^3-x_2^3\mp ix_4^3=0\}\subset\mathbb{P}^4.
$$
Finally, the summand $\mathbb{I}\oplus \mathbb{I}\oplus \mathbb{I}$ corresponds to the cubic threefolds
\begin{equation}
\label{equation:36-9-cubics}
ax_3(x_0x_2+x_1x_4)+bx_3^3+c(x_0^3+x_1^3+x_2^3+x_3^3+x_4^3)=0,
\end{equation}
where $[a:b:c]\in\mathbb{P}^2$. Thus, $X$ is given by \eqref{equation:36-9-cubics} with $c\ne 1$.
Hence, we may assume that $c=1$. Moreover, scaling $x_3$ appropriately, we may futher assume that $b=0$, so $[a:b:c]=[a:0:1]$, which simply means that $X=X_a$.
\end{proof}

If the cubic $X_a$ is smooth, then 
$$
-27\ne a^3\ne -\frac{27}{2}.
$$
Indeed, if $a^3=-27$, then $X_a$ has six nodes. If $a^3=-\frac{27}{2}$, then $X_a$ is the unique nine-nodal cubic threefold that admits an effective $G_{72,40}$-action \cite{CheltsovTschinkelZhang2025}, so $X_a$ is isomorphic to the cubic \eqref{equation:cubic-9-nodes}.

In this section, we will prove the following result, which implies both Theorems~\ref{theorem:36-9-smooth} and \ref{theorem:singular}, and it also implies \cite[Theorem 9.6]{CheltsovTschinkelZhang2025}.

\begin{theorem}
\label{theorem:36-9}
If $X_a$ is smooth or $a^3=-\frac{27}{2}$, then $X_a$ is $G$-birationally superrigid.
\end{theorem}

In the remaining part of this section, we will always assume that 
\begin{itemize}
\item either $X_a$ is smooth,
\item or $a^3=-\frac{27}{2}$ and $X_a$ has nine nodes.
\end{itemize}
In particular, $a^3\ne -27$. If $a^3=-\frac{27}{2}$, we may further assume that $a=-\frac{3}{\sqrt[3]{2}}$, where $\sqrt[3]{2}\in\mathbb{R}$.

\begin{remark}
If $a=-\frac{3}{\sqrt[3]{2}}$, the singular locus $\mathrm{Sing}(X_a)$ is the $G$-orbit of the point $[1:1:1:\sqrt[3]{2}:1]$, which consists of $9$ points. In this case, $X_a$ contains $9$ planes $\Pi_{i,j}$ given by
$$
\left\{\aligned
&2\zeta_3^ix_0+2\zeta_3^{2i}x_2+\sqrt[3]{4}x_3=0,\\
&2\zeta_3^jx_1+2\zeta_3^{2i}x_4+\sqrt[3]{4}x_3=0,
\endaligned
\right.
$$
where $i,j\in\{0,1,2\}$. Observe that each plane $\Pi_{i,j}$ contains $4$ singular points of the cubic $X_a$. Note~also that the group $G$ permutes these $9$ planes transitively, which implies that $\mathrm{rk}\,\mathrm{Cl}(X_a)^G=1$, so   $X_a$ is indeed a $G$-Mori fiber space (over a point) as claimed in \cite{Avilov2016a,CheltsovTschinkelZhang2025}.
\end{remark}

Let $H$ be a general hyperplane section of $X_{a}$, and let
$$
S=\{x_3=0\}\cap X_{a}.
$$
Then $S$ is a smooth $G$-invariant cubic surface. Note that $S$ is the~only $G$-invariant surface in $|H|$,
the surface $S$ is $G_{72,40}$-invariant, and the~group $G_{72,40}$ acts faithfully on $S$.

\begin{lemma}
\label{lemma:36-9-S-Pic}
Suppose that $G=G_{36,9}$. Then $\mathrm{Pic}(S)^G=\mathbb{Z}[-K_S]$.
\end{lemma}

\begin{proof}
The required assertion follows from \cite[Theorem 6.14]{DolgachevIskovskikh2009} or \cite{CheltsovTschinkelZhang2026}, and can be easily checked explicitly by looking how $G=G_{36,9}$ acts on the lines in $S$.
\end{proof}

If $G=G_{36,10}$, then $\mathrm{Pic}(S)^G\ne\mathbb{Z}[-K_S]$.
Indeed, let $\mathcal{L}_3$ be the union of the following lines:
\begin{align*}
\big\{x_3&=0, x_0+x_1=0, x_2+x_4=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3 x_1=0, x_2+\zeta_3^2x_4=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3^2x_1=0, x_2+\zeta_3x_4=0\big\},
\end{align*}
and let $\mathcal{L}_3^\prime$ be the union of the following lines:
\begin{align*}
\big\{x_3&=0, x_0+x_4=0, x_1+x_2=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3x_4=0, x_1+\zeta_3x_2=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3^2x_4=0, x_1+\zeta_3^2x_2=0\big\}.
\end{align*}
Then $\mathcal{L}_3$ and $\mathcal{L}_3^\prime$ are both $G_{36,10}$-irreducible curves in $S$, and each of these curves consists of three disjoint lines. In particular, if $G=G_{36,10}$, then $\mathrm{Pic}(S)^G\ne\mathbb{Z}[-K_S]$. Set $\mathcal{L}_6=\mathcal{L}_3+\mathcal{L}_3^\prime$. Then
\begin{equation}
\label{equation:39-9-L6}
\mathcal{L}_6=\big\{x_0x_2-x_1x_4=0\big\}\cap S,
\end{equation}
and $\mathcal{L}_6$ is $G_{36,9}$-irreducible.

\begin{lemma}
\label{lemma:36-10-S-Pic}
Suppose that $G=G_{36,10}$. Then $\mathrm{rk}\,\mathrm{Pic}(S)^G=2$, and there is a $G$-Sarkisov link
$$
\xymatrix{
&S\ar@{->}[ld]_\pi\ar@{->}[rd]^{\pi^\prime} & \\
Y\ar@{-->}[rr]_\tau  && Y}
$$
where $Y$ is the sextic smooth del Pezzo surface with $\mathrm{Pic}(Y)^G=\mathbb{Z}[-K_Y]$, $\pi$ and $\pi^\prime$ are birational contractions of the curves $\mathcal{L}_3$ and $\mathcal{L}_3^\prime$, respectively, and
$\tau$ is a $G_{36,10}$-birational involution.
\end{lemma}

\begin{proof}
Recall that $S$ contains $27$ lines. These $27$ lines form five $G_{39,10}$-irreducible curves, which can be described as follows: the curves  $\mathcal{L}_3$ and $\mathcal{L}_3^\prime$, the curve $\mathcal{L}_6^\prime$ consisting of the lines
\begin{align*}
\big\{x_3&=0, x_0+\zeta_3 x_1=0, x_2+\zeta_3 x_4=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3^2 x_1=0, x_2+x_4=0\big\},\\
\big\{x_3&=0, x_0+x_1=0, x_2+\zeta_3^2x_4=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3^2x_1=0, x_2+\zeta_3^2x_4=0\big\},\\
\big\{x_3&=0, x_0+x_1=0, x_2+\zeta_3x_4=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3x_1=0, x_2+x_4=0\big\},
\end{align*}
the curve $\mathcal{L}_6^{\prime\prime}$ consisting of the lines
\begin{align*}
\big\{x_3&=0, x_0+\zeta_3 x_4=0, x_1+\zeta_3 x_2=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3^2 x_4=0, x_1+x_2=0\big\},\\
\big\{x_3&=0, x_0+x_4=0, x_1+\zeta_3^2x_2=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3^2x_4=0, x_1+\zeta_3^2x_2=0\big\},\\
\big\{x_3&=0, x_0+x_4=0, x_1+\zeta_3x_2=0\big\},\\
\big\{x_3&=0, x_0+\zeta_3x_4=0, x_1+x_2=0\big\}.
\end{align*}
and the curve $\mathcal{L}_9$ consisting of the nine lines
$$
L_{i,j}=\big\{x_3=0, x_0+\zeta_3^ix_2=0, x_1+\zeta_3^jx_4=0\big\},
$$
where $i,j\in\{0,1,2\}$.
Note that $\mathcal{L}_3\cap\mathcal{L}_6^{\prime\prime}=\varnothing$ and $\mathcal{L}_3^\prime\cap\mathcal{L}_6^{\prime}=\varnothing$.
Moreover, we have 
$$
\mathcal{L}_3+\mathcal{L}_3^{\prime}\sim -K_S,
$$
and $\mathcal{L}_3+3\mathcal{L}_3^{\prime}\sim 2\mathcal{L}_6^{\prime}$,
$3\mathcal{L}_3+\mathcal{L}_3^{\prime}\sim 2\mathcal{L}_6^{\prime\prime}$,
$\mathcal{L}_6^{\prime}+\mathcal{L}_6^{\prime\prime}\sim -4K_S$, $\mathcal{L}_9\sim -3K_S$.
In particular, $\mathrm{rk}\,\mathrm{Pic}(S)^G=2$, since $\mathrm{Pic}(S)$ is generated by lines.
The existence of the $G$-Sarkisov link is known \cite{Iskovskikh1996,CheltsovTschinkelZhang2026}, and we can chose $\tau\in\mathrm{Bir}^G(Y)$ such that $\tau$ is induced by the involution $\iota\in G_{72,40}$ given by
$$
[x_0:x_1:x_2:x_3:x_4]\mapsto [x_0:x_4:x_2:x_3:x_1],
$$
because $\iota(\mathcal{L}_3)=\mathcal{L}_3^\prime$.
\end{proof}

Let us describe small $G_{36,9}$-orbits in $X_{a}$.
Let $\Sigma_6$ be the~$G_{36,9}$-orbit of the~point $[0:1:0:0:-1]$.
Then $\Sigma_6$ is a $G_{36,9}$-orbit of length $6$. Observe that
$$
[0:1:0:s:1]\in X_{a}\iff s^3+as+2=0.
$$
For every $s\in\mathbb{C}$ such that $s^3+as+2=0$, let $\Sigma_6^s$ be the~$G_{36,9}$-orbit of the~point $[0:1:0:s:1]$.
Note that the~discriminant of the~polynomial $s^3+as+2$ is $-4(a^3+27)\ne 0$ by our assumption, so there are exactly three possibilities for $\Sigma_6^s$. In each case, $\Sigma_6^s$ is a $G_{36,9}$-orbit of length $6$.

Let $\Sigma_9$ and $\Sigma_9^{\pm}$ be the~$G_{36,9}$-orbits of the points $[1:-1:1:0:-1]$ and $[1:\mp i:-1:0:\pm i]$, respectively.
Then $\Sigma_9$, $\Sigma_9^{+}$ and $\Sigma_9^{-}$ are $G_{36,9}$-orbits of length $9$. Observe that
$$
[1:1:1:t:1]\in X_{a}\iff t^3+2at+4=0.
$$
For every $t\in\mathbb{C}$ such that \mbox{$t^3+2at+4=0$}, let $\Sigma_9^t$ be the~$G_{36,9}$-orbit of the~point $[1:1:1:t:1]$.
Then $\Sigma_9^t$ is a $G_{36,9}$-orbit of length $9$. If $X_a$ is smooth ($a^3\ne-\frac{27}{2}$), the polynomial $t^3+2at+4$ does not have multiple roots, so there are three possibilities for  the~$G_{36,9}$-orbit $\Sigma_9^t$.
If $a^3=-\frac{27}{2}$, then 
$$
\mathrm{Sing}\big(X_a\big)=\mathrm{Orb}_G([a:a:a:-3:a])=\Sigma_9^{t}
$$
for $t=-\frac{3}{a}$. 
For instance, if $a=-\frac{3}{\sqrt[3]{2}}$, then 
$$
t^3+2at+4=(t+2\sqrt[3]{2})(t-\sqrt[3]{2})^2,
$$
and $\Sigma_9^{\sqrt[3]{2}}=\mathrm{Sing}(X_a)$, while $\mathrm{Orb}_G([1:1:1:-2\sqrt[3]{2}:1])\subset X_a\setminus\mathrm{Sing}(X_a)$.

Let $\Sigma_{12}$ be the~$G_{36,9}$-orbit of the~point $[1:-1:0:0:0]$.
Then $\Sigma_{12}$ has length $12$. Set
$$
\mathcal{R}_6=\big(\{x_0=0,x_2=0\}\cap X_{a}\big)\cup\big(\{x_1=0,x_4=0\}\cap X_{a}\big).
$$
Then $\mathcal{R}_6$ is a $G_{36,9}$-irreducible curve of degree $6$ that is a disjoint union of two cubic curves in $X_{a}$,
which are smooth \cite{ArtebaniDolgachev}, and they are contained in the smooth locus of the cubic $X_a$ in the case when $X_a$ is singular. These curves are the~fixed loci in $X_{a}$ of the~transformations $M_1$ and~$M_2$.
Thus, the length of the~$G_{36,9}$-orbit of every point of the curve $\mathcal{R}_6$ is at most $12$.

\begin{lemma}
\label{lemma:36-9-orbits}
Let $\Sigma$ be a $G_{36,9}$-orbit in $X_{a}$ of length $r\leqslant 12$. Then one of the~following cases holds:
\begin{itemize}
\item[$(\mathrm{a})$] $r=6$ and $\Sigma=\Sigma_6$;
\item[$(\mathrm{b})$] $r=6$ and $\Sigma=\Sigma_6^s$ for some $s\in\mathbb{C}$ such that $s^3+as+2=0$;
\item[$(\mathrm{c})$] $r=9$ and $\Sigma$ is one of the $G$-orbits $\Sigma_9$, $\Sigma_9^{+}$, $\Sigma_9^{-}$;
\item[$(\mathrm{d})$] $r=9$ and $\Sigma=\Sigma_9^t$ for some $t\in\mathbb{C}$ such that $t^3+2at+4=0$;
\item[$(\mathrm{e})$] $r=12$ and $\Sigma=\Sigma_{12}$;
\item[$(\mathrm{f})$] $r=12$, and $\Sigma\subset  \mathcal{R}_6$.
\end{itemize}
\end{lemma}

\begin{proof}
Let $P$ be a point in $\Sigma$, and let $G_P$ be its stabilizer in $G_{36,9}$.
Since $G_{36,9}$ fix no points in $X_{a}$, it follows from Lemma~\ref{lemma:36-9-subgroups} that that $G_P$ is isomorphic to $C_3\rtimes\mathfrak{S}_3$, $C_3^2$, $\mathfrak{S}_3$, $C_4$ or $C_3$. Then
$$
r\in\{2,4,6,9,12\}.
$$
In fact, if $G_P\simeq C_3\rtimes\mathfrak{S}_3$ or $G_P\simeq C_3^2$, then $G_P$ does not fix points in $X_{a}$, which is a contradiction. Therefore, we see that $r\ne 2$ and $r\ne 4$. Then one of the following three cases holds:
\begin{itemize}
\item $r=6$ and $G_P\simeq \mathfrak{S}_3$;
\item $r=9$ and $G_P\simeq C_4$;
\item $r=12$ and $G_P\simeq C_3$.
\end{itemize}
Let us deal with these cases one by one.

Suppose that $r=6$ and $G_P\simeq \mathfrak{S}_3$.
By Lemma~\ref{lemma:36-9-subgroups}, the group $G_{36,9}$ contains two subgroups isomorphic to $\mathfrak{S}_3$ up to conjugation, so we may assume that $G_P=\langle M_1,M_3^2\rangle$, or $G_P=\langle M_1M_2,M_3^2\rangle$.
If $G_P=\langle M_1M_2,M_3^2\rangle$, then $G_P$ does not fix points in $X_{a}$.
Thus, we conclude that $G_P=\langle M_1,M_3^2\rangle$. In this case, $G_P$ fixes the~following points:
$$
[0:1:0:0:-1], [0:1:0:s:1],
$$
where $s\in\mathbb{C}$ such that $s^3+as+2=0$. This gives cases $(\mathrm{a})$ or $(\mathrm{b})$.

Now, we suppose that $r=9$ and $G_P\simeq C_4$.
By Lemma~\ref{lemma:36-9-subgroups}, $G_{36,9}$ has a unique subgroup isomorphic to $C_4$ up to conjugation, so we may assume $G_P=\langle M_3\rangle$. Then $G_P$-fixed points are
$$
[1:-1:1:0:-1], [1:\mp i:-1:0:\pm i], [1:1:1:t:1],
$$
where $t\in\mathbb{C}$ such that $t^3+2at+4=0$. This gives cases $(\mathrm{c})$ or $(\mathrm{d})$.

Finally, we suppose that $r=12$ and $G_P\simeq C_3$. Up to conjugation, $G$ contains two subgroups isomorphic to $C_3$ up to conjugation.
Therefore, we may assume that $G_P=\langle M_1\rangle$ or $G_P=\langle M_1M_2\rangle$.
If $G_P=\langle M_1\rangle$, then $P\in  \mathcal{R}_6$, which gives case $(\mathrm{f})$.
If $G_P=\langle M_1M_2\rangle$, $G_P$ fixes the~following points:
\begin{align*}
[1:-1:0:0:0], [1:-\xi_3:0:0:0], [1:-\xi_3^2:0:0:0], \\
[0:0:1:0:-1], [0:0:1:0:-\xi_3], [0:0:1:0:-\xi_3^2].
\end{align*}
All of them are contained in $\Sigma_{12}$. This gives case $(\mathrm{e})$.
\end{proof}

\begin{corollary}[{cf. \cite[Theorem 1.12]{CheltsovWilson2013}}]
\label{corollary:36-9-alpha}
Suppose that $G=G_{36,9}$. Then $\alpha_G(S)=2$.
\end{corollary}

\begin{proof}
Since $|-K_S|$ contains a $G$-invariant curve, $\alpha_G(S)\leqslant 2$. Since $\mathrm{rk}\,\mathrm{Pic}(S)^G=1$ by Lemma~\ref{lemma:36-9-S-Pic}, and~$|-K_S|$  has no $G$-invariant curves, we get $\alpha_G(S)=2$ by \cite[Lemma 5.1]{Cheltsov2008} and Lemma~\ref{lemma:36-9-orbits}.
\end{proof}

Observe that $\Sigma_6$ and $\Sigma_6^s$ are also $G_{36,10}$-orbits of length $6$, where $s\in\mathbb{C}$ such that $s^3+as+2=0$.
Similarly, both $\Sigma_9$ and $\Sigma_9^t$ are $G_{36,10}$-orbits of length $9$, where $t\in\mathbb{C}$ such that $t^3+2at+4=0$.
Now, let~$\Sigma_{6}^\prime$ and $\Sigma_{6}^{\prime\prime}$ be the~$G_{36,10}$-orbits of the~points $[1:-1:0:0:0]$ and $[0:1:-1:0:0]$, respectively. Then both
$\Sigma_{6}^\prime$ and $\Sigma_{6}^{\prime\prime}$ are $G_{36,10}$-orbits of length $6$, and
$$
\Sigma_{12}=\Sigma_{6}^\prime\cup\Sigma_{6}^{\prime\prime}.
$$
Let $\Sigma_9^\prime$ and $\Sigma_9^{\prime\prime}$ be the~$G_{36,10}$-orbits of $[1:1:-1:0:-1]$ and $[1:-1:-1:0:1]$, respectively. Then $\Sigma_9^\prime$ and $\Sigma_9^{\prime\prime}$ are $G_{36,10}$-orbits of length $9$.

\begin{lemma}
\label{lemma:36-10-orbits}
Let $\Sigma$ be a $G_{36,10}$-orbit in $X_{a}$ of length $r\leqslant 9$, let $P$ be a point in $\Sigma$, and let $G_P$ be its stabilizer in $G_{36,10}$. Then one of the~following cases holds:
\begin{itemize}
\item[$(\mathrm{a})$] $r=6$, $G_P\simeq \mathfrak{S}_3$, and $\Sigma=\Sigma_6$;
\item[$(\mathrm{b})$] $r=6$, $G_P\simeq C_6$, and $\Sigma$ is one of the $G$-orbits $\Sigma_{6}^\prime$ or $\Sigma_6^{\prime\prime}$;
\item[$(\mathrm{c})$] $r=6$, $G_P\simeq\mathfrak{S}_3$, and $\Sigma=\Sigma_6^s$ for some $s\in\mathbb{C}$ such that $s^3+as+2=0$;
\item[$(\mathrm{d})$] $r=9$, $G_P\simeq C_2^2$, and $\Sigma$ is one of the $G$-orbits $\Sigma_{9}$, $\Sigma_9^\prime$, $\Sigma_9^{\prime\prime}$;
\item[$(\mathrm{e})$] $r=9$, $G_P\simeq C_2^2$, and $\Sigma=\Sigma_9^t$ for some $t\in\mathbb{C}$ such that $t^3+2at+4=0$.
\end{itemize}
\end{lemma}

\begin{proof}
Since $G_{36,10}$ does not fix points in $X_{a}$, we see that $r\ne 1$. If $r=2$ or $r=3$, then it follows from Lemma~\ref{lemma:36-10-subgroups} that $G_P$ is isomorphic to one of the following groups $\mathfrak{D}_6$, $C_3\times\mathfrak{S}_3$, $C_3\rtimes\mathfrak{S}_3$. In~each of these cases, the restriction of the representation $\mathbb{V}_4$ to the subgroup $G_P$ does not have one-dimensional subrepresentations, so $G_P$ does not fix points in $X_{a}$, which is a contradiction. Similarly, if $r=4$, then it follows from Lemma~\ref{lemma:36-10-subgroups} that $G_P=\langle M_1,M_2\rangle\simeq C_3^2$, so $G_P$ does not fix points in $X_{a}$ either. Hence, it follows from Lemma~\ref{lemma:36-10-subgroups} that one of the following cases holds:
\begin{itemize}
\item $r=6$ and $G_P\simeq C_6$;
\item $r=6$ and $G_P\simeq\mathfrak{S}_3$;
\item $r=9$ and $G_P\simeq C_2^2$.
\end{itemize}
Let us deal with these cases one by one.

Suppose that $G_P\simeq C_6$. By Lemma~\ref{lemma:36-10-subgroups}, we may assume that $G_P=\langle M_1M_5\rangle$ or $G_P=\langle M_1M_4M_5\rangle$. If $G_P=\langle M_1M_5\rangle$, then all $G_P$-fixed points are contained in $\Sigma_6^\prime$, so $\Sigma=\Sigma_6^\prime$. If $G_P=\langle M_1M_4M_5\rangle$, then all $G_P$-fixed points are contained in $\Sigma_6^{\prime\prime}$, so $\Sigma=\Sigma_6^{\prime\prime}$.

Suppose that $r=6$ and $G_P\simeq\mathfrak{S}_3$.
By Lemma~\ref{lemma:36-10-subgroups}, $G_{36,10}$ has five subgroups isomorphic to $\mathfrak{S}_3$ up to conjugation.
If $G_P$ is conjugated to $\langle M_1,M_4\rangle$, the $G_P$-fixed locus in $X_{a}$ consists of the points
$$
[0:1:0:0:-1], [0:1:0:s:1],
$$
where $s\in\mathbb{C}$ such that $s^3+as+2=0$. Thus, in this case, either $\Sigma=\Sigma_6$ or $\Sigma_6^s$. On the other hand, if $G_P$ is not conjugated to $\langle M_1,M_4\rangle$, the restriction of the representation $\mathbb{V}_4$ to $G_P$ does not have one-dimensional subrepresentations, so $G_P$ does not fix points in $X_{a}$, which is a contradiction.

Finally, we consider the case $r=9$. Then $G_P\simeq C_2^2$.
By Lemma~\ref{lemma:36-10-subgroups}, $G_{36,10}$ has a unique subgroup isomorphic to $C_2^2$ up to conjugation.
Thus, we may assume that $G_P=\langle M_4,M_5\rangle$. Then $G_P$ fixes the~following points in $X_{a}$:
$$
[1:-1:1:0:-1], [1:1:-1:0:-1], [1:-1:-1:0:1],  [1:1:1:t:1],
$$
where $t\in\mathbb{C}$ such that $t^3+2at+4=0$. Then $\Sigma$ is one of the $G$-orbits $\Sigma_{9}$, $\Sigma_9^\prime$, $\Sigma_9^{\prime\prime}$, $\Sigma_9^t$.
\end{proof}

\begin{corollary}
\label{corollary:36-9-36-10-curves-in-S}
Let $Z$ be a $G$-invariant curve in the cubic $X_{a}$ such that $Z\subset S$ and $\mathrm{deg}(Z)<6$. Then $G=G_{36,10}$, and either $Z=\mathcal{L}_3$ or $Z=\mathcal{L}_3^\prime$.
\end{corollary}

\begin{proof}
If $G=G_{36,9}$, then $\mathrm{deg}(Z)\geqslant 6$ by Lemma~\ref{lemma:36-9-S-Pic}, because $|-K_S|$ does not have $G$-invariant curves. Thus, we have $G=G_{36,10}$. Let us show that either $Z=\mathcal{L}_3$ or $Z=\mathcal{L}_3^\prime$. To do this, we may assume that $Z$ is $G$-irreducible. Suppose that $Z\ne\mathcal{L}_3$ and $Z\ne \mathcal{L}_3^\prime$. Then
$$
|Z\cap\mathcal{L}_3|+|Z\cap\mathcal{L}_3^\prime|\leqslant Z\cdot\mathcal{L}_3+Z\cdot\mathcal{L}_3^\prime=Z\cdot(\mathcal{L}_3+\mathcal{L}_3^\prime)=Z\cdot(-2K_S)=2\mathrm{deg}(X_{a})<12,
$$
because $\mathcal{L}_3+\mathcal{L}_3^\prime\sim -2K_S$ by \eqref{equation:39-9-L6}.
Thus, we see that either $|Z\cap\mathcal{L}_3|<6$ or $|Z\cap\mathcal{L}_3^\prime|<6$ (or both). Without loss of generality, we may assume that $|Z\cap\mathcal{L}_3|<6$. Then $Z\cap\mathcal{L}_3=\varnothing$ by Lemma~\ref{lemma:36-10-orbits}.

Let us use notations of Lemma~\ref{lemma:36-10-S-Pic}. Since $\mathrm{Pic}(Y)^G=\mathbb{Z}[-K_Y]$, there exists  $n\in\mathbb{Z}_{>0}$ such that
$$
\pi(Z)\sim n(-K_Y).
$$
Thus, since $Z$ is disjoint from $\mathcal{L}_3$, we have
$$
Z\sim \pi^*\big(n(-K_Y)\big)\sim n(-K_S)+n\mathcal{L}_3,
$$
which implies that $\mathrm{deg}(Z)=-K_S\cdot Z=6n\geqslant 6$, which contradicts our assumption.
\end{proof}

\begin{corollary}[{cf. \cite[Theorem 1.12]{CheltsovWilson2013}}]
\label{corollary:36-10-alpha}
Suppose that $G=G_{36,10}$. Then $\alpha_G(S)=2$.
\end{corollary}

\begin{proof}
Since $\mathcal{L}_3+\mathcal{L}_3^\prime\sim 2(-K_S)$ by \eqref{equation:39-9-L6}, we get $\alpha_G(S)\leqslant 2$. Suppose that $\alpha_G(S)<2$. Then there exists an effective $G$-invariant $\mathbb{Q}$-divisor $D$ on the surface $S$ such that
$$
D\sim_{\mathbb{Q}} -K_S,
$$
and  $(S,\lambda D)$ is strictly log canonical for some $\lambda<2$.
Thus, the locus $\mathrm{Nklt}(S,\lambda D)$ is $G$-invariant, and it is not empty. We claim that this locus consists of finitely many points.

Indeed, suppose that $\mathrm{Nklt}(S,\lambda D)$ contains a curve. Then $\mathrm{Nklt}(S,\lambda D)$ contains a $G$-irreducible curve $C$ such that
$$
\lambda D=C+D^\prime,
$$
where $D^\prime$ is an effective $\mathbb{Q}$-divisor on $S$ whose support does not contain the curve $C$. Then
$$
6>3\lambda=-K_S\cdot (\lambda D)=-K_S\cdot (C+D^\prime)=\mathrm{deg}(C)-K_S\cdot D^\prime\geqslant \mathrm{deg}(C),
$$
so either $C=\mathcal{L}_3$ or $C=\mathcal{L}_3^\prime$ by Corollary~\ref{corollary:36-9-36-10-curves-in-S}. Write
$$
D=a\mathcal{L}_3+b\mathcal{L}_3^\prime+\Delta_S,
$$
where $\Delta_S$ is an effective $\mathbb{Q}$-divisor on $S$ whose support does not contain the curve $\mathcal{L}_3$ and  $\mathcal{L}_3^\prime$, and $a$ and $b$ are non-negative rational numbers such that at least one of them is equal to $\frac{1}{\lambda}>\frac{1}{2}$. Then
\begin{align*}
3=\mathcal{L}_3\cdot D&=\mathcal{L}_3\cdot \big(a\mathcal{L}_3+b\mathcal{L}_3^\prime+\Delta_S)=9b-3a+\mathcal{L}_3\cdot \Delta_S\geqslant 9b-3a,\\
3=\mathcal{L}_3^\prime\cdot D&=\mathcal{L}_3^\prime\cdot \big(a\mathcal{L}_3+b\mathcal{L}_3^\prime+\Delta_S)=9a-3b+\mathcal{L}_3^\prime\cdot \Delta_S\geqslant 9a-3b.
\end{align*}
These inequalities imply that $a\leqslant\frac{1}{2}$ and $b\leqslant\frac{1}{2}$, which is a contradiction.

Hence, $\mathrm{Nklt}(S,\lambda D)$ is a finite $G$-invariant subset. Set $\Sigma=\mathrm{Nklt}(S,\lambda D)$, and let $\mathcal{I}_\Sigma$ be the ideal sheaf of the subset $\Sigma$. Then it follows from the Nadel vanishing \cite[Theorem 9.4.8]{Lazarsfeld2004} that
$$
h^1\big(S,\mathcal{I}_\Sigma\otimes\mathcal{O}_{S}(-K_S)\big)=0.
$$
Hence, we have the~following exact sequence of groups
$$
0\longrightarrow H^0\big(S,\mathcal{I}_\Sigma\otimes\mathcal{O}_{S}(-K_S)\big)\longrightarrow H^0\big(S,\mathcal{O}_{S}(-K_S)\big)\longrightarrow H^0\big(\Sigma,\mathcal{O}_{\Sigma}\big)\longrightarrow 0,
$$
which gives
$$
|\Sigma|=h^0\big(\Sigma,\mathcal{O}_{\Sigma}\big)=h^0\big(S,\mathcal{O}_{S}(-K_S)\big)-h^0\big(S,\mathcal{I}_\Sigma\otimes\mathcal{O}_{S}(-K_S)\big)=4-h^0\big(S,\mathcal{I}_\Sigma\otimes\mathcal{O}_{S}(-K_S)\big)\leqslant 4,
$$
which is impossible, because $S$ does not contain $G$-orbits of length $\leqslant 4$ by Lemma~\ref{lemma:36-10-orbits}.
\end{proof}

Observe that $|2H|$ contains a pencil consisting of $G_{72,40}$-invariant surfaces. This can be checked using the~following GAP code:
\begin{verbatim}
    G:=SmallGroup(72,40);
    T:=CharacterTable(G);
    Ir:=Irr(T);
    U:=Ir[1]+Ir[6];
    S:=SymmetricParts(T,[U],2);
    Print(MatScalarProducts(Ir,S));
\end{verbatim}
We denote this pencil by $\mathcal{P}$. Then $\mathcal{P}$ is free from fixed components, and $\mathcal{P}$ consists of surfaces
$$
S_\lambda=\big\{x_0x_2+x_1x_4+\lambda x_3^2=0\big\}\cap X_{a},
$$
where $\lambda\in\mathbb{C}\cup\{\infty\}$. Note also that $S_{\infty}=2S$. We set
$$
\mathcal{C}_6=\{x_0x_1+x_2x_4=0\}\cap S=\{x_3=0, x_0x_1+x_2x_4=0\}\cap X_{a}.
$$
Then $\mathcal{C}_6$ is a $G_{72,40}$-invariant smooth curve of genus $4$, so  $\mathcal{C}_6$ is $G_{36,9}$-invariant and $G_{36,10}$-invariant. The~smoothness of this curve can be checked using the following Magma code:
\begin{verbatim}
    Q:=RationalField();
    P4<x0,x1,x2,x3,x4>:=ProjectiveSpace(Q,4);
    C6:=Scheme(P4,[x3,x0*x2+x1*x4,x0^3+x1^3+x2^3+x3^3+x4^3]);
    IsNonsingular(C6);
\end{verbatim}
Moreover, since $\mathcal{C}_6$ is the~base locus of the~pencil $\mathcal{P}$ (taken with reduced structure), a general surface in the pencil $\mathcal{P}$ is a smooth K3 surface. Furthermore, using Lemma~\ref{lemma:36-9-orbits}, we obtain

\begin{corollary}
\label{corollary:36-9-Du-Val}
For every $\lambda\in\mathbb{C}$, the surface $S_\lambda$ has at most isolated ordinary double points (nodes).
Moreover, if the surface $S_\lambda$ is singular, then one of the following two cases holds:
\begin{itemize}
\item $\mathrm{Sing}(S_\lambda)=\Sigma_9^t$ for $t\in\mathbb{C}$ such that $t^3+2at+4=0$, and $\lambda=-\frac{2}{t^2}$;
\item $\mathrm{Sing}(S_\lambda)=\Sigma_6^s$ for $s\in\mathbb{C}$ such that $s^3+as+2$, and $\lambda=-\frac{1}{s^2}$.
\end{itemize}
\end{corollary}

\begin{proof}
Since $\mathcal{C}_6=S\cap S_\lambda$ and the curve $\mathcal{C}_6$ is smooth,
$S_\lambda$ is smooth at every point of this intersection, which implies, in particular, that $S_\lambda$ has isolated singularities, so $S_\lambda$ is normal.

Let $P$ be a singular point of $S_\lambda$ (if any). Then its $G_{36,9}$-orbit has at least $6$ points by Lemma~\ref{lemma:36-9-orbits}.
But $S_{\lambda}$ can have at most two non-Du Val singular points \cite{Umezu}, see also \cite[Theorem 6.9]{Shokurov92} and \cite{Fujino}.
Thus, we conclude that $S_\lambda$ has at most Du Val singularities.

Suppose that $S_\lambda$ is singular.
Let $\widetilde{S}_\lambda$ be its minimal resolution. Then $\widetilde{S}_\lambda$ is a K3 surface, so
$$
\mathrm{rk}\,\mathrm{Pic}\big(\widetilde{S}_\lambda\big)\leqslant 19,
$$
and $|\mathrm{Sing}(S_\lambda)|\leqslant 16$ by \cite{Nikulin}.
Hence, by Lemma~\ref{lemma:36-9-orbits}, one of the~following cases holds:
\begin{enumerate}
\item $\mathrm{Sing}(S_\lambda)$ is a $G_{36,9}$-orbit of length $12$ that is contained in $\mathcal{R}_6$.
\item $\mathrm{Sing}(S_\lambda)=\Sigma_9^t$ for $t\in\mathbb{C}$ such that  $t^3+2at+4=0$, and $\lambda=-\frac{2}{t^2}$;
\item $\mathrm{Sing}(S_\lambda)=\Sigma_6^s$ for $s\in\mathbb{C}$ such that $s^3+as+2=0$, and $\lambda=-\frac{1}{s^2}$.
\end{enumerate}
However, the first case is impossible, since otherwise we would have
$$
12=\mathcal{R}_6\cdot S_\lambda\geqslant 2\big|\mathrm{Sing}(S_\lambda)\big|\geqslant 24,
$$
because $ \mathcal{R}_6\not\subset S_\lambda$. If $\lambda=-\frac{2}{t^2}$ for $t\in\mathbb{C}$ such that $t^3+2at+4=0$, then $S_\lambda$ is singular at $\Sigma_9^t$, and its singularities are nodes (explicit computations).
If  $\lambda=-\frac{1}{s^2}$ for $s\in\mathbb{C}$ such that $s^3+as+2=0$, then $S_\lambda$ is singular at $\Sigma_6^s$, and its singularities are nodes.
\end{proof}

Recall that $\mathcal{C}_6$, $\mathcal{L}_6$, $\mathcal{R}_6$ are $G_{36,9}$-irreducible curves in $X_{a}$.
Let us present six $G_{36,9}$-irreducible curves of degree $9$ in $X_{a}$, whose irreducible components are lines, which are not contained in $S$. For~every $t\in\mathbb{C}$ such that $t^3+2at+4=0$, let $L_t^{\pm}$ be the line in $\mathbb{P}^4$ that is given by
$$
\left\{\aligned
&\pm itx_2-(1\pm i)x_3+tx_4=0,\\
&\mp itx_1\mp itx_4\pm 2ix_3=0,\\
&\mp itx_0-tx_1+(1\pm i)x_3=0.
\endaligned
\right.
$$
Then $L_t^{\pm}\subset X_{a}$, and $L^{\pm}$ contains the~points $[1:\mp i:-1:0:\pm i]\in \Sigma_9^{\pm}$ and $[1:1:1:t:1]\in \Sigma_9^t$.
Let $\mathcal{L}_t^{\pm}$ be the $G$-irreducible curve whose irreducible component is $L^\pm$.
Then $\mathcal{L}_t^{\pm}$ consists of $9$ lines, and these lines are disjoint.
Moreover, $\mathcal{L}_t^{\pm}$ contains $\Sigma_9^{\pm}$ and $\Sigma_9^t$, and
$$
\mathrm{Sing}(S_\lambda)=\Sigma_9^t\subset \mathcal{L}_t^{\pm}\subset S_\lambda,
$$
where $\lambda=-\frac{2}{t^2}$. Furthermore, we have
\begin{equation}
\label{equation:36-9-nine-lines}
\mathcal{L}_t^{+}+\mathcal{L}_t^{-}=\big\{6x_3(x_0x_2-x_1x_4)+t(x_0^3-x_1^3+x_2^3-x_4^3)=0\big\}\cap S_{\lambda}.
\end{equation}
Recall that $\mathcal{C}_6$ and $\mathcal{L}_6$ are contained in $S$. But $\mathcal{R}_6$ is not contained in any surface in  $\mathcal{P}$.

\begin{lemma}
\label{lemma:36-9-reducible-curves}
Let $Z$ be a reducible $G_{36,9}$-irreducible curve in $X_{a}$ of degree $d<12$ such that $Z\not\subset S$.
Then one of the following two cases holds:
\begin{itemize}
\item $d=6$ and $Z=\mathcal{R}_6$,
\item $d=9$ and either $Z=\mathcal{L}_t^{+}$ or $Z=\mathcal{L}_t^{-}$ for some $t\in\mathbb{C}$ such that $t^3+2at+4=0$.
\end{itemize}
\end{lemma}

\begin{proof}
The intersection $S\cap Z$ consists of $d<12$ points (counted with multiplicities), and this intersection is a union of $G_{36,9}$-orbits.
Hence, it follows from Lemma~\ref{lemma:36-9-orbits} that $d=6$ or $9$.

Let $C$ be an irreducible component of the curve $Z$, and let $G_C$ be its stabilizer in the group~$G_{36,9}$.
Then $Z$ cannot have $3$ irreducible components by Corollary~\ref{corollary:36-9-act}. Therefore, it follows from Lemma~\ref{lemma:36-9-subgroups} that one of the~following cases holds:
\begin{itemize}
\item $Z$ is a union of $6$ lines, and $G_C\simeq\mathfrak{S}_3$;
\item $Z$ is a union of $2$ irreducible curves of degree $3$, and $G_C\simeq C_3\rtimes\mathfrak{S}_3$;
\item $Z$ is a union of $9$ lines, $G_C\simeq C_4$.
\end{itemize}
Let us deal with these cases one by one.

Suppose that $Z$ is a union of $6$ lines.
Since $G_{36,9}$ has two subgroups isomorphic to $\mathfrak{S}_3$ up to conjugation,
we may assume that $G_C=\langle M_1,M_3^2\rangle$  or $G_C=\langle M_1M_2,M_3^2\rangle$.
But $C\cap S$ is fixed~by~$G_C$.
Then $G_C=\langle M_1M_2,M_3^2\rangle$, because  $\langle M_1,M_3^2\rangle$ does not fix points in $X_{a}$. Moreover, since $G_C$ fixes the~point $S\cap C$, it cannot act faithfully on  $C$, so the subgroup $\langle M_1\rangle\simeq C_3$ pointwise~fixes~$C$. Then~$C$~is an~irreducible component of the~cubic curve $\{x_0=x_2=0\}\cap X_{a}$, which is impossible, since this cubic curve is irreducible. Therefore, we conclude that $Z$ is not a union of $6$ lines.

Now, we suppose that $Z$ is a union of $2$ irreducible curves of degree $3$.
Then $G_C=\langle M_1,M_2,M_3^2\rangle$.
If $C$ is a twisted cubic,  $G_C$ acts faithfully on it, which is impossible,
since $\mathrm{PGL}_2(\mathbb{C})$ does not contain subgroups isomorphic to $C_3\rtimes\mathfrak{S}_3$.
Thus,~$C$ is a plane cubic. But $\{x_0=x_2=0\}$ and $\{x_1=x_4=0\}$ are the~only $G_C$-invariant planes in $\mathbb{P}^4$, so $Z=\mathcal{R}_6$.

Finally,we  suppose that the curve $Z$ is a union of $9$ lines.
Since $G$ contains a unique subgroup isomorphic to $C_4$ up to conjugation, we may assume that $G_C=\langle M_3\rangle$.
Since $C\cap S$ is fixed by $G_C$, it follows from Lemma~\ref{lemma:36-9-orbits} that this point is one of the~points
$$
[1:-1:1:0:-1], [1:-i:-1:0:i], [1:i:-1:0:-i]
$$
But $G_C$ fixes at least two points in the~line $C$.
Therefore, arguing as in the~proof of Lemma~\ref{lemma:36-9-orbits}, we see that $C$ contains the~points $[1:1:1:t:1]$ for some $t\in\mathbb{C}$ such that $(1+b)t^3+2at+4=0$.
If~$C$~contains $[1:-1:1:0:-1]$, then we have $a^3=-27$, which is impossible by our assumption.
Thus, we see that $C$ contains one of the points $[1:\mp i:-1:0:\pm i]$.
Then $Z=\mathcal{L}_t^+$ or $Z=\mathcal{L}_t^-$.
\end{proof}

Recall that $\mathcal{L}_3$, $\mathcal{L}_3^\prime$, $\mathcal{C}_6$, $\mathcal{R}_6$ are $G_{36,10}$-irreducible curves. Set
\begin{align*}
\mathcal{R}_6^\prime&=\big(\{x_0=0,x_1=0\}\cup\{x_2=0,x_4=0\}\big)\cap X_{a},\\
\mathcal{R}_6^{\prime\prime}&=\big(\{x_0=0,x_4=0\}\cup\{x_1=0,x_2=0\}\big)\cap X_{a}.
\end{align*}
Then $\mathcal{R}_6^\prime$ and $\mathcal{R}_6^{\prime\prime}$ are $G_{36,10}$-irreducible curves of degree $6$, and each of them is a union of two disjoint smooth plane cubic curves. Note that $\mathcal{R}_6^\prime$ and $\mathcal{R}_6^{\prime\prime}$ are contained in $S_0$. 

Similarly, we set
\begin{align*}
\mathcal{R}_9&=\big(\{x_0=x_1,x_2=x_4\}\cup\{x_0=\zeta_3 x_1,x_2=\zeta_3 x_4\}\cup\{x_0=\zeta_3^2 x_1,x_2=\zeta_3^2 x_4\}\big)\cap X_{a},\\
\mathcal{R}_9^{\prime}&=\big(\{x_0=x_4,x_1=x_2\}\cup\{x_0=\zeta_3 x_4,x_1=\zeta_3 x_2\}\cup\{x_0=\zeta_3^2 x_4,x_1=\zeta_3^2 x_2\}\big)\cap X_{a}.
\end{align*}
Then $\mathcal{R}_9$ and $\mathcal{R}_9^{\prime}$ are $G_{36,10}$-irreducible curves. If $X_a$ is smooth ($a^3\ne-\frac{27}{2}$), then $\mathcal{R}_9^\prime$ and $\mathcal{R}_9^{\prime}$ are unions of $3$ disjoint smooth plane cubic curves. On the other hand, if $X_a$ is singular ($a^3=-\frac{27}{2}$), then $\mathcal{R}_9$ and $\mathcal{R}_9^{\prime}$ are unions of $9$ lines, which are unions of three  disjoint reducible cubic curves consisting of three lines. For instance, if $a=-\frac{3}{\sqrt[3]{2}}$, an irreducible component of  $\mathcal{R}_9$ is the curve
$$
\big\{2x_0+2x_2+\sqrt[3]{4}x_3=0, 2x_1+2x_4+\sqrt[3]{4}x_3=0, x_2=x_4\big\}\subset\Pi_{0,0}.
$$
and an irreducible component of  $\mathcal{R}_9^{\prime}$ is the curve
$$
\big\{2x_0+2x_2+\sqrt[3]{4}x_3=0, 2x_1+2x_4+\sqrt[3]{4}x_3=0, x_0=x_4\big\}\subset\Pi_{0,0}.
$$
Moreover, if the cubic threefold $X_a$ is singular, then the curves $\mathcal{R}_9$ and $\mathcal{R}_9^\prime$ contains the singular locus of the cubic $X_a$. In fact, in this case, we have
$$
\mathrm{Sing}(\mathcal{R}_9)=\mathrm{Sing}(\mathcal{R}_9^{\prime})=\mathrm{Sing}(X_a).
$$
Note that $\mathcal{R}_9$ and $\mathcal{R}_9^{\prime}$ are not contained in any surface in $\mathcal{P}$.

Finally, let $\mathcal{K}_6$ be the union of the following three curves:
\begin{align*}
\big\{x_0+x_1=0, x_2+x_4&=0, 2ax_1x_4+x_3^2=0\big\},\\
\big\{x_0+\zeta_3 x_1=0, x_2+\zeta_3^2x_4&=0, 2ax_1x_4+(\zeta_3+1)x_3^2=0\big\},\\
\big\{x_0+\zeta_3^2x_1=0, x_2+\zeta_3x_4&=0, 2ax_1x_4+(\zeta_3^2+1)x_3^2=0\big\}.
\end{align*}
and let $\mathcal{K}_6^{\prime}$ be the union of the following three curves:
\begin{align*}
\big\{x_0+x_4=0, x_1+x_2&=0, 2ax_2x_4-x_3^2=0\big\},\\
\big\{x_0+\zeta_3x_4=0, x_1+\zeta_3x_2&=0, 2ax_2x_4+(\zeta_3+1)x_3^2=0\big\},\\
\big\{x_0+\zeta_3^2x_4=0, x_1+\zeta_3^2x_2&=0, 2ax_2x_4+(\zeta_3+1)x_3^2=0\big\}.
\end{align*}
If $a\ne 0$, then $\mathcal{K}_6$ and $\mathcal{K}_6^\prime$ are
$G_{36,10}$-irreducible curves of degree $6$ whose irreducible components are disjoint (smooth) conics.
In this case, we have
\begin{equation}
\label{equation:36-10-six-conics}
\mathcal{K}_6+\mathcal{K}_6^\prime=\big\{x_0x_2-x_1x_4=0\big\}\cap S_{\lambda},
\end{equation}
where $\lambda=\frac{1}{a}$, and $S_\lambda$ is smooth by Corollary~\ref{corollary:36-9-Du-Val}.
If $a=0$, then $\mathcal{K}_6=2\mathcal{L}_3$ and $\mathcal{K}_6^\prime=2\mathcal{L}_3^\prime$.

\begin{lemma}
\label{lemma:36-10-reducible-curves}
Let $Z$ be a reducible $G_{36,10}$-irreducible curve in $X_{a}$ of degree $d<12$ such that $Z\not\subset S$.
Then one of the following two cases holds:
\begin{itemize}
\item $d=9$, and $Z$ is one of the curves $\mathcal{R}_9$ or $\mathcal{R}_9^\prime$;
\item $d=6$, and $Z$ is one of the curves $\mathcal{R}_6$, $\mathcal{R}_6^\prime$ or $\mathcal{R}_6^{\prime\prime}$;
\item $d=6$, $a\ne 0$, and $Z$ is one of the curves $\mathcal{K}_6$ or $\mathcal{K}_6^{\prime}$.
\end{itemize}
\end{lemma}

\begin{proof}
The intersection $S\cap Z$ consists of $d<12$ points (counted with multiplicities), and this intersection is a union of $G_{36,10}$-orbits. Hence, it follows from Lemma~\ref{lemma:36-10-orbits} that $d=6$ or $9$.

Let $C$ be an irreducible component of the~curve $Z$, and let $G_C$ be its stabilizer in the group~$G_{36,10}$.
Then one of the~following holds:
\begin{itemize}
\item $Z$ is a union of $2$ plane cubic curves;
\item $Z$ is a union of $2$ twisted cubic curves;
\item $Z$ is a union of $3$ conics;
\item $Z$ is a union of $6$ lines;
\item $Z$ is a union of $3$ plane cubic curves;
\item $Z$ is a union of $3$ twisted cubic curves;
\item $Z$ is a union of $9$ lines.
\end{itemize}
Let us deal with these cases one by one.

Suppose that $Z$ is a union of $2$ irreducible curves of degree $3$.
Then it follows from Lemma~\ref{lemma:36-10-subgroups} that $G_C$ is a normal subgroup of $G$, and either $G_C\simeq C_3\times\mathfrak{S}_3$ or $G_C\simeq C_3\rtimes\mathfrak{S}_3$. In each case, the~restriction of the representation $\mathbb{V}_4$ to $G_C$ splits as a sum of two non-isomorphic two-dimensional irreducible representations of $G_C$, which can be verified by the following Magma code:
\begin{verbatim}
    G:=SmallGroup(36,10);
    K:=CyclotomicField(36);
    L:=IrreducibleModules(G,K);
    R:=L[9];
    S:=Subgroups(G);
    H:=S[21]`subgroup;
    print(IdentifyGroup(H));
    W:=Restriction(R,H);
    print(W);
    ID:=Decomposition(W);
    print(ID);
    IsIsomorphic(ID[1],ID[2]);
\end{verbatim}
Hence, $C$ is not a twisted cubic curve, because otherwise $C$ would be contained in $S$, which is not the case by assumption. Hence, $C$ is a plane cubic. If $G_C\simeq C_3\rtimes\mathfrak{S}_3$, then $G_C=\langle M_1,M_2,M_4\rangle$, and it follows from the proof of Lemma~\ref{lemma:36-9-reducible-curves} that we have $Z=\mathcal{R}_6$. Similarly, if $G_C\simeq C_3\times\mathfrak{S}_3$, then either $G_C=\langle M_1,M_2,M_5\rangle$ or $G_C=\langle M_1,M_2,M_4M_5\rangle$, so $C$ is one of the following curves:
\begin{align*}
\{x_0=0,x_1=0\}\cap X_{a},\\
\{x_2=0,x_4=0\}\cap X_{a},\\
\{x_0=0,x_4=0\}\cap X_{a},\\
\{x_1=0,x_2=0\}\cap X_{a}.
\end{align*}
Hence, in this case, either $Z=\mathcal{R}_6^\prime$ or $Z=\mathcal{R}_6^{\prime\prime}$.

Suppose that $Z$ is a union of $3$ irreducible conics. Then it follows from Lemma~\ref{lemma:36-10-subgroups} that $G_C\simeq\mathfrak{D}_6$. Up to conjugation, $G$ contains two subgroups isomorphic to $\mathfrak{D}_6$. In each case, the restriction of the representation $\mathbb{V}_4$ to $G_C$ splits as a sum of two non-isomorphic two-dimensional irreducible representations of the group $G_C$. Let $\Pi_C$ be the plane in $\mathbb{P}^4$ spanned by $C$. Then we have exactly two choices for the plane $\Pi_C$, and the plane $\Pi_C$ contains the $G_{36,10}$-fixed point $[0:0:0:1:0]\not\in X_a$, which implies that $\Pi_C\not\subset X_a$. Then
$$
X_{a}\cap\Pi_C=C+L,
$$
where $L$ is a line in $X_{a}$. Hence, it follows from Lemma~\ref{lemma:36-10-S-Pic} that $L$ is an irreducible component of one of the curves $\mathcal{L}_3$ or $\mathcal{L}_3^\prime$. This implies that $\Pi_C$ is one of the planes
\begin{align*}
\{x_0+x_1=0, x_2+x_4=0\},\\
\{x_0+\zeta_3 x_1=0, x_2+\zeta_3^2x_4=0\},\\
\{x_0+\zeta_3^2x_1=0, =x_2+\zeta_3x_4=0\},\\
\{x_0+x_4=0, =x_1+x_2=0\},\\
\{x_0+\zeta_3x_4=0, =x_1+\zeta_3x_2=0\},\\
\{x_0+\zeta_3^2x_4=0, =x_1+\zeta_3^2x_2=0\}.
\end{align*}
Intersecting these planes with $X_{a}$ and using the fact that $C$ is a smooth conic, we see that $a=0$, and either $Z=\mathcal{K}_6$ or $Z=\mathcal{K}_6^\prime$.

Suppose that $Z$ is a union of $6$ lines. Then $C\cap S$ is a $G_C$-fixed point, which implies that $G_C$ fixes a point $O\in C$ such that $O$ is not contained in $S$, since otherwise we would have $Z\subset S$, which contradicts our assumption. On the other hand, it follows from Lemma~\ref{lemma:36-10-subgroups} that 
\begin{itemize}
\item either $G_C\simeq C_6$, 
\item or $G_C\simeq\mathfrak{S}_3$. 
\end{itemize}
If $G_C\simeq C_6$, $S$ contains all $G_C$-fixed points  by Lemma~\ref{lemma:36-10-orbits}, which is a contradiction, since $O\not\in S$. Thus, $G_C\simeq \mathfrak{S}_3$. By Lemma~\ref{lemma:36-10-subgroups}, $G_{36,10}$ has five subgroups isomorphic to $\mathfrak{S}_3$ up to conjugation. But only one of them fixes a point in $X_{a}$. Hence, we may assume that $G_C=\langle M_1,M_4\rangle$. Then 
$$
C\cap S=[0:1:0:0:-1],
$$ 
and $O=[0:1:0:s:1]$, where $s\in\mathbb{C}$ such that $s^3+as+2=0$. But the line in $\mathbb{P}^4$ that passes through these points is not contained in $X_{a}$. Therefore, $Z$ is not a union of $6$ lines.

Now, we suppose that $Z$ is a union of $3$ irreducible cubic curves. Then $G_C\simeq\mathfrak{D}_6$ by Lemma~\ref{lemma:36-10-subgroups}. As we already mentioned above, the restriction of the representation $\mathbb{V}_4$ to $G_C$ splits as a sum of two non-isomorphic two-dimensional irreducible representations. Hence, $C$ is not a twisted cubic curve, because $C\not\subset S$. Thus, $C$ is a plane cubic. As above, we denote by $\Pi_C$ the plane in $\mathbb{P}^4$ that contains $C$. Then $\Pi_C$ is $G_C$-invariant. If $X_a$ is singular, $\Pi_C\not\subset X_a$, because
$$
\Pi_C\ne\Pi_{i,j}
$$
for any $i,j\in\{0,1,2\}$, since $G$ acts transitively on the planes $\Pi_{i,j}$. So, since $G$ contains $6$ subgroups isomorphic to $\mathfrak{D}_6$ by Lemma~\ref{lemma:36-10-subgroups}, we have exactly $12$ possibilities for the plane $\Pi_C$, and $6$ of them are already listed above --- they intersect $X_{a}$ by a reducible cubic consisting of a line an a conic. The remaining $6$ planes are
\begin{align*}
\{x_0=x_1&,  x_2=x_4\},\\
\{x_0=\zeta_3 x_1&,   x_2=\zeta_3^2x_4\},\\
\{x_0=\zeta_3^2x_1&,   x_2=\zeta_3x_4\},\\
\{x_0=x_4&,   x_1=x_2\},\\
\{x_0=\zeta_3x_4&,  x_1=\zeta_3x_2\},\\
\{x_0=\zeta_3^2x_4&,  x_1=\zeta_3^2x_2\}.
\end{align*}
They intersect $X_{a}$ in irreducible components of $\mathcal{R}_9+\mathcal{R}_9^\prime$, so either $Z=\mathcal{R}_9$ or $Z=\mathcal{R}_9^\prime$, which implies that $X_a$ is smooth in this case, since otherwise $\mathcal{R}_9$ and $\mathcal{R}_9^\prime$ would be consisting of $9$ lines.

Therefore, we may assume that $Z$ is a union of $9$ lines, so we have $G_C\simeq C_2\times C_2$ by Lemma~\ref{lemma:36-10-subgroups}, and~$G$ has one subgroup isomorphic to $C_2^2$ up to conjugation. We may assume that $G_C=\langle M_4,M_5\rangle$. Note that $C\cap S$ is a $G_C$-fixed point. Thus, this point is one of the~points
$$
[1:-1:1:0:-1], [1:1:-1:0:-1], [1:-1:-1:0:1]
$$
On the~other hand, the~group $G_C$ fixes at least two points in $C$.
Thus, by Lemma~\ref{lemma:36-10-orbits}, we have
$$
[1:1:1:t:1]\in C
$$
for some $t\in\mathbb{C}$ such that $t^3+2at+4=0$.
If $a^3\ne -\frac{27}{2}$, then $X_a$ does not contain the line in $\mathbb{P}^4$ passing through the point $[1:1:1:t:1]$ and any of the points 
\begin{center}
$[1:-1:1:0:-1]$, $[1:1:-1:0:-1]$, $[1:-1:-1:0:1]$.
\end{center}
Thus, $a^3=-\frac{27}{2}$, so $X_a$ is singular. Then 
\begin{itemize}
\item either $t=-\frac{3}{a}$ and $[1:1:1:t:1]$ is a singular point of $X_a$, 
\item or $t=-\frac{6}{a}$ and $[1:1:1:t:1]$ is a smooth point of $X_a$. 
\end{itemize}
This implies that either $C$ is the line passing through $[a:a:a:6:a]$ and $[1:1:-1:0:-1]$, or~$C$~is~the~line passing through  $[a:a:a:6:a]$ and $[1:-1:-1:0:1]$, so $Z=\mathcal{R}_9$ or $Z=\mathcal{R}_9^\prime$. 
\end{proof}

Now, we are ready to describe irreducible $G$-invariant curves of degree $<12$ in the cubic~$X_{a}$, which are not contained in the cubic surface $S$. If $G=G_{36,10}$, such curves do not exists:

\begin{lemma}
\label{lemma:36-9-irreducible-curves}
Let $C$ be an irreducible $G$-invariant curve in $X_{a}$.
Suppose that $d<12$ and $C\not\subset S$. Then $G=G_{36,9}$, $d=9$, the~curve $C$ is smooth, and its genus is either $1$ or $4$. Moreover, we have
$$
\mathrm{Sing}(S_\lambda)=\Sigma_{9}^t\subset C\subset S_\lambda
$$
where $t\in\mathbb{C}$ such that $t^3+2at+4=0$, and $\lambda=-\frac{2}{t^2}$.
\end{lemma}

\begin{proof}
Since $C$ is not contained in $S$, it is not contained in any hyperplane in $\mathbb{P}^4$,
which implies that $G$ acts faithfully on $C$. Let $\widehat{C}\to C$ be the~normalization, and let $g$ be the~genus of $\widehat{C}$. Then~the~$G$-action lifts to $\widehat{C}$. By Lemma~\ref{lemma:36-9-orbits}, if $G=G_{36,9}$, the length of a $G$-orbit in $\widehat{C}$ is $9$, $12$, $18$ or $36$. Similarly, by Lemma~\ref{lemma:36-10-orbits}, if $G=G_{36,10}$, the length of a $G$-orbit in $\widehat{C}$ is $6$, $12$, $18$ or $36$.

As in the~proof of Lemmas~\ref{lemma:36-9-reducible-curves} and \ref{lemma:36-10-reducible-curves}, we see that $S\cap C$ consists of $d<12$ points (counted with multiplicities). On the~other hand, the intersection $S\cap C$ is a union of $G$-orbits.
Hence, it follows from Lemmas~\ref{lemma:36-9-orbits} and \ref{lemma:36-10-orbits} that $d=6$ or $9$, and $S\cap C$ is a $G$-orbit of length $6$ or $9$, respectively. Moreover,  $S$ intersects $C$ transversally at each point of $S\cap C$, which implies that $C$ is smooth at every point of the~intersection $S\cap C$. In particular, if $d=6$, then $G=G_{36,10}$, because $\widehat{C}$ does not contain $G$-orbits of length $6$ in the case when $G=G_{36,9}$.
Similarly, if $d=9$, then $G=G_{36,9}$, because $\widehat{C}$ does not contain $G$-orbits of length $9$ in the case when $G=G_{36,10}$.

Let $P$ be a general point in $C$, and let $\Sigma$ be the~$G$-orbit of $P$. Then $|\Sigma|=36$, and there is $\lambda\in\mathbb{C}$ such that $P\in S_\lambda$. Moreover, $C\subset S_\lambda$, since otherwise we would have
$$
18\geqslant 2d=S_\lambda\cdot C\geqslant |S_\lambda\cap C|\geqslant |\Sigma|=36.
$$
By Corollary~\ref{corollary:36-9-Du-Val}, either the surface $S_\lambda$ is smooth, or  $\mathrm{Sing}(S_\lambda)$ is a $G$-orbit of length $6$ or $9$, and the singularities of $S_\lambda$ are nodes. Set $H_{S_{\lambda}}=H\vert_{S_{\lambda}}$. Then it follows from the Hodge index theorem that
$$
6C^2-d^2=\mathrm{det}\begin{pmatrix}
C^2 &d \\
d & 6
\end{pmatrix}=\mathrm{det}\begin{pmatrix}
C^2 &H_{S_{\lambda}}\cdot C \\
H_{S_{\lambda}}\cdot C & H_{S_{\lambda}}^2
\end{pmatrix}\leqslant 0.
$$
This gives $C^2\leqslant\frac{d^2}{6}$. Thus, if $d=9$, then $C^2\leqslant\frac{27}{2}$.
Similarly, if $d=6$, then $C^2\leqslant 6$.

Suppose that $C$ is contained in the~smooth locus of the surface $S_\lambda$.
Then, by adjunction formula, we have $C^2=2\mathrm{p}_a(C)-2$, where $\mathrm{p}_a(C)$ is the~arithmetic genus of the~curve $C$.
Thus, we have
$$
2g-2\leqslant 2\mathrm{p}_a(C)-2=C^2\leqslant
\left\{\aligned
&12\ \text{if $d=9$},\\
&6\ \text{if $d=6$}.
\endaligned
\right.
$$
Therefore, if $d=9$, then $g\leqslant \mathrm{p}_a(C)\leqslant 7$, so $g\in\{1,4\}$ by Lemma~\ref{lemma:36-9-curves}. If $d=6$, then $g\leqslant \mathrm{p}_a(C)\leqslant 4$, so it follows from Lemma~\ref{lemma:36-10-curves} that $g=\mathrm{p}_a(C)=4$. On the other hand, if the curve $C$ is singular, then $\mathrm{Sing}(C)$ is a union of $G$-orbits $\Sigma_1,\ldots\Sigma_r$, so
$$
g+\sum_{i=1}^rn_i|\Sigma_i|=\mathrm{p}_a(C)\leqslant\left\{\aligned
&7\ \text{if $d=9$},\\
&4\ \text{if $d=6$}.
\endaligned
\right.
$$
for some positive integers $n_1,\ldots,n_r$, which contradicts Lemmas~\ref{lemma:36-9-orbits} and \ref{lemma:36-10-orbits}.
Hence, we conclude that the curve $C$ is smooth, and one of the following three cases holds:
\begin{enumerate}
\item $G=G_{36,9}$, $d=9$ and $g=1$;
\item $G=G_{36,9}$, $d=9$ and $g=7$;
\item $G=G_{36,10}$, $d=6$ and $g=4$.
\end{enumerate}
If $G=G_{36,10}$, $d=6$ and $g=4$, then it follows from Lemma~\ref{lemma:36-9-curves} that $C$ has two $G$-orbits of length~$6$, but one of them is the intersection $C\cap S$, so the set $C\setminus(S\cap C)$ contains a $G$-orbit of length $6$, and the stabilizer in $G$ of a point in this orbit is isomorphic to $C_4$ (because it acts faithfully on the~tangent space to $C$ at this point), which contradicts Lemma~\ref{lemma:36-10-orbits}. Then $G=G_{36,9}$, $d=9$ and
\begin{enumerate}
\item either $g=1$;
\item or $g=7$.
\end{enumerate}
In both cases, $S\cap C$ is a $G$-orbit of length $9$, and $C$ has two $G$-orbits of length $9$ by Lemma~\ref{lemma:36-9-curves}. Hence, by Lemma~\ref{lemma:36-9-orbits}, the curve $C$ contains the~$G$-orbit $\Sigma_9^t$ for some $t\in\mathbb{C}$ such that $t^3+2at+4=0$, so it follows from Corollary~\ref{corollary:36-9-Du-Val} that $\lambda=-\frac{2}{t^2}$, and $\mathrm{Sing}(S_\lambda)=\Sigma_9^t$, which contradicts our assumption.

Hence, the surface $S_\lambda$ is singular, and $C$ contains some of its singular points.
By Corollary~\ref{corollary:36-9-Du-Val}, the surface $S_\lambda$ has isolated ordinary double points, and $\mathrm{Sing}(S_\lambda)$ is a $G$-orbit of length $6$ or $9$, so, in particular, the curve $C$ contains all singular points of the surface $S$.

Let $\widetilde{S}_\lambda\to S_\lambda$ be the~minimal resolution, let $\widetilde{C}$ be the~strict transform on $\widetilde{S}_\lambda$ of the~curve $C$, and let $\widetilde{C}_6$ be the~strict transform on $\widetilde{S}_\lambda$ of the~curve $C_6$. Then the~$G$-action lifts to $\widetilde{S}_\lambda$, and the~intersection $\widetilde{C}\cap\widetilde{C}_6$ is a $G$-orbit of length $d\in\{6,9\}$. Moreover, by Lemma~\ref{lemma:DuVal} and adjunction formula, we have
$$
2g-2\leqslant 2\mathrm{p}_a(\widetilde{C})-2=\widetilde{C}^2\leqslant C^2-\frac{|\mathrm{Sing}(S_\lambda)|}{2}\leqslant \frac{d^2}{6}-\frac{|\mathrm{Sing}(S_\lambda)|}{2}\leqslant\frac{27-|\mathrm{Sing}(S_\lambda)|}{2}\leqslant\frac{21}{2},
$$
where $\mathrm{p}_a(\widetilde{C})$ is the~arithmetic genus of the~curve $\widetilde{C}$.
Therefore, if $d=6$, then we must have $g<4$, which contradicts Lemma~\ref{lemma:36-10-curves}, since $G=G_{36,10}$ in this case.
Hence, $d=9$ and $G=G_{36,9}$. Then
$$
2g-2\leqslant 2\mathrm{p}_a(\widetilde{C})-2=\widetilde{C}^2\leqslant 10,
$$
Then $\mathrm{p}_a(\widetilde{C})\leqslant 6$. On the other hand, we know that $g\ne 0$ by Lemma~\ref{lemma:36-9-curves}.
Thus, arguing as above, we see that the curve $\widetilde{C}$ is smooth, and either $g=1$ or $g=4$.

By Lemma~\ref{lemma:36-9-curves}, $\widetilde{C}$ contains a $G$-orbit of length $9$ that is different from  $\widetilde{C}\cap\widetilde{C}_6$. Thus, it follows from Lemma~\ref{lemma:36-9-orbits} that the~image of this $G$-orbit in $C$ is a $G$-orbit of length $9$ that is not contained in the cubic surface $S$. Hence, it follows from Lemma~\ref{lemma:36-9-orbits} and Corollary~\ref{corollary:36-9-Du-Val} that
$$
\mathrm{Sing}(S_\lambda)=\Sigma_9^t\subset C\subset S_\lambda
$$
for some $t\in\mathbb{C}$ such that $t^3+2at+4=0$, as claimed.

To complete the~proof of the~lemma, we have to show that $C$ is smooth.
If $C$ is singular, then it follows from Lemma~\ref{lemma:DuVal} that
$$
0\leqslant 2g-2=\widetilde{C}^2\leqslant C^2-2|\mathrm{Sing}(S_\lambda)|=C^2-18\leqslant\frac{27}{2}-18<-2,
$$
which is absurd. Hence, we conclude that $C$ is smooth.
\end{proof}

Now, we are ready to prove the~main technical result of this section.

\begin{proposition}
\label{proposition:36-9-curves}
Let $Z$ be a $G$-irreducible curve in $X_{a}$,
and let $\mathcal{M}$ be a non-empty $G$-invariant mobile linear subsystem in $|nH|$ for some $n\in\mathbb{Z}_{>0}$. Then
$$
\mathrm{mult}_Z\big(\mathcal{M}\big)\leqslant\frac{n}{2}.
$$
\end{proposition}

\begin{proof}
Suppose that $\mathrm{mult}_Z(\mathcal{M})>\frac{n}{2}$. Let $M$ and $M^\prime$ be general surfaces in $\mathcal{M}$. Then
$$
\mathrm{mult}_Z\big(M\big)=\mathrm{mult}_Z\big(M^\prime\big)=\mathrm{mult}_Z\big(\mathcal{M}\big)>\frac{n}{2}.
$$
Then, in particular, we have
$$
M\cdot M^\prime=\mathrm{mult}^2_Z\big(\mathcal{M}\big)Z+\Delta,
$$
where $\Delta$ is an effective one-cycle (whose support may, a priori, contain $Z$). Then
$$
3n^2=H\cdot M\cdot M^\prime=\mathrm{mult}^2_Z\big(\mathcal{M}\big)\mathrm{deg}(Z)+H\cdot\Delta\geqslant\mathrm{mult}^2_Z\big(\mathcal{M}\big)\mathrm{deg}(Z)>\frac{n^2}{4}\mathrm{deg}(Z),
$$
where $H$ is a general hyperplane section of the~cubic $X_{a}\subset\mathbb{P}^4$. Therefore, we conclude $\mathrm{deg}(Z)<12$. Let us seek for a contradiction.

If $Z\subset S$, then the log pair $(S,\frac{2}{n}\mathcal{M}\vert_{S})$ is not log canonical along the curve $Z$, so $\alpha_G(S)<2$, because $\mathcal{M}\vert_{S}\sim n(-K_S)$, and  $\mathcal{M}\vert_{S}$ is $G$-invariant. However, $\alpha_G(S)=2$ by Corollaries~\ref{corollary:36-9-alpha} and \ref{corollary:36-10-alpha}.

Hence, it follows from Lemmas~\ref{lemma:36-9-reducible-curves}, \ref{lemma:36-10-reducible-curves} and \ref{lemma:36-9-irreducible-curves} that one of the following seven cases holds:
\begin{enumerate}
\item[$(\mathrm{i})$] $Z=\mathcal{R}_6$;
\item[$(\mathrm{ii})$] $G=G_{36,10}$, and $Z$ is one of the curves $\mathcal{R}_6^\prime$ or $\mathcal{R}_6^{\prime\prime}$;
\item[$(\mathrm{iii})$] $G=G_{36,10}$, and $Z$ is one of the curves $\mathcal{R}_9$ or $\mathcal{R}_9^\prime$;
\item[$(\mathrm{iv})$] $G=G_{36,10}$, $a\ne 0$, and $Z$ is one of the curves $\mathcal{K}_6$ or $\mathcal{K}_6^{\prime}$;
\item[$(\mathrm{v})$] $G=G_{36,9}$, and $Z=\mathcal{L}_t^{+}$ or $Z=\mathcal{L}_t^{-}$ for some $t\in\mathbb{C}$ such that $t^3+2at+4=0$;
\item[$(\mathrm{vi})$] $G=G_{36,9}$, and $Z$ is an irreducible smooth curve of genus $1$ and degree $9$;
\item[$(\mathrm{vii})$] $G=G_{36,9}$, and $Z$ is an irreducible smooth curve of genus $4$ and degree $9$.
\end{enumerate}
Moreover, $Z$ is smooth unless $X_a$ is singular, $G=G_{36,10}$, and $Z$ is one of the curves $\mathcal{R}_9$ or $\mathcal{R}_9^\prime$.

Let us exclude cases $(\mathrm{i})$, $(\mathrm{ii})$, $(\mathrm{iii})$. Suppose that $Z$ is one of the curves $\mathcal{R}_6$, $\mathcal{R}_6^\prime$, $\mathcal{R}_6^{\prime\prime}$, $\mathcal{R}_9$, $\mathcal{R}_9^\prime$. Then $Z$ is a disjoint union of plane cubic curves (if $X_a$ is singular, and $Z=\mathcal{R}_9$ or $Z=\mathcal{R}_9^\prime$, these cubic curves are reducible). Let $C$ be one of these cubic curves, and let $S_C$ be a general hyperplane section of the cubic $X_a$ that contains $C$. Then $S_C$ is a normal cubic surface, which is actually smooth unless $X_a$ is singular, and $Z=\mathcal{R}_9$ or $Z=\mathcal{R}_9^\prime$. If $X_a$ is singular, and $Z=\mathcal{R}_9$ or $Z=\mathcal{R}_9^\prime$, then $S_C$ has three ordinary double points, which are contained in $C$. Furthermore, we have
$$
n(-K_{S_C})\sim_{\mathbb{Q}}\mathcal{M}\big\vert_{S_C}=\mathcal{M}_{S_C}+\mathrm{mult}_Z\big(\mathcal{M}\big)C,
$$
where $\mathcal{M}_{S_C}$ is a non-empty mobile linear system on the cubic surface $S_C$. Note that $C\sim -K_{S_C}$, the intersection $S_C\cap (Z\setminus C)$ is contained in the smooth locus of the surface $S_C$, and this intersection consists of $\mathrm{deg}(Z)-3\geqslant 3$ distinct points, which are not contained in $C$. Hence, we have
$$
\mathrm{mult}_O\big(\mathcal{M}_{S_C}\big)>\frac{n}{2}
$$
for every point $O\in S_C\cap (Z\setminus C)$. Thus, if $M_{S_C}$ and $M_{S_C}^\prime$ are general curves in $\mathcal{M}_{S_C}$, then
\begin{multline*}
\frac{3n^2}{4}=\Big(n-\frac{n}{2}\Big)^2\big(-K_{S_C}\big)^2>\Big(n-\mathrm{mult}_Z\big(\mathcal{M}\big)\Big)^2\big(-K_{S_C}\big)^2=\\
=M_{S_C}\cdot M_{S_C}^\prime\geqslant\sum_{O\in S_C\cap (Z\setminus C)}\Big(M_{S_C}\cdot M_{S_C}^\prime\Big)_O\geqslant
\sum_{O\in S_C\cap (Z\setminus C)}\mathrm{mult}_O\big(M_{S_C}^\prime\big)\mathrm{mult}_O\big(M_{S_C}^\prime\big)=\\
=\sum_{O\in S_C\cap (Z\setminus C)}\mathrm{mult}^2_O\big(\mathcal{M}_{S_C}\big)>
\sum_{O\in S_C\cap (Z\setminus C)}\Big(\frac{n}{2}\Big)^2=\big|S_C\cap (Z\setminus C)\big|\frac{n^2}{4}\geqslant\frac{3n^2}{4},
\end{multline*}
which is a contradiction. Thus, we see that $Z$ is one of the curves $\mathcal{R}_6$, $\mathcal{R}_6^\prime$, $\mathcal{R}_6^{\prime\prime}$, $\mathcal{R}_9$, $\mathcal{R}_9^\prime$.

Let us present another way how to exclude cases $(\mathrm{i})$, $(\mathrm{ii})$, $(\mathrm{iii})$ in the case when $X_a$ is smooth, which exclude cases $(\mathrm{i})$, $(\mathrm{ii})$, $(\mathrm{iv})$ unconditionally. Namely, suppose that the curve $Z$ is smooth, and $Z\cap \mathrm{Sing}(X_a)=\varnothing$.
Let $\pi\colon \widetilde{X}_{a}\to X_{a}$ be the~blow up~of~$Z$, let $F$ be the $\pi$-exceptional divisor, let~$\widetilde{M}$ and $\widetilde{M}^\prime$ be the strict transforms on $\widetilde{X}_{a}$ of  $M$ and $M^\prime$, respectively. Set $\widetilde{H}=\pi^*(H)$. Then
$$
\widetilde{M}\sim \widetilde{M}^\prime\sim n\widetilde{H}-\mathrm{mult}_Z\big(\mathcal{M}\big)F.
$$
Note that $\widetilde{M}\cdot\widetilde{M}^\prime$ is an effective one-cycle.
Take $k\in\mathbb{Z}_{>0}$ such that $k\widetilde{H}-F$ is nef. Then
\begin{equation}
\label{equation:36-9-test-class}
0\leqslant \widetilde{M}\cdot \widetilde{M}^\prime\cdot \big(k\widetilde{H}-F\big)=3kn^2-2n\mathrm{deg}(Z)\mathrm{mult}_Z\big(\mathcal{M}\big)-(k\mathrm{deg}(Z)+F^3)\mathrm{mult}^2_Z\big(\mathcal{M}\big).
\end{equation}
Note that
$$
F^3=\left\{\aligned
&-12\ \text{if $Z=\mathcal{R}_6$, $Z=\mathcal{R}_6^\prime$ or $Z=\mathcal{R}_6^{\prime\prime}$},\\
&-18\ \text{if $Z=\mathcal{R}_9$ or $Z=\mathcal{R}_9^\prime$},\\
&-6\ \text{if $a=0$, and either $Z=\mathcal{K}_6$ or $Z=\mathcal{K}_6^{\prime}$},\\
&0\ \text{if $Z=\mathcal{L}_t^{+}$ or $Z=\mathcal{L}_t^{-}$},\\
&-18\ \text{if $Z$ is a curve of degree $9$ and genus $1$},\\
&-12\ \text{if $Z$ is a curve of degree $9$ and genus $4$}.
\endaligned
\right.
$$
If $Z$ is one of the curves $\mathcal{R}_6$  $\mathcal{R}_6^\prime$ or $\mathcal{R}_6^{\prime\prime}$, then the linear system $|2\widetilde{H}-F|$ is base point free and gives an elliptic fibration $\widetilde{X}_{a}\to\mathbb{P}^1\times \mathbb{P}^1$, so, in particular, we can set $k=2$. Then \eqref{equation:36-9-test-class} gives
$$
0\leqslant \widetilde{M}\cdot \widetilde{M}^\prime\cdot \big(2\widetilde{H}-F\big)=6n^2-12n\mathrm{mult}_Z\big(\mathcal{M}\big),
$$
which gives $\mathrm{mult}_Z(\mathcal{M})\leqslant\frac{n}{2}$, which contradicts our assumption. Similarly, if $Z=\mathcal{R}_9$ or $Z=\mathcal{R}_9^\prime$, the linear system $|3\widetilde{H}-F|$ is base point free, so we can set $k=3$. In this case,  \eqref{equation:36-9-test-class} gives
$$
0\leqslant \widetilde{M}\cdot \widetilde{M}^\prime\cdot \big(3\widetilde{H}-F\big)=9n^2-18n\mathrm{mult}_Z\big(\mathcal{M}\big),
$$
so $\mathrm{mult}_Z(\mathcal{M})\leqslant\frac{n}{2}$, which contradicts our assumption. Likewise, if $a=0$, and $Z=\mathcal{K}_6$ or $Z=\mathcal{K}_6^\prime$, then the divisor $3\widetilde{H}-F$ is nef, so applying \eqref{equation:36-9-test-class} with $k=3$, we get
$$
0\leqslant \widetilde{M}\cdot \widetilde{M}^\prime\cdot \big(3\widetilde{H}-F\big)=
3\Big(3n+2\mathrm{mult}_Z\big(\mathcal{M}\big)\Big)\Big(n-2\mathrm{mult}_Z\big(\mathcal{M}\big)\Big),
$$
which gives $\mathrm{mult}_Z(\mathcal{M})\leqslant\frac{n}{2}$ that contradicts our assumption.

Summarizing, we see that cases $(\mathrm{i})$, $(\mathrm{ii})$, $(\mathrm{iv})$  are impossible, because in these cases $Z$ is always smooth, and $Z$ is always contained in the smooth locus of the cubic $X_a$. Similarly, we again see that case $(\mathrm{iii})$ is impossible if $X_a$ is smooth. So, we conclude that $G=G_{36,9}$, the curve $Z$ is smooth, and one of the following cases holds:
\begin{enumerate}
\item[$(\mathrm{v})$] $Z=\mathcal{L}_t^{+}$ or $Z=\mathcal{L}_t^{-}$ for $t\in\mathbb{C}$ such that $t^3+2at+4=0$;
\item[$(\mathrm{vi})$] $Z$ is an irreducible curve of degree $9$ and genus $1$;
\item[$(\mathrm{vii})$] $Z$ is an irreducible curve of degree $9$ and genus $4$.
\end{enumerate}
Then it follows from Lemmas~\ref{lemma:36-9-reducible-curves} and \ref{lemma:36-9-irreducible-curves} that
$$
\mathrm{Sing}(S_\lambda)=\Sigma_{9}^t\subset Z\subset S_\lambda
$$
for $\lambda=-\frac{2}{t^2}$, where $t$ is a non-zero number such that $t^3+2at+4=0$. Furthermore, by Corollary~\ref{corollary:36-9-Du-Val}, the~singularities of the~surface $S_\lambda$ are (ordinary) nodes. Let $\varpi\colon \widetilde{S}_\lambda\to S_{\lambda}$ be the~minimal~resolution, let $E$ be the~$\varpi$-exceptional divisor (consisting of $9$ disjoint $(-2)$-curves), and let $\widetilde{Z}$ be the~strict transform on $\widetilde{S}_\lambda$ of the~curve $Z$. Then $\widetilde{S}_\lambda$ is a smooth K3 surface, the~$G$-action lifts to $\widetilde{S}_\lambda$, and
$$
\widetilde{Z}\sim_{\mathbb{Q}} \varphi^*(Z)-\frac{1}{2}E,
$$
because $Z$ is a smooth curve. On the surface  $\widetilde{S}_\lambda$, by adjunction formula, we have
$$
\widetilde{Z}^2=\left\{\aligned
&-18\ \text{if $Z=\mathcal{L}_t^{\pm}$},\\
&0\ \text{if $Z$ is a curve of genus $1$},\\
&6\ \text{if $Z$ is a curve of genus $4$}.
\endaligned
\right.
$$
Set $H_{S_{\lambda}}=H\vert_{S_{\lambda}}$ and $\widetilde{H}_{\widetilde{S}_{\lambda}}=\varpi^*(H_{S_{\lambda}})$.  Then the intersections of the divisors $\widetilde{Z}$, $\widetilde{H}_{S_{\lambda}}$, $\widetilde{E}$  are given in the following table:
\begin{center}
\centering\renewcommand{\arraystretch}{1.5}
\begin{tabular}{|c||c|c|c|}
\hline
$\bullet$         &  $\widetilde{Z}$    & $\widetilde{H}_{\widetilde{S}_{\lambda}}$ & $\widetilde{E}$\\
\hline\hline
$\widetilde{Z}$   & $\widetilde{Z}^2$ & $9$ & $9$  \\
\hline
$\widetilde{H}_{\widetilde{S}_{\lambda}}$   & $9$ & $6$& $0$\\
\hline
$\widetilde{E}$   & $9$ & $0$&  $-18$\\
\hline
\end{tabular}
\end{center}
Recall that $M$ is a general surface in $\mathcal{M}$. Then
$$
M\big\vert_{S_{\lambda}}=mZ+D
$$
for an integer $m\geqslant \mathrm{mult}_Z(\mathcal{M})$ and an effective divisor $D$ on the~surface $S_{\lambda}$. Let $\widetilde{D}$ be the strict transform of the divisor $D$ on the surface $\widetilde{S}_\lambda$. Then
$$
m\widetilde{Z}+\widetilde{D}\sim_{\mathbb{Q}} n\widetilde{H}_{\widetilde{S}_\lambda}-qE
$$
for some non-negative integer $q\leqslant 0$. Since $0\leqslant E\cdot \widetilde{D}=18q-9m$, we have $q\geqslant\frac{m}{2}$.

Consider the $G$-invariant linear system $|3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E|$. If $Z=\mathcal{L}_t^{\pm}$, it follows from \eqref{equation:36-9-nine-lines} that the~linear system $|3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E|$ contains the strict transform of the curve $\mathcal{L}_t^{\mp}$. This immediately implies that the divisor $5\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E$ is nef. Hence, if $Z=\mathcal{L}_t^{\pm}$, then
$$
0\leqslant \big(5\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E\big)\cdot \widetilde{D}=\big(5\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E\big)\cdot\big(n\widetilde{H}_{\widetilde{S}_\lambda}-qE-m\widetilde{Z})=21n-54m-9q,
$$
which gives that $m<\frac{n}{2}$, which contradicts to our assumption. Thus, we see that $Z\ne\mathcal{L}_t^{\pm}$. Then
$$
\big(3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E\big)^2=\widetilde{Z}^2\geqslant 0,
$$
so it follows from the~Riemann--Roch formula and Serre duality that
$$
h^0\big(\mathcal{O}_{\widetilde{S}_{\lambda}}(3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E)\big)\geqslant 2+\frac{\widetilde{Z}^2}{2}\geqslant 2.
$$
In particular, the~linear system $|3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E|$ has a non-empty mobile part.

We claim that the linear system $|3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E|$ does not have fixed curves. Indeed, write
$$
|3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E|=\mathscr{F}+|\mathscr{M}|,
$$
where $\mathscr{F}$ is the~fixed part of this linear system, and $\mathscr{M}$ is a  divisor  such that $|\mathscr{M}|$ is the mobile part of the linear system $|3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E|$. Then $\mathscr{F}$ is $G$-invariant, and
$$
9=\widetilde{H}_{\widetilde{S}_\lambda}\cdot\big(3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E\big)=\widetilde{H}_{\widetilde{S}_\lambda}\cdot\big(\mathscr{F}+\mathscr{M}\big)>\widetilde{H}_{\widetilde{S}_\lambda}\cdot\mathscr{F}=\mathrm{deg}\big(\varpi(\mathscr{F})\big).
$$
On the other hand, by Lemmas~\ref{lemma:36-9-reducible-curves} and \ref{lemma:36-9-irreducible-curves},  $S_\lambda$ does not contain $G$-invariant curves of degree~$<9$. Hence, we see that $\mathrm{deg}(\varpi(\mathscr{F}))=0$, which simply means that the divisor $\mathscr{F}$ is $\varpi$-exceptional. Then $\mathscr{F}=pE$ for some positive integer $p$, so
$$
0\leqslant \mathscr{M}^2=\big(3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E-\mathscr{F}\big)^2=\big(3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-(1+p)E\big)^2=\widetilde{Z}^2-18p^2-18p\leqslant -30,
$$
which is absurd. Thus, the linear system $|3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E|$ does not have fixed curves.

In particular, the divisor $3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E$ is nef. Observe that
\begin{multline*}
0\leqslant \big(3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E\big)\cdot\widetilde{D}=\big(3\widetilde{H}_{\widetilde{S}_\lambda}-\widetilde{Z}-E\big)\cdot\big(n\widetilde{H}_{\widetilde{S}_\lambda}-aE-m\widetilde{Z}\big)=\\
=9n+m(\widetilde{Z}^2-18)-9a\leqslant 9n+m(\widetilde{Z}^2-18)-\frac{9m}{2}=9n+m\Big(\widetilde{Z}^2-\frac{45}{2}\Big),
\end{multline*}
since $q\geqslant\frac{m}{2}$. So, if $\widetilde{Z}^2=0$, then $m<\frac{n}{2}$, which contradicts to our assumption. Then $Z$ is a smooth curve of genus $4$ and degree $9$, and $\widetilde{Z}^2=6$.

Now, we consider the $G$-invariant linear system $|5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E|$. Since $(5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E)^2=12$, it follows from the~Riemann--Roch formula and Serre duality that
$$
\mathrm{dim}(|5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E|)\geqslant 7.
$$
Moreover, we claim that  $|5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E|$ does not have fixed curves. Indeed, as above, we write
$$
|5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E|=\mathscr{F}^\prime+|\mathscr{M}^\prime|,
$$
where $\mathscr{F}^\prime$ is the~fixed part of the linear system $|5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E|$, and $\mathscr{M}^\prime$ is an effective divisor on the surface $\widetilde{S}_\lambda$ such that the complete linear system $|\mathscr{M}^\prime|$ is the mobile part of $|5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E|$. We may futher assume that $\mathscr{M}^\prime$ is a general member of the linear system $|\mathscr{M}^\prime|$. Then
$$
\widetilde{H}_{\widetilde{S}_\lambda}\cdot \mathscr{M}^\prime=\mathrm{deg}\big(\varpi(\mathscr{M}^\prime)\big)\geqslant 3,
$$
because $S_\lambda$ is not uniruled. Moreover, if $\widetilde{H}_{\widetilde{S}_\lambda}\cdot \mathscr{M}^\prime=3$, then $\varpi(\mathscr{M}^\prime)$ is a smooth irreducible cubic curve, so $\mathscr{M}^\prime$ is an irreducible smooth curve of genus $1$ and, therefore,  $|\mathscr{M}^\prime|$ is a pencil \cite{SD}, which is impossible, since
$$
\mathrm{dim}\big(|\mathscr{M}^\prime|\big)=\mathrm{dim}\big(|5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E|\big)\geqslant 7.
$$
Hence, we conclude that $\widetilde{H}_{\widetilde{S}_\lambda}\cdot \mathscr{M}^\prime\geqslant 4$. This gives $\mathrm{deg}(\varpi(\mathscr{F}^\prime))\leqslant 8$, because
$$
12=\widetilde{H}_{\widetilde{S}_\lambda}\cdot\big(5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E\big)=\widetilde{H}_{\widetilde{S}_\lambda}\cdot\big(\mathscr{F}^\prime+\mathscr{M}^\prime\big)\geqslant \widetilde{H}_{\widetilde{S}_\lambda}\cdot\mathscr{F}^\prime+4=\mathrm{deg}\big(\varpi(\mathscr{F}^\prime)\big)+4.
$$
Now, it follows from Lemmas~\ref{lemma:36-9-reducible-curves} and \ref{lemma:36-9-irreducible-curves} that $\mathscr{F}^\prime=p^\prime E$ for some $p^\prime\in\mathbb{Z}_{>0}$. Then
$$
0\leqslant (\mathscr{M}^\prime)^2=\big(5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E-\mathscr{F}^\prime\big)^2=\big(5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-(1+p^\prime)E\big)^2=12-18(p^\prime)^2\leqslant -6.
$$
Thus, $|5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E|$ does not have fixed curves, so $5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E$ is nef. Then
$$
0\leqslant \big(5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E\big)\cdot \widetilde{D}=\big(5\widetilde{H}_{\widetilde{S}_\lambda}-2\widetilde{Z}-E\big)\cdot\big(n\widetilde{H}_{\widetilde{S}_\lambda}-qE-m\widetilde{Z}\big)=12n-24m,
$$
which gives $m\leqslant\frac{n}{2}$, which is a contradiction. This completes the proof of Proposition~\ref{proposition:36-9-curves}.
\end{proof}

\begin{corollary}
\label{corollary:36-9-curves}
Let $n\in\mathbb{Z}_{>0}$, let $\mathcal{M}$ be a non-empty $G$-invariant mobile linear subsystem in $|nH|$, and let 
$$
\Sigma=\Big\{P\in X_{a}\ \text{such that $\Big(X_{a},\frac{2}{n}\mathcal{M}\Big)$ is not canonical at~$P$}\Big\}.
$$
Then $\Sigma$ is a finite subset, $\Sigma\cap S=\varnothing$ and $\Sigma\cap\mathcal{R}_6=\varnothing$.
If $G=G_{36,10}$, then 
$$
\Sigma\cap\big(\mathcal{R}_6^\prime\cup \mathcal{R}_6^{\prime\prime}\big)=\varnothing.
$$
If $G=G_{36,10}$ and $a\ne 0$, then $\Sigma\cap(\mathcal{K}_6\cup \mathcal{K}_6^\prime)=\varnothing$.
If $X_a$ is singular, then $\Sigma\cap\mathrm{Sing}(X_a)=\varnothing$.
\end{corollary}

\begin{proof}
If $\Sigma$ contains a $G$-irreducible curve $Z$, then it follows from \cite[Excercise 6.18]{KollarSmithCorti} that
$$
\mathrm{mult}_Z\big(\mathcal{M}\big)>\frac{n}{2},
$$
which is impossible by Proposition~\ref{proposition:36-9-curves}. Thus, we deduce that $\Sigma$ is a finite subset.

If there exists a point $P\in\Sigma$ such that $P$ is contained in the surface $S$, then it follows from the~inversion of adjunction \cite[Theorem 6.29]{KollarSmithCorti} that $(S,\frac{2}{n}\mathcal{M}\vert_{S})$ is not log canonical at $P$, which contradicts to Corollaries~\ref{corollary:36-9-alpha} and \ref{corollary:36-10-alpha}. Thus, $\Sigma\cap S=\varnothing$.

Let $M$ and $M^\prime$ be general surfaces in $\mathcal{M}$. Then it follows from \cite{Pukhlikov1998,Corti2000} that $$
\mathrm{mult}_P\big(M\cdot M^\prime)>n^2
$$ 
for every point $P\in \Sigma$ such that $X_a$ is smooth at $P$. Let us use this to show that $\Sigma\cap\mathcal{R}_6=\varnothing$.
Namely, suppose that $\Sigma$ contains a point $P\in \mathcal{R}_6$. Then $X_a$ is smooth at $P$, and $M\cdot M^\prime=m\mathcal{R}_6+\Delta$, where $\Delta$ is an effective one-cycle on $X_{a}$ whose support does not contain $\mathcal{R}_6$. Then $m\leqslant \frac{n^2}{2}$, because
$$
3n^2=H\cdot M\cdot M^\prime=H\cdot\big(m\mathcal{R}_6+\Delta\big)=6m+H\cdot\Delta\geqslant 6m,
$$
where $H$ is a general hyperplane section of the cubic $X_{a}$. Let $\Sigma_P$ be the $G$-orbit of the point $P$. Then, since $\mathcal{R}_6$ is a smooth curve, we have
$$
\mathrm{mult}_O(\Delta)>n^2-m.
$$
for every point $O\in\Sigma_P$. Let $\mathcal{D}$ be the linear subsystem in $|2H|$ consisting of surfaces that contains the curve $\mathcal{R}_6$, and let $D$ be its general member. Then the surface $D$ contains no curves in $\mathrm{Supp}(\Delta)$, and we have $|\Sigma_P|\leqslant 6$ by Lemmas~\ref{lemma:36-9-orbits} and \ref{lemma:36-10-orbits}. Thus, we have
$$
6n^2-12m=D\cdot\Delta\geqslant\sum_{O\in \Sigma_P}\big(D\cdot\Delta)_O\geqslant \sum_{O\in\Sigma_P}\mathrm{mult}_O(\Delta)\geqslant |\Sigma_P|(n^2-m)\geqslant 6(n^2-m),
$$
which is a contradiction. Thus, we conclude that $\Sigma\cap\mathcal{R}_6=\varnothing$. 

Similarly, if $G=G_{36,10}$, we see that $\Sigma\cap\big(\mathcal{R}_6^\prime\cup\mathcal{R}_6^{\prime\prime}\big)=\varnothing$.

Let us show that $\Sigma\cap\mathcal{K}_6=\varnothing$ if $G=G_{36,10}$ and $a\ne 0$. Suppose that $G=G_{36,10}$ and $\Sigma\cap\mathcal{K}_6\ne\varnothing$. As above, write $M\cdot M^\prime=m^\prime\mathcal{K}_6+\Delta^\prime$, where $\Delta^\prime$ is an effective one-cycle such that $\mathcal{K}_6\not\subset\mathrm{Supp}(\Delta^\prime)$. Then, arguing as above, we see that $m^\prime\leqslant \frac{n^2}{2}$. As above, we let $\Sigma_P$ be the $G$-orbit of the point $P$. If $a\ne 0$, then the curve $\mathcal{K}_6$ is smooth, which gives
$$
\mathrm{mult}_O(\Delta^\prime)>n^2-m^\prime.
$$
for every point $O\in\Sigma_P$. Let $\mathcal{D}^\prime$ be the linear subsystem in $|3H|$ that consists of all surfaces passing through the curve $\mathcal{K}_6$, and let $D^\prime$ be a general member of the linear system $\mathcal{D}^\prime$. Then $D^\prime$ does not contain curves in $\mathrm{Supp}(\Delta^\prime)$ that pass through any point in $\Sigma_P$, because we proved that $\Sigma_P\not\subset S$. Moreover, it follows from Lemmas~\ref{lemma:36-10-orbits} that $|\Sigma_P|\leqslant 9$. Thus, if $a\ne 0$, then 
$$
9n^2-18m=D^\prime\cdot\Delta^\prime\geqslant\sum_{O\in\Sigma_P}\mathrm{mult}_O(\Delta^\prime)\geqslant |\Sigma_P|(n^2-m)\geqslant 9(n^2-m),
$$
which is a contradiction. Thus, if $G=G_{36,10}$ and $a\ne 0$, then  $\Sigma\cap\mathcal{K}_6=\varnothing$.

Likewise, if  $G=G_{36,10}$ and $a\ne 0$, then  $\Sigma\cap\mathcal{K}_6^\prime=\varnothing$.

Finally, we suppose that the cubic $X_a$ is singular ($a^3=-\frac{27}{2}$). Let us show that $\Sigma\cap\mathrm{Sing}(X_a)=\varnothing$. Suppose that this is not true. Then
$$
\mathrm{Sing}(X_a)\subseteq\Sigma.
$$
Moreover, the cubic $X_a$ contains a plane $\Pi$ such that $\Pi$ contains $4$ singular points of the cubic $X_a$, which we denote  by $P_1$, $P_2$, $P_3$,~$P_4$. Let $f\colon\widetilde{X}_a\to X_a$ be the blow up of the~points $P_1$, $P_2$, $P_3$, $P_4$, let $E_1$, $E_2$, $E_3$, $E_4$ be the irreducible $f$-exceptional surfaces such that $f(E_1)=P_1$, $f(E_2)=P_2$, $f(E_3)=P_3$, $f(E_4)=P_4$, let $\widetilde{\mathcal{M}}$ be the strict transform on the threefold $\widetilde{X}_a$ of the linear system~$\mathcal{M}$, and let $\widetilde{M}$ be a general surface in $\widetilde{\mathcal{M}}$. Then
$$
\widetilde{M}\sim\widetilde{\mathcal{M}}\sim f^*\big(nH\big)-\mu(E_1+E_2+E_3+E_4)
$$
for some $\mu\in\mathbb{Z}_{>0}$. Moreover, one has $\mu>\frac{n}{2}$ by \cite[Theorem 3.10]{Corti2000} or \cite[Theorem 1.7.20]{Cheltsov2005}. Now, we let $\mathscr{C}$ be a general conic in $\Pi$ that passes through $P_1$, $P_2$, $P_3$, $P_4$, and let $\widetilde{\mathscr{C}}$ be its strict transform on $\widetilde{X}_a$. Then $\widetilde{\mathscr{C}}\not\subset\widetilde{M}$, because $\Pi$ is not contained in the base locus of $\mathcal{M}$. Then
$$
0\leqslant \widetilde{M}\cdot\widetilde{\mathscr{C}}=f^*\big(nH\big)\cdot\widetilde{\mathscr{C}}-\mu(E_1+E_2+E_3+E_4)\cdot\widetilde{\mathscr{C}}=2n-4\mu,
$$
which gives $\mu\leqslant\frac{n}{2}$. This is a contradiction.
\end{proof}

Now, we are ready to prove Theorem~\ref{theorem:36-9}.

\begin{proof}[Proof of Theorem~\ref{theorem:36-9}]
Suppose that $X_{a}$ is not $G$-birationally superrigid.
By \cite[Corollary 3.3.3]{CheltsovShramov2016}, there exists a non-empty $G$-invariant mobile linear subsystem
$\mathcal{M}\subset |nH|$, for a positive integer~$n$, such that the singularities of the log pair $(X_{a},\frac{2}{n}\mathcal{M})$ are not canonical. Set
$$
\Sigma=\Big\{P\in X_{a}\ \text{such that $\Big(X_{a},\frac{2}{n}\mathcal{M}\Big)$ is not canonical at~$P$}\Big\}.
$$
By Corollaries~\ref{corollary:36-9-curves} and \ref{corollary:36-9-curves}, the subset $\Sigma$ is finite, $\Sigma\cap\mathrm{Sing}(X_a)=\varnothing$, $\Sigma\cap S=\varnothing$, $\Sigma\cap\mathcal{R}_6=\varnothing$. Moreover, if $G=G_{36,10}$, then
$$
\Sigma\cap\big(\mathcal{R}_6^\prime\cup \mathcal{R}_6^{\prime\prime}\big)=\varnothing.
$$
Furthermore, if $G=G_{36,10}$ and $a\ne 0$, then $\Sigma\cap(\mathcal{K}_6\cup \mathcal{K}_6^\prime)=\varnothing$.

By \cite[Remark 3.6]{CheltsovSarikyanZhuang2024}, the~singularities of the~log pair $(X_{a},\frac{3}{n}\mathcal{M})$ are not log canonical at every point of $\Sigma$. Let us use this to show a contradiction.

Let $P$ be a point in $\Sigma$, and let $\Sigma_P$ is the $G$-orbit of the point $P$. Choose $\mu\in\mathbb{Q}_{>0}$ such that the~singularities of the log pair $(X_{a},\mu\mathcal{M})$ are strictly log canonical at $P$. Then $\mu\leqslant \frac{3}{n}$. Recall that 
$$
\mathrm{Nklt}(X_{a},\mu\mathcal{M})=\big\{P\in X\ \text{such that $\big(X_{a},\mu\mathcal{M}\big)$ is not Kawamata log terminal at $P$}\big\}.
$$
Then the locus $\mathrm{Nklt}(X_{a},\mu\mathcal{M})$ does not contain curves that are \textbf{not disjoint} from the $G$-orbit $\Sigma_P$. Indeed, let $M$ and $M^\prime$ be general surfaces in $\mathcal{M}$.
If $Z$ is a $G$-irreducible curve in $\mathrm{Nklt}(X_{a},\mu\mathcal{M})$, then it follows from \cite[Lemma 1.8]{Corti2000} that
$$
\big(M\cdot M^\prime\big)_Z\geqslant\frac{4}{\mu^2}>\frac{4n^2}{9}
$$
which gives $\mathrm{deg}(Z)\leqslant 6$, so it follows from Lemmas~\ref{lemma:36-9-reducible-curves} and \ref{lemma:36-9-irreducible-curves} that $Z\cap \Sigma_P=\varnothing$.

We see that the points of the $G$-orbit $\Sigma_P$ are isolated components of the locus $\mathrm{Nklt}(X_{a},\mu\mathcal{M})$. Now, using the~Nadel vanishing theorem as in the~proof of Theorem~\ref{theorem:405-15}, we see that
$$
\big|\Sigma_P\big|\leqslant h^0\big(X_{a},\mathcal{O}_{X_{a}}(H)\big)=5,
$$
which contradicts Lemma~\ref{lemma:36-9-orbits}. The~obtained contradiction completes the~proof of Theorem~\ref{theorem:36-9}.
\end{proof}


\appendix

\section{Magma codes}
\label{section:Magma}
In this appendix, we present  Magma codes used in the proof of Theorem~\ref{theorem:main} (Main Theorem).

\subsection{Plane free groups acting on Fermat cubic}
\label{subsection:plane-free-groups}
Let $X\subset \mathbb{P}^4$ be the~Fermat cubic \eqref{equation:Fermat}.
Then 
$$
\mathrm{Aut}(X)\simeq C_4^3\rtimes\mathfrak{S}_5,
$$
and the~action of this group on $X$ lifts to $\mathbb{P}^4$, so we consider $\mathrm{Aut}(X)$ as a subgroup in $\mathrm{PGL}_5(\mathbb{C})$.
The~following Magma code goes through all subgroups of the~group $\mathrm{Aut}(X)$, and construct a graph whose vertices are conjugacy classes of subgroups $G\subset\mathrm{Aut}(X)$ such that $\mathbb{P}^4$ contains no $G$-invariant planes (plane free groups), and edges correspond to inclusion of subgroups.
\smallskip

\begin{verbatim}
// Graph of plane-free subgroup classes
// Vertex i = i-th plane-free subgroup conjugacy class representative H_i.
// Directed edge i -> j iff H_i contains a subgroup conjugate to H_j.

K<zeta> := CyclotomicField(60);
w := RootOfUnity(3, K);
G5 := GL(5, K);
S5 := Sym(5);
p12    := PermutationMatrix(K, S5!(1,2));
p12345 := PermutationMatrix(K, S5!(1,2,3,4,5));

// Constructing Aut(X) - group of order 9720
d1 := DiagonalMatrix(K, [w, w^-1, 1, 1, 1]);
d2 := DiagonalMatrix(K, [w, 1, w^-1, 1, 1]);
d3 := DiagonalMatrix(K, [w, 1, 1, w^-1, 1]);
d4 := DiagonalMatrix(K, [w, 1, 1, 1, w^-1]);
AutX := sub< G5 | d1, d2, d3, d4, p12, p12345 >;

// Plane test 
function HasInvariantPlane(H)
    M := GModule(H);
    D := DirectSumDecomposition(M);
    dims := [ Dimension(U) : U in D ];
    sums := {0};
    for d in dims do
        sums := sums join { s + d : s in sums };
    end for;
    return (3 in sums);
end function;

// Plane-free classes 
S := Subgroups(AutX);
PlaneFreeRecs := [* *];
for rec in S do
    if not HasInvariantPlane(rec`subgroup) then
        Append(~PlaneFreeRecs, rec);
    end if;
end for;
n := #PlaneFreeRecs;
PlaneFreeSubs := [ PlaneFreeRecs[i]`subgroup : i in [1..n] ];
PlaneFreeOrders := [ PlaneFreeRecs[i]`order : i in [1..n] ];
print "Plane-free vertex count =", n;

//Locate to which vertex class the subgroup belongs to.
OrderToIdx := AssociativeArray();
for i in [1..n] do
    o := PlaneFreeOrders[i];
    if IsDefined(OrderToIdx, o) then
        Append(~OrderToIdx[o], i);
    else
        OrderToIdx[o] := [i];
    end if;
end for;

function VertexIndexOfSubgroup(K)
    o := #K;
    if not IsDefined(OrderToIdx, o) then
        return 0;
    end if;
    for j in OrderToIdx[o] do
        flag, _ := IsConjugate(AutX, K, PlaneFreeSubs[j]);
        if flag then
            return j;
        end if;
    end for;
    return 0; // should not happen if K is plane-free 
end function;

// Build adjacency sets 
Adj := [* {} : i in [1..n] *];
for i in [1..n] do
    Hi := PlaneFreeSubs[i];
    oi := PlaneFreeOrders[i];
    SHi := Subgroups(Hi); // all subgroup classes inside Hi
        for srec in SHi do
            if srec`order ge oi then 
                continue;
            end if;
            K := srec`subgroup; 
            if HasInvariantPlane(K) then // only consider plane-free K
                continue;
            end if;
        j := VertexIndexOfSubgroup(K);
        if j ne 0 then
            Include(~Adj[i], j);
        end if;
    end for;
end for;

// Create digraph object and print adjacency 
Edges := &cat[ [ [i, j] : j in Adj[i] ] : i in [1..n] ];
D := Digraph< n | Edges >;
print "";
print "Adjacency list (edges i -> j):";
for i in [1..n] do
    printf "%o -> %o\n", i, Sort(SetToSequence(Adj[i]));
end for;

// Print output 
print "";
print "========================================";
print "digraph PlaneFree {";
for i in [1..n] do
    o := PlaneFreeOrders[i];
    lab := Sprintf("%o: |G|=%o", i, o);
    if CanIdentifyGroup(o) then
        lab := lab cat Sprintf("\\nIdGroup=%o", IdentifyGroup(PlaneFreeSubs[i]));
    end if;
    printf " %o [label=\"%o\"];\n", i, lab;
end for;
for e in Edges do
    printf " %o -> %o;\n", e[1], e[2];
end for;
print "}";
print "========================================";
\end{verbatim}

\subsection{Qualifying subgroups of the~group $C_4^3\rtimes\mathfrak{S}_5$}
\label{subsection:qualifying-subgroups}
Let $X\subset \mathbb{P}^4$ be the~Fermat cubic \eqref{equation:Fermat}.
Then 
$$
\mathrm{Aut}(X)\simeq C_4^3\rtimes\mathfrak{S}_5.
$$
The~following Magma code goes through all subgroups of the~group $\mathrm{Aut}(X)$, and construct a graph whose vertices are conjugacy classes of subgroups $G\subset\mathrm{Aut}(X)$ such that $G$ has an irreducible representation of dimension $4$ or $5$ (qualifying subgroups), and the~edges corresponds to inclusion.
\smallskip

\begin{verbatim}
// Directed graph of containment among the qualifying subgroup classes
// Edge v_i -> v_j means G_i contains a proper subgroup conjugate in Aut(X) to G_j.

RF := recformat<lattice_elt,class_index,order,class_length,gap_id,deg45,Hperm,Hmat>;

function GAPIdString(H)
    o := Order(H);
    if CanIdentifyGroup(o) then
        id := IdentifyGroup(H);
        return Sprintf("(%o,%o)", id[1], id[2]);
    else
        return "N/A";
    end if;
end function;

function RelevantDegrees45(H)
    return [ t[1] : t in CharacterDegrees(H) | t[1] in {4,5} and t[2] gt 0 ];
end function;

function BuildAutModels()
    K<zeta60> := CyclotomicField(60);
    z3 := zeta60^20;   // primitive cube root of unity
    P4<x0,x1,x2,x3,x4> := ProjectiveSpace(K, 4);
    f := x0^3 + x1^3 + x2^3 + x3^3 + x4^3;
    X := Scheme(P4, [f]);
    S5 := Sym(5);
    d1 := GL(5,K)!DiagonalMatrix(K, [z3, 1, 1, 1, z3^-1]);
    d2 := GL(5,K)!DiagonalMatrix(K, [1, z3, 1, 1, z3^-1]);
    d3 := GL(5,K)!DiagonalMatrix(K, [1, 1, z3, 1, z3^-1]);
    d4 := GL(5,K)!DiagonalMatrix(K, [1, 1, 1, z3, z3^-1]);
    sM := GL(5,K)!PermutationMatrix(K, S5!(1,2));
    tM := GL(5,K)!PermutationMatrix(K, S5!(1,2,3,4,5));
    AutXmat := sub< GL(5,K) | d1, d2, d3, d4, sM, tM >;
    assert Order(AutXmat) eq 9720;
    assert IsInvariant(f, AutXmat);
    S15 := Sym(15);
    c1 := S15!(1,2,3);
    c2 := S15!(4,5,6);
    c3 := S15!(7,8,9);
    c4 := S15!(10,11,12);
    c5 := S15!(13,14,15);
    a1 := c1*c5^-1;
    a2 := c2*c5^-1;
    a3 := c3*c5^-1;
    a4 := c4*c5^-1;
    sP := S15!(1,4)(2,5)(3,6);
    tP := S15!(1,4,7,10,13)(2,5,8,11,14)(3,6,9,12,15);
    AutXperm := sub< S15 | a1, a2, a3, a4, sP, tP >;
    assert Order(AutXperm) eq 9720;
    perm_to_mat := hom< AutXperm -> AutXmat | d1, d2, d3, d4, sM, tM >;
    return X, AutXmat, AutXperm, perm_to_mat;
end function;

function BuildQualifiedData()
    X, AutXmat, AutXperm, perm_to_mat := BuildAutModels();
    classes := SubgroupClasses(AutXperm);
    L := SubgroupLattice(AutXperm);
    good := [* *];
    for i in [1..#classes] do
        Hperm := classes[i]`subgroup;
        deg45 := RelevantDegrees45(Hperm);
        if #deg45 gt 0 then
            Hmat := sub< AutXmat | [ perm_to_mat(g) : g in Generators(Hperm) ] >;
            Append(~good, rec< RF |
                lattice_elt := L!Hperm,
                class_index := i,
                order := classes[i]`order,
                class_length := classes[i]`length,
                gap_id := GAPIdString(Hperm),
                deg45 := deg45,
                Hperm := Hperm,
                Hmat := Hmat>);
        end if;
    end for;
    return X, AutXmat, AutXperm, L, good;
end function;

// Driver 
X, AutXmat, AutXperm, L, good := BuildQualifiedData();
labels := [ Sprintf("v%o", i) : i in [1..#good] ];
S := {@ labels[i] : i in [1..#labels] @};
// Full containment digraph on the selected vertices
edges := [ [ S[i], S[j] ] :
    i, j in [1..#good] |
    i ne j and good[i]`lattice_elt ge good[j]`lattice_elt ];

D := Digraph< S | edges : SparseRep := true >;

printf "Number of vertices: %o\n", #good;
printf "Number of directed edges: %o\n\n", #edges;
print "Vertex legend:";
for i in [1..#good] do
    r := good[i];
    printf "%o : class index = %o, order = %o, class length = %o, 
            GAP ID = %o, irreducible degrees in {4,5} = %o\n",
        labels[i], r`class_index, r`order, r`class_length, r`gap_id, r`deg45;
    print "    generators:";
    for g in Generators(r`Hmat) do
        print g;
    end for;
    print "";
end for;
print "Containment digraph:";
print D;
\end{verbatim}

\subsection{Qualifying subgroups of the~group $((C_3^2\rtimes C_3)\rtimes C_4)\times\mathfrak{S}_3$}
\label{subsection:weird-threefold-qualifying-subgroups}

The following Magma code goes through all subgroups of the~group $((C_3^2\rtimes C_3)\rtimes C_4)\times\mathfrak{S}_3$ with GAP ID [648,541], and construct a graph whose vertices are conjugacy classes of subgroups $G$ such that $G$ have an irreducible representation of dimension $4$ or $5$ (qualifying subgroups), and the~edges of this graph corresponds to inclusion of subgroups.
\smallskip

\begin{verbatim}
// Directed graph of containment among the qualifying subgroup classes
// Edge v_i -> v_j means G_i contains a proper subgroup conjugate in Aut(X) to G_j.

RF := recformat<lattice_elt,class_index,order,class_length,gap_id,deg45,Hperm,Hmat>;

function GAPIdString(H)
    o := Order(H);
    if CanIdentifyGroup(o) then
        id := IdentifyGroup(H);
        return Sprintf("(%o,%o)", id[1], id[2]);
    else
        return "N/A";
    end if;
end function;

function RelevantDegrees45(H)
    return [ t[1] : t in CharacterDegrees(H) | t[1] in {4,5} and t[2] gt 0 ];
end function;

function BuildAutModels()
    K<zeta60> := CyclotomicField(60);
    z3 := zeta60^20;   // primitive cube root of unity
    P4<x0,x1,x2,x3,x4> := ProjectiveSpace(K, 4);
    sq3:=SquareRoot(K!3);
    f := x0^3 + x1^3 + x2^3 + x3^3 + x4^3 + 3*(sq3-1)*x0*x1*x2;
    X := Scheme(P4, [f]);
    S5 := Sym(5);
    A1 := GL(5,K)!PermutationMatrix(K,S5!(1,2,3));
    A2 := GL(5,K)!DiagonalMatrix(K, [1, z3, z3^2, 1, 1]);
    A3 := GL(5,K)!Matrix(K,5,5,[1,1,1,0,0, 
                                1,z3,z3^2,0,0, 
                                1,z3^2,z3,0,0, 
                                0,0,0,sq3,0, 
                                0,0,0,0,sq3])*ScalarMatrix(5,K!(1/sq3));
    A4 := GL(5,K)!PermutationMatrix(K,S5!(4,5));
    A5 := GL(5,K)!DiagonalMatrix(K, [1, 1, 1, z3, z3^2]);
    AutXmat := sub< GL(5,K) | A1,A2,A3,A4,A5 >;
    assert Order(AutXmat) eq 648;
    S32 := Sym(32);
    a1 :=  S32!(9,10,11)(15,16,17)(18,21,23)(19,22,20)(24,29,27)(25,30,28);
    a2 :=  S32!(12,13,14)(15,18,20)(16,19,21)(24,25,26)(27,30,32)(28,31,29);
    a3 := S32!(9,12)(10,13,11,14)(15,24,16,25)(17,26)
              (18,27,21,30)(19,28,20,29)(22,31)(23,32);
    a4 :=  S32!(1,2)(3,4)(5,8)(6,7);
    a5 := S32!(3,5,7)(4,6,8);
    AutXperm := sub< S32 | a1, a2, a3, a4, a5 >;
    assert Order(AutXperm) eq 648;
    perm_to_mat := hom< AutXperm -> AutXmat | A1,A2,A3,A4,A5 >;
    return X, AutXmat, AutXperm, perm_to_mat;
end function;

function BuildQualifiedData()
    X, AutXmat, AutXperm, perm_to_mat := BuildAutModels();
    classes := SubgroupClasses(AutXperm);
    L := SubgroupLattice(AutXperm);
    good := [* *];
    for i in [1..#classes] do
        Hperm := classes[i]`subgroup;
        deg45 := RelevantDegrees45(Hperm);
        if #deg45 gt 0 then
            Hmat := sub< AutXmat | [ perm_to_mat(g) : g in Generators(Hperm) ] >;
            Append(~good, rec< RF |
                lattice_elt := L!Hperm,
                class_index := i,
                order := classes[i]`order,
                class_length := classes[i]`length,
                gap_id := GAPIdString(Hperm),
                deg45 := deg45,
                Hperm := Hperm,
                Hmat := Hmat>);
        end if;
    end for;
    return X, AutXmat, AutXperm, L, good;
end function;

// Driver
X, AutXmat, AutXperm, L, good := BuildQualifiedData();

labels := [ Sprintf("v%o", i) : i in [1..#good] ];
S := {@ labels[i] : i in [1..#labels] @};

// Full containment digraph on the~selected vertices
edges := [ [ S[i], S[j] ] :
    i, j in [1..#good] |
    i ne j and good[i]`lattice_elt ge good[j]`lattice_elt ];

D := Digraph< S | edges : SparseRep := true >;

printf "Number of vertices: %o\n", #good;
printf "Number of directed edges: %o\n\n", #edges;
print "Vertex legend:";
for i in [1..#good] do
    r := good[i];
    printf "%o : class index = %o, order = %o, class length = %o, 
            GAP ID = %o, irreducible degrees in {4,5} = %o\n",
        labels[i], r`class_index, r`order, r`class_length, r`gap_id, r`deg45;
    print "    generators:";
    for g in Generators(r`Hmat) do
        print g;
    end for;
    print "";
end for;
print "Containment digraph:";
print D;
\end{verbatim}



\bibliographystyle{alpha}
\bibliography{cubics}





\end{document}