Explicit analytic functions defining the images of wave-front singularities

Extracted source text · arXiv 2512.21518

\documentclass[a4paper,11pt, leqno]{amsart}
\pagestyle{plain}
\usepackage{graphicx}
\usepackage{amssymb}
\usepackage[usenames]{color}
\oddsidemargin=1.2cm
\evensidemargin=1.2cm
\textwidth=380pt
\textheight=615pt    
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Enumerations
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\renewcommand{\labelenumi}{{\rm (\arabic{enumi})}}
\renewcommand{\labelenumii}{{\rm (\alph{enumii})}}
\renewcommand{\theenumi}{{\rm (\arabic{enumi})}}
\renewcommand{\theenumii}{{\rm (\alph{enumii})}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Theorems
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\numberwithin{equation}{section}
\newtheorem{Theorem}{Theorem}[section]
\newtheorem{Thm}[Theorem]{Theorem}
\newtheorem{Fact}[Theorem]{Fact}
\newtheorem{Corollary}[Theorem]{Corollary}
\newtheorem{Cor}[Theorem]{Corollary}
 \newtheorem{Lemma}[Theorem]{Lemma}
\newtheorem{Proposition}[Theorem]{Proposition}
\newtheorem{Prop}[Theorem]{Proposition}
\theoremstyle{remark}
\newtheorem{Remark}[Theorem]{Remark}
\newtheorem{Rmk}[Theorem]{Remark}
\theoremstyle{definition}
\newtheorem{Definition}[Theorem]{Definition}
\newtheorem{Def}[Theorem]{Definition}
\newtheorem{Example}[Theorem]{Example}
\newtheorem{Exa}[Theorem]{Example}
\newtheorem*{acknowledgements}{Acknowledgements}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Symbols
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\inner}[2]{\left\langle{#1},{#2}\right\rangle}
\newcommand{\trans}[1]{\vphantom{#1}^{t}#1}
\newcommand{\imag}{\op{i}}
\newcommand{\R}{{\mathbb R}}
\newcommand{\Q}{{\mathbb Q}}
\newcommand{\N}{{\mathbb N}}
\newcommand{\Z}{{\mathbb Z}}
\newcommand{\C}{{\mathbb C}}
\newcommand{\K}{{\mathbb K}}
\newcommand{\mc}[1]{{\mathcal #1}}
\newcommand{\mb}[1]{{\mathbf #1}}
\newcommand{\pmt}[1]{{\begin{pmatrix} #1  \end{pmatrix}}}
\renewcommand{\phi}{\varphi}
\renewcommand{\epsilon}{\varepsilon}
\newcommand{\op}[1]{{\operatorname{ #1}}}
\newcommand{\dy}{\displaystyle}
\newcommand{\gauss}[1]{\left\lfloor #1\right\rfloor}
\renewcommand{\Re}{\operatorname{Re}}
\renewcommand{\Im}{\operatorname{Im}}
\renewcommand{\labelenumi}{{\rm(\arabic{enumi})}}
\renewcommand{\theenumi}{{\rm(\arabic{enumi})}}
\renewcommand{\mid}{\, ; \,}
\newcommand{\pt}[1]{\mathrm{#1}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Titles
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\title{
Explicit analytic functions defining the images 
of wave-front singularities
}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\author{K.~Saji}%{Kentaro Saji}
\address[Kentaro Saji]{
  Department of Mathematics,
  Faculty of Science,
  Kobe University,
  Rokko, Kobe 657-8501}
\email{saji@math.kobe-u.ac.jp}

\author{M.~Umehara}%{Masaaki Umehara}
\address[Masaaki Umehara]{%
Department of Mathematical and Computing Sciences,
Institute of Science Tokyo,
2-12-1-W8-34, O-okayama, Meguro-ku,
Tokyo 152-8552, Japan.
}
\email{umehara@comp.isct.ac.jp}

\author{K.~Yamada}%{Kotaro Yamada}
\address[Kotaro Yamada]{%
   Department of Mathematics\\
   Institute of Science Tokyo\\
   O-okayama, Meguro, Tokyo 152-8551, 
   Japan
}
\email{kotaro@math.sci.isct.ac.jp}

\date{December 25, 2025}
\keywords{main-analytic, wave front, generic singular point,
Morin singular point}
\subjclass[2020]{Primary 32B20; Secondary 58C25.}

\thanks{The first author was supported in part 
by Grant-in-Aid for Scientific Research (C)
No. 25K07001. The second and the third authors were
supported in part 
by Grant-in-Aid for Scientific Research
(B) No. 23K20794 and (B) No. 23K22392 respectively,
from Japan Society for the Promotion of Science.}

\usepackage[active]{srcltx}
\begin{document}
\setcounter{tocdepth}{1}

\maketitle
\begin{abstract}
We give explicit real-analytic functions whose zero sets characterize
the images of the standard maps of wave-front singularities.
Such functions are realizations of the
\emph{main-analytic} sets in the sense of Ishikawa-Koike-Shiota (1984).
More concretely, a subset of Euclidean space is called a 
\emph{global main-analytic set} if it can be described, up to a 
set of smaller Hausdorff dimension, as part of the zero set 
of a single real-analytic function, referred to as its 
\emph{main-analytic function}.

In this paper, we propose a general framework for constructing
main-analytic functions by a method 
based on explicit resultant computations.
In particular, we provide explicit formulas 
for the main-analytic functions 
associated with the standard maps of wave-front singularities 
of types $A$, $D$ and $E$.
\end{abstract}
\section*{Introduction}
Main-analyticity, introduced in a variant form by 
Ishikawa--Koike--Shiota~\cite{IKS}, provides a useful framework for describing
geometric images that are globally determined by the zero set of a single 
real-analytic function.  
A subset $S\subset\R^{n}$ is called a \emph{global main-analytic set} if there
exists a real-analytic function $\Theta$ on $\R^{n}$ whose zero set
$\mathcal Z(\Theta)$ contains~$S$, has the same Hausdorff dimension as~$S$, 
and differs from~$S$ only in a subset of strictly smaller Hausdorff dimension.
Such a $\Theta$ is called a \emph{main-analytic function} for~$S$.

Ishikawa--Koike--Shiota~\cite{IKS} gave a criterion ensuring 
the existence of main-analytic functions for the critical value 
sets of proper real-analytic maps.
Moreover, Appendix~D of this paper gives a useful criterion for 
main-analyticity of images of polynomial maps.  
Consequently, many images arising naturally in singularity theory are expected 
to be global main-analytic sets.

Morin singular points are the corank one stable singularities (cf.\ \cite{S}).
A useful criterion for Morin singular points is given in \cite{S}.
We first show the following as a demonstration of our method:

\begin{Theorem}\label{thm:M2}
Let $h:\R^{m}\to\R^{n}$ $(1\le m\le5<n)$ be the 
standard map defining a Morin singularity.  
Then $h$ is proper and the inverse image $h^{-1}(\mathbf{x})$ is finite.
Moreover, one can explicitly construct a main-analytic function showing that
the image $h(\R^{m})$ is a global main-analytic set in $\R^{n}$.
\end{Theorem}

The restriction $m\le5$ is expected to be unnecessary, 
but our present techniques do not allow a proof in full generality.

\medskip
Simple singularities of ADE type
play a central role in the geometry of 
wave-fronts and their discriminants.  
Arnol'd~\cite{A} classified simple critical points 
of analytic functions 
via the ADE Dynkin diagrams.  
Saito~\cite{Sa1} showed that the orbit space of a finite reflection group 
carries a canonical flat structure that naturally corresponds to the base of 
the semi-universal deformation of a simple singularity.  
Looijenga~\cite{Loo} studied the topology and real-analytic geometry of the 
discriminant, relating it to the reflection representation via the period map.  


\medskip
The main part of this paper concerns wave-front singularities of types 
$A$, $D$, and $E$.  
Each of these can be realized as the image
\[
   S = f(\R^{k-1}) \subset \R^{k},
\]
where $f$ is a standard polynomial map determined by a generating family.
The complexification $f^{\C}:\C^{k-1}\to\C^{k}$ satisfies
\[
   f^{\C}(\C^{k-1}) = \{\Delta_S = 0\},
\]
where $\Delta_S$ is the canonical discriminant polynomial, 
which is known to be squarefree.
Moreover, its real zero set 
$\mathcal Z_\C(\Delta_S)\cap \R^{k}$ has real singular 
locus of codimension at least two, by a result of Looijenga~\cite{Loo}.  
Hence $\Delta_S$ is a main-analytic function for $f(\R^{k-1})$.
For type~$A$, the explicit form of $\Delta_S$ 
has long been known. 
For type~$D$, by contrast, to the best of our knowledge,
no explicit formula has been written in the canonical coordinates on $\R^k$
as the target space of the image of the standard map.
Our main result provides such formulas using a new method based on 
resultant computations with small primes.

\begin{Theorem}\label{thm:M}
Let $h:\R^{k-1}\to\R^{k}$ be a standard map defining a wave-front
singularity of type $A$, $D$, or $E$.  
Then there exist explicit polynomials $A^{\mathbf{x}}(v)$ and 
$B^{\mathbf{x}}(v)$ in one variable $v$, depending polynomially on 
$\mathbf{x}\in\R^{k}$, such that a main-analytic function 
$\Theta$ of $h(\R^{k-1})$ in $\R^{k}$ is given by
\[
  \Theta(\mathbf{x})
   = \frac{\operatorname{Res}_v\!\bigl(A^{\mathbf{x}}(v),
      B^{\mathbf{x}}(v)\bigr)}{r(\mathbf{x})^{2}},
\]
where $r(\mb x)\in \R[\mathbf{x}]$.
Moreover, the polynomial $\Theta\in \R[\mathbf{x}]$ agrees with 
the discriminant polynomial $\Delta_S$, 
up to a nonzero scalar.
\end{Theorem}

When $h$ is of type $A$ or $D$, one may take 
$r(\mb x):=1$ and 
$B^{\mathbf{x}}(v):=dA^{\mathbf{x}}(v)/dv$, so that $\Theta$ is, 
up to a constant,
the discriminant of $A^{\mathbf{x}}(v)$.  
For the cases $E_6$, $E_7$, and $E_8$, the factor $r(\mathbf{x})$ is 
nontrivial.

If one is interested only in proving that $h(\R^{k-1})$ is a global 
main-analytic set without writing down an global main-analytic function explicitly,
the criterion of Appendix~D applies effectively, 
especially in the $E$-type cases.

\medskip
Section~1 recalls preliminaries, including localization of 
main-analyticity and standard maps of singularities.  
Section~2 treats Morin singularities.  
Sections~3 and~4 handle the $A$- and $D$-types, while 
Sections~5--7 deal with the $E_6$, $E_7$, and $E_8$ cases.  
Appendices~A--D discuss resultants, dimension estimates for 
common zero sets of polynomials, a unified treatment of the 
$E$-types, and the criterion for main-analyticity of polynomial maps.



\section{Preliminaries}
Consider a proper real analytic map
$
h_S:(\R^{k},o) \to (\R^{l},\mb 0),
$
($k\le l$)
where $o$ is the origin of $\R^k$
and $\mb 0$ is the origin of $\R^l$.
We think of $h_S$ as the \lq\lq standard map'' of a singular point 
$o$.

\begin{Definition}\label{Def:S}
We consider the map
$$
\tilde h_S:\R^k\times \R^{m-k}\ni 
(\mb x,\mb y)\mapsto (h_{S}(\mb x),\mb y,\mb 0)
\in \R^{l}\times \R^{m-k}\times \R^{n-l-m+k}=\R^{n}
$$
called the {\it suspension} of the map $h_S$,
where $k\le m,\,\,l\le n$.
\end{Definition}


\begin{Proposition}
If $h_S(\R^{k})$ 
is a global main-analytic set, then so is $\tilde h_S(\R^k\times \R^{m-k})$.
\end{Proposition}

\begin{proof}
In fact, the main-analytic function $\Theta_{\tilde h_S}$
of $\tilde h_S(\R^k\times \R^{m-k})$ is given by
$$
\Theta_{\tilde h_S}(\mb x,\mb y,\mb z):=\Theta_{h_S}(\mb x)^2+|\mb z|^2
\qquad (\mb z\in \R^{n-l-m+k}),
 $$
where $\Theta_{h_S}$ is the main-analytic function for the image of $h_S$ and
$$
|\mb z|^2:=\sum_{i=1}^{n-l-m+k}|z_i|^2.
$$
\end{proof}

\begin{Definition}
Let $X^{m}$ be a real-analytic $m$-manifold.
A real-analytic map
$
 f:X^{m}\to\R^{n}\,\, (m\le n),
$
is called an \emph{$m$-dimensional real analytic map} if $\mathrm{rank}\,df=m$
on a dense open subset of $X^{m}$.
\end{Definition}

The following fact is useful.

\begin{Prop}\label{prop:HDIM}
If $f : X^{m} \to \mathbb{R}^{n}$ is an $m$-dimensional real analytic map,
then $f(X^{m})$ has Hausdorff dimension $m$.
\end{Prop}

\begin{proof}
Indeed, by definition there exists a nonempty open subset 
$U \subset X^{m}$ on which $f$ is a real-analytic embedding, and 
so it preserves Hausdorff dimension.
Hence 
\[
   \dim_H f(X^m) \;\ge\; \dim_H f(U)=m,
\]
while the local Lipschitz property of $f$ implies $\dim_H f(X^m)\le m$.
\end{proof}


\begin{Definition}\label{Def418}
Let $f:X^m\to \R^n$ $(m\le n)$
be an $m$-dimensional real analytic map.
We say that  
a point $p\in X^m$ is 
a {\it singular point of $f$ modeled on 
$\tilde h_S$}
if there exist local real analytic diffeomorphisms
$$
\phi:(\R^{k}\times \R^{m-k},(o,\mb 0))\to (X^m,p),\quad
\Psi:(\R^l, \mb 0)\to (\R^n,f(p))
$$
such that
\begin{equation}\label{269}
\Psi^{-1}\circ f \circ \phi(\mb x,\mb y)=\tilde h_S(\mb x,\mb y)
\end{equation}
holds for each point $(\mb x,\mb y)$ lying in a
sufficiently small neighborhood of the origin
of $\R^m$.
\end{Definition}



We can localize the property of
\lq\lq main-analyticity'' as follows:

\begin{Fact}[cf. \cite{FKKRUYY2}]\label{prop:390}
Let $h:\R^m\to \R^n$ $(m\le n)$
be an $m$-dimensional real analytic map 
such that $h^{-1}(h(o))=\{o\}$ and
$h(\R^m)$ is a global main-analytic set of $\R^n$.
Let $f:X^m\to \R^n$ be an $m$-dimensional real analytic  map
and $p_0\in X^m$ 
a singular point of $f$ 
modeled on $h$.
If we fix a neighborhood $\mc W$ of $p_0$,
then  there exist
\begin{itemize}
\item 
 a neighborhood $\mc U$
of $x_0$, 
\item 
an
open neighborhood $\Omega(\subset \R^n)$
of $f(x_0)$, 
\item 
a real analytic function
$F:\Omega\to \R$, and
\item 
a subset $\mc L$ of $\Omega$ satisfying $\dim_H(\mc L)<m$
\end{itemize}
 such that 
\begin{enumerate}
\item 
$(f|_{\mc U})^{-1}(f(x_0))=\{x_0\}$, and
\item 
$\mc V:=(f|_{\mc U})^{-1}(\Omega)$
satisfies
$\mc Z(F)=f({\mc V})\cup \mc L$ and
$f({\mc V})\cap \mc L=\varnothing$.
In particular, $f(\mc V)$ is a global main-analytic set of $\Omega$.
\end{enumerate}
\end{Fact}

\begin{Rmk}\label{rmk:3510}
In \cite{FKKRUYY2}, an \emph{$m$-image-analytic point is defined to be
a point $p_0 \in X$ satisfying conditions {\rm(1)} and {\rm(2)} of
Fact~\ref{prop:390}.
As shown in \cite{FKKRUYY2}, if an $m$-dimensional real analytic map
$f \colon X \to \mathbb{R}^n$ $(m \le n)$ satisfies the
\emph{arc-properness} condition, which is a weak form of properness,
and admits only $m$-image-analytic points, then it has no nontrivial
$m$-dimensional real analytic extensions.}
\end{Rmk}


In this paper, 
we consider the following standard maps: 
\begin{enumerate}
\item 
We fix integers $n,m$ satisfying $n>m\ge 2$,
and an integer $r\ge 1$ satisfying
$m\ge r(n-m+1)$.
The standard map $h_M:\R^m\to \R^n$ 
of {\it $r$-Morin singularity} of type $(m,n)$
is given by
\begin{equation}\label{eq:457}
h_M(\mb x)=(x_1,\ldots,x_{m-1}, h_1(\mb x),\ldots,h_{n-m+1}(\mb x)),
\end{equation}
where $\mb x:=(x_1,\ldots,x_m)$ and
\begin{align}
&h_i(\mb x):=\sum_{j=1}^r x_{j+r(i-1)}x_m^j \qquad (i=1,\ldots,n-m), \\
&h_{n-m+1}(\mb x):=x_m^{r+1}+\sum_{j=1}^{r-1} x_{j+r(n-m)}x_m^j.
\end{align}
For example, 
$(m,n,r)=(2,3,1)$ is the case of the standard map
of a cross cap.
If $(m,n,r)=(4,5,1)$, we have
$$
h_M(x_1,x_2,x_3,x_4)=(x_1,x_2,x_3,x_1x_4,x_4^2),
$$
and, if $(m,n,r)=(6,7,3)$, we have
\begin{equation}\label{eq:R3}
h_M(x_1,\ldots,x_6)=(x_1,\ldots,x_5,x_1x_6+x_2x_6^2+x_3x_6^3,x_4x_6+x_5x_6^2+x_6^4).
\end{equation}
\item 
Fix an integer $k(\ge 2)$. 
The standard map $h_{A_k}:\R^{k-1}\to \R^{k}$ 
of  an {\it $A_k$-singular point}
is defined by 
\begin{equation}\label{eq:pA408}
h_{A_k}(v,\mb x_2):=
\Big(
k v^{k+1}+\sum_{i=2}^{k-1}(i-1)x_i v^i,
-(k+1)v^k-\sum_{i=2}^{k-1}ix_i v^{i-1},
 \mb x_2\Big),
\end{equation}
where $\mb x_2:=(x_2,\ldots,x_{k-1})$.
\item
We fix an integer $k$ satisfying $k\ge 4$. 
The standard map $h_{D_k}:\R^{k-1}\to \R^{k}$ 
of a {\it $D_k$-singular point}
is defined by
$$
h_{D_k}(u,v,\mb x_3):=
\Big(
h_0(u,v,\mb x_3),
h_1(u,v,\mb x_3),
h_2(u,v,\mb x_3),\mb x_3
\Big)
$$
and
\begin{align}\label{eq:374a}
h_0(u,v,\mb x_3)&:=
\pm 2u^2 v+(k-2)v^{k-1}+\sum_{i=3}^{k-1}(i-2)x_iv^{i-1}, \\
\label{eq:374b}
h_1(u,v,\mb x_3)&:=\mp 2uv, \\
\label{eq:374c}
h_2(u,v,\mb x_3)&:=
\mp u^2-(k-1)v^{k-2}-\sum_{i=3}^{k-1}(i-1)x_iv^{i-2},
\end{align}
where $\mb x_3:=(x_3,\ldots,x_{k-1})$.
The map $h:=h_{D_k}$ has a
$\pm$-ambiguity $h=h_+$ or $h=h_-$
(see Remark \ref{rmk:433}).

\item
The standard map $h_{E_6}:\R^5\to \R^{6}$ 
of an {\it $E_6$-singular point}
is defined by 
\begin{align*}
h_{E_6}(u,v, x_3,x_4,x_5)&:=
\Big(
2u^3+3v^4+v^2x_3+uvx_4+2uv^2x_5, \\
&\phantom{aaaaaaaaa}
 -3u^2-vx_4-v^2x_5, \\
&\phantom{aaaaaaaaaaaaaa}
-4 v^3-2vx_3-ux_4-2uv x_5,
 x_3, x_4,x_5\Big).
\end{align*}
Similarly, 
the standard maps $h_{E_7}:\R^6\to \R^{7}$,
and $h_{E_8}:\R^7\to \R^{8}$ 
of $E_7$ and $E_8$ singular points
are defined in Sections~6 and~7.
\end{enumerate}


\section{Our strategy to find main-analytic functions} 

\subsection*{Basic materials on resultants}

The key to our proof of Theorem~\ref{thm:M}
lies in the use of \lq\lq resultants'' of two polynomials (cf. Appendix~A). 

In \textit{Mathematica}, 
the built-in function
$$
 \op{Subresultants}\Big[\op{poly1},\op{poly2},\op{var}\Big]
$$
generates the list of the (principal) subresultant coefficients 
$$
\op{Res}_{\op{var}}^{(i)}(\op{poly1},\op{poly2}) 
\qquad (i=0,1,2,3,\ldots)
$$ 
of  the polynomials poly1 and poly2 with respect to the variable var.
Each (principal) subresultant coefficient
 is the determinant of a submatrix
of the Sylvester matrix.
In particular, when $i=0$,
$$
\op{Res}_{\op{var}}(\op{poly1},\op{poly2}):= 
\op{Res}_{\op{var}}^{(0)}(\op{poly1},\op{poly2}) 
$$
is the usual resultant between $\op{poly1}$ and $\op{poly2}$.
For two polynomials $a(v)$ and $b(v)$ whose 
leading coefficients 
are non-vanishing,  
 $a(v)$ 
and $b(v)$ have $k$ common roots in $\C$
if and only if
\begin{equation}
\op{Res}_{v}^{(0)}(a,b)=\ldots=\op{Res}_{v}^{(k-1)}(a,b)=0.
\end{equation}
This criterion will be useful in the
following discussion.
The definitions of the resultant $\op{Res}_{v}^{(0)}(a,b)$
(i.e. the $0$-th (principal) subresultant coefficient) 
and the first (principal) subresultant coefficient denoted by
\begin{equation}\label{eq:550}
\op{Psc}_{v}(a,b)
:=\op{Res}_{v}^{(1)}(a,b)
\end{equation}
are given in  Appendix A.
Precise properties of subresultants
are in Vega \cite{V}.

\medskip
Let
$
h_S:(\R^{k},o) \to (\R^{l},\mb 0)
$
be the standard map of a singular point $o$.
We now describe our strategy to find the main-analytic
function of the image of $h_S$:
We will find real polynomials $A^{\mb x}_i(v)$ ($i=1,\ldots,j$)
in $v$ with parameter $\mb x\in \R^k$
so that
$$
h_S(\R^k)=\{\mb x\in \R^k\,;\, \text{$\exists$ $v\in \R$
such that
$A^{\mb x}_i(v)=0$ for $i=1,\ldots,j$}
\},
$$
where $j$ is a positive integer.
We say that such a  family of polynomials
$\left(A^{\mb x}_i(v)\right)_{i=1}^{j}$ 
is the {\it characteristic family of polynomials}.
When $j=2$, we set $A^{\mb x}:=A^{\mb x}_1$ and $B^{\mb x}:=A^{\mb x}_2$, and
call $(A^{\mb x},B^{\mb x})$ the {\it characteristic pair of polynomials}
associated with the map $h$.
If $A^{\mb x}(v)=B^{\mb x}(v)=0$ for some non-real $v\in \C\setminus \R$,
then the first (principal) subresultant coefficient
$\op{Psc}_v(A^{\mb x},B^{\mb x})$ vanishes at $\mb x$
as well as $\op{Res}_v(A^{\mb x},B^{\mb x})$,
since the conjugate $\bar v$ of $v$ is another common root.
So, if 
\begin{align}\label{eq:Exp}
&\dim_H\{\mb x\in \R^k\,;\,
\op{Res}_v(A^{\mb x},B^{\mb x})=
\op{Psc}_v(A^{\mb x},B^{\mb x})=0\}\\
&\phantom{aaaaaa}\nonumber
<\dim_H\{\mb x\in \R^k\,;\,
\op{Res}_v(A^{\mb x},B^{\mb x})=0\}
\end{align}
holds, we can conclude that
\begin{equation}\label{eq:theta}
\mc R(\mb x):=\op{Res}_v(A^{\mb x},B^{\mb x})
\end{equation}
is a main-analytic function of $h_S(\R^k)$.
To check the inequality \eqref{eq:Exp},
we prepare useful methods 
in Appendix B.
However, in general,
$\mc R(\mb x)$ can be factorized into the form 
$\mc R(\mb x)=q(\mb x)\Theta(\mb x)$,
and $\Theta(\mb x)$ may be the desired main-analytic function.
This actually occurs in the case of the singularity of type $E$. 

We give here an elementary example.

\begin{Exa}\label{ex:623}
The real analytic map 
$
h:\R^2\ni (u,v)\mapsto (u,uv,v^2)\in \R^3
$
gives the standard map of the {\it cross cap}.
If we set $h(u,v)=(x,y,z)$,
then we have
$$
u=x,\quad uv=y,\quad v^2=z,
$$
which induce the two polynomials
$A(v):=xv-y$ and $B(v):=v^2-z$
giving the characteristic pair of polynomials, and
$\Theta(x,y,z):=y^2-zx^2$
coincides with the resultant
$
\op{Res}_v(A,B).
$ 
Since 
$
\mc S:=\op{Psc}_v(A,B)=x
$ (cf. \eqref{eq:550}),
$\mc Z(\Theta)\setminus h(\R^2)$ 
lies in the
zero set 
$$
\{x=y=0\}=\mc Z(\Theta)\cap 
\mc Z(\mc S),
$$
the resultant $\Theta$ 
gives a main-anlytic function of $h(\R^2)$.
\end{Exa}



\subsection*{The proof of Theorem~\ref{thm:M2}
}

We now fix integers $n,m$ satisfying $n>m\ge 2$. 
The standard map $h_{r,m}:\R^m\to \R^n$ 
of an $r$-Morin singular point
is defined when $m\ge r(n-m+1)$ (see \eqref{eq:457}).
If $m\le 5$, then we have
$$
r\le \frac{m}{n-m+1}\le \frac{m}{2}\le 2.
$$
Such a possibility for $r$ is $1$ or $2$.


\subsubsection*{\bf The case of $r=1$}

When $r=1$, each Morin map 
is obtained 
as the 
suspension (cf. Definition \ref{Def:S})
of the
standard $m$-dimensional cross cap 
$h_{1,m}:\R^m\to \R^{2m-1}$ ($m\ge 2$)
defined by
\begin{equation}
h_{1,m}(u_1,\ldots,u_m):=\left(u_1,\ldots,u_{m-1},u_1u_m,\ldots,u_{m-1}u_m,u_m^2\right).
\end{equation}
If $m=2$, then $h_{1,2}:\R^2\to \R^3$ is the standard map of the cross cap given
in Example \ref{ex:623}.
It can be easily checked 
that $h_{1,m}$ is a proper map and also an $m$-dimensional
real analytic map having exactly one isolated 
singular point at the origin in $\R^m$
satisfying $h_{1,m}^{-1}(\mb 0)=\{o\}$.
The image of $h_{1,m}$ is contained in the
following subset of $\R^{2m-1}$:
$$
\mc W_{1,m}:=\Big\{(x_1,\ldots,x_{2m-1})\in \R^{2m-1}\,;\, 
x_{m-1+i}^2=x_i^2 x_{2m-1}\,\,\, (i=1,\ldots,m-1)\Big\}.
$$
More precisely, it holds that
\begin{align}\label{eq:I-Cross-cap}
&\mc W_{1,m}=h_{1,m}(\R^m)\cup \mc L \\ \nonumber
&\phantom{aaaa}\mc L:=\Big\{(x_1,\ldots,x_{2m-1})
\in \R^{2m-1}\,;\, 
x_1
=\cdots=x_{2m-2}=0,\,\,
 x_{2m-1}<0\Big\}.
\end{align}
We remark that
$
\mc L
$
coincides with the image of the set of self-intersections of $h_{1,m}$.
This indicates that
$h_{1,m}(\R^m)$ is 
a global main-analytic subset of $\R^{2m-1}$, whose
main-analytic function is
$$
\Theta_{1,m}(x_1,\ldots,x_{2m-1}):=
\sum_{i=1}^{m-1}
\Big(x_{m-1+i}^2-x_i^2 x_{2m-1}\Big)^2.
$$



\subsubsection*{\bf The case of $r=2$}
Since $m\le 5$, 
the possibilities of Morin maps 
are
the standard map $h_{2,4}:\R^4\to \R^5$
defined by
\begin{align}\label{eq:h24}
&h_{2,4}(x_1,x_2,x_3,x_4)
=(x_1,x_2,x_3, x_1x_4+x_2x_4^2,x_3x_4+x_4^3)
\end{align}
and its suspension. So it is sufficient to consider the
main-analyticity of the map $h_{2,4}$.
It can be easily checked that 
$h_{2,4}$ 
is a $4$-dimensional real
analytic map.
We write $(h^1,\ldots,h^5):=h_{2,4}$
and set
\begin{align*}
a(y_1,\ldots,y_5,t)&:=-y_4+h^4(y_1,y_2,y_3,t)=
-y_4+y_1t+y_2t^2, \\
b(y_1,\ldots,y_5,t)&:=-y_5+h^5(y_1,y_2,y_3,t)=
-y_5+y_3t+t^3.
\end{align*}
Then 
the image of $h_{2,4}$ satisfies
\begin{equation}
h_{2,4}(\R^4)
=\{\mb y\in \mb \R^5\,;\, 
\text{there exists $t\in \R$ such that\,\,}
a(t,\mb y)=b(t,\mb y)=0\},
\end{equation}
that is, $(a,b)$ gives 
the characteristic pair of polynomials
associated with $h_{2,4}(\R^4)$.
Since the leading term of $b(t)$ is $t^3$,
we can apply the computation of resultant
as in Appendix A.
The resultant $\Theta_{2,4}(\mb y):=\op{Res}_t(a,b)$ of the
two polynomials $a,b$ in $t$ can be computed as
\begin{align*}
&\Theta_{2,4}(\mb y):=
-y_1^2 y_3 y_4 - y_2^2 y_3^2 y_4 - 2 y_2 y_3 y_4^2 - y_4^3 + y_1^3 y_5 \\
&\phantom{aaaaaaaaaaaaaaaaaaaaaa}
+ 
 y_1 y_2^2 y_3 y_5 + 3 y_1 y_2 y_4 y_5 + y_2^3 y_5^2.
\end{align*}



The proof of
Theorem~\ref{thm:M2} is accomplished by the following statement.

\begin{Prop}\label{thm:r2a}
The standard map
$(h:=)h_{2,4}:\R^4\to \R^5$ is a proper $4$-dimensional real analytic map whose
singular set is the two dimensional real analytic set
$$
\Sigma:=\{\mb x\in \R^4\,;\, x_1+2 x_2 x_4=x_3+3 x_4^2=0\}.
$$
Moreover, $h^{-1}(\mb y)$ is finite for each $\mb y\in \R^5$,
and $h(\R^4)$ is 
a global main-analytic set of $\R^{2m-1}$
with a main-analytic function $\Theta_{2,4}$.
\end{Prop}

\begin{proof}
We first prove the properness of $h$:
Let $\mc K$ be a compact subset of $\R^5$.
Consider a sequence $\{p_i\}_{i=1}^\infty$ of points  in $h^{-1}(\mc K)$.
It is sufficient to prove that $\{p_i\}_{i=1}^\infty$
has a convergent subsequence.
For each positive index $i$, we can write
$$
p_i=(\mb y^{(i)},t^{(i)}),\qquad 
h(p_i)=\Big(\mb y^{(i)} ,t^{(i)},h^4(\mb y^{(i)},t^{(i)}),
h^5(\mb y^{(i)},t^{(i)})\Big) \in \mc K,
$$
where $t^{(i)}\in \R$ and $\mb y^{(i)}=(y^{(i)}_1,y^{(i)}_2,y^{(i)}_3)\in \R^3$.
Since $\mc K$ is bounded,
there exists a constant $C$
not depending on $i$ such that the five components
of $h_{2,4}$ in \eqref{eq:h24} with $t:=x_4$
can be estimated by
$$
|y^{(i)}_1|,\,\,|y^{(i)}_2|,\,\, |y^{(i)}_3|,\,\,
|y^{(i)}_1t^{(i)}+y^{(i)}_2(t^{(i)})^2|,\,\,
|y^{(i)}_3t^{(i)}+(t^{(i)})^3|<C.
$$
We set $z_0:=y^{(i)}_3t^{(i)}+(t^{(i)})^3$.
Then any solution $t=\alpha$ of 
$
t^3+y^{(i)}_3t-z_0=0
$
satisfies
$
|\alpha|< 1+C
$ 
because of  Cauchy's upper bounds for roots
of the polynomials (see the remark below).
In particular, we have $|t^{(i)}|<1+C$, and
can conclude that $\{p_i\}_{i=1}^\infty$ is 
bounded in $\R^{4}$, proving the assertion.

The finiteness of
the set $h^{-1}(\mb y)$ 
can be proved easily.
Let $\mb y$ be a point in the set $\mc Z(\Theta_{2,4})$.
We examine the possibility of $t$
satisfying $h(y_1,y_2,y_3,t)=\mb y$.
We set $\mc S:=\op{Psc}_v(a,b)$.
If $t\in \C\setminus \R$,
the conjugate $\overline{t}$ is also a
common root of $a,b$. So
$\mb y$ must lie in the set
$
\mc L=\{\mb y\in \R^5\,;\, \Theta_{2,4}(\mb y)=\mc S(\mb y)=0\}.
$
We obtain
$$
\mc S(y_1,y_2,y_3,y_4,y_5)=y_4 y_2+y_1^2+y_2^2 y_3.
$$
If $y_2=0$, then $\mc S=0$ implies $y_1=0$.
Then $\Theta_{2,4}=0$ reduces to
$y_4=0$.
So we consider the case that $y_2\ne 0$.
Then $\mc S=0$ implies
$y_3=-(y_1^2 +y_4 y_2)/y_2^2$
and
$$
0=\Theta_{2,4}=\frac{(y_4y_1+y_5 y_2^2)^2}{y_2}.
$$
So we can conclude that $\dim_H(\mc L)=3$.\end{proof}

\begin{Rmk}[Cauchy's bound for complex roots]\label{rmk:Cauchy}
For a given polynomial
$t^n+a_{n-1}t^{n-1}+\cdots+a_1t+a_0$ in $t$
for complex coefficients,
any root $t=\alpha(\in \C)$ of the polynomial
satisfies $|\alpha|<1+\max_{i=0,\ldots,n-1}|a_i|$.
\end{Rmk}




\section{Singularities of type A}

As  defined in the previous section,
the standard map
$
h:=h_{A_k}:\R^{k-1}\to \R^{k}
$
for an $A_k$-singular point ($k\ge 2$)
is given by \eqref{eq:pA408}. 

\begin{Prop}\label{prop:A182}
The image of the map $h$
coincides with
the set
\begin{align*}
\mc W_{A_k} &:=\Big\{\mb x:=(x_0,\ldots,x_{k-1})\in \R^k\,;\, \\
& \phantom{aaaaaaaaa}
\text{the polynomial $A^{\mb x}(v)$ in $v$ has 
a real double root}\Big\},
\end{align*}
where
\begin{equation}\label{eq:F191}
A^{\mb x}(v):=v^{k+1}+x_{k-1}v^{k-1}+\cdots+x_1 v+x_0.
\end{equation}
\end{Prop}

\begin{proof}
We set 
\begin{equation}\label{eq:881}
B^{\mb x}(v) := \frac{dA^{\mb x}(v)}{dv}.
\end{equation}
We remark that $\mb x \in \mc W_{A_k}$  
if and only if there exists $v \in \R$ such that  
$A^{\mb x}(v) = B^{\mb x}(v) = 0$.  
So we assume the existence of such a $v \in \R$.
Since
$
B^{\mb x}(v)=(k+1)v^{k}+\sum_{i=1}^{k-1}i x_i v^{i-1},
$
we set
\begin{equation}\label{eq:481}
X_1(v,\mb x_2):=-(k+1)v^k-\sum_{i=2}^{k-1} ix_i v^{i-1},
\end{equation}
where $\mb x_2:=(x_2,\ldots,x_{k-1})$.
Substituting this into \eqref{eq:F191},
$A^{\mb x}(v)=0$ reduces to
\begin{align*}
x_0&=
-v^{k+1}+v\Big((k+1)v^k+\sum_{i=2}^{k-1} ix _i v^{i-1}\Big)
 -\sum_{i=2}^{k-1}x_i v^i \\
&=k v^{k+1}+\sum_{i=2}^{k-1} (i-1) x_i v^i.
\end{align*}
So, if we set
\begin{equation}\label{eq:481b}
X_0(v,\mb x_2):=k v^{k+1}+\sum_{i=2}^{k-1} (i-1) x_i v^i,
\end{equation}
then we have $h(v,\mb x_2)=(X_0(v,\mb x_2),X_1(v,\mb x_2),\mb x_2)$,
which proves the assertion. 
\end{proof}


\begin{Prop}\label{prop:1a}
The map $h:\R^{k-1}\to \R^k$ and its
complexification $h^\C:\C^{k-1}\to \C^k$ are proper.
\end{Prop}

\begin{proof}
By Proposition~\ref{prop:A182}, we have
$
h(v,\mb x_2)=\bigl(X_0(v,\mb x_2),\,X_1(v,\mb x_2),\,\mb x_2\bigr)
$.
Since the leading coefficients of $X_0$ and $X_1$
as polynomials in  $v$
are nonzero constant, $h$ is proper.
The same argument applies to $h^{\C}$.
\end{proof}

For the unified treatment, set $\K=\R$ or $\C$, and define
\begin{equation}\label{eq:hatH}
\hat h :=
\begin{cases}
h & \text{if } \K=\R,\\
h^\C & \text{if } \K=\C.
\end{cases}
\end{equation}

\begin{Prop}\label{prop:1b}
For each $\mb x\in \K^k$,
the inverse image $\hat h^{-1}(\mb x)$ 
consists of at most $k$ points.
Moreover, $\hat h^{-1}(\mb 0)=\{o\}$ holds,
where $o$ and $\mb 0$ are
the origins of $\K^{k-1}$ and $\K^{k}$, respectively.
\end{Prop}

\begin{proof}
By \eqref{eq:481}, the number of candidates for a real number $v$
satisfying $\hat h(v,\mb x_2)=(x_0,x_1,\mb x_2)$
is at most $k$.
For each such $v$, the corresponding $x_0$ is uniquely 
determined by \eqref{eq:481b}. 
This proves the first assertion.
The second assertion $\hat h^{-1}(\mb 0)=\{o\}$ is obvious.
\end{proof}

\begin{Prop}\label{prop:1c}
The map $h$ 
is a $(k-1)$-dimensional real analytic map.
\end{Prop}

\begin{proof}
We write
$h=(h_0,\ldots, h_{k-1})$.
If we remove the first two columns and
rows, the Jacobian matrix $J$ of $h$
gives the identity matrix of rank $k-2$. 
So, if the $1\times 2$-submatrix 
$
M:=\Big((h_0)_v ,(h_1)_v\Big)
$
of $J$ does not vanish at a certain point,
we obtain the conclusion: 
Since $h_0=x_0$ and $h_1=x_1$, 
by setting $(x_2,\ldots,x_{k-1})=(0,\ldots,0)$, 
$
M=k(k+1)v^{k-1}(v,-1)
$
is obtained, which does not vanish if $v\ne 0$.
\end{proof}

\begin{Theorem}\label{thm:A622}
The image $\mc W_{A_k}\,(=h(\R^{k-1}))$ $(k\ge 2)$
is a global main-analytic set of $\R^{k}$
whose main-analytic function is 
$\Theta_{A_k}:=\op{Res}_v(A^{\mb x},B^{\mb x})$
$($cf. \eqref{eq:881}$)$.
\end{Theorem}

\begin{proof}
Using $A^{\mb x}$
given in \eqref{eq:F191},
we have
\begin{equation}\label{eq:579}
h(\R^{k-1})=\mc Z(\Theta_{A_k})\setminus \mc L
\end{equation}
and
\begin{align*}
\mc L&:=\Big\{\mb x\in \mc Z(\Theta_{A_k})\,;\, 
\text{every common root $v\in\C$ of $A^{\mb x}$ and $B^{\mb x}$ is non-real}\Big\}.
\end{align*}
By \eqref{eq:579} with
Proposition~\ref{prop:HDIM},
we have
$$
\dim_H \mc Z(\Theta_{A_k})=\dim_H h(\R^{k-1})=k-1.
$$
So,
to prove that $\Theta_{A_k}$ is a main-analytic function
of the set $h(\R^{k-1})$,
it is sufficient to show that the dimension of the set
 $\mc L$ is less than $k-1$.

If $k=2$, then $h$ is the standard map of a cusp
and $\mc L$ is empty (see Example~\ref{exa:C}). 
In particular, the assertion
is obvious. So we may set $k\ge 3$.
Suppose that 
$(\alpha,x_2,\ldots,x_{k-1})\in \R^k$
satisfies $h(\alpha,x_2,\ldots,x_{k-1})\in \mc L$.
Then $\alpha$ is a non-real double root of $A^{\mb x}(v)$.
Since the coefficients of $A^{\mb x}$ are real numbers, 
the conjugate $\overline{\alpha}$ is also 
a non-real double root of $A^{\mb x}(v)$.
In particular, we can write
\begin{equation}\label{eq:613}
A^{\mb x}(v)=\Big(v^2-(\alpha+\bar \alpha)v+|\alpha|^2\Big)^2 
\left(v^{k-3}+\sum_{i=0}^{k-4} c_iv^i\right).
\end{equation}
Comparing coefficients in \eqref{eq:F191} and \eqref{eq:613},
we have $c_{k-4}=2(\alpha+\bar \alpha)$.
Thus, the expression \eqref{eq:613}
implies that
$\mb x\in \mc L$
depends only on the 
parameters $\alpha\in \C$ and $c_0,\ldots, c_{k-5}$.
So we have 
$$
\dim_H\,\mc L\le \dim_H \C+\dim_H \R^{k-4}=2+(k-4)=k-2,
$$
proving the assertion.
\end{proof}


\begin{Rmk}\label{rmk:a}
From the above discussion, in the complex setting,
one can easily verify that
$
h^\C(\C^{k-1})=\mathcal{Z}_\C(\Theta_{A_k}),
$
where $h^\C$ is the complexification of $h$, and
\[
\mathcal{Z}_\C(\Theta_{A_k})
:=\{\mb x\in \C^k \mid \Theta_{A_k}(\mb x)=0\}.
\]
Moreover, $\Theta_{A_k}$ is well-known 
to be irreducible over~$\C$. 
Therefore, $\Theta_{A_k}$ coincides with the defining polynomial of
$h^\C(\C^{k-1})$. 
In particular,
$\Theta_{A_k}$ coincides with 
the discriminant polynomial of type $A_k$
as in the introduction.
\end{Rmk}


\begin{Exa}\label{exa:C}
When $k=2$, the standard map $h:\R\to \R^2$ is
$
h(v)=(2v^3,-3v^2),
$
which has a cusp singular point at $v=0$.
In this case
\[
A^{\mb x}(v)=v^3+x_{1}v+x_0,\qquad  B^{\mb x}(v)=3v^2+x_1,
\]
and
\[
\Theta_{A_2}(x_0,x_1)
:=\op{Res}_v\!\big(A^{\mb x},B^{\mb x}\big)
=27 x_0^2+4 x_1^3
\]
is a main-analytic function of $h(\R)$.
In particular, $\mathcal{Z}(\Theta_{A_2})\setminus h(\R)=\varnothing$.
\end{Exa}

\begin{Exa}\label{exa:SW}
When $k=3$, the standard map $h:\R^2\to \R^3$ is
\[
h(v,x_2)=(3v^4+x_2v^2,\,-4v^3-2x_2v,\,x_2),
\]
which has a swallowtail singular point at $(v,x_2)=(0,0)$.
Here
\[
A^{\mb x}(v)=v^4+x_{2}v^2+x_{1}v+x_0,\qquad
B^{\mb x}(v)=4v^3+2x_2v+x_1,
\]
and the function
$
\Theta_{A_3}(x_0,x_1,x_2):=\op{Res}_v\!\big(A^{\mb x},B^{\mb x}\big)
$
is computed as
\[
\Theta_{A_3}
= 256\,x_0^3 
-128\,x_2^2\,x_0^2
+\bigl(144 x_1^2 x_2 + 16 x_2^4 \bigr)x_0
-\bigl(27 x_1^4 + 4 x_1^2 x_2^3 \bigr),
\]
which gives a main-analytic function of $h(\R^2)$.
If we set
$
\mathcal L:=\Bigl\{\bigl(\tfrac{t^2}4,\,0,\,t\bigr)\,;\; t>0\Bigr\},
$
then
$
\mc Z(\Theta_{A_3})\setminus h(\R^2)=\mathcal L.
$
\end{Exa}




\section{Singularities of type $D$}
The standard map
$
h:=h_{D_k}:\R^{k-1}\to \R^{k}\,\,(k\ge 4)
$
of $D_k$-singular points is defined by
$
h(u,v,\mb x_3):=
\Big(
h_0(u,v,\mb x_3),
h_1(u,v,\mb x_3),
h_2(u,v,\mb x_3),\mb x_3
\Big),
$
where 
$\mb x_3:=(x_3,\ldots,x_{k-1})$
and
$h_i$ ($i=0,1,2$)
are given in
\eqref{eq:374a}, \eqref{eq:374b} and \eqref{eq:374c}. 
As an analogue of 
Proposition~\ref{prop:A182},
the following assertion holds:

\begin{Prop}\label{prop:D644}
The image of the standard map $h$ coincides with
the set
\begin{align*}
\mc W_{D_k}&:=\Big\{\mb x:=(x_0,\ldots,x_{k-1})\in \R^k\,;\, 
\text{there exists $(u,v)\in \R^2$ such that}
\\
& \phantom{aaaaaaaaaaaaaaaaaaaaaaaa}
\text{$F(u,v,\mb x)=F_u(u,v,\mb x)=F_v(u,v,\mb x)=0$}\Big\},
\end{align*}
where
\begin{equation}\label{eq:F397}
F(u,v,\mb x):=\pm u^{2}v+v^{k-1}+x_1u+x_0+\sum_{i=2}^{k-1}x_{i} v^{i-1}.
\end{equation}
\end{Prop}


\begin{proof}
We assume $\mb x\in \mc W_{D_k}$.
Since
$
F_u=\pm 2uv+x_1
$,
we have $x_1=\mp 2uv$, 
which corresponds to \eqref{eq:374b}.
On the other hand,
since
$$
F_v=\pm u^{2}+(k-1)v^{k-2}+x_2+\sum_{i=3}^{k-1}(i-1)x_{i} v^{i-2},
$$
we have
\begin{equation}\label{eq:673}
x_2=\mp u^2-(k-1)v^{k-2}-\sum_{i=3}^{k-1}(i-1)x_iv^{i-2},
\end{equation}
which corresponds to \eqref{eq:374c}.
Substituting this and $x_1=\mp 2uv$
into the equation $F=0$, we have
$$
x_0=\pm 2u^2 v+(k-2)v^{k-1}+\sum_{i=3}^{k-1}(i-2)x_iv^{i-1}.
$$
Since this corresponds to \eqref{eq:374a}, we have
$\mc W_{D_k}\subset h(\R^{k-1})$.
By reversing the above argument, we also obtain
the converse inclusion 
$\mc W_{D_k}\supset h(\R^{k-1})$. 
\end{proof}

\begin{Rmk}\label{rmk:433}
The $\pm$-ambiguity of the map $h$
affects the type of singular points
only when $k$ is even. In fact, if $k$ is odd, 
then $k-1$ is even,
and hence the replacement of $v$ by $-v$ does not affect
the leading  $v^{k-1}$ of $F$ 
in \eqref{eq:F397}
as a polynomial in $v$, but
replaces the term
$\pm u^{2}v$ 
by $\mp u^{2}v$.
So $h$ for $+$
is right-left equivalent to $h$ for $-$
when $k$ is odd.
On the other hand, if $k$ is even,
$h$ for $+$ is not right-left equivalent to $h$ for $-$.
\end{Rmk}

Set $\K=\R$ or $\C$ and define $\hat h=h$ or $h^\C$ 
as in \eqref{eq:hatH}.

\begin{Prop}\label{prop:D709}
For each $\mb x\in \K^k$,
the inverse image  $\hat h^{-1}(\mb x)$ 
consists of at most $k$ points.
Moreover, $\hat h^{-1}(\mb 0)=\{o\}$ 
holds.
Furthermore, $\hat h:\K^{k-1}\to \K^k$ 
is proper.
\end{Prop}

\begin{proof}
The statement $\hat h^{-1}(\mb 0)=\{o\}$ is easily checked.
First consider the case $x_1=0$.
If $v\ne 0$, then $u=0$.
By \eqref{eq:374a}, the possible values of $v$ with $v\ne 0$ 
are at most $k-3$. Including $v=0$, the possibilities for $v$ are at most $k-2$.

Hence we may assume $x_1\ne 0$.
Then $v\ne 0$, and by \eqref{eq:374b} we have
\begin{equation}\label{eq:809a}
u=\mp \frac{x_1}{2v}\in \K .
\end{equation}
Substituting this into \eqref{eq:673} shows that the possible values of $v$
are at most $k$, proving the first assertion.

It remains to prove the properness of $\hat h$.
Let $\mc K\subset \K^k$ be compact and fix 
$$
\mb x=(x_0,\ldots,x_{k-1})\in \mc K.
$$
Then there exists $C>0$ such that $\max(|x_0|,\dots,|x_{k-1}|)<C$.
If $v=0$, then by \eqref{eq:374c} we have $|x_2|=|u|^2$, 
hence $|u|<\sqrt{C}$,
so $\{v=0\}\cap \hat h^{-1}(\mc K)$ is bounded.
Assume $v\ne 0$. By \eqref{eq:374b}, $u=\mp x_1/(2v)$.
Substituting this into the relation $x_0=h_0(u,v,\mb x_3)$ yields
\[
x_0 v=(k-2)v^k+\left(\sum_{i=3}^{k-1}(i-2)\,x_i\, v^i\right) 
\ \pm \ \frac{x_1^2 }2 .
\]
By Remark~\ref{rmk:Cauchy}, $|v|$ is bounded in terms of $C$. 
Then from $x_2=h_2(u,v,\mb x_3)$,
we also obtain a bound for $|u|$. 
Hence $\{v\ne 0\}\cap \hat h^{-1}(\mc K)$ 
is bounded.
Therefore, $\hat h^{-1}(\mc K)$ is bounded.
Since $\hat h$ is 
continuous and $\mc K$ is closed, $\hat h^{-1}(\mc K)$ is closed.
So $\hat h^{-1}(\mc K)$ is compact.
Thus $\hat h$ is proper.
\end{proof}


The following is an analogue of
Proposition~\ref{prop:1c}:

\begin{Proposition}\label{prop:1cd}
The map $h$ is a $(k-1)$-dimensional real analytic map.
\end{Proposition}

\begin{proof}
We set
$$
\mc M:=\pmt{(h_0)_u & (h_0)_v \\
(h_2)_u & (h_2)_v
}.
$$
Substituting $x_3=\cdots=x_{k-1}=0$, 
we have
$$
\mc M=
\pmt{\pm 4uv & \pm 2u^2+(k-1)(k-2)v^{k-2} \\
\mp 2u & -(k-1)(k-2)v^{k-3}
}.
$$
Removing  the first three columns and
rows of the Jacobian matrix of $h$,
we obtain the identity matrix of rank $k-3$. 
To prove the assertion, it is sufficient to
show that $\mc M$ is of rank two for some point.
In fact, $\mc M$ is of rank $2$ at $(u,0)$ if $u\ne 0$.
\end{proof}

Substituting \eqref{eq:809a} into 
the definition of $F(u,v,\mb x)$,  
we define
\begin{equation}
\label{eq:tF}
A^{\mb x}(v):= v \, F\!\left(\mp\frac{x_1}{2v},\, v,\, \mb x\right).
\end{equation}
To simplify notation, we denote the derivative of this polynomial 
with respect to $v$ by
\begin{equation}\label{eq:Bx}
B^{\mb x}(v) := \frac{dA^{\mb x}(v)}{dv}.
\end{equation}


\begin{Prop}\label{new:L}
Let $\K$ be either $\R$ or $\C$. 
For $\mb x\in \K^k$,
there exists $(u,v)\in \K\times (\K\setminus \{0\})$
such that $\hat h(u,v,\mb x_3)=\mb x$
if and only if
there exists $v\in \K\setminus \{0\}$ such that
$A^{\mb x}(v)=B^{\mb x}(v)=0$.
\end{Prop}

\begin{proof}
Throughout the proof we assume $v\neq 0$.
Since $F_u=\pm 2uv+x_1$, we set $u(v):=\pm x_1/(2v)$.
Then, we can write
$
A^{\mb x}(v)=v F(u(v),v).
$
Since $F_u(u(v),v)=0$, differentiating it, we have
\begin{equation}\label{eq:1612}
B^{\mb x}(v)=A^{\mb x}_v(v)=F(u(v),v)+v F_v(u(v),v). 
\end{equation}
So $A^{\mb x}=B^{\mb x}=0$ holds under the assumption that
$F=F_u=F_v=0$. 
Conversely, suppose that
there exists $v\in \K\setminus \{0\}$ such that
$A^{\mb x}(v)=B^{\mb x}(v)=0$ holds.
For such a $v\in \K$,
we set $u:=u(v)$. Then 
$F_u(u(v),v)=0$, that is, we have $u=u(v)$.
Moreover, $A^{\mb x}(v)=0$ implies $F(u(v),v)=0$.
Since \eqref{eq:1612} implies that
$
B^{\mb x}(v)=v F_v(u(v),v)
$,
we also have
$
F_v(u(v),v)=0.
$
\end{proof}

We set
$
\Theta_{D_k}(\mb x):=\op{Res}_v(A^{\mb x},B^{\mb x}).
$


\begin{Theorem}\label{thm:D}
The image $\mc W_{D_k}$ of the standard map $h:=h_{D_k}$ $(k\ge 4)$
is a global main-analytic set of $\R^{k}$
whose main-analytic function is $\Theta_{D_k}$.
\end{Theorem}

\begin{proof}
We set
$
\mc L_1:=\{\mb x:=(x_0,x_1,\ldots,x_{k-1})\in \R^k\,;\, x_0=x_1=0\}.
$
If $v=0$, then
$
h(u,0,\mb x_3)=(0,0,\mp u^2,\mb x_3).
$
In particular, we have 
$$
h(\{(u,0,\mb x_3)\in \R^{k-1}\,;\, \mb x_3\in \R^{k-3}\})
=\{(0,0,x_2,\ldots,x_{k-1})\in \R^k\,;\, 
\mp x_2\ge 0\}\subset \mc L_1.
$$
On the other hand, we set
$$
\mc L_2:=\{\mb x \in \R^k\,;\, \text{there exists $v\in \C\setminus \R$
such that $A^{\mb x}(v)=B^{\mb x}(v)=0$}\}.
$$
By Proposition~\ref{new:L}, we have
$
h(\R^{k-1})\subset \Big(\mc Z(\Theta_{D_k})\setminus \mc L_2\Big)\cup \mc L_1.
$
If $x_0=x_1=0$, then $v=0$ is a double root of the polynomial 
$A^{\mb x}(v)$,
and hence $\Theta_{D_k}(\mb x)$ vanishes.
In particular, we have
$
h(\R^{k-1})\subset \mc Z(\Theta_{D_k})\setminus \mc L_2
$.
It is sufficient to prove that 
the set $\mc L_2$  is of dimension less than
$k-1$ in $\R^k$:
The polynomial $A^{\mb x}$ can be explicitly written as
\begin{align}
A^{\mb x}(v)&=
\left.
\pm u^2v^2+ v^{k}+x_{k-1}v^{k-1}+\cdots+x_2 v^2+x_1uv+x_0v 
\right|_{u=\mp x_1/(2v)} \label{eq:AD}
\\
&=v^{k}+x_{k-1}v^{k-1}+\cdots+x_2 v^2+x_0v
\mp\frac{1}4 x_1^2. \nonumber
\end{align}
We fix a non-real double root $\alpha$ of 
the equation $A^{\mb x}(v)=0$.
Then  $\overline \alpha$ is also a double root, 
and we can write
$$
A^{\mb x}(v)
=\Big(v^2-(\alpha+\bar \alpha)v+|\alpha|^2\Big)^2 \left(v^{k-4}
+\sum_{i=0}^{k-5} c_iv^i\right),
$$
where $c_0,\ldots, c_{k-5}\in \R$.
This expression implies that
$\mb x(\in \mc L_2)$
depends only on the
parameters $\alpha$ and $c_0,\ldots, c_{k-5}$.
So we have 
$$
\dim_H\,\mc L_2\le \dim_H \C+\dim_H \R^{k-4}=2+(k-4)=k-2,
$$
proving the assertion.
\end{proof}

\begin{Rmk}\label{rmk:d}
Since Proposition~\ref{new:L} holds for $\K=\C$, we have 
$h^\C(\C^{k-1}\setminus\mc Z_\C(v))\subset\mc Z_\C(\Theta_{D_k})$.
As $h^\C$ is proper, 
$$ 
h^\C(\C^{k-1})
=h^\C\bigl(\overline{\C^{k-1}\setminus\mc Z_\C(v)}\bigr)
=\overline{h^\C(\C^{k-1}\setminus\mc Z_\C(v))}
\subset\mc Z_\C(\Theta_{D_k}).
$$ 
Moreover, $h^0(u,v,\mb x_3)=h^1(u,v,\mb x_3)=0$ when $v=0$, 
and this case occurs only for $x_0=x_1=0$.  
Hence Proposition~\ref{new:L} also gives 
the inclusion 
$$
\mathcal Z_\C(\Theta_{D_k})\setminus \mc L
\subset h^\C(\C^{k-1})
\qquad \Big(\mc L:=\mathcal Z_\C(x_0)\cap\mathcal Z_\C(x_1)\Big).
$$
By the properness of $h^\C$, the image $h^\C(\C^{k-1})$ 
is closed in $\C^k$. 
Since $\mc L$ has empty interior in $\mathcal Z_\C(\Theta_{D_k})$, 
we have
\[
\mathcal Z_\C(\Theta_{D_k})
=\overline{\mathcal Z_\C(\Theta_{D_k})\setminus \mc L
}\subset
\overline{h^\C(\C^{k-1})}=h^\C(\C^{k-1}),
\]
and $h^\C(\C^{k-1})=\mathcal Z_\C(\Theta_{D_k})$.

We here prove that $\Theta_{D_k}$ ($k\ge 4$) is squarefree:
We fix $\sigma\in\{\pm1\}$ and $x_1\neq 0$, and consider
\[
A_s(v)=G(v)\bigl((v-x_1/2)^2+s\bigr),
\]
where $G(v)$ is a monic polynomial of degree $k-2$ with real
coefficients such that $G(0)=-\sigma$ and $G(x_1/2)\neq 0$.
Since $G$ does not depend on $s$, all roots of $G$ remain fixed as $s$
varies. In particular they stay simple. Hence the only 
roots of $A_s$ depending on $s$
are
$
v_\pm(s)=x_1/2\pm i\sqrt{s},
$
and
\[
v_+(s)-v_-(s)=2i\sqrt{s}.
\]
All other factors in the Vandermonde product stay nonzero and smooth
in $s$, so
\[
\Delta(A_s)=C\,s + O(s^2), \qquad C\neq 0.
\]
Thus, the discriminant vanishes to order exactly one at $s=0$, which
shows that $\Theta_{D_k}$ is squarefree.
Therefore, $\Theta_{D_k}$ coincides, up to a nonzero scalar,  
with the defining polynomial of $h^\C(\C^{k-1})$.  
In particular, it agrees with the discriminant polynomial of type $D_k$,  
which is known to be irreducible over $\C$.
\end{Rmk}


\begin{Exa}
Consider the case $k=4$ and fix $\sigma\in\{\pm1\}$. Set
\[
F(u,v,\mathbf{x}):=\sigma\,u^{2}v+v^{3}+x_1u+x_0+x_2v+x_3v^2.
\]
Substituting $u:=-\,x_1/(2\sigma v)$, we obtain
\[
A^{\mathbf{x}}(v)=v^4+x_3v^3+x_2v^2+x_0v\ -\tfrac{\sigma}4\,x_1^2.
\]
\end{Exa}


%%
\section{Singularities of type $E_6$}
As mentioned in Section 1, the standard map
$
h:=h_{E_6}:\R^{5}\to \R^{6}
$
of an $E_6$-singular point
is defined by
$$
h(u,v,\mb x_3):=
\Big(
h_0(u,v,\mb x_3),
h_1(u,v,\mb x_3),
h_2(u,v,\mb x_3),\mb x_3
\Big),
$$
where 
$\mathbf{x}_3 := (x_3, x_4, x_5)$ 
is the subvector of 
$\mathbf{x}=(x_0,\ldots, x_5) \in \mathbb{R}^6$
and
\begin{align}\label{eq:1022a}
h_0(u,v,\mb x_3)&:=
2u^3+3v^4+v^2x_3+uv\, \delta_6(v), \\
\label{eq:1022b}
h_1(u,v,\mb x_3)&:=-3u^2-vx_4-v^2x_5, \\
\label{eq:1022c}
h_2(u,v,\mb x_3)&:=
-4 v^3-2vx_3-\delta_6(v)u, \\
\label{eq:eq:d6}
\delta_6(v)&:=x_4+2v x_5.
\end{align}
The following is an analogue of 
Propositions~\ref{prop:A182} and \ref{prop:D644}:

\begin{Prop} \label{prop:e6}
The image of $h$ coincides with
the following set
\begin{align*}
\mc W_{E_6}&:=\Big\{\mb x:=(x_0,\ldots,x_{5})\in \R^6\,;\, 
\text{there exists $(u,v)\in \R^2$ such that}
\\
& \phantom{aaaaaaaaaaaaaaaaaaaaaaaaa}
\text{$F(u,v,\mb x)=F_u(u,v,\mb x)=F_v(u,v,\mb x)=0$}\Big\},
\end{align*}
where 
\begin{equation}\label{eq:E1047}
F(u,v,\mb x):=
u^3+v^4+x_5 uv^2+x_4 uv+x_3 v^2+x_2 v+x_1 u+x_0.
\end{equation}
\end{Prop}

\begin{proof}
Computing $F_u$ (resp.  $F_v$) and 
replacing $x_1$ by $h_1(u,v,\mb x_3)$
(resp. $x_2$ by $h_2(u,v,\mb x_3)$)
we obtain \eqref{eq:1022b} and \eqref{eq:1022c}.
On the other hand, $F_u=F_v=0$ imply
$$
x_1=-3 u^2 - x_4 v - x_5 v^2,\qquad x_2=-4 v^3 - 2 x_3v - x_4u - 2x_5 u v.
$$
By substituting them into $F=0$, we obtain \eqref{eq:1022a}.
\end{proof}

As in
Propositions~\ref{prop:1c} and \ref{prop:1cd},
the following assertion can be proved easily.

\begin{Prop}\label{prop:1101}
The map $h$ is a $5$-dimensional real analytic map.
\end{Prop}

Set $\K=\R$ or $\C$ and define $\hat h=h$ or $h^\C$ as in
the case of type D:

\begin{Prop}\label{prop:E1057}
The map $\hat h:\K^{5}\to \K^6$ 
is proper.
Moreover, for each $\mb x\in \K^6$,
the inverse image $\hat h^{-1}(\mb x)$ is finite.
Furthermore, $\hat h^{-1}(\mb 0)=\{o\}$.
\end{Prop}

\begin{proof}
The fact $\hat h^{-1}(\mb 0)=\{o\}$ is immediate.
Write $x_i=h_i(u,v,x_3,x_4,x_5)$ ($i=0,1,2$) as 
in \eqref{eq:1022a} and \eqref{eq:eq:d6}.
From \eqref{eq:1022b}, we have
\begin{equation}\label{eq:194a}
u^2=\frac{-x_1-vx_4-v^2x_5}{3}.
\end{equation}
If $\delta_6(v):=x_4+2vx_5=0$, then \eqref{eq:1022c} 
gives $x_2=-4v^3-2vx_3$, so $v$ is a root of a cubic. 
Hence $v$ and then $u$ are finite in number.
If $\delta_6(v)\ne 0$, then from \eqref{eq:1022c}
\begin{equation}\label{eq:1034}
u=-\frac{4v^3+2vx_3+x_2}{\delta_6(v)}.
\end{equation}
Substituting \eqref{eq:1034} into \eqref{eq:194a}, we have
\begin{align}\label{eq:E6-A}
A^{\mb x}(v):={}&
48 v^6 + 4 (12 x_3 + x_5^3) v^4 + 8 (3 x_2 + x_4 x_5^2) v^3  \\
&\quad + (4 x_1 x_5^2 + 5 x_4^2 x_5 + 12 x_3^2) v^2  \nonumber \\
&+ (4 x_1 x_4 x_5 + 12 x_2 x_3 + x_4^3) v + x_1 x_4^2 + 3 x_2^2. \nonumber
\end{align}
The leading term is $48v^6$, so $v$ (hence $u$) has finitely many possibilities.
Thus $\hat h^{-1}(\mb x)$ is finite.

For properness, let $\mc K\subset\K^6$ be 
compact and assume $\mb x=\hat h(u,v,\mb x_3)\in \mc K$.
If $\delta_6(v)=0$, the cubic equation in $v$ above yields a bound 
on $|v|$ by Remark \ref{rmk:Cauchy}.
By
\eqref{eq:194a},
$|u|$ is bounded.
Therefore $\hat h^{-1}(\mc K)$ is bounded and closed, hence compact. 
Thus $\hat h$ is proper.
\end{proof}

Substituting  \eqref{eq:194a} and \eqref{eq:1034}
into the equation $h_0(u,v, x_3,\ldots,x_5)=x_0$
(cf. \eqref{eq:1022a}), we obtain
the equation
$B^{\mb x}(v)=0$, where $\mb x=(x_0,\ldots,x_5)\in \R^6$ and
\begin{align}
B^{\mb x}(v)&:=2 x_5v^5  +5 x_4v^4 +(8 x_1-2 x_3 x_5)v^3 
+(x_3 x_4-4 x_2x_5)v^2  \\
\nonumber
&\phantom{aaaaaaaaaaaaa}+(-6 x_0 x_5+4 x_1 x_3-x_2 x_4)v 
-3 x_0 x_4+2x_1 x_2.
\end{align}

If $h(u,v,\mb x_3)=\mb x$,
then the two polynomials 
$A^{\mb x}(v)$ and $B^{\mb x}(v)$ must have a common root in $\R$,
and the resultant
\begin{equation}\label{R6}
\mc R(x_0,\ldots,x_{5}):=\op{Res}_v(A^{\mb x},B^{\mb x})
\end{equation}
must vanish. For the latter discussions, 
we also set (cf. \eqref{eq:550})
\begin{equation}\label{S6}
\mc S(x_0,\ldots,x_{5}):=\op{Psc}_v(A^{\mb x},B^{\mb x}).
\end{equation}


Corresponding to \eqref{eq:1022a}, \eqref{eq:1022b} and \eqref{eq:1022c},
for each $\mb x:=(x_0,\ldots,x_5)\in \R^6$,  we set
\begin{align}\label{eq:1022a1}
g_0^{\mb x}(u,v)&:=2u^3+3v^4+v^2x_3-x_0+uv\, \delta_6(v), \\
\label{eq:1022b1}
g_1^{\mb x}(u,v)&:=3u^2+vx_4+v^2x_5+x_1, \\
\label{eq:1022c1}
g_2^{\mb x}(u,v)&:=-4 v^3-2vx_3-x_2-\delta_6(v)u.
\end{align}
By definition, we have the following:
Set $\K=\R$ or $\C$ and define $\hat h=h$ or $h^\C$.

\begin{Proposition}\label{prop:1742}
For $\mb x\in \K^6$ and $u,v\in \K$,
$\hat h(u,v,\mb x_3)=\mb x$ is equivalent to 
the simultaneous vanishing of 
$g_i^{\mb x}(u,v)$ $(i=0,1,2)$.
\end{Proposition}

We set
$
r_6(\mb x):=-\frac14\op{Res}_v(g_2^{\mb x},\delta_6),
$
then 
a direct computation with \textit{Mathematica} yields
\begin{equation}
r_6(\mb x)=x_4^3+2 x_3 x_5^2x_4-2 x_2 x_5^3.
\end{equation}


\begin{Prop}\label{prop:1366}
Fix $\mb x\in \K^6$. If $\mb x\in \hat{h}(\K^5)$, 
then there exists $v\in \K$ satisfying $A^{\mb x}(v)=B^{\mb x}(v)=0$.
Conversely, if
there exists $v\in \K$ satisfying $A^{\mb x}(v)=B^{\mb x}(v)=0$
and $\delta_6(v)\ne 0$,
then $\mb x\in \hat h(\K^5)$.
\end{Prop}

\begin{proof}
The first statement of the proposition follows from Lemma~\ref{LemC1}.
To prove the second statement,
we set
$G_i(u,v):=g_i^{\mb x}(u,v)$ 
($i=0,1,2$) and $\delta:=\delta_6$ for fixed $\mb x\in \K^6$ 
in Appendix~C.
Then the two polynomials $\alpha(v)$ and $\beta(v)$ given in 
\eqref{eq:C4} and \eqref{eq:C6} 
coincide with $A^{\mb x}(v)$ and $B^{\mb x}(v)$ respectively,
and Proposition~\ref{thm:Cm}
implies the conclusion.
In fact, $\mb x\in h(\K^5)$ if and only if
$\hat h(u,v,\mb x_3)=\mb x$, which is equivalent to 
the condition that $g_i^{\mb x}(u,v)=0$ ($i=0,1,2$).
\end{proof}

The following assertion follows from
Proposition~\ref{prop:1366} immediately:

\begin{Corollary}\label{eq:1366}
The image $h(\R^5)$ $($resp. $h^\C(\C^5))$
is a subset of $\mc Z(\mc R)$ $($resp. $\mc Z_\C(\mc R))$.
\end{Corollary}


In \eqref{R6} and \eqref{S6}, we defined the two polynomials
$\mc R(\mb x)$ and $\mc S(\mb x)$. Here we regard them as polynomials
in $x_0$. Then we obtain
\begin{align}
\label{eq:1316a}
\mc R(\mb x) &=
2^{20}3^{11} r_6(\mb x)^2 x_0^6
   + \text{(lower-order terms in $x_0$)}, \\
\label{eq:1316aa}
\mc S(\mb x) &=
-2^{21}3^{9} x_5^5 x_0^5
+ \text{(lower-order terms in $x_0$)}.
\end{align}
So, 
if we set
\begin{equation}\label{eq:1834}
\mc T:=
\{\mb x\in \R^6\,;\, r_6(\mb x)=x_5=0\}
=\{\mb x\in \R^6\,;\, x_4=x_5=0\},
\end{equation}
then the leading terms of
$\mc R(\mb x)$
and $\mc S(\mb x)$
as polynomials in $x_0$ vanish at the same time
if and only if $\mb x\in \mc T$.
From the right-hand side of 
\eqref{eq:1834}, 
it is clear that the dimension of $\mc T$ is less than $5$.

\begin{Thm}\label{thm:E6}
The dimension of the set 
$\mc E_1:=\mc T\cup \Big(\mc Z(\mc R)\cap \mc Z(\mc S)\Big)$
is less than $5$, and
\begin{equation}\label{eq:E1X}
\Xi_{E_6}:=\mc Z(\mc R)\setminus \mc E_1
\end{equation}
is an open subset of $\mc Z(\mc R)$.
Moreover, for each $\mb x\in \Xi_{E_6}$,
there exists a unique real number $v(\mb x)$ 
which is a common root of $A^{\mb x}(v)$ and $B^{\mb x}(v)$.
\end{Thm}

\begin{proof}
Corollary \ref{eq:1366} 
together with 
Proposition \ref{prop:1101}
implies $\dim_H \mc Z(\mc R)\ge 5$.
We remark that $r_6(\mb x)$ does not involve $x_0$.
Since $\mc R(\mb x)$ is a non-constant polynomial,
$\dim_H \mc Z(\mc R)\le 5$.
So, we can conclude that
\begin{equation}\label{eq:2058}
\dim_H \mc Z(\mc R)=5.
\end{equation}

We set
$
\phi(x_1,\ldots,x_5):=\op{Res}_{x_0}(\mc R,\mc S).
$
Then it can be computed that
$$
\phi(0,0,0,1,1)\equiv 2 \pmod{5}.
$$
By Corollary \ref{cor:A}
with \eqref{eq:1834}, we can conclude that
$$
\mc L:=
\{
\mb x\in \R^6\setminus \mc T\,;\, 
\mc R(\mb x)=\mc S(\mb x)=0
\}
$$
has dimension less than $5$.
Since we have already observed that $\dim_H(\mc T)<5$,
it follows that $\dim_H(\mc E_1)<5$ since $\mc E_1=\mc T\cup \mc L$.

In particular, $\Xi_{E_6}$ is an open subset of $\mc Z(\mc R)$.
We fix $\mb x\in \Xi_{E_6}$.
Since $\mb x\in \mc Z(\mc R)\setminus \mc T$,
there exists $v\in \C$ such that
$$
A^{\mb x}(v)=B^{\mb x}(v)=0.
$$
If $v$ were not unique, then we would have 
$\mc S(\mb x)=0$, which contradicts $\mb x\in \Xi_{E_6}$. 
Hence $v$ is uniquely determined, and we denote it by $v(\mb x)$.
Moreover, if $v(\mb x)$ were non-real, its complex conjugate would also be 
a common root of $A^{\mb x}(v)$ and $B^{\mb x}(v)$, leading again to 
$\mc S(\mb x)=0$, a contradiction. 
Therefore $v(\mb x)\in \mathbb{R}$.
\end{proof}

We have not yet determined the main-analytic function, 
but the following statement can already be established.


\begin{Corollary}\label{2079}
The set $h(\R^5)$ is a global main-analytic set of $\R^6$.
\end{Corollary}

\begin{proof}
Set $U := h^{-1}(\R^6 \setminus \mathcal E_1)$. 
By the last statement of Theorem~\ref{thm:E6}, the map $h|_U$ is injective.
To apply Proposition~\ref{thm:Coste-main-analytic-J},
it suffices to show that $U$ is dense in $\R^5$.
Otherwise, $h^{-1}(\mathcal E_1)$ would have a nonempty interior,
and hence $\dim_H(h^{-1}(\mathcal E_1)) = 5$.
Since $h$ is $m$-dimensional, this 
with Proposition~\ref{prop:HDIM}
implies $\dim_H(\mathcal E_1) = 5$,
contradicting the first statement of Theorem~\ref{thm:E6}.
\end{proof}


We next try to find an explicit formula for the main-analytic function.

\begin{Prop}\label{thm:E6main}
The polynomial $\mc R(\mb x)$ is divisible by $r_6(\mb x)^2$ and
$$
\Theta_{E_6}(\mb x):=\frac{\mc R(\mb x)}{r_6(\mb x)^2}
$$
is irreducible over $\Q$.
\end{Prop}

\begin{proof}
Using \textit{Mathematica}, we can check that
$\mc R(\mb x)$ is divisible by $r_6(\mb x)^2$,
and $\mc R(\mb x)/r_6(\mb x)^2$
is irreducible
over $\Q$.
\end{proof}

Since $r_6$ and $\mc R(\mb x)$ are both irreducible over $\Q$,
the following assertion follows from
Proposition~\ref{Prop:B0}
in the appendix:

\begin{Corollary}\label{eq:1770}
The dimension of the set
$
\mc E_0:=\mc Z(r_6)\cap \mc Z(\Theta_{E_6})
$
is less than $5$.
\end{Corollary}


\begin{Rmk}\label{rmk:1430}
Each coefficient of $A^{\mathbf{x}}(v)$ does not involve $x_0$, 
while each coefficient of $B^{\mathbf{x}}(v)$ involves $x_0$ 
but not $x_0^i$ for any $i \ge 2$.  
We define $B_0^{\mathbf{x}}(v)$ to be the polynomial obtained from $B^{\mathbf{x}}(v)$  
by retaining only those terms containing $x_0$, that is,
\begin{equation}\label{eq:B0}
B_0^{\mathbf{x}}(v):=x_0 \left(\frac{d}{dx_0}B^{\mathbf{x}}(v)\right).
\end{equation}
Then
$
  B_0^{\mathbf{x}}(v) = -3(2x_5v + x_4)x_0.
$
By this definition of $B_0^{\mathbf{x}}(v)$, in principle,  
the leading term of $\operatorname{Res}_v(A^{\mathbf{x}}, B_0^{\mathbf{x}})$  
(resp.\ $\operatorname{Res}^{(1)}_v(A^{\mathbf{x}}, B_0^{\mathbf{x}})$)  
with respect to $x_0$ must be a nonzero scalar multiple of the corresponding 
leading term of  
$\mathcal{R}(\mathbf{x})$ (resp.\ $\mathcal{S}(\mathbf{x})$).  
Indeed, by comparing
\begin{align*}
\operatorname{Res}_v(A^{\mathbf{x}}, B^{\mathbf{x}}_0) 
   &= 2^4 \cdot 3^{7} \, r_6(\mathbf{x})^2 \, x_0^6 +
     \text{(lower-order terms in $x_0$)}, \\[-0.5em]
\operatorname{Psc}_v(A^{\mathbf{x}}, B^{\mathbf{x}}_0) 
   &= -2^5 \cdot 3^{5} \, x_5^5 \, x_0^5
\end{align*}
with \eqref{eq:1316a} and \eqref{eq:1316aa},  
this principle can be verified directly.  

Since $B^{\mathbf{x}}_0$ is simpler than $B^{\mathbf{x}}$,  
this technique reduces the computation of the leading terms of  
$\mathcal{R}(\mathbf{x})$ and $\mathcal{S}(\mathbf{x})$.  
We will exploit this simplification in Section~7.
\end{Rmk}

If we solve $\delta_6(v)=0$, we have
$
v:=-{x_4}/(2x_5).
$
By substituting it into $g_0^{\mb x}(u,v)$ 
and $g_1^{\mb x}(u,v)$, the two polynomials
\begin{align*}
k^{\mb x}_0(u)&:=8 x_5^3u^3 -4 x_0 x_5^3-x_4^4+x_4^3 x_5, \\[-0.5em]
k^{\mb x}_1(u)&:=24 x_5^3u^2 - 8 x_1 x_5^3 - 4 x_4 x_5^3 - x_4^3
\end{align*}
are obtained, and their resultant
\begin{equation}\label{eq:H6}
\mc H(\mb x):=\op{Res}_u(k^{\mb x}_0,k^{\mb x}_1)\qquad (\mb x\in \R^6)
\end{equation}
is a polynomial not containing $x_2$ as its variable.

\begin{Thm}\label{thm:1954}
Fix $\mb x\in \Xi_{E_6}$ $($cf. \eqref{eq:E1X}$)$.
\begin{enumerate}
\item 
If $r_6(\mb x)\ne 0$, then
$\delta_6(v(\mb x))\ne 0$ holds,
and then there exists $u\in \R\setminus \{0\}$ 
such that $h(u,v(\mb x),\mb x_3)=\mb x$.
In particular, we have
\begin{equation}\label{eq:6Xb}
\Xi_{E_6}\setminus \mc Z(r_6)\subset h(\R^5).
\end{equation}
\item 
Suppose that
$\mb x\in h(\R^5)\cap \mc Z(r_6)$.
Then $\delta_6(v(\mb x))=0$ and $\mc H(\mb x)=0$ hold.
Moreover, the dimension of the set
$
h(\R^5)\cap \mc Z(r_6)
$
is less than $5$.
\end{enumerate}
\end{Thm}

\begin{proof}
Since $\mb x\in \Xi_{E_6}$, the component $x_5$ never vanishes, and
we can write
\begin{align}\label{eq:1948a}
A^{\mb x}(v)&=P_1^{\mb x}(v)\delta_6(v)+\frac{3r_6(\mb x)^2}{4 x_5^6}, \\[-0.5em]
\label{eq:1948b}
B^{\mb x}(v)&=P_2^{\mb x}(v)\delta_6(v)+\frac{(x_4^2-4x_1x_5)r_6(\mb x)}{4 x_5^4},
\end{align}
where $P_j^{\mb x}(v)$ ($j=1,2$) are certain polynomials.
We assume $r_6(\mb x)\ne 0$.
If $\delta_6(v(\mb x))=0$, then $v(\mb x)=-x_4/(2x_5)$.
Substituting $v:=v(\mb x)$ into $A^{\mb x}(v)$, we have
$$
0=A^{\mb x}(v(\mb x))=\frac{3r_6(\mb x)^2}{4 x_5^6},
$$
which implies $r_6(\mb x)=0$, a contradiction.
So  $\delta_6(v(\mb x))\ne 0$, and
the remaining assertions of (1) follow from 
the second assertion of Proposition~\ref{prop:1366}.

We next consider the case that 
$\mb x\in h(\R^5)\cap \mc Z(r_6)$.
Since $\delta_6(v)=0$ is a linear equation,
$\hat v(\mb x):=-x_4/(2 x_5)$ is the solution.
Then \eqref{eq:1948a} and \eqref{eq:1948b}
imply that  
$
A^{\mb x}(\hat v(\mb x))=B^{\mb x}(\hat v(\mb x))=0.
$
So the uniqueness of $v(\mb x)$ 
(cf. Theorem~\ref{thm:E6})
yields that 
$
v(\mb x)=\hat v(\mb x)=-x_4/(2 x_5).
$
Then we have $\delta_6(v(\mb x))=0$ and
$$
g^{\mb x}_2(u,v(\mb x))=-4 v^3 - 2 v x_3 - x_2\Big |_{v=-x_4/(2x_5)}
=\frac{-r_6(\mb x)}{2x_5^3}=0.
$$
If $\mb x\in h(\R^5)\cap \mc Z(r_6)$, then
by Proposition~\ref{prop:1742},
there exists $u\in \R$
satisfying
$$
k^{\mb x}_0(u)=g^{\mb x}_0(u,v(\mb x))=0,\qquad
k^{\mb x}_1(u)=g^{\mb x}_1(u,v(\mb x))=0,
$$
which imply that 
$\mc H(\mb x)=0$ (cf. \eqref{eq:H6}), proving the first part of (2).
Since $\mc H(\mb x)$ does not contain $x_2$ as its variable
but $r_6(\mb x)$ does,  Lemma~\ref{lem:B0} in the appendix yields that
the dimension of 
$\mc Z(r_6)\cap \mc Z(\mc H)$ is less than $6$.
Since
$$
h(\R^5)\cap \mc Z(r_6)\subset \Big(\mc Z(r_6)\cap \mc Z(\mc H)\Big)\cup \mc E_1,
$$
the first statement of Theorem~\ref{thm:E6} yields 
$
\dim_H (h(\R^5)\cap \mc Z(r_6))<5
$.
\end{proof}

\begin{Corollary}\label{cor:2024}
Fix $\mb x\in \mc Z_\C(\Theta_{E_6})$.
If $r_6(\mb x)\ne 0$, then
there exist $u,v\in \C$ 
such that $h(u,v,\mb x_3)=\mb x$.
\end{Corollary}

\begin{proof}
The condition $\Theta_{E_6}(\mb x)=0$ implies
the existence of $v\in \C$ satisfying
$A^{\mb x}(v)=B^{\mb x}(v)=0$.  
Since $r_6(\mb x)\ne 0$, we have $\delta_6(v)\ne 0$.  
Then we can set $u\in \C$ by
\eqref{eq:1034}.
Since $A^{\mb x}(v)=0$, this $u$ also satisfy
\eqref{eq:194a}.
This implies the existence of $u\in \C$ 
such that $h(u,v,\mb x_3)=\mb x$.
\end{proof}




We now arrive at the following assertion:

\begin{Theorem}
The polynomial $\Theta_{E_6}(\mathbf{x})$ is a main-analytic 
function of $h(\mathbb{R}^5)$.
\end{Theorem}

This provides an alternative proof of the main-analyticity of
$h(\R^5)$, without use of Appendix~D.

\begin{proof}
By Proposition \ref{thm:E6main}, 
$\mathcal{Z}(\mathcal{R})=\mathcal{Z}(r_6)\cup \mathcal{Z}(\Theta_{E_6})$ holds.
Thus, from \eqref{eq:E1X} and \eqref{eq:6Xb} we obtain
\[
\mathcal{Z}(\Theta_{E_6})\setminus (\mathcal{E}_0\cup \mathcal{E}_1)
=
\mathcal{Z}(\mathcal{R})\setminus (\mathcal{E}_1\cup \mathcal{Z}(r_6))
=\Xi_{E_6}\setminus \mathcal{Z}(r_6)\subset h(\mathbb{R}^5).
\]
Together with Corollary \ref{eq:1770} and
Theorem~\ref{thm:E6},
this yields
\begin{equation}\label{eq:2074}
\dim_H \bigl(\mathcal{Z}(\Theta_{E_6})\setminus h(\mathbb{R}^5)\bigr)<5.
\end{equation}
Moreover, Corollary~\ref{eq:1366} implies
\begin{equation}\label{eq:2023}
h(\mathbb{R}^5)\subset \mathcal{Z}(\Theta_{E_6})\cup \mathcal{E}_2,
\qquad \mathcal{E}_2:=h(\mathbb{R}^5)\cap \mathcal{Z}(r_6).
\end{equation}
By (2) of Theorem~\ref{thm:1954}, we know that $\dim_H(\mathcal{E}_2)<5$.
Suppose that $h^{-1}(\mathcal{E}_2)$ has non-empty interior.  
Since $h$ is a $5$-dimensional real analytic map, this would imply that
$
h\bigl(h^{-1}(\mathcal{E}_2)\bigr)
$
has dimension $5$ (cf.~Proposition~\ref{prop:HDIM}), 
contradicting $\dim_H(\mathcal{E}_2)<5$.

Now fix $\mathbf{x}\in \mathcal{E}_2$.  
Then there exists $u\in \mathbb{R}$ such that
$
h(u, v(\mathbf{x}), \mathbf{x}_3)=\mathbf{x}.
$
Since $h^{-1}(\mathcal{E}_2)$ has no interior points, we can choose a sequence
\[
\{(u_n,v_n,\mathbf{z}_n)\}_{n=1}^\infty 
\subset \mathbb{R}^5\setminus h^{-1}(\mathcal{E}_2)
\]
converging to $(u,v(\mathbf{x}),\mathbf{x}_3)$.  
By \eqref{eq:2023}, we then have
$
h(u_n,v_n,\mathbf{z}_n)\in \mathcal{Z}(\Theta_{E_6}),
$
which implies
\[
\mathbf{x}=\lim_{n\to\infty}h(u_n,v_n,\mathbf{z}_n)\in \mathcal{Z}(\Theta_{E_6}).
\]
Therefore, $\mathcal{E}_2\subset \mathcal{Z}(\Theta_{E_6})$.  
Combining this with \eqref{eq:2023}, we conclude that
$\Theta_{E_6}$ is a main-analytic function of $h(\R^5)$.
\end{proof}

\begin{Rmk}\label{rmk:e6}
By Proposition~\ref{prop:1366}, we have
$h^\C(\C^5\setminus \mc Z_\C(\delta_6))\subset \mc Z_\C(\Theta_{E_6})$.
Since $h^\C$ is proper, we obtain
\begin{equation}\label{eq:E6hc}
h^\C(\C^5)
=h^\C\bigl(\,\overline{\C^5\setminus \mc Z_\C(\delta_6)}\,\bigr)
=\overline{h^\C(\C^5\setminus \mc Z_\C(\delta_6))}
\subset \mc Z_\C(\Theta_{E_6}).
\end{equation}

Let $\mc L:=\mc Z_\C(\Theta_{E_6})\cap \mc Z_\C(r_6)$. 
Since the complex dimension of $\mc L$ is less than~$5$
(by the same reason as in the proof of 
Corollary~\ref{eq:1770}),
$\mc L$ has no interior point.
On the other hand,
Corollary~\ref{cor:2024} implies that
$\mc Z(\Theta_{E_6})\setminus \mc L\subset h^\C(\C^5)$.
By the properness of $h^\C$, it follows that
\begin{equation}\label{eq:E6hc2}
\mc Z_\C(\Theta_{E_6})=\overline{\mc Z(\Theta_{E_6})\setminus \mc L}
\subset \overline{h^\C(\C^5)}=h^\C(\C^5).
\end{equation}
Combining \eqref{eq:E6hc} and \eqref{eq:E6hc2}, we have
$
h^\C(\C^{5})=\mathcal{Z}_\C(\Theta_{E_6})\subset\C^{6}.
$

Since $\Theta_{E_6}$ is an irreducible polynomial over~$\Q$
(cf.~Proposition~\ref{thm:E6main}),
it coincides with the defining polynomial of
$h^\C(\C^{5})$. 
In particular, up to a nonzero scalar,
$\Theta_{E_6}$ coincides with
the discriminant polynomial of type $E_6$
which is known to be irreducible over $\C$.

Assigning weights $(12,8,9,6,5,2)$ to $(x_0,\ldots,x_5)$, we obtain:
\begin{itemize}
\item $r_6(\mb x)$ is a weighted homogeneous polynomial of degree $15$,
\item $\op{Res}_v(A^\mb x,B^\mb x)$ is weighted homogeneous of degree $102$,
\item the main-analytic  function $\Theta_{E_6}$ is weighted homogeneous of degree $72$.
\end{itemize}
\end{Rmk}




\section{Singularities of type $E_7$}

The standard map
$
h:=h_{E_7}:\R^{6}\to \R^{7}
$
of $E_7$-singular points
is defined by
$$
h(u,v,\mb x_3):=
\Big(
h_0(u,v,\mb x_3),
h_1(u,v,\mb x_3),
h_2(u,v,\mb x_3),\mb x_3
\Big),
$$
where 
$\mathbf{x}_3 := (x_3, x_4, x_5, x_6)$ is the subvector of 
$\mathbf{x} \in \mathbb{R}^7$,
and
\begin{align}\label{eq:1314a}
h_0(u,v,\mb x_3)&:=
2 u^3 +x_3v^2  + 2 x_4v^3  + u v\, \delta_7(v), \\
\label{eq:1314b}
h_1(u,v,\mb x_3)&:=-3 u^2 - v^3 - x_5v  -x_6 v^2, \\
\label{eq:1314c}
h_2(u,v,\mb x_3)&:=
- 2 x_3v - 3 x_4v^2  -u \, \delta_7(v), \\
\label{eq:d7}
\delta_7(v)&:=3 v^2+2 x_6v+x_5.
\end{align}
As an analogue of Proposition~\ref{prop:e6},
the following assertion holds:

\begin{Prop}\label{prop:e7}
The image of the standard map $h$ coincides with
the following set
\begin{align*}
\mc W_{E_7}&:=\Big\{\mb x:=(x_0,\ldots,x_{6})\in \R^7\,;\, 
\text{there exists $(u,v)\in \R^2$ such that}
\\
& \phantom{aaaaaaaaaaaaaaaaaaaaaaaa}
\text{$F(u,v,\mb x)=F_u(u,v,\mb x)=F_v(u,v,\mb x)=0$}\Big\},
\end{align*}
where 
\begin{align}
\label{eq:E1047b}
F(u,v,\mb x)&:= u^3+u v^3+x_6u v^2+x_5u v 
+x_4v^3 + 
x_3v^2 +x_2 v+x_1u+x_0.
\end{align}
\end{Prop}

As in
Propositions~\ref{prop:1c}, \ref{prop:1cd}
and \ref{prop:1101},
the following assertion holds:

\begin{Prop}\label{prop:1404}
The map $h$ is a $6$-dimensional real analytic map.
\end{Prop}


\begin{Prop}\label{prop:E1073}
The map $h:\R^{6}\to \R^7$ 
$($resp.\ $h^\C:\C^{6}\to \C^7)$
is proper.
Moreover, for each $\mb x\in \R^7$ $($resp.\ $\mb x\in \C^7)$, 
the inverse image $h^{-1}(\mb x)$ is finite.
Furthermore, $h^{-1}(\mb 0)=\{o\}$ 
$($resp.\ $(h^\C)^{-1}(\mb 0)=\{o\})$.
\end{Prop}

Set $\K=\R$ or $\C$ and define $\hat h$ as in the previous sections:


\begin{Prop}\label{prop:E7-proper}
The map $\hat h:\K^{6}\to \K^7$ is proper.
Moreover, for each $\mb x\in \K^7$, 
the inverse image $\hat h^{-1}(\mb x)$ is finite.
Furthermore, $\hat h^{-1}(\mb 0)=\{o\}$.
\end{Prop}

\begin{proof}
The fact $\hat h^{-1}(\mb 0)=\{o\}$ is obvious.
Write $x_i=h_i(u,v,x_3,x_4,x_5,x_6)$ ($i=0,1,2$).
If $\delta_7(v):=3v^2+2x_6v+x_5=0$, then $v$ is a 
root of a quadratic, hence finite. 
Since
\begin{equation}\label{eq:1364a}
u^2=\tfrac13(-v^3-x_6v^2-x_5v-x_1),
\end{equation}
the possibility of $u$ is finite.
If $\delta_7(v)\ne0$, we have
\begin{equation}\label{eq:1364b}
u=\frac{-x_2-2x_3v-3x_4v^2}{\delta_7(v)}.
\end{equation}
By \eqref{eq:1364a} and \eqref{eq:1364b}, we have
\begin{align}\label{eq:E7-A}
A^{\mb x}(v):={}&
9 v^7+21 x_6 v^6+(15 x_5+16 x_6^2)v^5 \\
&+(9 x_1+27 x_4^2+22 x_5 x_6+4 x_6^3)v^4 \nonumber\\
&+(12 x_1 x_6+36 x_3 x_4+7 x_5^2+8 x_5 x_6^2)v^3 \nonumber\\
&+(6 x_1 x_5+4 x_1 x_6^2+18 x_2 x_4+12 x_3^2+5 x_5^2 x_6)v^2 \nonumber\\
&+(4 x_1 x_5 x_6+12 x_2 x_3+x_5^3)v+x_1 x_5^2+3 x_2^2. \nonumber
\end{align}
Since the leading term is $9v^7$, 
the possibility of  $v$ (and $u$) is finite. 
Thus $h^{-1}(\mb x)$ is finite.

For properness, let $\mc K\subset\K^7$ be compact and 
assume $\mb x=\hat h(u,v,\mb x_3)\in \mc K$.
If $\delta_7(v)=0$, the quadratic bounds $|v|$ by
Remark \ref{rmk:Cauchy},
and then $|u|$ is bounded.
If $\delta_7(v)\ne0$, then $v$ satisfies \eqref{eq:E7-A} 
with leading term $9v^7$, hence $|v|$ is bounded; 
consequently $|u|$ is bounded.
Therefore $\hat h^{-1}(\mc K)$ is 
bounded and closed, hence compact. Thus $\hat h$ is proper.
\end{proof}


Substituting 
\eqref{eq:1364a} and \eqref{eq:1364b}
into the equation 
$h_0(u,v, x_3,\ldots,x_6)=x_0$
(cf.~\eqref{eq:1314a}) 
and eliminating the term $\delta_7(v)$, we obtain
the equation  $B^{\mb x}(v)=0$, where
$\mb x=(x_0,\ldots,x_6)$ and
\begin{align}
B^{\mb x}(v)&:=
-3 x_4 v^5-5 x_3 v^4+(-7 x_2-2 x_3 
x_6+3 x_4 x_5)v^3 \\ \nonumber
&\phantom{aaa}+(-9 x_0+6 x_1 x_4-4 x_2 x_6+x_3 x_5)v^2
\\ \nonumber
&\phantom{aaaaaa}
+(-6 x_0 x_6+4 x_1x_3-x_2 x_5)v -3 x_0 x_5+2 x_1 x_2.
\end{align}

By following the proof of Proposition~\ref{prop:1366},
we obtain the following:

\begin{Prop}\label{prop:1322e7}
Fix $\mb x\in \K^7$. If $\mb x\in \hat h(\K^6)$ holds, 
then there exists $v\in \K$ satisfying $A^{\mb x}(v)=B^{\mb x}(v)=0$.
Conversely, if
there exists $v\in \K$ satisfying $A^{\mb x}(v)=B^{\mb x}(v)=0$
and $\delta_7(v)\ne 0$,
then $\mb x\in \hat h(\K^6)$.
\end{Prop}

\begin{Corollary}\label{eq:1366e7}
The image $h(\R^6)$ $($resp. $h^\C(\C^6))$
is a subset of $\mc Z(\mc R)$ $($resp. $\mc Z_\C(\mc R))$.
\end{Corollary}

We set (cf. \eqref{eq:550})
$$
\mc R(\mb x):=\op{Res}_v(A^{\mb x},B^{\mb x}),\quad
\mc S(\mb x):=\op{Psc}_v(A^{\mb x},B^{\mb x})
\qquad (\mb x:=(x_0,\ldots,x_{6})\in \R^7).
$$
Since the leading term of $A^{\mb x}(v)$ with respect to $v$ is $9v^7$,  
the condition $\mc R(\mb x)=0$ (resp.~$\mc R(\mb x)=\mc S(\mb x)=0$) implies that  
$A^{\mb x}(v)$ and $B^{\mb x}(v)$ have at least one (resp.~two) common root(s).  
Corresponding to \eqref{eq:1314a}, \eqref{eq:1314b} and \eqref{eq:1314c}, we set
\begin{align}\label{eq:1022a7}
g^{\mb x}_0(u,v)&:=2u^3+2x_4v^3+x_3v^2-x_0+uv\, \delta_7(v), \\
\label{eq:1022b7}
g^{\mb x}_1(u,v)&:=3u^2-v^3-x_6 v^2-x_5v-x_1, \\
\label{eq:1022c7}
g^{\mb x}_2(u,v)&:=-3x_4v^2 -2x_3v-x_2-\delta_7(v)u.
\end{align}
By definition, we have:

\begin{Proposition}\label{prop:1742E7}
For $\mb x\in \K^7$ and $u,v\in \K$,
$\hat h(u,v,\mb x_3)=\mb x$ is equivalent to 
the simultaneous vanishing of 
$g_i^{\mb x}(u,v)$ $(i=0,1,2)$.
\end{Proposition}


We set
$
r_7(\mb x):=\op{Res}_v(g_2^{\mb x},\delta_7)/3,
$
then a direct computation with \textit{Mathematica} yields
$$
r_7(\mb x)= 3x_2^2
+(-6 x_4 x_5-4 x_3 x_6+4 x_4 x_6^2)x_2+x_5 (4 x_3^2+3 x_4^2 x_5-4 x_3 x_4 x_6)
$$
and
\begin{align}\label{T7}
\mc R(\mb x)&=3^{20} r_7(\mb x)^2 x_0^7+
      \text{(lower-order terms in $x_0$)}, \\
\mathcal S(\mb x)&=2^2 3^{18} \label{S7}
(x_3 - x_4 x_6) (3 x_2 - 3 x_4 x_5 - 2 x_3 x_6 + 2 x_4 x_6^2)
 x_0^6 \\\nonumber
&\phantom{aaaaaaaaaaaaaaaaaaaaaaaaa}    + \text{(lower-order terms in $x_0$)}. 
\end{align}






Regarding \eqref{T7} and \eqref{S7}, set
\begin{align*}
\mc T_1&:=
\Big\{
\mb x\in \R^7\,;\, 
r_7(\mb x)=x_3 - x_4 x_6=0\Big\}, \\
\mc T_2&:=
\Big\{
\mb x\in \R^7\,;\, 
r_7(\mb x)=3 x_2 - 3 x_4 x_5 - 2 x_3 x_6 + 2 x_4 x_6^2=0\Big\},
\end{align*}
and $\mc T := \mc T_1 \cup \mc T_2$.
Then $\dim_H(\mc T)<6$, and we have:

\begin{Thm}\label{thm:E7}
The set 
$
\mc E_1:=\mc T\cup \Big(\mc Z(\mc R)\cap \mc Z(\mc S)\Big)
$
has dimension less than $6$, and 
$
\Xi_{E_7}:=\mc Z(\mc R)\setminus \mc E_1
$
is an open subset of $\mc Z(\mc R)$.
Moreover, for each $\mb x \in \Xi_{E_7}$, 
there exists a unique real number $v(\mb x)$ 
that is a common root of $A^{\mb x}(v)$ and $B^{\mb x}(v)$.
\end{Thm}

\begin{proof}
By \eqref{T7}, $\mc R(\mb x)$ is a non-constant polynomial, and hence
$
\dim_H \mc Z(\mc R)\le 6.
$
By Corollary \ref{eq:1366e7}, together with Proposition~\ref{prop:1404}, 
we obtain  
\begin{equation}
\dim_H \mc Z(\mc R)=6.
\end{equation}
Setting
$
\xi_1:=(x_0,1,1,0,1,1,1) \in \R^7
$,
one checks that
$$
\op{Res}_{x_0}\big(\mc R(\xi_1),\mc S(\xi_1)\big)\equiv 1 \quad (\mod 5).
$$
Hence
$
\mc L:=\{\mb x\notin \mc T\ ;\ \mc R(\mb x)=\mc S(\mb x)=0\}
$
has $\dim_H(\mc L)<6$. Therefore $\dim_H(\mc E_1)<6$, and $\Xi_{E_7}$ 
is open in $\mc Z(\mc R)$.
Uniqueness and reality of $v(\mb x)$ are proved as in Theorem~\ref{thm:E6}.
\end{proof}

As in Corollary~\ref{2079}, we obtain the following result.

\begin{Corollary}\label{2079b}
The image $h(\R^6)$ is a global main-analytic set of $\R^7$.
\end{Corollary}

\begin{proof}
The proof is the same as that of Corollary~\ref{2079}.
\end{proof}

We now try to find an explicit formula for the
main-analytic function.
As an analogue of
Proposition~\ref{thm:E6main},
we prove the following:

\begin{Prop}\label{thm:E7main}
The polynomial $\mc R(\mb x)$ is divisible by  $r_7(\mb x)^2$ and
$$
\Theta_{E_7}(\mb x):=\frac{\mc R(\mb x)}{r_7(\mb x)^2}
$$
is irreducible over $\Q$.
\end{Prop}

\begin{proof}
As in the case of $E_6$, we proved this fact 
using \textit{Mathematica}.
\end{proof}

\begin{Corollary}\label{eq:1770e7}
The dimension of the set
$
\mc E_{0}:=\mc Z(r_7)\cap \mc Z(\Theta_{E_7})
$
is less than $6$.
\end{Corollary}

\begin{Remark}\label{rmk:1770e7b}
Even without assuming the irreducibility of $\Theta_{E_7}$ over $\Q$, 
this statement can be verified by applying 
Corollary~\ref{cor:A}, taking $a:=r_7$ and $b:=\Theta_{E_7}$ 
as polynomials in $x_2$. 
For instance, with $p=5$, one may consider the sampling point 
$(x_0,\ldots,x_6)=(0,1,x_2,1,0,0,0)$.
\end{Remark}



\begin{Thm}\label{thm:2506}
Fix $\mb x\in \Xi_{E_7}$.
\begin{enumerate}
\item 
If $r_7(\mb x)\ne 0$, then
$\delta_7(v(\mb x))\ne 0$ holds
and there exists $u\in \R\setminus \{0\}$ 
such that $h(u,v(\mb x),\mb x_3)=\mb x$.
In particular,
\begin{equation}\label{eq:e7r6}
\Xi_{E_7}\setminus \mc Z(r_7)\subset h(\R^6).
\end{equation}
\item 
Suppose that $\mb x\in h(\R^6)\cap \mc Z(r_7)$.
Then  $\delta_7(v(\mb x))=0$ and $\mc H(\mb x)=0$ hold.
Moreover, 
$
\mc E_2:=h(\R^6)\cap \mc Z(r_7)
$
has dimension less than $6$.
\end{enumerate}
\end{Thm}

\begin{proof}
All symbolic computations below were carried out with
\textit{Mathematica}.
We obtain
\begin{align}
&\op{Res}_v(A^{\mb x},\delta_7)=3^7 r_7(\mb x)^2 \label{eq:7a}, \\
&\op{Res}_v(B^{\mb x},\delta_7)=12 r_7(\mb x) \label{eq:7b}
\Big(27 x_1^2-18 x_1 x_5 x_6+4 x_1x_6^3+4 x_5^3-x_5^2 x_6^2\Big).
\end{align}
Thus if $\delta_7(v(\mb x))=0$ and $A^{\mb x}(v(\mb x))=0$, 
then $r_7(\mb x)=0$, a contradiction.
Hence $\delta_7(v(\mb x))\neq 0$, and the rest of (1) 
follows from Proposition~\ref{prop:1322e7}.

For (2), assume $\mb x\in h(\R^6)\cap \mc Z(r_7)$.
Dividing $g_2^{\mb x}(u,v)$ by $\delta_7(v)$ gives
\begin{equation}\label{eq:2527}
g_2^{\mb x}(u,v)=-(u+x_4)\delta_7(v)-2\Delta(\mb x)v-x_2+x_4 x_5,
\end{equation}
where $\Delta(\mb x):=x_3-x_4 x_6$.
Set
$
\hat v(\mb x):=({-x_2+x_4 x_5})/({2 \Delta(\mb x)}).
$
Then, substituting $v=\hat v(\mb x)$ into $\delta_7(v)$,
we obtain
\begin{equation}\label{eq:2592}
\delta_7(\hat v(\mb x))=\frac{r_7(\mb x)}{4\Delta(\mb x)^2}=0,
\end{equation}
since $r_7(\mb x)=0$.
So $g_2^{\mb x}(u,\hat v(\mb x))=0$ for any $u$.
Using Lemma~\ref{lem:B0} 
and Proposition~\ref{prop:BB}, one concludes $\dim_H(\mc E_2)<6$.
\end{proof}

\begin{Corollary}\label{cor:2024e7}
Fix $\mb x\in \mc Z_\C(\Theta_{E_7})$.
If $r_7(\mb x)\ne 0$, then
there exist $u,v\in \C$ 
such that $h(u,v,\mb x_3)=\mb x$.
\end{Corollary}

\begin{proof}
The condition $\Theta_{E_7}(\mb x)=0$ implies
the existence of $v\in \C$ satisfying
$A^{\mb x}(v)=B^{\mb x}(v)=0$.
Since $r_7(\mb x)\ne 0$, we have $\delta_7(v)\ne 0$.  
Then we can set $u\in \C$ by
\eqref{eq:1364b}.
Since $A^{\mb x}(v)=0$, this $u$ also satisfy
\eqref{eq:1364a}.
This implies the existence of $u\in \C$ 
such that $h(u,v,\mb x_3)=\mb x$.
\end{proof}



\begin{Thm}\label{Thm:M7}
The polynomial $\Theta_{E_7}(\mb x)$ is a  
main-analytic function on $h(\R^6)$.
\end{Thm}

This provides an alternative proof of the main-analyticity of
$h(\R^6)$, without invoking Appendix~D.


\begin{proof}
Since $\mathcal{Z}(\mathcal{R})=\mathcal{Z}(r_7)\cup \mathcal{Z}(\Theta_{E_7})$, 
\eqref{eq:e7r6}
yields
\[
\mathcal{Z}(\Theta_{E_7})\setminus (\mathcal{E}_0\cup \mathcal{E}_1)
=
\mathcal{Z}(\mathcal{R})\setminus (\mathcal{E}_1\cup \mathcal{Z}(r_7))
=\Xi_{E_7}\setminus \mathcal{Z}(r_7)\subset h(\mathbb{R}^6).
\]
Hence
$
\dim_H \bigl(\mathcal{Z}(\Theta_{E_7})\setminus h(\mathbb{R}^6)\bigr)<6.
$
On the other hand,
by (2) of Theorem~\ref{thm:2506},
\begin{equation}\label{eq:2023c}
h(\R^6)\subset \mc Z(\Theta_{E_7})\cup \mc E_2,\qquad
\dim_H(\mc E_2)<6.
\end{equation}
If $h^{-1}(\mathcal{E}_2)$ had an interior point, the image would have dimension $6$, a contradiction. A limiting argument as in the $E_6$ case then shows $\mc E_2\subset \mc Z(\Theta_{E_7})$, proving the claim.
\end{proof}




\begin{Rmk}\label{rmk:e7}
By Proposition~\ref{prop:1322e7}, we have
$h^\C(\C^6\setminus \mc Z_\C(\delta_7))\subset \mc Z_\C(\Theta_{E_7})$.
Since $h^\C$ is proper, we obtain
\begin{equation}\label{eq:E7hc}
h^\C(\C^6)
=h^\C\bigl(\,\overline{\C^6\setminus \mc Z_\C(\delta_7)}\,\bigr)
=\overline{h^\C(\C^6\setminus \mc Z_\C(\delta_7))}
\subset \mc Z_\C(\Theta_{E_7}).
\end{equation}

Let $\mc L:=\mc Z_\C(\Theta_{E_7})\cap \mc Z_\C(r_7)$. 
Since the complex dimension of $\mc L$ is less than~$6$
by the same reason as in the proof of 
Corollary~\ref{eq:1770e7},
$\mc L$ has no interior point.
On the other hand,
Corollary~\ref{cor:2024e7}
implies that
$\mc Z_\C(\Theta_{E_7})\setminus \mc L\subset h^\C(\C^6)$.
By the properness of $h^\C$, it follows that
\begin{equation}\label{eq:E7hc2}
\mc Z_\C(\Theta_{E_7})=\overline{\mc Z_\C(\Theta_{E_7})\setminus \mc L}
\subset \overline{h^\C(\C^6)}=h^\C(\C^6).
\end{equation}
Combining \eqref{eq:E7hc} and \eqref{eq:E7hc2}, we have
$
h^\C(\C^{6})=\mathcal{Z}_\C(\Theta_{E_7})\subset\C^{7}.
$

Since $h^\C$ is an immersion on an open dense subset of $\C^{6}$,
the hypersurface $h^\C(\C^{6})$ is smooth on an open dense subset. 
As $\Theta_{E_7}$ is
an irreducible polynomial over~$\Q$ (cf.~Proposition~\ref{thm:E7main}), 
it coincides with the defining polynomial of
$h^\C(\C^{6})$.
In particular, $\Theta_{E_7}$ coincides with
the discriminant polynomial of type $E_7$,  
which is known to be irreducible over $\C$.

Assigning weights $(9,7,6,5,3,4,2)$ to $(x_0,\ldots,x_6)$, we obtain:
\begin{itemize}
\item $r_7(\mb x)$ is a weighted homogeneous polynomial of degree $14$,
\item $\op{Res}_v(A^\mb x,B^\mb x)$ is weighted homogeneous of degree $91$,
\item the main-analytic function $\Theta_{E_7}$ is weighted homogeneous of degree $63$.
\end{itemize}
\end{Rmk}

\section{Singularities of type $E_8$}

Since this is parallel to Sections~5 and 6, 
we only record the  main points.
The standard map
$
h:=h_{E_8}:\R^{7}\to \R^{8}
$
of an
$E_8$-singular point
is defined by
$$
h(u,v,\mb x_3):=
\Big(
h_0(u,v,\mb x_3),
h_1(u,v,\mb x_3),
h_2(u,v,\mb x_3),\mb x_3
\Big),
$$
where 
$\mathbf{x}_3 := (x_3, x_4, x_5, x_6, x_7)$ is the subvector 
of $\mathbf{x} \in \mathbb{R}^8$,
and
\begin{align}\label{eq:1636a}
h_0(u,v,\mb x_3)&:=
2 u^3 + 4 v^5 + x_3v^2 + 2x_4 v^3 
+ u v\, \delta_8(v), \\
\label{eq:1636b}
h_1(u,v,\mb x_3)&:=-3 u^2 - x_5 v - x_6 v^2 - x_7 v^3, \\
\label{eq:1636c}
h_2(u,v,\mb x_3)&:=
-5 v^4 - 2x_3v- 3x_4v^2 -u \, \delta_8(v), \\
\label{eq:1687a}
\delta_8(v)&:=3x_7 v^2+2 x_6 v+x_5.
\end{align}

\begin{Prop}
The image of the standard map $h$ coincides with
the set
\begin{align*}
\mc W_{E_8}&:=\Big\{\mb x:=(x_0,\ldots,x_{7})\in \R^8\,;\, 
\text{there exists $(u,v)\in \R^2$ such that}
\\
& \phantom{aaaaaaaaaaaaaaaaaaaaaa}
\text{$F(u,v,\mb x)=F_u(u,v,\mb x)=F_v(u,v,\mb x)=0$}\Big\},
\end{align*}
where 
\begin{align}
\label{eq:E1660}
F(u,v,\mb x)&:= 
u^3 + v^5 + x_7 u v^3 + x_6 u v^2 + x_5 u v \\
&\phantom{aaaaaaaaaa}+ x_4 v^3 + x_3 v^2 + x_2 v + 
 x_1 u + x_0. \nonumber
\end{align}
\end{Prop}



The following assertion can be easily verified:

\begin{Prop}\label{prop:1719}
The map $h$ is a $7$-dimensional real analytic map.
\end{Prop}

Set $\K=\R$ or $\C$ and define $\hat h$ as in the previous sections:

\begin{Prop}\label{prop:E1673}
The map $\hat h:\K^{7}\to \K^8$  is proper.
Moreover, for each $\mb x\in \K^8$, 
the inverse image $\hat h^{-1}(\mb x)$ is finite.
Furthermore, $\hat h^{-1}(\mb 0)=\{o\}$.
\end{Prop}

\begin{proof}
The assertion $\hat h^{-1}(\mb 0)=\{o\}$ is immediate.
Fix $\mb x=(x_0,\ldots,x_7)\in \K^8$ and write 
$x_i=h_i(u,v,x_3,\ldots,x_7)$ for $i=0,1,2$.
Recall  that
\begin{align}
x_1&=h_1(u,v,\mb x_5)= -3u^2 - x_5 v - x_6 v^2 - x_7 v^3, \label{eq:E8-1}\\
x_2&=h_2(u,v,\mb x_5)= -5v^4 - 2x_3 v - 3x_4 v^2 - u\,\delta_8(v), \label{eq:E8-2}
\end{align}
where $\delta_8(v):=x_5+2x_6 v+3x_7 v^2$.

\smallskip
If $\delta_8(v)=0$, then \eqref{eq:E8-2} reduces to
$
x_2=-5v^4-2x_3 v - 3x_4 v^2,
$
so $v$ is a root of a quartic equation over $\K$. Hence the possibilities 
for $v$ are finite, and then
\eqref{eq:E8-1} shows that the possibilities for $u$ are also finite.

Assume now $\delta_8(v)\ne 0$. From \eqref{eq:E8-1} and \eqref{eq:E8-2} we obtain
\begin{align}
u^2&=\frac{-x_1 - x_5 v - x_6 v^2 - x_7 v^3}{3}, \label{eq:E8-u2}\qquad
u=\frac{-5 v^4-3 x_4 v^2-2 x_3 v - x_2}{\delta_8(v)}. 
\end{align}
Eliminating $u$ between \eqref{eq:E8-u2} yields a polynomial equation
$A^{\mb x}(v)=0$, where
\begin{align}\label{eq:E8-A}
A^{\mb x}(v):={}&
75 v^8+9 x_7^3v^7+3 \bigl(30 x_4+7 x_6 x_7^2\bigr)v^6 \\
&+\bigl(60 x_3+x_7 \bigl(15 x_5 x_7+16 x_6^2\bigr)\bigr)v^5  \notag\\
&+\bigl(9 x_1 x_7^2+30 x_2+27 x_4^2+22 x_5 x_6 x_7+4 x_6^3\bigr)v^4  \notag\\
&+\bigl(12 x_1x_6 x_7+36 x_3 x_4+7 x_5^2 x_7+8 x_5 x_6^2\bigr)v^3  \notag\\
&+\bigl(6 x_1 x_5 x_7+4 x_1 x_6^2+18 x_2 x_4+12 x_3^2+5 x_5^2 x_6\bigr)v^2  \notag\\
&+\bigl(4 x_1 x_5 x_6+12 x_2 x_3+x_5^3\bigr)v+x_1 x_5^2+3 x_2^2. \notag
\end{align}
Since the leading term of $A^{\mb x}(v)$ is $75 v^8$, $v$ has 
finitely many possibilities. So, \eqref{eq:E8-u2} gives 
finitely many possibilities for $u$.
Therefore, for every $\mb x\in \K^8$, the inverse image 
$\hat h^{-1}(\mb x)$ is finite.

The properness of the map $\hat h$ can be proved
by imitating the case of
$E_6$ and $E_7$.
\end{proof}

Substituting 
\eqref{eq:E8-u2}
into the equation $h_0(v,\mb x_3)=x_0$
(cf. \eqref{eq:1636a}), we obtain
the equation  $B^{\mb x}(v)=0$, where
\begin{align*}
B^{\mb x}(v)&:=
x_7 v^7+4 x_6 v^6+(7 x_5-3 x_4 x_7)v^5 +5 (2 x_1-x_3 x_7)v^4 \\
&\phantom{aaaaa}+(-7 x_2 x_7-2 x_3 x_6+3 x_4 x_5)v^3 \\
&\phantom{aaaaaaaa}+(-9 x_0 x_7+6 x_1 x_4
-4 x_2 x_6+x_3 x_5)v^2  \\
&\phantom{aaaaaaaaaa}+(-6 x_0 x_6+4 x_1 x_3-x_2 x_5)v-3 x_0 x_5+2 x_1 x_2.
\end{align*}
As in the cases of $E_6$ and $E_7$,
we set
$$
\mc R(\mb x):=\op{Res}_v(A^{\mb x},B^{\mb x}),\qquad
\mc S(\mb x):=\op{Psc}_v(A^{\mb x},B^{\mb x}).
$$
With $g^{\mb x}_i(u,v):=x_i-h_i(u,v,\mb x_3)$ $(i=0,1,2)$, 
like as the cases of $E_6$ and $E_7$, we have the following two assertions:

\begin{Proposition}\label{prop:1742E8}
For $\mb x\in \K^8$ and $u,v\in \K$,
$\hat h(u,v,\mb x_3)=\mb x$ is equivalent to 
$g_i^{\mb x}(u,v)=0$ $(i=0,1,2)$.
\end{Proposition}

\begin{Prop}\label{prop:1322e8}
Fix $\mb x\in \K^8$. If $\mb x\in \hat h(\K^7)$, 
then there exists $v\in \K$ with $A^{\mb x}(v)=B^{\mb x}(v)=0$.
Conversely, the existence of $v\in \K$ satisfying $\delta_8(v)\ne 0$
implies $\mb x\in \hat h(\K^7)$.
\end{Prop}

\begin{Corollary}\label{eq:1366e8}
The image $h(\R^7)$ $($resp. $h^\C(\C^7))$
is a subset of $\mc Z(\mc R)$ $($resp. $\mc Z_\C(\mc R))$.
\end{Corollary}

The resultant
$
r_8(\mb x):=\op{Res}_v(g^{\mb x}_2,\delta_8)
$
can be computed as
\begin{align*}
r_8(\mb x)
&= 25\,x_5^4
 \;-\; 90\,x_4 x_7\,x_5^3 \\
&\quad
 + \bigl(60\,x_4 x_6^2 + 81\,x_4^2 x_7^2 + 180\,x_3 x_6 x_7 
+ 90\,x_2 x_7^2\bigr)\,x_5^2 \\
&\quad
 + \bigl(-80\,x_3 x_6^3 - 240\,x_2 x_6^2 x_7 - 108\,x_3 x_4 x_6 x_7^2
          + 108\,x_3^2 x_7^3 - 162\,x_2 x_4 x_7^3\bigr)\,x_5 \\
&\quad\quad
 + x_2\bigl(80\,x_6^4 + 108\,x_4 x_6^2 x_7^2
 - 108\,x_3 x_6 x_7^3 + 81\,x_2 x_7^4\bigr).
\end{align*}

Set  $B^{\mb x}_0:=x_0 (dB^{\mb x}/dx_0)$ (cf. \eqref{eq:B0}).
Then we have
\[
B^{\mb x}_0(v):=-3 x_0 x_5-6 x_0 x_6v-9x_0 x_7 v^2.
\]
By Remark \ref{rmk:1430},
$\op{Res}_v(A^{\mb x},B^{\mb x})$ and
$\op{Psc}_v(A^{\mb x},B^{\mb x})$ are non-zero constant 
multiples of
$\op{Res}_v(A^{\mb x},B^{\mb x}_0)$ and
$\op{Psc}_v(A^{\mb x},B^{\mb x}_0)$,
respectively.
Furthermore, we have
\begin{align*}
\op{Res}_v(A^{\mb x},B^{\mb x}_0)
  &= 3^{10} r_8(\mb x)^2 x_0^8 
     + \text{(lower-order terms in $x_0$)}, \\
\op{Psc}_v(A^{\mb x},B^{\mb x}_0)
  &= -2^{2} 3^8 p_1(\mb x) p_2(\mb x) x_0^7
     + \text{(lower-order terms in $x_0$)},
\end{align*}
where
\begin{align*}
p_1(\mb x) 
&= 20\,x_6^3 
  + (27 x_4 x_7^2 - 30 x_5 x_7)\,x_6 
  - 27 x_3 x_7^3,
\\
p_2(\mb x) 
&= 40\,x_6^4 
  + (54 x_4 x_7^2 - 120 x_5 x_7)\,x_6^2 
  - 54 x_3 x_7^3\,x_6 \\ \nonumber
&\phantom{aaaaaaaaaaaaaaaaaaaa} + \bigl(45 x_5^2 x_7^2 
- 81 x_4 x_5 x_7^3 + 81 x_2 x_7^4\bigr).
\end{align*}
It holds that
\[
\op{Res}_{x_6}(r_8,p_i)\Big|_{x_5=1,x_7=0}\equiv 1 \pmod{7}\quad \qquad (i=1,2).
\]
Hence each
\[
\mc T_i:=\{\mb x\in\R^8\ ;\ r_8(\mb x)=p_i(\mb x)=0\}\qquad \quad (i=1,2)
\]
has $\dim_H(\mc T_i)<7$, so 
$\mc T:=\mc T_1\cup\mc T_2$ also satisfies $\dim_H(\mc T)<7$.

\begin{Thm}\label{thm:E8}
$\mc E_1:=\mc T\cup \bigl(\mc Z(\mc R)\cap \mc Z(\mc S)\bigr)$ 
has $\dim_H(\mc E_1)<7$, and 
$\Xi_{E_8}:=\mc Z(\mc R)\setminus \mc E_1$ is open in $\mc Z(\mc R)$. 
Moreover, for each $\mb x\in \Xi_{E_8}$ there is a 
unique $v(\mb x)\in\R$ with $A^{\mb x}(v(\mb x))=B^{\mb x}(v(\mb x))=0$.
\end{Thm}

\begin{proof}
Set
$
\xi_1:=(x_0,1,0,0,0,1,0,1).
$
Using \textit{Mathematica}, we have 
\[
\mc R(\xi_1)\equiv
3 \bigl(5 + 3 x_0 + 6 x_0^2 + x_0^4\bigr)\bigl(4 + 3 x_0^2 
+ 2 x_0^3 + x_0^4\bigr) \pmod{7}
\]
and
\[
\mc S(\xi_1)\equiv
3+2 x_0+3 x_0^2+x_0^3+3 x_0^4+4 x_0^5+x_0^6 \pmod{7}.
\]
Moreover, the resultant of the two polynomials 
$\mc R(\xi_1)$ and $\mc S(\xi_1)$ in $x_0$
is equal to $1$ modulo $7$.
Thus, by Corollary \ref{cor:A} in the appendix, we conclude that
\[
\mc L:=
\{
\mb x\in \R^8\setminus \mc T \; ; \;
\mc R(\mb x)=\mc S(\mb x)=0
\}
\]
has dimension less than $7$.
Hence
$
Z:=\{
\mb x\in \R^8 \; ; \; 
\mc R(\mb x)=\mc S(\mb x)=0
\}
\subset \mc L\cup \mc T,
$
which implies that $Z$ has dimension less than $7$.
Thus the first claim follows, and
$\Xi_{E_8}$ is an open subset of $\mc Z(\mc R)$.
Fix $\mb x\in \Xi_{E_8}$.
Then, as in the cases of $E_6$ and $E_7$,
there exists a unique $v\in \R$ such that
$
A^{\mb x}(v)=B^{\mb x}(v)=0.
$
We denote this unique solution by $v(\mb x)$. 
This is exactly the $v(\mb x)$ claimed in the theorem.
\end{proof}

As in Corollaries \ref{2079} and \ref{2079b}, we obtain the following:

\begin{Corollary}\label{2079c}
$h(\R^7)$ is a global main-analytic set of $\R^8$.
\end{Corollary}

We next show the following:

\begin{Thm}\label{thm:E8main}
The polynomial $\mc R(\mb x)$ is divisible by $r_8(\mb x)^2$, and 
$
\Theta_{E_8}(\mb x):={\mc R(\mb x)}/{r_8(\mb x)^2}
$
is irreducible over $\Q$.
\end{Thm}

\begin{proof}
By using \textit{Mathematica}, we computed
$
\Theta_{E_8}(\mb x)=\operatorname{Res}_v\!\left(A^{\mb x}(v),\,B^{\mb x}(v)\right).
$
The resulting expression occupies about 13 megabytes when stored as a file, 
and the computation itself required approximately 7.5 hours on a standard laptop.  
From this explicit output we verified that $\Theta_{E_8}(\mb x)$ 
is divisible by $r_8(\mb x)^2$.  
Moreover, the quotient $\Theta_{E_8}(\mb x)/r_8(\mb x)^2$ 
is checked to be irreducible over $\Q$.
\end{proof}



\begin{Corollary}\label{eq:1770e8}
$\dim_H\bigl(\mc Z(r_8)\cap \mc Z(\Theta_{E_8})\bigr)<7$.
\end{Corollary}

\begin{Remark}\label{rmk:2765}
Even without assuming the irreducibility of $\Theta_{E_8}$, 
this statement can be verified by applying 
Corollary~\ref{cor:A}, taking $a:=r_8$ and $b:=\Theta_{E_8}$ as polynomials in $x_5$. 
For instance, with $p=7$, one may consider the sampling point 
$(x_0,\ldots,x_7)=(0,1,0,1,1,x_5,1,0)$.
\end{Remark}




The following assertion is an analogue of Theorems~\ref{thm:1954}
and \ref{thm:2506}.

\begin{Thm}\label{thm:3170}
Fix $\mb x\in \Xi_{E_8}$.
\begin{enumerate}
\item 
If $r_8(\mb x)\ne 0$, then
$\delta_8(v(\mb x))\ne 0$ and there exists $u\in \R\setminus \{0\}$ 
such that $h(u,v(\mb x),\mb x_3)=\mb x$. In particular,
$
\Xi_{E_8}\setminus \mc Z(r_8)\subset h(\R^7).
$
\item 
If $\mb x\in h(\R^7)\cap \mc Z(r_8)$, then $\delta_8(v(\mb x))=0$ and $\mc H(\mb x)=0$.
Moreover, 
$\mc E_2:=h(\R^7)\cap \mc Z(r_8)$ has $\dim_H(\mc E_2)<7$.
\end{enumerate}
\end{Thm}

\begin{proof}
All symbolic computations were carried out with {\it Mathematica}. One finds
\[
\op{Res}_v(A^{\mb x},\delta_8)=9\, r_8(\mb x)^2,\qquad
\op{Res}_v(B^{\mb x},\delta_8)=4x_7\, J(\mb x)\, r_8(\mb x),
\]
where $J$ is a certain polynomial.
So $\delta_8(v(\mb x))=0$ and $A^{\mb x}(v(\mb x))=0$ 
would force $r_8(\mb x)=0$, proving (1).  
For (2), the reconstruction of $\hat v(\mb x)$ via division of
 $g_2^{\mb x}$ by $\delta_8$ proceeds exactly as in the $E_7$ 
case, with the dimension bounds $<7$
 obtained by Lemma~\ref{lem:B0} and Corollary~\ref{cor:A}. 
\end{proof}

By the same argument as in the proof of Corollary~\ref{cor:2024e7},
 we obtain the following.

\begin{Corollary}\label{cor:2024e8}
Fix $\mb x\in \mc Z_\C(\Theta_{E_8})$.
If $r_8(\mb x)\ne 0$, then
there exist $u,v\in \C$ 
such that $h(u,v,\mb x_3)=\mb x$.
\end{Corollary}


\begin{Thm}\label{Thm:M8}
The polynomial $\Theta_{E_8}(\mb x)$ is a  
main-analytic function of $h(\R^7)$.
\end{Thm}

\begin{proof}
Since $\mathcal{Z}(\mathcal{R})=\mathcal{Z}(r_8)\cup \mathcal{Z}(\Theta_{E_8})$, 
we get
\[
\mathcal{Z}(\Theta_{E_8})\setminus (\mathcal{E}_0\cup \mathcal{E}_1)
=
\mathcal{Z}(\mathcal{R})\setminus (\mathcal{E}_1\cup \mathcal{Z}(r_8))
=\Xi_{E_8}\setminus \mathcal{Z}(r_8)\subset h(\mathbb{R}^7).
\]
Hence $\dim_H \bigl(\mathcal{Z}(\Theta_{E_8})\setminus h(\mathbb{R}^7)\bigr)<7$.  
Combined with Corollary \ref{eq:1366e7} and Theorem~\ref{thm:3170}, 
the standard limiting argument (as in $E_7$) shows 
$h(\R^7)\subset \mc Z(\Theta_{E_8})\cup \mc E_2$ 
with $\dim_H(\mc E_2)<7$ and $\mc E_2\subset \mc Z(\Theta_{E_8})$.
\end{proof}

\begin{Rmk}\label{rmk:E8}
As in the cases of $E_6$ and $E_7$, we have
$
h^\C(\C^{7})=\mathcal{Z}_\C(\Theta_{E_8})\subset\C^{8}.
$
Since $h^\C$ is an immersion on an open dense subset of $\C^{7}$,
the same reasoning as in the $E_6$ and $E_7$ cases shows that 
$\Theta_{E_8}$ coincides with the discriminant polynomial of type $E_8$
up to a nonzero scalar, which is known to be irreducible over $\C$.

Assigning weights $(15,10,12,9,6,7,4,1)$ to $(x_0,\ldots,x_7)$, 
we obtain:
\begin{itemize}
\item $r_8(\mb x)$ is a weighted homogeneous polynomial of degree $28$,
\item $\op{Res}_v(A^\mb x,B^\mb x)$ is weighted homogeneous of degree $176$,
\item the discriminant $\Theta_{E_8}$ 
is weighted homogeneous of degree $120$.
\end{itemize}
\end{Rmk}


\begin{acknowledgements}
The authors thank Professors 
Goulwen Fichou,   Toshizumi Fukui,
Atsufumi Honda, Goo Ishikawa, Satoshi Koike and Toru Ohmoto for
 valuable comments and for fruitful discussions.
\end{acknowledgements}


\appendix
\section{Resultants of two polynomials}

Let $a(t), b(t)\in\R[t]$ be nonzero polynomials of degrees $n,m\ge 1$.  
We denote by $\op{Res}_t(a,b)$ the resultant and by $\op{Psc}_t(a,b)$ the
first principal subresultant coefficient (cf. \eqref{eq:550}), 
both defined via 
Sylvester-type matrices. 
It is classical that if $a_0b_0\neq 0$, then
\begin{enumerate}
\item $\op{Res}_t(a,b)=0$ holds if and only if 
$a,b$ have a common root in $\C$,
\item $\op{Res}_t(a,b)=\op{Psc}^{(1)}_t(a,b)=0$
if and only if $a,b$ have at least two common roots in $\C$.
\end{enumerate}
The same conclusions remain valid whenever $(a_0,b_0)\neq (0,0)$:
if, say, $b_0=0$, then 
\[
\op{Res}_t(a,b)=a_0\op{Res}_t(a,\hat b), \qquad
\op{Psc}_t(a,b)=a_0\op{Psc}_t(a,\hat b),
\]
where $\hat b$ is obtained by removing the highest vanishing coefficients of $b$.
Thus the claim follows by induction.

\section{
Estimating the dimension of 
common zero sets of two polynomials}

We begin with the following fundamental fact.

\begin{Proposition}\label{Prop:B0}
Let $a,b \in \Q[x_1,\ldots,x_n]$ be nonzero polynomials.  
If $a$ and $b$ have no nontrivial common factor in $\Q[x_1,\ldots,x_n]$,  
then the real zero set $\mathcal Z(a)\cap \mathcal Z(b)\subset \R^n$  
has dimension strictly less than $n-1$. 
\end{Proposition}

In this appendix, we present a more efficient method for estimating
the dimension of a common zero set by focusing on a single variable
and employing the resultant with respect to that variable.

\begin{Lemma}\label{lem:B0}
Let 
$$
a(t,\mb x)=\alpha(\mb x)t^j+\text{\rm $($lower terms$)$}
\qquad (j\ge1)
$$
be a polynomial belonging to $\R[x_0,\dots,x_{n-1}][t]$,
where $\mb x=(x_0,\ldots,x_{n-1})$, and let 
$b(\mb x)\in\R[x_0,\ldots,x_{n-1}]$ be nonzero.  
Then
\begin{align*}
S&:=\{(t,\mb x)\in\R^{n+1}\,;\, a(t,\mb x)
=b(\mb x)=0,\ \alpha(\mb x)\ne0\}, \\
\tilde S&:=\{\mb x \in\R^{n}\,;\, 
\exists\,t\ \text{such that}\,\,
a(t,\mb x)=b(\mb x)=0,\ \alpha(\mb x)\ne0\}
\end{align*}
are semianalytic and have dimension $<n$.
\end{Lemma}

\begin{proof}
For each $\mb x$ with $b(\mb x)=0$ and $\alpha(\mb x)\ne0$, the polynomial 
$t\mapsto a(t,\mb x)$ has finitely many real roots. 
Hence,
\[
\dim_H S\le \dim_H \{\mb x\in \R^n \,;\, b(\mb x)=0\} \le n-1.
\]
Since the canonical projection $\pi:\R^{n+1}\to \R^n$
is a Lipschitz map and satisfies $\pi(S)=\tilde S$,
it follows that
$
\dim_H \tilde S\le \dim_H S \le n-1.
$
\end{proof}

Let $a,b\in\R[x_0,\dots,x_{n-1}][t]$ with $\deg_t a,\deg_t b\ge1$ 
and leading coefficients $\alpha(\mathbf x),\beta(\mathbf x)$.
Set
$\mathcal T:=\{\mathbf x\in\R^n:\ \alpha(\mathbf x)=\beta(\mathbf x)=0\}$
and
$\mathcal R(\mathbf x):=\operatorname{Res}_t(a,b)$.

\begin{Proposition}\label{prop:BB}
If there exists $\mathbf c_0\in \R^n$ 
satisfying $\mathcal R(\mathbf c_0)\neq0$,
then the sets
\begin{align*}
S:=\{(t,\mathbf x)\in\R^{n+1}:\ a=b=0,\ \mathbf x\notin\mathcal T\},\\
\tilde S:=\{\mathbf x\in\R^n\setminus\mathcal T:\ \exists\,t\ \text{such that}\,\, a=b=0\}
\end{align*}
satisfy $\dim_H S<n$ and $\dim_H\tilde S<n$.
\end{Proposition}

\begin{proof}
Since $\tilde S\subset\mathcal Z(\mathcal R)$ and $\mathcal R\not\equiv0$, 
we have $\dim_H\mathcal Z(\mathcal R)<n$.
By
Lemma \ref{lem:B0},
\[
S_1:=\{a=\mathcal R=0,\ \alpha\neq0\},\qquad
S_2:=\{b=\mathcal R=0,\ \beta\neq0\}
\]
satisfy 
$\dim_H S_i<n$  for $i=1,2$.
Since $S\subset S_1\cup S_2$, 
we get $\dim_H S<n$ and  also $\dim_H\tilde S<n$.
\end{proof}

\begin{Corollary}\label{cor:A}
If the coefficients of $a,b$ are integers and there exist a prime $p$ 
and $\mathbf c_0\in\Z^n$ with
$\mathcal R(\mathbf c_0)\not\equiv0\pmod p$, then the same dimension estimate holds.
\end{Corollary}

In practical applications, one usually chooses the 
smallest possible prime $p$ to simplify computations.



\section{A property of a system of three polynomial equations}

Set $\K=\R$ or $\C$.
To investigate properties of the 
images of the standard maps $h_{E_6}, h_{E_7}$ 
and $h_{E_8}$ in a unified manner,
we consider the following three polynomials
\begin{align}
G_0(u,v)&:=2u^3+\gamma_0(v)+uv\delta(v) \label{eq:C0}, \\ 
G_1(u,v)&:=-3u^2+\gamma_1(v) \label{eq:C1}, \qquad
G_2(u,v):=\gamma_2(v)-u \delta(v),
\end{align}
where $\delta(v)$ and $\gamma_i(v)$ ($i=0,1,2$)
 are polynomials in $v$ with real coefficients.

If there exists $u_0,v_0\in \K$ such that
$G_i(u_0,v_0)=0$ for $i=0,1,2$,
then  \eqref{eq:C1} can be rewritten as
\begin{equation}\label{eq:C2b}
3u_0^2=\gamma_1(v_0), \qquad u_0=\gamma_2(v_0)/\delta(v_0),
\end{equation}
where the second equation makes sense whenever $\delta(v_0)\ne 0$.
If we set
\begin{equation}\label{eq:C4}
\alpha(v):=3\gamma_2(v)^2-\delta(v)^2 \gamma_1(v) 
\qquad (v\in \K),
\end{equation}
then $\alpha(v_0)=0$ holds.
Under the assumptions  $G_i(u_0,v_0)=0$ ($i=0,1,2$),
\eqref{eq:C0}, \eqref{eq:C1} and \eqref{eq:C2b} yield
\begin{equation}
0=G_0(u_0,v_0)
=\frac{\gamma_1(v_0)}3\frac{\gamma_2(v_0)}{\delta(v_0)}+\gamma_0(v_0)+u_0v_0
\delta(v_0).
\end{equation}
So, the polynomial
\begin{equation}\label{eq:C6}
\beta(v):=\gamma_1(v)\gamma_2(v)+3 \gamma_0(v)\delta(v)
+\,3v\,\gamma_2(v)\delta(v)
\qquad (v\in \K)
\end{equation}
vanishes at $v=v_0$.

\begin{Lemma}\label{LemC1}
If there exist $u_0,v_0\in \K$ such that
$G_i(u_0,v_0)=0$ for all $i=0,1,2$
then $\alpha(v_0)=\beta(v_0)=0$ holds.
\end{Lemma}
 
\begin{proof}
We have already observed that the assertion holds if $\delta(v_0)\ne 0$.
So we consider the case $\delta(v_0)=0$.
Then $G_2(u_0,v_0)=0$ implies that $\gamma_2(v_0)=0$, and
$\alpha(v_0)=\beta(v_0)=0$ hold obviously.
\end{proof}

Conversely, we can show the following:

\begin{Proposition}\label{thm:Cm}
If $\delta(v_0)\ne 0$ and $\alpha(v_0)=\beta(v_0)=0$
for some $v_0\in \K$, 
then there exists $u_0\in \K$ such that
$G_i(u_0,v_0)=0$ holds for $i=0,1,2$.
\end{Proposition}

\begin{proof}
Since $\delta(v_0)\ne 0$,
we can set
$
u_0:={\gamma_2(v_0)}/{\delta(v_0)},
$
then $G_2(u_0,v_0)=0$ holds.
Moreover, $\alpha(v_0)=0$ implies 
$u_0^2={\gamma_1(v_0)}/3$. So we have $G_1(u_0,v_0)=0$.
Then $\beta(v_0)=0$ implies $G_0(u_0,v_0)=0$.
\end{proof}


\section{A criterion for global main-analyticity}

To give a criterion for global main-analyticity as mentioned
in the introduction, we here recall the following fact:

\begin{Fact}\label{fact:Coste}
Let 
$
    f : \mathbb{R}^m \to \mathbb{R}^n \,\, (m \le n)
$
be a polynomial map.  
Assume that $f$ is \emph{generically injective}, that is, 
there exists a Zariski open dense subset 
$U \subset \mathbb{R}^m$ such that the restriction $f|_U$ is injective.
Put $S := f(\mathbb{R}^m)$ and let 
$\overline{S}^{\,Z} \subset \mathbb{R}^n$ denote the Zariski closure of $S$.
Then the complement $\overline{S}^{\,Z} \setminus S$ has $($algebraic$)$ 
dimension
strictly less than $m$.
\end{Fact}

This fact follows from standard results in real algebraic geometry
(see, for instance, Bochnak--Coste--Roy~\cite[\S 2.8 and \S 3.3]{BCR98}). 
In our setting, it is obtained as a special case of 
Coste~\cite[Corollary~3.19]{Coste2007}, applied to the regular map
$f : \mathbb{R}^m \to \overline{S}^{\,Z}$, where $\mathbb{R}^m$ is regarded 
as a nonsingular real algebraic variety.

\begin{Prop}
\label{thm:Coste-main-analytic-J} 
Let 
$
    f : \mathbb{R}^m \to \mathbb{R}^n \,\, (m \le n)
$
be a polynomial map.  
Assume that there exists an open dense subset $U \subset \mathbb{R}^m$ such that
\begin{enumerate}
\item[(1)]
$f$ is an immersion on $U$, i.e.\ $\operatorname{rank}(df_x)=m$ for all $x\in U$,
\item[(2)]
$f|_U$ is injective.
\end{enumerate}
Then $S := f(\mathbb{R}^m)$ is a global main-analytic set.
Moreover, one may take a main-analytic function $\Theta$ to be a real 
polynomial on $\mathbb{R}^n$.
\end{Prop}

\begin{proof}
By assumption (2), there exists an open dense subset 
$U \subset \mathbb{R}^m$ on which $f$ is injective.  
Let
\[
T := \{x\in\mathbb{R}^m \mid \exists x'\neq x,\ f(x)=f(x')\}.
\]
Then $T$ is semialgebraic and contained in the nowhere dense set
$\mathbb{R}^m\setminus U$, hence $\dim T<m$, where $\dim$ denotes
the algebraic dimension of $T$.
In particular, the Zariski closure $\overline{T}^{\,Z}$ is a proper algebraic
subset of $\mathbb{R}^m$, and $f$ is generically injective in the sense of
Fact~\ref{fact:Coste}.

By the Tarski--Seidenberg theorem, the image $S=f(\mathbb{R}^m)$ is a
semialgebraic subset of $\mathbb{R}^n$, and 
$\overline{S}^{\,Z}$ is a real algebraic set. 
Since $f|_U$ is an $m$-dimensional real
analytic map, $\dim_H(S)=m$ holds (cf.\ Proposition \ref{prop:HDIM}).
Since $S$ is semi-algebraic, its Hausdorff dimension coincides with its
(algebraic) dimension; thus $\dim S = m$.
In particular, the algebraic
dimension of $\overline{S}^{\,Z}$ also equals $m$.

There exists a real polynomial $\Theta$ defining the algebraic set
$\overline{S}^{\,Z}$ so that
$
   \mathcal Z(\Theta) = \overline{S}^{\,Z} \supset S.
$
Since $\overline{S}^{\,Z}$ has (algebraic) dimension $m$, 
its Hausdorff dimension also equals $m$. Hence
$
   \dim_H(S) = \dim_H(\mathcal{Z}(\Theta)) = m.
$

Moreover, by Fact~\ref{fact:Coste}, the complement
$\overline{S}^{\,Z}\setminus S$ has (algebraic) 
dimension strictly less than
$m$.  This set is semialgebraic, 
so its Hausdorff dimension coincides with its
(algebraic) dimension.  Therefore
\[
   \dim_H\bigl(\mathcal{Z}(\Theta)\setminus S\bigr)
   = \dim_H\bigl(\overline{S}^{\,Z}\setminus S\bigr)
   < m = \dim_H(S).
\]
So $S$ is a global main-analytic set with $\Theta$ as a main-analytic
function on $\mathbb{R}^n$.
\end{proof}

\begin{Rmk}
If $f$ is a real-analytic map but not a polynomial,
then the conclusion of Proposition~\ref{thm:Coste-main-analytic-J}
may fail.
For example, the Osgood map
$$
f_O(u,v) := (u,\, uv,\, uve^v) \qquad (u,v\in\R)
$$
is injective on its regular set, but its image cannot be globally
main-analytic.
Indeed, it is a classical fact that any real-analytic function
vanishing on $f_O(\R^2)$ near the origin must be identically zero.
Hence no nontrivial real-analytic function can have $f_O(\R^2)$ in its
zero set.
\end{Rmk}


\begin{thebibliography}{00}

\bibitem{A}
V.\,I.~Arnol'd,
{\it Normal forms of functions near degenerate critical points, 
the Weyl groups $A_k$, $D_k$, $E_k$ and Lagrangian singularities},
Funct.\ Anal.\ Appl.\ {\bf 6} (1972), 254--272.


\bibitem{BCR98}
J.~Bochnak, M.~Coste, and M.-F.~Roy,
\textit{Real Algebraic Geometry},
Ergebnisse der Mathematik und ihrer Grenzgebiete (3), Vol.~36,
Springer-Verlag, Berlin, 1998.

\bibitem{Coste2007}
M.~Coste,
\newblock \emph{Real Algebraic Sets},
\newblock in: T.~Fukui, L.~van~den~Dries, J.~Koll\'ar (eds.),
\newblock \emph{Arc Spaces and Additive Invariants 
in Real Algebraic and Analytic Geometry},
Panoramas et Synth\`eses~24,
Soci\'et\'e Math\'ematique de France,
(2007), 109--156.

\bibitem{FKKRUYY2}
S.~Fujimori, Y.~Kawakami, M.~Kokubu, W.~Rossman,
M.~Umehara, K.~Yamada and S.-D.~Yang,
{\it Unextendability of real analytic map images}, 
in preparation.

\bibitem{IKS}
G.~Ishikawa, S.~Koike and M.~Shiota,
{\it Critical value sets of generic mappings},
Pacific J.\ Math.\ {\bf 114} (1984), 165--174.

\bibitem{Loo}
E.~Looijenga,
{\it The discriminant of a real simple singularity},
Compositio Math.\ {\bf 37} (1978), 51--82.


\bibitem{Sa1}
K.~Saito,
{\it On a linear structure of the quotient variety by a finite reflection group},
Publ.\ Res.\ Inst.\ Math.\ Sci.\ {\bf 29} (1975/76), 535--579.


\bibitem{S}
K.~Saji,
{\it Criteria for Morin singularities into higher dimension 
$($The theory of singularities of smooth mappings and around it$)$},
Surikaiseki Kenkyusho Kokyuroku Bessatsu~B55 (2016), 205--224.



\bibitem{V}	
L.\,G.~Vega,
{\it A subresultant theory for multivariate polynomials},
Proc.\ ISSAC~'91 (1991), 79--85.

\end{thebibliography}


\end{document}