A Second Main Theorem for Entire Curves Intersecting Three Conics
Extracted source text · arXiv 2512.03948
\documentclass[english, 11pt]{amsart}
%\usepackage{showkeys}
\usepackage{listings}
\usepackage{xcolor}
\lstset{
backgroundcolor=\color{white},
basicstyle=\ttfamily\footnotesize,
breaklines=true,
frame=none,
numbers=none,
numberstyle=\tiny,
keywordstyle=\color{blue},
commentstyle=\color{green!60!black},
stringstyle=\color{red},
showstringspaces=false
}
\usepackage{mathptmx}
\usepackage{tikz}
\usepackage{aliascnt}
\usepackage{mathrsfs}
\usepackage[all,poly,knot]{xy}
\usepackage{hyperref}
\usepackage{comment}
\usepackage{csquotes}
\usepackage{amssymb,amsbsy,amsmath,amsfonts,amssymb,amscd,
graphics,color,footmisc,fancyhdr,multicol,fancybox,
graphicx,mathrsfs,rotating,ifthen,wasysym}
\usepackage{tikz-cd}
\usepackage{nameref}
\usepackage{times}
\usepackage{pdfpages}
\usepackage{multirow}
\usepackage{nccrules}
%\usepackage{stmaryrd}
\usepackage{textcomp}
\usepackage{comment}
\usepackage{framed}
\usepackage[all]{xy}
\usepackage{graphicx}
\usepackage{pgf,tikz}
\usepackage{mathrsfs}
\usetikzlibrary{arrows}
\usepackage{lipsum}
\usepackage{mathtools}
\DeclarePairedDelimiter\ceil{\lceil}{\rceil}
\DeclarePairedDelimiter\floor{\lfloor}{\rfloor}
\usepackage{geometry} % 在 \documentclass 之后、\begin{document} 之前添加这行
\geometry{
a4paper, % 论文尺寸
top=20mm, % 上边距
bottom=25mm, % 下边距。通常比上边距稍大，以容纳页码。
left=25mm, % 左边距
right=20mm, % 右边距。考虑到装订，左边距通常稍大。
headheight=15pt % 确保页眉高度足够，避免警告
}
\makeatletter\newcommand{\leqnomode}{\tagsleft@true}
\newcommand{\reqnomode}{\tagsleft@false}\makeatother
\newcommand{\medcap}{\mathbin{\scalebox{1.5}{\ensuremath{\cap}}}}
\newcommand{\medcup}{\mathbin{\scalebox{1.5}{\ensuremath{\cup}}}}
\DeclareMathOperator{\clu}{Clu}
\DeclareMathOperator{\Sym}{Sym}
\DeclareMathOperator{\pic}{Pic}
\DeclareMathOperator{\vol}{vol}
\DeclareMathOperator{\reg}{reg}
\DeclareMathOperator{\pr}{pr}
\DeclareMathOperator{\dif}{d}
\DeclareMathOperator{\leng}{Length}
\DeclareMathOperator{\hol}{Hol}
\DeclareMathOperator{\rank}{rank}
\DeclareMathOperator{\lcm}{lcm}
\DeclareMathOperator{\im}{Im}
\DeclareMathOperator{\codim}{codim}
\DeclareMathOperator{\wron}{Wron}
\DeclareMathOperator{\ins}{in}
\DeclareMathOperator{\ric}{Ric}
\DeclareMathOperator{\fin}{fin}
\DeclareMathOperator{\supp}{supp}
\DeclareMathOperator{\ord}{ord}
\DeclareMathOperator{\sign}{sign}
\newcommand{\hty}{\rightharpoonup}
\newcommand{\rong}{\emptyset}
\newcommand{\NN}{\mathbb{N}}
\newcommand{\ZZ}{\mathbb{Z}}
\newcommand{\RR}{\mathbb{R}}
\newcommand{\CC}{\mathbb{C}}
\newcommand{\BB}{\mathbb{B}}
%\newcommand{\vf}{\vfill\end{document}}
\newcommand{\C}{\mathbb{C}}
\newcommand{\N}{\mathbb{N}}
\renewcommand{\P}{\mathbb{P}}
\newcommand{\Z}{\mathbb{Z}}
\newcommand{\style}[1]{{\sf #1}}
\newcommand{\Jet}{\style{Jet}}
\newcommand{\GL}{\style{GL}}
\newcommand{\smallbullet}{{\scriptscriptstyle{\bullet}}}
\def\log{\mathrm{log}\,}
\def\blue{\textcolor[rgb]{0,0,1}}
\newcommand{\lowerromannumeral}[1]{\romannumeral#1\relax}
\def\green{\textcolor[rgb]{0,1,0}}
\def\red{\textcolor[rgb]{1,0,0}}
\def\blue{\textcolor[rgb]{0,0,1}}
\theoremstyle{plain}
\newtheorem{thmx}{Theorem}
\renewcommand{\thethmx}{\Alph{thmx}}
\newtheorem{thm}{Theorem}[section]
\newtheorem{prethm}[thm]{{Pretheorem}}
\newtheorem{cor}[thm]{{Corollary}}
\newtheorem{corx}[thmx]{Corollary}
\newtheorem{lem}[thm]{{Lemma}}
\newtheorem{prelem}[thm]{{Prelemma}}
\newtheorem{pro}[thm]{Proposition}
\newtheorem{preprop}[thm]{Preproposition}
\newtheorem{conj}[thm]{{Conjecture}}
\newtheorem{conjx}[thmx]{Conjecture}
\newtheorem{ques}[thm]{{Question}}
\newtheorem{rem}[thm]{Remark}
\newtheorem{defi}[thm]{Definition}
\newtheorem{obs}[thm]{Observation}
\newtheorem*{claim}{Claim}
\newtheorem{KeyObservation}[equation]{Key Observation}
\newtheorem{Observation}[equation]{Observation}
\newtheorem{Proposition}[equation]{Proposition}
\newtheorem{MiracleFactorizations}[equation]{Miracle Factorizations}
\newtheorem{Lemma}[equation]{Lemma}
\newtheorem{KeyVanishingTheorem}[equation]{Key Vanishing Lemma}
\theoremstyle{remark}
\newtheorem{rmk}[thm]{Remark}
\newtheorem{nota}[thm]{Notation}
\theoremstyle{definition}
\newtheorem{Hypothesis}[equation]{Hypothesis}
\numberwithin{equation}{section}
\usepackage[hyperpageref]{backref}
\renewcommand{\backref}[1]{$\uparrow$~#1}
\usepackage{xcolor}
\hypersetup{
colorlinks,
linkcolor={red!50!black},
citecolor={blue!62!black},
urlcolor={blue!80!black}
}
\theoremstyle{plain}
\newcommand{\thistheoremname}{}
\newtheorem*{genericthm*}{\thistheoremname}
\newenvironment{namedthm*}[1]{\renewcommand{\thistheoremname}{#1}%
\begin{genericthm*}}
{\end{genericthm*}}
\newtheorem*{theorem}{Theorem}
\newtheoremstyle{named}{}{}{\itshape}{}{\bfseries}{.}{.5em}{\thmnote{#3's }#1}
\theoremstyle{named}
\newtheorem*{namedthm}{Theorem}
\allowdisplaybreaks[4]
\makeatletter
\newcommand\thankssymb[1]{\textsuperscript{\@fnsymbol{#1}}}
%\newcommand\thankssymb[1]{\lowercase{\textsuperscript{\@alph{#1}}}}
%\newcommand\thankssymb[1]{\textsuperscript{#1}}
\makeatother
\begin{document}
\title[Second Main Theorem for three conics]{\bf A Second Main Theorem for Entire Curves Intersecting Three Conics}
\subjclass[2020]{32H30, 32Q45}
\keywords{
Nevanlinna theory, Second Main Theorem, entire curves, jet differentials, Siu's slanted vector fields, Demailly-Semple tower}
\author{Lei Hou}
\address{Academy of Mathematics and Systems Sciences, Chinese Academy of Sciences, Beijing 100190, China}
\email{houlei@amss.ac.cn}
\author{Dinh Tuan Huynh}
\address{Department of Mathematics, University of Education, Hue University, Hue City 530000, Vietnam}
\email{dinhtuanhuynh@hueuni.edu.vn}
\author{
Joël Merker}
\address{ Laboratoire de Math\'ematiques d’Orsay, CNRS, Universit\'e Paris-Saclay, 91405 Orsay
Cedex, France}
\email{joel.merker@universite-paris-saclay.fr}
\author{Song-Yan Xie}
\address{State Key Laboratory of Mathematical Sciences, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing 100190, China; School of Mathematical Sciences, University of Chinese Academy of Sciences, Beijing 100049, China}
\email{xiesongyan@amss.ac.cn}
\date{\today}
\begin{abstract}
We establish a Second Main Theorem for entire holomorphic curves \( f: \mathbb{C} \to \mathbb{P}^2 \) intersecting a generic configuration of three conics \( \mathcal{C}_1, \mathcal{C}_2, \mathcal{C}_3 \) in the complex projective plane $\mathbb{P}^2$. Using invariant logarithmic $2$-jet differentials with negative twists, we prove the estimate
\[
T_f(r) \leqslant 5 \sum_{i=1}^3 N_f^{[1]}(r, \mathcal{C}_i) + o\big(T_f(r)\big)\quad\parallel,
\]
where \( T_f(r) \) is the Nevanlinna characteristic function, and \( N_f^{[1]}(r, \mathcal{C}_i) \) is the \(1\)-truncated counting function.
\end{abstract}
\thanks{ Appendix~\ref{appendix:results of Pengchao Wang} was authored by Pengchao Wang.}
\maketitle
\section{\bf Introduction}
\subsection{Motivation}
In the complex projective plane \(\mathbb{P}^2\), the logarithmic Kobayashi conjecture posits that if \(\mathcal{C} \subset \mathbb{P}^2\) is a \emph{generic} algebraic curve of degree \(d \geqslant 5\), then the complement \(\mathbb{P}^2 \setminus \mathcal{C}\) should be hyperbolic \cite{Kobayashi1970}. Here, hyperbolicity means that \(\mathbb{P}^2 \setminus \mathcal{C}\) does not contain the image of any nonconstant holomorphic map \(f: \mathbb{C} \rightarrow \mathbb{P}^2\). This conjecture generalizes the classical Little Picard Theorem on the hyperbolicity of \(\mathbb{P}^1 \setminus \{\text{3 points}\}\), and is supported by explicit examples constructed by Zaidenberg \cite{Zaidenberg1988}. For generic curves \(\mathcal{C}\) of sufficiently large degree in \(\mathbb{P}^2\), the conjecture was confirmed by Siu and Yeung \cite{siu_yeung1996} by constructing explicit logarithmic $2$-jet differentials with negative twists. Later, building on the work \cite{Demailly-Elgoul2000}, El Goul \cite{Elgoul2003} improved the degree bound to \(15\) for \emph{very generic} \(\mathcal{C}\). Combining this with Siu's strategy of using slanted vector fields \cite{Siu2004} and a technique of Păun~\cite{mihaipaun2008}, Rousseau \cite{Rousseau2009} further reduced the degree bound to \(14\).
These results for high-degree curves naturally extend to configurations with multiple components. In the case where the curve \(\mathcal{C} = \cup_{i=1}^q \mathcal{C}_i\) decomposes into the union of \(q \geqslant 2\) generic algebraic curves \(\mathcal{C}_i\) of degree \(d_i\), the hyperbolicity of the complement \(\mathbb{P}^2 \setminus \mathcal{C}\) has been established with smaller degree bounds \(d = \sum_{i=1}^q d_i\) in various works. For instance, a result in \cite{babets1984}
settles the case where \(q \geqslant 5\). The case \((q = 4, \sum_{i=1}^4 d_i \geqslant 5)\) was treated in \cite{DSW1995-AJM}. The configuration of three conics was considered in \cite{Grauert1989, DSW1995-AJM, Francois-Duval-2001} by different methods. Especially, in \cite{DSW1995-duke}, the authors employed the symmetric differentials $1$-forms to deal with the case of two conics and one cubic. Rousseau \cite{Rousseau2009} addressed the case \(q = 2\) under the assumptions \((d_1 = 1, d_2 \geqslant 11)\), \((d_1 = 2, d_2 \geqslant 8)\), \((d_1 = 3, d_2 \geqslant 5)\), or \((d_1 = 4, d_2 \geqslant 4)\). Notably, for $q \geqslant 3$, Noguchi, Winkelmann, and Yamanoi \cite{NWA2007} established the \textit{algebraic degeneracy} of every entire curve in $\mathbb{P}^2 \setminus \mathcal{C}$ under the mild condition $\sum_{i=1}^q d_i \geqslant 4$. Their result also extends to the corresponding higher-dimensional setting.
%In the complex projective plane \(\mathbb{P}^2\), the logarithmic Kobayashi conjecture posits that if \(\mathcal{C} \subset \mathbb{P}^2\) is a \emph{generic} algebraic curve of degree \(d \geqslant 5\), then the complement \(\mathbb{P}^2 \setminus \mathcal{C}\) should be hyperbolic \cite{Kobayashi1970}. Here, hyperbolicity means that \(\mathbb{P}^2 \setminus \mathcal{C}\) does not contain the image of any nonconstant holomorphic map \(f: \mathbb{C} \rightarrow \mathbb{P}^2\). This conjecture generalizes the classical Little Picard Theorem, which asserts the hyperbolicity of \(\mathbb{P}^1 \setminus \{\text{3 points}\}\), and is supported by explicit examples constructed by Zaidenberg \cite{Zaidenberg1988} (see also \cite{azukawa-suzuki1980, masuda_noguchi1996, siu_yeung1997, shiffman_zaidenberg2002_pn} for further examples). For generic curves \(\mathcal{C}\) of sufficiently large degree in \(\mathbb{P}^2\), the conjecture was confirmed by Siu and Yeung \cite{siu_yeung1996} by constructing explicit special 2-jets. Later, building on the work of \cite{Demailly-Elgoul2000}, using invariant 2-jets, El Goul \cite{Elgoul2003} improved the degree bound to \(15\) for \emph{very generic} \(\mathcal{C}\). Combining this with Siu's technique of slanted vector fields \cite{Siu2004}, Rousseau \cite{Rousseau2009} further reduced the degree bound to \(14\).
%In the more tractable case where the curve \(\mathcal{C} = \cup_{i=1}^q \mathcal{C}_i\) decomposes into the union of \(q \geqslant 2\) generic algebraic curves \(\mathcal{C}_i\) of degree \(d_i\), the hyperbolicity of the complement \(\mathbb{P}^2 \setminus \mathcal{C}\) has been established with smaller degree bounds \(d = \sum_{i=1}^q d_i\) in various works. For instance, results in \cite{babets1984, Grauert1989, dethloff_schumacher_wong1995, Francois-Duval-2001} cover cases where \((q \geqslant 5)\), \((q = 4, \sum_{i=1}^4 d_i \geqslant 5)\), and \((q = 3, d_i \geqslant 2)\). Rousseau \cite{Rousseau2009} addressed the case \(q = 2\) under the assumptions \((d_1 = 1, d_2 \geqslant 11)\), \((d_1 = 2, d_2 \geqslant 8)\), \((d_1 = 3, d_2 \geqslant 5)\), or \((d_1 = 4, d_2 \geqslant 4)\). Notably, for \(q \geqslant 3\), the \textit{algebraic degeneracy} of every holomorphic curve into the complement \(\mathbb{P}^2 \setminus \mathcal{C}\) was established by Noguchi, Winkelmann, and Yamanoi \cite{NWA2007} under the mild condition \(\sum_{i=1}^q d_i \geqslant 4\).
Thus, in the aforementioned settings, every algebraically nondegenerate entire curve \(f: \mathbb{C} \rightarrow \mathbb{P}^2\) must intersect the curve \(\mathcal{C}\). In this article, we aim to quantify these geometric results by means of Nevanlinna theory. Specifically, we aim to establish a Second Main Theorem type estimate:
\begin{equation}
\label{SMT}
T_f(r) \leqslant c \sum_{i=1}^q N_f^{[k]}(r, \mathcal{C}_i) + o\big(T_f(r)\big) \quad\parallel,
\end{equation}
where
\[
N^{[k]}(r, \mathcal{C}_i) := \int_1^r \sum_{|z| < t} \min\{k, \ord_z f^* \mathcal{C}_i\} \frac{\dif t}{t} \quad (i = 1, \dots, q; \, r > 1)
\]
are the \textsl{\(k\)-truncated counting functions}, measuring the frequency of intersections \(f(\mathbb{C}) \cap \mathcal{C}_i\); and
\[
T_f(r) := \int_1^r \frac{\dif t}{t} \int_{|z| < t} f^* \omega_{\text{FS}} \quad (r > 1)
\]
is the \textsl{order function}, capturing the area growth rate of the holomorphic discs \(f(\{|z| < t\})\) with respect to the Fubini-Study metric \(\omega_{\text{FS}}\) on $\mathbb{P}^2$. The constant \(c\) on the right-hand side of~\eqref{SMT} is independent of \(r\), and the notation ``\(\parallel\)'' indicates that the inequality \eqref{SMT} holds for all \(r > 1\) outside a set of finite Lebesgue measure.
When \(\mathcal{C}\) has at least \(4\) irreducible components, such a Second Main Theorem type estimate with truncation at level \(2\) was established by H. Cartan \cite{Cartan1933} in the linear case where \(\mathcal{C}_i\) are lines:
\[
T_f(r) \leqslant \frac{1}{q-3} \sum_{i=1}^q N_f^{[2]}(r, \mathcal{C}_i) + o\big(T_f(r)\big)\quad \parallel.
\]
In the nonlinear case, analogous results (without truncation) were obtained by Eremenko and Sodin \cite{Eremenko-Sodin1992} via pluripotential theory, and by Ru \cite{Minru2009} through a synthesis of Cartan's method and Diophantine geometry. These result also extends to the corresponding higher-dimensional setting. For further background and recent advances in Nevanlinna theory, see \cite{Noguchi-Winkelmann2014, Ru21}.
%%The problem of establishing \eqref{SMT} remains largely open when \(\mathcal{C}\) has \(q \leqslant 3\) components. In \cite{Huynh-Vu-Xie-2017}, using methods developed for the Green-Griffiths conjecture, we addressed the case \(q = 1\) with a degree bound \(d \geqslant 660\). When the degree of \(\mathcal{C}\) is low, the problem becomes more difficult to track, according to the motto "the lower the degree the harder the hyperbolicity" \cite[p.~3]{Duval2021}.
In the prior work \cite{CHSX2025}, the authors established a Second Main Theorem for $n+1$ hypersurfaces $\{D_i\}_{i=1}^{n+1}$ in $\mathbb{P}^n$ in general position and intersecting transversally, proving the defect relation
\begin{equation}
\label{CHSX-defect}
\sum_{i=1}^{n+1} \delta_f(D_i) < n+1
\end{equation}
for any algebraically nondegenerate entire curve $f: \mathbb{\mathbb{C}}\rightarrow \mathbb{P}^n$
under the optimal total degree assumption $\sum_{i=1}^{n+1} \deg D_i \geqslant n+2$.
Here,
each defect $\delta_f(D_i)$ is defined by
\[
\delta_f(D_i)
\coloneqq 1-\limsup_{r \to \infty} \frac{N_f(r, D_i)}{\deg(D_i) \, T_f(r)}.
\]
The proof combined parabolic Nevanlinna theory \cite{Siboni-Paun2014} with the Second Main Theorems for semi-Abelian varieties \cite{Noguchi-Yamanoi-Winkelmann2002} and for $n+2$ hypersurfaces~\cite{Ru2004}.
Subsequently, Ru and Wang \cite{Ru-Wang2025} extended the result~\eqref{CHSX-defect} to moving targets with truncation level~$1$, while preserving the upper bound $n+1$ on the right-hand side.
In the case of $n=2$, the defect relation~\eqref{CHSX-defect} might be viewed as the weakest, albeit nontrivial, form of a Second Main Theorem in the vein of~\eqref{SMT} with noneffective $c < \infty$. The primary objective of this paper is to establish an effective Second Main Theorem, with truncation level $1$,
for three generic conics with an explicit constant $c = 5$.
\subsection{Two Key Theorems}
The pursuit of a Second Main Theorem encounters a profound ontological obstacle: the very nature of entire curves exhibits an inherent and wild flexibility in large-scale geometry, as evidenced by the exotic examples in \cite{HX2021, Xie2024, WX2025, CFX2025}. Our strategy constrains this flexibility by leveraging additional algebro-geometric structures to extract the underlying rigidity of entire curves, building on the following two key results.
\begin{thm}[McQuillan~\cite{Mcquillan1998}]
\label{McQuillan's result}
Let $\mathcal{C}=\sum_{i=1}^q\mathcal{C}_i$ be a normal crossings divisor in $\mathbb{P}^2$, where each $\mathcal{C}_i$ is a smooth algebraic curve of degree $d_i$ with $d=\sum_{i=1}^q d_i\geqslant 4$. Let $f:\mathbb{C}\rightarrow\mathbb{P}^2$ be an algebraically nondegenerate entire holomorphic curve. If the lifting $f_{[1]}:\mathbb{C}\rightarrow X_1=\mathbb{P}\,T_{\mathbb{P}^2}(-\log \mathcal{C})$ is algebraically degenerate, then the following Second Main Theorem type estimate holds:
\[
(d-3)\,T_f(r)\leqslant N^{[1]}_f(r,\mathcal{C}) + o\big(T_f(r)\big)\quad\parallel.
\]
\end{thm}
\begin{thm}[{\cite[Theorem 3.1]{Huynh-Vu-Xie-2017}}]
\label{smt-form-logarithmic-diff-jet}
Let \(\mathcal{C} = \sum_{i=1}^q \mathcal{C}_i\) be a normal crossings divisor in \(\mathbb{P}^2\). For any logarithmic jet differential
\[
\omega \in H^0\big(\mathbb{P}^2, E_{k,m} T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big) \quad (k, m, t \geqslant 1),
\]
and any entire holomorphic curve \(f: \mathbb{C} \rightarrow \mathbb{P}^2\) such that \(f^*\omega \not\equiv 0\), the following Second Main Theorem type estimate holds:
\begin{equation}
\label{SMT2}
T_f(r) \leqslant \frac{m}{t} \sum_{i=1}^q N_f^{[1]}(r, \mathcal{C}_i) + O\big(\log T_f(r) + \log r\big)\quad\parallel.
\end{equation}
\end{thm}
For the definition of logarithmic jet differential bundles $E_{k,m} T_{\mathbb{P}^2}^*(\log \mathcal{C})$, see Section~\ref{sect: preliminaries}.
Theorem~\ref{smt-form-logarithmic-diff-jet} can be seen as a quantitative refinement of the Fundamental Vanishing Theorem for entire curves (cf.~\cite[pp.~55-59]{Green-Griffiths1980}, \cite[Lemma 1.2]{Siu-Yeung1996} and \cite[Corollary 7.9]{Demailly1997}). Although the original statement in \cite{Huynh-Vu-Xie-2017} is formulated for hypersurfaces in $\mathbb{P}^n$, the proof extends directly to the case of simple normal crossing divisors in projective manifolds (cf.~\cite[Theorem A10.7.2]{Ru21}).
%Theorem~\ref{smt-form-logarithmic-diff-jet} enables a direct application. Specifically, whenever the existence of sufficiently many nonzero logarithmic jet differentials ensures the complex hyperbolicity of the complement $\mathbb{P}^2 \setminus \mathcal{C}$, it facilitates the derivation of a \emph{quantitative} Second Main Theorem for entire curves in $\mathbb{P}^2$ intersecting $\mathcal{C}$. This passage from qualitative hyperbolicity to precise analytic inequalities represents a key step in our approach.
\subsection{Our First Result}
The first result of this paper is an application of Theorem~\ref{smt-form-logarithmic-diff-jet} in the case $k = 1$.
\begin{thm}
\label{smt for three components, case large degrees}
Let $\mathcal{C}=\cup_{i=1}^3\mathcal{C}_i$ be a union of three smooth curves $\mathcal{C}_i$ of degrees $d_i$ with simple normal crossings. Assuming that $d_1\geqslant d_2\geqslant d_3\geqslant 2$ and $d_1\geqslant 3$, so that the total degree $d=d_1+d_2+d_3\geqslant 7$. For a nonconstant entire holomorphic curve $f\colon \mathbb{C}\rightarrow\mathbb{P}^2$ whose image is not contained in any $\mathcal{C}_i$, one has the estimate
\begin{equation*}
%%\label{eq:smt-three-components, case large degrees}
T_f(r) \leqslant c \sum_{i=1}^3 N_f^{[1]}(r, \mathcal{C}_i) + o\big(T_f(r)\big) \quad\parallel,
\end{equation*}
for any positive constant $c$ satisfying
\[
c > \frac{1}{(d-3)\delta_1}, \quad \text{where } \quad { \delta_1 = \frac{1}{2} - \frac{\sqrt{9(d-1)^2 - 12(d_1d_2 + d_2d_3 + d_3d_1)}}{6(d-3)}}.
\]
\end{thm}
Here, the numerical assumption on $d_1, d_2, d_3$ is made to guarantee that the Euler characteristics are positive:
\[
\chi\big(S^mT_{\mathbb{P}^2}^*(\log \mathscr{C})\big)
= O(1)^+ \cdot m^3
+O(m^2)
> 0, \quad \forall\, m \gg 1.
\]
\begin{rem}\rm
In the case where $\mathcal{C}$ consists of two conics and one cubic, a direct computation shows that $\delta_1$ takes the value $\frac{6-\sqrt{33}}{12}$.
This leads to a Second Main Theorem of the form
\[
T_f(r) \leqslant c \sum_{i=1}^3 N_f^{[1]}(r, \mathcal{C}_i) + o\big(T_f(r)\big)\quad \parallel,
\]
valid for any positive constant $c$ satisfying
$
c > \frac{3}{6-\sqrt{33}} \approx 11.74\ldots$.
This quantifies the hyperbolicity result of $\mathbb{P}^2\setminus \mathcal{C}$ due to Dethloff--Schumacher--Wong \cite{DSW1995-duke}.
\end{rem}
\subsection{Proof of Theorem~\ref{smt for three components, case large degrees}}
First, if the image of \(f\) is contained in an irreducible curve \(\mathcal{D}\), then \(\mathcal{D}\) must intersect the configuration \(\{\mathcal{C}_i\}\) in at least three points~\cite[Prop.~4.1]{DSW1995-AJM}. In this special case, stronger results are already available in the literature~\cite[Chapter 4]{Ru21}.
We now assume that \(f\) is algebraically nondegenerate.
The key step is to find a nonzero global negatively twisted logarithmic symmetric differential form
\begin{equation}
\label{key 1-jet differential existence}
\omega \in H^0\big(\mathbb{P}^2,\, S^m\, T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}(-t)\big),
\quad m \geqslant 1,\ t \geqslant 1,\ m/t \leqslant c,\ \omega \not\equiv 0.
\end{equation}
\begin{proof}[Proof of Theorem~\ref{smt for three components, case large degrees} assuming~\eqref{key 1-jet differential existence}]
If $f^*\omega \not\equiv 0$, then Theorem~\ref{smt-form-logarithmic-diff-jet} yields the desired estimate. On the other hand, if $f^*\omega \equiv 0$, then we can apply McQuillan's Theorem~\ref{McQuillan's result} to obtain the stronger inequality
\[
(d - 3) \, T_f(r) \leqslant N_f^{[1]}(r, \mathcal{C}) + o(T_f(r))\quad \parallel.
\]
In either case, the conclusion of the theorem holds.
\end{proof}
The proof of~\eqref{key 1-jet differential existence} will be given in Section~\ref{section: The case of three curves with large degrees}, based on Riemann--Roch computations and a Bogomolov-type vanishing theorem (Theorem~\ref{thm:Bogomolov vanishing theorem}).
\subsection{Our Second, Main Result: Second Main Theorem for Three Conics.}\label{subsection: main theorem}
Having established an effective Second Main Theorem for configurations of large degree, we now turn to the more challenging case of three generic conics $\mathcal{C}$ in $\mathbb{P}^2$. While the $1$-jet method proves successful in high-degree regimes, it encounters a fundamental obstruction in this low-degree setting, as the Euler characteristics are now negative:
\[
\chi\big(S^mT_{\mathbb{P}^2}^*(\log \mathscr{C})\big)
= O(1)^- \cdot m^3 + O(m^2)
< 0, \quad \forall\, m \gg 1.
\]
\begin{ques}\rm
\label{prob:three-conics}
For a generic configuration of three conics $\mathcal{C} = \cup_{i=1}^3 \mathcal{C}_i$ in $\mathbb{P}^2$, does there exist some integer $m \geqslant 1$ such that
\[
H^0\bigl( \mathbb{P}^2, \, S^{m} T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-1) \bigr) \neq 0?
\]
Here and throughout, ``generic'' means outside a proper Zariski-closed subset of the parameter space.
\end{ques}
It is instructive to compare this with known borderline cases. On the one hand, if we replace the three conics $\mathcal{C}$ by certain special six lines, such an $m$ does exist (see~\cite[p.~133]{Hirzebruch} and also~\cite[Theorem 1.4]{VZ2000}). On the other hand, if $\mathcal{C}$ is a generic sextic curve, then no such $m$ exists~\cite[Lemma 1.4.1]{Elgoul2003}.
Since the answer to Question~\ref{prob:three-conics} is currently unknown,~\footnote{\ By applying the approach developed in this work, we can verify the vanishing for $1\leqslant m\leqslant 5$ for generic three conics.} the key assumption~\eqref{key 1-jet differential existence} cannot be verified a priori.
This situation vividly illustrates the principle that ``\textit{the lower the degree, the harder the hyperbolicity}''~\cite[p.~3]{Duval2021}.
To circumvent this obstacle, we employ invariant logarithmic $2$-jet differentials with negative twists.
\begin{thm}[Second Main Theorem for Three Conics]
\label{main theorem 3 conics}
For a generic configuration of three conics $\{\mathcal{C}_i\}_{i=1}^3$ in $\mathbb{P}^2$ and a nonconstant entire holomorphic curve $f \colon \mathbb{C} \to \mathbb{P}^2$ whose image is not contained in any $\mathcal{C}_i$, the following inequality holds:
\begin{equation}
\label{eq:smt-three-conics}
T_f(r) \leqslant 5 \sum_{i=1}^3 N_f^{[1]}(r, \mathcal{C}_i) + o\big(T_f(r)\big) \quad\parallel.
\end{equation}
\end{thm}
%Moreover, the term ``generic'' admits an explicit characterization: the three conics must form a simple normal crossing divisor and satisfy the vanishing condition~\eqref{key vanishings to define generic} specified below. In Section~\ref{section 7}, we develop an effective method which, in principle, can check whether condition~\eqref{key vanishings to define generic} holds for any given explicit equations of three simple normal crossing conics; the underlying algorithm, however, would require extensive computational resources in general. For brevity and to illustrate the method concretely, we therefore carry out the detailed computations only for some tractable Fermat-type case, see Appendix~\ref{app:maple}.
\begin{rmk}
The constant $5$ in this theorem is smaller than the constant ${\approx}11.74$ appearing in Theorem~\ref{smt for three components, case large degrees} for the case of two conics and one cubic. This may seem counterintuitive at first, since the larger total degree $2+2+3 > 2+2+2$ would generally suggest a smaller constant. The explanation lies in the method of proof: here we employ not only $1$-jet techniques but also the more refined $2$-jet approach. A detailed analysis using the latter is expected to yield sharper constants not only for the two-conics-and-one-cubic case but also for other configurations covered by Theorem~\ref{smt for three components, case large degrees}.
\end{rmk}
We now outline the main ideas underlying the proof of Theorem~\ref{main theorem 3 conics}. As in the proof of Theorem~\ref{smt for three components, case large degrees}, we can assume that \(f\) is algebraically nondegenerate.
The core idea of the proof is to seek two ``independent'', negatively twisted invariant logarithmic $2$-jet differentials $\omega_1$ and $\omega_2$ such that:
\begin{itemize}
\item If $f^*\omega_1 \not\equiv 0$ or $f^*\omega_2 \not\equiv 0$, then Theorem~\ref{smt-form-logarithmic-diff-jet} directly implies the desired Second Main Theorem type estimate.
\item If $f^*\omega_1 \equiv 0$ and $f^*\omega_2 \equiv 0$, then the ``independence'' of $\omega_1$ and $\omega_2$ ensures that the lifting \(f_{[1]}(\mathbb{C})\subset X_1\) is algebraically degenerate, where $X_1$ is the first level of the logarithmic Demailly–Semple tower (see Subsection~\ref{subsec:log DS tower}) for $(\mathbb{P}^2, \mathcal{C}=\cup_{i=1}^3\mathcal{C}_i)$. Hence McQuillan's Theorem~\ref{McQuillan's result} applies.
\end{itemize}
In either case, we obtain the desired Second Main Theorem type estimate.
\medskip
In practice, however, this conceptual framework faces substantial challenges. Controlling the common base locus of $\omega$ and $\omega'$---essential for ensuring their independence---requires detailed geometric information that is typically accessible only through explicit constructions of negatively twisted logarithmic jet differentials. While such constructions are feasible when the curve $\mathcal{C}$ is of high degree, as demonstrated by Siu and Yeung~\cite{siu_yeung1996}, the low-degree case presents significantly greater difficulties. To date, not a single such differential has been explicitly described for low-degree curves, due to the intricate algebraic complexity of jet differentials (cf. \cite{Merker2008}). Moreover, the subsequent elimination step introduces further difficulties, rendering a direct approach effectively intractable.
A notable illustration of these difficulties arises even in the $1$-jet case. Debarre~\cite{Debarre2005} conjectured that for a generic choice of high-degree hypersurfaces $H_1, \dots, H_c \subset \mathbb{P}^N$ with $c \geqslant N/2$, the complete intersection $X := H_1 \cap \cdots \cap H_c$ should have an ample cotangent bundle $T_X^*$, which is equivalent to the existence of sufficiently many negatively twisted $1$-jet differentials on $X$ with empty common base locus. The first proof of Debarre's ampleness conjecture was established by Xie~\cite{Xie2018}, using explicit $1$-jet differentials obtained by Brotbek~\cite{Brotbek2016Symmetricdifferentialforms} for Fermat-type hypersurfaces. A key step in Xie's approach is a core lemma~\cite[p.~969]{Xie2018} that overcomes substantial difficulties in the elimination process. Shortly thereafter, Brotbek and Darondeau~\cite{BD2018} found an alternative proof, which instead applies Nakamaye's theorem~\cite{Nakamaye} to tackle the elimination.
\subsection{Roadmap for Theorem~\ref{main theorem 3 conics}: Towards an Effective Finite Constant $c$}
We now outline the proof of our main result, focusing first on establishing an explicit finite constant $c$ in the inequality~\eqref{eq:smt-three-conics}. The argument proceeds in several steps.
\medskip
First, for generic three conics $\mathcal{C} = \mathcal{C}_1 + \mathcal{C}_2 + \mathcal{C}_3$, standard Riemann--Roch computations (detailed in Section~\ref{Section 3}) guarantee the existence of an irreducible, negatively twisted invariant logarithmic $2$-jet differential
\begin{equation}
\label{omega1, m, t}
\omega_1 \in H^0\big(\mathbb{P}^2, E_{2,m}T^*_{\mathbb{P}^2}(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big),
\end{equation}
with weighted degree $m \geqslant 1$, vanishing order $t \geqslant 1$.
If $f^*\omega_1 \not\equiv 0$, then Theorem~\ref{smt-form-logarithmic-diff-jet} immediately yields the desired Second Main Theorem with a finite constant $c$. We may therefore assume $f^*\omega_1 \equiv 0$ in what follows.
\smallskip
The subsequent strategy depends on the values of $(m, t)$:
\smallskip
\smallskip\noindent \textbf{Case 1: $t > 7$}. We apply Siu's strategy~\cite{Siu2004} of slanted vector fields, which realized in dimension two by Pa\u{u}n \cite{mihaipaun2008} (compact case) and Rousseau \cite{Rousseau2009} (log case), to differentiate $\omega_1$ by a slanted vector field with twist $\mathcal{O}_{\mathbb{P}^2}(7)$, thereby obtaining an ``independent'' nonzero invariant logarithmic $2$-jet differential
\[
\omega_2 \in H^0\big(\mathbb{P}^2, E_{2,m}T^*_{\mathbb{P}^2}(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t+7)\big).
\]
Consequently, the proof can be completed by standard arguments (see Section~\ref{section: Siu's fields}).
\medskip
\noindent\textbf{Case 2:\, $t < m$.}
Following an approach of Demailly–El Goul~\cite{Demailly-Elgoul2000} (see Section~\ref{sect: refined Demailly-EG}), we can show the existence of a nonzero invariant logarithmic $2$-jet differential $\omega_2$ with negative twist via Riemann-Roch computations. However, the definition of $\omega_2$ is subtle: it is not defined on the whole space $\mathbb{P}^2$, but only on the zero locus $\{\omega_1 = 0\}$ within the second level $X_2$ of the Demailly-Semple tower for $(\mathbb{P}^2, \mathcal{C})$; see Proposition~\ref{prop: bigness on Z}. We then distinguish two subcases:
\begin{itemize}
\item If $f^*\omega_2 \not\equiv 0$, we can derive a Second Main Theorem-type estimate; see Theorem~\ref{pro:SMT on Z}. The arguments are analogous to those for Theorem~\ref{smt-form-logarithmic-diff-jet} but are refined in a crucial way, using the structure of Wronskians (see Proposition~\ref{Wronskian's magic}). This refinement represents a new contribution of this paper.
\smallskip
\item If $f^*\omega_2 \equiv 0$, then the simultaneous vanishing of $f^*\omega_1$ and $f^*\omega_2$ implies that the lifting $f_{[1]}(\mathbb{C})$ is algebraically degenerate in the first level $X_1$ of the logarithmic Demailly--Semple tower for $(\mathbb{P}^2, \mathcal{C})$, and hence McQuillan's Theorem~\ref{McQuillan's result} applies.
\end{itemize}
\medskip
\noindent
\textbf{Case 3: The Remaining Possibilities.}
For the remaining pairs $(m, t)$ not covered by the previous two cases, we establish the following Key Vanishing Lemma I to rule out the existence of any nonzero $\omega_1$ in these settings. Note that vanishing for $(m, t)$ implies vanishing for all $(m, t')$ with $t' > t$. Thus we complete the proof of a Second Main Theorem for three generic conics with a finite constant $c$. Moreover, careful numerical computation shows that we can take $c = 19$; see Subsection~\ref{computing c}.
\subsection{New Vanishing Results and Proof Strategy}
\begin{namedthm*}{Key Vanishing Lemma I}\label{thm:vanishing1}
For a generic configuration of three conics $\mathcal{C} = \cup_{i=1}^3 \mathcal{C}_i$ in $\mathbb{P}^2$, one has:
\[
H^0\big(\mathbb{P}^2, E_{2, m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big) = 0
\]
for all $(m,t)$ in the set
$
\{ (3,3), (4,4), (5,5), (6,6), (7,7)\}.
$
\end{namedthm*}
While such vanishing results have long been recognized as fundamental in complex hyperbolicity theory, a proof has remained out of reach for lack of a viable method. The principal innovation of this paper is a novel methodological framework that resolves this impasse.
\smallskip
Our proof of Key Vanishing Lemma I proceeds in three steps, detailed in Subsection~\ref{subsection 7.2}:
\smallskip
\begin{itemize}
\item \textbf{Algebraic Reduction:} We begin by transforming the problem of constructing invariant logarithmic jet differentials into that of solving a specific, tractable linear system, leveraging the GAGA principle to apply algebraic techniques in an analytic setting. The central challenge then becomes establishing bounds on the exponents in the denominators of the associated coefficients. This challenge is addressed by a judicious choice of basis and an analysis of the transition matrices between different frames.
\smallskip
\item \textbf{Computational Verification:} For a set of well-chosen explicit conics, we provide a computer-assisted proof that the associated linear systems possess only the trivial zero solution, as verified via symbolic computation.
\smallskip
\item \textbf{Geometric Extension:} A semi-continuity argument globalizes the computationally verified vanishing results from a specific choice of three conics to generic ones, concluding the proof.
\end{itemize}
%The approach of Demailly and El Goul~\cite[Proposition 5.4]{Demailly-Elgoul2000} established the vanishing of global $2$-jet differentials of weighted degrees $3$, $4$, and $5$ for generic smooth surfaces of degree $d \geqslant 6$ in $\mathbb{P}^3$. This result was subsequently extended by El Goul~\cite[Lemma 1.4.4]{Elgoul2003} to the logarithmic case of the complement of a smooth curve of degree $d \geqslant 7$ in $\mathbb{P}^2$. Both proofs rely fundamentally on Pa\u{u}n's ``proportionality lemma'' and Nadel's construction of meromorphic connections~\cite{Nadel}. The differences in the underlying setting, however, render these approaches insufficient for proving our Key Vanishing Lemma I.
%Another possible approach is to adapt the argument of Pa\u{u}n \cite[pp.~886-889]{mihaipaun2008} and Rousseau \cite[Proposition 17]{Rousseau2009} by employing Siu's slanted vector fields with a twist by $\mathcal{O}_{\mathbb{P}^2}(3)$, aiming to show that a certain associated ordinary differential equation admits no nontrivial solution. If successful, this method would significantly reduce the number of vanishing results required (only $(m, t)=(3, 3)$ would be needed). However, in the case of three conics, we were unable to overcome the substantial algebraic complexity that arises in the analysis of the corresponding ODE; see also Remark~\ref{how to use O(3)}.
\begin{rmk}
An alternative approach to establishing such vanishing results is through cohomological diagram chasing based on the filtration structure~\eqref{filtration of E_2,m}. This method enables systematic construction of long exact sequences in cohomology, thereby facilitating stepwise computation of
$H^0\big(\mathbb{P}^2, E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C}) \otimes \mathcal{O}(-t)\big)$. When $m = t$ is not divisible by $3$, the vanishing can be reduced to that of certain logarithmic $1$-jet differentials with negative twists. This observation is due to Pengchao Wang; see Appendix~\ref{appendix:results of Pengchao Wang}.\footnote{\
If the answer to Question~\ref{prob:three-conics} is affirmative, McQuillan's Theorem~\ref{McQuillan's result} immediately yields an effective second main theorem for three generic conics. Should the answer be negative, Proposition~\ref{pro:vanishing of E_2,3q+r} automatically establishes the cases $(m, t)=(4, 4), (5, 5), (7, 7)$ in Key Vanishing Lemma I.
}
However, in more general cases—such as when $m > t$ or when $3$ divides $m = t$—this diagram-chasing approach becomes substantially more complicated and its feasibility remains uncertain. Determining the essential information required for the diagram chasing presents a significant challenge. In particular, computing $H^i$ for $i = 0, 1, 2$ for each term in the long exact sequence requires either analyzing the connecting homomorphisms from $H^i$ to $H^{i+1}$, or computing other $H^{i}$ groups that typically do not vanish, thereby introducing considerable additional complexity.
\end{rmk}
\begin{rmk}
To demonstrate the efficacy of our approach, we apply the systematic framework developed in this article to establish new vanishing lemmas about $2$-jet differentials, which in turn yield improved degree bounds for the Kobayashi hyperbolicity conjecture in complex dimension two. Specifically, we refine prior results by lowering Păun's~\cite{mihaipaun2008} bound from $18$ to $16$ for generic surfaces in $\mathbb{P}^3$, and Rousseau's~\cite{Rousseau2009} bound from $14$ to $12$ for complements of generic curves in $\mathbb{P}^2$. It should be noted that the proofs of these vanishing results require subtle and novel adaptations of our framework, and their application toward the improved bounds relies on
other ingredients. Complete details will be presented in a forthcoming paper.
\footnote{\ Moreover, our method shows promising potential for further improving these degree bounds to $15$ and $11$, respectively. The main obstacle lies in the substantial computational resources required to verify the necessary vanishing lemmas, which exceeds the capacity of standard computers and calls for parallel computing. The implementation is theoretically feasible but requires porting and optimizing our Maple code in C—a task we deem accessible but have not yet pursued.}
\end{rmk}
\subsection{Proof Blueprint of Theorem~\ref{main theorem 3 conics} with an Effective Constant $c=5$}\label{subsec:intro more vanishing results}
We first show that in the existence result for $\omega_1$ in~\eqref{omega1, m, t}, the ratio $t/m$ can be made close to $\frac{4-\sqrt{10}}{3}$; see Proposition~\ref{estimate of 2-jet threshold}. If $f^*\omega_1 \not\equiv 0$, then Theorem~\ref{smt-form-logarithmic-diff-jet} yields the estimate~\eqref{eq:smt-three-conics} with constant $c$ around $\frac{3}{4-\sqrt{10}} = 3.58\ldots$. This value represents the theoretical limit of our current approach. Our goal of achieving $c=5$ is strategically set near this limit.
\smallskip
In {\bf Case 1} ($t > 7$), to ensure that the application of Theorem~\ref{smt-form-logarithmic-diff-jet} to both $\omega_1$ and $\omega_2$ yields $c = 5$, we strengthen the conditions to $m/t \leqslant 5$ and $m/(t-7) \leqslant 5$, which is equivalent to $t \geqslant m/5 + 7$.
\smallskip
In {\bf Case 2} ($t < m$), we refine Demailly–El Goul's approach~\cite{Demailly-Elgoul2000} to estimate the ratio of the vanishing order of $\omega_2$ to its weighted degree. Through a detailed and refined analysis in Section~\ref{sect: refined Demailly-EG}, we show that the constant $c$ can be expressed as $2/ \tau_1(\frac{t}{m})$, where
{\footnotesize
\[
\tau_1(\tfrac{t}{m}) = \frac{3}{2}\left(4 - \frac{t}{m}\right) - \frac{\sqrt{3}}{2} \sqrt{32 - 8\frac{t}{m} + 3\left(\frac{t}{m}\right)^2}.
\]}
A direct computation shows that $2/ \tau_1(\frac{t}{m}) \leqslant 5$ is equivalent to $t \leqslant \frac{92}{135}m$.
\smallskip
Although most positive integer pairs $(m,t)$ satisfy either $t \leqslant \frac{92}{135}m$ or $t \geqslant m/5 + 7$, exceptional cases remain where neither condition holds. To address these, we need the following new vanishing results:
\begin{namedthm*}{Key Vanishing Lemma II}\label{thm:vanishing}
For a generic configuration of three conics $\mathcal{C} = \cup_{i=1}^3 \mathcal{C}_i$ in $\mathbb{P}^2$, one has:
\begin{equation}
\label{key vanishings to define generic}
H^0\big(\mathbb{P}^2, E_{2, m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big) = 0
\end{equation}
for all $(m,t)\in
\{ (3,3), (4,3), (5,4), (6,5), (7,5), (8,6), (9,7), (10,7), (11,8), (12,9), (13,9)\}$.
\end{namedthm*}
The proof (see Subsection~\ref{subsection 7.3}) introduces two new ingredients: the utilization of Fermat-type conics to facilitate computation; and a modulo $p=5$ technique that leverages Maple to effectively perform Gaussian elimination, thus proving that the associated system of linear equations with integer coefficients has only the trivial solution.
\medskip
\subsection{A Challenge}
To avoid the use of any Key Vanishing Lemma and directly obtain, for three generic conics $\mathcal{C}$, a constant
\[
c \approx \frac{1}{\overline{\theta}_2(\mathbb{P}^2, \mathcal{C})} < 3.58\ldots\quad\text{(see Proposition~\ref{estimate of 2-jet threshold})}
\]
in the Second Main Theorem, it suffices to verify the following conjecture.
\begin{conj}\label{key question}
Let $\mathcal{C} \subset \mathbb{P}^2$ be a configuration of three conics with simple normal crossings. For any $\epsilon > 0$, there exists a nonzero negatively twisted invariant logarithmic $2$-jet differential
\[
\omega \in H^0\!\left( \mathbb{P}^2,\; E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t) \right)
\]
with $ \overline{\theta}_2(X, \mathcal{C}) - \epsilon<\tfrac{t}{m}<\overline{\theta}_2(X, \mathcal{C})$, such that its zero locus $\{\omega = 0\} \subset X_2$ is irreducible and reduced modulo $\Gamma_2$.
\end{conj}
We expect that such a \emph{multiplicity-one} property to remain valid in a more general context, in particular for generic members $\mathcal{C}$ of a sufficiently varying family.
\medskip
\subsection{Structure of the Paper}
Section~\ref{sect: preliminaries} provides necessary background. Section~\ref{section: The case of three curves with large degrees} presents detailed Riemann–Roch computations that form the core of the proof of Theorem~\ref{smt for three components, case large degrees}. Section~\ref{Section 3} advances to more sophisticated Riemann–Roch computations, specifically designed to establish the existence of negatively twisted invariant logarithmic $2$-jet differentials for three generic conics. Section~\ref{section: Siu's fields} details the application of Siu's slanted vector fields in \textbf{Case 1}. Section~\ref{sect: refined Demailly-EG} then introduces a refinement of the Demailly–El Goul argument, adapted to handle \textbf{Case 2}. The core contributions of the paper are contained in Section~\ref{section 7}, which presents the proofs of Key Vanishing Lemmas~I and~II. Finally, Appendix~\ref{app:maple} provides the Maple code used to verify these key vanishing results. Appendix~\ref{appendix:results of Pengchao Wang} presents an observation by Pengchao Wang.
\medskip
\section{\bf Preliminaries}
\label{sect: preliminaries}
%In this section, we briefly recall some fundamental concepts and results concerning jet differentials, the Demailly–Semple tower, and a Bogomolov-type vanishing theorem. For a more comprehensive treatment, we refer the reader to \cite{Bloch1926, Green-Griffiths1980, Noguchi1986, Siu-Yeung1996, siu_yeung1997, Demailly1997, Dethloff-Lu2001, Elgoul2003, Siu2004, mihaipaun2008, Rousseau2009}.
\subsection{Logarithmic Jet Bundles and Invariant Differential Forms}
Let $X$ be an $n$-dimensional complex manifold. The $k$-th jet bundle $J_k(X)$ is defined by considering holomorphic germs $f \colon (\mathbb{C}, 0) \to (X, x)$, where two germs are considered equivalent if their Taylor expansions coincide up to order $k$ in some local coordinate system. The equivalence class of a germ $f$, denoted by $j_k(f)$, is said to be regular if $f'(0) \ne 0$.
The total space
$
J_k(X) = \cup_{x \in X} J_k(X)_x
$
carries a natural topology and, together with the natural projection $\pi_k \colon J_k(X) \to X$, forms a holomorphic fiber bundle. Note that $J_1(X)$ coincides with the tangent bundle $T_X$.
Now let $D \subset X$ be a normal crossing divisor. The logarithmic cotangent bundle $T_X^*(\log D)$ is a locally free sheaf generated by meromorphic $1$-forms with at most logarithmic poles along $D$. Precisely, for $x \in X \setminus D$, the stalk $T_{X,x}^*(\log D)$ coincides with the stalk $T_{X,x}^*$ of the usual cotangent sheaf $T_X^*$ at $x$. For $x \in D$, take a neighborhood $U$ of $x$ and irreducible holomorphic functions $\eta_1, \dots , \eta_{\ell}$ on $U$ such that $D \cap U = \{ \eta_1 \cdots \eta_{\ell} = 0 \}$; then the stalk is given by $T_{X,x}^*(\log D) = \sum_{j=1}^{\ell} \mathcal{O}_{X,x} \frac{\mathrm{d} \eta_j}{\eta_j} + T_{X,x}^*$. Its dual, the logarithmic tangent bundle $T_X(-\log D)$, consists of holomorphic vector fields tangent to each component of $D$ (cf.~\cite{Noguchi1986, Elgoul2003}).
Consider an open subset $U \subset X$, a section $\omega \in H^0(U, T_X^*(\log D))$, and a jet $j_k(f)$ in $J_k(X)|_U$. The pullback $f^*\omega$ takes the form $A(z) \mathrm{d}z$ for some meromorphic function $A$, and the derivatives $A^{(j)}(0)$ for $0 \leqslant j \leqslant k-1$ depend only on the jet $j_k(f)$. This construction induces a well-defined map:
\begin{equation}\label{trivialization-jet}
\tilde\omega \colon J_k(X)|_U \to (\mathbb{C} \cup \{\infty\})^k, \quad j_k(f) \mapsto \big( A(0), A'(0), \dots, A^{(k-1)}(0) \big).
\end{equation}
A holomorphic section $s \in H^0(U, J_k(X))$ is called a logarithmic $k$-jet field on $U$ if for every open subset $U' \subset U$ and every $\omega \in H^0(U', T_X^*(\log D))$, the composition $\tilde\omega \circ s$ is holomorphic on $U'$, where $\tilde\omega$ is defined as in \eqref{trivialization-jet}. These logarithmic $k$-jet fields form a subsheaf of $J_k(X)$, which corresponds to the sections of a holomorphic fiber bundle over $X$ known as the logarithmic $k$-jet bundle along $D$, denoted $J_k(X, -\log D)$ (cf.~\cite{Noguchi1986, Elgoul2003, Rousseau2009}).
Furthermore, given a local basis $\omega_1, \dots, \omega_n$ of $T_X^*(\log D)$ on $U$, we obtain a trivialization:
\begin{equation*}
H^0(U, J_k(X, -\log D)) \to U \times (\mathbb{C}^k)^n, \quad \sigma \mapsto \big( \pi_k \circ \sigma;\; \tilde\omega_1 \circ \sigma, \dots, \tilde\omega_n \circ \sigma \big).
\end{equation*}
A logarithmic jet differential of order $k$ and weighted degree $m$ at a point $x \in X$ is a polynomial $Q(f^{(1)}, \dots, f^{(k)})$ on the fiber of $J_k(X, -\log D)$ over $x$, satisfying the weighted homogeneity condition:
\begin{equation*}
Q \big( \lambda f', \lambda^2 f'', \dots, \lambda^k f^{(k)} \big) = \lambda^m Q \big( f', f'', \dots, f^{(k)} \big),
\end{equation*}
for all $\lambda \in \mathbb{C}^*$ and all $(f', f'', \dots, f^{(k)}) \in J_k(X, -\log D)_x$.
The vector space $E_{k,m}^{GG}T_X^*(\log D)_x$ consists of all such jet differential polynomials defined at the point $x \in X$. Then the total space
\begin{equation*}
E_{k,m}^{GG}T_X^*(\log D) := \cup_{x \in X} E_{k,m}^{GG}T_X^*(\log D)_x,
\end{equation*}
acquires the structure of a holomorphic vector bundle over $X$ by Faà di Bruno’s formula (cf. e.g.~\cite{Constantine1996, Merker2015}), referred to as the logarithmic Green–Griffiths bundle of jet differentials of order $k$ and weighted degree $m$.
Let $\mathbb{G}_k$ denote the group of germs of $k$-jets of biholomorphisms of $(\mathbb{C}, 0)$:
\begin{equation*}
\mathbb{G}_k \coloneqq \Big{\{} \varphi \colon t \mapsto \sum_{j=1}^k c_j t^j \;\Big|\; c_1 \in \mathbb{C}^*,\; c_j \in \mathbb{C} \text{ for } j \geqslant 2 \Big{\}},
\end{equation*}
where the group operation is defined by composition modulo $t^{k+1}$. This group acts naturally on $J_k(X)$ via the right action:$\varphi \cdot j_k(f) \coloneqq j_k(f \circ \varphi)$.
A logarithmic jet differential operator $P \in \Gamma\big(U, E_{k,m}^{GG}T_X^*(\log D)\big)$ over an open set $U \subset X$ is said to be invariant under the reparametrization group $\mathbb{G}_k$ if for every $\varphi \in \mathbb{G}_k$ the following identity holds:
\begin{equation}
\label{define invariant jet differentials}
P \big( (f \circ \varphi)', (f \circ \varphi)'', \dots, (f \circ \varphi)^{(k)} \big) = (\varphi')^m \cdot P \big( f', f'', \dots, f^{(k)} \big).
\end{equation}
The subbundle $E_{k,m} T_X^*(\log D) \subset E_{k,m}^{GG}T_X^*(\log D)$ is then defined as the holomorphic vector bundle whose fibers consist of all such invariant logarithmic jet differential polynomials.
This notion was initially introduced by Demailly \cite{Demailly1997} in the compact case, and subsequently generalized to the logarithmic case by Dethloff-Lu \cite{Dethloff-Lu2001}. %As noted in \cite[p.~4]{Demailly1997}, The subbundle $E_{k,m}$ turns out to have better positivity properties than $E_{k,m}^{GG}$. At least in the case of surfaces of general type, the conditions involved for the existence of sections in $E_{k,m}$ are better than those for $E_{k,m}^{GG}$.
Throughout this paper, we only concern the situation $k=1$ or $2$. In the case that $\dim_{\mathbb{C}} X = 2$, the bundle of invariant logarithmic jet differentials of order $2$ admits a filtration \cite{Demailly1997}:
\begin{equation}\label{filtration of E_2,m}
\mathrm{Gr}^{\bullet} E_{2,m} T_X^*(\log D) = \bigoplus_{0 \leqslant j \leqslant \lfloor m/3 \rfloor} S^{m-3j} T_X^*(\log D) \otimes \overline{K}_X^{j},
\end{equation}
where the logarithmic conincal line bundle $\overline{K}_X \coloneqq K_X \otimes \mathcal{O}_X(D)$. This filtration corresponds to the expression of an invariant logarithmic jet differential $P(f', f'')$ outside $D$ as:
\begin{equation*}
P(f', f'') = \sum_{0 \leqslant j \leqslant \lfloor m/3 \rfloor} \;
\sum_{\substack{\alpha_1 + \alpha_2 = m - 3j \\ \alpha_1, \alpha_2\geqslant 0}}
R_{\alpha_1, \alpha_2, j}(f_1, f_2) \cdot
(f_1')^{\alpha_1} (f_2')^{\alpha_2} (f_1' f_2'' - f_2' f_1'')^j,
\end{equation*}
where $f=(f_1, f_2)$ represents the holomorphic germs in local coordinates, and where $R_{\bullet}$ are holomorphic functions.
\subsection{Logarithmic Demailly-Semple Tower}\label{subsec:log DS tower}
The main reference of this subsection is~\cite{Demailly1997, Dethloff-Lu2001}.
For simplicity, consider the log-direct manifold $(X_0,D_0,V_0)$ associated with three generic conics $\mathcal{C}_1$, $\mathcal{C}_2$, $\mathcal{C}_3$ in the complex projective plane $\mathbb{P}^2$, where $X_0=X:=\mathbb{P}^2$, $D_0:=\mathcal{C}=\sum_{i=1}^3\mathcal{C}_i$ is a simple normal crossing divisors on $X$ whose irreducible components $\mathcal{C}_i$ are smooth conics and $V_0:=T_X(-\log D_0)$. Define $X_1:=\mathbb{P}(V_0)$ with the natural projection $\pi_{1,0}:X_1\rightarrow X_0$. Set $D_1:=\pi_{1,0}^*D_0$. The subbundle $V_1\subset T_{X_1}(-\log D_1)$ is defined fiberwise by:
\begin{equation}
\label{define V_1}
V_{1,(x,[v])}:=\{\xi\in T_{X_1,(x,[v])}(-\log D_1)\, |\, \left(\pi_{1,0}\right)_{*}\xi\in \mathbb{C} \cdot v\},
\end{equation}
with $\mathbb{C} \cdot v \subset V_{0,x} = T_{X_{0},x}(-\log D_0)$. Thus we obtain the first level $(X_1, D_1, V_1)$ of the logarithmic Demailly-Semple tower for $(\mathbb{P}^2, \mathcal{C})$.
Proceeding likewise, we can construct the second level of logarithmic Demailly–Semple tower:
\[
(X_2, D_2, V_2)\rightarrow (X_1,D_1,V_1)\rightarrow(X_0,D_0,V_0),
\]
together with the projections
$\pi_{2,1} \colon X_2\rightarrow X_1$ and
$\pi_{2,0} \colon X_2\rightarrow X_0$.
Let $\mathcal{O}_{X_{1}}(-1)$ denote the tautological line bundle of $X_{1} = \mathbb{P}(V_0)$ such that $\mathcal{O}_{X_{1}}(-1)_{(x,[v])} = \mathbb{C} \cdot v$. By definition, there is a short exact sequence
\begin{equation}\label{SES:V_1}
\xymatrix{
0 \ar[r] & T_{X_{1} / X_{0}} \ar[r] & V_{1} \ar[r]^(0.35){\left(\pi_{1,0}\right)_{*}} & \mathcal{O}_{X_{1}}(-1) \ar[r] & 0
},
\end{equation}
where $T_{X_{1} / X_{0}}$ is the relative tangent bundle of the fibration $\pi_{1,0} \colon X_1 \rightarrow X_0$. And there is a relative version of the Euler sequence of the tangent bundle of the fiber $\mathbb{P}(V_{0,x})$:
\begin{equation}\label{ses:relative Euler sequence}
\xymatrix{
0 \ar[r] & \mathcal{O}_{X_1} \ar[r] & \pi_{1,0}^{*} V_{0} \otimes \mathcal{O}_{X_1}(1) \ar[r] & T_{X_{1} / X_{0}} \ar[r] & 0
}.
\end{equation}
Furthermore, there exists a canonical injection $\mathcal{O}_{X_{2}}(-1) \hookrightarrow \pi_{2,1}^{*} V_{1}$. Combining this with the above sequence yields a natural morphism of line bundles:
\begin{equation*}
\xymatrix@=1.5cm{
\mathcal{O}_{X_{2}}(-1) \ar@{^{(}->}[r] & \pi_{2,1}^{*} V_{1} \ar[r]^(0.4){ \pi_{2,1}^{*} \left(\pi_{1,0}\right)_{*}} & \pi_{2,1}^{*} \mathcal{O}_{X_{1}}(-1)
},
\end{equation*}
whose zero divisor is the hyperplane subbundle:
\begin{equation}\label{equ:def of Gamma_2}
\Gamma_2 \coloneqq \mathbb{P}(T_{X_{1} / X_{0}}) \subset \mathbb{P}(V_1) = X_2.
\end{equation}
Consequently, we can obtain the isomorphism:
\begin{equation}
\label{gamma_2=O(-1, 1)}
\mathcal{O}_{X_2}(\Gamma_2)
\cong
%%\mathcal{O}_{X_{2}}(-1,1),
\pi_{2,1}^{*} \mathcal{O}_{X_{1}}(-1)\otimes \mathcal{O}_{X_{2}}(1).
\end{equation}
\medskip
Let $f \colon \mathbb{D}_R \rightarrow X_0$ be a nonconstant holomorphic map from the disc $\mathbb{D}_R=\{z\in\mathbb{C}:|z|<R\}$. To define its liftings to the first level $X_1$ of the logarithmic Demailly–Semple tower, we consider the following two cases for $f'(t) \in T_{X_0 , f(t)}$:
When $0 \neq f'(t) \in T_{X_0 , f(t)} \left( - \log D_0 \right)$, The point $[f'(t)] \in \mathbb{P} \left( T_{X_0 , f(t)} \left( - \log D_0 \right) \right)$ is well defined;
When $f'(t_0) = 0$ or $f'(t_0) \in T_{X_0 , f(t_0)} \setminus T_{X_0 , f(t_0)} \left( - \log D_0 \right)$, we can define $[f'(t_0)] \in \mathbb{P} \left( T_{X_0 , f(t_0)} \left( - \log D_0 \right) \right)$ as following: near $t_0 \in \mathbb{D}_R$, choose a local frame $(e_1, e_2)$ for $T_{X_0}(-\log D_0)$ in a neighborhood of $f(t_0)$. In this frame, the derivative $f'(t)$ has a local expression $f'(t) = u_1(t) e_1 + u_2(t) e_2$, where $u_1(t), u_2(t)$ are meromorphic functions with at most finite-order poles. Let $\ell \in \mathbb{Z}$ be the minimum vanishing order (equivalently, the negative of the maximal pole order) of $u_1(t)$ and $u_2(t)$ at $t_0$. Factoring out $(t - t_0)^\ell$, we can write
\[
f'(t) = (t - t_0)^\ell \bigl( v_1(t) e_1 + v_2(t) e_2 \bigr),
\]
where $v_1(t)$ and $v_2(t)$ are holomorphic near $t_0$ and not simultaneously vanishing at $t_0$. We define the projective class:
\[
[f'(t_0)] \coloneqq [v_1(t_0) : v_2(t_0)] \in \mathbb{P}(T_{X_0, f(t_0)}(-\log D_0)).
\]
This definition is independent of the choice of local frame, so $[f'(t_0)] \in \mathbb{P} \left( T_{X_0 , f(t_0)} \left( - \log D_0 \right) \right)$ is well defined.
With this convention, the first lifting $f_{[1]} = (f; [f'])$ of $f$ is a well-defined holomorphic map from $\mathbb{D}_R$ to $X_1$. Repeating the process for $f_{[1]}$, the second lifting $f_{[2]} = (f_{[1]}; [f_{[1]}'])$ of $f$ is also well-defined and holomorphic from $\mathbb{D}_R$ to $X_2$.
Moreover, from the definition of $\Gamma_2$ and the identity $\pi_{1,0} \circ f_{[1]} = f$, it follows that $f_{[2]}(t) \in \Gamma_2$ if and only if $(\pi_{1,0})_* f_{[1]}'(t) = f'(t) = 0$.
Demailly~\cite[Corollary 5.12]{Demailly1997} observed that, for any $w_0 \in X_{2}$, there exists an open neighborhood $U_{w_0} \subset X_2$ such that for all the $w \in U_{w_0}$, one can find a holomorphic family of germs $f_{w} \colon (\mathbb{C},0) \rightarrow U$ such that $\left( f_{w} \right)_{[2]} (0) = w$ and $\left( f_{w} \right)_{[1]}'(0) \neq 0$. By the construction of the Demailly–Semple tower, we have $f_{w}'(t) \in \mathcal{O}_{X_1}(-1)_{\left( f_{w} \right)_{[1]}(t)}$ and $\left( f_{w} \right)_{[1]}'(t) \in \mathcal{O}_{X_2}(-1)_{\left( f_{w} \right)_{[2]}(t)}$. Then for any $P \in E_{2,m} T_X^*(\log D)_x$ with $x \in \pi_{2,0}(U_{w_0}) \subset X_0$, this induces a holomorphic section $\sigma_P (w)$ of $\mathcal{O}_{X_2}(m)$ over the fiber $X_{2,x}$ defined by:
\begin{equation*}
\sigma_P (w) \coloneqq P(f_{w}', f_{w}'')(0) \left( \left( f_{w} \right)_{[1]}'(0) \right)^{-m}.
\end{equation*}
More generally, the following fundamental isomorphism holds:
\begin{namedthm*}{Direct image formula}[cf.~{\cite[Theorm 6.8]{Demailly1997}}, {\cite[Propsition 3.9]{Dethloff-Lu2001}}]
The direct image sheaf
\begin{equation}\label{direct image formula}
\left( \pi_{2,0} \right)_{*} \mathcal{O}_{X_2} (m) \cong \mathcal{O} \left( E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \right)
\end{equation}
is isomorphic to the sheaf of invariant logarithmic $2$-jet differentials of weighted degree $m$.
\end{namedthm*}
\subsection{A Bogomolov-type Vanishing Theorem}
Later in Sections~\ref{section: The case of three curves with large degrees} and \ref{Section 3}, to establish the existence of negatively twisted logarithmic $1$-jet, and $2$-jet, differentials via Riemann--Roch, it suffices to verify the vanishing of the cohomology groups $H^2\big(\mathbb{P}^2, S^{m}T^*_{\mathbb{P}^2}(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big)$ and $H^2\big(\mathbb{P}^2, E_{2,m}T^*_{\mathbb{P}^2}(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big)$, respectively. Using Serre duality and the filtration structure in~\eqref{filtration of E_2,m}, this problem reduces to a Bogomolov-type vanishing theorem for a log pair $(X,D)$ of log general type, where $X$ is a smooth projective surface and $D$ is a reduced simple normal crossing divisor.
\begin{thm}[cf.~{\cite[Theorem 4.4]{Chen2025}}]\label{thm:Bogomolov vanishing theorem}
Let $(X,D)$ be a log pair where $X$ is a smooth projective surface and $D$ is a reduced simple normal crossing divisor.
Assume that $(X,D)$ is of general type (i.e., $K_{X} + D$ is big) and $K_{X} + D$ is nef. Then
\begin{equation*}
H^0 \big( X , S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D)) \big) = 0,
\end{equation*}
for all integers $p, q$ satisfying $p > 2q$.
\end{thm}
In the compact case, Bogomolov \cite{Bogomolov1978} established a seminal vanishing theorem under the assumption that tangent bundle $T_{X}$ is semi-stable. Later, Enoki \cite{Enoki1987} showed that $T_{X}$ is $K_{X}$-semi-stable for minimal K\"ahler space $X$ with big $K_{X}$, and Tsuji \cite{Tsuji1988} proved that $T_{X}$ is $K_{X}$-semi-stable for smooth minimal algebraic variety (i.e., $K_{X}$ is nef). Combining these results, the Bogomolov-type vanishing theorem holds for projective variety $X$ with big and nef $K_{X}$; see~\cite[Theorem 14.1]{Demailly1997}.
%For the logarithmic case, the proof of the theorem \ref{thm:Bogomolov vanishing theorem} also ties to the semi-stability of $T_{X}(- \log D)$. Under the assumptions that $K_{X} + D$ is ample and $T_{X}(- \log D)$ is semi-stable, a brief proof is provided in the Appendix. In our three conics case, $K_{\mathbb{P}^2}(\mathcal{C})$ is ample, so $T_{\mathbb{P}^2}(-\log\mathcal{C})$ is semi-stable with respect to $K_{\mathbb{P}^2}(\mathcal{C})$ by Tsuji's result \cite[Theorem 5.1]{Tsuji1988} (see also \cite{Guenancia2016}).
\medskip
\section{\bf Proof of Theorem~\ref{smt for three components, case large degrees}}
\label{section: The case of three curves with large degrees}
Let $\mathcal{C} = \sum_{i=1}^3 \mathcal{C}_i$ be a simple normal crossing divisor of total degree $d = d_1 + d_2 + d_3$, where $d_i = \deg \mathcal{C}_i$ satisfy
$
d_1 \geqslant d_2 \geqslant d_3 \geqslant 2$ and $d_1 \geqslant 3.
$
To derive an effective Second Main Theorem, we seek nontrivial global sections of the twisted sheaf
\[
S^m T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(1)^{-(d-3)\delta m}
\cong
S^m T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{\, -\delta m},
\]
where $\delta > 0$ is a small rational number to be chosen, and $m$ is taken sufficiently divisible.
\begin{defi}[cf.~{\cite[p.~517]{Demailly-Elgoul2000}}, {\cite[Definition 1.3.1]{Elgoul2003}}]
\normalfont
The \emph{$1$-jet log threshold} of $(\mathbb{P}^2,\mathcal{C})$ is defined by
\[
\overline{\theta}_1(\mathbb{P}^2,\mathcal{C})
\coloneqq
\sup\left\{
\frac{t}{m} \,\middle|\, m, t > 0,\;
H^0\big(\mathbb{P}^2, S^m T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big) \neq 0
\right\}.
\]
If no such pair $(m, t)$ exists, the threshold is defined to be zero.
\end{defi}
\begin{pro}[Proof of~\eqref{key 1-jet differential existence}]
\label{existence of symmetric differentials 1 form, three components case}
Let $\mathcal{C}=\cup_{i=1}^3\mathcal{C}_i$ be a union of three smooth curves $\mathcal{C}_i$ of degrees $d_i$ with simple normal crossings. Assume that $d_1\geqslant d_2\geqslant d_3\geqslant 2$ and $d_1\geqslant 3$. Then one has the estimate
\[
0 < \frac{1}{\overline{\theta}_1(\mathbb{P}^2,\mathcal{C})} \leqslant \frac{1}{(d-3)\delta_1},
\]
where{\footnotesize
\begin{equation}
\label{definition of delta 1, three components case with large degree}
\delta_1 = \frac{1}{2}-\dfrac{\sqrt{9(d-1)^2-12(d_1d_2+d_2d_3+d_3d_1)}}{6(d-3)}>0.
\end{equation}}
\end{pro}
\begin{proof}
We follow the standard Riemann-Roch argument (cf.~\cite{Demailly-Elgoul2000, Elgoul2003}).
By Serre duality, we obtain
\begin{align*}
\label{use Serre duality_1 jet}
H^2\big(\mathbb{P}^2,S^mT_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\big)
& = H^0\big(\mathbb{P}^2,K_{\mathbb{P}^2}\otimes S^mT_{\mathbb{P}^2}(\log\mathcal{C})\otimes \overline{K}_{\mathbb{P}^2}^{\delta m}\big)\\
&=
H^0\big(\mathbb{P}^2, S^mT_{\mathbb{P}^2}(\log\mathcal{C})\otimes \overline{K}_{\mathbb{P}^2}^{\delta m-\frac{3}{(d-3)}}\big).
\end{align*}
For any rational number $\delta\in (0,\frac{1}{2})$ and sufficiently large $m$, the inequality
$
m > 2(\delta m - \frac{3}{d-3})
$
holds. Applying the Bogomolov-type Vanishing Theorem \ref{thm:Bogomolov vanishing theorem}, we obtain
\[
H^2\big(\mathbb{P}^2,S^mT_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\big)=0
\]
which yields
\[
\dim H^0\Big(\mathbb{P}^2, S^mT_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\Big)
\geqslant \chi\Big(S^mT_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\Big),
\]
for large $m$. A standard Riemann-Roch computation gives
\begin{align*}
\chi\Big(S^mT_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\Big)
&=
\dfrac{1}{6}\big((m-\delta m)^3-(-\delta m)^3\big)\bar{c}_1^2
-
\dfrac{1}{6}m^3\bar{c}_2+O(m^2)\\
&=\dfrac{1}{2}m^3\bigg((d-3)^2\delta^2-(d-3)^2\delta+\frac{1}{3}(d_1d_2+d_2d_3+d_3d_1)-d+2\bigg)+O(m^2).
\end{align*}
The leading coefficient $\varphi(\delta)=(d-3)^2\delta^2-(d-3)^2\delta+\frac{1}{3}(d_1d_2+d_2d_3+d_3d_1)-d+2$ has two positive roots given by
{\footnotesize
\[
\delta_1 = \frac{1}{2}-\dfrac{\sqrt{9(d-1)^2-12(d_1d_2+d_2d_3+d_3d_1)}}{6(d-3)}, \quad \quad \delta_2 = \frac{1}{2}+\dfrac{\sqrt{9(d-1)^2-12(d_1d_2+d_2d_3+d_3d_1)}}{6(d-3)}>0.
\]} We can check the
positivity of $\delta_1$ directly. Hence $\varphi(\delta) > 0$ for all $\delta \in (0,\delta_1)$. Thus we conclude the proof.
\end{proof}
\medskip
\section{\bf Twisted Logarithmic $2$-Jet Differentials and Threshold}
\label{Section 3}
To obtain a constant $c$ as small as possible in the Second Main Theorem for three conics $\mathcal{C} \subset \mathbb{P}^2$,
by Theorem~\eqref{smt-form-logarithmic-diff-jet}, we seek negatively twisted invariant logarithmic $2$-jet differentials
with relatively high vanishing order $t$ compared to the degree $m$. This motivates the study of the following threshold:
\begin{defi}[cf.~{\cite[p.~517]{Demailly-Elgoul2000}}, {\cite[Definition 1.3.1]{Elgoul2003}}]
\normalfont
The \emph{$2$-jet log threshold} of $(\mathbb{P}^2,\mathcal{C})$ is defined by
\[
\overline{\theta}_2(\mathbb{P}^2, \mathcal{C})
\coloneqq
\sup\left\{
\frac{t}{m} \,\middle|\, m, t > 0,\;
H^0\big(X_2, \mathcal{O}_{X_2}(m) \otimes \pi_{2,0}^* \mathcal{O}_X(-t)\big) \neq 0
\right\}.
\]
If no such pair $(m, t)$ exists, the threshold is defined to be zero.
\end{defi}
If $\overline{\theta}_2 > 0$, then for any sufficiently small $\epsilon > 0$, there exists a negatively twisted logarithmic $2$-jet differential
\[
\omega \in H^0\big(\mathbb{P}^2, E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big)
\]
with ratio $\overline{\theta}_2 - \epsilon < \frac{t}{m} \leqslant \overline{\theta}_2$.
Furthermore, if $f^* \omega \not\equiv 0$ also holds, then Theorem~\ref{smt-form-logarithmic-diff-jet} yields the Second Main Theorem with constant
\[
\frac{1}{\overline{\theta}_2} \leqslant c = \frac{m}{t} < \frac{1}{\overline{\theta}_2 - \epsilon}.
\]
The significance of the $2$-jet log threshold stems from its role as the fundamental barrier that dictates the possibility of constructing jet differentials with a prescribed ratio of vanishing orders to weighted degrees.
The following result establishes an explicit upper bound for $\frac{1}{\overline{\theta}_2}$ in the case of three smooth conics with simple normal crossings:
\begin{pro}\label{estimate of 2-jet threshold}
Let $\mathcal{C}=\cup_{i=1}^3\mathcal{C}_i$ be a union of three smooth conics $\mathcal{C}_i$ with simple normal crossings. Then one has the estimate
\[
0 < \frac{1}{\overline{\theta}_2(\mathbb{P}^2, \mathcal{C})} < \frac{3}{4-\sqrt{10}}=3.58\ldots.
\]
\end{pro}
The proof employs a Riemann–Roch calculation, which is standard (cf. \cite{Demailly-Elgoul2000}).
\begin{proof}
We seek nonzero invariant logarithmic $2$-jet differential forms with high vanishing order, which correspond to global sections of the twisted vector bundle
\[
E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C})\otimes\mathcal{O}_{\mathbb{P}^2}(1)^{-3\delta m}
\cong
E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m},
\]
where the logarithmic canonical line bundle $$\overline{K}_{\mathbb{P}^2}=K_{\mathbb{P}^2}\otimes\mathcal{O}(\mathcal{C})\cong \mathcal{O}_{\mathbb{P}^2}(-3)\otimes \mathcal{O}_{\mathbb{P}^2}(6)\cong\mathcal{O}_{\mathbb{P}^2}(3),$$ with $\delta>0$ a rational number to be chosen and $m$ sufficiently divisible.
The standard approach begins with Serre duality:
\begin{align}
\label{use Serre duality}
H^2\big(\mathbb{P}^2,E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\big) = H^0\big(\mathbb{P}^2,K_{\mathbb{P}^2}\otimes E_{2,m}T_{\mathbb{P}^2}(-\log\mathcal{C})\otimes \overline{K}_{\mathbb{P}^2}^{\delta m}\big).
\end{align}
The bundle $K_{\mathbb{P}^2}\otimes E_{2,m}T_{\mathbb{P}^2}(-\log\mathcal{C})\otimes \overline{K}_{\mathbb{P}^2}^{\delta m}$ admits a filtration, which is induced by the dual of~\eqref{filtration of E_2,m} with a twist, with graded pieces
\[
S^{m-3j}T_{\mathbb{P}^2}(-\log\mathcal{C})\otimes \overline{K}_{\mathbb{P}^2}^{-j}\otimes K_{\mathbb{P}^2}\otimes\overline{K}_{\mathbb{P}^2}^{\delta m}
= S^{m-3j}T_{\mathbb{P}^2}(-\log\mathcal{C})\otimes \overline{K}_{\mathbb{P}^2}^{\delta m-j-1}
\]
for $0\leqslant j\leqslant \frac{m}{3}$.
For any rational number $\delta\in (0,\frac{1}{3})$ and sufficiently divisible $m\gg 1$, the inequality
\[
m - 3j > 2(\delta m - j - 1)
\]
automatically holds for all $0 \leqslant j \leqslant \frac{m}{3}$. By applying the Bogomolov-type Vanishing Theorem~\ref{thm:Bogomolov vanishing theorem}, we deduce:
\[
H^0\Big(\mathbb{P}^2, S^{m-3j}T_{\mathbb{P}^2}(-\log\mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{\delta m - j - 1}\Big) = 0 \quad \,\,\,\text{for all }\quad 0 \leqslant j \leqslant \tfrac{m}{3}.
\]
This establishes the vanishing of the right-hand side in~\eqref{use Serre duality}, which consequently implies the vanishing of the left-hand side as well. We thus obtain the lower bound
\[
\dim H^0\Big(\mathbb{P}^2, E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\Big)
\geqslant \chi\Big(E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\Big),
\]
since the Euler characteristic decomposes as
\[
\chi(\cdots) = \dim H^0(\cdots) - \dim H^1(\cdots) + \dim H^2(\cdots) = \dim H^0(\cdots) - \dim H^1(\cdots).
\]
To estimate the asymptotic growth of the Euler characteristic, we employ the Schur bundle filtration (see~\eqref{filtration of E_2,m}):
\begin{align}\label{equ:filtration of 2-jet}
\mathrm{Gr}^{\bullet}
E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C})
&= \bigoplus_{a+3b=m}\Gamma^{(a+b,\,b)}T_{\mathbb{P}^2}^*(\log \mathcal{C})
\qquad \text{(cf.~\cite[p.~67]{Demailly1997}, \cite[p.~474]{Elgoul2003})} \notag \\
&= \bigoplus_{0\leqslant b\leqslant\frac{m}{3}}\Gamma^{(m-2b,\, b)}T_{\mathbb{P}^2}^*(\log \mathcal{C}).
\end{align}
Together with the Pieri formula~(cf.~\cite[p.~455]{FultonHarris1991}, \cite[p.~66]{Demailly1997}):
\[
\Gamma^{(\ell_1,\, \ell_2)}T_{\mathbb{P}^2}^*(\log \mathcal{C})\otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}
= \Gamma^{(\ell_1-\delta m,\, \ell_2-\delta m)}T_{\mathbb{P}^2}^*(\log \mathcal{C}),
\]
we derive the following computation (cf.~\cite[p.~77]{Merker2015}):
\begin{align*}
&\chi\big(E_{2,m}T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{-\delta m}\big) \\
=\,& \sum_{b=0}^{\frac{m}{3}} \chi\big(\Gamma^{(m-2b-\delta m,\, b-\delta m)}T_{\mathbb{P}^2}^*(\log \mathcal{C})\big) \\
=\, &\bigg(\frac{1}{6} \int_{0}^{\frac{m}{3}} \big[(m-2b-\delta m)^3-(b-\delta m)^3\big] \mathrm{d} b\bigg) \bar{c}_1^2 \,-\, \bigg(\frac{1}{6} \int_0^{\frac{m}{3}} (m-3b)^3 \mathrm{d} b\bigg) \bar{c}_2 + O(m^3) \\
=\,& \frac{m^2(m+3)}{108}(54\delta^2 m-48\delta m+13m-18\delta+15)\bar{c}_1^2
- \frac{m^4}{72}\bar{c}_2 + O(m^3) \\
=\,& \frac{m^4}{648}\big((54\delta^2-48\delta+13)\bar{c}_1^2 - 9\bar{c}_2\big) + O(m^3).
\end{align*}
Here the logarithmic Chern numbers are determined by the formula $\bar{c}_1^2=(d-3)^2=9$ and $\bar{c}_2=d^2 - 3d + 3 - \sum_{i\neq j}d_id_j=9$, where $d_1=d_2=d_3=2$ and $d=d_1+d_2+d_3=6$.
The quadratic leading coefficient (after removing the common factor $\frac{9}{648}$)
\[
\varphi(\delta) := (54\delta^2 - 48\delta + 13) - 9 = 54\delta^2 - 48\delta + 4
\]
has two positive roots given by
$
\delta_1 = \frac{4-\sqrt{10}}{9}$ and $ \delta_2 = \frac{4+\sqrt{10}}{9}$.
Hence, $\varphi(\delta) > 0$ for all $\delta \in (0,\delta_1)$. It follows that $\overline{\theta}_2 \geqslant 3\delta_1=0.279\ldots$, which completes the proof of Proposition~\ref{estimate of 2-jet threshold}.
\end{proof}
To apply Siu's strategy of slanted vector fields, we follow the argument of Siu~\cite[pp.~558--559]{Siu2004} to extend a negatively twisted logarithmic $2$-jet differential as a holomorphic family, invoking the following standard result in algebraic geometry:
\begin{thm}[{\cite[p.~288]{HartshorneGTM52}}]\label{thm:extension of sections}
Let $\tau \colon \mathcal{Z} \rightarrow S$ be a flat holomorphic family of compact complex spaces and let $\mathcal{L} \rightarrow \mathcal{Z}$ be a holomorphic vector bundle. Then there exists a proper subvariety $Z \subset S$ such that for each $s_0 \in S \backslash Z$, the restriction map $H^0 \big{(} \tau^{-1} \left( U_{s_0} \right) , \mathcal{L} \big{)} \rightarrow H^0 \big ( \tau^{-1} ( s_0) , \mathcal{L}|_{\tau^{-1} ( s_0 )} \big )$ is onto, for some Zariski-dense open set $U_{s_0} \subset S$ containing $s_0$.
\end{thm}
Now we apply Demailly--El Goul's argument (cf.~\cite[p.~532]{Demailly-Elgoul2000}, \cite[Lemma 1.3.2]{Elgoul2003}) to show the existence of ``irreducible'' negatively twisted invariant logarithmic $2$-jet differentials.
\begin{pro}\label{prop:step1}
For some positive integer $m, t \geqslant 1$ with $ {m}<{5}t$, there exists a Zariski-open subset $U^{m,t}$ in the parameter space $\mathbb{P}^5 \times \mathbb{P}^5 \times \mathbb{P}^5$ (see~\eqref{family of curves in P^2}) of three conics $\mathcal{C}$ in \(\mathbb{P}^2\), such that:
\begin{enumerate}
\item For each parameter $a \in U^{m,t}$, the corresponding \(\mathcal{C}_a\) forms a simple normal crossing divisor;
\item there is a holomorphic family
\begin{equation*}
\left\{ \omega_{1,a} \; \middle| \; 0 \not\equiv \omega_{1,a} \in H^0\left(\mathbb{P}^2, E_{2,m}T^*_{\mathbb{P}^2}(\log \mathcal{C}_a) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\right) , \; a \in U^{m,t} \right\},
\end{equation*}
which varies holomorphically with respect to the variable $a \in U^{m,t}$. Moreover, each zero locus $\{\omega_{1,a} = 0\}$ on the second level $X_{2, a}$ of the logarithmic Demailly-Semple tower for $(\mathbb{P}^2, \mathcal{C}_{a})$ is irreducible and reduced modulo $\Gamma_{2, a}$ (see \eqref{equ:def of Gamma_2}).
\end{enumerate}
\end{pro}
\begin{proof}%%[Proof of Proposition \ref{prop:step1}]
First, for any integers $M,T \geqslant 1$, Theorem~\ref{thm:extension of sections} guarantees that there exists a proper Zariski-closed subset $Z_{M,T}$ in the parameter space $\mathbb{P}^5 \times \mathbb{P}^5 \times \mathbb{P}^5$ of three conics, such that, for any parameter $a$ outside $Z_{M,T}$, every
nonzero global section of
$E_{2,M}T^*_{\mathbb{P}^2}(\log \mathcal{C}_a) \otimes \mathcal{O}_{\mathbb{P}^2}(-T)$ extends to a holomorphic family parameterized by some Zariski open neighborhood of $a$. Thus, to ensure the local extendability of any nonzero negatively twisted logarithmic $2$-jet differential on $(\mathbb{P}^2,\mathcal{C}_{a_0})$, we choose a point $a_0$ in $(\mathbb{P}^5 \times \mathbb{P}^5 \times \mathbb{P}^5) \setminus \cup_{M,T \geqslant 1} Z_{M,T}$ such that $\mathcal{C}_{a_0}$ is a simple normal crossing divisor.
The existence of such $a_0$ is guaranteed by the Bertini theorem.
By Proposition \ref{estimate of 2-jet threshold}, noting that $5 > \frac{3}{4-\sqrt{10}} = 3.58\ldots$, for some integer $M, T \geqslant 1$ with ratio $\frac{1}{5} < \frac{T}{M} \leqslant \overline{\theta}_2(\mathbb{P}^2, \mathcal{C}_{a_0})$, there exists a nonzero negatively twisted invariant logarithmic $2$-jet differential
\begin{equation*}
\omega_0 \in H^0\big(X, E_{2, M} T_{\mathbb{P}^2}^*(\log \mathcal{C}_{a_0}) \otimes \mathcal{O}_{\mathbb{P}^2}(-T)\big).
\end{equation*}
Next, we decompose the zero locus \(\{\omega_{0}=0\}\) in the second level $X_{2, a_0}$ of the Demailly-Semple tower of $(\mathbb{P}^2, \mathcal{C}_{a_0})$ into irreducible components. By the same argument as Demailly and El Goul~\cite[{p.~532}]{Demailly-Elgoul2000}, \cite[{Lemma 1.3.2}]{Elgoul2003}, we can extract an irreducible and reduced component $Z$, which in fact ``corresponds'' to some irreducible negatively twisted invariant logarithmic $2$-jet differential \begin{equation}\label{define omega_1}
\omega_1 \in H^0\big(X, E_{2, m} T_{\mathbb{P}^2}^*(\log \mathcal{C}_{a_0}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\big),\quad \text{with}\
\tfrac{t}{m} \geqslant \tfrac{T}{M} > \tfrac{1}{5},
\end{equation}
in the sense that $\{\omega_1=0\}=Z+b\Gamma_{2, a_0}$ (see \eqref{equ:def of Gamma_2} and \eqref{gamma_2=O(-1, 1)}) for some integer $b>0$. The argument is as follows.
Let $u_1 = \pi_{2,1}^* \mathcal{O}_{X_{1,a_0}}(1)$, $u_2 = \mathcal{O}_{X_{2,a_0}}(1)$ and $h = \mathcal{O}_{\mathbb{P}^2}(1)$. Then the zero divisor $Z_{\omega_{0}}$ of $\omega_{0}$ on $X_{2, a_0}$ is linearly equivalent to
\begin{equation*}
M u_2 - T \pi_{2,0}^* h \in \pic \left( X_{2, a_0} \right) .
\end{equation*}
Let $Z_{\omega_{0}} = \sum p_j Z_j$ be the irreducible decomposition of the divisor $Z_{\omega_{0}}$, with each $Z_j$ irreducible and reduced. Since $\operatorname{Pic}(X_{2, a_0}) \cong \operatorname{Pic}(\mathbb{P}^2) \oplus \mathbb{Z} u_1 \oplus \mathbb{Z} u_2$ and $\operatorname{Pic}(\mathbb{P}^2) \cong \mathbb{Z}$, each component satisfies
\begin{equation*}
Z_j \sim b_{1,j} u_1 + b_{2,j} u_2 - t_j \pi_{2,0}^* h , \qquad b_{1,j} \; , \; b_{2,j} \; , \; t_j \in \mathbb{Z}.
\end{equation*}
By \cite[Lemma 3.3]{Demailly-Elgoul2000}, the effectiveness of $Z_j$ implies that
exactly one of the following mutually exclusive cases holds:
\begin{itemize}
\item $\left( b_{1,j} , b_{2,j} \right) = (0,0)$ and $Z_j \in \pi_{2,0}^* \pic \left( \mathbb{P}^2 \right)$, $-t_j > 0$;
\item $\left( b_{1,j} , b_{2,j} \right) = (-1,1)$, then $Z_j$ contains $\Gamma_{2, , a_0}$, where $\Gamma_{2, , a_0}$ is an irreducible divisor on $X_{2, a_0}$. Thus $Z_j = \Gamma_{2, a_0}$ and $t_j = 0$;
\item $b_{1,j} \geqslant 2 b_{2,j} \geqslant 0$ and $m_j \coloneqq b_{1,j} + b_{2,j} > 0$.
\end{itemize}
In the third case, the irreducible and reduced divisor $Z_j$ corresponds a section
\begin{equation*}
\omega_{j} \in H^0\big(
X_{2, a_0}, \pi_{2,1}^{*} \mathcal{O}_{X_{1, a_0}}(b_{1,j}) \otimes \mathcal{O}_{X_{2, a_0}}(b_{2,j})\otimes\pi_{2,0}^*\mathcal{O}_{\mathbb{P}^2}(-t_j)\big).
\end{equation*}
Noting that $\pi_{2,1}^{*} \mathcal{O}_{X_{1, a_0}}(b_1) \otimes \mathcal{O}_{X_{2, a_0}}(b_2) \cong \mathcal{O}_{X_{2, a_0}}\left(b_1 + b_2\right) \otimes \mathcal{O}_{X_{2, a_0}}\left(-b_1 \Gamma_{2, a_0} \right)$ (see \eqref{gamma_2=O(-1, 1)}), we may regard
\begin{equation*}
\omega_{j} \in H^0\big(
X_{2, a_0}, \mathcal{O}_{X_{2, a_0}}(m_{j})\otimes\pi_{2,0}^*\mathcal{O}_{\mathbb{P}^2}(-t_j)\big),
\end{equation*}
whose zero divisor is $Z_j + b_{1,j} \Gamma_2$. Since both $Z_j$ and $\Gamma_2$ are irreducible and reduced, $\omega_{j}$ can not be factorized into a product of other logarithmic $2$-jet differentials.
Let $\max \{ {t_j}/{m_j} \}$ denote the maximum ratio among those sections obtained from the third case. Since $M = \sum p_j m_j$ and $T = \sum p_j t_j$, we have $\max \{ {t_j}/{m_j} \} \geqslant {T}/{M} > 1/5$. Without loss of generality, we may assume that $t_1 / m_1 = \max \{ t_j / m_j \}$, and denote $m \coloneqq m_1$ and $t \coloneqq t_1$. By applying the direct image formula \eqref{direct image formula} and abusing the notation for $\omega_1$, we thus obtain \eqref{define omega_1}.
By our choice of $a_0\notin \cup_{M,T \geqslant 1} Z_{M,T}$, automatically $\omega_1$ extends to a holomorphic family
\begin{equation*}
\left\{ \omega_{1,a} \; \middle| \; 0 \not\equiv \omega_{1,a} \in H^0\left(\mathbb{P}^2, E_{2,M}T^*_{\mathbb{P}^2}(\log \mathcal{C}_a) \otimes \mathcal{O}_{\mathbb{P}^2}(-T)\right) , \; a \in U_{a_0}^{m,t} \right\},
\end{equation*}
which varies holomorphically with respect to the parameter $a$ in some Zariski-open neighborhood $U_{a_0}^{m,t}$ of $a_0$, with $\omega_{1,a_0} = \omega_1$. By shrinking to a smaller Zariski-open subset $U^{m,t} \subset U_{a_0}^{m,t}$ if necessary, we may ensure that each $a \in U^{m,t}$ defines three conics $\mathcal{C}_a$ with simple normal crossings while the zero locus $\{\omega_{1,a} = 0\}$ on the second level $X_{2, a}$ of the logarithmic Demailly-Semple tower for $(\mathbb{P}^2, \mathcal{C}_{a})$ is irreducible and reduced modulo $\Gamma_{2, a}$.
\end{proof}
\section{\bf Generating Independent Twisted Jet Differentials via Slanted Vector Fields}\label{section: Siu's fields}
To construct an additional family of independent negatively twisted jet differentials, we employ Siu's slanted vector field method~\cite{Siu2004}. This approach generalizes to $k$-jet spaces the original technique developed for $1$-jet spaces by Clemens~\cite{Clemens1986}, Ein \cite{Ein1988}, and Voisin~\cite{Voisin1996} in their study of rational curves.
The method was first implemented by P\u{a}un~\cite{mihaipaun2008} for compact surfaces in $\mathbb{P}^{3}$. Rousseau~\cite{Rousseau2009} subsequently established the logarithmic theory for pairs $(\mathbb{P}^{2},D)$ with curve configurations $D$, while in~\cite{Rousseau2007cpt,Rousseau2007log} he further extended both P\u{a}un's setting and the logarithmic case to threefolds. For hypersurfaces $H\subset\mathbb{P}^n$ of arbitrary dimension, Merker~\cite{Merker2009} established the complete generalization, thereby realizing Siu's original vision concerning generic global generation. The logarithmic counterpart for pairs $(\mathbb{P}^n,H)$ was subsequently obtained by Darondeau~\cite{Darondeau_vectorfield}.
This section adapts the standard technique of Siu's slanted vector fields—twisted by $\mathcal{O}_{\mathbb{P}^2}(7)$—to the case of three generic conics in $\mathbb{P}^2$. The content is entirely classical and no originality is claimed.
\subsection{Logarithmic Slanted Vector Fields for Three Conics in $\mathbb{P}^2$}\label{subsec:groundwork of slanted vector fields}
The framework presented in this subsection specializes Rousseau's logarithmic setting in $\mathbb{P}^2$ to configurations of three conics. We emphasize that this subsection contains no original contributions and is included solely for completeness and the reader's convenience.
Let
\begin{equation}\label{family of curves in P^2}
\mathcal{X} = \cup_{\lambda=1}^3 \mathcal{X}_{\lambda} \subset \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_{\lambda}}}
\end{equation}
denote the universal family of algebraic curves in \(\mathbb{P}^2\) consisting of three components of degrees \(d_1\), \(d_2\), and \(d_3\), respectively. Here, each \(\mathcal{X}_{\lambda}\) is the universal curve of degree \(d_{\lambda}\), defined by the equation $\sum_{|\alpha| = d_{\lambda}} A_{\alpha}^{\lambda} Z^{\alpha}=0$, where \([A^{\lambda}] = [A_{\alpha}^{\lambda}]_{|\alpha|=d_{\lambda}} \in \mathbb{P}^{N_{d_{\lambda}}}\) are the homogeneous coefficients and \([Z] = [Z_0 : Z_1 : Z_2] \in \mathbb{P}^2\) are the projective coordinates. For a multi-index \(\alpha = (\alpha_0, \alpha_1, \alpha_2) \in \mathbb{N}^3\) with length \(|\alpha| := \alpha_0 + \alpha_1 + \alpha_2\), we employ the notation
$
Z^{\alpha} := Z_0^{\alpha_0} Z_1^{\alpha_1} Z_2^{\alpha_2}$.
The dimension of the parameter space for each component is given by
$
N_{d_{\lambda}} = \binom{d_{\lambda} + 2}{2} - 1$.
For any integer $k \geqslant 1$, the \emph{vertical logarithmic $k$-jet space} is then defined as
\[
\overline{J_k^v}\left(\mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}\right)
:= \ker\left(\mathrm{pr}_2\right)_*
\subset J_k\left(\mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}, -\log \mathcal{X} \right),
\]
where $\mathrm{pr}_2 \colon \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}} \longrightarrow {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}$ denotes the natural projection onto the parameter space and $J_k \big(\mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}, -\log \mathcal{X} \big)$ is the logarithmic $k$-jet bundle along the universal family $\mathcal{X}$.
We define the universal intersection family
\[
\mathcal{Y} = \cap_{\lambda=1}^3 \mathcal{Y}_{\lambda} \subset \mathbb{P}^5 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_{\lambda}} + 1},
\]
where for each \(1 \leqslant \lambda \leqslant 3\), the hypersurface \(\mathcal{Y}_{\lambda}\) is the universal family of degree \(d_\lambda\) hypersurfaces in \(\mathbb{P}^5\), given explicitly by $\mathcal{Y}_{\lambda} := \left\{
A_{\mathbf{0}}^{\lambda} Z_{2+\lambda}^{d_{\lambda}} + \sum_{|\alpha| = d_{\lambda}} A_{\alpha}^{\lambda} Z^{\alpha} = 0
\right\} \subset \mathbb{P}^5 \times \mathbb{P}^{N_{d_{\lambda}} + 1}$. Here, the space \(\mathbb{P}^5\) has homogeneous coordinates $[Z : Z_3 : Z_4 : Z_5] = [Z_0 : \cdots : Z_5]$, and the coefficient space \(\mathbb{P}^{N_{d_{\lambda}} + 1}\) is equipped with homogeneous coordinates $[A^{\lambda} : A_{\mathbf{0}}^{\lambda}] = [\ldots : A_{\alpha}^{\lambda} : \ldots : A_{\mathbf{0}}^{\lambda}]_{|\alpha| = d_{\lambda}}$.
For notational convenience, we denote the exceptional locus $\{Z = \mathbf{0}\} \cup \cup_{\lambda=1}^{3} \{A^{\lambda} = \mathbf{0}\}$, where $\{Z = \mathbf{0}\}$ denotes the subset of $\mathbb{P}^5$ defined by $Z_0 = Z_1 = Z_2 = 0$ and $\{A^{\lambda} = \mathbf{0}\}$ refers to the locus in $\mathbb{P}^{N_{d_{\lambda}} + 1}$ where all coordinates vanish except $A_{\mathbf{0}}^{\lambda}$. There exists a natural forgetful map
\[
\pi \colon \mathbb{P}^{5} \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_{\lambda}} + 1} \setminus \left(\{Z = \mathbf{0}\} \cup \cup_{\lambda=1}^{3} \{A^{\lambda} = \mathbf{0}\}\right) \longrightarrow \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_{\lambda}}}
\]
that forgets the auxiliary coordinates. Explicitly, $\pi$ is given by
\[
\left([Z:Z_3:Z_4:Z_5], [A^1:A_{\mathbf{0}}^1], [A^2:A_{\mathbf{0}}^2], [A^3:A_{\mathbf{0}}^3]\right) \longmapsto \left([Z], [A^1], [A^2], [A^3]\right).
\]
We observe the containment relation
\begin{equation}
\label{containment relation}
\mathcal{Y} \cap \left( \{ Z = \mathbf{0} \} \cup \cup_{\lambda=1}^{3} \{ A^{\lambda} = \mathbf{0} \} \right)
\subset \mathcal{Y} \cap \left( \cup_{\lambda=1}^{3} \left( \{ A_{\mathbf{0}}^{\lambda} = 0 \} \cup \{ A^{\lambda} = \mathbf{0} \} \right) \right),
\end{equation}
from the inclusion $\mathcal{Y} \cap \{ Z = \mathbf{0} \} \subset \mathcal{Y} \cap \cup_{\lambda=1}^{3} \{ A_{\mathbf{0}}^{\lambda} = 0 \}$, which is deduced by the definition of $\mathcal{Y}$.
Now we consider $\mathcal{Y}^* \coloneqq \mathcal{Y} \setminus \left( \cup_{\lambda=1}^{3} \left( \{ A_{\mathbf{0}}^{\lambda} = 0 \} \cup \{ A^{\lambda} = \mathbf{0} \} \right) \right)$ on which the restricted forgetful map $\pi|_{\mathcal{Y}^*} \colon \mathcal{Y}^* \rightarrow \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}$ is well-defined according to the
containment relation~\eqref{containment relation}. Furthermore, we obtain the identification of the preimage $\left( \pi|_{\mathcal{Y}^*} \right)^{-1}(\mathcal{X}) = \{ Z_3 Z_4 Z_5 = 0 \} =: \mathcal{D}$.
This shows that \(\pi\) induces a logarithmic morphism from $\left( \mathcal{Y}^*, \mathcal{D} \right)$ to $\big( \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}, \mathcal{X} \big)$, which consequently gives rise to a dominant map between jet spaces:
\begin{equation}\label{pi_[2]}
\pi_{[2]} \colon \overline{J_2^v}(\mathcal{Y}^*) \longrightarrow \overline{J_2^v}\left( \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}} \right).
\end{equation}
Here, $\overline{J_2^v}(\mathcal{Y}^*)$ represents the space of vertical logarithmic 2-jets with respect to the normal crossing divisor $\mathcal{D} = \{ Z_3 Z_4 Z_5 = 0 \}$. This construction allows us to perform all local computations on the more tractable space $\mathcal{Y}^*$ through pullback via $\pi$, where we have access to explicit coordinate expressions.
On the affine chart $\{Z_0 \neq 0\} \times {\textstyle\prod}_{\lambda=1}^3 \{A_{\mathbf{0}}^{\lambda} \neq 0\}$, we employ the affine coordinates $z_i := Z_i/Z_0$ for $i = 1,\dots,5$ and $a_{\alpha}^{\lambda} := A_{\alpha_0 \alpha_1 \alpha_2}^{\lambda}/A_{\mathbf{0}}^{\lambda}$ with $\alpha_0 = d_{\lambda} - \alpha_1 - \alpha_2$. To streamline notation, for each jet of order $j = 1,2$ we define:
\begin{equation*}\label{notation xi of 2-jets}
\xi_i^{(j)} := \begin{cases}
z_i^{(j)} & \text{for } i = 1,2, \\
(\log z_{2+\lambda})^{(j)} & \text{for } i = 3,4,5 \ (\lambda = i-2).
\end{cases}
\end{equation*}
Then local coordinates for $\overline{J_2}\big( \{ Z_0 \neq 0 \} \times {\textstyle\prod}_{\lambda=1}^3 \{ A_{\mathbf{0}}^{\lambda} \neq 0 \} \big) \subset J_2\big( \mathbb{P}^5 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda} + 1} , -\log \mathcal{D} \big)$ are given by the coordinate tuple $\left( z_i , a_{\alpha}^{\lambda} , \xi_{i_1}' , \xi_{i_2}'' \right)$.
\medskip
Following the strategy of Siu \cite{Siu2004}
and the realizations of Păun~\cite{mihaipaun2008} and Rousseau~\cite{Rousseau2009}, we consider the following family of Siu's slanted vector fields with pole order $7$:
\begin{thm}\label{thm:globally generated vector field of pole order 7}
Let $\Sigma_0$ be the zero locus of the $2 \times 2$ Wronskian $\det ( \xi_i^{(j)} )_{1 \leqslant i,j \leqslant 2}$:
\begin{equation*}
\Sigma_0 = \left\{ \left( z_i , a_{\alpha}^{\lambda} , \xi_{i_1}' , \xi_{i_2}'' \right) \,\middle\vert\, \det \left( \xi_i^{(j)} \right)_{1 \leqslant i,j \leqslant 2} = 0 \right\} \subset \overline{J_2^v} \left( \mathcal{Y} \cap \left( \{ Z_0 \neq 0 \} \times {\textstyle\prod}_{\lambda=1}^3 \{ A_{\mathbf{0}}^\lambda \neq 0 \} \right) \right),
\end{equation*}
and $\Sigma$ be the closure of $\Sigma_0$ in $\overline{J_2^v}(\mathcal{Y}^*)$. Then the twisted tangent bundle
\begin{equation}\label{the twisted tangent bundle}
T_{ \overline{J_2^v}\left( \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}} \right) } \otimes p^* \mathrm{pr}_1^* \mathcal{O}_{\mathbb{P}^2}(7) \otimes p^* \mathrm{pr}_2^* \mathcal{O}_{\prod_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}}(1,1,1),
\end{equation}
is generated by its global sections at every point of $\overline{J_2^v} ( \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}} ) \setminus \left( \pi_{[2]} \left( \Sigma \right) \cup p^{-1} \left( \mathcal{X} \right) \right)$, where $\pi_{[2]}$ is introduced in \eqref{pi_[2]}, $p$ is the natural projection
\begin{equation*}
p
\colon
\overline{J_2^v} \left( \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}} \right)
\longrightarrow
\mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}},
\end{equation*}
and $\mathrm{pr}_1$ (resp. $\mathrm{pr}_2$) is the projection from $\mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}$ to $\mathbb{P}^2$ (resp. ${\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}}$).
\end{thm}
\medskip
\subsection{Producing an Independent $2$-Jet via Siu's Slanted Vector Fields of Pole Order~$7$}
\begin{pro}
Given a holomorphic family $\{ \omega_{1, a} \}_{a\in U}$ of negatively twisted invariant logarithmic $2$-jet differentials as in Proposition \ref{prop:step1} with vanishing order $t \geqslant 8$,
one can find some slanted vector field $V$ of pole order $7$ as a global section of~\eqref{the twisted tangent bundle}, such that,
the derivative of $\{ \omega_{1, a} \}$ along $V$ gives a new family of holomorphic invariant logarithmic $2$-jet differentials $\omega_{2, a}$ with vanishing order $t-7\geqslant 1$. Furthermore, the common zero set $\{\omega_{1,a} = 0\} \cap \{\omega_{2,a} = 0\}$ in the second level $\left( \mathbb{P}^2_{a} \right)_2$ of the logarithmic Demailly–Semple tower of the log pair $\left( \mathbb{P}^2 , \mathcal{C}_a \right)$ has codimension at least $2$ modulo $\Gamma_2$ for a generic parameter $a \in U$.
\end{pro}
This kind of proposition is due to Siu~\cite{Siu2004}. For the sake of completeness, we provide an argument here.
\begin{proof}
%By proposition \ref{prop:step1}, there exists a Zariski open set $U$ in the parameter space$\prod_{{\lambda}=1}^3 \mathbb{P}^{N_{d_{\lambda}}}$of three conics, such that for any $a\in U$, The zero locus $\{\omega_{1,a} = 0\}$ in the second level $X_{a,2}$ of the logarithmic Demailly-Semple tower for $(\mathbb{P}^2, \mathcal{C}_a)$ is irreducible and reduced modulo $\Gamma_2$, where we regard
%\begin{equation*} \omega_{1,a} \in H^0 \big( (\mathbb{P}_a^2)_2,\mathcal{O}_{(\mathbb{P}^2_a)_2}(m)\otimes\pi_{2,0}^*\mathcal{O}_{\mathbb{P}^2}(-t) \big) \quad \text{(by~\eqref{direct image formula})}.\end{equation*}
We regard $\{ \omega_{1, a}\}_{a\in U} $ as a holomorphic map
\begin{equation*}
P \colon \overline{J_2^v} \left( \mathbb{P}^2 \times {\textstyle\prod}_{{\lambda}=1}^3 \mathbb{P}^{N_{d_{\lambda}}} \right) \Big|_{\mathrm{pr}_2^{-1} \left( U \right)} \longrightarrow p^* \mathrm{pr}_1^* \mathcal{O}_{\mathbb{P}^2}(-t) |_{\mathrm{pr}_2^{-1} \left( U \right)} ,
\end{equation*}
For any global section $V$ of \eqref{the twisted tangent bundle}, which is a meromorphic global vector field on $\overline{J_2^v}\left( \mathbb{P}^2 \times {\textstyle\prod}_{\lambda=1}^3 \mathbb{P}^{N_{d_\lambda}} \right)$ with values in $p^* \mathrm{pr}_1^* \mathcal{O}_{\mathbb{P}^2}(7)$, as shown in \cite{mihaipaun2008, Rousseau2009}, differentiating $P$ by $V$ yields a new holomorphic map:
\begin{equation*}
\mathrm{d} \, P(V) \colon \overline{J_2^v} \left( \mathbb{P}^2 \times {\textstyle\prod}_{{\lambda}=1}^3 \mathbb{P}^{N_{d_{\lambda}}} \right) \Big|_{\mathrm{pr}_2^{-1} \left( U \right)} \longrightarrow p^* \mathrm{pr}_1^* \mathcal{O}_{\mathbb{P}^2}(7-t) |_{\mathrm{pr}_2^{-1} \left( U \right)} ,
\end{equation*}
which corresponds to a new holomorphic family of invariant logarithmic $2$-jet differentials with vanishing order $t-7\geqslant 1$.
By Theorem \ref{thm:globally generated vector field of pole order 7}, the twisted tangent bundle \eqref{the twisted tangent bundle} is globally generated outside $\pi_{[2]} \left( \Sigma \right) \cup p^{-1} \left( \mathcal{X} \right)$. Noting that $\{P|_{a} = \omega_{1,a} = 0 \}$ is irreducible and reduced modulo $\Gamma_2$, whence we can find some global section $V$ of \eqref{the twisted tangent bundle} such that the intersection $\{P|_{a} = 0 \} \cap \{\mathrm{d} \, P( V ) |_{a} = 0 \}$ has codimension $2$ modulo $\Gamma_2$ for a generic parameter $a \in U$.
\end{proof}
%\begin{rmk}\label{how to use O(3)}\rm
%We hypothesize that an alternative strategy, which deepens the technical machinery of Paŭn \cite[pp.~886-889]{mihaipaun2008} and Rousseau \cite[Proposition 17]{Rousseau2009} via a refined application of Siu's slanted vector fields with twists by $\mathcal{O}_{\mathbb{P}^2}(3)$, $\mathcal{O}_{\mathbb{P}^2}(4)$, $\mathcal{O}_{\mathbb{P}^2}(5)$, or $\mathcal{O}_{\mathbb{P}^2}(6)$, could demonstrate the nonexistence of nontrivial solutions for a certain associated ODE. Achieving this would substantially lower the number of required vanishing results. Nevertheless, in the setting of three conics, the substantial algebraic complexity encountered in the ODE analysis precluded further progress.
%\end{rmk}
\section{\bf Refinement of Demailly--El Goul's Arguments}\label{sect: refined Demailly-EG}
\subsection{More Riemann-Roch Computations}
\label{subsection:4.1}
In the preceding Section~\ref{Section 3}, we show the existence of a nonzero negatively twisted invariant logarithmic $2$-jet differential
\[
\omega_1 \in H^0\big(X_2, \mathcal{O}_{X_2}(m) \otimes \pi_{2,0}^* \mathcal{O}_X(-t)\big),
\]
for some positive integers $m, t \geqslant 1 $ with $t/m\geqslant 1/5$, whose zero divisor on the second level of the Demailly-Semple tower $X_2$ is of the form $Z + b_1 \Gamma_2$, where $Z\subset X_2$ is irreducible and reduced. Moreover, we have the linear equivalence
\begin{equation}\label{equivalent class of Z}
Z \sim b_1 u_1 + b_2 u_2 - t \pi_{2,0}^* h, \quad (b_1, b_2) \in \mathbb{Z}^2, \quad b_1 \geqslant 2b_2 > 0, \quad b_1 + b_2 = m, \quad t \in \mathbb{Z}_{+},
\end{equation}
where $u_1 = \pi_{2,1}^* \mathcal{O}_{X_1}(1)$, $u_2 = \mathcal{O}_{X_2}(1)$, and $h = \mathcal{O}_{\mathbb{P}^2}(1)$.
Hereafter, we use the standard notation:
\begin{equation*}
\mathcal{O}_{X_{2}}(b_1 , b_2) \coloneqq \pi_{2,1}^{*} \mathcal{O}_{X_{1}}(b_1) \otimes \mathcal{O}_{X_{2}}(b_2).
\end{equation*}
By the same argument as Demailly--El Goul~\cite[Proposition 3.4]{Demailly-Elgoul2000}, one can show that $\mathcal{O}_{X_2}(2,1)|_{Z}$ is big when $t < m$. Now, our goal is to determine an effective small rational number $\tau > 0$ such that the $\mathbb{Q}$-line bundle
$
\big( \mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau) \big)|_{Z}
$
remains big. This will be achieved through Riemann-Roch computations and the following proposition observed by Demailly~\cite[p.~73]{Demailly1997}.
\begin{pro}\label{pro:the higher direct images vanishing}
For all integers $a_2 > -2$ and $a_1 > 1$, the following higher direct images vanish:
\begin{equation*}
R^{q} (\pi_{2,0})_{*} \mathcal{O}_{X_2}(a_1,a_2) = 0, \quad \forall q\geqslant 1.
\end{equation*}
\end{pro}
Since no readily available reference for this result exists in the literature, we provide a proof here.
\begin{proof}
We first show that each fiber of $\pi_{2,0}: X_2 \to X_0$ is isomorphic to the Hirzebruch surface (cf.~\cite[p.~40]{Beauville1996})
\[
\mathbb{F}_3 = \mathbb{P}\bigl(\mathcal{O}_{\mathbb{P}^1} \oplus \mathcal{O}_{\mathbb{P}^1}(3)\bigr).
\]
Let $x \in X_0$ be an arbitrary point and denote by $F_1 \cong \mathbb{P}^1$ the fiber of $\pi_{1,0}: X_1 \to X_0$ over $x$. From the construction of the logarithmic Demailly--Semple tower (Subsection~\ref{subsec:log DS tower}) we know that the fiber $F_2$ of $\pi_{2,0}$ over $x$ satisfies
\[
F_2 \cong \mathbb{P}\bigl(V_1|_{F_1}\bigr),
\]
and the exact sequence \eqref{SES:V_1} gives $V_1|_{F_1} \cong T_{X_1/X_0}|_{F_1} \oplus \mathcal{O}_{\mathbb{P}^1}(-1)$.
Using the relative Euler sequence \eqref{ses:relative Euler sequence} and taking determinants, we obtain
\[
T_{X_1/X_0} \cong \pi_{1,0}^* \,\overline{K}_{X}^* \otimes \mathcal{O}_{X_1}(2).
\]
Restricting to $F_1$ yields $T_{X_1/X_0}|_{F_1} \cong \mathcal{O}_{\mathbb{P}^1}(2)$. Hence
\[
V_1|_{F_1} \cong \mathcal{O}_{\mathbb{P}^1}(2) \oplus \mathcal{O}_{\mathbb{P}^1}(-1),
\]
and we conclude that $F_2 \cong \mathbb{F}_3$.
On such a fiber we identify the restrictions of the relevant line bundles:
\begin{itemize}
\item $\mathcal{O}_{X_2}(1)|_{F_2}$ is the tautological bundle, which we denote by $\mathcal{O}_{\mathbb{F}_3}(\hbar)$.
\item $\pi_{2,1}^* \mathcal{O}_{X_1}(1)|_{F_2} \cong \mathcal{O}_{\mathbb{F}_3}(f)$, where $f$ is a fiber of $F_2\rightarrow F_1$.
\end{itemize}
Consequently,
\[
\mathcal{O}_{X_2}(a_1,a_2)|_{F_2} \cong \mathcal{O}_{\mathbb{F}_3}(a_2 \hbar + a_1 f).
\]
Recall that the canonical bundle of $\mathbb{F}_3$ is
\[
K_{\mathbb{F}_3} \cong \mathcal{O}_{\mathbb{F}_3}(-2\hbar + f).
\]
Therefore,
\[
\mathcal{O}_{\mathbb{F}_3}(a_2 \hbar + a_1 f)
\cong \mathcal{O}_{\mathbb{F}_3}\bigl((a_2+2)\hbar + (a_1-1)f\bigr) \otimes K_{\mathbb{F}_3}.
\]
Note that the line bundle $\mathcal{O}_{\mathbb{F}_3}\bigl((a_2+2)\hbar + (a_1-1)f\bigr)$ is ample precisely when
\[
a_2+2 > 0 \quad\text{and}\quad a_1-1 > 0.
\]
Under these conditions, the Kodaira Vanishing Theorem gives
\[
H^q\bigl(\mathbb{F}_3,\, \mathcal{O}_{\mathbb{F}_3}(a_2 \hbar + a_1 f)\bigr) = 0,
\qquad \forall\,q \geqslant 1.
\]
Hence
\[
H^q\bigl(F_2,\, \mathcal{O}_{X_2}(a_1,a_2)|_{F_2}\bigr) = 0, \qquad\forall\,x \in X_0,\ F_2=\pi_{2, 0}^{-1}(x),\ q \geqslant 1.
\]
Finally, Grauert's Direct Image Theorem implies that the vanishing holds for the direct images on $X_0$:
\[
R^{q} (\pi_{2,0})_* \mathcal{O}_{X_2}(a_1,a_2) = 0,
\qquad \forall\,q \geqslant 1.
\]
\end{proof}
\begin{pro}\label{prop: bigness on Z}
If $0 < \frac{t}{m} < 1$ and $0 \leqslant \tau < \tau_1(\frac{t}{m})$, where
{\footnotesize
\[
\tau_1 \left( \frac{t}{m} \right) = \frac{3}{2} \left(4 - \frac{t}{m}\right) - \frac{\sqrt{3}}{2} \sqrt{32 - 8 \frac{t}{m} + 3 \left( \frac{t}{m} \right)^2},
\]}
then the restricted $\mathbb{Q}$-line bundle
$
\big( \mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau) \big)|_{Z}
$
is big on $Z$.
\end{pro}
\begin{proof}
We begin by recalling some key intersection formulas on the Demailly–Semple tower. By~\cite[p.~478]{Elgoul2003}, we have the identities:
\[
u_1^2 + \bar{c}_1 u_1 + \bar{c}_2 = 0 \quad \text{and} \quad u_2^2 + \bar{c}_1(V_1) u_2 + \bar{c}_2(V_1) = 0,
\]
where $\bar{c}_1(V_1) = \bar{c}_1 + u_1$ (see~\eqref{define V_1}) and $\bar{c}_2(V_1) = \bar{c}_2 - u_2^2 = 2\bar{c}_2 + \bar{c}_1 u_1$.
On the second level $X_2$ of the Demailly-Semple tower, the following identities hold:
\[
u_{1}^{4} = 0, \quad u_{1}^{3} u_{2} = \bar{c}_{1}^{2} - \bar{c}_{2}, \quad u_{1}^{2} u_{2}^{2} = \bar{c}_{2}, \quad u_{1} u_{2}^{3} = \bar{c}_{1}^{2} - 3\bar{c}_{2}, \quad u_{2}^{4} = 5\bar{c}_{2} - \bar{c}_{1}^{2}.
\]
Furthermore, for any pullback $F \in \pi_{2,0}^* \pic (\mathbb{P}^2)$, we have:
\[
u_{1}^{3} \cdot F = 0, \quad u_{1}^{2} u_{2} \cdot F = -\bar{c}_{1} \cdot F, \quad u_{1} u_{2}^{2} \cdot F = 0, \quad u_{2}^{3} \cdot F = 0.
\]
A crucial part of our approach relies on the following additional identities:
\[
u_{1}^2 \cdot F_1 \cdot F_2 = 0, \quad u_{1} u_{2} \cdot F_1 \cdot F_2 = F_1 \cdot F_2, \quad u_{2}^2 \cdot F_1 \cdot F_2 = - F_1 \cdot F_2,
\qquad
\forall\, F_1, F_2 \in \pi_{2,0}^* \pic (\mathbb{P}^2).
\]
Then by direct computation, we can obtain
\begin{equation}\label{equ:self intersection on Z}
\left( 2 u_1 + u_2 - \tau \pi_{2,0}^* h \right)^3 \cdot Z = m\left( 13\overline{c}_{1}^{2} -9\overline{c}_{2}\right) + 12t(\overline{c}_{1} \cdot h) +12m(\overline{c}_{1} \cdot h) \tau +3m\left( h^{2}\right) \tau^{2} + 9\left( h^{2}\right) t \tau .
\end{equation}
In our setting of three generic conics, we have $\overline{c}_{1} = -3h$ and $\overline{c}_{2} = 9 h^2$. Substituting these into~\eqref{equ:self intersection on Z}, together with $h^2 = 1$, yields:
\begin{equation}\label{equ:self intersection on Z for 3 conics}
\left( 2 u_1 + u_2 - \tau \pi_{2,0}^* h \right)^3 \cdot Z = 3 \left( m\tau ^{2} - 3(4m-t) \tau +12(m-t) \right).
\end{equation}
This quadratic polynomial has two real roots:
{\footnotesize
\begin{equation}
\label{what is tau}
\tau_1 \left( \frac{t}{m} \right) = \frac{3}{2} \left(4 - \frac{t}{m}\right) - \frac{\sqrt{3}}{2} \sqrt{32 - 8 \frac{t}{m} + 3 \left( \frac{t}{m} \right)^2}
\,
<\, \tau_2 \left( \frac{t}{m} \right) = \frac{3}{2} \left(4 - \frac{t}{m}\right) + \frac{\sqrt{3}}{2} \sqrt{32 - 8 \frac{t}{m} + 3 \left( \frac{t}{m} \right)^2},
\end{equation}
}
because $32 - 8 \frac{t}{m} + 3 ( \frac{t}{m} )^2 > 0$. Moreover, $\tau_1 ( \frac{t}{m} ) > 0$ if and only if $\frac{t}{m} < 1$.
%For $0 < \frac{t}{m} < 1$, both functions $\tau_1(\frac{t}{m})$, $\tau_2(\frac{t}{m})$ are monotonically decreasing with respect to $\frac{t}{m}$, with $0 < \tau_1(\frac{t}{m}) < 6-2 \sqrt{6}$ and $9 < \tau_2(\frac{t}{m}) < 6+2 \sqrt{6}$. Furthermore, if $0 < \tau < \tau_1(\frac{t}{m})$ or $\tau > \tau_2(\frac{t}{m})$, then
%\begin{equation*}
%\left( 2 u_1 + u_2 - \tau \pi_{2,0}^* h \right)^3 \cdot Z = 3 m \left( \tau - \tau_1\left(\frac{t}{m}\right) \right) \left( \tau - \tau_2\left(\frac{t}{m}\right) \right) > 0.
%\end{equation*}
Thus when $0<\frac{t}{m}<1$ and $0 \leqslant \tau < \tau_1(\frac{t}{m})$, we have $( 2 u_1 + u_2 - \tau \pi_{2,0}^* h )^3 \cdot Z > 0$. Applying the Riemann-Roch Theorem then gives, for sufficiently divisible integer $p\gg 1$, that:
\begin{align*}
&h^0 \big( Z, (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} |_{Z} \big) + h^2 \big( Z, (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} |_{Z} \big) \\
\geqslant&
\chi \big( Z, (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} |_{Z} \big)
=
\frac{\big( 2 u_1 + u_2 - \tau \pi_{2,0}^* h \big)^3 \cdot Z}{3!} p^3 + O(p^2)\gg 1.
\end{align*}
Now we show that
\[
h^2 \big( Z, (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} |_{Z} \big) = 0
\]
for all sufficiently divisible $p \gg 1$.
Consider the short exact sequence
\begin{equation*}
\begin{tikzcd}
0 \arrow[r] &
\mathcal{O}_{X_2}(-Z) \arrow[r] &
\mathcal{O}_{X_2} \arrow[r] &
\mathcal{O}_{Z} \arrow[r] &
0,
\end{tikzcd}
\end{equation*}
and tensor it with $(\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p}$. The associated long exact cohomology sequence implies the dimension estimate:
\begin{align*}
h^2 \big( Z, (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} |_{Z} \big)
&\leqslant
h^2 \big( X_2, (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \big) \\
&\quad +
h^3 \big( X_2, \mathcal{O}_{X_2}(-Z) \otimes (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \big).
\end{align*}
First, we prove the vanishing
\begin{equation}\label{H^2 of the direct image of O(2p,p) tensoring O(-p tau)}
h^2 \big( X_2, (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \big) = 0.
\end{equation}
Consider the direct image sheaf $(\pi_{2, 0})_*(\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p}$ on $\mathbb{P}^2$. By Proposition~\ref{pro:the higher direct images vanishing}, the higher direct images vanish: $R^q (\pi_{2,0})_* (\mathcal{O}_{X_2}(2,1))^{p} = 0$ for $q \geqslant 1$. By the projection formula, we obtain:
\[
R^q (\pi_{2,0})_* (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \cong R^q (\pi_{2,0})_* (\mathcal{O}_{X_2}(2,1))^{p} \otimes \mathcal{O}_{\mathbb{P}^2}(-p \tau) = 0.
\]
Hence, Leray's spectral sequence implies
\begin{equation*}
H^2 \big( X_2, (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \big)
\cong
H^2 \big( \mathbb{P}^2, (\pi_{2,0})_* (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \big).
\end{equation*}
Now, by the isomorphism (cf.~\cite[Lemma 3.3 (c)]{Demailly-Elgoul2000})
\[
(\pi_{2,0})_* (\mathcal{O}_{X_2}(2,1))^{p} \cong E_{2,3p} T_{\mathbb{P}^2}^*(\log \mathcal{C}),
\]
we have
\begin{equation*}
H^2 \big( \mathbb{P}^2, (\pi_{2,0})_* (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \big)
\cong
H^2 \big( \mathbb{P}^2, E_{2,3p} T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-p \tau) \big).
\end{equation*}
The same argument as in the proof of Proposition~\ref{estimate of 2-jet threshold} shows that
\[
H^2 \big( \mathbb{P}^2, E_{2,3p} T_{\mathbb{P}^2}^*(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-p \tau) \big) = 0.
\]
Indeed, in the setting of Proposition~\ref{estimate of 2-jet threshold}, taking $m = 3p$ and $\delta = \tau/9$, the condition $0 \leqslant \delta < 1/3$ is satisfied because $0 < \frac{t}{m} < 1$ and $0 \leqslant \tau < \tau_1(\frac{t}{m})$ while the later is bounded from above by $6 - 2\sqrt{6} \approx 1.101$ by elementary calculation.
This establishes \eqref{H^2 of the direct image of O(2p,p) tensoring O(-p tau)}.
Next, we prove the vanishing
\begin{equation}\label{H^3 of the left term}
H^3 \big( X_2, \mathcal{O}_{X_2}(-Z) \otimes (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \big) = 0.
\end{equation}
By \eqref{equivalent class of Z}, we have the isomorphism
\[
\mathcal{O}_{X_2}(-Z) \otimes \left(\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau)\right)^{p} \cong \mathcal{O}_{X_2}(2p-b_1,\,p-b_2) \otimes \pi_{2,0}^{*} \mathcal{O}_{\mathbb{P}^2}(-(\tau p + t)),
\]
with the inequality $2p - b_1 \leqslant 2(p - b_2)$.
As before, by Proposition~\ref{pro:the higher direct images vanishing}, for all sufficiently divisible $p \gg 1$, the higher direct images vanish:
\[
R^q (\pi_{2,0})_* \left( \mathcal{O}_{X_2}(-Z) \otimes (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \right) = 0 \quad (q \geqslant 1).
\]
Thus, by Leray's spectral sequence, we conclude that
\begin{align*}
H^3 &\big( X_2, \mathcal{O}_{X_2}(-Z) \otimes (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} \big) \\
&\cong H^3 \big( \mathbb{P}^2, (\pi_{2,0})_* ( \mathcal{O}_{X_2}(-Z) \otimes (\mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau))^{p} ) \big) = 0,
\end{align*}
for all sufficiently divisible $p \gg 1$. This establishes \eqref{H^3 of the left term}, which
completes the proof.
\end{proof}
\subsection{Geometric Interpretation of Wronskians}
Let $\mathcal{C} = \mathcal{C}_1 + \mathcal{C}_2 + \mathcal{C}_3$ be a normal crossing divisor of three smooth conics in $\mathbb{P}^2$, defined by homogeneous quadratic polynomials $A(\mathbf{Z})$, $B(\mathbf{Z})$, and $C(\mathbf{Z})$, where $\mathbf{Z} = (Z_0,Z_1,Z_2)$.
On the affine chart $U_0 \coloneqq \{Z_0 \neq 0\}$ with coordinates $(x_1,x_2) = (Z_1/Z_0,Z_2/Z_0)$, let $a(x_1,x_2)$, $b(x_1,x_2)$, and $c(x_1,x_2)$ be the dehomogenized polynomials of $A, B, C$. Denote $a_i = \partial a / \partial x_i$ and $b_j = \partial b / \partial x_j$ for $i,j=1,2$.
Define the Zariski-open sets:
\[
U_{0,C} \coloneqq U_0 \cap \{ c (a_1 b_2 - b_1 a_2) \neq 0 \}, \,\,
U_{0,B} \coloneqq U_0 \cap \{ b (a_1 c_2 - c_1 a_2) \neq 0 \}, \,\,
U_{0,A} \coloneqq U_0 \cap \{ a (b_1 c_2 - c_1 b_2) \neq 0 \}.
\]
For three generic smooth conics, we have
\[
\{ c (a_1 b_2 - b_1 a_2) = 0 \} \cap \{ b (a_1 c_2 - c_1 a_2) = 0 \} \cap \{ a (b_1 c_2 - c_1 b_2) = 0 \} = \varnothing,
\]
so that $U_0 = \cup_{Q \in \{A,B,C\}} U_{0,Q}$. Similarly, we define Zariski-open sets $U_{i,Q}$ for $i=1,2$ and $Q \in \{A,B,C\}$. These sets $U_{i,Q}$ form an open covering of $\mathbb{P}^2$.
On $U_{0,C}$, noting that $c$ is nowhere vanishing, the logarithmic $1$-forms $(\log a)'$ and $(\log b)'$ generate the logarithmic cotangent bundle $T_{\mathbb{P}^2}^{*}(\log \mathcal{C})$. Let $f \colon (\mathbb{C},0) \ni t \mapsto f(t) \in U_{0,C} \subset \mathbb{P}^2$ be a germ of a holomorphic curve, with local representation $f = (f_1,f_2)$ in the coordinates $(x_1, x_2)$. Denote $a(f(t)) \coloneqq (f^{*} a) (t) = a(f_1(t),f_2(t))$ and $b(f(t)) \coloneqq (f^{*} b) (t) = b(f_1(t),f_2(t))$.
Consider the Wronskian operator
\begin{equation}\label{Wronskian of (log a)' (log b)' (preliminary)}
\begin{vmatrix}
(\log a(f))' & (\log b(f))' \\
(\log a(f))'' & (\log b(f))''
\end{vmatrix}.
\end{equation}
Under a biholomorphic reparametrization $\varphi \colon (\mathbb{C},0) \rightarrow (\mathbb{C},0)$, the chain rule gives:
\begin{align*}
(\log a(f \circ \varphi))' &= \varphi' \cdot (\log a(f))' \circ \varphi, \\
(\log a(f \circ \varphi))'' &= (\varphi')^2 \cdot (\log a(f))'' \circ \varphi + \varphi'' \cdot (\log a(f))' \circ \varphi,
\end{align*}
with analogous expressions for $(\log b(f \circ \varphi))'$ and $(\log b(f \circ \varphi))''$. A direct computation then shows:
\[
\begin{vmatrix}
(\log a(f \circ \varphi))' & (\log b(f \circ \varphi))' \\
(\log a(f \circ \varphi))'' & (\log b(f \circ \varphi))''
\end{vmatrix}
= (\varphi')^3 \cdot
\begin{vmatrix}
(\log a(f))' & (\log b(f))' \\
(\log a(f))'' & (\log b(f))''
\end{vmatrix}
\circ \varphi.
\]
Hence, by Definition~\eqref{define invariant jet differentials}, the Wronskian operator \eqref{Wronskian of (log a)' (log b)' (preliminary)} defines an invariant logarithmic $2$-jet differential on $U_{0,C}$ of weighted degree $3$.
\begin{pro}
\label{Wronskian's magic}
Over the Zariski-open subset $\pi_{2,0}^{-1}(U_{0,C})$, the Wronskian operator \eqref{Wronskian of (log a)' (log b)' (preliminary)} corresponds to a nowhere vanishing holomorphic section $w_{A,B,0}$ of the line bundle $\mathcal{O}_{X_2}(2,1)$.
\end{pro}
While such properties are classical in the compact case (see, e.g., \cite{Demailly1997}), to our knowledge, no convenient reference covers the logarithmic setting with three generic conics. For this reason, we include an explicit proof.
\begin{proof}
Since the Wronskian operator \eqref{Wronskian of (log a)' (log b)' (preliminary)} defines an invariant logarithmic $2$-jet differential of weighted degree $3$ on $U_{0,C}$, the direct image formula
\[
(\pi_{2,0})_{*} \mathcal{O}_{X_2}(m) \cong \mathcal{O}(E_{2,m} T_{\mathbb{P}^2}^{*}(\log \mathcal{C}))
\]
implies that it corresponds to a nonzero section $w_{A,B,0}$ of $\mathcal{O}_{X_2}(3)$ over $U_{0,C}$.
We now determine the vanishing order of $w_{A,B,0}$ along $\Gamma_2$ (see~\eqref{equ:def of Gamma_2}). For any $(x_1, x_2) \in U_{0,C}$, we construct a germ $f \colon (\mathbb{C},0) \rightarrow (U_{0,C}, (x_1, x_2))$ of a logarithmic singular curve satisfying
\[
(f_1(0), f_2(0)) = (x_1,x_2) \quad \text{and} \quad ((\log a(f))'(0), (\log b(f))'(0)) = (0,0),
\]
by recovering $(f_1, f_2)$ from the ODE system:
\begin{equation}\label{singular log jet (peliminary)}
\begin{pmatrix}
(\log a(f))'(t) \\
(\log b(f))'(t)
\end{pmatrix}
=
\begin{pmatrix}
a_1/a & a_2/a \\
b_1/b & b_2/b
\end{pmatrix}
\begin{pmatrix}
f_1'(t) \\
f_2'(t)
\end{pmatrix}
=
\begin{pmatrix}
t h_a(t) \\
t h_b(t)
\end{pmatrix},
\end{equation}
where $h_a$ and $h_b$ are holomorphic functions to be chosen. Solving gives:
\[
\begin{pmatrix}
f_1'(t) \\
f_2'(t)
\end{pmatrix}
= t \cdot
\begin{vmatrix}
a_1 & b_1 \\
a_2 & b_2
\end{vmatrix}^{-1}
\begin{pmatrix}
a b_2 h_a(t) - b a_2 h_b(t) \\
-a b_1 h_a(t) + b a_1 h_b(t)
\end{pmatrix}.
\]
Since the middle determinant is nonvanishing on $U_{0,C}$, the holomorphic functions $(f_1, f_2)$ are determined by integration from the initial condition $(f_1(0), f_2(0)) = (x_1, x_2)$.
Substituting \eqref{singular log jet (peliminary)} into the Wronskian \eqref{Wronskian of (log a)' (log b)' (preliminary)} yields:
\[
\begin{vmatrix}
(\log a(f))'(t) & (\log b(f))'(t) \\
(\log a(f))''(t) & (\log b(f))''(t)
\end{vmatrix}
=
\begin{vmatrix}
t h_a(t) & t h_b(t) \\
h_a(t) + t h_a'(t) & h_b(t) + t h_b'(t)
\end{vmatrix}
= t^2
\begin{vmatrix}
h_a(t) & h_b(t) \\
h_a'(t) & h_b'(t)
\end{vmatrix}.
\]
By choosing $h_a$, $h_b$ such that the last determinant is nonvanishing, we see that $w_{A,B,0}$ vanishes to order exactly $2$ along $\Gamma_2$. Therefore, by~\eqref{gamma_2=O(-1, 1)}, we can regard $w_{A,B,0}$ as a nowhere vanishing section of
\begin{equation}
\label{what is O(2, 1)}
\mathcal{O}_{X_2}(3) \otimes \mathcal{O}(-2\Gamma_2) \cong \mathcal{O}_{X_2}(2,1)
\end{equation}
on $\pi_{2,0}^{-1}(U_{0,C})$.
\end{proof}
\subsection{A Refined Estimate}
The following estimate, formulated in the notation of Subsection~\ref{subsection:4.1}, is a key step in proving our main theorem.
\begin{thm}\label{pro:SMT on Z}
Let $g \colon \mathbb{C} \rightarrow \mathbb{P}^2$ be an entire curve whose lift $g_{[2]} \colon \mathbb{C} \rightarrow X_2$ satisfies $g_{[2]}(\mathbb{C}) \subset Z$. Suppose there exists a section
\begin{equation}
\label{sigma of thm 5.3}
\sigma \in H^0\Big(Z, \big( \mathcal{O}_{X_2}(2,1)^{\tilde{m}} \otimes \pi_{2,0}^*\mathcal{O}_X(-\tilde{t}) \big)|_Z \Big) \cong H^0\Big(Z, \big( \mathcal{O}_{X_2}(3) \otimes \mathcal{O}(- 2 \Gamma_2) \big)^{\tilde{m}} \otimes \pi_{2,0}^*\mathcal{O}_X(-\tilde{t}) \big|_Z \Big),
\end{equation}
such that
$
\sigma(g_{[2]}) \cdot ( g_{[1]}' )^{3\tilde{m}} \not\equiv 0.
$
Then the following Second Main Theorem type estimate holds:
\[
T_{g}(r) \leqslant \frac{2\tilde{m}}{\tilde{t}} \sum_{i=1}^{3} N_{g}^{(1)}(r,\mathcal{C}_i) + o\big(T_g(r)\big) \quad \|.
\]
\end{thm}
Our proof builds on the geometric interpretation of Wronskian operators and incorporates techniques from the proof of \cite[Theorem~3.1]{Huynh-Vu-Xie-2017}. This approach allows us to improve the constant on the right-hand side from $\frac{3 \tilde{m}}{\tilde{t}}$ (check Theorem~\ref{smt-form-logarithmic-diff-jet}) to $\frac{2\tilde{m}}{\tilde{t}}$. This refinement constitutes a key contribution of the present work and underscores the advantage of using invariant jet differentials over normal ones.
\begin{proof}
%%On the Zariski-open set $U_{0 , C}$, the logarithmic $1$-forms $\left(\log a \right)'$ and $\left(\log b \right)'$ generate the logarithmic cotangent bundle $T_{\mathbb{P}^2}^{*} \left( \log \mathcal{C} \right)$. Thus we can choose the dual basis $\mathbf{e}_a$ and $\mathbf{e}_b$ of the logarithmic tangent bundle $T_{\mathbb{P}^2} \left( - \log \mathcal{C} \right)$.
%%Let $f \colon (\mathbb{C},0) \ni t \mapsto f(t) \in U_{0 , C} \subset \mathbb{P}^2$ be a germ of holomorphic curve, and $f = (f_1,f_2)$ be the local representation of $f$ over the open set $U_{0 , C}$. Then we can write $f' = f_1' \partial / \partial x_1 + f_2' \partial / \partial x_2 = \left(\log a(f) \right)' \mathbf{e}_a + \left(\log b(f) \right)' \mathbf{e}_b$.
%%The construction of Demailly-Semple tower yields $f'(t) \in \mathcal{O}_{X_1}(-1)_{f_{[1]}(t)}$ and $f_{[1]}'(t) \in \mathcal{O}_{X_2}(-1)_{f_{[2]}(t)}$. Moreover, by the definition of $\Gamma_{2}$ and $\pi_{1,0} \circ f_{[1]} = f$, we can conclude that $f_{[2]}(t) \in \Gamma_{2}$ if and only if ${\pi_{1,0}}_{*} f_{[1]}'(t) = f'(t) = 0$.
Consider the Wronskian operator \eqref{Wronskian of (log a)' (log b)' (preliminary)} which corresponds to a nonvanishing section $w_{A,B,0}$ of $\mathcal{O}_{X_2}(2,1)$ on $\pi_{2,0}^{-1} (U_{0 , C})$. Hence it induces a trivialization $\mathcal{O}_{X_2}\cdot w_{A,B,0} $ of $\mathcal{O}_{X_2}(2,1)$ on $\pi_{2,0}^{-1} (U_{0 , C})$. Taking the $\tilde{m}$-power thus gives a trivialization
$\mathcal{O}_{X_2}\cdot w_{A,B,0}^{\tilde{m}} $ of $\mathcal{O}_{X_2}(2,1)^{\tilde{m}}$ on $\pi_{2,0}^{-1} (U_{0 , C})$.
Therefore, on $\pi_{2,0}^{-1} (U_{0 , C}) \cap Z$, we can write the section $\sigma$ as
$\mathcal{R} \cdot w_{A,B,0}^{\tilde{m}} \cdot ( \pi_{2,0}^{*} e )^{\tilde{t}}$ for some regular function $\mathcal{R}$ on ${\pi_{2,0}^{-1} (U_{0 , C}) \cap Z}$, where $e$ is a local basis of $\mathcal{O}_{X}(-1)$ on $U_{0,C}$.
Now we equip $\mathcal{O}_{X}(-1)$ with the standard norm induced by the Euclidean metric on $\mathbb{C}^{n+1}$, so that the Chern curvature form $c_1(\mathcal{O}_{X}(-1))=-\omega_{\mathrm{FS}}$ is exactly the negative of the Fubini-Study form. This induces a norm on $\mathcal{O}_{X}(-\tilde{t})$.
By~\eqref{sigma of thm 5.3} and $g_{[1]}'(t) \in \mathcal{O}_{X_2}(-1)_{g_{[2]}(t)}$, we can view $g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}}$ as a nontrivial meromorphic section of $g_{[2]}^{*} ( \pi_{2,0}^*\mathcal{O}_X(-\tilde{t})) = g^{*} ( \mathcal{O}_{\mathbb{P}^2}(-\tilde{t}))$. Thus, on the open subset $g^{-1}(U_{0 , C} \cap \pi_{2,0}(Z))$ of $\mathbb{C}$,
\begin{align*}
\left\| g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} \right\|_{g_{[2]}^{*} ( \pi_{2,0}^*\mathcal{O}_X(-\tilde{t}) ) }
&=
\left| \left( g_{[2]}^{*} \mathcal{R} \right) \left( g_{[2]}^{*} w_{A,B,0} \right)^{\tilde{m}} \cdot \left( g_{[1]}' \right)^{3 \tilde{m}} \right| \left\| g_{[2]}^{*} \pi_{2,0}^{*} e \right\|_{g_{[2]}^{*} ( \pi_{2,0}^*\mathcal{O}_X(-1) )}^{\tilde{t}}
\\
\text{[by $g=\pi_{2,0}\circ g_{[2]}$ and functoriality]}
\quad
&=
\left| \left( g_{[2]}^{*} \mathcal{R} \right)
\begin{vmatrix}
\left(\log a(g) \right)' & \left(\log b(g) \right)' \\
\left(\log a(g) \right)'' & \left(\log b(g) \right)''
\end{vmatrix}^{\tilde{m}} \right| g^{*} \left\| e \right\|_{\mathcal{O}_{\mathbb{P}^2}(-1) }^{\tilde{t}}.
\end{align*}
Note that $g_{[2]}^{*} \mathcal{R}$ is a holomorphic function on $g^{-1}(U_{0 , C} \cap \pi_{2,0}(Z))$.
Let $D_{\sigma , g_{[2]}}$ be the pole divisor of $g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}}$. Then by the Poincar\'{e}-Lelong formula, we obtain that
\begin{align}
\mathrm{dd}^c \log \Vert g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}}\Vert_{_{g_{[2]}^{*} ( \pi_{2,0}^*\mathcal{O}_X(-\tilde{t}) ) }}^2
&
=
-\tilde{t} g_{[2]}^{*} c_1 \left( \pi_{2,0}^*\mathcal{O}_X(-1) \right) + \left[ \mathrm{div} ( g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} ) \right] \label{inequ:dd^c log} \\
&\geqslant
\tilde{t} g^* \omega_{\mathrm{FS}} - [ D_{\sigma , g_{[2]}} ], \notag
\end{align}
where
functoriality guarantees that
$g_{[2]}^{*} c_1 \big( \pi_{2,0}^*\mathcal{O}_X(-1)\big)=-g^*\omega_{\mathrm{FS}}$.
Note that for $k=1,2$, the pole order of $\left(\log a(g) \right)^{(k)}$ (resp. $\left(\log b(g) \right)^{(k)}$) at $z \in g^{-1}(U_{0 , C}) \subset \mathbb{C}$ is at most $k \min \{ \ord_{z} A(g) , 1 \}$ (resp. $k \min \{ \ord_{z} B(g) , 1 \}$). Thus
%%\begin{equation*}
%% D_{\sigma , g_{[2]}} |_{g^{-1}(U_{0 , C} \cap \pi_{2,0}(Z))} \leqslant 3 \tilde{m} \sum_{z \in g^{-1}(U_{0 , C} \cap \pi_{2,0}(Z))} \min \{ \ord_{z} AB(g) , 1 \} z.
%%\end{equation*}
\begin{equation*}
D_{\sigma , g_{[2]}} |_{g^{-1}(U_{0 , C} )} \leqslant \tilde{m} \Big( \sum_{z \in g^{-1}(U_{0 , C} )} 2 \big( \min \{ \ord_{z} A(g) , 1 \} + \min \{ \ord_{z} B(g) , 1 \} \big)\, z \Big).
\end{equation*}
Note that $X_2$ can be covered by $\pi_{2,0}^{-1} (U_{i,Q})$ for $i=0,1,2$ and $Q \in \{A,B,C\}$. Applying similar argument on $U_{i,Q}$, we can obtain analogous pole order estimate on $g^{-1}(U_{i,Q})$. Combining these results, we get
\begin{equation*}
D_{\sigma , g_{[2]}} \leqslant 2 \tilde{m} \sum_{z \in \mathbb{C}} \Big( \sum_{Q \in \{A,B,C\}} \min \{ \ord_{z} Q(g) , 1 \} \Big) z.
\end{equation*}
Substituting this in the inequality \eqref{inequ:dd^c log} and applying Jensen's formula, we obtain
\begin{equation*}
\tilde{t}
\cdot T_{g}(r) \leqslant \int_{\partial \mathbb{D}_r} \log \Vert g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}}\Vert_{_{g_{[2]}^{*} ( \pi_{2,0}^*\mathcal{O}_X(-\tilde{t}) ) }} \frac{\mathrm{d} \, \theta}{2 \pi} + 2 \tilde{m} \, \sum_{i=1}^{3} N_{g}^{(1)}(r,\mathcal{C}_i) + O(1).
\end{equation*}
%%Moreover, note that $\min \{ \ord_{z} ABC(g) , 1 \} \leqslant \sum_{Q \in \{A,B,C\}} \min \{ \ord_{z} Q(g) , 1 \}$, then $N_{g}^{(1)}(r,\mathcal{C}) \leqslant \sum_{i=1}^{3} N_{g}^{(1)}(r,\mathcal{C}_i)$.
Therefore, to conclude the proposition, we need to show that
\begin{equation}\label{equ:key estimate in SMT for Z}
\int_{\partial \mathbb{D}_r} \log \Vert g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}}\Vert_{_{g_{[2]}^{*} ( \pi_{2,0}^*\mathcal{O}_X(-\tilde{t}) ) }} \frac{\mathrm{d} \, \theta}{2 \pi} = o\big(T_g(r)\big)\quad \parallel.
\end{equation}
The remaining argument follows the same line as the proof of (3.3) in \cite[Theorem 3.1]{Huynh-Vu-Xie-2017} and we omit it here. Hence we conclude the proof.
\end{proof}
\subsection{An Effective Constant $c$ in the Second Main Theorem for Three Conics}\label{computing c}
Suppose there exists a nonzero negatively twisted invariant logarithmic $2$-jet differential
\[
\omega_1 \in H^0\bigl(\mathbb{P}^2,\, E_{2,m}T^*_{\mathbb{P}^2}(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-t)\bigr)
\]
satisfying
$
0 < t<m
$
and such that the divisor $Z$ of $\omega_1$ on the second level $X_2$ of the logarithmic Demailly--Semple tower for $(\mathbb{P}^2, \mathcal{C})$ is irreducible and reduced modulo $\Gamma_2$. Then, by the argument in Section~\ref{subsection:4.1}, for any rational number $\tau \in (0, \tau_1(t/m))$, we obtain a nontrivial global section
\[
\sigma \in H^0\left( Z,\, \left( \mathcal{O}_{X_2}(2,1) \otimes \pi_{2,0}^* \mathcal{O}_{\mathbb{P}^2}(-\tau) \right)^{\otimes \ell} \right)
\]
for sufficiently divisible $\ell \gg 1$ (depending on $\tau$).
Furthermore, if $f^* \omega_1 \equiv 0$ but $f_{[2]}^* \sigma \not\equiv 0$, then Theorem~\ref{pro:SMT on Z} yields the estimate~\eqref{eq:smt-three-conics} with constant (see~\eqref{what is tau})
$
c = \frac{2}{\tau} > \frac{2}{\tau_1(t/m)}.
$
By the arbitrariness of $\tau$, the estimate~\eqref{eq:smt-three-conics} holds for any constant $c > 2/\tau_1(\frac{t}{m})$.
Direct computation shows that
$
\inf_{0<x<1} \frac{2}{\tau_1(x)}
=
\frac{3+\sqrt{6}}{3}.
$
Thus
$
\frac{2}{\tau_1(t/m)} < c
$
is equivalent to $
c > \frac{3+\sqrt{6}}{3}$ and $0 < \frac{t}{m} < \frac{2(3c^2 - 6c + 1)}{3c(2c-1)}.
$
Meanwhile, in {\bf Case 2}, to ensure that
$
\omega_{2} \in H^0\big(\mathbb{P}^2,\, E_{2,m}T^*_{\mathbb{P}^2}(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(7-t)\big)
$
satisfies $m/(t-7) \leqslant c$, the positive integers $m$ and $t$ associated to the first family $\omega_1$ must satisfy
$
t \geqslant \frac{m}{c} + 7.
$
\smallskip
Taking $c = 19$, we obtain the desired estimate~\eqref{eq:smt-three-conics} for positive integer pairs $(m,t)$ satisfying
$
0 < t < \frac{1940}{2109}m$ or $\quad t \geqslant \frac{m}{19} + 7.$
Note that $\frac{1940}{2109} \times 8 \approx 7.36$ and $\frac{1940}{2109} \times 7 < 7$. To exclude the exceptional cases of $(m, t)$, it suffices to prove the Key Vanishing Lemma~I.
\smallskip
Taking $c = 5$, we obtain the desired estimate~\eqref{eq:smt-three-conics} for positive integer pairs $(m,t)$ satisfying
$
0 < t < \frac{92}{135}m $ or $ t \geqslant \frac{m}{5} + 7$.
One may verify that for $m \geqslant 14$, every $t \geqslant 1$ satisfies one of these inequalities, since
$
\left\lceil \frac{92}{135}m \right\rceil \geqslant \left\lceil \frac{m}{5} + 7 \right\rceil $ for $ m \geqslant 14$.
\smallskip
For $3 \leqslant m \leqslant 14$, the following table provides the relevant data.
\begin{table}[!h]
\centering
\small
\renewcommand{\arraystretch}{1.4} % 增大行高，1.5为伸缩因子，可根据需要调整
\begin{tabular}{|c|*{12}{c|}}
\hline
$m$ & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 & 14 \\
\hline
$\frac{92}{135}m \approx$ & 2.04 & 2.73 & 3.41 & 4.09 & 4.77 & 5.45 & 6.13 & 6.81 & 7.50 & 8.18 & 8.86 & 9.54 \\
\hline
$\left\lfloor \frac{92}{135}m \right\rfloor$ & 2 & 2 & 3 & 4 & 4 & 5 & 6 & 6 & 7 & 8 & 8 & 9 \\
\hline
\end{tabular}
\end{table}
Thus, to exclude the exceptional cases of $(m, t)$, we only need to establish the Key Vanishing Lemma~II.
\section{\bf Proof of the Key Vanishing Lemmas I and II}\label{section 7}
\subsection{Basic Calculations}
Let $\mathcal{C} = \mathcal{C}_1 + \mathcal{C}_2 + \mathcal{C}_3$ be a divisor of three smooth conics in $\mathbb{P}^2$, defined by homogeneous quadratic polynomials:
\begin{align*}
A(Z) &= A_{200}Z_0^2 + A_{020}Z_1^2 + A_{002}Z_2^2 + A_{110}Z_0Z_1 + A_{101}Z_0Z_2 + A_{011}Z_1Z_2, \\
B(Z) &= B_{200}Z_0^2 + B_{020}Z_1^2 + B_{002}Z_2^2 + B_{110}Z_0Z_1 + B_{101}Z_0Z_2 + B_{011}Z_1Z_2, \\
C(Z) &= C_{200}Z_0^2 + C_{020}Z_1^2 + C_{002}Z_2^2 + C_{110}Z_0Z_1 + C_{101}Z_0Z_2 + C_{011}Z_1Z_2,
\end{align*}
where $Z = (Z_0,Z_1,Z_2)$ and all coefficients $A_{\bullet}, B_{\bullet}, C_{\bullet} \in \mathbb{C}$ are generic.
On the affine chart $U_0 = \{Z_0 \neq 0\}$ with coordinates $(x_1,x_2) = (Z_1/Z_0,Z_2/Z_0)$, the dehomogenized polynomials are:
\begin{align*}
a(x_1,x_2) &= A_{200} + A_{020}x_1^2 + A_{002}x_2^2 + A_{110}x_1 + A_{101}x_2 + A_{011}x_1x_2, \\
b(x_1,x_2) &= B_{200} + B_{020}x_1^2 + B_{002}x_2^2 + B_{110}x_1 + B_{101}x_2 + B_{011}x_1x_2, \\
c(x_1,x_2) &= C_{200} + C_{020}x_1^2 + C_{002}x_2^2 + C_{110}x_1 + C_{101}x_2 + C_{011}x_1x_2.
\end{align*}
Their partial derivatives are given by:
\begin{align*}
a_1 &:= \frac{\partial a}{\partial x_1} = 2A_{020}x_1 + A_{110} + A_{011}x_2, \quad
a_2 := \frac{\partial a}{\partial x_2} = 2A_{002}x_2 + A_{101} + A_{011}x_1, \\
b_1 &:= \frac{\partial b}{\partial x_1} = 2B_{020}x_1 + B_{110} + B_{011}x_2, \quad
b_2 := \frac{\partial b}{\partial x_2} = 2B_{002}x_2 + B_{101} + B_{011}x_1, \\
c_1 &:= \frac{\partial c}{\partial x_1} = 2C_{020}x_1 + C_{110} + C_{011}x_2, \quad
c_2 := \frac{\partial c}{\partial x_2} = 2C_{002}x_2 + C_{101} + C_{011}x_1.
\end{align*}
The Jacobian determinant is the homogeneous cubic polynomial:
\[
J(Z) :=
\begin{vmatrix}
\frac{\partial}{\partial Z_0} A & \frac{\partial}{\partial Z_0} B & \frac{\partial}{\partial Z_0} C \\
\frac{\partial}{\partial Z_1} A & \frac{\partial}{\partial Z_1} B & \frac{\partial}{\partial Z_1} C \\
\frac{\partial}{\partial Z_2} A & \frac{\partial}{\partial Z_2} B & \frac{\partial}{\partial Z_2} C
\end{vmatrix}.
\]
Its vanishing locus $\{J=0\}$ is the ramification divisor of the endormorphism:
\begin{equation}
\label{what is sigma}
\sigma \colon \mathbb{P}^2 \rightarrow \mathbb{P}^2 , \quad [Z_0:Z_1:Z_2] \mapsto [A:B:C].
\end{equation}
The utility of such Jacobian curves has been well-documented in the literature, cf. e.g.~\cite{Francois-Duval-2001,Corvaja-Zannier-2008,Guo-Sun-Wang-2021,Guo-Sun-Wang-2022,CHSX2025}.
Under generic coefficient assumptions, the Jacobian curve $\{J=0\}$ satisfies the following transversality properties (cf. e.g.~\cite{Guo-Sun-Wang-2022,CHSX2025}):
\begin{itemize}
\item[{\bf (TP1)}] Transverse intersection with each defining conic $\mathcal{C}_i$ for $i=1,2,3$
\item[{\bf (TP2)}] Transverse intersection with each coordinate hyperplane $\{Z_i=0\}$ for $i=0,1,2$
\end{itemize}
For the Key Vanishing Lemma I (but not for II), we will work with parameters satisfying these transversality conditions.
\begin{pro}\label{prop:log-generating}
Let $\mathcal{H} := \{Z_0=0\} + \{Z_1=0\} + \{Z_2=0\}$
be a divisor consisting of the three coordinate hyperplanes in $\mathbb{P}^2$. Then
the logarithmic cotangent bundle $T^*_{\mathbb{P}^2}(\log \mathcal{H})$
is globally generated by the sections
$(\log\frac{Z_0}{Z_2})'$ and $(\log\frac{Z_1}{Z_2})'$.
\end{pro}
\begin{proof}
The verification is straightforward by direct computation on the three affine charts $\{Z_i \neq 0\}$, $i = 0, 1, 2$.
\end{proof}
\begin{pro}\label{E23 for hyperplanes}
The logarithmic $2$-jet differential bundle $E_{2,m}T^*_{\mathbb{P}^2}(\log \mathcal{H})$ of weight $3$ is globally generated by the following sections:
\begin{equation}\label{basis about Z0/Z2, Z1/Z2}
\left(\left(\log\frac{Z_0}{Z_2}\right)'\right)^k\left(\left(\log\frac{Z_1}{Z_2}\right)'\right)^{m-3j-k}
\begin{vmatrix}
\left(\log\frac{Z_0}{Z_2}\right)' & \left(\log\frac{Z_1}{Z_2}\right)' \\
\left(\log\frac{Z_0}{Z_2}\right)'' & \left(\log\frac{Z_1}{Z_2}\right)''
\end{vmatrix}^j
\quad (0 \leqslant j \leqslant \lfloor m/3 \rfloor , \; 0 \leqslant k \leqslant m-3j).
\end{equation}
\end{pro}
\begin{proof}
By a similar direct computation on each affine chart $\{Z_i \neq 0\}$ as in Proposition~\ref{prop:log-generating}.
\end{proof}
The endomorphism $\sigma$ defined in~\eqref{what is sigma} is in fact a morphism
$
\sigma : (\mathbb{P}^2, \mathcal{C}) \to
(\mathbb{P}^2, \mathcal{H})
$
between two log pairs. The functoriality immediately yields the relations between
logarithmic
differential $1$-forms
\[
\Big(\log\frac{A}{C}\Big)' = \sigma^* \Big(\log\frac{Z_0}{Z_2}\Big)'
\quad\text{and}\quad
\Big(\log\frac{B}{C}\Big)' = \sigma^* \Big(\log\frac{Z_1}{Z_2}\Big)'.
\]
Combining this with Proposition~\ref{prop:log-generating}, we obtain:
\begin{pro}\label{prop:log-generating-proper-transform}
On the Zariski-open set $\{J \neq 0\} \subset \mathbb{P}^2$, the logarithmic 1-forms
$(\log\frac{A}{C})'$ and $(\log\frac{B}{C})'$
generate the sheaf $T^*_{\mathbb{P}^2}(\log \mathcal{C})$.
\end{pro}
We complement this conceptual proof with explicit computations. While the functorial approach is elegant, the detailed coordinate computations reveal specific algebraic structures that will play an essential role in our later analysis.
\begin{proof}
The sections $\left(\log\frac{A}{C}\right)'$ and $\left(\log\frac{B}{C}\right)'$ are clearly holomorphic on $\{J \neq 0\}$. We prove generation by working on affine charts.
On $U_0 = \{Z_0 \neq 0\}$ with coordinates $(x_1,x_2)$, consider the transformation matrix between differentials:
\begin{equation*}
\label{equ:2 times 2 transition formula}
\begin{pmatrix}
\left(\log\frac{a}{c}\right)' \\ \left(\log\frac{b}{c}\right)'
\end{pmatrix}
= M \begin{pmatrix}
x_1' \\ x_2'
\end{pmatrix}, \quad
M = \begin{pmatrix}
\frac{\partial}{\partial x_1}\log\frac{a}{c} & \frac{\partial}{\partial x_2}\log\frac{a}{c} \\
\frac{\partial}{\partial x_1}\log\frac{b}{c} & \frac{\partial}{\partial x_2}\log\frac{b}{c}
\end{pmatrix}.
\end{equation*}
A direct computation gives the determinant of $M$:
\begin{equation}
\label{what is D}
\det M = \frac{D}{abc}, \quad \text{where} \quad D = \begin{vmatrix}
a & b & c \\
a_1 & b_1 & c_1 \\
a_2 & b_2 & c_2
\end{vmatrix} = \frac{J}{2Z_0^3}.
\end{equation}
Thus the matrix $M$ is invertible on the Zariski open set $U_0 \cap \{J\neq 0\} \cap \{ABC \neq 0\}$, with its inverse given explicitly by:
\[
M^{-1} = \frac{1}{D}
\begin{pmatrix}
a(c b_2 - b c_2) & b(a c_2 - c a_2) \\
a(b c_1 - c b_1) & c(b a_1 - a b_1)
\end{pmatrix}.
\]
Hence inverting the relation~\eqref{what is D} yields the explicit coordinate transformation:
\begin{equation}
\label{equ:inverse of 2 times 2 transition formula}
\begin{pmatrix}
x_1' \\ x_2'
\end{pmatrix}
= \frac{1}{D}
\begin{pmatrix}
a(c b_2 - b c_2) & b(a c_2 - c a_2) \\
a(b c_1 - c b_1) & c(b a_1 - a b_1)
\end{pmatrix}
\begin{pmatrix}
\left(\log\frac{a}{c}\right)' \\ \left(\log\frac{b}{c}\right)'
\end{pmatrix}.
\end{equation}
This establishes that the logarithmic differential forms $\left(\log\frac{A}{C}\right)'$ and $\left(\log\frac{B}{C}\right)'$ generate the sheaf of logarithmic cotangent bundle $T^*_{\mathbb{P}^2}(\log\mathcal{C})$ over the Zariski-open set $\{J\neq 0\}\cap \{ABC\neq 0\}$, as the transformation matrix consists of regular functions on this domain.
Furthermore, the differential $(\log c)'$ can be expressed in terms of our generators:
\begin{align*}
(\log c)' &= \frac{c'}{c} = \begin{pmatrix}
\frac{c_1}{c} & \frac{c_2}{c}
\end{pmatrix}
\begin{pmatrix}
x_1' \\ x_2'
\end{pmatrix} \\
\text{[use~\eqref{equ:inverse of 2 times 2 transition formula}]}\quad
&= \frac{1}{D}\begin{pmatrix}
a \begin{vmatrix}
c_1 & b_1 \\
c_2 & b_2
\end{vmatrix} &
b \begin{vmatrix}
a_1 & c_1 \\
a_2 & c_2
\end{vmatrix}
\end{pmatrix}
\begin{pmatrix}
\left(\log\frac{a}{c}\right)' \\ \left(\log\frac{b}{c}\right)'
\end{pmatrix}.
\end{align*}
Thus sections $\left(\log\frac{A}{C}\right)'$ and $\left(\log\frac{B}{C}\right)'$ continue to generate the logarithmic cotangent bundle $T^*_{\mathbb{P}^2}(\log\mathcal{C})$ in a neighborhood of $\{J\neq 0\}\cap\{C=0\}$ inside $\{J\neq 0\}$, as the transformation matrix remains regular when approaching this locus.
The analogous argument holds on charts $U_1=\{Z_1 \neq 0\}$ and $U_2=\{Z_2 \neq 0\}$, completing the proof.
\end{proof}
Using Proposition~\ref{E23 for hyperplanes} and considering the morphism $\sigma : (\mathbb{P}^2, \mathcal{C}) \to (\mathbb{P}^2, \mathcal{H})$, by functoriality, we obtain:
\begin{pro}\label{prop4}
The bundle $E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C})$ is freely generated over $\{J \neq 0\}$ by the logarithmic jet differentials:
\begin{equation}\label{basis about A/C, B/C}
\left(\left(\log\frac{A}{C}\right)'\right)^k\left(\left(\log\frac{B}{C}\right)'\right)^{m-3j-k}
\begin{vmatrix}
\left(\log\frac{A}{C}\right)' & \left(\log\frac{B}{C}\right)' \\
\left(\log\frac{A}{C}\right)'' & \left(\log\frac{B}{C}\right)''
\end{vmatrix}^j
\quad (0 \leqslant j \leqslant \lfloor m/3 \rfloor , \; 0 \leqslant k \leqslant m-3j).
\qed
\end{equation}
\end{pro}
Let $P \in \Gamma \big( \{J \neq 0\} , E_{2,m}T^*{\mathbb{P}^2}(\log\mathcal{C}) \otimes \mathcal{O}(-t) \big)$. To check and verify the global extendability of $P$, it suffices to demonstrate that $P$ admits a holomorphic extension across the Jacobian divisor $\{J = 0\}$. Note that the intersection $\{J = 0\} \cap \{Z_0ABC = 0\}$ constitutes a Zariski-closed subset of codimension $2$ in $\mathbb{P}^2$. By Hartog's extension theorem, it is thus sufficient to extend $P$ to $\{J = 0\} \cap \{Z_0ABC \neq 0\}$. This extension problem necessitates an analysis of the transition formula for $E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C}) \otimes \mathcal{O}(-t)$ from $\{J \neq 0\}$ to $\{Z_0ABC \neq 0\}$.
Our carefully chosen basis \eqref{basis about A/C, B/C} induces a highly homogeneous representation of the logarithmic jet differentials, which substantially mitigates the complications arising from inhomogeneous coordinate transitions on affine charts. %Indeed, this basis maintains consistency across all affine charts $U_{i} = \{Z_{i} \neq 0 \}$ for $i=0,1,2$. As a result, it suffices to analyze the transition formula for the restriction from $U_0 \cap \{J \neq 0\}$ to $U_0 \cap \{ABC \neq 0\}$, thereby significantly reducing the computational complexity.
\medskip
To convey the core ideas, we demonstrate the proof of Key Vanishing Lemma I in the “advanced case” $(m, t) = (3, 1)$. The other cases, while verifiable by similar means, are omitted here because they follow systematically from the Key Vanishing Lemma II.
\subsection{Exemplary Case $m=3, t=1$}\label{subsection 7.2}
For any global section $P \in H^0(\mathbb{P}^2, E_{2,3}T^*_{\mathbb{P}^2}(\log\mathcal{C})\otimes\mathcal{O}(-1))$, after suitable normalization, we may view $P$ as a global section of $E_{2,3}T^*_{\mathbb{P}^2}(\log\mathcal{C})$ vanishing along a divisor $\{\zeta=0\}$ for some fixed $\zeta \in H^0(\mathbb{P}^2, \mathcal{O}(1))$. For simplicity, we take $\zeta := Z_2$. On $\{J \neq 0\}$, by Proposition~\ref{prop4}, $P$ has the local decomposition:
\begin{equation*}\label{equ:P on D(J)}
P = R \begin{vmatrix}
\left(\log\frac{A}{C}\right)' & \left(\log\frac{B}{C}\right)' \\
\left(\log\frac{A}{C}\right)'' & \left(\log\frac{B}{C}\right)''
\end{vmatrix}
+ \sum_{k=0}^3 R^{k,3-k} \left(\left(\log\frac{A}{C}\right)'\right)^k \left(\left(\log\frac{B}{C}\right)'\right)^{3-k},
\end{equation*}
where, by Serre's GAGA principle~\cite{MR82175} applied to the algebraic vector bundle \(E_{2,3}T_{\mathbb{P}^{2}}^{*}(\log\mathcal{C})\), the coefficients \(R,R^{k,3-k}\) are algebraic rational functions on \(\{J \neq 0\}\); consequently they can be written in the form
\begin{equation}\label{shape of R}
\frac{K}{J^\ell} \quad \text{with} \quad
\begin{cases}
K \in \mathbb{C}[Z_0,Z_1,Z_2]_{3\ell} \text{ homogeneous}, \\
\ell \geqslant 0 \text{ integer}, \\
\deg K = \deg J^\ell = 3\ell.
\end{cases}
\end{equation}
The core technical advances of our approach involve two principal components:
\begin{itemize}
\item[(i)] Employing algebro-geometric techniques to establish the uniform bound of $\ell$, thereby reducing the infinite-dimensional problem to a computationally tractable finite-dimensional linear algebra framework;
\item[(ii)] Implementing a verification scheme (executed in Maple) that confirms the identical vanishing $P \equiv 0$ through explicit computation with judiciously chosen specializations of the coefficients $A_{\bullet}$, $B_{\bullet}$, and $C_{\bullet}$.
\end{itemize}
The semi-continuity principle then guarantees the extension of this vanishing property to all generic coefficient choices $A_{\bullet}$, $B_{\bullet}$, $C_{\bullet}$.
\medskip
On the affine chart $U_0 \cap \{J \neq 0\} = \{Z_0 \neq 0\} \cap \{J \neq 0\}$ with coordinates $(x_1,x_2) = (Z_1/Z_0,Z_2/Z_0)$, the section $P$ has the local expression:
\begin{equation}\label{equ:P on U_0 cap D(J)}
P = R_0(x_1,x_2)
\begin{vmatrix}
\left(\log\frac{a}{c}\right)' & \left(\log\frac{b}{c}\right)' \\
\left(\log\frac{a}{c}\right)'' & \left(\log\frac{b}{c}\right)''
\end{vmatrix}
+ \sum_{k=0}^3 R_0^{k,3-k}(x_1,x_2) \left(\left(\log\frac{a}{c}\right)'\right)^k \left(\left(\log\frac{b}{c}\right)'\right)^{3-k},
\end{equation}
where the coefficients $R_0, R_0^{k,3-k} \in \mathbb{C}(x_1,x_2)$ are rational functions whose denominators consist of powers of the discriminant $D$ defined in~\eqref{what is D}, and which vanish to order $1$ along the divisor $\{x_2=0\}$.
When restricted to $U_0 \cap \{ABC \neq 0\}$, the sheaf $E_{2,3}T^*_{\mathbb{P}^2}(\log\mathcal{C})$ becomes free with basis:
\begin{equation}
\label{new basis}
\begin{vmatrix}
x_1' & x_2' \\
x_1'' & x_2''
\end{vmatrix}
\quad \text{and} \quad
(x_1')^k (x_2')^{3-k} \quad (0 \leqslant k \leqslant 3).
\end{equation}
Consequently, the section $P$ vanishing along $\{Z_2=0\}$ with order $1$ can be rewritten as:
\begin{equation}\label{equ:P on U_0 cap ABC neq 0}
P = \widetilde{R}_{0}(x_1,x_2)
\begin{vmatrix}
x_1' & x_2' \\
x_1'' & x_2''
\end{vmatrix}
+ \sum_{k=0}^3 \widetilde{R}_0^{k,3-k}(x_1,x_2) (x_1')^k (x_2')^{3-k},
\end{equation}
where $\widetilde{R}_{0}, \widetilde{R}_0^{k,3-k}$ are regular on $U_0 \cap \{abc \neq 0\}$ and vanish to order $1$ along the divisor $\{x_2=0\}$.
Comparing the expressions~\eqref{equ:P on U_0 cap D(J)} and~\eqref{equ:P on U_0 cap ABC neq 0} on $U_0 \cap \{ABC \neq 0\} \cap \{J \neq 0\}$ gives the coefficient transformation:
\begin{equation}
\label{comparing the coefficients}
\begin{pmatrix}
\widetilde{R}_0 \\
\widetilde{R}_0^{30} \\
\widetilde{R}_0^{21} \\
\widetilde{R}_0^{12} \\
\widetilde{R}_0^{03}
\end{pmatrix}
= \Phi
\begin{pmatrix}
R_0 \\
R_0^{30} \\
R_0^{21} \\
R_0^{12} \\
R_0^{03}
\end{pmatrix}
=
\begin{pmatrix}
(abc)^{-1}D & \mathbf{0}_{1\times4} \\
\Delta & \Psi
\end{pmatrix}
\begin{pmatrix}
R_0 \\
R_0^{30} \\
R_0^{21} \\
R_0^{12} \\
R_0^{03}
\end{pmatrix},
\end{equation}
where $\Phi$ is the $5\times5$ transition matrix between bases~\eqref{basis about A/C, B/C} and~\eqref{new basis}, with:
\begin{itemize}
\item $\Delta = (abc)^{-2} \delta = (abc)^{-2}(\delta^{30}, \delta^{21}, \delta^{12}, \delta^{03})^\mathsf{T}$,
where $\delta^{30}, \delta^{21}, \delta^{12}, \delta^{03}$ are some polynomials in $\mathbb{C}[A_{\bullet}, B_{\bullet}, C_{\bullet}][x_1, x_2]$. To avoid diverting from our current focus, we omit their explicit expressions here.
\smallskip
\item $\Psi$ being the $4\times4$ submatrix with factorization:
\begin{equation}
\label{what is Psi}
\Psi = \psi \cdot \mathrm{diag}\left(-\frac{1}{a^3c^3}, -\frac{1}{a^2bc^3}, -\frac{1}{ab^2c^3}, -\frac{1}{b^3c^3}\right),
\end{equation}
where the last term means a diagonal matrix with the specified entries, and where $\psi$ is the matrix:
\[
\begin{pmatrix}
\alpha_1^3 & \alpha_1^2\beta_1 & \alpha_1\beta_1^2 & \beta_1^3 \\
3\alpha_1^2\alpha_2 & \alpha_1(2\alpha_2\beta_1 + \alpha_1\beta_2) & \beta_1(2\beta_2\alpha_1 + \beta_1\alpha_2) & 3\beta_1^2\beta_2 \\
3\alpha_1\alpha_2^2 & \alpha_2(2\alpha_1\beta_2 + \alpha_2\beta_1) & \beta_2(2\beta_1\alpha_2 + \beta_2\alpha_1) & 3\beta_1\beta_2^2 \\
\alpha_2^3 & \alpha_2^2\beta_2 & \alpha_2\beta_2^2 & \beta_2^3
\end{pmatrix}
\]
with $\alpha_i = a c_i - a_i c$ and $\beta_i = b c_i - b_i c$ for $i=1,2$.
\end{itemize}
From equation~\eqref{comparing the coefficients}, we immediately obtain the relation:
\begin{equation}
\label{compare two R_0}
\widetilde{R}_{0} = (abc)^{-1} D R_0.
\end{equation}
Under our standing assumption that the Jacobian curve $\{J=0\}$ intersects the three smooth conics $\mathcal{C}_1, \mathcal{C}_2, \mathcal{C}_3$ transversally, the discriminant $D$ is coprime to $a$, $b$, and $c$ in $\mathbb{C}[x_1,x_2]$ (cf.~\cite{Guo-Sun-Wang-2022,CHSX2025}).
By the form of $R_0$ given in~\eqref{shape of R}, we can write:
\[ R_0 = \frac{t_0}{D^\ell} \]
where $\ell \geqslant 0$, $t_0 \in \mathbb{C}[x_1,x_2]$ is coprime to $D$, and $\deg t_0 \leqslant 3\ell$.
\medskip
\textbf{Claim.} The exponent satisfies $\ell \leqslant 1$.
\begin{proof}
Assume, for contradiction, that $\ell \geqslant 2$. Through equation~\eqref{compare two R_0}, we obtain the representation
\[
\widetilde{R}_{0} = \frac{t_0}{abc D^{\ell-1}}.
\]
The denominator reveals that $\widetilde{R}_{0}$ must have poles on the Zariski-open set $U_0 \cap \{ABC \neq 0\}$ along $\{D=0\}$. This conclusion follows from:
\begin{itemize}
\item Our geometric transversality assumption that $\{J=0\}$ meets each conic $\mathcal{C}_i$ transversally
\item The consequent coprimality between $D$ and $abc$
\end{itemize}
However, this contradicts the required regularity of $\widetilde{R}_{0}$. We therefore conclude that $\ell \leqslant 1$ must hold.
\end{proof}
Thus we can write
\begin{equation}\label{equ:R_0 = r_0 / D}
R_0 = \frac{r_0}{D}, \quad \text{with } r_0 \in \mathbb{C}[x_1,x_2] \text{ and } \deg r_0 \leqslant 3.
\end{equation}
The remaining coefficients satisfy the transformation:
\begin{equation*}
\begin{pmatrix}
\widetilde{R}_0^{30} \\
\widetilde{R}_0^{21} \\
\widetilde{R}_0^{12} \\
\widetilde{R}_0^{03}
\end{pmatrix}
= (abc)^{-2} R_0
\begin{pmatrix}
\delta^{30} \\
\delta^{21} \\
\delta^{12} \\
\delta^{03}
\end{pmatrix}
+ \Psi
\begin{pmatrix}
R_0^{30} \\
R_0^{21} \\
R_0^{12} \\
R_0^{03}
\end{pmatrix}.
\end{equation*}
Substituting $R_0 = r_0/D$ into the above equation yields:
\begin{equation}\label{equ:R_0^k,3-k}
\begin{pmatrix}
R_0^{30} \\
R_0^{21} \\
R_0^{12} \\
R_0^{03}
\end{pmatrix}
= \frac{1}{(abc)^2 D} \Psi^{-1} \left( (abc)^2 D
\begin{pmatrix}
\widetilde{R}_0^{30} \\
\widetilde{R}_0^{21} \\
\widetilde{R}_0^{12} \\
\widetilde{R}_0^{03}
\end{pmatrix}
- r_0
\begin{pmatrix}
\delta^{30} \\
\delta^{21} \\
\delta^{12} \\
\delta^{03}
\end{pmatrix}
\right).
\end{equation}
To analyze the pole orders of the rational functions $R_0^{30},
R_0^{21},
R_0^{12},
R_0^{03}$ along $\{D=0\}$, we compute $\Psi^{-1}$ explicitly using the factorization in~\eqref{what is Psi}. The key observation is (see~\eqref{what is D}):
\[
\begin{vmatrix}
\alpha_1 & \beta_1 \\
\alpha_2 & \beta_2
\end{vmatrix} = abc^2 \det M = cD,
\]
which leads to:
{\small
\[
\Psi^{-1} = \frac{-1}{D^3} \operatorname{diag}(a^3, a^2b, ab^2, b^3)
\begin{pmatrix}
\beta {_{2}}^{3} & -\beta _{1} \beta {_{2}}^{2} & \beta {_{1}}^{2} \beta _{2} & -\beta {_{1}}^{3}\\
-3\alpha _{2} \beta {_{2}}^{2} & \beta _{2}\left( 2\beta _{1} \alpha _{2} +\beta _{2} \alpha _{1}\right) & -\beta _{1}\left( 2\beta _{2} \alpha _{1} +\beta _{1} \alpha _{2}\right) & 3\beta {_{1}}^{2} \beta _{2}\\
3\alpha {_{2}}^{2} \beta _{2} & -\alpha _{2}\left( 2\alpha _{1} \beta _{2} +\alpha _{2} \beta _{1}\right) & \alpha _{1}\left( 2\alpha _{2} \beta _{1} +\alpha _{1} \beta _{2}\right) & -3\beta _{1} \beta {_{2}}^{2}\\
-\alpha {_{2}}^{3} & \alpha _{1} \alpha {_{2}}^{2} & -\alpha {_{1}}^{2} \alpha _{2} & \alpha {_{1}}^{3}
\end{pmatrix}
\]
}
by direct computations.
The pole structure of the left-hand side in equation~\eqref{equ:R_0^k,3-k} is governed by three components of its right-hand side:
\begin{itemize}
\item The prefactor $\frac{1}{(abc)^2 D}$ contributes a simple pole along $\{D=0\}$.
\item The inverse matrix $\Psi^{-1} = -\frac{1}{D^3}[\cdots]$ introduces a triple pole along $\{D=0\}$.
\item The vector $\Psi^{-1} (\delta^{30}, \delta^{21}, \delta^{12}, \delta^{03})^\mathsf{T} = \frac{1}{D^2} (\cdots)^{\mathsf{T}}$ introduces a double pole along $\{D=0\}$ by direct computations. The key aspect of our argument is to bound the pole orders along $\{D=0\}$ by effective small numbers (not necessary optimal), thereby managing the computational complexity through computer-assisted calculations subsequently.
\item The numerator terms $(abc)^2D\widetilde{R}_0^{k,3-k}$ and $r_0\delta^{k,3-k}$ remain regular on $\{D=0\}\cap\{abc\neq 0\}\cap U_0$.
\end{itemize}
Therefore, our transversality assumption --- $\{J=0\}$ meets both the conics $\mathcal{C}_i$ and coordinate hyperplanes transversally --- ensures that:
\begin{equation}\label{equ:R_0^pq = r_0^pq / D^3}
R_0^{k,3-k} = \frac{r_0^{k,3-k}}{D^3}, \quad \text{where } r_0^{k,3-k} \in \mathbb{C}[x_1,x_2].
\end{equation}
%%The maximal pole order 4 results from summing the triple pole of $\Psi^{-1}$ with the simple pole of the prefactor.
Furthermore, the degree bound $\deg r_0^{k,3-k} \leqslant 9$ is naturally constrained by the cubic nature of $D$.
Substituting the expressions from~\eqref{equ:R_0 = r_0 / D} and~\eqref{equ:R_0^pq = r_0^pq / D^3} into~\eqref{comparing the coefficients} yields the transformed coefficients:
\begin{equation}
\label{final matrix}
\begin{pmatrix}
\widetilde{R}_0 \\
\widetilde{R}_0^{30} \\
\widetilde{R}_0^{21} \\
\widetilde{R}_0^{12} \\
\widetilde{R}_0^{03}
\end{pmatrix}
= \frac{-1}{(abc)^3 D^3}
\begin{pmatrix}
\widetilde{r}_0 \\
\widetilde{r}_0^{30} \\
\widetilde{r}_0^{21} \\
\widetilde{r}_0^{12} \\
\widetilde{r}_0^{03}
\end{pmatrix},
\end{equation}
where the numerator polynomials are determined by:
\begin{equation}\label{equ:formula of numerator}
\begin{pmatrix}
\widetilde{r}_0 \\
\widetilde{r}_0^{30} \\
\widetilde{r}_0^{21} \\
\widetilde{r}_0^{12} \\
\widetilde{r}_0^{03}
\end{pmatrix}
=
\begin{pmatrix}
-abc D^3 r_0 \\
\psi \cdot \operatorname{diag}(b^3, ab^2, a^2b, a^3) \cdot \mathbf{r} - abc D^2 r_0 \mathbf{\delta}
\end{pmatrix},
\end{equation}
with $\mathbf{r} = (r_0^{30}, r_0^{21}, r_0^{12}, r_0^{03})^\mathsf{T}$ and $\mathbf{\delta} = (\delta^{30}, \delta^{21}, \delta^{12}, \delta^{03})^\mathsf{T}$.
The vanishing of $R_0$ and $R_0^{k,3-k}$ along $\{x_2=0\}$ to order $1$, together with the expressions in~\eqref{equ:R_0 = r_0 / D} and~\eqref{equ:R_0^pq = r_0^pq / D^3}, forces the polynomials $r_0$ and $r_0^{k,3-k}$ to adopt the specific forms:
\begin{equation}\label{equ:all the r_0 and lambda}
r_0 = x_2 \left( \sum_{\substack{p+q \leqslant 2}} \lambda_{p,q} x_1^p x_2^q \right), \quad
r_0^{k,3-k} = x_2 \bigg( \sum_{\substack{p+q \leqslant 8}} \lambda_{p,q}^{k,3-k} x_1^p x_2^q \bigg)
\quad (\lambda_{p,q}, \lambda_{p,q}^{k,3-k} \in \mathbb{C}).
\end{equation}
The regularity conditions for $\widetilde{R}_{0}$ and $\widetilde{R}_0^{k,3-k}$ on $U_0 \cap \{abc \neq 0\}$, combined with their first-order vanishing along $\{x_2=0\}$, impose the following divisibility requirement by~\eqref{final matrix}: The numerators $\widetilde{r}_{0}$ and $\widetilde{r}_0^{k,3-k}$ must be divisible by $x_2 D^3$ in $\mathbb{C}[x_1,x_2]$.
From \eqref{equ:formula of numerator} and \eqref{equ:all the r_0 and lambda}, it is clear that $\widetilde{r}_{0} = - (abc)^2 x_2 D^3 \left( \sum_{\substack{p+q \leqslant 2}} \lambda_{p,q} x_1^p x_2^q \right)$ satisfies the required divisibility condition. It therefore suffices to verify the divisibility condition for the remaining components $\widetilde{r}_0^{k,3-k}$ with $k=0,1,2,3$, which are given explicitly by:
\begin{equation}\label{equ:formula of numerator2}
\begin{pmatrix}
\widetilde{r}_0^{30} \\
\widetilde{r}_0^{21} \\
\widetilde{r}_0^{12} \\
\widetilde{r}_0^{03}
\end{pmatrix}
= \psi \cdot \operatorname{diag}(b^3, ab^2, a^2b, a^3) \cdot \mathbf{r} - abc D^2 r_0 \mathbf{\delta},
\end{equation}
where $\mathbf{r} = (r_0^{30}, r_0^{21}, r_0^{12}, r_0^{03})^\mathsf{T}$ and $\mathbf{\delta} = (\delta^{30}, \delta^{21}, \delta^{12}, \delta^{03})^\mathsf{T}$.
We consider the complex vector space $V_3$ spanned by the $\binom{2+2}{2} + 4 \times \binom{8+2}{2} = 186$ coefficients $\lambda_{p,q}, \lambda_{p,q}^{k,3-k}$ from \eqref{equ:all the r_0 and lambda}. The target space $\widetilde{V}_3$ parameterizes the coefficients of four $x_2$-divisible polynomials in $\mathbb{C}[x_1,x_2]$ with degree at most $24$, having dimension $\dim \widetilde{V}_3 = 4 \times \binom{24-1+2}{2} = 1200$.
The expression \eqref{equ:formula of numerator2} naturally defines a linear transformation $\mathcal{F}_3: V_3 \to \widetilde{V}_3$. The additional $D^3$-divisibility condition (see~\eqref{final matrix}) specifies a subspace $W_3 \subset \widetilde{V}_3$ consisting of elements that factor as $\widetilde{r}_0^{k,3-k} = x_2 D^3 q_0^{k,3-k}$ where $\deg q_0^{k,3-k} \leqslant 14$, giving $\dim W_3 = 4 \times \binom{14+2}{2} = 480$.
By dimensional counting, we compare:
\[
\dim \mathcal{F}_3(V_3) + \dim W_3 \leqslant 186+480 = 666 < 1200 = \dim \widetilde{V}_3.
\]
Hence it is ``highly likely'' that $\mathcal{F}_3(V_3) \cap W_3 = \{0\}$.
Indeed, we can confirm this conclusion through Maple computations with specific three conics:
\begin{align*}
A(Z) = 2 Z_0^2 + Z_1^2 + Z_2^2 + Z_0Z_1, \quad
B(Z) = Z_0^2 + Z_1^2 + 2 Z_2^2 + Z_0Z_2, \quad
C(Z) = Z_0^2 + 2 Z_1^2 + Z_2^2 + Z_1Z_2.
\end{align*}
Thus by semi-continuity argument, we establish the the vanishing result:
\[
H^0\big(\mathbb{P}^2, E_{2,3}T^*_{\mathbb{P}^2}(\log\mathcal{C})\otimes\mathcal{O}(-1)\big)=0\]
for generic three conics $\mathcal{C}$ in $\mathbb{P}^2$.
\subsection{Proof of the Key Vanishing Lemma II}\label{subsection 7.3}
Verifying the required divisibility conditions via linear algebra becomes computationally intricate for larger values of $m$, even with Maple. The main challenge lies in constructing the linear system that encodes these divisibility constraints.
To streamline explicit computation in Maple, we adopt a strategic simplification by choosing specific coefficients $A_{\bullet}, B_{\bullet}, C_{\bullet}$ such that the three conics $\mathcal{C}$ are of Fermat-type, for example:
\begin{align}
\label{Fermat conics}
A(Z) = 2 Z_0^2 + Z_1^2 + Z_2^2,\quad
B(Z) = Z_0^2 + 2Z_1^2 + Z_2^2, \quad
C(Z) = Z_0^2 + Z_1^2 + 2Z_2^2.
\end{align}
This choice ensures that the Jacobian $J = 32 Z_0 Z_1 Z_2$ takes a particularly simple form, which significantly facilitates the verification of divisibility conditions in Maple.
A minor drawback is that the transversality condition {\bf (TP2)} is no longer satisfied. However, this can be readily overcome using Hartogs' extension theorem: we perform the analysis by transitioning from $\{J \neq 0\}$ not only to $\{Z_0 ABC\neq 0\}$ but also to $\{Z_1 ABC\neq 0\}$, and verify the divisibility in both charts.
We simplify the computation by working modulo a prime $p=5$. The absence of nontrivial solutions in the reduced system implies the same for the original integer system, while being significantly easier to verify.
Using the Maple program provided in Appendix~\ref{app:maple}, we can verify all instances of Key Vanishing Lemma~II. The parameters \( 1 \leqslant m \leqslant 13 \) and \( t \geqslant 1 \) on top of the programe can be adjusted manually to test different cases.\footnote{\ Detailed explanations of these computations
are available at:
\url{https://www.imo.universite-paris-saclay.fr/~joel.merker/d-m.pdf}}
For the most challenging case \( (m, t) = (13, 9) \), the system comprised:
\begin{align*}
\text{Equations:} & \quad \mathbf{35,378}, \\
\text{Variables:} & \quad \mathbf{12,550},
\end{align*}
and was solved in Maple with the following resources:
\begin{align*}
\text{Computation Time:} & \quad \mathbf{22,117} \text{ seconds }, \\
\text{Memory Usage:} & \quad \mathbf{994.27} \text{ MB}.
\end{align*}
%\begin{rmk}
%\label{m=5 for three Fermat-type conics}
%Our Maple computations for generic Fermat-type three conics $\mathcal{C}$ yield, as a byproduct, the expected vanishing for all $(m, t) = (1, 1), (2, 1), (3, 1), (4, 1), (5, 2)$. In contrast, a nonvanishing result occurs unexpectedly for $(m, t) = (5, 1)$:
%\[
%\dim\, E_{2,5}T^*_{\mathbb{P}^2}(\log \mathcal{C}) \otimes \mathcal{O}_{\mathbb{P}^2}(-1) = 9.
%\]
%By the strategy below Remark~\ref{rmk: 1.7}, we can conclude a Second Main Theorem for such Fermat-type three conics $\mathcal{C}$ with constant~$5$. It should be noted that this result does not follow directly from our main theorem, as the ``generic'' assumption therein may exclude the specific Fermat-type configuration considered here. Moreover, our explicit computations enable us to identify concrete examples of such Fermat-type three conics $\mathcal{C}$ satisfying this Second Main Theorem, such as the selection~\eqref{Fermat conics}.
%\end{rmk}
\medskip
\section*{\bf Acknowledgements}
We are grateful to Julien Duval for introducing us a decade ago to the problem of establishing a Second Main Theorem for three conics in $\mathbb{P}^2$. We also thank Junjiro Noguchi for sharing his private notes, and Duc Viet Vu for offering valuable insights from a pluripotential perspective.
S.-Y. Xie would like to thank Daniel Barlet, Fr\'ed\'eric Campana, Yunling Chen, Tao Cui, Hanlong Fang, Jie Liu, Mihai P\u{a}un, Erwan Rousseau, Yum-Tong Siu, Ruiran Sun, and Kang Zuo for inspiring discussions.
S.-Y. Xie and L. Hou also wish to thank Tianyuan Mathematics Research Center for its excellent environment.
We also thank Yunling Chen, Alexandre Eremenko, and Yi C. Huang for their suggestions on an earlier version.
The results of this paper were first presented by S.-Y. Xie at {\sl The Seminar on Several Complex Variables and Complex Geometry} (held from July 27 to August 2, 2025), which was organized by Xiangyu Zhou, Yum-Tong Siu, and John Erik Fornæss.
\section*{\bf Funding}
S.-Y. Xie is partially supported by the National Key R\&D Program of China under Grants No. 2023YFA1010500 and No. 2021YFA1003100, as well as by the National Natural Science Foundation of China under Grants No. 12288201 and No. 12471081.
D.T. Huynh is supported by the Vietnam National Foundation for Science and Technology Development (NAFOSTED) under the grant number 101.04-2025.19
\medskip
\appendix
\addcontentsline{toc}{section}{Appendix}
\section{Maple Code for the Key Vanishing Lemmas}
\label{app:maple}
\begin{lstlisting}
restart: with(LinearAlgebra): with(Groebner):
m := 13;
twist := - m + 4;
p := 5;
dR := 3*(m-0) + twist:
dS := 3*(m-2) + twist:
dU := 3*(m-4) + twist:
dV := 3*(m-6) + twist:
dW := 3*(m-8) + twist:
AA := 2*T^2 + 1*x^2 + 1*y^2;
BB := 1*T^2 + 2*x^2 + 1*y^2;
CC := 1*T^2 + 1*x^2 + 2*y^2;
AT := diff(AA,T): BT := diff(BB,T): CT := diff(CC,T):
AX := diff(AA,X): BX := diff(BB,X): CX := diff(CC,X):
AY := diff(AA,Y): BY := diff(BB,Y): CY := diff(CC,Y):
ATT := diff(AA,T,T): ATX := diff(AA,T,X): ATY := diff(AA,T,Y):
AXX := diff(AA,X,X): AXY := diff(AA,X,Y): AYY := diff(AA,Y,Y):
BTT := diff(BB,T,T): BTX := diff(BB,T,X): BTY := diff(BB,T,Y):
BXX := diff(BB,X,X): BXY := diff(BB,X,Y): BYY := diff(BB,Y,Y):
CTT := diff(CC,T,T): CTX := diff(CC,T,X): CTY := diff(CC,T,Y):
CXX := diff(CC,X,X): CXY := diff(CC,X,Y): CYY := diff(CC,Y,Y):
MatDD := Matrix([[AT,BT,CT],
[AX,BX,CX],
[AY,BY,CY]]):
DD := factor(Determinant(MatDD));
Ap := AT*T1 + AX*X1 + AY*Y1:
Bp := BT*T1 + BX*X1 + BY*Y1:
Cp := CT*T1 + CX*X1 + CY*Y1:
App := + AT*T2 + AX*X2 + AY*Y2
+ ATT*T1*T1 + 2*ATX*T1*X1 + 2*ATY*T1*Y1 + AXX*X1*X1 + 2*AXY*X1*Y1 + AYY*Y1*Y1:
Bpp := + BT*T2 + BX*X2 + BY*Y2
+ BTT*T1*T1 + 2*BTX*T1*X1 + 2*BTY*T1*Y1 + BXX*X1*X1 + 2*BXY*X1*Y1 + BYY*Y1*Y1:
Cpp := + CT*T2 + CX*X2 + CY*Y2
+ CTT*T1*T1 + 2*CTX*T1*X1 + 2*CTY*T1*Y1 + CXX*X1*X1 + 2*CXY*X1*Y1 + CYY*Y1*Y1:
logACp := Ap/AA - Cp/CC:
logBCp := Bp/BB - Cp/CC:
logACpp := App/AA - Ap^2/AAA^2 - (Cpp/CC - Cp^2/CCC^2):
logBCpp := Bpp/BB - Bp^2/BB^2 - (Cpp/CC - Cp^2/CCC^2):
Carte_xy := {T = 1, X = x, Y = y,
T1 = 0, X1 = x1, Y1 = y1,
T2 = 0, X2 = x2, Y2 = y2}:
xyAA := subs(Carte_xy, AA):
xyBB := subs(Carte_xy, BB):
xyCC := subs(Carte_xy, CC):
xyDD := subs(Carte_xy, DD):
xylogACp := subs(Carte_xy, logACp):
xylogBCp := subs(Carte_xy, logBCp):
xylogACpp := subs(Carte_xy, logACpp):
xylogBCpp := subs(Carte_xy, logBCpp):
Carte_tx := {T = t, X = x, Y = 1,
T1 = t1, X1 = x1, Y1 = 0,
T2 = t2, X2 = x2, Y2 = 0}:
txAA := subs(Carte_tx, AA):
txBB := subs(Carte_tx, BB):
txCC := subs(Carte_tx, CC):
txDD := subs(Carte_tx, DD):
txlogACp := subs(Carte_tx, logACp):
txlogBCp := subs(Carte_tx, logBCp):
txlogACpp := subs(Carte_tx, logACpp):
txlogBCpp := subs(Carte_tx, logBCpp):
Jet_ABC := + (1/(T^X^Y)^(m-0))
* add(R[m-0-j,j] * logACp^(m-0-j) * logBCp^j *
(logACp*logBCpp-logACpp*logBCp)^0, j=0..m-0)
+ (1/(T^X^Y)^(m-2))
* add(S[m-3-j,j] * logACp^(m-3-j) * logBCp^j *
(logACp*logBCpp-logACpp*logBCp)^1, j=0..m-3)
+ (1/(T^X^Y)^(m-4))
* add(U[m-6-j,j] * logACp^(m-6-j) * logBCp^j *
(logACp*logBCpp-logACpp*logBCp)^2, j=0..m-6)
+ (1/(T^X^Y)^(m-6))
* add(V[m-9-j,j] * logACp^(m-9-j) * logBCp^j *
(logACp*logBCpp-logACpp*logBCp)^3, j=0..m-9)
+ (1/(T^X^Y)^(m-8))
* add(W[m-12-j,j] * logACp^(m-12-j) * logBCp^j *
(logACp*logBCpp-logACpp*logBCp)^4, j=0..m-12):
xyJet_ABC := factor(+ (1/(x*y)^(m-0))
* add(Rxy[m-0-j,j] * xylogACp^(m-0-j) * xylogBCp^j *
(xylogACp*xylogBCpp-xylogACpp*xylogBCp)^0, j=0..m-0)
+ (1/(x*y)^(m-2))
* add(Sxy[m-3-j,j] * xylogACp^(m-3-j) * xylogBCp^j *
(xylogACp*xylogBCpp-xylogACpp*xylogBCp)^1, j=0..m-3)
+ (1/(x*y)^(m-4))
* add(Uxy[m-6-j,j] * xylogACp^(m-6-j) * xylogBCp^j *
(xylogACp*xylogBCpp-xylogACpp*xylogBCp)^2, j=0..m-6)
+ (1/(x*y)^(m-6))
* add(Vxy[m-9-j,j] * xylogACp^(m-9-j) * xylogBCp^j *
(xylogACp*xylogBCpp-xylogACpp*xylogBCp)^3, j=0..m-9)
+ (1/(x*y)^(m-8))
* add(Wxy[m-12-j,j] * xylogACp^(m-12-j) * xylogBCp^j *
(xylogACp*xylogBCpp-xylogACpp*xylogBCp)^4, j=0..m-12)):
txJet_ABC := + (1/(t*x)^(m-0))
* add(Rtx[m-0-j,j] * txlogACp^(m-0-j) * txlogBCp^j *
(txlogACp*txlogBCpp-txlogACpp*txlogBCp)^0, j=0..m-0)
+ (1/(t*x)^(m-2))
* add(Stx[m-3-j,j] * txlogACp^(m-3-j) * txlogBCp^j *
(txlogACp*txlogBCpp-txlogACpp*txlogBCp)^1, j=0..m-3)
+ (1/(t*x)^(m-4))
* add(Utx[m-6-j,j] * txlogACp^(m-6-j) * txlogBCp^j *
(txlogACp*txlogBCpp-txlogACpp*txlogBCp)^2, j=0..m-6)
+ (1/(t*x)^(m-6))
* add(Vtx[m-9-j,j] * txlogACp^(m-9-j) * txlogBCp^j *
(txlogACp*txlogBCpp-txlogACpp*txlogBCp)^3, j=0..m-9)
+ (1/(t*x)^(m-8))
* add(Wtx[m-12-j,j] * txlogACp^(m-12-j) * txlogBCp^j *
(txlogACp*txlogBCpp-txlogACpp*txlogBCp)^4, j=0..m-12):
NFxy := proc(P,j,i)
expand(mtaylor(P,x,j) + mtaylor(P,y,i) - mtaylor(mtaylor(P,y,i), x,j)):
end proc:
NFix := proc(P,j,i)
expand(mtaylor(P,t,j) + mtaylor(P,x,i) - mtaylor(mtaylor(P,x,i), t,j)):
end proc:
xyJet := mtaylor(
factor((x*y*xyAAA*xyBB*yyyCC)^m * (subs(y2=(xyW+x2*y1)/x1, xyJet_ABC)))
, [x1,y1,xyW], 100):
txJet := mtaylor(
factor((t*x*txAAA*txBB*txCC)^m * (subs(x2=(txW+t2*x1)/t1, txJet_ABC)))
, [t1,x1,txW], 100):
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
xyEEEq[m-j-3*k,j,k] := factor(coeftayl(xyJet, [x1,y1,xyW]={0,0,0}, [m-j-3*k,j,k])):
od:
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
txEEEq[m-j-3*k,j,k] := factor(coeftayl(txJet, [t1,x1,txW]={0,0,0}, [m-j-3*k,j,k])):
od:
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
xyEEEq[m-j-3*k,j,k] := mtaylor(
factor(
(1/(x^((m-j-3*k)*y^)*(x*y)^*(3*k))) *
(1/((xyAA*xyBB*xyCC)^*(2*k))) *
xyEEEq[m-j-3*k,j,k])
, [x,y], 1000):
od:
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
txEEEq[m-j-3*k,j,k] := mtaylor(
factor(
(1/(t^m-j-3*k)*x^j*(t*x)^(3*k))) * txEEEq[m-j-3*k,j,k])
, [t,x], 1000):
od:
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
xyEEq[m-j-3*k,j,k] := NFxy(xyEEEq[m-j-3*k,j,k], j,m-j-3*k) mod p:
od:
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
txEEq[m-j-3*k,j,k] := NFtx(txEEEq[m-j-3*k,j,k], j,m-j-3*k) mod p:
od:
od:
for j from 0 to m-0 do
Rxy[m-0-j,j] := add(add(cR[m-0-j,j,dR-q-r,q,r]*1^dR-q-r)*x^q*y^r, q=0..dR-r), r=0..dR):
od:
for j from 0 to m-3 do
Sxy[m-3-j,j] := add(add(cS[m-3-j,j,dS-q-r,q,r]*1^dS-q-r)*x^q*y^r, q=0..dS-r), r=0..dS):
od:
for j from 0 to m-6 do
Uxy[m-6-j,j] := add(add(cU[m-6-j,j,dU-q-r,q,r]*1^dU-q-r)*x^q*y^r, q=0..dU-r), r=0..dU):
od:
for j from 0 to m-9 do
Vxy[m-9-j,j] := add(add(cV[m-9-j,j,dV-q-r,q,r]*t^((dV-q-r)*x^q*y^r, q=0..dV-r), r=0..dV):
od:
for j from 0 to m-12 do
Wxy[m-12-j,j] := add(add(cW[m-12-j,j,dW-q-r,q,r]*t^((dW-q-r)*x^q*y^r, q=0..dW-r), r=0..dW):
od:
for j from 0 to m-0 do
Rtx[m-0-j,j] := add(add(cR[m-0-j,j,dR-q-r,q,r]*t^((dR-q-r)*x^q*1^r, q=0..dR-r), r=0..dR):
od:
for j from 0 to m-3 do
Stx[m-3-j,j] := add(add(cS[m-3-j,j,dS-q-r,q,r]*t^((dS-q-r)*x^q*1^r, q=0..dS-r), r=0..dS):
od:
for j from 0 to m-6 do
Utx[m-6-j,j] := add(add(cU[m-6-j,j,dU-q-r,q,r]*t^((dU-q-r)*x^q*1^r, q=0..dU-r), r=0..dU):
od:
for j from 0 to m-9 do
Vtx[m-9-j,j] := add(add(cV[m-9-j,j,dV-q-r,q,r]*t^((dV-q-r)*x^q*1^r, q=0..dV-r), r=0..dV):
od:
for j from 0 to m-12 do
Wtx[m-12-j,j] := add(add(cW[m-12-j,j,dW-q-r,q,r]*t^((dW-q-r)*x^q*1^r, q=0..dW-r), r=0..dW):
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
xyEq[m-j-3*k,j,k] := NFxy(xyEEq[m-j-3*k,j,k], j,m-j-3*k):
od:
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
txEq[m-j-3*k,j,k] := NFtx(txEEq[m-j-3*k,j,k], j,m-j-3*k):
od:
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
xyEquations[m-j-3*k,j,k] := {coeffs(xyEq[m-j-3*k,j,k], [x,y])}:
od:
od:
printlevel := 2:
for k from 0 to m/3 do
for j from 0 to m-3*k do
txEquations[m-j-3*k,j,k] := {coeffs(txEq[m-j-3*k,j,k], [t,x])}:
od:
od:
xyEquations_Add := {}
for k from 0 to m/3 do
for j from 0 to m-3*k do
xyEquations_Add := xyEquations_Add union xyEquations[m-j-3*k,j,k]:
od:
od:
Equations_xy := xyEquations_Add:
txEquations_Add := {};
for k from 0 to m/3 do
for j from 0 to m-3*k do
txEquations_Add := txEquations_Add union txEquations[m-j-3*k,j,k]:
od:
od:
Equations_tx := txEquations_Add:
Equations := Equations_xy union Equations_tx mod p:
Nombre_Equations := nops(Equations);
Variables := {seq(seq(seq(cR[m-0-j,j,dR-q-r,q,r], q=0..dR-r), r=0..dR), j=0..m-0)}
union {seq(seq(seq(cS[m-3-j,j,dS-q-r,q,r], q=0..dS-r), r=0..dS), j=0..m-3)}
union {seq(seq(seq(cU[m-6-j,j,dU-q-r,q,r], q=0..dU-r), r=0..dU), j=0..m-6)}
union {seq(seq(seq(cV[m-9-j,j,dV-q-r,q,r], q=0..dV-r), r=0..dV), j=0..m-9)}
union {seq(seq(seq(cW[m-12-j,j,dW-q-r,q,r], q=0..dW-r), r=0..dW), j=0..m-12)}:
Nombre_Variables := nops(Variables);
Solutions := msolve(Equations,p):
Verification := seq(op(2,op(i,Solutions)),i=1..nops(Solutions)):
indets({Verification}):
nops(indets({Verification})):
assign(Solutions):
for j from 0 to m-0 do
Voir_R[m-0-j,j]:= add(add(cR[m-0-j,j,dR-q-r,q,r]*T^((dR-q-r))*X^q*Y^r, q=0..dR-r), r=0..dR):
od:
for j from 0 to m-3 do
Voir_S[m-3-j,j]:= add(add(cS[m-3-j,j,dS-q-r,q,r]*T^((dS-q-r))*X^q*Y^r, q=0..dS-r), r=0..dS):
od:
for j from 0 to m-6 do
Voir_U[m-6-j,j]:= add(add(cU[m-6-j,j,dU-q-r,q,r]*T^((dU-q-r))*X^q*Y^r, q=0..dU-r), r=0..dU):
od:
for j from 0 to m-9 do
Voir_V[m-9-j,j]:= add(add(cV[m-9-j,j,dV-q-r,q,r]*T^((dV-q-r))*X^q*Y^r, q=0..dV-r), r=0..dV):
od:
for j from 0 to m-12 do
Voir_W[m-12-j,j]:= add(add(cW[m-12-j,j,dW-q-r,q,r]*T^((dW-q-r))*X^q*Y^r, q=0..dW-r), r=0..dW):
od:
Voir_R_13,0 := 0:
Voir_R_12,1 := 0:
Voir_R_11,2 := 0:
Voir_R_10,3 := 0:
Voir_R_9,4 := 0:
Voir_R_8,5 := 0:
Voir_R_7,6 := 0:
Voir_R_6,7 := 0:
Voir_R_5,8 := 0:
Voir_R_4,9 := 0:
Voir_R_3,10 := 0:
Voir_R_2,11 := 0:
Voir_R_1,12 := 0:
Voir_R_0,13 := 0:
Voir_S_10,0 := 0:
Voir_S_9,1 := 0:
Voir_S_8,2 := 0:
Voir_S_7,3 := 0:
Voir_S_6,4 := 0:
Voir_S_5,5 := 0:
Voir_S_4,6 := 0:
Voir_S_3,7 := 0:
Voir_S_2,8 := 0:
Voir_S_1,9 := 0:
Voir_S_0,10 := 0:
Voir_U_7,0 := 0:
Voir_U_6,1 := 0:
Voir_U_5,2 := 0:
Voir_U_4,3 := 0:
Voir_U_3,4 := 0:
Voir_U_2,5 := 0:
Voir_U_1,6 := 0:
Voir_U_0,7 := 0:
Voir_V_4,0 := 0:
Voir_V_3,1 := 0:
Voir_V_2,2 := 0:
Voir_V_1,3 := 0:
Voir_V_0,4 := 0:
Voir_W_1,0 := 0:
Voir_W_0,1 := 0:
\end{lstlisting}
\section{A Cohomological Observation}\label{appendix:results of Pengchao Wang}
\noindent
\textit{This appendix was authored by Pengchao Wang} \\
\textit{(School of Mathematical Sciences, Peking University, Beijing 100871, P. R. China).} \\
\textit{Email: 2200010773@stu.pku.edu.cn.}
\vspace{1em}
Let $\mathcal{C}\subset \mathbb{P}^2$ be three generic smooth conics with simple normal crossings.
Let $$0\subset
S^m\,T^*_{\mathbb{P}^2}(\log\mathcal{C})=
F^0(E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C}))\subset F^1(E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C})) \subset
\cdots
\subset F^{\lfloor m/3\rfloor}(E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C}))
\subset
E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C})$$ be
the ascending filtration of subsheaves that defines the
grading~\eqref{filtration of E_2,m}. We set $F^k(E_{2,m}T^*_{\mathbb{P}^2}(\log\mathcal{C})) = 0$ for $k \leqslant -1$. From~\eqref{filtration of E_2,m} we obtain:
\begin{pro}
\label{thm:general SES of E2m}
For every $0\leqslant k\leqslant m/3$,
the following sequence is exact:
\begin{equation*}
\xymatrix{
0 \ar[r] & F^{k-1}(E_{2,m} T^*_{\mathbb{P}^2}(\log\mathcal{C})) \ar[r] & F^k(E_{2,m} T^*_{\mathbb{P}^2}(\log\mathcal{C})) \ar[r] & S^{m-3k} T^*_{\mathbb{P}^2}(\log\mathcal{C}) \otimes \overline{K}_{\mathbb{P}^2}^{\otimes k} \ar[r] & 0.
}
\end{equation*}
\end{pro}
For convenience, here we write $\overline{T}^*_{\mathbb{P}^2} \coloneqq T^*_{\mathbb{P}^2}(\log\mathcal{C})$ and $\mathcal{O}(1) \coloneqq \mathcal{O}_{\mathbb{P}^2}(1)$.
\begin{pro}\label{pro:vanishing of E_2,3q+r}
For integer $q\geqslant 0$ and $r\in \{1, 2 \}$, the following vanishing holds:
\begin{equation*}
H^0 \big( \mathbb{P}^2 , E_{2,3q+r} \overline{T}^*_{\mathbb{P}^2} \otimes \mathcal{O}_{\mathbb{P}^2}(-(3q+r)) \big) = 0,
\end{equation*}
provided that $H^0 \big( \mathbb{P}^2 , S^{3\ell+r} \overline{T}^*_{\mathbb{P}^2} \otimes \mathcal{O}_{\mathbb{P}^2}(-(3\ell+r)) \big) = 0$ for all $\ell=0,1, \dots ,q$.
\end{pro}
\begin{proof}
By Proposition~\ref{thm:general SES of E2m}, for every integer $k=0,1, \dots ,q$, we have a short exact sequence:
\begin{equation*}
0
\rightarrow
F^{k-1}(E_{2,3q+r} \overline{T}^*_{\mathbb{P}^2}) \otimes \mathcal{O}(-(3q+r))
\rightarrow
F^k(E_{2,3q+r} \overline{T}^*_{\mathbb{P}^2}) \otimes \mathcal{O}(-(3q+r))
\rightarrow
S^{3q+r-3k} \overline{T}^*_{\mathbb{P}^2} \otimes \overline{K}_{\mathbb{P}^2}^{\otimes k} \otimes \mathcal{O}(-(3q+r))
\rightarrow
0.
\end{equation*}
Note that the last term $S^{3q+r-3k} \overline{T}^*_{\mathbb{P}^2} \otimes \overline{K}_{\mathbb{P}^2}^{\otimes k} \otimes \mathcal{O}(-(3q+r)) \cong S^{3(q-k)+r} \overline{T}^*_{\mathbb{P}^2} \otimes \mathcal{O}(-(3(q-k)+r))$ has no global section by our assumption.
Considering the induced long exact sequence in cohomology,
we obtain
\begin{equation*}
H^0 \big( \mathbb{P}^2 , F^k(E_{2,3q+r} \overline{T}^*_{\mathbb{P}^2}) \otimes \mathcal{O}(-(3q+r)) \big) \cong H^0 \big( \mathbb{P}^2 , F^{k-1}(E_{2,3q+r} \overline{T}^*_{\mathbb{P}^2}) \otimes \mathcal{O}(-(3q+r)) \big).
\end{equation*}
By iterating backwards for $k=q, q-1, \dots, 1$, we conclude the desired vanishing since $
H^0 \big( \mathbb{P}^2 , S^{3q+r} \overline{T}^*_{\mathbb{P}^2} \otimes \mathcal{O}(-(3q+r)) \big) = 0$ by assumption.
\end{proof}
\begin{rmk}
The above argument fails for $r=0$ because when $q=k$,
the term $$S^{3q+r-3k} \overline{T}^*_{\mathbb{P}^2} \otimes \overline{K}_{\mathbb{P}^2}^{\otimes k} \otimes \mathcal{O}(-(3q+r)) \cong \mathcal{O}_{\mathbb{P}^2}$$
has one-dimensional global sections.
\end{rmk}
\bigskip
\addcontentsline{toc}{chapter}{Bibliography}
\bibliographystyle{plain}
%\thispagestyle{empty}
\bibliography{article}
%\input{tltkhao.tex}
\end{document}
\section{A Bogomolov Type Vanishing Theorem}\label{appendix:Bogomolov Vanishing Thm}
We claim no originality for the content of this appendix and thank Yunling Chen for graciously permitting us to reproduce her argument here.
\begin{thm}[cf. {\cite[Theorem 4.4]{Chen2025}}]
Let $(X,D)$ be a log pair where $X$ is a smooth projective surface and $D$ is a reduced simple normal crossing divisor.
Assume that $K_{X} + D$ is ample and $T_X(- \log D)$ is semi-stable. Then
\begin{equation*}
H^0 \big( X , S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D)) \big) = 0,
\end{equation*}
for all integers $p, q$ satisfying $p - 2q > 0$.
\end{thm}
\begin{proof}
Suppose, for contradiction, that there exists a nonzero section
\begin{equation*}
s \in H^0 \big( X , S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D)) \big),
\end{equation*}
which induces a morphism
\begin{equation*}
\xymatrix{
\mathcal{O}_X \ar[r]^(0.2){\otimes s} & S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D)).
}
\end{equation*}
Hence, the nonzero section $s$ generates a rank $1$ subbundle $L$ of $S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D))$. Since $s$ is a nonzero holomorphic section and $K_{X} + D$ is ample, one has
\begin{equation*}
\mu_{K_{X} + D}(L) = \frac{c_1(L) \cdot (K_{X} + D)}{\rank L} \geqslant 0.
\end{equation*}
Then the semi-stability of $T_{X}(- \log D)$ yields that
\begin{align*}
0 \leqslant \mu_{K_{X} + D}(L) &\leqslant \mu_{K_{X} + D} \big( S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D)) \big) \\
&= \frac{c_1 \big( S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D)) \big) \cdot (K_{X} + D)}{\rank \big( S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D)) \big)} \\
&= \frac{c_1 \big( S^p T_{X}(- \log D) \otimes \mathcal{O}_X(q(K_{X} + D)) \big) \cdot (K_{X} + D)}{p+1} \\
&= \frac{ \big( c_1 ( S^p T_{X}(- \log D) ) + (p+1) c_1 ( \mathcal{O}_X(q(K_{X} + D)) ) \big) \cdot (K_{X} + D)}{p+1} \\
&= \frac{ \big( c_1 ( S^p T_{X}(- \log D) ) + (p+1)q (K_{X} + D) \big) \cdot (K_{X} + D)}{p+1} \\
\text{(cf.~\cite[Lemma 10.1]{Bogomolov1978})} \qquad &= \frac{ \big( \frac{p(p+1)}{2!} c_1 ( T_{X}(- \log D) ) + (p+1)q (K_{X} + D) \big) \cdot (K_{X} + D)}{p+1} \\
&= \frac{ \big( - \frac{p(p+1)}{2!} (K_{X} + D) + (p+1)q (K_{X} + D) \big) \cdot (K_{X} + D)}{p+1} \\
&= -\frac{p-2q}{2} (K_{X} + D)^2 < 0.
\end{align*}
The last inequality holds because $p-2q > 0$ and $K_{X} + D$ is ample. This yields a contradiction.
\end{proof}
\medskip
\section{Remaining Details in the Proof of Theorem~\ref{pro:SMT on Z}}\label{appendix:application of Log Derivative Lemma}
To simplify, we denote
\begin{equation*}
\left\Vert g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} \right\Vert \coloneqq \left\Vert g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} \right\Vert_{_{g_{[2]}^{*} ( \pi_{2,0}^*\mathcal{O}_X(-\tilde{t}) ) }}.
\end{equation*}
Let $\{ \chi_{i,Q} \}$ be a compactly supported partition of unity subordinated to $\{ \pi_{2,0}^{-1}(U_{i,Q}) \}$ where $i=0,1,2$ and $Q \in \{A,B,C\}$. Then there is some constants $C_{i,Q} > 0$ such that
\begin{align*}
|| g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} ||
&=
|| \sum_{i,Q} g_{[2]}^{*} ( \chi_{i,Q} \sigma ) \cdot ( g_{[1]}' )^{3 \tilde{m}} || \\
&\leqslant
\sum_{i,Q} ( \chi_{i,Q} \circ g_{[2]} ) || g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} || \\
&\leqslant
\sum_{i,Q} ( \chi_{i,Q} \circ g_{[2]} ) C_{i,Q} \big| ( g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} ) |_{g^{-1}(U_{i,Q} \cap \pi_{2,0}(Z))} \big|.
\end{align*}
Note that for $s,s_1,\dots,s_N \in \mathbb{R}_{>0}$,
\[
\log s = \log^{+} s - \log^{+} \frac{1}{s} \leqslant \log^{+} s, \quad
\log^{+} \sum_{i=1}^{N} s_i \leqslant \sum_{i=1}^{N} \log^{+} s_i + \log N, \quad
\log^{+} {\textstyle\prod}_{i=1}^{N} s_i \leqslant \sum_{i=1}^{N} \log^{+} s_i.
\]
Hence we have
\begin{align*}
\log || g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} ||
&\leqslant
\log^{+} || g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} || \\
&\leqslant
\log^{+} \sum_{i,Q} ( \chi_{i,Q} \circ g_{[2]} ) C_{i,Q} \big| ( g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} ) |_{g^{-1}(U_{i,Q} \cap \pi_{2,0}(Z))} \big| \\
&\leqslant
\sum_{i,Q} \log^{+} ( \chi_{i,Q} \circ g_{[2]} ) C_{i,Q} \big| ( g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} ) |_{g^{-1}(U_{i,Q} \cap \pi_{2,0}(Z))} \big| + \log 9 \\
&\leqslant
\sum_{i,Q} \log^{+} ( \chi_{i,Q} \circ g_{[2]} ) \big| ( g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} ) |_{g^{-1}(U_{i,Q} \cap \pi_{2,0}(Z))} \big| + \sum_{i,Q} \log^{+} C_{i,Q} + \log 9.
\end{align*}
Without loss of generality, we consider the term with $i=0$ and $Q=C$:
\begin{align*}
&\log^{+} ( \chi_{0,C} \circ g_{[2]} ) \big| ( g_{[2]}^{*} \sigma \cdot ( g_{[1]}' )^{3 \tilde{m}} ) |_{g^{-1}(U_{0,C} \cap \pi_{2,0}(Z))} \big| \\
=
&\log^{+} ( \chi_{0,C} \circ g_{[2]} ) \left| g_{[2]}^{*} \mathcal{R} \right|
\left| \begin{vmatrix}
\left(\log a(g) \right)' & \left(\log b(g) \right)' \\
\left(\log a(g) \right)'' & \left(\log b(g) \right)''
\end{vmatrix} \right|^{\tilde{m}} \\
\leqslant
&\log^{+} C_{0,C}' ( \chi_{0,C} \circ g_{[2]} ) \left| \begin{vmatrix}
\left(\log a(g) \right)' & \left(\log b(g) \right)' \\
\left(\log a(g) \right)'' & \left(\log b(g) \right)''
\end{vmatrix} \right|^{\tilde{m}} \\
\leqslant
&\log^{+} ( \chi_{0,C} \circ g_{[2]} ) \left| \begin{vmatrix}
\left(\log a(g) \right)' & \left(\log b(g) \right)' \\
\left(\log a(g) \right)'' & \left(\log b(g) \right)''
\end{vmatrix} \right|^{\tilde{m}}
+ \log^{+} C_{0,C}' \\
=
&\log^{+} ( \chi_{0,C} \circ g_{[2]} ) \left| \sum_{k=0}^{\tilde{m}} (-1)^{k} \binom{\tilde{m}}{k} \left( \left(\log a(g) \right)' \left(\log b(g) \right)'' \right)^{\tilde{m}-k} \left( \left(\log b(g) \right)' \left(\log a(g) \right)'' \right)^{k} \right| + \log^{+} C_{0,C}' \\
\leqslant
&\log^{+} \sum_{k=0}^{\tilde{m}} \binom{\tilde{m}}{k} ( \chi_{0,C} \circ g_{[2]} ) \left| \left( \left(\log a(g) \right)' \left(\log b(g) \right)'' \right)^{\tilde{m}-k} \left( \left(\log b(g) \right)' \left(\log a(g) \right)'' \right)^{k} \right| + \log^{+} C_{0,C}' \\
\leqslant
&\sum_{k=0}^{\tilde{m}} \log^{+} \binom{\tilde{m}}{k} ( \chi_{0,C} \circ g_{[2]} ) \left| \left( \left(\log a(g) \right)' \left(\log b(g) \right)'' \right)^{\tilde{m}-k} \left( \left(\log b(g) \right)' \left(\log a(g) \right)'' \right)^{k} \right| + \log \left(\tilde{m}+1\right) + \log^{+} C_{0,C}' \\
\leqslant
&\sum_{k=0}^{\tilde{m}} \log^{+} ( \chi_{0,C} \circ g_{[2]} ) \left| \left( \left(\log a(g) \right)' \left(\log b(g) \right)'' \right)^{\tilde{m}-k} \left( \left(\log b(g) \right)' \left(\log a(g) \right)'' \right)^{k} \right| \\
&+ \sum_{k=0}^{\tilde{m}} \log^{+} \binom{\tilde{m}}{k} + \log \left(\tilde{m}+1\right) + \log^{+} C_{0,C}' \\
\leqslant
&\sum_{k=0}^{\tilde{m}} \log^{+} \left| \left( \left(\log a(g) \right)' \left(\log b(g) \right)'' \right)^{\tilde{m}-k} \left( \left(\log b(g) \right)' \left(\log a(g) \right)'' \right)^{k} \right| \\
&+ \sum_{k=0}^{\tilde{m}} \log^{+} \binom{\tilde{m}}{k} + \log \left(\tilde{m}+1\right) + \log^{+} C_{0,C}' \\
\leqslant
&\sum_{k=0}^{\tilde{m}} \left( \log^{+} \left| \left(\log a(g) \right)' \right|^{\tilde{m}-k} + \log^{+} \left| \left(\log b(g) \right)'' \right|^{\tilde{m}-k} + \log^{+} \left| \left(\log b(g) \right)' \right|^{k} + \log^{+} \left| \left(\log a(g) \right)'' \right|^{k} \right) \\
&+ \sum_{k=0}^{\tilde{m}} \log^{+} \binom{\tilde{m}}{k} + \log \left(\tilde{m}+1\right) + \log^{+} C_{0,C}' .
\end{align*}
Therefore, we can apply Logarithmic Derivative Lemma to $a(g)$, $b(g)$, and then for $j=1,2$ and $0 \leqslant k \leqslant \tilde{m}$, we have
\begin{align*}
\int_{\partial \mathbb{D}_r} \log^{+} \left| \left(\log a(g) \right)^{(j)} \right|^{k} \frac{\mathrm{d} \, \theta}{2 \pi} = o\big(T_g(r)\big)\quad \parallel, \\
\int_{\partial \mathbb{D}_r} \log^{+} \left| \left(\log b(g) \right)^{(j)} \right|^{k} \frac{\mathrm{d} \, \theta}{2 \pi} = o\big(T_g(r)\big)\quad \parallel.
\end{align*}
Combining all previous estimates, we obtain (\ref{equ:key estimate in SMT for Z}) and finally finish the proof.
\section{Some Computational Details for the Key Vanishing Lemma I}\label{appendix:facts for m=3}
%\subsection*{Technical Lemma and Facts}
Let $f(x_1,x_2)'$ and $g(x_1,x_2)'$ be two $1$-jets on $\mathbb{C}^2$. With the coordinate $(x_1,x_2)$ of $\mathbb{C}^2$, we have
\begin{equation*}\label{equ:expression of f', g' by x_1'. x_2'}
\begin{pmatrix}
f(x_{1} ,x_{2} )'\\
g(x_{1} ,x_{2} )'
\end{pmatrix}
=
\begin{pmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial f}{\partial x_{2}}\\
\frac{\partial g}{\partial x_{1}} & \frac{\partial g}{\partial x_{2}}
\end{pmatrix}
\begin{pmatrix}
x_{1} '\\
x_{2} '
\end{pmatrix},
\end{equation*}
and
\begin{equation*}
\begin{pmatrix}
x_{1} '\\
x_{2} '
\end{pmatrix}
=
\begin{pmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial f}{\partial x_{2}}\\
\frac{\partial g}{\partial x_{1}} & \frac{\partial g}{\partial x_{2}}
\end{pmatrix}^{-1}
\begin{pmatrix}
f(x_{1} ,x_{2} )'\\
g(x_{1} ,x_{2} )'
\end{pmatrix}
=
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}^{-1}
\begin{pmatrix}
\frac{\partial g}{\partial x_{2}} & -\frac{\partial f}{\partial x_{2}}\\
-\frac{\partial g}{\partial x_{1}} & \frac{\partial f}{\partial x_{1}}
\end{pmatrix}
\begin{pmatrix}
f(x_{1} ,x_{2} )'\\
g(x_{1} ,x_{2} )'
\end{pmatrix}.
\end{equation*}
Let
\begin{equation*}
h(x_{1} ,x_{2} )'' \coloneqq \frac{\partial h}{\partial x_{1}} x_{1} ''+\frac{\partial h}{\partial x_{2}} x_{2} ''+\frac{\partial ^{2} h}{\partial x{_{1}}^{2}}( x_{1} ')^{2} +2\frac{\partial ^{2} h}{\partial x_{1} \partial x_{2}} x_{1} 'x_{2} '+\frac{\partial ^{2} h}{\partial x{_{2}}^{2}}( x_{2} ')^{2},
\end{equation*}
be the $2$-jet corresponding to the second order derivative of germ of the $1$-jet $h(x_{1} ,x_{2} )'$. Then by direct computation, we have
\begin{equation*}
\begin{vmatrix}
f(x_{1} ,x_{2} )' & g(x_{1} ,x_{2} )'\\
f(x_{1} ,x_{2} )'' & g(x_{1} ,x_{2} )''
\end{vmatrix}
=
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}
\begin{vmatrix}
x_{1} ' & x_{2} '\\
x_{1} '' & x_{2} ''
\end{vmatrix}
+
\begin{pmatrix}
( x_{1} ')^{3} & ( x_{1} ')^{2} x_{2} ' & x_{1} '( x_{2} ')^{2} & ( x_{2} ')^{3}
\end{pmatrix}
\begin{pmatrix}
\Delta ^{30}\\
\Delta ^{21}\\
\Delta ^{12}\\
\Delta ^{03}
\end{pmatrix}
\end{equation*}
where
\begin{align*}
\Delta ^{30}
&\coloneqq
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial ^{2} f}{\partial x{_{1}}^{2}} & \frac{\partial ^{2} g}{\partial x{_{1}}^{2}}
\end{vmatrix}
\\
\Delta ^{21}
&\coloneqq
2
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial ^{2} f}{\partial x_{1} \partial x_{2}} & \frac{\partial ^{2} g}{\partial x_{1} \partial x_{2}}
\end{vmatrix}
+
\begin{vmatrix}
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}\\
\frac{\partial ^{2} f}{\partial x{_{1}}^{2}} & \frac{\partial ^{2} g}{\partial x{_{1}}^{2}}
\end{vmatrix}
\\
\Delta ^{12}
&\coloneqq
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial ^{2} f}{\partial x{_{2}}^{2}} & \frac{\partial ^{2} g}{\partial x{_{2}}^{2}}
\end{vmatrix}
+
2
\begin{vmatrix}
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}\\
\frac{\partial ^{2} f}{\partial x_{1} \partial x_{2}} & \frac{\partial ^{2} g}{\partial x_{1} \partial x_{2}}
\end{vmatrix}
\\
\Delta ^{03}
&\coloneqq
\begin{vmatrix}
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}\\
\frac{\partial ^{2} f}{\partial x{_{2}}^{2}} & \frac{\partial ^{2} g}{\partial x{_{2}}^{2}}
\end{vmatrix}.
\end{align*}
Moreover, we also have
\begin{align*}
&
\begin{pmatrix}
( f(x_{1} ,x_{2} )')^{3} & ( f(x_{1} ,x_{2} )')^{2} g(x_{1} ,x_{2} )' & f(x_{1} ,x_{2} )'( g(x_{1} ,x_{2} )') & ( g(x_{1} ,x_{2} )')^{3}
\end{pmatrix} \\
=
&
\begin{pmatrix}
( x_{1} ')^{3} & ( x_{1} ')^{2} x_{2} ' & x_{1} '( x_{2} ')^{2} & ( x_{2} ')^{3}
\end{pmatrix}
\Psi,
\end{align*}
where
\begin{equation*}
\Psi
\coloneqq
\begin{pmatrix}
\left(\frac{\partial f}{\partial x_{1}}\right)^{3} & \left(\frac{\partial f}{\partial x_{1}}\right)^{2}\frac{\partial g}{\partial x_{1}} & \frac{\partial f}{\partial x_{1}}\left(\frac{\partial g}{\partial x_{1}}\right)^{2} & \left(\frac{\partial g}{\partial x_{1}}\right)^{3}\\
3\left(\frac{\partial f}{\partial x_{1}}\right)^{2}\left(\frac{\partial f}{\partial x_{2}}\right) & \frac{\partial f}{\partial x_{1}}\left(\frac{\partial f}{\partial x_{1}}\frac{\partial g}{\partial x_{2}} +2\frac{\partial f}{\partial x_{2}}\frac{\partial g}{\partial x_{1}}\right) & \frac{\partial g}{\partial x_{1}}\left(\frac{\partial g}{\partial x_{1}}\frac{\partial f}{\partial x_{2}} +2\frac{\partial g}{\partial x_{2}}\frac{\partial f}{\partial x_{1}}\right) & 3\left(\frac{\partial g}{\partial x_{1}}\right)^{2}\left(\frac{\partial g}{\partial x_{2}}\right)\\
3\left(\frac{\partial f}{\partial x_{1}}\right)\left(\frac{\partial f}{\partial x_{2}}\right)^{2} & \frac{\partial f}{\partial x_{2}}\left(\frac{\partial f}{\partial x_{2}}\frac{\partial g}{\partial x_{1}} +2\frac{\partial f}{\partial x_{1}}\frac{\partial g}{\partial x_{2}}\right) & \frac{\partial g}{\partial x_{2}}\left(\frac{\partial g}{\partial x_{2}}\frac{\partial f}{\partial x_{1}} +2\frac{\partial g}{\partial x_{1}}\frac{\partial f}{\partial x_{2}}\right) & 3\left(\frac{\partial g}{\partial x_{1}}\right)\left(\frac{\partial g}{\partial x_{2}}\right)^{2}\\
\left(\frac{\partial f}{\partial x_{2}}\right)^{3} & \left(\frac{\partial f}{\partial x_{2}}\right)^{2}\frac{\partial g}{\partial x_{2}} & \frac{\partial f}{\partial x_{2}}\left(\frac{\partial g}{\partial x_{2}}\right)^{2} & \left(\frac{\partial g}{\partial x_{2}}\right)^{3}
\end{pmatrix}.
\end{equation*}
Similarly, we can get
\small{
\begin{equation*}
\Psi ^{-1}
=
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}^{-3}
\begin{pmatrix}
\left(\frac{\partial g}{\partial x_{2}}\right)^{3} & -\left(\frac{\partial g}{\partial x_{2}}\right)^{2}\frac{\partial g}{\partial x_{1}} & \frac{\partial g}{\partial x_{2}}\left(\frac{\partial g}{\partial x_{1}}\right)^{2} & -\left(\frac{\partial g}{\partial x_{1}}\right)^{3}\\
-3\left(\frac{\partial g}{\partial x_{2}}\right)^{2}\left(\frac{\partial f}{\partial x_{2}}\right) & \frac{\partial g}{\partial x_{2}}\left(\frac{\partial f}{\partial x_{1}}\frac{\partial g}{\partial x_{2}} +2\frac{\partial f}{\partial x_{2}}\frac{\partial g}{\partial x_{1}}\right) & -\frac{\partial g}{\partial x_{1}}\left(\frac{\partial g}{\partial x_{1}}\frac{\partial f}{\partial x_{2}} +2\frac{\partial g}{\partial x_{2}}\frac{\partial f}{\partial x_{1}}\right) & 3\left(\frac{\partial g}{\partial x_{1}}\right)^{2}\left(\frac{\partial f}{\partial x_{1}}\right)\\
3\left(\frac{\partial g}{\partial x_{2}}\right)\left(\frac{\partial f}{\partial x_{2}}\right)^{2} & -\frac{\partial f}{\partial x_{2}}\left(\frac{\partial f}{\partial x_{2}}\frac{\partial g}{\partial x_{1}} +2\frac{\partial f}{\partial x_{1}}\frac{\partial g}{\partial x_{2}}\right) & \frac{\partial f}{\partial x_{1}}\left(\frac{\partial g}{\partial x_{2}}\frac{\partial f}{\partial x_{1}} +2\frac{\partial g}{\partial x_{1}}\frac{\partial f}{\partial x_{2}}\right) & -3\left(\frac{\partial g}{\partial x_{1}}\right)\left(\frac{\partial f}{\partial x_{1}}\right)^{2}\\
-\left(\frac{\partial f}{\partial x_{2}}\right)^{3} & \left(\frac{\partial f}{\partial x_{2}}\right)^{2}\frac{\partial f}{\partial x_{1}} & -\frac{\partial f}{\partial x_{2}}\left(\frac{\partial f}{\partial x_{1}}\right)^{2} & \left(\frac{\partial f}{\partial x_{1}}\right)^{3}
\end{pmatrix}.
\end{equation*}
}
\begin{lem}
\begin{equation*}
\Psi ^{-1}
\begin{pmatrix}
\Delta ^{30}\\
\Delta ^{21}\\
\Delta ^{12}\\
\Delta ^{03}
\end{pmatrix}
=
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}^{-2}
\begin{pmatrix}
\Theta ^{30}\\
\Theta ^{21}\\
\Theta ^{12}\\
\Theta ^{03}
\end{pmatrix}
\end{equation*}
where
\begin{align*}
\Theta ^{30}
&\coloneqq
\left(\frac{\partial g}{\partial x_{1}}\right)^{2}\frac{\partial ^{2} g}{\partial x{_{2}}^{2}} -2\frac{\partial g}{\partial x_{1}}\frac{\partial g}{\partial x_{2}}\frac{\partial ^{2} g}{\partial x_{1} \partial x_{2}} +\left(\frac{\partial g}{\partial x_{2}}\right)^{2}\frac{\partial ^{2} g}{\partial x{_{1}}^{2}} \\
\Theta ^{21}
&\coloneqq
-\left[ 2\frac{\partial f}{\partial x_{1}}
\begin{vmatrix}
\frac{\partial g}{\partial x_{1}} & \frac{\partial g}{\partial x_{2}}\\
\frac{\partial ^{2} g}{\partial x_{1} \partial x_{2}} & \frac{\partial ^{2} g}{\partial x{_{2}}^{2}}
\end{vmatrix}
-2\frac{\partial f}{\partial x_{2}}
\begin{vmatrix}
\frac{\partial g}{\partial x_{1}} & \frac{\partial g}{\partial x_{2}}\\
\frac{\partial ^{2} g}{\partial x{_{1}}^{2}} & \frac{\partial ^{2} g}{\partial x_{1} \partial x_{2}}
\end{vmatrix}
+
\left(\frac{\partial g}{\partial x_{1}}\right)^{2}\frac{\partial ^{2} f}{\partial x{_{2}}^{2}} -2\frac{\partial g}{\partial x_{1}}\frac{\partial g}{\partial x_{2}}\frac{\partial ^{2} f}{\partial x_{1} \partial x_{2}} +\left(\frac{\partial g}{\partial x_{2}}\right)^{2}\frac{\partial ^{2} f}{\partial x{_{1}}^{2}}\right] \\
\Theta ^{12}
&\coloneqq
2\frac{\partial g}{\partial x_{1}}
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial f}{\partial x_{2}}\\
\frac{\partial ^{2} f}{\partial x_{1} \partial x_{2}} & \frac{\partial ^{2} f}{\partial x{_{2}}^{2}}
\end{vmatrix}
-2\frac{\partial g}{\partial x_{2}}
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial f}{\partial x_{2}}\\
\frac{\partial ^{2} f}{\partial x{_{1}}^{2}} & \frac{\partial ^{2} f}{\partial x_{1} \partial x_{2}}
\end{vmatrix}
+
\left(\frac{\partial f}{\partial x_{1}}\right)^{2}\frac{\partial ^{2} g}{\partial x{_{2}}^{2}} -2\frac{\partial f}{\partial x_{1}}\frac{\partial f}{\partial x_{2}}\frac{\partial ^{2} g}{\partial x_{1} \partial x_{2}} +\left(\frac{\partial f}{\partial x_{2}}\right)^{2}\frac{\partial ^{2} g}{\partial x{_{1}}^{2}} \\
\Theta ^{03}
&\coloneqq
-\left[\left(\frac{\partial f}{\partial x_{1}}\right)^{2}\frac{\partial ^{2} f}{\partial x{_{2}}^{2}} -2\frac{\partial f}{\partial x_{1}}\frac{\partial f}{\partial x_{2}}\frac{\partial ^{2} f}{\partial x_{1} \partial x_{2}} +\left(\frac{\partial f}{\partial x_{2}}\right)^{2}\frac{\partial ^{2} f}{\partial x{_{1}}^{2}}\right].
\end{align*}
\end{lem}
\begin{proof}
direct computation, see Maple code
\end{proof}
Then, we have
\begin{align*}
\begin{vmatrix}
x_{1} ' & x_{2} '\\
x_{1} '' & x_{2} ''
\end{vmatrix}
&=
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}^{-1}
\left[
\begin{vmatrix}
f(x_{1} ,x_{2} )' & g(x_{1} ,x_{2} )'\\
f(x_{1} ,x_{2} )'' & g(x_{1} ,x_{2} )''
\end{vmatrix}
-
\begin{pmatrix}
( x_{1} ')^{3} & ( x_{1} ')^{2} x_{2} ' & x_{1} '( x_{2} ')^{2} & ( x_{2} ')^{3}
\end{pmatrix}
\begin{pmatrix}
\Delta ^{30}\\
\Delta ^{21}\\
\Delta ^{12}\\
\Delta ^{03}
\end{pmatrix}
\right]\\
&=
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}^{-1}
\begin{vmatrix}
f(x_{1} ,x_{2} )' & g(x_{1} ,x_{2} )'\\
f(x_{1} ,x_{2} )'' & g(x_{1} ,x_{2} )''
\end{vmatrix} \\
&-
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}^{-1}
\begin{pmatrix}
( f(x_{1} ,x_{2} )')^{3} & ( f(x_{1} ,x_{2} )')^{2} g(x_{1} ,x_{2} )' & f(x_{1} ,x_{2} )'( g(x_{1} ,x_{2} )') & ( g(x_{1} ,x_{2} )')^{3}
\end{pmatrix}
\Psi ^{-1}
\begin{pmatrix}
\Delta ^{30}\\
\Delta ^{21}\\
\Delta ^{12}\\
\Delta ^{03}
\end{pmatrix} \\
&=
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}^{-1}
\begin{vmatrix}
f(x_{1} ,x_{2} )' & g(x_{1} ,x_{2} )'\\
f(x_{1} ,x_{2} )'' & g(x_{1} ,x_{2} )''
\end{vmatrix} \\
&-
\begin{vmatrix}
\frac{\partial f}{\partial x_{1}} & \frac{\partial g}{\partial x_{1}}\\
\frac{\partial f}{\partial x_{2}} & \frac{\partial g}{\partial x_{2}}
\end{vmatrix}^{-3}
\begin{pmatrix}
( f(x_{1} ,x_{2} )')^{3} & ( f(x_{1} ,x_{2} )')^{2} g(x_{1} ,x_{2} )' & f(x_{1} ,x_{2} )'( g(x_{1} ,x_{2} )') & ( g(x_{1} ,x_{2} )')^{3}
\end{pmatrix}
\begin{pmatrix}
\Theta ^{30}\\
\Theta ^{21}\\
\Theta ^{12}\\
\Theta ^{03}
\end{pmatrix}.
\end{align*}
Combining the facts above with the following equation:
\begin{equation*}
\begin{vmatrix}
F_{1} -H_{1} & G_{1} -H_{1}\\
F_{2} -H_{2} & G_{2} -H_{2}
\end{vmatrix}
=
\begin{vmatrix}
1 & 1 & 1\\
F_{1} & G_{1} & H_{1}\\
F_{2} & G_{2} & H_{2}
\end{vmatrix},
\end{equation*}
we can get the formulas in Section \ref{section 7}.
\medskip
\section{Maple Code for the Key Vanishing Lemma II: Case $(m, t) = (7, 5)$}
\includepdf[pages={1-13},scale=0.95]{mod-p-E-2-m-inexistence.pdf}
\medskip
\end{document}