Positivity of line bundles on general blow-ups of abelian surfaces

Extracted source text · arXiv 1711.03764

\documentclass[a4paper,11pt]{amsart}
\usepackage{amsmath,amsfonts,amssymb,amsthm,enumerate}
\usepackage[left=2.5cm,right=2.5cm,top=3cm,bottom=3cm,a4paper]{geometry}
\usepackage[onehalfspacing]{setspace}
\usepackage{mathrsfs}
\usepackage[pdftex]{graphicx}
\usepackage[all]{xy}
\usepackage{tikz}
\usepackage{float}
\usepackage{tikz-3dplot}
%\usepackage{notcite, notref}

\usetikzlibrary{positioning}
\tikzset{>=stealth}

\theoremstyle{plain}
\newtheorem{thm}{Theorem}[section]
\newtheorem{lem}[thm]{Lemma}
\newtheorem{prop}[thm]{Proposition}
\newtheorem{cor}[thm]{Corollary}

\theoremstyle{definition}
\newtheorem{defn}[thm]{Definition}
\newtheorem{conj}[thm]{Conjecture} 
\newtheorem{exmp}[thm]{Example} 
\newtheorem{notation}[thm]{Notation}
\newtheorem{rmk}[thm]{Remark}
\newtheorem{note}[thm]{Note}
\newtheorem{ques}[thm]{Question}
\newtheorem{caution}[thm]{Caution}

\newcommand{\CC}{\mathbb{C}}
\newcommand{\RR}{\mathbb{R}}
\newcommand{\QQ}{\mathbb{Q}}
\newcommand{\ZZ}{\mathbb{Z}}
\newcommand{\NN}{\mathbb{N}}
\newcommand{\Ss}{\boldsymbol{S}}
\newcommand{\Pp}{\boldsymbol{P}}
\newcommand{\Kk}{\boldsymbol{K}}
\newcommand{\SO}{\boldsymbol{SO}}
\newcommand{\Oo}{\boldsymbol{O}}
\newcommand{\Uu}{\boldsymbol{U}}
\newcommand{\RP}{\mathbb{R}\boldsymbol{P}}
\newcommand{\CP}{\mathbb{C}\boldsymbol{P}}
\newcommand{\nz}{\backslash\{0\}}
\newcommand{\OO}{\mathcal{O}}
\newcommand{\FF}{\mathcal{F}}
\newcommand{\GG}{\mathcal{G}}
\newcommand{\HH}{\mathcal{H}}
\newcommand{\Ac}{\mathcal{A}}
\newcommand{\Ab}{\mathcal{A}b}
\newcommand{\II}{\mathcal{I}}
\newcommand{\Hom}{\mathrm{Hom}}
\newcommand{\Id}{\mathrm{Id}}
\newcommand{\colim}{\mathrm{colim}}
\newcommand{\limit}{\mathrm{lim}}
\newcommand{\x}{^\times}
\newcommand{\av}{\mathbf{e}}
\newcommand{\xv}{\mathbf{x}}
\newcommand{\Ga}{\Gamma}
\newcommand{\Gr}{\mathcal{G}}
\newcommand{\GL}{\mathrm{GL}}
\newcommand{\im}{\mathrm{im}}
\newcommand{\coker}{\mathrm{coker}}
\newcommand{\Coker}{\mathrm{Coker}}

%%%%% 추가 %%%%%%
\newcommand{\ceil}[1]{\lceil #1 \rceil}
\newcommand{\floor}[1]{\lfloor #1 \rfloor}
%%%%%%%%%%%%%%%%%%

\title{Positivity of line bundles on general blow-ups of abelian surfaces}
\author{Sanghyeon Lee and Jaesun Shin}
\date{}

\address{Department of Mathematical Sciences, Seoul National University, GwanAkRo 1, Seoul 08826, Korea}

\email{tlrehrdl@snu.ac.kr}

\address{Department of Mathematical Sciences, KAIST, 291 Daehak-ro, Yuseong-gu, Daejeon 305-701, Korea}

\email{jsshin1991@kaist.ac.kr}

\begin{document}
\maketitle

\begin{abstract}
Let $(S,L_{S})$ be a polarized abelian surface, and let $M = c \cdot \pi^*L_S - \alpha \cdot \sum_{i=1}^r E_i$ be a line bundle on ${\rm Bl}_{r}(S)$, where $\pi:{\rm Bl}_{r}(S) \rightarrow S$ is the blow-up of $S$ at $r$ general points with exceptional divisors $E_{1},\dots,E_{r}$. In this paper, we provide a criterion for $k$-very ampleness of $M$. Also, we deal with the case when $S$ is an arbitrary surface of Picard number one with a numerically trivial canonical divisor.  
\end{abstract}



\begin{section} {Introduction}

Let $(S,L_{S})$ be a complex polarized surface, and let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $r$ general points $p_{1},\dots,p_{r}$ with exceptional divisors $E_{1},\dots,E_{r}$. Line bundles of the form 
\begin{align*}
M = c \cdot \pi^*L_S - \alpha \cdot \sum_{i=1}^r E_i
\end{align*}
have been addressed by many authors with respect to different properties. Here we are interested in their positivity. The positivity of $M$ provides positivity data of $L_{S}$ along $r$ general points $p_{1},\dots,p_{r}$.

A classical way to study the positivity of $M$ is to check its global generation and its very ampleness. The notions of the global generation and the very ampleness were generalized in several ways. Here we focus on the $k$-very ampleness, which generalizes in a natural way two classical notions:

\begin{defn} \label{k-very ampleness}
A line bundle $L$ on a smooth projective variety $S$ is $k$-very ample if the restriction map $H^0(S,L) \to H^0(S,L|_{Z})$ is surjective for any $0$-dimensional subscheme $Z \subset S$ with length $k+1$.
\end{defn}

In fact, $L$ is $0$-very ample if and only if it is generated by global sections, and $L$ is $1$-very ample if and only if it is very ample. Geometrically, if $L$ is very ample and embeds $S$ in $\mathbb{P}^{h^{0}(L)-1}$, then $L$ is $k$ -very ample if and only if $S$ has no $(k+1)$-secant $(k-1)$ planes. 

Due to its importance, there are many studies about local positivity of line bundle $L$ on an algebraic surface $S$ of various types. When $S$ is an arbitrary smooth projective surface, K\"uchle \cite{K2} proved that lines bundles of the form $M=c\cdot \pi^* L - \alpha \cdot \sum_{i=1}^r E_i$ are ample when $c \geq 3$, $\alpha=1$ and $M^2 > 0$. Tutaj-Gasi\'nska \cite{T2} and Szemberg-Tutaj-Gasi\'nska \cite{ST2} carried out studies for the case where $S$ is a ruled surface. When $S=\mathbb{P}^2$, d'Almeida-Hirschowitz \cite{DH} obtained a complete answer for the very ampleness of a line bundle $c \cdot \pi^{*} \OO_{\mathbb{P}^{2}}(1)-\sum_{i=1}^{r}E_{i}$, and Hanumanthu (\cite{H}) studied the global generation, ampleness, and very ampleness of a line bundle $c \cdot \pi^{*} \OO_{\mathbb{P}^{2}}(1)-\sum_{i=1}^{r}m_{i}E_{i}$. Moreover, Szemberg-Gasi\'nska \cite{ST3} studied a criterion for $k$-very ampleness of the line bundle $M$. In general, for a projective space $\mathbb{P}^n$, d'Almeida-Hirschowski \cite{DH} and Coppens \cite{C} studied sufficient conditions for very ampleness of $M$. Finally, Tutaj-Gasi\'nska \cite{T} dealt with the $k$-very ampleness of $M$ when $S$ is an abelian surface of Picard number $1$, where $c=1$ and $\alpha=k$. Szemberg-Tutaj-Gasi\'nska \cite{ST1} worked on the very ampleness of $M$, where $c=1$ and $\alpha=1$, for an arbitrary Picard number. 

From now on, let $(S,L_{S})$ be a polarized abelian surface, and let $\rho(S)$ be the Picard number of $S$ unless otherwise specified. In this paper, we continue the investigations initiated in \cite{T} and \cite{ST1} to find a satisfactory result of positivity for $M$. To be more specific, our aim is to study $k$-very ampleness of $M$ in a fully generalized setting: arbitrary $c$, $\alpha \ge k$, and $\rho(S)$. Note that the condition $\alpha \ge k$ is necessary for the $k$-very ampleness of $c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r}E_{i}$ since $(c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r}E_{i}).E_{j}=\alpha$ for each $j=1,\dots,r$. In this direction, we prove the following results:

\begin{thm} (=Theorem \ref{thm:higher Picard number one}) \label{thm:introduction 1}
Let $(S,L_{S})$ be a polarized abelian surface of type $(1,d)$ with $\rho(S)=1$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $r$ general points with $1 \le r \le \frac{2d-(4\ceil{\frac{k}{c}}+5)}{(\ceil{\frac{\alpha}{c}}+1)^{2}}$ for integers $\alpha \ge k \ge 0$ and $c \in \mathbb{N}$. Then the line bundle $M_{c,\alpha}=c \cdot \pi^{*}L_{S}- \alpha \cdot \sum_{i=1}^{r}E_{i}$ is $k$-very ample. 
\end{thm}

\begin{thm} (=Theorem \ref{thm:any Picard number}) \label{thm:introduction 2}
Let $(S,L)$ be a polarized abelian surface of type $(1,d)$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $2 \le r < \frac{2d}{(\ceil{\frac{\alpha}{c}}+1)^{2}}-2$ general points.
\begin{enumerate}[(1)]
\item Assume that $S$ contains no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}} +1$ with respect to $L$. Then the line bundle $M_{c,\alpha}=c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is globally generated when $\alpha \ge 0$.
\item Assume that $S$ contains no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}} +2$ with respect to $L$. Then the line bundle $M_{c,\alpha}=c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is very ample when $\alpha \ge 1$. 
\end{enumerate}
\end{thm}

\begin{thm} (=Theorem \ref{thm:k-very ampleness on abelian surface}) \label{thm:introduction 3}
Let $(S,L)$ be a polarized abelian surface of type $(1,d)$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $2 \le r <\frac{2d}{(\ceil{\frac{\alpha}{c}}+\ceil{\frac{k}{c}}+1)^{2}}-2$ general points for integers $\alpha \ge k \ge 0$. Assume that $S$ contains no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}}+\ceil{\frac{k}{c}}+1$ with respect to $L$. Then $M=c\cdot\pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ is $k$-very ample. 
\end{thm}

In particular, when restricted to the cases of \cite{T} and \cite[Theorem 1]{ST1}, Theorem \ref{thm:introduction 1}, Theorem \ref{thm:introduction 2}, and Theorem \ref{thm:introduction 3} are improved relative to the previous results.

If $S$ is a simple abelian surface, then the condition for elliptic curves is not needed. On most abelian surfaces, however, the condition on the lower bound of degrees of elliptic curves may be a nuisance. Regarding this, we also give a criterion for $k=0,1$-very ampleness without the condition for elliptic curves (cf. Corollary \ref{cor:application 1} and Corollary \ref{cor:application 2}). 

In addition, we deal with the case when $S$ is an arbitrary surface of Picard number one with a numerically trivial canonical divisor: 

\begin{prop} (=Proposition \ref{prop:picard number 1 case}) \label{prop:introduction 4}
Let $S$ be a surface of $\rho(S)=1$ with a numerically trivial canonical divisor and $L$ the ample generator. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up at $2 \le r \le \frac{L^{2}}{(\ceil{\frac{\alpha}{c}}+2)^{2}}$ general points for integers $\alpha \ge k \ge 0$ and $c \in \mathbb{N}$. Then the line bundle $M=c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ is $k$-very ample. 
\end{prop}

%We remark that Reider-type criterion for $k$-very ampleness of a line bundle, which is proved by Beltrametti-Sommese \cite{BS} is heavily used to the proofs of above theorems.
%Also, the informations about Seshadri constant of Abelian varieties and general projective varieties are also heavily used. 

%Seshadri constant of Abelian surfaces are studied in many works, Bauer \cite{B}, Bauer-Szemberg \cite{BS1}, Nakamaye \cite{N} and Garc\'ia \cite{F}. 
%For more general projective variety, the work of Szemberg \cite{Sz} dealt with Seshadri constant of surfaces with picard number one, and K\"uchle \cite{K}, studied about Seshadri constant of an arbitrary smooth projective varieties.
%\\ \\
%Although we focus on the case when a surface $S$ is Abelian surface in this paper, we also remark about many studies about local positivity of line bundles $L$ on various smooth projective variety $S$.
%When $S$ is arbitrary smooth projective surface, K\"uchle \cite{K2} proved that lines bundle of the form $M=a\cdot \pi^* L - b \sum_{i=1}^r E_i$ is ample when $a \geq 3$, $b=1$ and $M^2 > 0$. 
%When $S$ is a ruled surface, about Seshadri constant, there were works of L. F. Garc\'ia \cite{F2} and Hanumanthu-Mukhopadhyay \cite{HM}. And about very ampleness and $k$-very ampleness of the line bundle $M$, there were works of Gasi\'nska \cite{T2} and Szemberg-Gasi\'nska \cite{ST2}. When $S$ is $\mathbb{P}^2$, D'Almeida-Hirschowitz \cite{DH} obtained a complete answer for the very ampleness of the line bundle $M$. Szemberg-Gasi\'nska \cite{ST3} studied about criterion for $k$-very ampleness of the line bundle $M$. When $S$ is a projective space $\mathbb{P}^n$, D'Almeida-Hirschowski \cite{DH} and Coppens \cite{C} studied about sufficient conditions for very ampleness of $M$.



%Let $S$ be an algebraic surface. In this paper, we deal with the case when $S$ is a polarized abelian surface $(S,L_S)$. We study about local positivity of the $L_S$. For any line bundle $L$ on $S$, there is an invariant called Seshadri constant which measures how positive the line bundle $L$ is near a smooth point $x \in S$, which is introduced by Demailly \cite{D}. Seshadri constant $\epsilon(L,x)$ is defined by :
%\[
%\epsilon(L,x) = sup\{\alpha \in \mathbb{R}_{\geq 0}| \pi^*L - \alpha \cdot E\textrm{ is nef}\}
%\]
%where $\pi$ is a blow-up morphism $\pi : \widetilde{S} \to S$ along $x\in S$ and $E$ is the exceptional divisor. As a generalization, we can also define a multi-point Seshadri constant $\epsilon(L,x_1,...,x_r)$ for $r$ smooth points :
%\[
%\epsilon(L,x) = sup\{\alpha \in \mathbb{R}_{\geq 0}| \pi^*L - \alpha \cdot \sum_{i=1}^r E_i\textrm{ is nef}\}
%\]

%On the other hand, if we know that a line bundle $\pi^*L - \alpha \sum_{i=1}^r E_i$ is very ample, then we can gives lower bound of the Seshadri constant, i.e. $\epsilon(L,x_1,...,x_r) \geq \alpha$. In this viewpoint, very ampleness of line bundles of the form $\pi^*L - \alpha \sum_{i=1}^r E_i$ also measures local positivity of $L$ near the point $x$.
 
%Return to the case that $S$ is a polarized abelian surface $(S, L_S)$. In Szemberg-Gasi\'nska \cite{ST1} and Gasi\'nska \cite{T}, the authors $k$-very ampleness of line bundles of the forms :
%\begin{equation}\label{line bundle}
%M = \pi^*L_S - k \sum_{i=1}^r E_i
%\end{equation}
%$k$-very ampleness is a generalization of globally generatedness and very ampleness, which are $0$ and $1$-very ampleness respectively. $k$-very ampleness is defined by the following :
%\begin{defn}\label{k-very ampleness}
%A line bundle $L$ on a smooth projective variety $S$ is $k$-very ample if the restriction map $H^0(S,L) \to H^0(S,L|_{Z})$ is surjective for any $0$-dimensional subscheme $Z \subset S$ with length $k+1$.
%\end{defn}

%In our paper, we study about $k$-ampleness of line bundles of more general form :
%\begin{equation}\label{line bundle 2}
%M=c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}
%\end{equation}
%where $c$ is a general positive integer and $\alpha \geq k$.
%\\
%When the Abelian surface $S$ has picard number one, then we have the following result.

%\begin{thm}\label{main 1}
%Let $S$ be a surface of $\rho(S)=1$ with numerically trivial canonical divisor and $L$ the ample generator. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up at $2 \le r \le \frac{L^{2}}{(\ceil{\frac{\alpha}{c}}+2)^{2}}$ general points for an integer $\alpha \ge k \ge 0$ and $c \in \mathbb{N}$. Then the line bundle $M=c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ is $k$-very ample on ${\rm Bl}_{r}(S)$. 
%\end{thm}

%We also studied about the case when the Abelian surface $S$ has arbitrary picard number with type $(1,d)$, with some restriction about lower bound of degree of elliptic curves in $S$ In our paper, we gives a criterion for $k=0,1$-very ampleness, i.e. globally generatedness and very ampleness. We note that condition about lower bound of degree of elliptic curves is necessary to determine lower bound of Seshadri constant, which is important ingredient in the proof of the next theorem :



%Especially when $c=1$, $\alpha=k$, $k=1$, the above theorem gives a stronger result than \cite[Theorem 1]{ST1}.


%Based on Theorem \ref{thm:introduction 2}, we present a question about a sufficient condition on $k$-very ampleness for arbitrary $k$:

%\begin{ques}
%Let $(S,L)$ be a polarized abelian surface of type $(1,d)$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow up of $S$ at $2 \le r < \frac{2d}{(\ceil{\frac{\alpha}{c}}+1)^{2}}-2$ general points for an integer $\alpha \ge k \ge 0$ and $c \in \mathbb{N}$. Assume that $S$ contains no elliptic curve of degree $\le \alpha+k+1$ with respect to $L$. Then the line bundle $c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ is $k$-very ample?
%\end{ques}

%Towards the above question, we give a partial answer:


The main tool used in our study is a Reider-type criterion \cite{BS} for $k$-very ampleness, which is proved by Beltrametti-Sommese, and the results for Seshadri constants of abelian surfaces and those of some projective varieties (\cite{B, BS1, F, K, N, Sz}). 

Concerning the organization of the paper, we begin in Section \ref{section:2} by fixing notation and collecting useful facts about Seshadri constants and a Reider-type criterion for $k$-very ampleness. In Section \ref{section:3}, we prove Theorem \ref{thm:introduction 1}. Using a similar method, we also give a criterion for $k$-very ampleness on general blow-ups of surfaces of Picard number $1$ with a numerically trivial canonical divisor (Proposition \ref{prop:introduction 4}). Finally, Section \ref{section:4} is devoted to the proofs of Theorem \ref{thm:introduction 2} and Theorem \ref{thm:introduction 3} as well as to their applications.  \\

\textbf{Acknowledgements.}
Research of the second author supported by NRF (National Research Foundation of Korea) Grant funded by the Korean Government (NRF-2016-Fostering Core Leaders of the Future Basic Science Program/Global Ph.D. Fellowship Program). 

%Seshadri constant of abelian surfaces are studied in many works, Bauer \cite{B}, Bauer-Szemberg \cite{BS1}, Nakamaye \cite{N}. Also the work of Szemberg \cite{Sz} which deals with Seshadri constant of surfaces with picard number one is also helpful for our study. 
%\\ \\
%Although we focus on the case when a surface $S$ is Abelian surface, we also note that there were many studies about local positivity of line bundles $L$ on various algebraic surface $S$. When $S$ is a ruled surface. About Seshadri constant, there were works of L. F. Garc\'ia \cite{F2} and Hanumanthu-Mukhopadhyay \cite{HM}. And about very ampleness and $k$-very ampleness of line bundles of the form $a\cdot \pi^* L - b \sum_{i=1}^r E_i$, there were works of Gasi\'nska \cite{T2} and Szemberg-Gasi\'nska \cite{ST2}. When $S$ is $\mathbb{P}^2$, D'Almeida-Hirschowitz \cite{DH} obtained a complete answer for the very ampleness of line bundles of the form $a\cdot \pi^* L - b \sum_{i=1}^r E_i$ when the blow-up locus $x_1,...,x_r$ are at general position.


\end{section}


\begin{section}{Preliminaries} \label{section:2}

\begin{subsection} {Reider type criterion}

For the discussion from now on, we recall the concept of $k$-very ampleness, which is a generalization of global generation and very ampleness:

\begin{defn}
Let $X$ be a smooth projective variety and $Z$ a $0$-dimensional subscheme of $X$ of length $k+1$ (,i.e. $h^{0}(\OO_{Z})=k+1$). A line bundle $L$ on $X$ is $k$-very ample at $X$ if the restriction map 
\begin{align*}
H^{0}(X,L) \rightarrow H^{0}(X,L \otimes \OO_{Z})
\end{align*}
is surjective. If $L$ is $k$-very ample at every $0$-dimensional subscheme of length $k+1$, we say $L$ is $k$-very ample on $X$. 
\end{defn}

It is immediate that a line bundle is $0$-very ample if and only if it is globally generated, and it is $1$-very ample if and only if it is very ample. \\

We also comment the Reider's celebrated result on global generation and very ampleness of line bundles (\cite{R}):

\begin{thm} \label{thm:Reider}
Let $X$ be a smooth projective surface, and let $N$ be a nef line bundle on $X$. 
\begin{enumerate}[(i)]
\item If $N^{2} \ge 5$, then either $K_{X}+N$ is globally generated or there exists an effective divisor $D$ on $X$ that satisfies one of the numerical conditions:
\begin{enumerate}[(1)]
\item $D.N=0$ and $D^{2}=-1$.
\item $D.N=1$ and $D^{2}=0$. 
\end{enumerate}
\item If $N^{2} \ge 10$, then either $K_{X}+N$ is very ample or there exists an effective divisor $D$ on $X$ that satisfies one of the numerical conditions:
\begin{enumerate}[(1)]
\item $D.N=0$ and $D^{2}=-1$ or $-2$. 
\item $D.N=1$ and $D^{2}=0$ or $-1$. 
\item $D.N=2$ and $D^{2}=0$.
\end{enumerate}
\end{enumerate}
\end{thm}

In the sequel, we will also make use of the following criterion proved by Beltrametti and Sommese (\cite{BS}). It is a Reider type criterion for higher order embeddings. 

\begin{thm} \label{thm:criterion}
Let $X$ be a smooth projective surface, and let $Z$ a $0$-dimensional subscheme of $X$ of length $k+1$. Let $L$ be a nef line bundle on $X$ such that $L^{2} \ge 4k+5$. Then either $K_{X}+L$ is $k$-very ample at $Z$ or there exists an effective divisor $D$ on $X$ satisfying the following properties:
\begin{enumerate}[(i)]
\item $L-2D$ is $\mathbb{Q}$-effective.
\item $D$ contains $Z$ and $K_{X}+L$ is not $k$-very ample at $Z$. 
\item $(L.D)-k-1 \le D^{2} < \frac{1}{2}(L.D)<k+1$.
\end{enumerate}
\end{thm}

\end{subsection}

\begin{subsection}{Seshadri constants}
The (multi-poiont) Seshadri constant of a nef and big line bundle $L$ at $x_{1}, \dots, x_{r}$ is the real number
\begin{align*}
\epsilon(L;x_{1},\dots,x_{r})=\sup\{a \ge 0 \text{ } | \text{ } \pi^{*}L-a \cdot \sum_{i=1}^{r}E_{i} \text{ is nef on $\tilde{X}$}\},
\end{align*}
where $\pi:\tilde{X} \rightarrow X$ is the blow-up of a smooth projective variety $X$ at $x_{1},\dots,x_{r} \in X$ (cf. \cite[Definition 1.5 and 1.8]{BDHKKSS}). Customary, for the Seshadri constant of a nef line bundle $L$ on $r$ very general points, we write briefly $\epsilon(L;r)$. We recall some crucial facts about Seshadri constants for later use. 

\begin{lem} (\cite{EKL}) \label{lem:general}
Let $X$ be a smooth projective variety, and let $N$ be a nef line bundle on $X$. Then for any $\delta \in \mathbb{R}_{>0}$, there exists a non-empty open subset $U_{\delta} \subseteq X$ such that $\epsilon(L;x_{1},\dots,x_{r}) \ge \epsilon(L;r)-\delta$ for any $\{x_{1},\dots,x_{r}\} \subseteq U_{\delta}$. 
\end{lem}

The following criterion is due to K\"uchle (\cite{K}). 

\begin{prop} \label{prop:multi lower bound}
Let $L$ be a nef and big line bundle on a smooth projective $n$-fold. For $r \ge 2$, we have
\begin{align*}
\epsilon(L;r) \ge \min\{\epsilon(L;1), \frac{\sqrt[n]{L^{n}}}{2}, \frac{\sqrt[n]{L^{n}(r-1)^{n-1}}}{r}\}.
\end{align*}
\end{prop}

Note that on abelian surface $S$, $\epsilon(L;1)=\epsilon(L,x)$ for any $x \in S$. From \cite{BS1}, the following lower bound can be obtained: 

\begin{prop} \label{prop:one lower bound}
Let $(S,L_{S})$ be a polarized abelian surface of type $(1,d)$. Then 
\begin{align*}
\epsilon(L;1) \ge \min\{\epsilon_{0},\frac{\sqrt{7}}{2} \sqrt{d}\},
\end{align*}
where $\epsilon_{0}$ is the minimal degree of an elliptic curve in $S$ with respect to $L_{S}$. 
\end{prop}

Moreover, if $(S, L_{S})$ is a polarized abelian surface of type $(1,d)$ with Picard number $1$, its Seshadri constant at $r$ general points is sufficiently close to its maximal value. 

\begin{prop} (\cite[Corollary 2.6]{F}) \label{prop:lower bound Picard number one}
Let $(S,L_{S})$ be a polarized abelian surface of type $(1,d)$ with $\rho(S)=1$, and let $x_{1},\dots,x_{r}$ be $r$ general points of $X$. Then:
\begin{enumerate}[(i)]
\item If $\sqrt{\frac{2d}{r}} \in \mathbb{Q}$, then $\epsilon(L_{S};x_{1},\dots,x_{r})=\sqrt{\frac{2d}{r}}$. 
\item If $\sqrt{\frac{2d}{r}} \notin \mathbb{Q}$, then $\epsilon(L_{S};x_{1},\dots,x_{r}) \ge 2d \cdot \frac{k_{0}}{l_{0}}=\sqrt{1-\frac{1}{l_{0}^{2}}} \sqrt{\frac{L_{S}^{2}}{r}}$, where $(l_{0},k_{0})$ is the primitive solution of Pell's equation $l^{2}-2rdk^{2}=1$. 
\end{enumerate}
\end{prop}

\end{subsection}

\end{section}



\begin{section}{Positivity on general blow-ups of surfaces with $\rho(S)=1$.} \label{section:3}

\begin{subsection}{Abelian surfaces with $\rho(S)=1$.}

Let $(S,L_{S})$ be a polarized abelian surface of type $(1,d)$ with $\rho(S)=1$, and let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $r$ general points $p_{1},\dots,p_{r}$. Our first result is about the $k$-very ampleness of $M_{c,\alpha}=c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r}E_{i}$. 

\begin{lem} \label{lem:0 Picard number one}
Let $(S,L_{S})$ be a polarized abelian surface of type $(1,d)$ with $\rho(S)=1$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $1 \le r \le \frac{2d-5}{{(\ceil{\frac{\alpha}{c}}+1)}^{2}}$ general points for $\alpha \in \mathbb{Z}_{\ge 0}$ and $c \in \mathbb{N}$. Then the line bundle $M_{c,\alpha}=c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is globally generated. 
\end{lem}

\begin{proof}
Let $N_{c,\alpha}=M_{c,\alpha}-K_{{\rm Bl}_{r}(S)}=c \cdot \pi^{*}L_{S}-(\alpha+1) \cdot \sum E_{i}$. First, we consider the case when $c=1$. Note that $N_{c,\alpha}^{2}=2d-r(\alpha+1)^{2} \ge 5$. Moreover, $N_{c,\alpha}$ is ample: in fact, $\epsilon(L_{S};r) \ge \sqrt{1-\frac{1}{l_{0}^{2}}} \sqrt{\frac{L_{S}^{2}}{r}}$, where $l_{0}$ is as in Proposition \ref{prop:lower bound Picard number one}-$(ii)$. Since $l_{0}^{2}=1+2rdk_{0}^{2} \ge 1+2rd$, 
\begin{align*}
\sqrt{1-\frac{1}{l_{0}^{2}}} \sqrt{\frac{L_{S}^{2}}{r}} \ge \sqrt{\frac{2rd}{1+2rd}}{\sqrt{\frac{2d}{r}}} =\frac{2d}{\sqrt{1+2rd}}.
\end{align*}
Since $\frac{2d}{\sqrt{1+2rd}}$ is a decreasing function on $r$, it is enough to check the inequality $\frac{2d}{\sqrt{1+2rd}}>\alpha+1$ for maximal $r=\frac{2d-5}{{(\alpha+1)}^{2}}$. By our assumption on $d$, it is easy to see that $\frac{2d}{\sqrt{(\alpha+1)^{2}+4d^{2}-10d}}>1$, that is, $\epsilon(L_{S};r)>\alpha +1$. Combining Lemma \ref{lem:general} with the fact that $\pi$ is the blow-up at $r$ general points, $N_{c,\alpha}$ is ample. \

Now, we are ready to use Theorem \ref{thm:Reider}. Suppose $M_{c,\alpha}$ is not globally generated. Then there exists an effective divisor $D$ on ${\rm Bl}_{r}(S)$ such that $D.N_{c,\alpha}=0$, $D^{2}=-1$, or $D.N_{c,\alpha}=1$, $D^{2}=0$. Since $N_{c,\alpha}$ is ample, there exists no effective divisor $D$ such that $D.N_{c,\alpha}=0$. So $D$ should satisfy $D.N_{c,\alpha}=1$ and $D^{2}=0$. Since $D^{2}=0$, $D$ should be of the form $\pi^{*}D_{S}-\sum a_{i}E_{i}$, where $D_{S}$ is a nontrivial effective divisor on $S$. Since $\rho(S)=1$, $D_{S}=aL_{S}$ for some $a \in \mathbb{N}$. The two numerical conditions give that $D^{2}=2a^{2}d-\sum a_{i}^{2}=0$ and $D.N_{c,\alpha}=2ad-(\alpha+1) \cdot \sum a_{i}=1$. So $(2ad-1)^{2}=(\alpha+1)^{2}(\sum a_{i})^{2} \le (\alpha+1)^{2} \cdot r \sum a_{i}^{2}$, that is, 
\begin{align*}
4a^{2}d^{2}-4ad \le (2ad-1)^{2} \le (\alpha+1)^{2}r \cdot 2a^{2}d.
\end{align*}
Dividing both sides by $2ad$, we obtain $2ad-2 \le r(\alpha+1)^{2} \cdot a$. Since $r \le \frac{2d-5}{{(\alpha+1)}^{2}}$, we have
\begin{align*}
2ad-2 \le r(\alpha+1)^{2} \cdot a \le (2d-5)a,
\end{align*}
i.e. $5a \le 2$, which is a contradiction. Thus $M_{c,\alpha}$ is globally generated. 

Next, we consider an arbitrary $c\in \mathbb{N}$. Let $\alpha = c\floor{\frac{\alpha}{c}} + \alpha'$, where $0\le \alpha' < c$. Then we have $\alpha = (c-\alpha')\floor{\frac{\alpha}{c}}+\alpha'\ceil{\frac{\alpha}{c}}$, so we can write
\begin{align*}
M_{c,\alpha}=(\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i})^{\otimes c-\alpha'}\otimes (\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i})^{\otimes \alpha'}. 
\end{align*}
By the proof of $c=1$ case, $\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ and $\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ are globally generated, i.e. $0$-very ample. Thus, by \cite[Theorem 1.1]{HTT}, $M_{c,\alpha}$ is globally generated. 
\end{proof}


Now, we can derive the following result on higher embeddings of a polarized abelian surface with Picard number one, which is a generalization of \cite[Theorem 2]{ST1} and \cite[Theorem 7]{T}. Note that the condition $\alpha \ge k$ is necessary for the $k$-very ampleness of $c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r}E_{i}$ since $(c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r}E_{i}).E_{j}=\alpha$ for each $j=1,\dots,r$. 

\begin{thm} \label{thm:higher Picard number one}
Let $(S,L_{S})$ be a polarized abelian surface of type $(1,d)$ with $\rho(S)=1$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $r$ general points such that $1 \le r \le \frac{2d-(4\ceil{\frac{k}{c}}+5)}{(\ceil{\frac{\alpha}{c}}+1)^{2}}$ for integers $\alpha \ge k \ge 0$ and $c \in \mathbb{N}$. Then the line bundle $M_{c,\alpha}=c \cdot \pi^{*}L_{S}- \alpha \cdot \sum_{i=1}^{r}E_{i}$ is $k$-very ample. 
\end{thm}

Our main tools are the Reider type criterion (Theorem \ref{thm:criterion}) and Proposition \ref{prop:lower bound Picard number one}. From now on, we write $M$ instead of $M_{c,\alpha}$ by fixing $c$ and $\alpha$ if there is no confusion. Let $L$ be a line bundle numerically equivalent to $M-K_{{\rm Bl}_{r}(S)}$, that is, $L \equiv c \cdot \pi^{*}L_{S}-(\alpha+1) \cdot \sum_{i=1}^{r}E_{i}$. 

\begin{lem} \label{lem:4k+5}
When $c=1$, $L$ is ample with $L^{2} \ge 4k+5$ for any $r$, $d$, $k$, and $\alpha$ in Theorem \ref{thm:higher Picard number one}. 
\end{lem}

\begin{proof}
It is obvious that $L^{2} \ge 4k+5$, so we focus on the ampleness of $L$. It is sufficient to show that $\epsilon(L_{S};r)>\alpha+1$. In fact, we claim that 
\begin{align*}
\epsilon(L_{S};r)>\frac{d(\alpha+1)}{d-k}. 
\end{align*}
Then applying Lemma \ref{lem:general}, we obtain the ampleness of $L$. By Proposition \ref{prop:lower bound Picard number one}, we need to check 
\begin{align*}
\sqrt{1-\frac{1}{l_{0}^{2}}} \sqrt{\frac{L_{S}^{2}}{r}} > \frac{d(\alpha+1)}{d-k},
\end{align*}
where $l_{0}$ is defined as in Proposition \ref{prop:lower bound Picard number one}-$(ii)$. Since $l_{0}^{2}=1+2rdk_{0}^{2} \ge 1+2rd$ and $\sqrt{1-\frac{1}{l_{0}^{2}}}$ is an increasing function for $l_{0}^{2}$, it is enough to check it for $l_{0}^{2}=1+2rd$. Since $L_{S}^{2}=2d$, it is equivalent to 
\begin{align*}
\frac{1}{\sqrt{1+2rd}}>\frac{\alpha+1}{2d-2k}.
\end{align*}
Finally, since $\frac{1}{\sqrt{1+2rd}}$ is a decreasing function for $r$, it is enough to check for maximal $r$, that is, 
\begin{align*}
\frac{1}{\sqrt{(\alpha+1)^{2}+2d(2d-4k-5)}}>\frac{1}{2d-2k}.
\end{align*}
Now, it is easy to check that this inequality holds since 
\begin{align*}
(4d^{2}-8dk+4k^{2})-(\alpha^{2}+2\alpha+1+4d^{2}-8kd-10d)=10d+4k^{2}-(\alpha+1)^{2}>0.
\end{align*}
\end{proof}

\begin{proof}[Proof of Theorem \ref{thm:higher Picard number one}]
For $k=0$, the conclusion is a special case of Lemma \ref{lem:0 Picard number one}, so we may assume that $k \ge 1$. We first deal with the case when $c=1$. By Lemma \ref{lem:4k+5}, it is enough to show the non-existence of an effective divisor $D$ on ${\rm Bl}_{r}(S)$ satisfying Theorem \ref{thm:criterion}-$(iii)$. If $D=\sum a_{i}E_{i}$ with $a_{i} \ge 0$, $L.D-(k+1) \le D^{2}$ implies $\alpha<k$, which is a contradiction. \

So set $D=\pi^{*}D_{S}-\sum m_{i}E_{i}$. In the proof of Lemma \ref{lem:4k+5}, we already proved that $\epsilon(L_{S};x_{1},\dots,x_{r})>\frac{d(\alpha+1)}{d-k}$ for general $x_{1},\dots,x_{r}$. Since $D$ is effective, 
\begin{align*}
(\pi^{*}L_{S}-\frac{d(\alpha+1)}{d-k} \cdot \sum E_{i}.D)>0 \text{, that is, } (\alpha+1) \cdot \sum m_{i}<\frac{(L_{S}.D_{S})(d-k)}{d}.
\end{align*}
Hence, we obtain $L.D=L_{S}.D_{S}-(\alpha+1) \cdot \sum m_{i} > (L_{S}.D_{S}) \cdot \frac{k}{d}$. Since $\rho(S)=1$, $D_{S}=aL_{S}$ for some $a \in \mathbb{N}$. So the above inequalities, combined with Theorem \ref{thm:criterion}-$(iii)$, give only one possibility:
\begin{enumerate}
\item[] $a=1$, i.e. $D_{S}=L_{S}$ and $L.D=2k+1$.
\end{enumerate}

Since $L.D=2k+1$, $D^{2}=k$ holds, so that $\sum m_{i}^{2}=2d-k$ and $\sum m_{i}=\frac{2d-2k-1}{\alpha+1}$. From $L.D<2k+2$, we obtain:
\begin{align*}
4d^{2}=(L_{S}.D_{S})^{2} &<4(k+1)^{2}+4(k+1)(\alpha+1)\sum m_{i}+(\alpha+1)^{2}(\sum m_{i})^{2} \\
& \le 4(k+1)^{2}+4(k+1)(2d-2k-1)+(\alpha+1)^{2}r \cdot \sum m_{i}^{2} \\
& \le 4(k+1)^{2}+4(k+1)(2d-2k-1)+(2d-4k-5)(2d-k) \\
&=4d^{2}-2d-2kd+k,
\end{align*}
i.e. $2d(k+1) < k$, which excludes the possibility. Hence we conclude the result for $c=1$. 

For an arbitrary $c\in \mathbb{N}$, let $\alpha = c\floor{\frac{\alpha}{c}} + \alpha'$, where $0\le \alpha' < c$. Since $\alpha = (c-\alpha')\floor{\frac{\alpha}{c}}+\alpha'\ceil{\frac{\alpha}{c}}$, we can write 
\begin{align*}
M_{c,\alpha}=(\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i})^{\otimes c-\alpha'}\otimes (\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i})^{\otimes \alpha'}.
\end{align*}
Next, let $k=c\floor{\frac{k}{c}}+k'$, that is, $k = (c-k')\floor{\frac{k}{c}}+k'\ceil{\frac{k}{c}}$. Since $\alpha \ge k$, either $\alpha' \ge k'$ or $\floor{\frac{\alpha}{c}} \ge \floor{\frac{k}{c}}+1$ holds.

If $\alpha'\ge k'$, then by the proof of $c=1$ case, $\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ is $\floor{\frac{k}{c}}$-very ample, and $\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ is $\ceil{\frac{k}{c}}$-very ample. Then by \cite[Theorem 1.1]{HTT}, we have that $M_{c,\alpha}$ is $((c-c')\floor{\frac{k}{c}}+c'\ceil{\frac{k}{c}})$-very ample. Since $\alpha' \ge k'$, we obtain $((c-\alpha')\floor{\frac{k}{c}}+\alpha'\ceil{\frac{k}{c}})\ge ((c-k')\floor{\frac{k}{c}}+k'\ceil{\frac{k}{c}})=k$. Therefore $M_{c,\alpha}$ is $k$-very ample.

If $\floor{\frac{\alpha}{c}} \ge \floor{\frac{k}{c}}+1\ge \ceil{\frac{k}{c}}$, then $\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ is $\floor{\frac{k}{c}}$ and $\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ are both $\ceil{\frac{k}{c}}$-ample. Therefore, again by \cite[Theorem 1.1]{HTT}, $M_{c,\alpha}$ is $c\ceil{\frac{k}{c}}$-very ample. Since $c\ceil{\frac{k}{c}}\ge k$, the conclusion follows. 
\end{proof}

\end{subsection}

\begin{subsection}{Surfaces of Picard number one with a numerically trivial canonical divisor.} 

The proof of Theorem \ref{thm:higher Picard number one} is also useful for analyzing higher embeddings of an arbitrary surface with Picard number one. As a special case, we observe the $k$-very ampleness of a polarized surface $(S,L)$ of $\rho(S)=1$ with the numerically trivial canonical divisor. 

\begin{prop}\label{prop:picard number 1 case}
Let $S$ be a surface of $\rho(S)=1$ with the numerically trivial canonical divisor, and let $L$ be the ample generator. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up at $2 \le r \le \frac{L^{2}}{(\ceil{\frac{\alpha}{c}}+2)^{2}}$ general points for integers $\alpha \ge k \ge 0$ and $c \in \mathbb{N}$. Then the line bundle $M=c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ is $k$-very ample. 
\end{prop}

\begin{proof}
%%%%%%%%%%%%%%%%%%
%Note that $K_{{\rm Bl}_{r}(S)} \equiv \sum E_{i}$ since $K_{S} \equiv \OO_{S}$. Let $L^{2}=l$ and $N=M-K_{{\rm Bl}_{r}(S)} \equiv c \cdot \pi^{*}L-(\alpha+1) \cdot \sum E_{i}$. We claim that $N$ is ample with $N^{2} \ge 4k+5$. Since $N^{2} \ge l-r(\alpha+1)^{2} \ge \frac{2\alpha+3}{(\alpha+2)^{2}}l \ge 4\alpha+6 >4k+5$, we focus on the proof of the ampleness of $N$. Again, for the ampleness, we may assume that $c=1$.
%%%%%%%%%%%%%%%%%%%%%%%%%%% 
First, we assume that $c=1$. Note that $K_{{\rm Bl}_{r}(S)} \equiv \sum E_{i}$ since $K_{S} \equiv \OO_{S}$. Let $L^{2}=l$ and $N=M-K_{{\rm Bl}_{r}(S)} \equiv c \cdot \pi^{*}L-(\alpha+1) \cdot \sum E_{i}$. We claim that $N$ is ample with $N^{2} \ge 4k+5$ since $N^{2} \ge l-r(\alpha+1)^{2} \ge \frac{2\alpha+3}{(\alpha+2)^{2}}l \ge 4\alpha+6 >4k+5$. We focus on the proof of the ampleness of $N$. 
For $r$ general points $x_{1},\dots,x_{r} \in S$, we prove the inequality 
\begin{align}
\epsilon(L;x_{1},\dots,x_{r}) > \frac{l(\alpha+1)}{l-k}.
\end{align}
For the proof, note that $\epsilon(L;r) \ge \lfloor \sqrt{\frac{l}{r}} \rfloor \ge \alpha+2$ by \cite[Theorem 3.2]{Sz}. So it is sufficient to show that $\alpha+2 > \frac{l(\alpha+1)}{l-k}$. It follows from the following inequality:
\begin{align*}
\alpha+2-\frac{l(\alpha+1)}{l-k}=\frac{l-k(\alpha+2)}{l-k} \ge \frac{2(\alpha+2)^{2}-k(\alpha+2)}{l-k} >0. 
\end{align*}
Then the inequality $(1)$ follows from Lemma \ref{lem:general} so that $N$ is ample. \

%%%%%%%%%%%%%%%%%%
%First, we fix $c=1$. 
%%%%%%%%%%%%%%%%%%%%%%%%
The argument in the proof of Lemma \ref{lem:0 Picard number one} gives the global generation of $M$, so we may assume that $k \ge 1$. Suppose that $M$ is not $k$-very ample. Then Theorem \ref{thm:criterion} implies that there exists an effective divisor $D$ on ${{\rm Bl}_{r}}(S)$ such that $N.D-k-1 \le D^{2}<\frac{N.D}{2} <k+1$. If $D=\sum a_{i}E_{i}$ with $a_{i} \ge 0$, we have $(\alpha+1) \sum a_{i}-k-1=N.D-k-1 \le D^{2}=-\sum a_{i}^{2}$, that is, $(\alpha+2)\sum a_{i} \le k+1$. Since $\alpha \ge k$, it is a contradiction. \

So we may assume that $D=\pi^{*}D_{S}-\sum a_{i}E_{i}$ with $D_{S}=aL$ for some $a \in \mathbb{N}$. So the inquality $(1)$ implies that 
\begin{align*}
(\pi^{*}L-\frac{l(\alpha+1)}{l-k} \cdot \sum E_{i}.D)=(L.D_{S})-\frac{l(\alpha+1)}{l-k} \sum a_{i}>0,
\end{align*}
i.e. $(\alpha+1)\sum a_{i}<\frac{(L.D_{S})(l-k)}{l}$. Then $2k+1 \ge (N.D)=(L.D_{S})-(\alpha+1)\sum a_{i}>\frac{k(L.D_{S})}{l}=ak$. So there are two possibilities:
\begin{enumerate} [(i)]
\item $a=1$, i.e. $D_{S}=L$ and $k<N.D < 2k+2$.
\item $a=2$, i.e. $D_{S}=2L$ and $N.D=2k+1$.
\end{enumerate}

For the case ${\rm (i)}$, consider the inquality $D^{2}=l-\sum a_{i}^{2}<\frac{1}{2}(N.D)=\frac{1}{2}(l-(\alpha+1)\sum a_{i})$ so that 
\begin{align*}
(\alpha+1) \sum a_{i}<-l+2\sum a_{i}^{2}.
\end{align*}
Since $k<(N.D)$, we have $-1 <(N.D)-k-1 \le D^{2}=l-\sum a_{i}^{2}$, i.e. $\sum a_{i}^{2} \le l$. Moreover, the inequality $l-(\alpha+1)\sum a_{i}=N.D<2k+2$ gives
\begin{align*}
l<2(k+1)+(\alpha+1)\sum a_{i}.
\end{align*}
By summing up the above inequalities, we have:
\begin{align*}
l^{2} &<4(k+1)^{2}+4(k+1)(\alpha+1)\sum a_{i}+(\alpha+1)^{2}(\sum a_{i})^{2} \\
&<4(k+1)^{2}+4(k+1)(-l+2 \sum a_{i}^{2})+(\alpha+1)^{2}r(\sum a_{i}^{2}) \\
&\le 4(k+1)^{2}+4(k+1)l+\frac{(\alpha+1)^{2}}{(\alpha+2)^{2}}l^{2}.
\end{align*}
So $(4\alpha+6)l \le \frac{2\alpha+3}{(\alpha+2)^{2}}l^{2}<4(k+1)^{2}+4(k+1)l$. It implies the following:
\begin{align*}
4(k+2)^{2} \le 4(\alpha+2)^{2} \le 2l \le (4\alpha-4k+2)l <4(k+1)^{2},
\end{align*}
which is a contradiction. Thus we exclude the case ${\rm (i)}$. \

For the case ${\rm (ii)}$, consider the inequalities $2l-(\alpha+1)\sum a_{i}=(N.D)=2k+1$ and $4l-\sum a_{i}^{2}=D^{2}=k$ so that $(\alpha+1)\sum a_{i}=2l-2k-1$ and $\sum a_{i}^{2}=4l-k$. Again, from $(N.D)<2k+2$, we have:
\begin{align*}
4l^{2} &<4(k+1)^{2}+4(k+1)(\alpha+1) \sum a_{i}+(\alpha+1)^{2}(\sum a_{i})^{2} \\
&\le 4(k+1)^{2}+4(k+1)(2l-2k-1)+(\alpha+1)^{2}(4l-k) \cdot \frac{l}{(\alpha+2)^{2}} \\
&=-4k^{2}-4k+8kl+8l-\frac{(\alpha+1)^{2}}{(\alpha+2)^{2}}kl+\frac{(\alpha+1)^{2}}{(\alpha+2)^{2}}4l^{2} \\
&<(8k+8)l+\frac{(\alpha+1)^{2}}{(\alpha+2)^{2}}4l^{2}.
\end{align*}
Thus we have $\frac{2\alpha+3}{(\alpha+2)^{2}}l<2k+2$, i.e. $4\alpha+6<2k+2$, which contradicts our assumption on $\alpha$. 

Finally, the similar argument in the proof of Theorem \ref{thm:higher Picard number one} yields the desired result for an arbitrary $c \in \mathbb{N}$, so we are done.  
\end{proof}

\end{subsection}

\end{section}


\begin{section}{Positivity on general blow-ups of abelian surfaces.} \label{section:4}

In this section, we show how \cite[Theorem 1]{ST1} can be extended to higher order embeddings of line bundles of the form $c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ for a non-negative integer $\alpha$ and $c \in \mathbb{N}$. 

\begin{thm} \label{thm:any Picard number}
Let $(S,L)$ be a polarized abelian surface of type $(1,d)$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $2 \le r < \frac{2d}{(\ceil{\frac{\alpha}{c}}+1)^{2}}-2$ general points. %for an integer $\alpha \ge 1$ and $c \in \mathbb{N}$.  
\begin{enumerate}[(1)]
\item Assume that $S$ contains no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}} +1$ with respect to $L$. Then the line bundle $M_{c,\alpha}=c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is globally generated when $\alpha \ge 0$.
\item Assume that $S$ contains no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}} +2$ with respect to $L$. Then the line bundle $M_{c,\alpha}=c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is very ample when $\alpha \ge 1$. 
\end{enumerate}
\end{thm}

\begin{proof}
%We claim that $L$ is globally generated. Note that the assumption $2 \le \frac{2d}{(\alpha+1)^{2}}-2$ gives that $L^{2}=2d \ge 6$. Moreover, since there is no elliptic curve $E$ on $S$ satisfying the inequality $(L.E)=1$, the global generation of $L$ can be obtained by \cite[Theorem 1.1]{T'}. So we may assume that $c=1$ (\cite{HTT}). 

$(1)$ We deal with the case $c=1$ first. If $\alpha$=0, $M_{c,\alpha}=\pi^*L_S$. Since we obtain $L^2=2d\ge 6$ from the asusmption $2 < \frac{2d}{(\alpha+1)^2}-2$, and there is no elliptic curve $E$ on $S$ satisfying the inequality $(L.E)=1$, the global generation of $L_S$ follows from \cite[Theorem 1.1]{T'}. Therefore $M_{c,\alpha}=\pi^*L_S$ is globally generated. So we may assume that $\alpha \ge 1$.
Let $N=M_{c,\alpha}-K_{X}=\pi^{*}L-(\alpha+1) \cdot \sum_{i=1}^{r}E_{i}$. Assume that $S$ contains no elliptic curve of degree $\le \alpha+1$ with respect to $L$. We claim that $\epsilon(L;r)>\alpha+1$. From Proposition \ref{prop:multi lower bound} and Proposition \ref{prop:one lower bound}, it is sufficient to check that $\epsilon_{0}>\alpha+1$ and $\frac{\sqrt{2d(r-1)}}{r}>\alpha+1$, where $\epsilon_{0}$ is defined as in Propositon \ref{prop:one lower bound}. The first one is obtained directly by the assumption on $S$, so we focus on the second one. However, it is also immediate since the assumption on $r$ implies that 
\begin{align*}
\frac{\sqrt{2d(r-1)}}{r} > \frac{(\alpha+1) \sqrt{(r-1)(r+2)}}{r} \ge \alpha+1.
\end{align*}
Thus we have $\epsilon(L;r)>\alpha+1$. By Lemma \ref{lem:general}, the generality condition gives that $N$ is ample on ${\rm Bl}_{r}(S)$. Moreover, we have $N^{2} =2d-r(\alpha+1)^{2} \ge 2(\alpha+1)^{2} \ge 5$. Applying Theorem \ref{thm:Reider}, we are done once we exclude the following two cases: there exists an effective divisor $D$ on ${\rm Bl}_{r}(S)$ such that
\begin{enumerate}[i)]
\item $D.N=0$, $D^{2}=-1$. 
\item $D.N=1$, $D^{2}=0$.
\end{enumerate}

The case ${\rm i)}$ is excluded by the ampleness of $N$, so consider the case ${\rm ii)}$. Since $D^{2}=0$, we may assume that $D=\pi^{*}D_{S}-\sum m_{i}E_{i}$, where $D_{S}$ is an effective divisor on $S$. By Hodge index theorem, 
\begin{align*}
2d \sum m_{i}^{2}=L^{2} \cdot D_{S}^{2} \le (L.D_{S})^{2}&=1+(\alpha+1)^{2}(\sum m_{i})^{2}+2(\alpha+1)\sum m_{i} \\
&\le 1+(2d-2(\alpha+1)^{2})\sum m_{i}^{2}+2(\alpha+1)\sum m_{i}^{2},
\end{align*}
i.e. $2(\alpha+1)\alpha\sum m_{i}^{2} \le 1$. This gives that $\sum m_{i}^{2}=0$. Hence $D_{S}$ is an elliptic curve of degree $1$ with respect to $L$, which contradicts our assumption on $S$. 

For an arbitrary $c\in \mathbb{N}$, write $\alpha = c\floor{\frac{\alpha}{c}} + \alpha'$ where $0\le \alpha' < c$. Then $\alpha = (c-\alpha')\floor{\frac{\alpha}{c}}+\alpha'\ceil{\frac{\alpha}{c}}$, so we can write 
\begin{align*}
M_{c,\alpha}=(\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i})^{\otimes c-\alpha'}\otimes (\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i})^{\otimes \alpha'}.
\end{align*}
By the proof of $c=1$ case, $\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ and $\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ are globally generated, i.e. $0$-very ample. Thus, by \cite[Theorem 1.1]{HTT}, we obtain that $M_{c,\alpha}$ is globally generated.

%This ends the proof of $(1)$. 

$(2)$ Next, assume that $S$ contains no elliptic curve of degree $\le \alpha+2$ with respect to $L$. As before, we deal with $c=1$ case first. For $\alpha=1$, it is just \cite[Theorem 1]{ST1}. So we may assume that $\alpha \ge 2$. Let $N$ be as above. Then the proof of $(1)$ shows that $N$ is ample with $N^{2}>2(\alpha+1)^{2} \ge 10$. Applying Reider's theorem (Theorem \ref{thm:Reider}), we need to exclude the existence of an effective divisor $D$ on ${\rm Bl}_{r}(S)$ satisfying one of the following cases:
\begin{enumerate}[i)]
\item $D.N=0$, $D^{2}=-1$ or $-2$. 
\item $D.N=1$, $D^{2}=0$ or $-1$. 
\item $D.N=2$, $D^{2}=0$.
\end{enumerate}

By the ampleness of $N$, the case ${\rm i)}$ can be excluded. First, we claim that any such $D$, if exists, cannot contain some of the exceptional divisors. Suppose that $D=D'+aE$, where $a \ge 1$, $E=E_{i}$ for some $i$, and $D'$ does not contain $E$. Then 
\begin{align*}
2 \ge D.N=D'.N+a(\alpha+1),
\end{align*}
which is a contradiction. Thus we may assume that $D=\pi^{*}D_{S}-\sum m_{i}E_{i}$, where $D_{S}$ is an effective divisor on $S$ and $m_{i} \ge 0$ for all $i$. \

The argument in the proof of $(1)$ eliminates the existence of $D$ satisfying $D.N=1$ and $D^{2}=0$, so it remains to deal with the last two cases. First, suppose the case when $D.N=1$ and $D^{2}=-1$. It is immediate that 
\begin{align*}
(L.D_{S})=1+(\alpha+1)\sum m_{i} \text{ and } D_{S}^{2}=-1+\sum m_{i}^{2}.
\end{align*}
By Hodge index theorem, we have
\begin{align*}
2d(-1+\sum m_{i}^{2})=L^{2} \cdot D_{S}^{2} \le (L.D_{S})^{2} &= 1+2(\alpha+1) \sum m_{i}+(\alpha+1)^{2}(\sum m_{i})^{2} \\
&< 1+2(\alpha+1) \sum m_{i}^{2}+(2d-2(\alpha+1)^{2})\sum m_{i}^{2},
\end{align*}
that is, $2\alpha(\alpha+1) \sum m_{i}^{2}<2d+1$. Now, we claim that $\sum m_{i}^{2} \ge 5$. Since $D_{S}$ is an effective divisor on $S$, $D_{S}^{2} \ge 0$, i.e. $\sum m_{i}^{2} \ge 1$. Moreover, by adjunction formula, $D_{S}^{2}$ should be an even number. So if $\sum m_{i}^{2}<5$, it should be either $1$ or $3$. If $\sum m_{i}^{2}=1$, then $D_{S}$ is an elliptic curve of degree $\alpha+2$ with respect to $L$, which contradicts our assumption on $S$. Next, if $\sum m_{i}^{2}=3$, $D_{S}^{2}=2$. Then $D_{S}$ is a principal polarization(,i.e. of type $(1,1)$) passing through $3$ general points in $S$. However, three points on a principal polarization are not general so that $\sum m_{i}^{2} \ge 5$. \

Now, we claim that $L-(\alpha+1)D_{S}$ is ample. By \cite[Lemma 4.3.2]{BL}, it is sufficient to show that $(L-(\alpha+1)D_{S}.L)>0$ and $(L-(\alpha+1)D_{S})^{2}>0$. Note that 
\begin{align*}
(L-(\alpha+1)D_{S}.L)&=2d-(\alpha+1)(1+(\alpha+1)\sum m_{i}) \\
&>2\alpha(\alpha+1) \sum m_{i}^{2}-1-(\alpha+1)^{2}\sum m_{i}-(\alpha+1) \\
&\ge (\alpha^{2}-1) \sum m_{i}^{2}-(\alpha+2) \\
&>0.
\end{align*}
For the inequality $(L-(\alpha+1)D_{S})^{2}>0$, we claim that either $\sum m_{i}^{2}>5$ or $\sum m_{i}^{2}>\sum m_{i}$ holds. Suppose that $\sum m_{i}^{2} \le 5$. Since $\sum m_{i}^{2} \ge 5$, it means that $\sum m_{i}^{2}=5$. Then there are two possibilities; first one is $m_{1}=2$, $m_{2}=1$, $m_{3}=\dots=m_{r}=0$, and the second one is $m_{1}=\dots=m_{5}=1$, $m_{6}=\dots=m_{r}=0$ by renumbering if necessary. The first one implies $\sum m_{i}^{2}> \sum m_{i}$, so we need to exclude the second one. In this case, since $D_{S}^{2}=4$, it is a polarization of type $(1,2)$ passing through $5$ general points. However, it is impossible by the generality assumption, which proves the claim. Now, consider the following inequalities:
\begin{align*}
(L-(\alpha+1)D_{S})^{2}&=2d+(\alpha+1)^{2}(-1+\sum m_{i}^{2})-2(\alpha+1)(1+(\alpha+1)\sum m_{i}) \\
&>2\alpha(\alpha+1)\sum m_{i}^{2}-1+(\alpha+1)^{2}\sum m_{i}^{2}-2(\alpha+1)^{2} \sum m_{i}-\alpha^{2}-4\alpha -3 \\
&\ge (\alpha^{2}-1) \sum m_{i}^{2}-\alpha^{2}-4\alpha-4 \\
&\ge 5(\alpha^{2}-1)-\alpha^{2}-4\alpha-4 \\
&=4\alpha^{2}-4\alpha-9 \\
&\ge -1. 
\end{align*}
However, the above claim implies that at least one of the second and the third inequalities is strict. Hence $(L-(\alpha+1)D_{S})^{2}>0$, i.e. $L-(\alpha+1)D_{S}$ is ample. 

So we have the following inequalities:
\begin{align*}
0<(L-(\alpha+1)D_{S}.D_{S})&=1+(\alpha+1)\sum m_{i}-(\alpha+1)(-1+\sum m_{i}^{2}) \\
&=-(\alpha+1) \sum m_{i}^{2}+(\alpha+1)\sum m_{i}+\alpha+2 \\
&=(\alpha+2)+(\alpha+1)(\sum m_{i}-\sum m_{i}^{2}).
\end{align*}
Thus all $m_{i}$ are either $0$ or $1$. Let $\beta$ be the number of $1$'s. Then we have 
\begin{align*}
2\alpha(\alpha+1) \beta < 2d+1 \text{, } L.D_{S}=1+(\alpha+1)\beta \text{, and } D_{S}^{2}=-1+\beta.
\end{align*}
Again, Hodge index theorem gives 
\begin{align*}
(\beta-1)((2\alpha^{2}+2\alpha)\beta-1)<L^{2} \cdot D_{S}^{2} \le (D_{S}.L)^{2}=(1+(\alpha+1)\beta)^{2},
\end{align*}
that is, $(\alpha^{2}-1)\beta<2\alpha^{2}+4\alpha+3$. Since $m_{i}$'s are either $0$ or $1$, $\beta=\sum m_{i}^{2} \ge 7$. So $7(\alpha^{2}-1) \le \beta(\alpha^{2}-1)<2\alpha^{2}+4\alpha+3$, i.e. $5\alpha^{2}-4\alpha-10<0$, which is a contradiction since $\alpha \ge 2$. Therefore, we can exclude the case when $D.N=1$ and $D^{2}=-1$. \

Finally, suppose that $D.N=2$ and $D^{2}=0$. Then the Hodge index theorem gives that
\begin{align*}
2d\sum m_{i}^{2} \le (2+(\alpha+1)\sum m_{i})^{2}&=4+4(\alpha+1)\sum m_{i}+(\alpha+1)^{2}(\sum m_{i})^{2} \\
&<4+4(\alpha+1) \sum m_{i}^{2}+(2d-2(\alpha+1)^{2})\sum m_{i}^{2},
\end{align*}
i.e. $(\alpha^{2}-1) \cdot \sum m_{i}^{2}<2$. Since $\alpha \ge 2$, we have $\sum m_{i}^{2}=0$. Then $D_{S}$ is an elliptic curve of degree $2$ with respect to $L$, which is a contradiction. So $\pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ is very ample as desired. 

For an arbitrary $c\in \mathbb{N}$, write $\alpha = c\floor{\frac{\alpha}{c}} + \alpha'$, where $0\le \alpha' < c$. If $\alpha \ge c$, then $\floor{\frac{\alpha}{c}}\ge 1$.
Thus $\alpha = (c-\alpha')\floor{\frac{\alpha}{c}}+\alpha'\ceil{\frac{\alpha}{c}}$, so we can write 
\begin{align*}
M_{c,\alpha}=(\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i})^{\otimes c-\alpha'}\otimes (\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i})^{\otimes \alpha'}.
\end{align*}
By the proof of $c=1$ case, $\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ and $\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ are very ample, i.e. $1$-very ample. Thus, by \cite[Theorem 1.1]{HTT}, $M_{c,\alpha}$ is $c$-very ample, therefore it is very ample.

If $1 \le \alpha < c$, then $\alpha'>0$. In this case, $\pi^*L_S-\floor{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ is at least globally generated, i.e. $0$-very ample, and $\pi^*L_S-\ceil{\frac{\alpha}{c}}\cdot\sum_{i=1}^{r}E_{i}$ is very ample, i.e. $1$-very ample. Therefore, by \cite[Theorem 1.1]{HTT}, $M_{c,\alpha}$ is $\alpha'$-very ample so that it is very ample.
\end{proof}

If $S$ is a simple abelian surface, then the condition for elliptic curves is not needed. On most abelian surfaces, however, this may be a nuisance. Regarding this, the following corollary tells us that the condition is redundant if we choose a large $d$ in the case of a very general abelian surface (in the sense of \cite[Theorem 1.(b)]{B}). 

\begin{cor} \label{cor:application 1}
Let $(S,L)$ be a very general polarized abelian surface of type $(1,d)$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $2 \le r < \frac{2d}{(\ceil{\frac{\alpha}{c}}+1)^{2}}-2$ general points.  
\begin{enumerate}[(1)] 
\item Assume that $d>\frac{81}{16}(\ceil{\frac{\alpha}{c}}+1)^{2}$. Then the line bundle $c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is globally generated when $\alpha \ge 0$.
\item Assume that $d>\frac{81}{16}(\ceil{\frac{\alpha}{c}}+2)^{2}$. Then the line bundle $c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is very ample when $\alpha \ge 1$. 
\end{enumerate}
\end{cor}

\begin{proof}
When $\alpha=0$, the global generation of a given line bundle is clear by \cite[Theorem 1.1]{T'}. So we may assume that $\alpha \ge 1$. By \cite[Theorem 4.5]{S}, the very generality of $(S,L)$ and the inequality $d>\frac{81}{16}(\ceil{\frac{\alpha}{c}}+1)^{2}$ give the $N_{\ceil{\frac{\alpha}{c}}-1}$ property of $L$. It implies that $S$ contains no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}}+1$ with respect to $L$ by \cite[Theorem 1.1]{KL1509}. Now the first statement is a direct consequence of Theorem \ref{thm:any Picard number}-(1). \

The second statement can be proven similarly. 
\end{proof}

On a polarized abelian surface $(S,L)$, there are many results about the lower bound of its Seshadri constant (e.g., \cite{B, BDHKKSS, BS1, L, N}). So it is sometimes easier to deal with the Seshadri constant than the degree of an elliptic curve on $(S,L)$. The following result says that we can change from the condition for elliptic curves in Theorem \ref{thm:any Picard number} to that for the Seshadri constant. 

\begin{cor} \label{cor:application 2}
Let $(S,L)$ be a polarized abelian surface of type $(1,d)$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $2 \le r < \frac{2d}{(\ceil{\frac{\alpha}{c}}+1)^{2}}-2$ general points. 
\begin{enumerate}[(1)]
\item Assume that $\epsilon(L;1)>\frac{5-\sqrt{5}}{2}(\ceil{\frac{\alpha}{c}}+1)$ and $d \ge \frac{5}{2}(\ceil{\frac{\alpha}{c}}+1)^{2}$. Then the line bundle $c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is globally generated when $\alpha \ge 0$.
\item Assume that $\epsilon(L;1)>\frac{5-\sqrt{5}}{2}(\ceil{\frac{\alpha}{c}}+2)$ and $d \ge \frac{5}{2}(\ceil{\frac{\alpha}{c}}+2)^{2}$. Then the line bundle $c \cdot \pi^{*}L_{S}-\alpha \cdot \sum_{i=1}^{r} E_{i}$ is very ample when $\alpha \ge 1$. 
\end{enumerate}
\end{cor}

\begin{proof}
Since the proofs of $(1)$ and $(2)$ are similar, we only give the proof of $(2)$. By Theorem \ref{thm:any Picard number}, it is sufficient to show that there exists no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}}+2$ with respect to $L$. Suppose not. Since $L^{2}=2d \ge 5(\ceil{\frac{\alpha}{c}}+2)^{2}$, $L$ does not satisfy property $N_{\ceil{\frac{\alpha}{c}}}$ by \cite[Theorem 1.1]{KL1509}. By \cite[Theorem 1.1]{S}, we obtain:
\begin{align*}
2d \cdot (\epsilon(L;1)-\ceil{\frac{\alpha}{c}}-2)-(\ceil{\frac{\alpha}{c}}+2) \cdot \epsilon(L;1)^{2} \le 0.
\end{align*}
Note that this inequality is equivalent to the following one:
\begin{align*}
\epsilon(L;1)^{2}-(\frac{2d}{\ceil{\frac{\alpha}{c}}+2}) \cdot \epsilon(L;1)+2d \ge 0.
\end{align*}
It implies that either $\epsilon(L;1) \le \frac{d-\sqrt{d^{2}-2d(\ceil{\frac{\alpha}{c}}+2)^{2}}}{\ceil{\frac{\alpha}{c}}+2}$ or $\epsilon(L;1) \ge \frac{d+\sqrt{d^{2}-2d(\ceil{\frac{\alpha}{c}}+2)^{2}}}{\ceil{\frac{\alpha}{c}}+2}$ holds. \

Now, we claim that $\sqrt{2d}<\frac{d+\sqrt{d^{2}-2d(\ceil{\frac{\alpha}{c}}+2)^{2}}}{\ceil{\frac{\alpha}{c}}+2}$. For this claim, it is sufficient to show that 
\begin{align*}
\sqrt{2} \cdot (\ceil{\frac{\alpha}{c}}+2) < \sqrt{d}+\sqrt{d-2(\ceil{\frac{\alpha}{c}}+2)^{2}},
\end{align*}
which is clear by our assumption on $d$. However, since $\epsilon(L;1) \le \sqrt{2d}$, the inequality $\epsilon(L;1) \ge \frac{d+\sqrt{d^{2}-2d(\ceil{\frac{\alpha}{c}}+2)^{2}}}{\ceil{\frac{\alpha}{c}}+2}$ can be excluded. Moreover, it is easy to see that 
\begin{align*}
\frac{d-\sqrt{d^{2}-2d(\ceil{\frac{\alpha}{c}}+2)^{2}}}{\ceil{\frac{\alpha}{c}}+2} \le \frac{5-\sqrt{5}}{2}(\ceil{\frac{\alpha}{c}}+2)
\end{align*}
since $d \ge \frac{5}{2}(\ceil{\frac{\alpha}{c}}+2)^{2}$. However, since $\epsilon(L;1) >\frac{5-\sqrt{5}}{2}(\ceil{\frac{\alpha}{c}}+2)$, it is a contradiction. Hence, there exists no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}}+2$ with respect to $L$, so the conclusion follows. 
\end{proof}

From Theorem \ref{thm:any Picard number}, we present the following question:

\begin{ques} \label{ques:abelian surface with any Picard number}
Let $(S,L)$ be a polarized abelian surface of type $(1,d)$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $2 \le r < \frac{2d}{(\ceil{\frac{\alpha}{c}}+1)^{2}}-2$ general points for integers $\alpha \ge k \ge 0$ and $c \in \mathbb{N}$. Assume that $S$ contains no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}}+\ceil{\frac{k}{c}}+1$ with respect to $L$. Then is the line bundle $c \cdot \pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ $k$-very ample?
\end{ques}

As a result of our effort, we provide a partial answer to Question \ref{ques:abelian surface with any Picard number}.

\begin{thm} \label{thm:k-very ampleness on abelian surface}
Let $(S,L)$ be a polarized abelian surface of type $(1,d)$. Let $\pi:{\rm Bl}_{r}(S) \rightarrow S$ be the blow-up of $S$ at $2 \le r <\frac{2d}{(\ceil{\frac{\alpha}{c}}+\ceil{\frac{k}{c}}+1)^{2}}-2$ general points for integers $\alpha \ge k \ge 0$. Assume that $S$ contains no elliptic curve of degree $\le \ceil{\frac{\alpha}{c}}+\ceil{\frac{k}{c}}+1$ with respect to $L$. Then $M=c\cdot\pi^{*}L-\alpha \cdot \sum_{i=1}^{r}E_{i}$ is $k$-very ample. 
\end{thm}

\begin{lem}\label{key lemma}
Let $N \equiv \pi^{*}L-(\alpha+1) \sum_{i=1}^{r}E_{i}$. Then there exists no effective divisor $D$ on ${\rm Bl}_{r}(S)$ satisfying the inequality:
\begin{align*}
(N.D)-k-1 \le D^{2}<\frac{1}{2}(N.D)<k+1.
\end{align*}
\end{lem}

\begin{proof}
Suppose not. Note that $N$ is ample with $N^{2} > 4k+5$ by the proof of Theorem \ref{thm:any Picard number}. If $D=D'+aE$ such that $a \ge 1$, $E=E_{i}$ for some $i$, and $D'$ does not contain $E$, then 
\begin{align*}
2k+1 \ge D.N=D'.N+a(\alpha+1) \ge D'.N+a(k+1).
\end{align*}
So we may assume that $a=1$. In this case, 
\begin{align*}
D^{2} \ge N.D-k-1 \ge D'.N+k+1-k-1 \ge 1
\end{align*}
so that $D^{2}=D'^{2}-1 \ge 1$, that is, $D'^{2} \ge 2$. By Hodge index theorem, 
\begin{align*}
2k+1 \ge D'.N+k+1 \ge \sqrt{D'^{2}} \cdot \sqrt{N^{2}}+k+1>\sqrt{2} \cdot \sqrt{2}(k+1)+k+1=3k+3,
\end{align*}
which is a contradiction. Thus we may assume that $D=\pi^{*}D_{S}-\sum m_{i}E_{i}$, where $D_{S}$ is an effective divisor on $S$ and $m_{i} \ge 0$ for all $i$. If $D^{2} \ge 1$, again by Hodge index theorem, 
\begin{align*}
(2k+1)^{2} \ge (D.N)^{2} \ge D^{2} \cdot N^{2} \ge 2d-r(\alpha+1)^{2} &\ge (r+2)(\alpha+k+1)^{2}-r(\alpha+1)^{2} \\
&=rk(2\alpha+2+k)+2(\alpha+k+1)^{2},
\end{align*}
which is a contradiction. So $D^{2} \le 0$. 

Now, we use the induction on $k$. For $k=0, 1$, we are done by Theorem \ref{thm:any Picard number}. Suppose it holds for $k-1 \ge 1$. If our $D$ satisfies $N.D-k \le D^{2}$, then the induction hypothesis implies that $N.D \ge 2k$. However, it means $k \le N.D-k \le D^{2} \le 0$, which is a contradiction. So $D^{2}=N.D-k-1$ should hold. 

Moreover, since $\alpha+k+1=(\alpha+1)+(k-1)+1$, the induction hypothesis gives the $(k-1)$-very ampleness of $N \equiv \pi^{*}L-(\alpha+1) \cdot \sum E_{i}$. Since $D$ is effective, $N.D \ge k-1$. Combining it with $D^{2} \le 0$, there exist only three possible cases:
\begin{enumerate}[i)]
\item $D^{2}=-2$ and $N.D=k-1$. 
\item $D^{2}=-1$ and $N.D=k$.
\item $D^{2}=0$ and $N.D=k+1$. 
\end{enumerate}

We deal with the case iii) first. Note that $D_{S}^{2}=\sum m_{i}^{2}$ and $L.D_{S}=k+1+(\alpha+1) \cdot \sum m_{i}$. By Hodge index theorem, 
\begin{align*}
2d(\sum m_{i}^{2}) \le (L.D_{S})^{2}&=(\alpha+1)^{2}(\sum m_{i})^{2}+2(k+1)(\alpha+1) \cdot \sum m_{i}+(k+1)^{2} \\
&<(k+1)^{2}+2(k+1)(\alpha+1)\sum m_{i}^{2}+(2d-(2\alpha+k+1)^{2})\sum m_{i}^{2}. 
\end{align*}
Hence $(2(\alpha+k+1)^{2}-2(k+1)(\alpha+1))\sum m_{i}^{2}<(k+1)^{2}$ holds. However, since $\sum m_{i}^{2} \ge 1$, this is impossible. 

Next, consider the case ii). Note that $D_{S}^{2}=\sum m_{i}^{2}-1$ and $L.D_{S}=k+(\alpha+1) \cdot \sum m_{i}$. Then Hodge index theorem gives 
\begin{align*}
2d(-1+\sum m_{i}^{2}) \le (L.D_{S})^{2}&=k^{2}+2k(\alpha+1)\sum m_{i}+(\alpha+1)^{2}(\sum m_{i})^{2} \\
&< k^{2}+2k(\alpha+1) \sum m_{i} +(2d-2(\alpha+k+1)^{2})\sum m_{i}^{2}. 
\end{align*}
So we have $2d > 2(\alpha+k+1)^{2} \sum m_{i}^{2}-2k(\alpha+1) \sum m_{i}-k^{2}$. Suppose that there exists $m_{i} \ge 2$. Then $\sum m_{i}^{2} \ge \sum m_{i}+3$ so that
\begin{align*}
2d &>2(\alpha+k+1)^{2} \sum m_{i}^{2}-2k(\alpha+1)\sum m_{i}-k^{2} \\
&\ge (2(\alpha+k+1)^{2}-2k(\alpha+1))\sum m_{i}+6(\alpha+k+1)^{2}-k^{2}.
\end{align*}
Now, we claim that $L-(\alpha+1)D_{S}$ is ample. As in the proof of Theorem \ref{thm:any Picard number}, it is sufficient to show that $(L-(\alpha+1)D_{S}.L)>0$ and $(L-(\alpha+1)D_{S})^{2} >0$. Note that 
\begin{align*}
(L-(\alpha+1)D_{S}.L)&=2d-(\alpha+1)(k+(\alpha+1)\sum m_{i}) \\
&\ge (2(\alpha+k+1)^{2}-2k(\alpha+1)-(\alpha+1)^{2})\sum m_{i}+6(\alpha+k+1)^{2}-k^{2}-k(\alpha+1) \\
&>0 
\end{align*}
and 
\begin{align*}
(L-(\alpha+1)D_{S})^{2}&=2d+(\alpha+1)^{2}(-1+\sum m_{i}^{2})-2(\alpha+1)(k+(\alpha+1)\sum m_{i}) \\
&\ge 2d+(\alpha+1)^{2}(2+\sum m_{i})-2(\alpha+1)(k+(\alpha+1)\sum m_{i}) \\
&>0.
\end{align*}
Hence $L-(\alpha+1)D_{S}$ is ample. However, we have:
\begin{align*}
0<(L-(\alpha+1)D_{S}.D_{S})&=k+(\alpha+1)\sum m_{i}-(\alpha+1)(-1+\sum m_{i}^{2}) \\
&=(\alpha+1)(\sum m_{i}-\sum m_{i}^{2})+\alpha+k+1 \\
& \le -3(\alpha+1)+\alpha+k+1 \\
&<0,
\end{align*}
which is a contradiction. So all $m_{i}$'s are either $0$ or $1$. 

Let $\beta$ be the number of $1$'s among $m_{i}$'s, that is, $\beta=\sum m_{i}=\sum m_{i}^{2}$. Since $D_{S}^{2}$ is even, $\beta$ should be odd. If $\beta=1$, $D_{S}$ is an elliptic curve of degree $\alpha+k+1$ with respect to $L$, which is impossible. So $\beta \ge 3$. However, it means that $D_{S}$ passes through $\beta=D_{S}^{2}+1=2h^{0}(D_{S})+1$ general points, which is also impossible. So the case ii) is excluded. 

Finally, it remains to eliminate the case i). Note that $D_{S}^{2}=-2+\sum m_{i}^{2}$ and $L.D_{S}=k-1+(\alpha+1) \cdot \sum m_{i}$. The similar argument as in the case ii) shows that all $m_{i}$'s are either $0$ or $1$. Let $\beta$ be the number of $1$'s among $m_{i}$'s. Since $D_{S}^{2}$ is even, $\beta$ should be even. If $\beta \ge 4$, the generality condition induces a contradiction as above. So let $\beta=2$, i.e. $D_{S}$ is an elliptic curve passing through $2$ general points on $S$. By Riemann-Roch, $h^{0}(\OO_{S}(D_{S}))=h^{1}(\OO_{S}(D_{S}))$. Consider the following short exact sequence:
\begin{align*}
0 \rightarrow \OO_{S}(-D_{S}) \rightarrow \OO_{S} \rightarrow \OO_{D_{S}} \rightarrow 0.
\end{align*}
Then it induces the long exact sequence:
\begin{align*}
0 \rightarrow H^{0}(\OO_{S}(-D_{S})) \rightarrow H^{0}(\OO_{S}) \overset{\psi}{\rightarrow} H^{0}(\OO_{D_{S}}) \rightarrow H^{1}(\OO_{S}(-D_{S})) \overset{\phi}{\rightarrow} H^{1}(\OO_{S}) \rightarrow H^{1}(\OO_{D_{S}}).
\end{align*}
Since $H^{0}(\OO_{S}(-D_{S}))=0$ and $H^{0}(\OO_{S}) \cong \mathbb{C} \cong H^{0}(\OO_{D_{S}})$, $\psi$ is an isomorphism. Thus $\phi$ is injective. Since $h^{1}(\OO_{S})=2$, we have $h^{0}(\OO_{S}(D_{S}))=h^{1}(\OO_{S}(D_{S}))=h^{1}(\OO_{S}(-D_{S})) \le h^{1}(\OO_{S})=2$, that is, ${\rm dim} \lvert D_{S} \rvert \le 1$. Hence $D_{S}$ cannot pass through $2$ general points on $S$, which is a contradiction. Thus we exclude the case i). 
\end{proof}

\begin{proof}[Proof of Theorem \ref{thm:k-very ampleness on abelian surface}] When $c=1$, it is trivial by Lemma \ref{key lemma} and Theorem \ref{thm:criterion}. For an arbitrary $c$, the same procedure as in the proof of Theorem \ref{thm:higher Picard number one} and of Theorem \ref{thm:any Picard number} leads to the conclusion.
\end{proof}


\end{section}


\bibliographystyle{abbrv}
\bibliography{Library}



\end{document}

\begin{thebibliography}{10}

\bibitem{B}
T.~Bauer.
\newblock Seshadri constants and periods of polarized abelian varieties.
\newblock {\em Mathematische Annalen}, 312(4):607--623, 1998.

\bibitem{BDHKKSS}
T.~Bauer, S.~D. Rocco, B.~Harbourne, M.~Kapustka, A.~Knutsen, W.~Syzdek, and
  T.~Szemberg.
\newblock {A primer on Seshadri constants}.
\newblock {\em Contemporary Mathematics}, 496:33, 2009.

\bibitem{BS1}
T.~Bauer and T.~Szemberg.
\newblock {Seshadri constants of abelian surfaces. Appendix to: Seshadri
  constants and periods of polarized abelian varieties}.
\newblock {\em Mathematische Annalen}, 312(4):607--623, 1998.

\bibitem{BS}
M.~Beltrametti and A.~J. Sommese.
\newblock Zero cycles and k-th order embeddings of smooth projective surfaces.
\newblock In {\em Problems in the theory of surfaces and their classification,
  Symposia Math}, volume~32, pages 33--48, 1991.

\bibitem{BL}
C.~Birkenhake and H.~Lange.
\newblock {\em Complex abelian varieties}, volume 302.
\newblock Springer Science \& Business Media, 2013.

\bibitem{C}
M.~Coppens.
\newblock Embeddings of general blowing-ups at points.
\newblock {\em Journal fur die Reine und Angewandte Mathematik}, pages
  179--179, 1995.

\bibitem{DH}
J.~d'Almeida and A.~Hirschowitz.
\newblock Quelques plongements projectifs non speciaux de surfaces
  rationnelles.
\newblock {\em Mathematische Zeitschrift}, 211(1):479--483, 1992.

\bibitem{EKL}
L.~Ein, O.~K{\"u}chle, and R.~Lazarsfeld.
\newblock Local positivity of ample line bundles.
\newblock {\em arXiv preprint alg-geom/9408003}, 1994.

\bibitem{F}
L.~F. Garc{\'\i}a.
\newblock Seshadri constants in finite subgroups of abelian surfaces.
\newblock {\em Geometriae Dedicata}, 127(1):43--48, 2007.

\bibitem{H}
K.~Hanumanthu.
\newblock Positivity of line bundles on general blow ups of $\mathbb{P}^2$.
\newblock {\em Journal of Algebra}, 461:65--86, 2016.

\bibitem{HTT}
Y.~Hinohara, K.~Takahashi, and H.~Terakawa.
\newblock On tensor products of $k$-very ample line bundles.
\newblock {\em Proceedings of the American Mathematical Society},
  133(3):687--692, 2005.

\bibitem{K2}
O.~K{\"u}chle.
\newblock Ample line bundles on blown up surfaces.
\newblock {\em Mathematische Annalen}, 304(1):151--155, 1996.

\bibitem{K}
O.~K{\"u}chle.
\newblock {Multiple point Seshadri constants and the dimension of adjoint
  linear series}.
\newblock In {\em Annales de l'institut Fourier}, volume~46, pages 63--72.
  Chartres: L'Institut, 1950-, 1996.

\bibitem{KL1509}
A.~K{\"u}ronya and V.~Lozovanu.
\newblock {A Reider-type theorem for higher syzygies on abelian surfaces}.
\newblock {\em arXiv preprint arXiv:1509.08621}, 2015.

\bibitem{L}
R.~Lazarsfeld.
\newblock {Lengths of periods and Seshadri constants of abelian varieties}.
\newblock {\em arXiv preprint alg-geom/9606012}, 1996.

\bibitem{N}
M.~Nakamaye.
\newblock Seshadri constants on abelian varieties.
\newblock {\em American Journal of Mathematics}, pages 621--635, 1996.

\bibitem{R}
I.~Reider.
\newblock Vector bundles of rank 2 and linear systems on algebraic surfaces.
\newblock {\em Annals of Mathematics}, 127(2):309--316, 1988.

\bibitem{S}
J.~Shin.
\newblock Higher syzygies on abelian surfaces.
\newblock {\em arXiv preprint arXiv:1608.06052}, 2016.

\bibitem{Sz}
T.~Szemberg.
\newblock {Bounds on Seshadri constants on surfaces with Picard number $1$}.
\newblock {\em Communications in Algebra}, 40(7):2477--2484, 2012.

\bibitem{ST1}
T.~Szemberg and H.~Tutaj-Gasi{\'n}ska.
\newblock Embeddings of general blowups of abelian surfaces.
\newblock {\em Manuscripta Mathematica}, 103(2):183--190, 2000.

\bibitem{ST2}
T.~Szemberg and H.~Tutaj-Gasi{\'n}ska.
\newblock General blowups of ruled surfaces.
\newblock {\em Abhandlungen aus dem Mathematischen Seminar der Universit{\"a}t
  Hamburg}, 70(1):93--103, 2000.

\bibitem{ST3}
T.~Szemberg and H.~Tutaj-Gasi{\'n}ska.
\newblock General blow-ups of the projective plane.
\newblock {\em Proceedings of the American Mathematical Society},
  130(9):2515--2524, 2002.

\bibitem{T'}
H.~Terakawa.
\newblock {The $k$-very ampleness and $k$-spannedness on polarized abelian
  surfaces}.
\newblock {\em Mathematische Nachrichten}, 195(1):237--250, 1998.

\bibitem{T2}
H.~Tutaj-Gasi{\'n}ska.
\newblock Higher embeddings of general blowups of ruled surfaces.
\newblock {\em Monatshefte f{\"u}r Mathematik}, 133(3):255--263, 2001.

\bibitem{T}
H.~Tutaj-Gasi{\'n}ska.
\newblock A note on general blow-ups of abelian surfaces.
\newblock {\em Journal of Pure and Applied Algebra}, 176(1):81--88, 2002.

\end{thebibliography}