The partial data Calder\'on problem in dimension three

Extracted source text · arXiv 2606.10247

\documentclass[11pt]{amsart}
%%%packages%%%
\usepackage{amssymb}
\usepackage{amsfonts}
\usepackage{amscd}
\usepackage{amsmath}
\usepackage{mathrsfs}
\usepackage{curves}
\usepackage{epsfig}
\usepackage[margin=1.1in, centering]{geometry}
\usepackage{graphicx}
\usepackage{verbatim}
\usepackage{latexsym}
\usepackage{eucal} 
 
%%%environments%%% 
%\setlength{\topmargin}{.25in}
%\setlength{\textheight}{608pt}
%\renewcommand{\baselinestretch}{1.0} 
\numberwithin{equation}{section}

%%%Thms%%%
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{prop}[theorem]{Proposition}
\newtheorem{cor}[theorem]{Corollary}
\newtheorem{conj}[theorem]{Conjecture}
\newtheorem{assump}[theorem]{Assumption}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{example}[theorem]{Example}
\newtheorem{remark}[theorem]{Remark}

%%%%letters%%%
\def \mca {{\mathcal A}}
\def \mcb {{\mathcal B}}
\def \mcc {{\mathcal C}}
\def \mcd {{\mathcal D}}
\def \mce {{\mathcal E}}
\def \mcf {{\mathcal F}}
\def \mcg {{\mathcal G}}
\def \mch {{\mathcal H}}
\def \mci {{\mathcal I}}
\def \mcj {{\mathcal J}}
\def \mck {{\mathcal K}}
\def \mcl {{\mathcal L}}
\def \mcm {{\mathcal M}}
\def \mcn {{\mathcal N}}
\def \mco {{\mathcal O}}
\def \mcp {{\mathcal P}}
\def \mcq {{\mathcal Q}}
\def \mcr {{\mathcal R}}
\def \mcs {{\mathcal S}}
\def \mct {{\mathcal T}}
\def \mcu {{\mathcal U}}
\def \mcv {{\mathcal V}}
\def \mcw {{\mathcal W}}
\def \mcz {{\mathcal Z}}

\def \mse {\mathscr E}
\def \msd {\mathscr D}

\def \mbb {{\mathbb B}}
\def \mbc {{\mathbb C}}
\def \mbh {{\mathbb H}}
\def \mbn {{\mathbb N}}
\def \mbr {{\mathbb R}}
\def \mbs {{\mathbb S}}
\def \mbz {{\mathbb Z}}

%%%operator names%%%
\def \id {\operatorname{Id}}
\def \comp {\operatorname{comp}}
\def \loc {\operatorname{loc}}
\def \wf {\operatorname{WF}}
\def \tr {\operatorname{Tr}}
\def \im {\operatorname{Im}}
\def \re {\operatorname{Re}}
\def \diag{\operatorname{Diag}}
\def \supp {\operatorname{supp }}
\def \defeq {\stackrel{\operatorname{def}}{=}}
\def \beqq {\begin{equation}}
\def \eeqq {\end{equation}}
\def \WF {\text{WF}}
\def \dim {\text{dim}}
\def \bpf {\begin{proof}}
\def \epf {\end{proof}}
\def \beq {\begin{equation*}}
\def \eeq {\end{equation*}}

%%%symbols%%%
\def \mrn {{\mathbb R}^n}
\def \msn {{\mathbb S}^n}
\def \eps {\epsilon}   
\def \la {\lambda}   
\def \La {\Lambda}    
\def \del {\delta}   
\def \lap {\Delta}
\def \p {\partial}
\def \ha {\frac{1}{2}}
\def \sig {\sigma}
\def \tilde {\widetilde}
\def \grad  {\operatorname{grad}} 
\def \sym {\text{Sym}}
\def \op {\text{Op}}
\def \msm {\mathscr{M}}
\def \msx {\mathscr{X}}
\def \msc {\mathscr{C}}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\begin{document}
\title[]{The partial data Calder\'on problem in dimension three}
\author{Gunther Uhlmann}
\address{Gunther Uhlmann
\newline
\indent Department of Mathematics, University of Washington}
\email{gunther@math.washington.edu}
\author{Yiran Wang}
\address{Yiran Wang
\newline
\indent Department of Mathematics, Emory University}
\email{yiran.wang@emory.edu}
\begin{abstract} 
We consider an inverse boundary value problem for the time-independent Schr\"odinger equation in dimension three. We prove that the local Dirichlet-to-Neumann map defined near a boundary point uniquely  determines the potential in a neighborhood of the boundary point in the interior. In particular, we show that the uniqueness question can be reduced to the injectivity of a weighted X-ray transform, which links inverse boundary value problems to integral geometry.  
\end{abstract}
\date{\today.}
%\dedicatory{\today}
\maketitle
%%%%%%%%%%%%%%%%%%
\section{Introduction}\label{sec-intro}
Let $\mcm$ be a simply connected domain in $\mbr^3$ with smooth   boundary $\p \mcm$.   Let $\lap$ denote the Laplacian on $\mbr^3$ and $V$ be a smooth function on $\mcm$. We consider the Dirichlet problem 
\beqq\label{eq-dtn}
\begin{gathered}
(-\lap + V)   u  = 0  \text{ in } \mcm,\\
u = f \text{ at } \p \mcm.
\end{gathered}
\eeqq
Suppose that $0$ is not a Dirichlet eigenvalue. For $f\in C^\infty(\p\mcm),$ we let $u$ be the unique $C^\infty(\overline\mcm)$ solution of \eqref{eq-dtn}. The Dirichlet-to-Neumann (DtN) map $\La: C^\infty(\p \mcm)\rightarrow C^\infty(\p\mcm)$ is defined by 
\beqq\label{eq-dtn11}
\La f = \p_\nu u|_{\p \mcm},
\eeqq
 where $\nu$ is the unit outward normal to $\p \mcm$.  The inverse problem we study is the determination of $V$ from $\La$. This problem is closely related to the inverse conductivity problem studied by Calder\'on in \cite{Cal}. It was solved by Sylvester and Uhlmann in \cite{SyUh} using Complex Geometric Optics (CGO) solutions. In this article, we prove a uniqueness result for a local problem, see \cite{KeSa} for a review of this inverse problem. For $z\in \p \mcm$, we let $\Omega$ be a neighborhood of $z$ in $\mcm$ and $\Gamma = \Omega \cap \p\mcm$.  We consider Dirichlet data $f$ compactly supported in $\Gamma$, and define a local DtN map  $\La^\Gamma: C_0^\infty(\Gamma)\rightarrow C^\infty(\Gamma)$ as 
\beqq\label{eq-dtn-p}
\La^\Gamma f = \p_\nu u|_{\Gamma}. 
\eeqq
Our main result is the unique determination of $V$ in $\Omega$ from $\La^\Gamma.$ 
\begin{theorem}\label{thm-main1}
Let $\mcm$ be a strictly convex domain. For any $z\in \p \mcm$, there exists a neighborhood $\Omega$ in $\mcm$ such that for any $V, \tilde V\in C^\infty(\mcm)$ with $\supp V\cap \Gamma = \emptyset,$ $\supp \tilde V\cap \Gamma = \emptyset$, if $0$ is not a Dirichlet eigenvalue for \eqref{eq-dtn} for $V, \tilde V$, and $\La^\Gamma = \tilde \La^\Gamma$, then $V = \tilde V$ in $\Omega.$  
\end{theorem}

We remark that we assumed dimension three and $V, \tilde V$ supported away from $\Gamma$ to avoid some technicalities in the analysis. We believe that these assumptions can be removed. The partial data problem for the Calder\'on problem and other elliptic equations has been extensively studied in the literature starting with the paper \cite{BuUh} in the articles \cite{AmUh, DKSU, IUY1, IUY2, Isa, KrUh1, KrUh2, LiUh} to name a few. The method employed in this article is completely different, doesn't use CGO solutions or unique continuation. In fact, we show that the uniqueness question can be reduced to the injectivity of weighted X-ray transforms, which was solved in \cite{UhVa}. The method also applies to the inverse problem with full data $\La$ in \eqref{eq-dtn11}. Thus, we have a new proof of the global uniqueness theorem in \cite{SyUh} without using CGO solutions.  To demonstrate the new method, we prove the following version of the global uniqueness theorem. 

 \begin{theorem}\label{thm-main}
Let $\mcm$ be a strictly convex domain and $V, \tilde V\in C_0^\infty(\mcm)$.  %with $\|V\|_{C^2},$ $\|\tilde V\|_{C^2} < M$. 
Assume that $0$ is not a Dirichlet eigenvalue for \eqref{eq-dtn} for  $V, \tilde V$. Let $\La, \tilde \La: C^\infty(\p \mcm)\rightarrow C^\infty(\p \mcm)$ be the corresponding Dirichlet-to-Neumann maps. If $\La = \tilde \La,$ then $V = \tilde V.$
 \end{theorem}

This paper is organized as follows. In Section \ref{sec-prep}, we derive an integral identity for a family of approximate solutions of \eqref{eq-dtn}. The main analysis is in Section \ref{sec-pf-main} and Section \ref{sec-pf-main1}, where we construct a specific family of solutions to extract some weighted X-ray transforms of the difference of potentials from the leading order terms of the integral identity. Then we estimate the remainder terms in Section \ref{sec-pf-rem}. Finally, we complete the proof of Theorems \ref{thm-main} and \ref{thm-main1} in Section \ref{sec-pf}.  
 
%=====================%
\section{Preliminaries}\label{sec-prep}
Let $V, \tilde V$ be two potentials as in Theorem \ref{thm-main}. Let $u, \tilde u$ be the unique smooth solutions of 
\beqq\label{eq-schro1}
\begin{gathered}
(-\lap + V)   u  = 0  \text{ in } \mcm\\
u = f \text{ at } \p \mcm
\end{gathered}
\eeqq
and 
\beqq\label{eq-schro2} 
\begin{gathered}
(-\lap + \tilde V)   \tilde u  = 0  \text{ in } \mcm\\
\tilde u = \tilde f \text{ at } \p \mcm
\end{gathered}
\eeqq
with $f, \tilde f \in C^\infty(\p\mcm)$.  From $\La = \tilde \La$, we have the following integral identity (see e.g.\ \cite{FSU}) 
 \beqq\label{eq-ide0}
 \int_{\mcm} \delta V(z) u(z)\tilde u(z)dz = 0, \text{ where } \delta V(z) = V(z) - \tilde V(z). 
 \eeqq
We will construct special solutions in the identity to recover $\delta V.$


Let $h\in (0, 1)$ be a small parameter. Let $u$ be the unique solution of \eqref{eq-dtn} with $f\in C^\infty(\p \mcm).$ Then we observe that $u$ satisfies  
\beqq\label{eq-dtn0}
\begin{gathered}
h^2(-\lap + V)u   -  u  =  -u   \text{ in } \mcm\\
u = f \text{ at } \p \mcm
\end{gathered}
\eeqq 
for all $h\in (0, 1)$. We use \eqref{eq-dtn0} to derive an expression of $u$ for small $h.$ We remark that the existence and regularity of $u$ is known from \eqref{eq-dtn}, and we are merely using \eqref{eq-dtn0} to get an expression of $u.$ First, we remove $V$ from the equation and consider the Dirichlet problem
\beqq\label{eq-dtn1}
\begin{gathered}
h^2(-\lap)u_0   -  u_0  =  -u  \text{ in } \mcm\\
u_0 = f + f_*  \text{ at } \p \mcm. 
\end{gathered}
\eeqq
We show below that one can choose $f_*$ so that $u_0$ becomes an approximation of $u$. Hereafter, we use $C$ for a generic constant that does not depend on $h.$ The constant can depend on $V$ but it is not an issue and it is not necessary to keep track of this dependency. 

\begin{lemma}\label{lm-est}
Let $u$ be the unique smooth solution of \eqref{eq-dtn} with $f\in C^\infty(\p\mcm)$. Then there is a solution $u_0\in C^\infty(\overline \mcm)$ of \eqref{eq-dtn1} with   $f_*\in C^\infty(\p\mcm)$ such  that  $v = u-u_0$ satisfies  
\beqq\label{eq-vest0}
 \|v\|_{L^2(\mcm)}\leq Ch\|f\|_{L^2(\p\mcm)}, \quad \|v\|_{H^1(\mcm)} \leq C \|f\|_{L^2(\p\mcm)}, \quad  \|v\|_{H^2(\mcm)} \leq Ch^{-1} \|f\|_{L^2(\p\mcm)}. 
\eeqq
In particular, $f_*$ satisfies the following estimates
\beq
\|f_*\|_{H^{1/2}(\p\mcm)}\leq C\|f\|_{L^2(\p\mcm)}. 
%\|f_*\|_{H^{1/2}(\p\mcm)}\leq  C \|f\|_{L^2(\p\mcm)}, \quad \|f_*\|_{H^1(\p \mcm)}\leq C h^{-1/2}\|f\|_{L^2(\p\mcm)}. 
\eeq
\end{lemma}
\bpf
From the standard elliptic PDE theory, we know that the solution of the Dirichlet problem \eqref{eq-schro1} satisfies  
\beqq\label{eq-uest}
\|u\|_{H^{1/2}(\mcm)}\leq C\|f\|_{L^2(\p\mcm)}. 
\eeqq
See e.g.\ \cite[Section 1 of Chapter 5]{Tay}.  Next, we use \eqref{eq-dtn0} and \eqref{eq-dtn1} to get 
\beqq\label{eq-v}
\begin{gathered}
(-h^2\lap -1)v = - h^2 Vu \text{ in } \mcm,  \\
 v = f_* \text{ on } \p \mcm. 
\end{gathered}
\eeqq 
Consider the semiclassical resolvent $R(h) = (-h^2\lap - 1)^{-1}$ on the whole space $\mbr^3$.  Let $\chi_1, \chi_2$ be smooth cut-off functions on $\mbr^3$ such that (i) $\chi_2$ is supported in $\mcm$ and $\chi_2 = 1$ on the support of $V$;   (ii) $\chi_1 = 1$ on $\mcm.$  The following estimate for the cut-off resolvent $\chi_1 R(h)\chi_2$  can be seen from Theorem of \cite{VaZw}: 
\beqq\label{eq-semi-est}
\begin{gathered}
\|\chi_1 R(h) \chi_2\|_{H^m(\mbr^3)\rightarrow H^{m}(\mbr^3)} \leq C_m  h^{-1},\\
\|\chi_1 R(h) \chi_2\|_{H^m(\mbr^3)\rightarrow H^{m+1}(\mbr^3)} \leq C_m  h^{-2},\\
\|\chi_1 R(h) \chi_2\|_{H^m(\mbr^3)\rightarrow H^{m+2}(\mbr^3)} \leq C_m  h^{-3}, 
\end{gathered}
\eeqq
for any $m\in \mbr$. Here, $C_m$ is some constant depending on $m.$ Also, we point out that the Sobolev spaces in \eqref{eq-semi-est} are classical Sobolev spaces (as opposed to the semiclassical Sobolev spaces used in \cite{VaZw}\footnote{The authors thank Andr\'as Vasy for clarifying the semiclassical nature of the resolvent estimate in \cite{VaZw}.}.) We let $v =  \chi_1 R(h)\chi_2 (-h^2 Vu)$. Then $v\in C^\infty(\mbr^3)$ and it satisfies the equation in \eqref{eq-v} on $\mcm$ with boundary data $f_* = v|_{\p\mcm} \in C^\infty(\p \mcm)$. Finally, we have   
\beq
\begin{gathered} 
\|v\|_{L^2(\mcm)}\leq C h^{-1}  \|h^2Vu\|_{L^2(\mcm)} \leq Ch  \|f\|_{L^2(\p\mcm)}, 
 \end{gathered}
\eeq 
using \eqref{eq-uest} and \eqref{eq-semi-est}. The higher order estimates  in \eqref{eq-vest0} follow similarly. The estimate of $f_*$ follows from \eqref{eq-vest0} and the trace theorem. This completes the proof.  
\epf

Together with the estimate of $u$ in \eqref{eq-uest}, we derive from Lemma \ref{lm-est} that  
\beqq\label{eq-u0est}
\|u_0\|_{H^{1/2}(\mcm)}\leq C\|f\|_{L^2(\p\mcm)}.
\eeqq 

Second, it is well-known that the Green's function of $-h^2\lap - 1$ on $\mbr^3$ is  
\beqq\label{eq-fund}
G_0(z, z'; h) = \frac{e^{i|z - z'|/h}}{4\pi h^2 |z - z'|}, \quad z, z'\in \mbr^3, 
\eeqq
where $|\cdot|$ denotes the Euclidean norm. Then the solution $u_0$ of \eqref{eq-dtn1} can be expressed as 
\beqq\label{eq-u1}
\begin{gathered}
u_0(z)  = -  \int_{\mcm} G_0(z, z'; h)u(z') dz' - h^2 \int_{\p\mcm} G_0(z, z'; h)\p_\nu u_0(z') dz' \\
+ h^2 \int_{\p\mcm}( f(z') + f_*(z')) \p_\nu G_0(z, z'; h)dz', 
\end{gathered}
\eeqq
see e.g.\ equation (10.14) of \cite{Tre}.  In the last term, $\nu$ denotes the outward pointing unit normal vector at $z'\in \p \mcm$. We find that 
\beq
\begin{split}
\p_\nu G_0(z, z'; h) &=     -\frac{e^{i|z - z'|/h}}{4\pi h^2} \frac{\p_\nu |z - z'|}{|z - z'|^2} +  \frac{i}{h} \p_\nu |z - z'| \frac{e^{i|z - z'|/h}}{4\pi h^2 |z - z'|}  \\
 &=   e^{i|z - z'|/h} \frac{\nu \cdot (z - z')}{4\pi h^2 |z - z'|^3}  - e^{i|z - z'|/h}\frac{ i \nu \cdot (z - z') }{4\pi h^3 |z - z'|^2}  \\
& \doteq G_{\nu, 1}(z, z'; h) + G_{\nu, 2}(z, z'; h). 
 \end{split}
\eeq
%We can split the last term in \eqref{eq-u1} to 
%\beq
%\begin{gathered}
%\int_{\p \mcm} \p_\nu G_0(z, z'; h)f(z')dz' = \int_{\p \mcm}  G_{\nu, 1}(z, z'; h)f(z')dz' + \int_{\p \mcm}  G_{\nu, 2}(z, z'; h)f(z')dz'
% \end{gathered}
%\eeq
We can split the last term in \eqref{eq-u1} and write $u_0$ as    
\beqq\label{eq-Is}
\begin{gathered}
u_0(z) = \sum_{j = 1}^6 I_j(z), \text{ where }\\
I_1(z) = -\int_{\mcm} G_0(z, z'; h)u(z') dz', \quad I_2(z) = - \int_{\p\mcm} h^2 G_0(z, z'; h)\p_\nu u_0(z') dz', \\
I_3(z) = \int_{\p \mcm}  h^2 G_{\nu, 1}(z, z'; h)f(z')dz', \quad I_4(z) =  \int_{\p \mcm}  h^2 G_{\nu, 2}(z, z'; h)f(z')dz', \\
I_5(z) = \int_{\p \mcm}  h^2 G_{\nu, 1}(z, z'; h)f_*(z')dz', \quad I_6(z) =  \int_{\p \mcm}  h^2 G_{\nu, 2}(z, z'; h)f_*(z')dz'. 
\end{gathered}
\eeqq

Now we can use the constructed solutions in the main identity \eqref{eq-ide0}.  Applying Lemma \ref{lm-est} to both \eqref{eq-schro1} and \eqref{eq-schro2}, we write $u = u_0 + v,  \tilde u = \tilde u_0 + \tilde v$, where $v, \tilde v$ satisfy the estimates \eqref{eq-vest0}, and 
 $u_0, \tilde u_0$ are given by \eqref{eq-Is} in the form 
\beq
u_0 = \sum_{j = 1}^6 I_j, \quad \tilde u_0 = \sum_{j = 1}^6 \tilde I_j. 
\eeq
We get from \eqref{eq-ide0} that  
\beqq\label{eq-mainid}
\begin{gathered}
0 =    \int_{\mcm} \delta V(z) u_0(z) \tilde u_0(z) dz  + Y_{12} + Y_{21} + Y_{22}, 
\end{gathered}
\eeqq
where 
\beqq\label{eq-Y}
\begin{gathered}
Y_{12} = \int_{\mcm} \delta V(z) u_0(z) \tilde v(z) dz, \quad Y_{21} =  \int_{\mcm} \delta V(z) v(z) \tilde u_0(z) dz,\\
Y_{22} =  \int_{\mcm} \delta V(z) v(z) \tilde v(z) dz. 
\end{gathered}
\eeqq
%For the integral in \eqref{eq-mainid}, 
%we have 
%\beq
%u_0(z)\tilde u_0(z) = \sum_{j, k = 1}^6 I_j(z)\tilde I_k(z).
%\eeq
For $j, k = 1, \cdots, 6$, we define the integrals 
\beqq\label{eq-X}
X_{jk} \doteq \int_{\mcm} \delta V(z) I_j(z)\tilde I_k(z) dz.
\eeqq
Then we write \eqref{eq-mainid} as 
\beqq\label{eq-mainid1}
0 = \sum_{j, k = 1}^6 X_{jk} + Y_{12} + Y_{21} + Y_{22}. 
\eeqq
Note that all the $X_\bullet, Y_\bullet$ terms depend on $h$. Our goal is to identify the leading order terms as $h\rightarrow 0.$ 

 
 %=====================%
\section{The X-ray transform}\label{sec-pf-main}
In this section, we analyze $X_{44}$ which contains many key ingredients of the proof.  We have
\beq
\begin{split}
X_{44} =& \int_{\p\mcm}\int_{\p \mcm} \int_\mcm \delta V(z) h^2 G_{\nu, 2}(z, z'; h) f(z') h^2 G_{\nu, 2}(z, z''; h)\tilde f(z'')dz dz' dz''\\
 =& \int_{\p\mcm}\int_{\p \mcm} (\int_\mcm \delta V(z)   e^{i|z - z'|/h} \frac{ i \nu_{z'} \cdot (z - z') }{4\pi h |z - z'|^2}   e^{i|z - z''|/h}\frac{ i \nu_{z''} \cdot (z - z'') }{4\pi h |z - z''|^2}     dz) f(z')  \tilde f(z'') dz' dz''. 
\end{split}
\eeq
Here, we use $\nu_{\bullet}$ to denote the unit outward pointing normal vector at $\bullet\in \p \mcm.$ Note that $\delta V$ is supported away from $\p\mcm$ by the assumption of Theorem \ref{thm-main} so the integrand is integrable and we can change the order of integration. We first analyze the integral in $z$ which is 
\beqq\label{eq-mI}
I(z', z'') = \int_\mcm \delta V(z) \frac{ i^2 ( \nu_{z'} \cdot (z - z')) (  \nu_{z''} \cdot (z - z'') ) }{16 \pi^2 h^2 |z - z'|^2|z - z''|^2}   e^{i|z - z'|/h + i|z - z''|/h}  dz 
\eeqq
for $z', z''\in \p\mcm$. Again, the integrand 
\beq
F(z) = \delta V(z) \frac{ (\nu_{z'} \cdot (z - z'))( \nu_{z''} \cdot (z - z''))  }{|z - z'|^2|z - z''|^2}  (-\frac{1}{16\pi^2 h^2})
\eeq
is smooth for $z\in \mcm.$ 
 
To simplify the calculation of \eqref{eq-mI}, for fixed $z', z''\in \p \mcm, z' \neq z''$, we work in coordinates  $x = (x_1, x_2, x_3)\in \mbr^3$ so that $x(z') = 0$ and $x(z'') - x(z')= |z'' - z'|(1, 0, 0)$. For $z\in \mcm$, we write it as $z = (a, x')$ with $a\in \mbr, x' = (x_2, x_3) \in \mbr^{2}$. We have  
\beq
|z - z'| = (a^2 + |x'|^2)^\ha, \quad |z - z''| = ((|z' - z''| - a)^2 + |x'|^2)^\ha.
\eeq
For fixed $a$, we consider the phase function in \eqref{eq-mI}
\beq
\Phi(a, x') = (a^2 + |x'|^2)^\ha +  ((|z' - z''| - a)^2 + |x'|^2)^\ha.
\eeq
We claim that for fixed $a$, $\Phi(a, x')$ has non-degenerate critical points at $x' =0$. First, 
\beq
\nabla_{x'} \Phi(a, x') =   \frac{x'}{(a^2 + |x'|^2)^\ha} +  \frac{x'}{((|z' - z''| - a)^2 + |x'|^2)^\ha}.
\eeq
So $\nabla_{x'} \Phi(a, x') = 0$ at $x' = 0$. Next, we compute the Hessian  
\beq
\begin{gathered}
 \Phi''(a, x') = 
 \begin{pmatrix}
 \frac{a^2 + x_3^2}{(a^2 + |x'|^2)^{3/2}} +  \frac{(|z' - z''| - a)^2 + x_3^2}{((|z - z''| - a)^2 + |x'|^2)^{3/2}} & -\frac{x_2x_3}{(a^2 + |x'|^2)^{3/2}} -  \frac{x_2x_3}{((|z' - z''| - a)^2 + |x'|^2)^{3/2}} \\
  -\frac{x_2x_3}{(a^2 + |x'|^2)^{3/2}} - \frac{x_2x_3}{((|z' - z''| - a)^2 + |x'|^2)^{3/2}} & \frac{a^2 + x_2^2}{(a^2 + |x'|^2)^{3/2}} +  \frac{(|z - z''| - a)^2 + x_2^2}{((|z' - z''| - a)^2 + |x'|^2)^{3/2}} 
 \end{pmatrix}. 
   \end{gathered}
\eeq
At $x' = 0$, we find that 
\beq
\det \Phi''(a, 0) = (\frac{1}{|a|} + \frac{1}{||z' - z''| - a|})^2\neq 0.
\eeq
Thus we have proved the claim. Now we can apply the stationary phase argument  (e.g.\ Theorem 7.7.5 of \cite{Ho1}) to get  (with $n = 3$)
\beq
\begin{split}
I(z', z'')& =  \int_{0}^{|z' - z''|} \det(\Phi''(a, 0)/(2\pi i))^{-\ha} h^{(n-1)/2} F(a, 0)  e^{\frac{i}{h}(|a| + ||z' - z''| - a|)} da  
+ O_{L^\infty}(h^{(n+1)/2} )\\
 &=  e^{\frac{i}{h}|z' - z''|}h  \int_{0}^{|z' - z''|} \det(\Phi''(a, 0)/(2\pi i))^{-\ha}  F(a, 0)   da  + O_{L^\infty}(h^{2}  ). 
\end{split} 
\eeq
For $z', z''\in \p \mcm$, we parametrize the line segment from $z'$ to $z''$ by 
\beq
\gamma_{z', z''}(t) = z' + t (z'' - z')/|z'' - z'|, \quad t\in [0, |z''-z'|].
\eeq
On the line segment, $F(z)$ becomes
\beq
F(t) = \delta V(\gamma_{z', z''}(t)) \frac{1}{t(|z' - z''| -t)} (\nu_{z'}\cdot  \frac{(z'' - z')}{|z' - z''|}) (\nu_{z''}\cdot  \frac{(z' - z'')}{|z' - z''|})(\frac{-1}{16\pi^2h^2})(2\pi i). 
\eeq
Let 
\beq
W_{44}(z', z'', t) =  (\frac{1}{t} + \frac{1}{|z' - z''| - t} )^{-1}  \frac{1}{t(|z' - z''| -t)}. 
\eeq
Note that $W_{44}$ is positive. %because $-\nu_{z'}$ and $(z''-z')/|z''-z'|$ are both inward pointing at $z'$.  
Let $\varphi\in C_0^\infty(\mcm)$. We define the weighted X-ray transform of $\varphi(z)$ as
\beq
X^{W_{44}} \varphi(z', z'') = \int_{0}^{|z'-z''|} W_{44}(z', z'', t)  \varphi(\gamma_{z', z''}(t))   dt, \quad z', z''\in \p \mcm. 
\eeq
Thus we proved that 
\beqq\label{eq-x44}
\begin{gathered}
X_{44} = \int_{\p \mcm}\int_{\p \mcm} e^{i|z' - z''|/h} h^{-1} (-\frac{(2\pi i)}{16\pi^2})  (\nu_{z'}\cdot  \frac{(z'' - z')}{|z' - z''|}) (\nu_{z''}\cdot  \frac{(z' - z'')}{|z' - z''|}) X^{W_{44}}\delta V(z', z'') \\
\cdot f(z')  \tilde f(z'')dz'dz'' +  \int_{\p \mcm}\int_{\p \mcm} e^{i|z' - z''|/h} Z(z', z''; h) f(z') \tilde f(z'')dz'dz'', 
 \end{gathered}
\eeqq 
where $Z(z', z''; h)$ is smooth on $\p\mcm\times \p\mcm$ and the $L^\infty$ norm is bounded for small $h.$ The term comes from the stationary phase argument and involves second order derivatives of $X^{W_{44}}\delta V$. \\

 
Next, we extract $X^{W_{44}}\delta V$ from the first integral in \eqref{eq-x44}. For simplicity, we drop the scalar factors and consider 
\beqq\label{eq-estf1} 
\begin{gathered}
\tilde X_{44} \doteq \int_{\p \mcm}\int_{\p \mcm} e^{i|z' - z''|/h} X^{W_{44}}\delta V(z', z'') f(z')  \tilde f(z'')dz'dz''. 
 \end{gathered}
\eeqq
Note that normally the integral is very small for $h$ small because of the high oscillation. But we have the freedom to choose the boundary data    to suspend the oscillation. In particular, we will choose $f, \tilde f \in C^\infty(\p\mcm)$ so that they approximate the delta function supported at $z_0', z_0''\in \p \mcm$ plus certain oscillations. 

We fix $z_0', z_0''\in \p \mcm, z_0' \neq z_0''$. Near $z_0'$, we use local coordinates 
\beqq\label{eq-xco}
\begin{gathered}
x = (x_1, x_2, x_3) \in \mbr^3 
\text{ so that $x(z_0') = 0$ and $\p \mcm$ is given by $x_1 = 0$.}
\end{gathered}
\eeqq 
%In particular, we can arrange the coordinate so that the Jacobian $J(x)$ of the change of variable satisfies $J(0)=1.$ 
Near $z_0''$, we use local coordinates 
\beqq\label{eq-yco}
y = (y_1, y_2, y_3)\in \mbr^3 \text{ so that $y(z_0'') = 0$ and $\p \mcm$ is given by $y_1 = 0$.}
\eeqq 
We denote the Jacobians of theses changes of variables by $J(x)$ and $\tilde J(y)$. 
%We also arrange the Jacobian $\tilde J(y)$ so that $\tilde J(0) =1$. 
We write $x' = (x_2, x_3), y' = (y_2, y_3)$ as coordinates on the boundary.  Let $r>0$. We consider $\mcu = \{x'\in \mbr^2: |x'|< r\},$ $\tilde \mcu = \{y'\in \mbr^2: |y'|<r\}$ as neighborhood of $z_0', z_0''$ on $\p\mcm.$ With these choices of coordinates, we see that 
\beq
(X^{W_{44}} \delta V)(z'(x'), z''(y')) = (X^{W_{44}} \delta V)(z_0', z_0'') + \psi(x', y'), 
\eeq
where $\psi$ is smooth on $\mcu\times \tilde \mcu$ and $\psi(0, 0) = 0$. 
Next, we can write 
%\beq
%z'(x') = z_0' + \alpha(x'), \quad z''(y') = z_0'' + \beta(y')
%\eeq
%where $\alpha, \beta$ are smooth on $\mcu$, $\tilde \mcu$ respectively, and $\alpha(0) = \beta(0) = 0$. Consider  
\beq
|z'(x') - z''(y')| = |z_0' - z_0''| + \sigma(x') + \tilde \sigma(y') + \sigma_0(x', y'), 
\eeq
in which $\sigma, \tilde \sigma$ are linear in $x', y'$ respectively with $\sigma(0) = \tilde \sigma(0) = 0$, and $\sigma_0$ is smooth on $\mcu\times \tilde \mcu$ such that $|\sigma_0(x', y')|\leq C(|x'||y'| + |x'|^2 + |y'|^2).$ 

Let $\phi$ be a $C_0^\infty(\mbr^{2})$ function supported in $B(0, 1) = \{x'\in \mbr^2: |x'|<1\}$ such that $\phi\geq 0$ and $\int_{\mbr^{2}} \phi(x')dx' = 1$. For $N >0,$ we let $\Phi_N(x') = N^2 \phi(N x')$. Then $\int_{\mbr^2} \Phi_N(x') dx' = 1$. It is known that $\lim_{N\rightarrow \infty} \Phi_N = \delta_0$ in the sense of distributions.  For $N > 0 $, we define   
\beqq\label{eq-f1}
f_N(x') =  N^2\phi(N x') e^{-i\sigma(x')/h},  
\eeqq
and 
\beqq\label{eq-f2}
\tilde f_N(y') = N^2\phi(N y')e^{-i\tilde \sigma(y')/h}. 
\eeqq
Note that both $f_N$ and  $\tilde f_N$ are smooth and are supported on $\mcu, \tilde \mcu$ respectively. 
 
Now for the integral in \eqref{eq-estf1}, we use $f = f_N, \tilde f = \tilde f_N$ and  make changes of variables to get that 
\beqq\label{eq-x44-1}
\begin{split} 
\tilde X_{44} = &\int_{\mbr^{2}}\int_{\mbr^{2}} e^{i|z_0' - z_0''|/h}e^{i\sigma_0(x', y')/h} [(X^{W_{44}}\delta V)(z_0', z_0'')  + \psi(x', y')] \\
&\cdot N^2 \phi(N x')  N^2  \phi(Ny') J(0, x')\tilde J(0, y') dx' dy'\\ 
= & \int_{\mbr^{2}}\int_{\mbr^{2}} e^{i|z_0' - z_0''|/h}e^{i\sigma_0(x'/N, y'/N)/h} [(X^{W_{44}}\delta V)(z_0', z_0'') + \psi(x'/N, y'/N)] \\
&\cdot   \phi(x')   \phi(y') J(0, x'/N)\tilde J(0, y'/N) dx' dy'\\ 
= &\int_{\mbr^{2}}\int_{\mbr^{2}}  e^{i|z_0' - z_0''|/h} (X^{W_{44}}\delta V)(z_0', z_0'')   \phi(x')   \phi(y')  J(0)\tilde J(0) dx' dy'\\
&+ \int_{\mbr^{2}}\int_{\mbr^{2}}  e^{i|z_0' - z_0''|/h} (X^{W_{44}}\delta V)(z_0', z_0'')   \phi(x')   \phi(y') (J(0, x'/N)\tilde J(0, y'/N) - J(0)\tilde J(0))  dx' dy'\\
&+ \int_{\mbr^{2}}\int_{\mbr^{2}} e^{i|z_0' - z_0''|/h} (e^{i\sigma_0(x'/N, y'/N)/h} - 1) (X^{W_{44}}\delta V)(z_0', z_0'')    \phi(x')   \phi(y')   J(0, x'/N)\tilde J(0, y'/N)  dx' dy'\\ 
&+ \int_{\mbr^{2}}\int_{\mbr^{2}} e^{i|z_0' - z_0''|/h} e^{i\sigma_0(x'/N, y'/N)/h} \psi(x'/N, y'/N)   \phi(x') \phi(y') J(0, x'/N)\tilde J(0, y'/N)  dx' dy'. 
\end{split}
\eeqq
Choosing $N = h^{-2/3}$, we see that $N^2 h = h^{-1/3}$ and 
\beq
|\sigma_0(x'/N, y'/N)|/h\leq C/(N^2 h)\leq C h^{1/3}. 
\eeq
So the high oscillation   is suspended. We can estimate the last three integrals in \eqref{eq-x44-1} directly by noticing that 
\beq
\begin{gathered}
|J(0, x'/N)\tilde J(0, y'/N) - J(0)\tilde J(0)|\leq Ch^{2/3}, \\
|e^{i\sigma_0(x'/N, y'/N)/h} - 1| \leq Ch^{1/3}, \\
|\psi(x'/N, y'/N)|\leq C h^{2/3}.
\end{gathered}
\eeq
Thus, we get from \eqref{eq-x44-1} that 
\beq
 -e^{i|z_0' - z_0''|/h} X^{W_{44}}\delta V(z_0', z_0'')J(0)\tilde J(0) = \tilde X_{44} + O(h^{1/3}).  
\eeq
The second integral in \eqref{eq-x44} can be analyzed similarly but with an extra $h$ factor. Thus we conclude that 
\beqq\label{eq-rayest}
 -\frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} \beta_1 \tilde \beta_1 X^{W_{44}}\delta V(z_0', z_0'')J(0)\tilde J(0)  = hX_{44} + O(h^{1/3}), 
\eeqq 
where
\beqq\label{eq-beta}
\begin{gathered}
\beta_1 = (\nu_{z'}\cdot  \frac{(z'' - z')}{|z' - z''|})|_{z' = z_0', z'' = z''_0} = -\p_{\nu_{z'}} |z'' - z'| |_{z' = z_0', z'' = z''_0} < 0, \\
\tilde \beta_1 =  (\nu_{z''}\cdot  \frac{(z' - z'')}{|z' - z''|})|_{z' = z_0', z'' = z_0''}  = -\p_{\nu_{z''}} |z' - z''| |_{z' = z_0', z'' = z''_0} < 0. 
\end{gathered}
\eeqq
We remark that we have arranged the weighted X-ray transform so that the weight is positive. This is done purposely and we will keep track of the scalar factor in front of $X^{W_{44}}$. 


%========================+%
\section{The leading order terms}\label{sec-pf-main1}
The leading order terms in \eqref{eq-mainid1} for small $h$ are contained in $X_{44}, X_{11}, X_{14}$,$ X_{41}$ and $X_{22}, X_{21},$ $X_{12}, X_{24}, X_{42}$. For each term, we will get an expansion similar to \eqref{eq-rayest} involving some weighted X-ray transform. We first analyze $X_{11}, X_{14}, X_{41}$ which requires some new ingredients. The rest of the analysis will follow the same pattern. Note that these three terms all contain $I_1(z)$ or $\tilde I_1(z)$. We recall that 
\beq 
I_1(z) = -\int_{\mcm} G_0(z, z'; h)u(z') dz'. 
\eeq
Here, $u$ is the solution of $-\lap u + Vu = 0$ in $\mcm$ with Dirichlet data $f$ given in \eqref{eq-f1}. Our idea is to construct an approximate solution that concentrates near boundary points. Then the integral in $I_1(z)$ is only effective near the boundary. These techniques are often used in the boundary determination problem, see Section 3.1 of \cite{FSU}. 

\subsection{The concentrating solution}\label{sec-con}
We still use the local coordinates \eqref{eq-xco} near $z_0'\in \p \mcm$  and write 
\beq
f =    N^2 \phi(N x') e^{-i\sigma(x')/h}, \quad N = h^{-2/3}. 
\eeq
The analysis near $z_0''\in \p \mcm$ is the same. Because $\sigma$ is linear in $x$, we can write $\sigma(x') = \alpha \cdot x'$ where $\alpha \in \mbr^2$ is a tangent vector at $0.$ Note that $\alpha$ can be regarded as a function of $z_0', z_0''$ and $\alpha$ could be zero! In fact, this is unavoidable: for any $z_0'\in \p \mcm$, there exists $z_0''\in \p\mcm$ such that the corresponding $\alpha = 0$ at $z_0''$. For the construction of concentrating solutions, it is crucial that $\alpha$ is non-zero. We can show that this happens ``generically". 
\begin{lemma}\label{lm-geo}
There is a dense open set $\mcs$ of $\p\mcm\times \p\mcm$ such that for $(z_0', z_0'')\in   \mcs$, we have $\alpha \neq 0$.  
\end{lemma}
\bpf
For fixed $z'_0, z''_0\in \p\mcm$ and $z_0'\neq z_0''$, we let $\mcv', \mcv''$ be small neighborhood of $z_0', z_0''$ on $\p\mcm$, respectively such that $\mcv'\cap \mcv''=\emptyset.$  
Consider the function $\Theta(z', z'') = \nabla_{z'}r(z', z'')$ for $z'\in \mcv', z''\in \mcv''$. We know that $\Theta$ is continuous on $\mcv'\times \mcv''$ so the set $\mcw \doteq \{(z', z'')\in \mcv'\times \mcv'': \Theta(z', z'') =0\}$ is closed. We claim that the set has no interior. We argue by contradiction and assume without loss of generality that $\Theta(z', z'') = 0$ for $z'$ close to $z_0'$ and $z''$ close to $z_0''$. Then $\Theta(z', z_0'') = 0$ for $z'$ in some neighborhood $\tilde \mcv'$ of $z'_0$ on $\p\mcm$. Thus $r(z', z_0'') = \text{constant}$ for $z'\in \mcv'$, and $\mcv'$ must be a part of a sphere centered at $z_0''$. Then for $z''$ close to $z_0''$ on $\p\mcm$ with $z''\neq z_0''$, we cannot have $\Theta(z', z'')$ identically zero for $z'\in \tilde\mcv'$. We reached a contradiction. Therefore, we proved that the set $\mcw$ is a closed set with empty interior so the complement $\mcw^c$ is a dense open set in $\mcv'\times \mcv''$. We can finish the proof of the lemma by repeating the argument for $(z_0', z_0'')\in \p\mcm\times \p\mcm$ and using the fact that $\p\mcm$ is compact. 
\epf

Below, we assume that $(z_0', z_0'')$ is a pair of boundary points such that $\alpha\neq 0$ and $\tilde \alpha \neq0$. Here, $\tilde \alpha$ is defined in the same way as $\alpha$. It is possible to construct directly approximate solutions near boundary points that vanish to high order in $h$ for $h>0$ small, but it involves lengthy calculations. Instead, we adapt Proposition 3.4.2 of \cite{FSU} to our setting. Then we rescale to get the approximate solution.  Near $z_0',$ we use coordinates \eqref{eq-xco}. In fact, as discussed in Proposition 3.3.1 and Lemma 3.3.5 of \cite{FSU}, we can further choose the coordinates so that $-\lap$ is transformed to 
\beqq\label{eq-gamma}
P = -\nabla\cdot (\gamma \nabla), \text{ where $\gamma$ is of the form }  \gamma = \begin{pmatrix}
c & 0\\
0 & cH
\end{pmatrix}, 
\eeqq
where $c$ is a positive scalar function and $H$ is a positive definite matrix. For $r>0$, we let 
\beq
\begin{gathered}
\Omega_r = B(0, r)\cap \mcm = \{x\in B(0, r): x_1>0\},\\
\Gamma_r = B(0, r)\cap \p \mcm = \{x\in B(0, r): x_1 = 0\}.
\end{gathered}
\eeq 
The following is adapted from Proposition 3.4.2 of \cite{FSU}. 
\begin{lemma}\label{lm-uapp0}
Let $K>0$. For $L>1$ large and some small $\delta >0$, there is $v_L \in C^\infty(\overline\mcm)$ satisfying
\beq
v_L(0, x') = e^{-iL x'\cdot \alpha} \eta(x') \text{ on } \Gamma_r, \quad \supp v_L \subset [0, \delta]\times \Gamma_r  
\eeq
and 
\beqq\label{eq-vest}
\|v_L\|_{H^1(\mcm)}\leq C L^{-1/2}, \quad 
\|P v_L\|_{L^2(\mcm)}\leq C L^{-K + 3/2}, 
\eeqq
where $C$ is independent of $L$. Further, the function takes the form 
\beq
v_L(x) = e^{L \Phi(x)}(a_0(x) + L^{-1} a_{-1}(x) + \cdots + L^{-K} a_{-K}(x)), 
\eeq
where in local coordinates so \eqref{eq-gamma} holds, we have 
\begin{enumerate}
\item $\Phi$ is a smooth complex function satisfying for some $\tau>0$
\beq
\begin{gathered}
\Phi(0, x') = -i x'\cdot \alpha, \quad \p_{x_1} \Phi(0, x') = -\zeta(x'), \text{ for } x'\in \Gamma_r \\
\re(\Phi(x_1, x')) \leq -\tau x_1, \text{ for } (x_1, x') \in [0, \delta]\times \Gamma_r, 
\end{gathered}
\eeq
where $\zeta(x')$ is a positive function depending on $\alpha.$   In particular, $\zeta(x') = 0$ if $\alpha = 0.$

\item $a_0, \cdots, a_{-K}$ are smooth complex functions independent of $L$, supported in $[0, \delta]\times \Gamma_r$ and they satisfy
\beq
a_0(0, x') = \eta(x'), \quad a_{-l}(0, x') = 0, l\geq 1, \quad x'\in \Gamma_r. 
\eeq
\end{enumerate}
\end{lemma}

Then we prove
\begin{lemma}\label{lm-uapp}
Let $u$ be the unique solution  of \eqref{eq-dtn}. Then we can write $u = w_0 + w_1$ where 
\begin{enumerate}
\item $w_0\in C^\infty(\overline \mcm)$ is supported near $z_0'\in \p \mcm$ and 
\beqq\label{eq-w0est}
\|w_0\|_{L^2(\mcm)}\leq C h^{-1/6}
\eeqq
for $h$ small. More precisely, in local coordinate \eqref{eq-xco}, we have 
\beq
w_0(0, x') = N^2 e^{-ix'\cdot \alpha/h} \phi(N x') \text{ on } \Gamma_{h^{2/3}r}, \quad \supp w_0 \subset  [0, h^{2/3}\delta]\times \Gamma_{h^{2/3}r}
\eeq
and the function takes the form 
\beq
w_0(x) = e^{\Phi(N x)/h^{1/3}}N^2 (a_0(Nx) + h^{1/3} a_{-1}(Nx) + \cdots + h^{2K/3} a_{-K}(Nx)),
\eeq
for some integer $K\geq 20$. Here, $\Phi$ is a smooth complex function satisfying for some $\tau>0$
\beq
\begin{gathered}
\Phi(0, N x') = -i N x'\cdot \alpha, \quad \p_{x_1} \Phi(0, N x') = -N \zeta(Nx'), \text{ for } x'\in \Gamma_{h^{2/3}r} \\
\re(\Phi(Nx_1, N x')) \leq -N \tau x_1, \text{ for } x\in [0, h^{2/3}\delta)\times \Gamma_{h^{2/3}r}
\end{gathered}
\eeq
and $a_0, \cdots, a_{-K}$ are smooth complex functions independent of $N$, supported in $[0, \delta]\times \Gamma_r$ and they satisfy
\beq
a_0(0, N x') = \phi(N x'), \quad a_{-l}(0, N x') = 0, l\geq 1, \quad x'\in \Gamma_{h^{2/3}r}. 
\eeq

\item $w_1\in C^\infty(\overline \mcm)$ and it satisfies 
\beqq\label{eq-w1est}
\|w_1\|_{H^1(\mcm)} \leq C h^{4}. 
\eeqq
\end{enumerate}
\end{lemma}
\bpf
We take $L = h^{-1/3}$ in Lemma \ref{lm-uapp0} and make a change of variable $x \rightarrow Nx = h^{-2/3}x$ with $|x|< h^{2/3}\delta$. Let $w_0(x)$ be $N^2 v_L(Nx)$ in Lemma \ref{lm-uapp0}. We get the claims about  $w_0$ in (1).  

To estimate $w_1$, we know that  $w_1$ satisfies 
\beqq\label{eq-v1}
\begin{gathered}
-\lap w_1 + Vw_1 =  \lap w_0 - V w_0 \text{ in } \mcm\\
w_1 = 0 \text{ on } \p \mcm. 
\end{gathered}
\eeqq
By the standard elliptic type estimate, we get 
\beq
\|w_1\|_{H^1(\mcm)}\leq  C  \|\lap w_0 - V w_0\|_{L^2(\mcm)} \leq C\|\lap w_0\|_{L^2(\mcm)} + C\|V w_0\|_{L^2(\mcm)}
\eeq
Because $V$ is supported away from $\p\mcm$ and $w_0$ is supported near $\p \mcm$, we know that 
$\|V w_0\|_{L^2(\mcm)} = 0$ for $h$ sufficiently small.  Then we use \eqref{eq-vest} to find that 
\beq
\|\lap w_0\|_{L^2(\mcm)}\leq C h^{K/3 - 1/2} h^{-5/3} = C h^{K/3 - 13/6}. 
\eeq
We take $K\geq 19$ to get $\|w_1\|_{H^1(\mcm)} \leq C h^4$. 
\epf



%==============================+%
\subsection{The analysis of $X_{11}$} \label{eq-X11}
We use Lemma \ref{lm-uapp} to write $u = w_0 + w_1$. Applying the construction of concentrating solutions near  $z_0''$, we can find a decomposition $\tilde u = \tilde w_0 + \tilde w_1$ where $\tilde w_0, \tilde w_1$ have the same properties  as $w_0, w_1$ in Lemma \ref{lm-uapp}. Then we can write 
\beq
\begin{gathered}
X_{11} = \sum_{l = 1}^4 X_{11, l}, \text{ where }\\
X_{11, 1} = \int_{\mcm} \int_{\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h)w_0(z')\tilde w_0(z'')dzdz'dz'', \\
X_{11, 2} = \int_{\mcm} \int_{\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h) w_0(z')\tilde w_1(z'')dzdz'dz'', \\
X_{11, 3} = \int_{\mcm}  \int_{\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h) w_1(z')\tilde w_0(z'')dzdz'dz'',\\
X_{11, 4} = \int_{\mcm}  \int_{\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h) w_1(z')\tilde w_1(z'')dz dz' dz''.  
\end{gathered}
\eeq

First we can easily estimate the last three terms with the help of Lemma \ref{lm-uapp}. For example, we have 
\beq
\begin{split}
|X_{11, 2}| &\leq C |\int_{\mcm} \int_{\mcm} \int_{\mcm} |\delta V(z)| h^{-4} \frac{1}{|z - z'||z-z''|} |w_0(z')| |\tilde w_1(z'')| dzdz'dz''\\
&\leq C \big|\int_{\mcm} \int_{\mcm} h^{-4} |w_0(z')| |\tilde w_1(z'')| dz'dz''\big|, 
\end{split}
\eeq
where we used that for the integration in $z$, the singularities in the integrand is integrable. Then we continue to get 
\beq
\begin{gathered}
|X_{11, 2}|  \leq C h^{-4} \|w_0\|_{L^2} \|\tilde w_1\|_{L^2} \leq C h^{-4+4-1/6}\leq C h^{-1/6}, 
\end{gathered}
\eeq
where we used the estimates in Lemma \ref{lm-uapp}. 
By the same argument, we get 
\beq
|X_{11, 3}|\leq C h^{-1/6}, \quad |X_{11, 4}|\leq C h^{4}.  
\eeq




To analyze $X_{11, 1}$, we note that $w_0, \tilde w_0$ are supported near $\p \mcm$ by Lemma \ref{lm-uapp}. By taking $h>0$ sufficiently small, we can assume that $\supp w_0 \cap \supp \delta V = \emptyset$ and $\supp \tilde w_0\cap \supp \delta V = \emptyset.$ Then we can use the stationary phase argument in Section \ref{sec-pf-main}. In particular, we write  
\beq
\begin{gathered}
X_{11, 1}  =  \int_{\mcm}\int_{\mcm} \frac{1}{16\pi^2 h^4} (\int_\mcm \delta V(z)   e^{i|z - z'|/h} \frac{ 1}{|z - z'|}   e^{i|z - z''|/h}\frac{1}{|z - z''|} dz) 
 w_0(z') \tilde w_0(z'') dz' dz''. 
\end{gathered}
\eeq
Let $I_{11}(z', z'')$ be the integration in $z$ where $z', z''$ are in the support of $w_0, \tilde w_0.$ We use the stationary phase argument as in Section \ref{sec-pf-main} and follow the notations there to get 
\beq
\begin{gathered}
I_{11}(z', z'')  
 =  e^{\frac{i}{h}|z' - z''|}h  \int_{0}^{|z' - z''|} \det(\Phi''(a, 0)/(2\pi i))^{-\ha}  F_{11}(a, 0)   da  
+ O_{L^\infty}(h^{2}), 
\end{gathered}
\eeq
where 
\beq
F_{11}(a, 0) =   \delta V(\gamma_{z', z''}(a)) \frac{1}{ a(|z' - z''| - a)}(\frac{1}{16\pi^2h^4})(2\pi i). 
\eeq
Along the line segment $\gamma_{z', z''}(t)$, we define  
\beq
W_{11}(z', z'', t) =  (\frac{1}{t} + \frac{1}{|z' - z''| - t} )^{-1}  \frac{1}{t(|z' - z''| -t)}  
\eeq
and let $X^{W_{11}}$ be the X-ray transform with weight $W_{11}$. We proved that 
\beqq\label{eq-x11}
\begin{gathered}
X_{11, 1} = \int_{\mcm}\int_{\mcm} e^{i|z' - z''|/h} h^{-3} \frac{ (2\pi i)}{16\pi^2}  (X^{W_{11}}\delta V(z', z'')  + O_{L^\infty}(h)) w_0(z')  \tilde w_0(z'')dz'dz''. 
 \end{gathered}
\eeqq  
Note that the integration is not on $\p\mcm$ but in a small neighborhood of $\p \mcm$. Because $w_0, \tilde w_0$ decays exponentially fast away from $\p \mcm$, we can still reduce the integral to the boundary with a small error term. We show the reduction for the integration in $z'$. 

We compute in local coordinate \eqref{eq-xco} that 
\beq 
\begin{gathered}
 \int_{\mcm}  e^{i|z' - z''|/h}  (X^{W_{11}} \delta V(z', z'')+O_{L^\infty}(h))w_0(z')  dz' \\
  = \int_{\mbr^2} \int_{0}^{\delta} e^{i|z'(x) - z''|/h}  (X^{W_{11}} \delta V(z'(x), z'')+O_{L^\infty}(h)) 
 N^2 \phi(Nx') e^{\Phi(Nx)/h^{1/3}} J(x) 
  dx_1 dx' \\
  +   \int_{\mbr^2} \int_{0}^{\delta} e^{i|z'(x) - z''|/h}  (X^{W_{11}} \delta V(z'(x), z'')+O_{L^\infty}(h)) 
O_{L^\infty}(h) e^{\Phi(Nx)/h^{1/3}} J(x) 
  dx_1 dx'.
 \end{gathered}
\eeq
We remark that the higher order expansion terms $a_k, k\geq 1$ in $w_0$ belongs to $O_{L^\infty}(h)$.  We are interested in the integration in $x_1$ for the moment. We write 
 \beq
|z'(x) - z''| = |(0, x') - z''| + \kappa(x_1, x'). 
\eeq
Then we compute 
\beq 
\begin{split}
&  \int_{0}^{\delta} e^{i|(0, x') - z''|/h} e^{i\kappa(x_1, x')/h}  (X^{W_{11}} \delta V(z', z'')N^2 \phi(Nx')  +O_{L^\infty}(h)) e^{\Phi(N x)/h^{1/3}} J(x) dx_1 \\
   =  & \int_{0}^{\delta}  e^{i|(0, x') - z''|/h}  (X^{W_{11}} \delta V(z', z'')N^2 \phi(Nx')  +O_{L^\infty}(h))   \\
  & \cdot  \frac{ J(x)}{\p_1 \Phi(N x) /h^{1/3} + i \p_1\kappa'(x_1, x')/h} d e^{i\kappa(x_1, x')/h + \Phi(Nx)/h^{1/3}} \\
    =  &  -e^{i|(0, x') - z''|/h - i x'\cdot \alpha/h}  (X^{W_{11}} \delta V((0, x'), z'') N^2 \phi(Nx') +O_{L^\infty}(h)) \\
& \cdot \frac{ h J(0, x')}{-  \zeta(N x') + i\p_1\kappa(0, x')}  + O_{L^\infty}(h^2) 
 \end{split}
\eeq
In the last line, the first term comes from the boundary term at $x_1 = 0$ after integration by parts. We actually need to use the results of $\Phi$ in Lemma \ref{lm-uapp} at $x_1 = 0$. The second term comes from two sources. The first is the boundary term at $x_1 = \delta$ and $e^{\Phi(Nx)/h^{1/3}}$ decays exponentially fast for $h$ small, see Lemma \ref{lm-uapp}. The second source is the integral after integration by parts which is $O_{L^\infty}(h)$. But we can repeat the integration by parts one more time to get $O_{L^\infty}(h^2)$. We also remark that $\p_1\kappa(0, x')<0$. 

Putting the integral in $x_1$ back to $X_{11, 1}$ and repeating the argument for the integration in $z''$, we get 
 \beq 
\begin{split}
X_{11, 1} &= \int_{\mbr^2} \int_{\mbr^2}   e^{i|(0, x') - (0, y')|/h - i\sigma(x')/h - i\tilde \sigma(y')/h} \frac{1}{(-\zeta(Nx') + i\p_1\kappa(0, x'))( -\tilde \zeta(Ny') + i\p_1\tilde\kappa(0, y'))} \\
&\cdot \frac{ (2\pi i)}{16\pi^2}   \frac{1}{h}  X^{W_{11}} \delta V((0, x'), (0,y')) N^2\phi(Nx') N^2\tilde \phi(Ny') J(0, x')\tilde J(0, y') dx'dy'
\\
&+  \int_{ \mcm} \int_{ \mcm}  O_{L^\infty}(1) w_0(z')  \tilde w_0(z'') dz'dz''. 
\end{split}
\eeq
The second integral can be estimated by the $L^2$ norms of $w_0, \tilde w_0$    as 
\beq
\begin{gathered}
|\int_{ \mcm} \int_{ \mcm}  O_{L^\infty}(1) w_0(z')  \tilde w_0(z'') dz'dz''  |\leq C \|w_0\|_{L^2}\|\tilde w_0\|_{L^2} \leq C h^{-1/3}. 
\end{gathered}
\eeq
For the first integral, we can extract the weighted X-ray transform as in the analysis for \eqref{eq-x44}.  
Thus we get 
\beq
\frac{1}{(-\beta_0 + i \beta_1)(-\tilde \beta_0+ i\tilde \beta_1)} \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} X^{W_{11}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{11, 1} + O(h^{2/3}), 
\eeq
where we note that $\p_1\kappa(0),  \p_1 \tilde \kappa(0)$ are actually $\beta_1, \tilde \beta_1$  defined in \eqref{eq-beta} and they are both negative. Also, $\beta_0 = \zeta(0), \tilde \beta_0 = \zeta(0)$ are both positive. This completes the analysis for $X_{11, 1}$. \\


Finally, we can summarize all the estimates for $X_{11, k}, k = 1, 2, 3, 4$ and  conclude that  
\beqq\label{eq-rayest11}
\frac{1}{(-\beta_0 + i\beta_1) (-\tilde\beta_0 + i\tilde \beta_1)} \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} X^{W_{11}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{11} + O(h^{2/3}).
\eeqq

%========================================%
\subsection{The analysis of $X_{14}$ and $X_{41}$}\label{sec-x14}
Note that there is an apparent symmetry in $X_{14}$ and $X_{41}$ so we only analyze $X_{14}$ below. Again, we use Lemma \ref{lm-uapp} to write 
\beq
\begin{gathered}
X_{14} = X_{14, 1} + X_{14, 2}, \text{ where }\\
X_{14, 1} =  -\int_{\mcm} \int_{\mcm} \int_{\p \mcm} h^2 \delta V(z)G_0(z, z'; h) G_{\nu, 2}(z, z''; h) w_0(z') \tilde f(z'')dzdz'dz'', \\ 
X_{14, 2} =  -\int_{\mcm} \int_{\mcm} \int_{\p \mcm} h^2 \delta V(z)G_0(z, z'; h) G_{\nu, 2}(z, z''; h) w_1(z') \tilde f(z'')dzdz'dz''. 
\end{gathered}
\eeq
The second term can be estimated by 
\beq
\begin{split}
|X_{14, 2}|& \leq C h^{-3}  \int_{\mcm} \int_{\mcm} \int_{\p \mcm}  |\delta V(z)| \frac{1}{|z - z'|} \frac{ |\nu_{z''} \cdot (z - z'')|  }{|z - z''|^2}  |w_1(z')| |\tilde f(z'')| dzdz'dz''\\
& \leq Ch^{-3} \int_{\mcm}  \int_{\p \mcm} |w_1(z')| |\tilde f(z'')|  dz'dz'' \leq Ch^{-3} \|w_1\|_{L^2} \|\tilde f\|_{L^2} \leq C h^{1/3}. 
\end{split}
\eeq
Here, we used that for $\tilde f$ in \eqref{eq-f2},  
\beq
\begin{gathered}
\|\tilde f\|_{L^2}^2 \leq C \int_{\mbr^2} |N^2\phi(N y')|^2dy'= C \int_{\mbr^2}N^2 \phi^2(y') dy'  = C N^2. 
\end{gathered}
\eeq
Thus with the choice $N = h^{-2/3}$, we have $\|\tilde f\|_{L^2} \leq C h^{-2/3}.$ 

For the first term, we write 
\beq
\begin{gathered}
X_{14, 1} =  \int_{\p \mcm}  \int_{\mcm} h^{-3} \frac{i}{16\pi^2}  (\int_{\mcm} \delta V(z) e^{i|z - z'|/h}\frac{1}{|z - z'|} e^{i|z - z''|/h}\frac{\nu_{z''}\cdot (z - z'')}{|z - z''|^2} dz) w_0(z') \tilde f(z'') dz'dz''. 
\end{gathered}
\eeq
The integration in $z$ can be analyzed exactly as in $X_{11, 1}$ or $X_{44}$.  
Let $I_{14}(z', z'')$ be the integration in $z$ where $z'$ is in the support of $w_0$ and $z_0''\in \p \mcm.$ We use the stationary phase argument to get 
\beq
\begin{gathered}
I_{14}(z', z'')  
 = e^{\frac{i}{h}|z' - z''|}h  \int_{0}^{|z' - z''|} \det(\Phi''(a, 0)/(2\pi i))^{-\ha}  F_{14}(a, 0)   da   + O_{L^\infty}(h^{2}), 
\end{gathered}
\eeq
where 
\beq
F_{14}(a, 0) =   \delta V(\gamma_{z', z''}(a)) \frac{1}{ a (|z' - z''| - a)} (\nu_{z''}\cdot \frac{(z' - z'')}{|z' - z''|})(\frac{i}{16\pi^2 h^3})(2\pi i). 
\eeq
Along the line segment $\gamma_{z', z''}(t)$, we define  
\beqq\label{eq-w14}
W_{14}(z', z'', t) =  (\frac{1}{t} + \frac{1}{|z' - z''| - t} )^{-1}  \frac{1}{ t (|z' - z''| - t)}. 
\eeqq
Note that $W_{14}$ is positive. Let $X^{W_{14}}$ be the X-ray transform with weight $W_{14}$. We proved that 
\beq 
\begin{gathered}
X_{14, 1} = \int_{\mcm}\int_{\p \mcm} e^{i|z' - z''|/h} h^{-2} \frac{ i (2\pi i)}{16\pi^2}(\nu_{z''}\cdot \frac{(z' - z'')}{|z' - z''|})  (X^{W_{14}}\delta V(z', z'')  + O_{L^\infty}(h)) w_0(z')  \tilde f(z'')dz'dz''. 
 \end{gathered}
\eeq
Now we can use the argument for $X_{11, 1}$ to reduce the integral in $z'$ to the boundary. In fact, the calculation is the same and we get in local coordinate that 
 \beqq\label{eq-x141}
\begin{split}
X_{14, 1} = & \int_{\mbr^2} \int_{\mbr^2}   e^{i|(0, x') - (0, y')|/h} \frac{-i}{-\zeta(Nx') + i \p_1\kappa(0, x')} \frac{ (2\pi i)}{16\pi^2}   \frac{1}{h}  
((\nu_{z''}\cdot \frac{(z' - z'')}{|z' - z''|})|_{z' = (0, x'), z'' = (0, y')} )\\
& \cdot X^{W_{14}} \delta V((0, x'), (0, y'))  N^2\phi(Nx')  \tilde f(y') J(0, x')\tilde J(0, y')dx'dy'  \\
& +  \int_{ \mcm} \int_{ \p \mcm}  O_{L^\infty}(1) w_0(z')  \tilde f_0(z'') dz'dz''. 
\end{split}
\eeqq
The second term can be estimated by 
\beq
|\int_{ \mcm} \int_{ \p \mcm}  O_{L^\infty}(1) w_0(z')  \tilde f_0(z'') dz'dz''  |\leq C \|w_0\|_{L^2}\|\tilde f\|_{L^2} \leq C h^{-1/6} h^{-2/3} = Ch^{-5/6}. 
\eeq

For the first integral in \eqref{eq-x141}, we can extract the weighted X-ray transform as in \eqref{eq-x44} and get 
\beq
\frac{-i\tilde \beta_1}{-\beta_0+ i\beta_1} \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h}  X^{W_{14}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{14, 1} + O(h^{1/6}).
\eeq
Together with the estimate for $X_{14, 2}$, we have 
\beqq\label{eq-rayest14}
\frac{-i\tilde \beta_1}{-\beta_0 + i\beta_1} \frac{ (2\pi i)}{16\pi^2}   e^{i|z_0' - z_0''|/h} X^{W_{14}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{14} + O(h^{1/6}).
\eeqq


The analysis for $X_{41}$ is identical. We will not repeat it. The final result is that 
\beqq\label{eq-rayest41}
\frac{-i\beta_1}{-\tilde \beta_0 + i\tilde \beta_1} \frac{ (2\pi i)}{16\pi^2}   e^{i|z_0' - z_0''|/h} X^{W_{41}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{41} + O(h^{1/6}).
\eeqq 
where $W_{41}$ is a positive weight function. Now we observe that all the weights $W_{44} = W_{11} = W_{14} = W_{41}$ are equal. Hereafter, we denote the weight by $W.$ However, it is helpful to write them as $W_\bullet$ to keep track of the terms. 


%========================================%
\subsection{The analysis of $X_{22}$ and $X_{24}, X_{42}, X_{12}, X_{21}$}\label{sec-x22}
We first recall from \eqref{eq-X} and \eqref{eq-Is} that  
\beq
X_{22} = \int_\mcm \delta V(z) I_2(z) \tilde I_2(z) dz, 
\eeq
in which 
\beq
I_2(z) = - \int_{\p\mcm} h^2 G_0(z, z'; h)\p_\nu u_0(z') dz', \quad \tilde I_2(z) = - \int_{\p\mcm} h^2 G_0(z, z'; h)\p_\nu \tilde u_0(z') dz'. 
\eeq
Roughly speaking, $I_2, \tilde I_2$ should be treated as $I_4, \tilde I_4$. In fact, for our choice of the boundary data $f, \tilde f$ in \eqref{eq-f1}, \eqref{eq-f2}, we can use the concentrating solution to find that $\p_\nu u_0$ and $\p_\nu \tilde u_0$ on $\p\mcm$ essentially only differ from $f, \tilde f$ by some scalar factors plus an $h^{-1}$ factor. The $h^{-1}$ factor is the reason why $X_{22}$ contributes to the leading term as $h\rightarrow 0.$ Nevertheless, the analysis of $X_{22}$ eventually will follow the same calculation we did for $X_{44}$ once we figure out the differences. 

 
%According to Lemma \ref{lm-est}, we know that $\|\p_\nu v\|_{L^2(\p\mcm)}  = \|\p_\nu u - \p_\nu u_0\|_{L^2(\p\mcm)}\leq C h\|f\|_{L^2(\p\mcm)}.$ 
We replace $u_0$ by $u$ in $X_{22}$ and decompose it as 
\beq
\begin{gathered}
X_{22} = X_{22, 1} + X_{22, 2} + X_{22, 3} + X_{22, 4}, \text{ where }\\ 
X_{22, 1} = h^4 \int_{\p\mcm} \int_{\p\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h)\p_\nu u(z') \p_\nu \tilde u(z'')dzdz'dz'', \\
X_{22, 2} = h^4 \int_{\p\mcm} \int_{\p\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h) \p_\nu u(z') \p_\nu \tilde v(z'')dzdz'dz'', \\
X_{22, 3} = h^4 \int_{\p\mcm}  \int_{\p\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h) \p_\nu v(z') \p_\nu \tilde u(z'')dzdz'dz'',\\
X_{22, 4} = h^4 \int_{\p\mcm}  \int_{\p\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h) \p_\nu v(z') \p_\nu \tilde v(z'')dz dz' dz''.  
\end{gathered}
\eeq

First, consider $X_{22, 1}$. We use Lemma \ref{lm-uapp} to write $u = w_0 + w_1$. Then we see that 
\beq
\begin{split}
\p_{x_1}w_0(x) &= (\p_{x_1} e^{\Phi(N x)/h^{1/3}}) N^2 (a_0(Nx) + h^{1/3} a_{-1}(Nx) + \cdots + h^{2K/3} a_{-K}(Nx)) \\
&+ e^{\Phi(N x)/h^{1/3}} N^2 (\p_{x_1} a_0(Nx) + h^{1/3} \p_{x_1 }a_{-1}(Nx) + \cdots + h^{2K/3} \p_{x_1}a_{-K}(Nx)) \\
 &= \frac{\p_{x_1}\Phi(N x)}{h^{1/3}} e^{\Phi(N x)/h^{1/3}} N^2 (b_0(Nx) + h^{1/3} b_{-1}(Nx) + \cdots + h^{2K/3} b_{-K}(Nx))
\end{split}
\eeq
with proper $b_{-i}, i = 1, \cdots, K$. In particular, at $x_1 = 0$ we have 
\beqq\label{eq-neww0}
\begin{gathered}
\p_{x_1}w_0(x)|_{x_1 = 0}  = \frac{-\zeta(Nx')}{h} e^{-ix'\cdot \alpha/h} N^2  \phi(N x'). 
\end{gathered}
\eeqq
So $\p_{x_1}w_0$ has the same structure as $w_0$ at $\p\mcm$. We see that $\|\p_\nu w_0\|_{L^2(\p\mcm)}\leq C h^{-1}\|f\|_{L^2(\p\mcm)}$ and $\|\p_{\nu}w_1\|_{L^2(\p\mcm)}\leq Ch^{4}$ by Lemma \ref{lm-uapp}. Thus, we can replace $\p_\nu u$ in $X_{22, 1}$ by $\p_\nu w_0$ and $\p_\nu \tilde u$ by $\p_\nu \tilde w_0$ which will produce an error term of order $h^{-1/3}$ (in fact, order $h^{7/3}$). So it suffices to analyze 
\beq
\int_{\p\mcm} \int_{\p\mcm} \int_{\mcm} \delta V(z) G_0(z, z'; h)G_0(z, z''; h)\p_\nu w_0(z') \p_\nu \tilde w_0(z'')dzdz'dz''. 
\eeq
Because $\p_\nu w_0|_{\p\mcm}$ has the same structure as $f$, the same calculation for $X_{44}$ yields that 
\beqq\label{eq-rayest22-1}
 \beta_0 \tilde \beta_0  \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} X^{W_{22}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{22, 1} + O(h^{1/3}),
\eeqq
where $W_{22} = W$. We will not show the computations here. 

It remains to estimate $X_{22, j}, j = 2, 3, 4.$. From the equation \eqref{eq-dtn} for $u$, we know from standard elliptic estimate that $\|\p_\nu u\|_{L^2(\p\mcm)} \leq C\|f\|_{H^1(\p\mcm)}$. Using the form of  $f$ in \eqref{eq-f1}, we see that 
\beq
\|\p_\nu u\|_{L^2(\p\mcm)} \leq Ch^{-1}\|f\|_{L^2(\p\mcm)} \leq Ch^{-5/3}.
\eeq
Now we estimate $X_{22, 2}$. For the integration in $z$, we can apply the stationary phase argument to get a factor of $h$ so  
\beqq\label{eq-x22-2}
\begin{split}
X_{22, 2} &=  \int_{\p\mcm}\int_{\p \mcm} (\int_\mcm  \delta V(z) \frac{e^{i|z - z'|/h}e^{i|z - z''|/h}}{16\pi^2   |z - z'||z - z''|}  dz) \p_\nu  u (z')  \p_\nu \tilde v(z'') dz'dz''\\
 &=  \int_{\p\mcm} \int_{\p\mcm} e^{i|z' - z''|/h} h F(z', z'';h) \p_\nu u (z')  \p_\nu \tilde v(z'')dz'dz'', 
 \end{split}
\eeqq  
where $F$ is a smooth function in $z', z''$ and bounded in $h$. Write $\p_\nu u = \p_\nu w_0 + \p_\nu w_1$. We observe that $\|f\|_{L^2(\p\mcm)}$ is of order $h^{-2/3}$ but $\|f\|_{L^1(\p\mcm)}$ is bounded in $h.$ So in \eqref{eq-x22-2}, the integration in $z'$ yields  
\beq 
\begin{split}
|X_{22, 2}| &\leq  \int_{\p\mcm}  |\tilde F(z'';h)|  | \p_\nu \tilde v(z'')|  dz'' \leq C \|\p_\nu \tilde v\|_{L^2(\p\mcm)}. 
 \end{split}
\eeq
%\beq
%\begin{gathered}
%|X_{22, 2}|  = |\int_{\p\mcm}\int_{\p \mcm} (\int_\mcm  \delta V(z) \frac{e^{i|z - z'|/h}e^{i|z - z''|/h}}{16\pi^2   |z - z'||z - z''|}  dz) \p_\nu  u (z')  \p_\nu \tilde v(z'') dz'dz''|\\
%\leq C   h \|\p_\nu u\|_{L^2} \|\p_\nu \tilde v\|_{L^2}  \leq Ch h^{-5/3} h^{-1/2} \|\tilde f\|_{L^2}\leq C h^{-5/6}. 
%\end{gathered}
%\eeq
Here, $\tilde F$ is a smooth function in $z''$ bounded in $h.$ Now we use Lemma \ref{lm-est} and interpolation to get 
\beqq\label{eq-pnuv-est}
\|\p_\nu \tilde v\|_{L^2(\p\mcm)} \leq C\|\tilde v\|_{H^{3/2}(\mcm)} \leq C h^{-1/2}\|\tilde f\|_{L^2(\p\mcm)}\leq C h^{-5/6}. 
\eeqq
Thus we proved that $|X_{22, 2}|\leq C h^{-5/6}$. By the same argument, we have $|X_{22, 3}|\leq C h^{-5/6}$. Finally, for $X_{22, 4}$, we can estimate directly after the stationary phase argument that 
\beq
\begin{gathered}
|X_{22, 4}|  = |\int_{\p\mcm}\int_{\p \mcm} (\int_\mcm  \delta V(z) \frac{e^{i|z - z'|/h}e^{i|z - z''|/h}}{16\pi^2   |z - z'||z - z''|}  dz) \p_\nu  v (z')  \p_\nu \tilde v(z'') dz'dz''|\\
\leq C   h \|\p_\nu v\|_{L^2(\p\mcm)} \|\p_\nu \tilde v\|_{L^2(\p \mcm)}  \leq Ch h^{-5/6}  h^{-5/6} \leq C h^{-2/3}. 
\end{gathered}
\eeq
In conclusion, we proved that 
\beqq\label{eq-rayest22}
 \beta_0 \tilde \beta_0  \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} X^{W_{22}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{22} + O(h^{1/6}). 
\eeqq

The analysis for $X_{24}, X_{42}$ and $X_{12}, X_{21}$ then follow the same calculations we did for $X_{44}$ and $X_{14}, X_{41}$. So we have 
\beqq\label{eq-rayest24}
\begin{gathered}
- \beta_0 (i \tilde \beta_1)  \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} X^{W_{24}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{24} + O(h^{1/6}), \\
- i \beta_1 \tilde \beta_0  \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} X^{W_{42}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{42} + O(h^{1/6}), 
 \end{gathered}
\eeqq
and 
\beqq\label{eq-rayest12}
\begin{gathered}
 \frac{\tilde \beta_0}{-\beta_0 + i\beta_1} \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} X^{W_{12}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{12} + O(h^{1/6}), \\
  \frac{\beta_0}{-\tilde \beta_0 + i\tilde \beta_1}  \frac{ (2\pi i)}{16\pi^2}  e^{i|z_0' - z_0''|/h} X^{W_{21}}\delta V(z_0', z_0'') J(0)\tilde J(0) = hX_{21} + O(h^{1/6}).
 \end{gathered}
\eeqq
Actually, all the weights $W_{24}, W_{42}, W_{12}, W_{21}$ are equal to $W$. 
 

 %======================%
\subsection{The conclusion}\label{sec-con}
We summarize the analysis for the leading order terms  in \eqref{eq-mainid1}. So far we have obtained the expansion for $X_{44}$ in \eqref{eq-rayest}, $X_{11}$ in \eqref{eq-rayest11}, $X_{14}$ in \eqref{eq-rayest14}, $X_{41}$ in \eqref{eq-rayest41}, $X_{22}$ in \eqref{eq-rayest22}, $X_{24}, X_{42}$ in \eqref{eq-rayest24} and $X_{12}, X_{21}$ in \eqref{eq-rayest12}, all of which involve the X-ray transform with positive weight $W$, the factor $(2\pi i)/(16\pi^2) e^{i|z_0' - z_0''|/h}$, the Jacobian factors $J(0)\tilde J(0)$ and some extra scalar factors. %We remark that the weights are positive for any fixed $z_0', z_0''\in \p \mcm$ and $z_0'\neq z_0''$.  
We focus on the X-ray transform and the scalar factors and consider the sum 
\beq 
\begin{gathered}
 - \beta_1 \tilde \beta_1 X^{W_{44}}\delta V(z_0', z_0'')  + \frac{-i\tilde \beta_1}{-\beta_0 + i\beta_1}  X^{W_{14}}\delta V(z_0', z_0'')  + \frac{-i\beta_1}{-\tilde \beta_0 + i\tilde \beta_1}    X^{W_{41}}\delta V(z_0', z_0'') \\
  + \frac{1}{(-\beta_0 + i\beta_1) (-\tilde \beta_0 + i\tilde \beta_1)}  X^{W_{11}}\delta V(z_0', z_0'') \\
  +  \beta_0 \tilde \beta_0   X^{W_{22}}\delta V(z_0', z_0'')  -  \beta_0 (i \tilde \beta_1)  X^{W_{24}}\delta V(z_0', z_0'') -  i \beta_1 \tilde \beta_0   X^{W_{42}}\delta V(z_0', z_0'')  \\
 +  \frac{\tilde \beta_0}{-\beta_0 + i\beta_1}  X^{W_{12}}\delta V(z_0', z_0'')  +   \frac{\beta_0}{-\tilde \beta_0 + i\tilde \beta_1}    X^{W_{21}}\delta V(z_0', z_0''). 
 \end{gathered}
\eeq
Note that all the weights are equal to $W$. After simplifying the coefficients, we see that the above is equal to 
\beq
\begin{gathered}
\frac{(1- A^2)(1- \tilde A^2)}{A\tilde A} X^{W}\delta V(z_0', z_0''), 
\end{gathered}
\eeq
where $A = -\beta_0 + i   \beta_1, \tilde A = -\tilde \beta_0 + i \tilde \beta_1.$ 
Note that the factor $1- A^2$ cannot be zero because $\im(1- A^2) = 2\beta_0\beta_1<0$. Similarly, $1- \tilde A^2\neq0$. Thus we get  from the sum of the leading terms that 
\beqq\label{eq-rayest1}
\begin{gathered}
|X^W\delta V(z_0', z_0'')|\leq C h |X_{44}  + X_{11} + X_{14} + X_{41} + X_{22} + X_{24} + X_{42} + X_{12} + X_{21}| + Ch^{1/6}.  
\end{gathered}
\eeqq
Now we can use the main identity \eqref{eq-mainid1} to get 
\beqq\label{eq-rayest1}
\begin{gathered}
|X^W\delta V(z_0', z_0'')|\leq   C \sum_{j = 1, 2, 4, k = 3, 5, 6} h|X_{jk}|  + C \sum_{j =  3, 5, 6, k = 1, 2, 4} h|X_{jk}| \\
+ C \sum_{j, k = 3, 5, 6} h|X_{jk}| +   C h(|Y_{12}| + |Y_{21}| + |Y_{22}|) + C h^{1/6}.
\end{gathered}
\eeqq
We remark that the constant $C$ depends on $z_0', z_0''$ but not on $h.$ We estimate the remaining terms in the next section. 



%=============================%
\section{The remaining estimates}\label{sec-pf-rem}
In the end of the proof, we will take $h\rightarrow 0$ in \eqref{eq-rayest1}, and it suffices to show that the $X_\bullet, Y_\bullet$ terms there are $o(h^{-1})$. In the follows, we achieve the goal but we do not look for the optimal estimates. We start from the $Y$ terms. 
\begin{lemma}
$|Y_{12}|, |Y_{21}|, |Y_{22}| %\leq Ch \|f\|_{L^2}\|\tilde f\|_{L^2} 
\leq C h^{-1/3}$.
\end{lemma}
\bpf
Using the expression of $Y_{12}$, we have  
\beq
|\int_{\mcm} \delta V(z) u_0(z)\tilde v(z) dz|\leq \|\delta V u_0\|_{L^2(\mcm)} \|\tilde v\|_{L^2(\mcm)}\leq C \| f\|_{L^2(\p\mcm)}  h\|\tilde f\|_{L^2(\p\mcm)}, 
\eeq
where we used  \eqref{eq-vest} and \eqref{eq-u0est} and the fact that $\delta V$ is compactly supported in $\mcm.$ Also, $C$ depends on the bound $M$ of $V, \tilde V$ but not $h.$ For $f, \tilde f$ defined in \eqref{eq-f1}, \eqref{eq-f2}, we have 
\beqq\label{eq-fest} 
\|f\|_{L^2(\p\mcm)}, \|\tilde f\|_{L^2(\p\mcm)} \leq C h^{-2/3}. 
\eeqq  
So the estimate follows.  The other two estimates are similar.
\epf


 
To estimate the $X_{jk}$ terms, we observe that there is an apparent symmetry in $j, k$  so it suffices to analyze $X_{jk}$ for $j\leq k$.
\begin{lemma}
For $j, k =  3, 5, 6$, we have $|X_{jk}| \leq C h^{-1/3}.$  
\end{lemma}
\bpf
These terms can be estimated as $X_{44}$ because $I_j, j =  3, 5, 6$ only involve boundary integrals. In particular, we can apply the stationary phase argument to get an extra $h$ which is enough. We show the estimates for $X_{33}$. We have 
\beq
\begin{split}
X_{33}  &= \int_{\p\mcm}\int_{\p\mcm} (\int_\mcm  \delta V(z) \frac{e^{i|z - z'|/h}e^{i|z - z''|/h}\nu_{z''} \cdot (z - z'') \nu_{z'} \cdot (z - z') }{16\pi^2   |z - z'|^3|z - z''|^3}  dz)  f(z')   \tilde f(z'') dz'dz''\\
 &= \int_{\p\mcm} \int_{\p\mcm} e^{i|z' - z''|/h} h F(z', z'';h) f (z')   \tilde f(z'') dz'dz'', 
\end{split}
\eeq
where $F$ is a smooth function in $z', z''$ and bounded in $h$. %For the integration in $z''$, we can make a change of variable so $\tilde f$ does not contribute any negative powers of $h$. We remark that  Note that $\|f\|_{L^2(\p\mcm)}, \|\tilde f\|_{L^2(\p\mcm)}$ are of order $h^{-2/3}$. 
Using \eqref{eq-fest},  we estimate that  
\beq
\begin{gathered}
|X_{33} |\leq C h h^{-4/3} \leq  C h^{-1/3}. 
\end{gathered}
\eeq
The other estimates are similar, noting that $\|f_*\|_{L^2(\p \mcm)}, \|\tilde f_*\|_{L^2(\p \mcm)}$ are of order $h^{-1/6}$.  In fact, 
\beq
\|f_*\|_{L^2(\p\mcm)}\leq C\|v\|_{H^{1/2}(\mcm)} \leq Ch^{1/2} \|f\|_{L^2(\p\mcm)}, 
\eeq
where we used interpolation of the estimates for $v$ in Lemma \ref{lm-est}.
\epf


\begin{lemma}
For $j = 1, k = 3, 5, 6$, we have $|X_{jk}|, |X_{kj}|  \leq C h^{-2/3}$. 
\end{lemma}
\bpf
Consider $X_{13}$. We follow the calculation for $X_{14}$ in Section \ref{sec-x14}. We have 
\beq
\begin{gathered}
X_{13} = X_{13, 1} + X_{13, 2}, \text{ where }\\
X_{13, 1} =  -\int_{\mcm} \int_{\mcm} \int_{\p \mcm} h^2 \delta V(z)G_0(z, z'; h) G_{\nu, 1}(z, z''; h) w_0(z')  \tilde f(z'')dzdz'dz'', \\ 
X_{13, 2} =  -\int_{\mcm} \int_{\mcm} \int_{\p \mcm} h^2 \delta V(z)G_0(z, z'; h) G_{\nu, 1}(z, z''; h) w_1(z')  \tilde f(z'')dzdz'dz''. 
\end{gathered}
\eeq
The second term can be estimated by 
\beq
\begin{split}
|X_{13, 2}| &\leq C h^{-2}  \int_{\mcm} \int_{\mcm} \int_{\p \mcm}  |\delta V(z)| \frac{1}{|z - z'|} \frac{1 }{|z - z''|}  |w_1(z')| |\tilde f(z'')| dzdz'dz''\\
&\leq Ch^{-2} \int_{\mcm}  \int_{\p \mcm} |w_1(z')| |\p_\nu \tilde f(z'')|  dz'dz'' \\
&\leq Ch^{-2} \|w_1\|_{L^2(\p\mcm)} \|\tilde f\|_{L^2(\p\mcm)} \leq C h^{-2 + 5 - 2/3} = C h^{7/3}. 
\end{split}
\eeq
For the first term, we can repeat the calculation as for $X_{14, 1}$. The  structures of the integral kernels are the same and the only difference is that now the kernel has  a factor $h^{-2}$ instead of $h^{-3}$. In particular, we can get a  similar expression as in \eqref{eq-x141}  of the form 
 \beqq\label{eq-x121}
\begin{gathered}
X_{13, 1} = \int_{\p \mcm} \int_{\p \mcm} F(z', z'', h) f(z') \tilde f(z'') dz'dz'' 
+  \int_{\mcm} \int_{\p \mcm} Z(z', z'', h) w_1(z') \tilde f(z'') dz'dz'', 
\end{gathered}
\eeqq
where $F = O_{L^\infty}(1)$ and $Z = O_{L^\infty}(h)$ for small $h$. By the $L^2$ estimates of $w_1, \tilde f$, we get that 
\beq
|X_{13, 1}|\leq C h^{-2/3} + C h h^{-1/6} h^{-2/3} = Ch^{-2/3}. 
\eeq
So finally we get $|X_{13}|\leq C h^{-2/3}$. The estimates for $X_{1j}, j = 5, 6$ are similar. 
\epf

 
\begin{lemma}
For $j = 2, 4, k = 3, 5, 6$, we have $|X_{jk}|, |X_{kj}|  \leq C h^{-2/3}$. 
\end{lemma}
\bpf 
These terms can be estimated as $X_{44}$. We again apply the stationary phase argument to get an extra $h$. Note that $I_k, k = 3, 5, 6$ contain extra factors of $h^{1/2}$ in some way compared to $I_{2}, I_4$ and this is enough. We consider $X_{23}$ as an example. Similar to the analysis of $X_{22},$ we decompose   
\beq
\begin{split}
X_{23} = & -\int_{\p \mcm} \int_{\p \mcm} \int_{\mcm} h^4 \delta V(z)G_0(z, z'; h) G_{\nu, 1}(z, z''; h) \p_\nu u_0(z')  \tilde f(z'')dzdz'dz''\\
 = & -\int_{\p \mcm} \int_{\p \mcm} \int_{\mcm} h^4 \delta V(z)G_0(z, z'; h) G_{\nu, 1}(z, z''; h) \p_\nu u(z')  \tilde f(z'')dzdz'dz''\\
 & + \int_{\p \mcm} \int_{\p \mcm} \int_{\mcm} h^4 \delta V(z)G_0(z, z'; h) G_{\nu, 1}(z, z''; h) \p_\nu v(z')  \tilde f(z'')dzdz'dz'' = X_{23, 1} + X_{23, 2}. 
 \end{split}
\eeq
For $X_{23, 2}$, we estimate after stationary phase argument that 
\beq
\begin{split}
|X_{23, 2}| & \leq  |\int_{\p \mcm} \int_{\p \mcm}  e^{i|z' - z''|/h} h F(z', z'';h)   \p_\nu v(z')  \tilde f(z'') dz'dz'' |\\
& \leq  C h \|\p_\nu v\|_{L^2(\p\mcm)}\|\tilde f\|_{L^2(\p\mcm)}\leq Ch h^{-5/6} h^{-2/3} = C h^{-1/2}. 
\end{split}
\eeq
Hereafter, we use $F$ for a generic smooth function in $z', z''$ and bounded in $h$. Also, we used \eqref{eq-pnuv-est}. For $X_{23, 1}$, we can use stationary phase argument to get 
\beq 
X_{23, 1} %&=  -\int_{\p \mcm} \int_{\p \mcm} \int_{\mcm} h^4 \delta V(z)G_0(z, z'; h) G_{\nu, 1}(z, z''; h) \p_\nu u(z')  \tilde f(z'')dzdz'dz''\\
 %&=   \int_{\p \mcm} \int_{\p \mcm}   (\int_\mcm  \delta V(z) \frac{e^{i|z - z'|/h}e^{i|z - z''|/h}\nu_{z''} \cdot (z - z'') \nu_{z'} \cdot (z - z') }{16\pi^2   |z - z'| |z - z''|^3}  dz)  \p_\nu u_0(z')  \tilde f(z'')dzdz'dz''\\
 =  \int_{\p\mcm} \int_{\p\mcm} e^{i|z' - z''|/h} h F(z', z'';h) \p_\nu u (z')   \tilde f(z'') dz'dz''. 
\eeq  
Now we repeat the argument for $X_{22, 2}$. For the integration in $z''$, we use that $\|\tilde f\|_{L^1(\p\mcm)}$ is bounded in $h$.  Then we estimate the integration in $z'$ to get  that 
\beq
\begin{gathered}
|X_{23, 1}| \leq C  h \|\p_\nu u\|_{L^2(\p\mcm)} \leq C h  h^{-1}h^{-2/3}  = Ch^{-2/3}. 
 \end{gathered}
\eeq 
Thus, we proved that $|X_{23}|\leq C h^{-2/3}.$ The estimate for the rest of the terms follow from the same arguments. 
\epf

%==========================%
\section{Completion of the proof}\label{sec-pf}
Using the estimates of $X_\bullet$, $Y_\bullet$ terms in Section \ref{sec-pf-rem}, we get from \eqref{eq-rayest1} that 
\beqq\label{eq-rayest2}
\begin{gathered}
|X^W\delta V(z_0', z_0'')| %\leq \sum_{j, k = 1, 2, 3, 4, (j, k)\neq (4, 4)} h^{5}|X_{jk}| + C h^{1/3}\\
\leq C h^{1-2/3} + Ch^{1/6} \leq Ch^{1/6}. 
\end{gathered}
\eeqq
Now we can take $h\rightarrow 0$ to get  
\beq
X^W\delta V(z_0', z''_0) = 0.
\eeq 
This holds for any $(z_0', z_0'')$ in a dense open set in $\p \mcm \times \p \mcm$ on which $\alpha \neq0, \tilde \alpha\neq 0$, see Lemma \ref{lm-geo}. Because  $\delta V$ is smooth, we know that $X^W\delta V$ is smooth. So we actually have $X^W\delta V(z_0', z''_0) = 0$ for any $z_0', z_0''\in \p \mcm. $ Finally,  the weight $W$ is non-vanishing and smooth. It is known that $X^W$ is injective. This can be seen from the proof of Theorem in \cite{UhVa} by adding a non-vanishing weight to the geodesic ray transform. We conclude that $\delta V = 0$ so $V = \tilde V.$ This completes the proof of Theorem \ref{thm-main}. \\

For the proof of Theorem \ref{thm-main1}, we first consider the local weighted X-ray transform $X^W$ near some boundary point $z\in \p \mcm$.  We know from the Theorem in \cite{UhVa} that  for any $z \in \p \mcm$, there exists neighborhood $\Omega$ of $z$ in $\mcm$ such that $X^W$ is injective on $C^\infty(\overline\Omega)$. Let $\Gamma = \Omega\cap \p\mcm$. We observe that the identity \eqref{eq-ide0} holds if $\La^\Gamma = \tilde \La^\Gamma$. Then we note that  the proof of Theorem \ref{thm-main} is local in the sense that it works for any $z_0', z_0''\in \p \mcm$ satisfying Lemma \ref{lm-geo} and $\delta V$ supported away from $z_0', z_0''$. So we can still derive $X^W\delta V(z_0', z''_0) = 0$ for $z_0', z_0''\in \Gamma$. Thus we derive $\delta V = 0$ in $\Omega$ so $V = \tilde V$ in $\Omega.$  


\section*{Acknowledgement}
GU would like to thank Yuchao Yi for his assistance. Both GU and YW are supported by NSF. 


 %%===============================REFERENCE==========================================%
\begin{thebibliography}{99}
%\bibitem{Ale} G. Alessandrini. {\em Stable determination of conductivity by boundary measurements.} Applicable Analysis 27.1-3 (1988): 153-172.
%\bibitem{AlVe} G. Alessandrini, S. Vessella. {\em Lipschitz stability for the inverse conductivity problem.} Adv. Appl. Math. 35 207-41, (2015). 
\bibitem{AmUh} H. Ammari, G. Uhlmann. {\em Reconstruction of the potential from partial Cauchy data for the Schr\"odinger equation.} Indiana University Mathematics Journal (2004): 169-183.
%\bibitem{Bug} A. Bukhgeim. {\em Recovering a potential from Cauchy data in the two-dimensional case.} J. Inverse Ill-Posed Probl. 16(1), 19-33 (2008).
\bibitem{BuUh} A. Bukhgeim, G. Uhlmann. {\em Recovering a potential from partial Cauchy data.} Communications in Partial Differential Equations 27.3-4 (2002): 653-668.
\bibitem{Cal}  A. Calder\'on. {\em On an inverse boundary value problem.} Seminar on Numerical Analysis and its Applications to Continuum Physics (Rio de Janeiro: Sociedade Brasileira de Matem\'atica) pp 65-73, 1980. 
\bibitem{DKSU} F. David Dos Santos, C. Kenig, J Sj\"ostrand, G. Uhlmann.  {\em Determining a magnetic Schr\"odinger operator from partial Cauchy data.} Communications in Mathematical Physics 271.2 (2007): 467-488.
\bibitem{FSU} J. Feldman, M. Salo, G. Uhlmann. {\em The Calder\'on Problem - An Introduction to Inverse Problems.} American Mathematical Scociety, 2025.
%\bibitem{FKSU}  D. Dos Santos Ferreira, C. Kenig,  M. Salo, G. Uhlmann. {\em Limiting Carleman weights and anisotropic inverse problems.} Inventiones Mathematicae, 178(1), 119-171. 2009.
%\bibitem{FKLS} D.  Dos Santos Ferreira, Y. Kurylev, M. Lassas, M. Salo. {\em The Calder\'on problem in transversally anisotropic geometries.} Journal of the European Mathematical Society, 18(11), 2579-2626. (2016)
%\bibitem{Dui} J. Duistermaat. {\em Fourier Integral Operators.} Springer Science \& Business Media, 2010.
%\bibitem{GHL} S. Gallot, D. Hulin, J. Lafontaine. {\em Riemannian geometry.} Vol. 3. Berlin: Springer-Verlag, 1990.
%\bibitem{GiTr} D. Gilbarg, N. Trudinger. {\em Elliptic Partial Differential Equations of Second Order.} Berlin: Springer; 1977.
%\bibitem{Gu} V. Guillemin. {\em Sojourn times and asymptotic properties of the scattering matrix.} Publications of the Research Institute for Mathematical Sciences 12.Supplement (1977): 69-88.
%\bibitem{GuTz} C. Guillarmou, L. Tzou. {\em Calder\'on inverse problem for the Schrödinger operator on Riemann surfaces.} In The AMSI--ANU Workshop on Spectral Theory and Harmonic Analysis, pp. 129-141.
%\bibitem{Ho} L. H\"ormander. {\em Pseudo-differential operators and non-elliptic boundary problems.} Annals of Mathematics (1966): 129-209.
%\bibitem{Hor} Hormander, nonlinear hyperbolic 
\bibitem{Ho1}  L. H\"ormander. {\em The Analysis of Linear Partial Differential Operators I: Distribution Theory and Fourier Analysis. Second Edition}. Springer-Verlag, 1990. 
\bibitem{IUY1} O. Imanuvilov, G. Uhlmann, M. Yamamoto. {\em The Calder\'on problem with partial data in two dimensions.} Journal of the American Mathematical Society 23.3 (2010): 655-691.
\bibitem{IUY2} O. Imanuvilov, G. Uhlmann, M. Yamamoto. {\em The Neumann-to-Dirichlet map in two dimensions.} Advances in Mathematics 281 (2015): 578-593.
\bibitem{Isa} V. Isakov. {\em On uniqueness in the inverse conductivity problem with local data.} Inverse Problems and Imaging 1.1 (2007): 95.
%\bibitem{Isa} V. Isakov, S. Nagayasu, G.  Uhlmann, J. N. Wang. (2014). {\em Increasing stability of the inverse boundary value problem for the Schr\"odinger equation.} Contemp. Math, 615, 131-141.
\bibitem{KeSa} C. Kenig, M. Salo. {\em Recent progress in the Calder\'on problem with partial data.} Inverse Problems and Applications, Contemp. Math., Vol. 615, Amer. Math. Soc., Providence, RI, 2014, pp. 193-222. 
%\bibitem{KSU} C. Kenig, M. Salo, G. Uhlmann. {\em Inverse problems for the anisotropic Maxwell equations.} Duke Mathematical Journal 157.2 (2011): 369-419.
%\bibitem{KUW} P. Z. Kow, G. Uhlmann, J. N. Wang. {\em Optimality of increasing stability for an inverse boundary value problem.} arXiv:2102.11532 (2021).
\bibitem{KrUh1} K. Krupchyk, G. Uhlmann. {\em The Calder\'on problem with partial data for conductivities with $3/2$ derivatives.} Communications in Mathematical Physics 348.1 (2016): 185-219.
\bibitem{KrUh2} K. Krupchyk, G. Uhlmann. {\em Stability estimates for partial data inverse problems for Schr\"odinger operators in the high frequency limit.} Journal de Math\'ematiques Pures et Appliqu\'ees 126 (2019): 273-291.
%\bibitem{LaTh} S. Larsson,  V. Thom\'ee. {\em Partial differential equations with numerical methods.} Vol. 45. Springer Science \& Business Media, 2008.
%\bibitem{LTU} M. Lassas,  M. Taylor, G. Uhlmann. {\em The Dirichlet-to-Neumann map for complete Riemannian manifolds with boundary. }Communications in Analysis and Geometry, 11(2), 207-221. 2003.
%\bibitem{LaUh} M. Lassas, G. Uhlmann. {\em On determining a Riemannian manifold from the Dirichlet-to-Neumann map.} Annales Scientifiques de l’Ecole Normale Sup\'erieure. Vol. 34. No. 5.  2001.
%\bibitem{LeUh} J. Lee, G. Uhlmann. {\em Determining anisotropic real‐analytic conductivities by boundary measurements.} Communications on Pure and Applied Mathematics 42.8 (1989): 1097-1112.
\bibitem{LiUh} X. Li, G. Uhlmann. {\em Inverse problems with partial data in a slab.} Inverse Problems and Imaging 4.3 (2010): 449-462.
%\bibitem{Man} N. Mandache. {\em Exponential instability in an inverse problem for the Schrödinger equation.} Inverse Problems 17.5 (2001): 1435.
%\bibitem{Mel} R.  Melrose. {\em  Geometric scattering theory.}
%\bibitem{MSV} R. Melrose, A. S\'a Barreto, A. Vasy. {\em Analytic continuation and semiclassical resolvent estimates on asymptotically hyperbolic spaces.} Communications in Partial Differential Equations 39.3 (2014): 452-511.
%\bibitem{MSS} G. Mockenhaupt, A. Seeger, C. Sogge. {\em Local smoothing of Fourier integral operators and Carleson-Sj\"olin estimates.} Journal of the American Mathematical Society 6.1 (1993): 65-130.
%\bibitem{SaWa} A. S\'a Barreto, Y. Wang. {\em The semiclassical resolvent on conformally compact manifolds with variable curvature at infinity.}  Communications in Partial Differential Equations 41, no.8, 1230-1302, (2016).
%\bibitem{Sha} V. Sharafutdinov. {\em Integral geometry of tensor fields.} Vol. 1. Walter de Gruyter, 2012.
%\bibitem{Shi} N. Shimakura. {\em Partial Differential Operators of Elliptic Type.} American Mathematical Society, 1992. 
%\bibitem{So} C. Sogge. {\em Fourier integrals in classical analysis.} Vol. 210. Cambridge University Press, 2017.
%\bibitem{StUh0} P. Stefanov, G. Uhlmann. {\em Rigidity for metrics with the same lengths of geodesics.} Mathematical Research Letters 5.1 (1998): 83-96
% \bibitem{StUh} P. Stefanov, G. Uhlmann. {\em Boundary rigidity and stability for generic simple metrics.} Journal of the American Mathematical Society 18.4 (2005): 975-1003
%\bibitem{SUV} P. Stefanov, G. Uhlmann, A. Vasy. {\em Boundary rigidity with partial data.} Journal of the American Mathematical Society 29.2 (2016): 299-332.
\bibitem{SyUh} J. Sylvester, G. Uhlmann. {\em A global uniqueness theorem for an inverse boundary value problem.}  Annals of Mathematics (1987): 153-169.
\bibitem{Tay} M. Taylor. {\em Partial Differential Equations I Basic Theory, 2nd Edition.} Applied Mathematical Sciences, Volume 115, Springer, 2011. 
\bibitem{Tre} F. Tr\`eves. {\em Basic Linear Partial Differential Equations.}  Academic Press, 1975.
%\bibitem{Tre1} F. Tr\`eves. {\em Introduction to pseudodifferential and Fourier integral operators: Pseudodifferential operators.} Springer Science \& Business Media, 2013.
%\bibitem{Uh} G. Uhlmann. {\em The Cauchy data and the scattering relation.} Geometric methods in inverse problems and PDE control. Springer, New York, NY, 2004. 263-287.
%\bibitem{Uh} G. Uhlmann. {\em Inverse problems: seeing the unseen.} Bulletin of Mathematical Sciences 4.2 (2014): 209-279.
\bibitem{UhVa} G. Uhlmann, A. Vasy. {\em The inverse problem for the local geodesic ray transform.} Inventiones Mathematicae 205.1 (2016): 83-120.
\bibitem{VaZw} A. Vasy, M. Zworski. {\em Semiclassical estimates in asymptotically Euclidean scattering.} Communications in Mathematical Physics 212 (2000): 205-217. 
%\bibitem{UhWa} G. Uhlmann, Y. Wang. {\em The anisotropic Calder\'on problem for fixed high frequency.} SIAM Journal on Mathematical Analysis. 
%\bibitem{Zwo} M. Zworski. {\em Semiclassical Analysis.} Vol. 138. American Mathematical Soc., 2012.
\end{thebibliography}
\end{document}