Fisher's combined probability test for cross-sectional independence in panel data models with serial correlation
Extracted source text ยท arXiv 2309.08543
\documentclass[12pt]{article}
\usepackage{amsthm,amsmath,amsfonts,amssymb}
\usepackage{natbib}%% uncomment this for author-year citations
\usepackage[colorlinks,citecolor=blue,urlcolor=blue]{hyperref}
\usepackage{graphicx}
\usepackage[table,xcdraw]{xcolor}
\usepackage{enumerate}
\usepackage{color}
\usepackage{url}
\usepackage[space]{grffile}
\usepackage{latexsym}
\usepackage{textcomp}
\usepackage{longtable}
\usepackage{tabulary}
\usepackage{booktabs,array,multirow}
\usepackage[figuresright]{rotating}
\usepackage[margin=1in]{geometry}
\renewcommand{\baselinestretch}{1.25}
\newcommand\blfootnote[1]{%
\begingroup
\renewcommand\thefootnote{}\footnote{#1}%
\addtocounter{footnote}{-1}%
\endgroup
}
\long\def\symbolfootnote[#1]#2{\begingroup%
\def\thefootnote{\fnsymbol{footnote}}\footnote[#1]{#2}\endgroup}
\usepackage{times}
\def\mE{\mathbb E}
\def\tr{\mathrm {tr}}
\def\cov{\mathrm {cov}}
\def\var{\mathrm {var}}
\def\mE{\mathbb{E}}
\def\x{{\mathbf x}}
\def\U{{\mathbf U}}
\def\n{\nonumber}
\def\cd{\mathop{\rightarrow}\limits^{d}}
\def\cp{\mathop{\rightarrow}\limits^{p}}
\def\mE{\mathbb E}
\def\tr{\mathrm {tr}}
\def\cov{\mathrm {cov}}
\def\var{\mathrm {var}}
\def\mE{\mathbb{E}}
\def\x{{\mathbf x}}
\def\U{{\mathbf U}}
\def\n{\nonumber}
\def\x{\mathbf{x}}
\def\hx{{\hat{\bm{x}} }}
\def\he{\hat{e}}
\def\tr{\mathrm {tr}}
\def\hep{\hat{\epsilon}}
\def\ep{{\epsilon}}
\def\trans{\top}
\def\P{\mathbb{P}}
\def\be{\begin{eqnarray}}
\def\ee{\end{eqnarray}}
\def\bse{\begin{eqnarray*}}
\def\ese{\end{eqnarray*}}
\def\cd{\mathop{\rightarrow}\limits^{d}}
\def\cp{\mathop{\rightarrow}\limits^{p}}
\numberwithin{equation}{section}
\newtheorem{simulation}{Setting}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{corollary}{Corollary}[section]
\newtheorem{example}{Example}[section]
\newtheorem{assumption}{Assumption}
\newtheorem{remark}{Remark}
\allowdisplaybreaks[4]
%\def\boxit#1{\vbox{\hrule\hbox{\vrule\kern6pt\vbox{\kern6pt#1\kern6pt}\kern6pt\vrule}\hrule}}
%\def\liucomment#1{\vskip 2mm\boxit{\vskip 2mm{\color{red}\bf#1}
%{\color{blue}\bf -- LIU\vskip 2mm}}\vskip 2mm}
\theoremstyle{plain}
\newtheorem{lemma}{Lemma}[section]
\begin{document}
\title{\bf Fisher's combined probability test for cross-sectional independence in panel data models with serial correlation}
\vspace{0.5in}
\author{Hongfei Wang$^{1}$, Binghui Liu$^{1}$, Long Feng$^{2}$ and Yanyuan Ma$^{3}$\\
\fontsize{10}{10}\selectfont\itshape
$^{1}$\, School of Mathematics and Statistics and KLAS, Northeast Normal University, Renmin Street, Changchun, Jilin, China. \\
\fontsize{10}{10}\selectfont\itshape
$^{2}$\,School of Statistics and Data Science, LPMC, LEBPS and KLMDASR, Weijin Road, Tianjin, China.\\
\fontsize{10}{10}\selectfont\itshape
$^{3}$\, Department of Statistics, Pennsylvania State University, Old Main, University Park, Pennsylvania, United States of America.
}
\date{}
\maketitle
\begin{abstract}
Testing cross-sectional independence in panel data models is of
fundamental importance in econometric analysis with high-dimensional
panels. Recently, econometricians began to turn their attention to the problem in the
presence of serial dependence.
The existing procedure for testing cross-sectional independence with
serial correlation is based on the sum of the sample
cross-sectional correlations, which generally performs well when
the alternative has dense cross-sectional correlations, but suffers from
low power against sparse alternatives. To deal with sparse
alternatives, we propose a test based on the maximum of the
squared sample cross-sectional correlations. Furthermore, we propose
a combined test to combine the p-values of the max based and sum based tests, which
performs well under both dense and sparse
alternatives. The combined test relies on the
asymptotic independence of the max based and sum based test
statistics, which we show rigorously.
We show that the proposed max based and
combined tests have attractive theoretical
properties and demonstrate the superior performance via extensive
simulation results. We apply the two new tests to
analyze the weekly returns on the
securities in the S\&P 500 index under the Fama-French three-factor
model, and confirm the usefulness of the proposed combined test in
detecting cross-sectional independence.
\end{abstract}
\noindent
{\it Keywords:} Asymptotic independence; Cross-sectional dependence; Fisher's combined probability test;
Heterogeneous panel data models; High dimensionality.
\vfill
\newpage
\baselineskip=24.5pt
\section{Introduction}\label{sec:intro}
In this paper, we consider the problem of testing cross-sectional
independence in heterogeneous panel data models.
In statistics and econometrics, panel data occur
frequently, which contain observations
of various types obtained over multiple time periods for
any single unit. In the study of
panel data models, the cross-sectional dependency is an important
concept, described as the
interaction between cross-sectional units, which could arise from the
behavioral interaction
between units \citep{breusch1980l,feng2020}. To make theoretical study easier,
experts often assume cross-sectional independence in the model
setups. If data across units are dependent,
inferences under the assumption of cross-sectional independence would
be inaccurate and misleading; see literature on spatial
econometrics, such as
\cite{anselin1998,kelejian1999,kapoor2007,lee2007,lee2010}
for examples of cross-sectional dependence.
Therefore, testing the existence of cross-sectional dependence is
important and attracts increasing attention.
A large number of literatures on testing cross-sectional dependence
are available, among which
the most widely known is likely the Lagrange Multiplier (LM)
test proposed by \cite{breusch1980l}.
The LM test is based on the sum of the squared cross-sectional
correlations of residuals, which is applicable when
the sample size $T$ is large and the dimension $N$ is finite, but
is not a valid test when $N\rightarrow \infty$.
To develop tests applicable when both $N$ and $T$ are
large, two limit schemes have been considered
in the literature.
One is the sequential limit scheme: $T\rightarrow \infty,$ followed by
$N\rightarrow \infty$, and the other is the
simultaneous limit scheme: $N$ and $T$ tend to infinity
simultaneously, i.e. $(N,T)\rightarrow \infty$.
Under the sequential limit scheme, \cite{pesaran2004} proposed a
scaled version of the LM test, as well as
a test based on the sum of the residual correlations,
instead of the squared sum,
to address the issue that the former test may suffer from substantial
size distortions in the case of large $N$ and small $T$;
\cite{pesaran2008} proposed a bias-adjusted LM test based on the scaled LM test.
Later, under the simultaneous limit scheme, \cite{pesaran2015}
established the asymptotical properties of the sum based
test by \cite{pesaran2004};
\cite{feng2020} established the asymptotical properties of the
bias-adjusted LM test proposed by \cite{pesaran2008}. In addition to
these tests,
there are also works on testing cross-sectional independence under
other related models, such as the fixed effect panel data models;
see, for instance, \cite{baltagi2012,feng2020fixed}.
All the above tests are sum based tests, i.e., they are based
on the sum of the correlations or squared correlations of the residuals.
These tests generally perform well under dense alternatives, but may
suffer from low power against sparse alternatives. To deal with sparse
alternatives,
\cite{feng2020} proposed a max based test for testing cross-sectional
dependence, and further developed a combined test that integrates the
advantages of the max based and sum based
tests, by establishing the asymptotical independence between the test statistics.
All the tests mentioned above make the common
assumption that the errors in the panel data models are independent
across time.
However, the existence of serial dependence
is likely to be the rule rather than the exception; see, for
instance, \cite{wei2006,Hong2010,box2015}.
In many applications, serial dependence may have great impact on
statistical inference, such as leading to deviation of the limiting
spectral distribution of the sample covariance matrix
\citep{gao2017high}. Hence, in analyzing panel data with serially
correlated errors, using tests for cross-sectional independence
that are based on the assumption that the errors are serially
independent may lead to wrong conclusions.
To solve this problem, \cite{baltagi2016} proposed a test for
cross-sectional correlation under heterogeneous panel
data models with serial correlations by adjusting the sum based test
by \cite{pesaran2004}.
Similarly, \cite{lan2017} proposed a test for cross-sectional
independence under fixed effects panel data models with serial
correlations. Both
tests are sum based, hence the scope of their
application is limited to dense alternatives. As far as we know,
research on max based tests for sparse alternatives or
combined tests regardless of whether alternatives are sparse or not is
not yet available for data with serial dependence.
We aim to fill this gap in this work.
To this end, for testing cross-sectional independence under
heterogeneous panel data models with serial correlations, we propose a
max based test based on the maximum of the squared cross-sectional
residual correlations to deal with sparse alternatives. The method
follows the framework of \cite{chen2018} for testing
independence of correlated samples, while relaxing their Gaussian
sample assumption.
Furthermore, we propose a Fisher's combined probability test by
combining the p-values of our proposed max based test and the sum based
test by \cite{baltagi2016}. This combined test is applicable regardless
the alternatives are sparse or sparse. We derive the
asymptotic null distribution of the proposed combined test, by first
rigorously establishing the asymptotic independence of the max based and
sum based test statistics.
In summary, there are two main contributions of our work.
\begin{enumerate}
\item[(1)]
We propose a max based test for testing cross-sectional
independence
in models with serial correlation and sub-Gaussian error. The new test
is
powerful in detecting sparse alternatives.
\item[(2)]
We establish the asymptotic independence between the sum based and
max based test statistics.
We propose a combined test for testing cross-sectional independence
in models with serial correlation and sub-Gaussian error. The new test is
powerful overall.
\end{enumerate}
The rest of the paper is organized as follows. We propose two new
tests and establish their asymptotic properties in Section
\ref{sec:statistic}.
Simulation results of the proposed tests and their comparison to some
existing methods are demonstrated in Section \ref{sec:simu}, followed
by an empirical application in Section \ref{sec:data}. We
conclusion the paper in Section \ref{sec:discuss} and relegate
the technical proofs to Supplementary Materials.
\textsc{Notations.} For any square matrix $\mathbf{A},$
$(\mathbf{A})_{ij}$ denotes the $(i,j)$-th entry of $\mathbf{A}$,
$\tr(\mathbf{A})$ denotes the trace of $\mathbf{A}$,
$\lambda_{\max}(\mathbf{A})$ and $\lambda_{\min}(\mathbf{A})$ denote the
maximum and the minimum eigenvalues of $\mathbf{A}$, respectively,
$\|\mathbf{A}\|_{\mathrm{F}}$ denotes the Frobenius norm of $\mathbf{A}$, and
$\mathbf{A}^{1/2}$ denotes the principal square root matrix of
$\mathbf{A}$ if $\mathbf{A}$ is a positive definite matrix.
$\mathbf{I}_{n}$ denotes the $n\times n$ identity matrix for each positive integer $n$.
For any two real numbers $x$ and $y$, let $ x \vee y=\max (x, y)$ and
$x \wedge y=\min (x, y).$
For any vector $v,$ $\|v\|$ denotes the Euclidean norm of $v.$
Let $\mathcal{N}(a,b)$ denote the normal distribution with mean
$a\in\mathbb{R}$ and variance $b\geq0$, $t_{v}$ denote the
$t$-distribution with degree of freedom $v$ and $\chi_{v}^2$
denote the chi-square distribution with degree of freedom $v.$
Let $\Phi(\cdot)$ denote the cumulative distribution function of the
standard normal distribution.
We use $(N, T) \rightarrow \infty$ to denote the joint convergence of
$N$ and $T$ to infinity, i.e. $\min(N,T)\to\infty$.
\section{ The proposed tests}\label{sec:statistic}
\subsection{Problem description}
We consider the heterogeneous panel data model taking the form
\begin{align}\label{model}
y_{i t}=x_{i t}^{\prime} \beta_{i}+\epsilon_{it},\,\, 1\leq i\leq N,\,\,1\leq t\leq T,
\end{align}
where $i$ indexes the cross-sectional units, and $t$ indexes the time dimension.
$y_{it}\in \mathbb{R}$ is the dependent variable, $x_{i
t}\in\mathbb{R}^{p}$ is the exogenous regressors with slope
parameters $\beta_{i}\in
\mathbb{R}^{p}$ that are allowed to vary across $i$ and
$\epsilon_{it}\in\mathbb{R}$ is the corresponding idiosyncratic error
term.
For each $1\leq i\leq N$, let
$\mathbf{x}_{i}=(x_{i1},\cdots,x_{iT})^{\prime}\in\mathbb{R}^{T \times
p},$
${y}_{i}=(y_{i1},\cdots,y_{iT})^{\prime}\in\mathbb{R}^{T }$ and
$\epsilon_{i\cdot}=(\epsilon_{i1},\cdots,\epsilon_{iT})^{\prime}\in\mathbb{R}^{T}$.
For each $1\leq t\leq T$, let $\epsilon_{\cdot
t}=(\epsilon_{1t},\cdots,\epsilon_{Nt})^{\prime}$.
The null hypothesis of cross-sectional independence can be written as
\begin{align}\label{null hypothesis}
H_{0}: \epsilon_{1\cdot}, \epsilon_{2\cdot}, \cdots, \epsilon_{N\cdot}
\text{ are
independent random vectors.}
\end{align}
\subsection{Related works}\label{secR}
Early studies on testing cross-sectional independence in \eqref{null
hypothesis} are based on the assumption that there is no serial
dependence in $\{\epsilon_{\cdot t}\}_{t=1}^T$, where
$\epsilon_{\cdot t}$ is assumed to be iid over time $t$. The earliest
work is the LM test \citep{breusch1980l},
with test statistic
$$
\mathrm{L M_{B P}}=T \sum_{j=2}^{N} \sum_{i=1}^{j-1} \hat{\rho}_{i j}^{2},
$$
where $\hat{\rho}_{i j}$ is the sample correlation constructed by the
OLS residuals $\hat{\epsilon}_{i t}=y_{i t}-x_{i t}^{\prime}
\hat{\beta}_{i}$,
with
$$
\hat{\beta}_{i}=\left(\mathbf{x}_{i}^{\prime}
\mathbf{x}_{i}\right)^{-1} \mathbf{x}_{i}^{\prime} {y}_{i}
\text{~~and~~} \hat{\rho}_{i j}=\frac{\sum_{t=1}^{T} \hat{\epsilon}_{i
t} \hat{\epsilon}_{j t}}{\sqrt{\sum_{t=1}^{T} \hat{\epsilon}_{i
t}^{2} \sum_{t=1}^{T} \hat{\epsilon}_{j t}^{2}}}.
$$
The asymptotic null distribution of $\mathrm{LM}_{\mathrm{BP}}$ is a
chi-squared distribution with $N(N-1) / 2$ degrees of freedom, which
is established when $N$ is fixed and $T$ diverges to infinity. Hence,
it is not applicable to the case of large $N$. To overcome the size
distortions of the scaled version of the $\mathrm{LM_{BP}}$ test
proposed by \cite{pesaran2004} for large $N$ and small $T$,
\cite{pesaran2008} proposed a bias-adjusted test, with
test statistic
$$
\mathrm{LM_{PUY}}=\sqrt{\frac{2}{N(N-1)}} \sum_{j=2}^{N}
\sum_{i=1}^{j-1}\frac{(T-p) \hat{\rho}_{i j}^{2}-\mu_{T i j}}{v_{T i
j}},
$$
where
\begin{align*}
&\mu_{T i j}=\frac{1}{T-p} \tr\left[\mE\left(\mathbf{P}_{i}
\mathbf{P}_{j}\right)\right],~~v_{T i
j}^{2}=\left[\tr^{2}\left\{\mE\left(\mathbf{P}_{i}
\mathbf{P}_{j}\right)\right\}\right] a_{1 T}+2
\tr\left[\left\{\mE\left(\mathbf{P}_{i}
\mathbf{P}_{j}\right)\right\}^{2}\right] a_{2 T},\\
&a_{1 T}=a_{2 T}-\frac{1}{(T-p)^{2}} \text{~~and~~} a_{2
T}=3\left\{\frac{(T-p-8)(T-p+2)+24}{(T-p+2)(T-p-2)(T-p-4)}\right\}^{2}.
\end{align*}
Here, for each $1\leq i\leq N$,
$\mathbf{P}_{i}=\mathbf{I}_{T}-\mathbf{x}_{i}\left(\mathbf{x}_{i}^{\prime}
\mathbf{x}_{i}\right)^{-1} \mathbf{x}_{i}^{\prime}$,
hence $\hat{\epsilon}_{i\cdot}=\left(\hat{\epsilon}_{i 1}, \cdots,
\hat{\epsilon}_{i T}\right)^{\prime}=\mathbf{P}_{i} \epsilon_{i\cdot}$.
The asymptotic null distribution of $\mathrm{LM_{PUY}}$ is
$\mathcal{N}(0,1)$, which is established when $T \rightarrow \infty$
first and then $N \rightarrow \infty$.
Later, \cite{feng2020} established that $\mathrm{LM_{PUY}}\to
\mathcal{N}(0,1)$ in distribution when $\min(N,T)\to\infty$,
and also established that
$\mathrm{LM_{FJLX}}\to \mathcal{N}(0,1)$ in distribution when $\min(N,T)\to\infty$, where
\begin{align*}
\mathrm{LM_{FJLX}}=\frac{\sum_{j=2}^{N} \sum_{i=1}^{j-1}T \hat{\rho}_{i j}^{2}-\mu_{N}}{N} \text{ and } \mu_{N}=\frac{T}{(T-p)^2} \sum_{1 \leq i<j \leq N} \tr\left(\mathbf{P}_i \mathbf{P}_j\right).
\end{align*}
These tests are all based on the sum of squared sample correlations
$\sum\sum_{1\le i<j\le N}\hat{\rho}_{i j}^{2}$.
In contrast, \cite{pesaran2004} proposed a test for cross-sectional
independence directly based on the sum of sample correlations,
with test statistic
$$
\mathrm{C D_{P}}=\sqrt{\frac{2 T}{N(N-1)}} \sum_{j=2}^{N} \sum_{i=1}^{j-1} \hat{\rho}_{i j}.
$$
\cite{pesaran2015} established the asymptotic null distribution of
$\mathrm{C D_{P}}$ to be $\mathcal{N}(0,1)$ when $\min(N,T)\to\infty$.
Recently, to test cross-sectional correlation with serially
correlated errors, \cite{baltagi2016} proposed a new test with test
statistic
\begin{equation}\label{sumtest}
S_{N}=\sqrt{\frac{2}{N(N-1)}}\underset{1\leq i<j\leq N}{\sum\sum}\hat{\rho}_{i j},
\end{equation}
which is very similar to $\mathrm{CD_{P}}$, but has different
asymptotic variance. Under certain assumptions, they established that
under the null hypothesis,
\begin{equation}\label{sumtest0}
S_{N}/{\hat{\sigma}}_{S_{N}}\to \mathcal{N}(0,1) \text{~~in
distribution when~~} \min(N,T)\to\infty,
\end{equation}
hence a level-$\alpha$ test can be performed by rejecting $H_0$
when $S_N/\hat{\sigma}_{S_N}$ is larger than the $(1-\alpha)$-quantile
$z_{\alpha}= \Phi^{-1}(1-\alpha).$ Here,
\begin{align}\label{sigmasumhat}
\hat{\sigma}^2_{S_{N}}=\frac{2}{N(N-1)}\underset{1\leq i<j\leq
N}{\sum\sum} v_{j}^{\prime}\left(v_{i}-\bar{v}_{ij}\right)
v_{i}^{\prime}\left(v_{j}-\bar{v}_{i j}\right),
\end{align}
$\bar{v}_{ij}=\sum_{1<k \neq i, j<N} v_{k}/(N-2)$ and
$v_{k}=\hat{\epsilon}_{k\cdot}/\|\hat{\epsilon}_{k\cdot}\|$ for all $1\leq
k\leq N.$
\subsection{Max based test}
All tests mentioned in Section \ref{secR} are sum based tests, which
generally perform very well under dense alternatives,
i.e. alternatives with dense cross-sectional correlation matrices,
but may suffer from low power against sparse alternatives,
i.e. alternatives with sparse cross-sectional correlation matrices. To
deal with sparse alternatives,
we now propose a max based test based on the maximum of the squares of the
sample cross-sectional correlations. The test statistic we propose is
\begin{equation}\label{maxtest}
L_{N}=\max_{1\leq i<j\leq N}\hat{\rho}^2_{ij}.
\end{equation}
Max based tests have been widely studied in testing independence
among variables, e.g., \cite{lx2015}, \cite{chen2018} and
\cite{feng2020}. Specifically, \cite{feng2020} used it to test cross-sectional
independence under the assumption of no serial correlation between the
errors, and established that
$TL_{N}-4 \log N+\log\log N \to G(y)$ in distribution, where
$G(y)=\exp \big\{- \exp \left(-y/2\right)/\sqrt{8 \pi}\big\}$.
However, this test may perform poorly if blindly used in
the situation where serial correlation exists.
To utilize the test statistic in \eqref{maxtest}
for data with serial correlations, we must reinvestigate
the asymptotic properties of $L_{N}$.
We impose the following assumptions.
\begin{assumption}\label{assum:E distribution}
Assume that
$\mathbf{E}=(\epsilon_{1\cdot},\cdots,\epsilon_{N\cdot})^{\prime}=\mathbf{U}^{1/2}\mathbf{Z}\mathbf{\Sigma}^{1/2},$
where $\mathbf{U}\in \mathbb{R}^{N \times N}$ and $\mathbf{\Sigma}\in
\mathbb{R}^{T \times T}$ are positive definite matrices,
all elements of $\mathbf{Z}\in \mathbb{R}^{N \times T}$ are iid
variables with mean zero and variance one, the density function of
$(\mathbf{Z})_{it}$ is symmetric and the sub-Gaussian norm of
$(\mathbf{Z})_{it}$ is bounded by $K$,
i.e. $\mE\exp{(\mathbf{Z})^2_{it}/K^2}\le 2$, for each $1\leq i\leq N$
and $1\leq t\leq T$.
\end{assumption}
\begin{assumption}\label{assum:x_it}
(i) Assume that $p >0$ is fixed, and the regressors $x_{i t}$ are strictly exogenous, such that $$
\mE\left(\epsilon_{i t} | \mathbf{x}_{i}\right)=0, \text { for all } 1\leq i\leq N \text { and } 1\leq t\leq T
.$$ (ii)Assume that $\mathbf{x}_{i}^{\prime}\mathbf{x}_{i}/T=
\sum_{t=1}^{T} {x}_{i t} {x}_{i t}^{\prime}/T$ is non-singular and $\mathbf{x}_{i}\big(\mathbf{x}_{i}^{\prime}\mathbf{x}_i/T\big)^{-1} \mathbf{x}_{i}^{\prime}$
is stochasticly bounded for all $1\leq i\leq N.$
\end{assumption}
\begin{assumption}\label{assum:matrix}
(i) Assume that $C^{-1} \leq \lambda_{\min }(\mathbf{\Sigma}) \leq \lambda_{\max }(\mathbf{\Sigma})
\leq C$ and $C^{-1} \leq \lambda_{\min }(\mathbf{U}) \leq$
$\lambda_{\max }(\mathbf{U}) \leq C$ for some constant $C>0$.
(ii) Assume that $\max_{j=1, \dots T}\sum_{k=1}^{T}\left|(\mathbf{\Sigma})_{j
k}\right|^{\tau} \leq C'$
and $\max_{j=1, \dots
N}\sum_{k=1}^{N}\left|(\mathbf{U})_{j k}\right|^{\tau} \leq C'$
for some $0<\tau<2$ and $C'>0$.
(iii) Assume that all diagonal elements of $\mathbf{U}$ are one.
\end{assumption}
Assumption \ref{assum:E distribution} writes the error into
$\mathbf{E}=\mathbf{U}^{1/2}\mathbf{Z}\mathbf{\Sigma}^{1/2}$, which is
investigated by \cite{hornstein2018joint} in the case that the random
matrix $\mathbf{Z}$ has independent sub-Gaussian entries.
$\mathbf{U}$ represents the row covariance matrix of $\mathbf{E}$,
containing information on the cross-sectional correlation between
$\{\epsilon_{1\cdot},\cdots,\epsilon_{N\cdot}\}$, whereas
$\mathbf{\Sigma}$
represents the column covariance matrix of $\mathbf{E}$, containing
the serial correlation information between $\{\epsilon_{\cdot 1},\cdots,\epsilon_{\cdot
T}\}$. It is known that when $(\mathbf{Z})_{it}\overset{iid}{\sim}
\mathcal{N}(0,1)$, $\mathbf{E}$ has the matrix-variate normal
distribution $\mathcal{N}({\bf 0}, \mathbf{U}\otimes\mathbf{\Sigma})$,
where $\otimes$ denotes the Kronecker product.
Assumption \ref{assum:x_it}(i) is used in \cite{pesaran2008}, which is
a common condition for panel data model in \eqref{model}.
Assumption \ref{assum:x_it}(ii) is imposed to ensure that the
difference between the distribution of the max based statistic based on
the residuals and that based on the errors is negligible.
Assumption \ref{assum:matrix}(i) is the same as Condition (C1) in
\cite{chen2018}, which is a common eigenvalue assumption in
high-dimensional inference literature such as in \cite{crz2016} and
contains many important types of covariance matrices, including the
bandable, Toeplitz and sparse covariance matrices.
Assumption \ref{assum:matrix}(ii) is the same as Condition (C2) in
\cite{chen2018}, which assumes the sparsity of $\mathbf{U}$ and
$\mathbf{\Sigma}$.
Note that the correlation between the rows in $\mathbf{E}$ is the
correlation we are interested in testing, which corresponds to
the diagonality of $\U$. The
sparsity of $\U$ characterizes the sparse alternative. On the other
hand, the sparsity of $\mathbf{\Sigma}$ is imposed to ensure good
estimation property. To eliminate the complexity
caused by the serial correlation across $t$ within each row, ideally
we hope to work with
$\mathbf{E}\mathbf{\Sigma}^{-1/2}=\mathbf{U}^{1/2}\mathbf{Z}$, and to
test if $\U$ has zero off-diagonal elements. Thus, in our test
procedure, we will need to estimate
a $\mathbf{\Sigma}$ dependent quantity
$\tr^2(\mathbf{\Sigma})/\|\mathbf{\Sigma}\|_{\mathrm{F}}^2$. Since
$\mathbf{\Sigma}$ is a high dimensional matrix, sparsity is a common
assumption to regularize the properties of the related estimation.
Finally,
Assumption \ref{assum:matrix}(iii) is imposed to make the
error model identifiable.
Throughout the text, we let $\gamma$ be a positive finite constant.
\begin{theorem}\label{th:max null}
Under Assumptions \ref{assum:E distribution}-\ref{assum:matrix} and
the null hypothesis in \eqref{null hypothesis},
when $\min(N,T)\to\infty$ and $N/T\to\gamma$,
for any
$y\in\mathbb{R},$
we have
$$\mathbb{P}\left(\frac{\tr^2({\mathbf{\Sigma}})}{\|{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}
L_{N}-4 \log N+\log\log N \leq y\right) \rightarrow G(y),$$
where $G(y)=\exp \big\{- \exp \left(-y/2\right)/\sqrt{8 \pi}\big\}$
is a type-I Gumbel distribution function.
\end{theorem}
The column covariance matrix $\mathbf{\Sigma}$, which is needed in the
construction of the max based test statistic, can be
assessed via a procedure similar to that proposed in
\cite{chen2018} for testing independence among variables using
correlated samples. First, define the column sample covariance matrix
$\hat{\mathbf{\Sigma}}=\left(\hat{\sigma}_{i j}\right)_{1 \leq i, j \leq T}$ with
$$\hat{\sigma}_{i j}=\frac{1}{N-1} \sum_{l=1}^{N}\left(\hat{\epsilon}_{l i}-
\bar{\hat{\epsilon}}_{\cdot i}\right)\left(\hat{\epsilon}_{l j}-\bar{\hat{\epsilon}}_{\cdot j}
\right),\quad \bar{\hat{\epsilon}}_{\cdot j}=\frac{1}{N}
\sum_{l=1}^{N}{\hat{\epsilon}}_{lj} \text{~~and~~} \theta_{i
j}=\frac{\hat{\sigma}_{i j}}{\sqrt{\hat{\sigma}_{i i}
\hat{\sigma}_{j j}}}$$
for each $1 \leq i, j \leq T$. Then, define
$\tilde{\mathbf{\Sigma}}=\left(\tilde{\sigma}_{i j}\right)_{1 \leq i,
j \leq T}$ with
\begin{equation}\label{v>1.42}
\tilde{\sigma}_{i j}=\left\{
\begin{aligned}
&\hat{\sigma}_{i j} I\bigg\{\frac{|{\theta}_{i j}|}{1-{\theta}_{i j}^{2}} \geq
\nu \sqrt{\frac{\hat{P}_{N} \log T}{N}}\bigg\} , & i\neq j, \\
&\hat{\sigma}_{i i} , & i=j
\end{aligned}
\right.
\end{equation}
for each $1 \leq i, j \leq T$. Here, $\nu>\sqrt{2}$ and $
\hat{P}_{N}=\left[\|\hat{\mathbf{U}}\|_{\mathrm{F}}^{2}-\frac{1}{T}
\{\tr(\hat{\mathbf{U}})\}^{2}\right]/N,$
where $(\hat{\mathbf{U}})_{ij}=\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot}/
\tr(\hat{\mathbf{\Sigma}})$ for each $1 \leq i, j \leq N.$
Based on $\tilde{\mathbf{\Sigma}}$, $\tr^2(\mathbf{\Sigma})$ and
$\|{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}$
are estimated by $\tr^2(\tilde{\mathbf{\Sigma}})$ and
$\|{\tilde{\mathbf{\Sigma}}}\|_{\mathrm{F}}^{2}$, respectively.
Hence, ${\tr^2({\mathbf{\Sigma}})}/{\|{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}$
is estimated by ${\tr^2(\tilde{\mathbf{\Sigma}})}/{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}$,
which is a ratio-consistent estimator of
${\tr^2({\mathbf{\Sigma}})}/{\|{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}$
as indicated in the following theorem.
\begin{theorem}\label{th:max cons}
Under Assumptions \ref{assum:E distribution}-\ref{assum:matrix} and
the null hypothesis in \eqref{null hypothesis}, for any $\nu > \sqrt{2}$,
\begin{align}
\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}
\frac{\|{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}{\tr^2({\mathbf{\Sigma}})}
=1+O_{\mathbb{P}}\left\{\Big(\frac{\log T}{N}\Big) ^{ \frac{1}{2}\wedge(1-\frac{\tau}{2})}\right\},
\end{align}
when $\min(N,T)\to\infty$ and
$N/T\to\gamma$.
\end{theorem}
Combining Theorems \ref{th:max null} and \ref{th:max cons}, we have that for any $y\in\mathbb{R},$
\begin{align}\label{maxtest0}
\mathbb{P}\left(\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}
L_{N}-4 \log N+\log \log N \leq y\right) \rightarrow G(y).
\end{align}
Based on \eqref{maxtest0}, for a given significance level $\alpha,$
the null hypothesis in \eqref{null hypothesis} will be rejected by the
established max based test when
$L_{N}{\tr^2(\tilde{\mathbf{\Sigma}})}/{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}
\geq w_{\alpha}+ 4 \log N-\log \log N$, where $w_{\alpha}$ is the $1-\alpha$ quantile of
the type I extreme value distribution with the cumulative
distribution function $G(y)$ and has the specific form of
$w_{\alpha}=\log (8 \pi)-2 \log \log (1-\alpha)^{-1}.$
Next, we turn to the power analysis of the proposed max based test.
\begin{theorem}\label{th:maxpower}
Under Assumptions \ref{assum:E distribution}-\ref{assum:matrix},
suppose that for some $\delta>2,$ some $\nu>0,$ and sufficiently large $N$ and $T$,
$
u_{N}=\max _{1 \leq i<j \leq N}\left|(\mathbf{U})_{i j}\right| \geq
\delta \sqrt{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2} \log
N/\tr^{2}(\mathbf{\Sigma})},
$
then
\begin{align}\label{maxpower}
\mathbb{P}\left(\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}
L_{N}-4 \log N+\log \log N> w_{\alpha}\right) \rightarrow 1,
\end{align}
as $\min(N,T)\to\infty$ and $N/T\to\gamma$.
\end{theorem}
Theorem \ref{th:maxpower} indicates that the proposed
max based test is consistent under the sparse alternative in which
the maximum non diagonal entries of $\mathbf{U}$ is
sufficiently large. More specifically, because
$\sqrt{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}
/\tr^{2}(\mathbf{\Sigma})}\asymp T^{-1/2}$ under Assumption
\ref{assum:matrix}, the test is able to
detect the dependence as long as a single covariance is at the order
of $(\log N/T)^{1/2}$, which leads to an overall detection
rate of $(\log N/T)^{1/2}$. In contrast, the sum based test
can detect the departure from null if each covariance
reaches $T^{-1/2}N^{-1}$
(Theorem 4 of \cite{baltagi2016}), hence the overall detection rate is $N/T^{1/2}$.
This theoretical result is consistent with the simulation results in
Section \ref{sec:simu}, where the max based test performs
better than the sum based test in terms of empirical power
under sparse alternatives.
The presence of the possible serial correlation makes it
more difficult to establish asymptotic properties of
$L_{N}$. For example, we need to reestablish the
asymptotic variance of $L_{N}$ that depends on $\mathbf{\Sigma}$. This is
the reason why we replace the test statistic $TL_{N}-4 \log N+\log\log
N$ in \cite{feng2020}, where no serial correlation is considered, by
$L_{N}{\tr^2({\mathbf{\Sigma}})}/{\|{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}-4
\log N+\log\log N$ and
$L_{N}{\tr^2(\tilde{\mathbf{\Sigma}})}/{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}-4
\log N+\log\log N$.
On the other hand, we note that the max based test for testing
independence among normally distributed variables with correlated
samples has been studied in \cite{chen2018}, whereas in this paper we
consider the max based test under panel data models with
strictly
exogenous regressors, and we relax the Gaussian assumption to
sub-Gaussian. The proof under sub-Gaussianity
is much more challenging because many properties associated with
Gaussianity, such as rotation invariance, can no longer
be used.
To establish the theoretical results under sub-Gaussian distributions,
more advanced technical tools, such as the
Hanson-Wright inequality \citep{rudelson2013}, need to be engaged in deriving
the asymptotical distribution of the max based test statistic
$L_{N}$. In addition, the expressions of the moments of the quadratic
forms under sub-Gaussian distributions are also much more complex than that
under Gaussian distributions, which further complicate the proof.
\subsection{Fisher's combined probability test}
It is intuitively expected that the sum based test $S_N$ performs well
under
dense alternatives, whereas the max based test $L_N$ performs well
under sparse alternatives. However, in practice, it is usually
unknown whether the correlation matrix of the errors is sparse or
not, hence it is difficult to decide which test to
use. For this reason, in this subsection, we propose using Fisher's
combined probability test by combining the sum based and max based
tests, which is expected to take advantage of both
tests. To construct the test, we first
need to establish the asymptotic
independence between the two statistics $S_N$ and $L_N$ under the
null hypothesis.
\begin{theorem}\label{th:sum-max}
Under Assumptions \ref{assum:E distribution}-\ref{assum:matrix} and
the null hypothesis in \eqref{null hypothesis},
$L_{N}{\tr^2(\tilde{\mathbf{\Sigma}})}/{\|
\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}
-4 \log N+\log \log N $ and $S_{N}/\hat{\sigma}_{S_{N}}$ are
asymptotically independent, as $\min(N,T)\to\infty$ and
$N/T\to\gamma$.
\end{theorem}
Note that in many related studies, such as \cite{lx2015} and \cite{feng2020},
the asymptotic independence between the sum based and max based
statistics is established under the assumption that all components of
the random vectors concerned are independent and identically
distributed. In contrast, we establish the asymptotic independence in
the situation where the error vectors may have serial correlation,
which makes the proof of the asymptotic independence much more
challenging and requires more complex technical treatments
and tools.
Let $P_{L_{N}} = 1-G\Big\{L_{N}{\tr^2(\tilde{\mathbf{\Sigma}})}/
{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}
-4\log N+\log \log N \Big\}$ and
$P_{S_{N}} =1-\Phi\left(S_{N} / \hat{\sigma}_{S_{N}}\right)$. We
construct Fisher's combined probability test statistic as
\begin{align}\label{testSUMMAX}
T_{C}=-2\log(P_{L_{N}})-2\log(P_{S_{N}}).
\end{align}
By Theorem \ref{th:sum-max}, \eqref{sumtest0} and \eqref{maxtest0},
we see that $P_{L_{N}}$ and $P_{S_{N}}$
are asymptotically independent under
the null hypothesis and each has limit distribution $U[0,
1]$, the uniform distribution on $[0, 1]$. We thus obtain
the following corollary.
\begin{corollary}\label{Tc}
Assume that the assumptions in Theorem \ref{th:sum-max} hold,
then we have $T_{ C } \rightarrow \chi_{4}^{2}$ in distribution when
$\min(N,T)\to\infty$ and $N/T\to\gamma$.
\end{corollary}
According to Corollary \ref{Tc}, we proposed Fisher's combined
probability tes at level $\alpha$ by rejecting the null hypothesis in \eqref{null
hypothesis} if $T_{C} \geq q_{\alpha}$, where
$q_{\alpha}$ is the $1-\alpha$ quantile of $\chi_{4}^{2}$.
\section{Simulation studies}\label{sec:simu}
We now conduct simulations to investigate the finite sample performance of
the two tests proposed in this paper, i.e.
the test based on $L_{N}$ and the Fisher's combined probability test based
on $T_{C}$. For comparison, we also implement three
other existing tests, i.e. the test based on $S_{N}$
proposed by \cite{baltagi2016}, the $\mathrm{LM_{PUY}}$ test
proposed by \cite{pesaran2008} and the $\mathrm{CD_{P}}$ test proposed by \cite{pesaran2004}.
Here, the max based test is implemented by setting $\nu = 1.42$ in
\eqref{v>1.42} as in \cite{chen2018}.
For simplicity, we will abbreviate these five tests as $L_{N}$,
$T_{C}$, $S_{N}$, $\mathrm{LM_{PUY}}$ and $\mathrm{CD_{P}}$,
respectively.
We consider the data generating process
\begin{align}\label{simu model}
y_{i t}=\alpha_{i}+\sum_{l=2}^{p}x_{li,t} \beta_{li}+\epsilon_{i t},\,\, 1\leq i\leq N,\,\,1\leq t\leq T,
\end{align}
where $x_{it}=\left(1, x_{2 i ,t}, \cdots, x_{pi,t }\right)^{\prime}
\in \mathbb{R}^{p}$ and $\beta_{i}= \left(\alpha_{i}, \beta_{ 2i}, \cdots, \beta_{ pi }\right)^{\prime} \in \mathbb{R}^{p}$.
We generate $\alpha_{i} \mathop{\sim}\limits^{iid} \mathcal{N}(0,1)$ and
$\beta_{li} \mathop{\sim}\limits^{iid}\mathcal{N}(1,0.04)$ for $2 \leq l \leq p.$
The strictly exogenous regressors are generated by $x_{li ,t }=0.6 x_{li, t-1 }+v_{li ,t }$
for $1\leq i\leq N ,$ $-50 \leq t \leq T$ and $2\leq l \leq p$, where
$x_{ li,- 51}=0$ and $v_{li, t } \mathop{\sim}\limits^{iid}\mathcal{N}\left(0, \psi_{li }^{2}
/\left(1-0.6^{2}\right)\right)$ with $\psi_{li }^{2} \mathop{\sim}\limits^{iid}
\chi_{6}^{2} / 6 $.
Consider the following two settings of serial correlation of the errors $\epsilon_{i t}$.
\begin{enumerate}[(i)]
\item The errors follow an
auto-regressive (AR) model of order one over time, i.e. AR(1): $\epsilon_{i 1}=e_{i 1}$
and $\epsilon_{i t}=0.6 \epsilon_{i t-1}+e_{i t}$
for $2\leq t\leq T$ and $1\leq i\leq N.$ \label{H01}
\item The errors follow an auto-regressive and
moving average (ARMA) model of order (1,1) over time, i.e. ARMA(1,1): $\epsilon_{i 1}=e_{i 1}$
and
$\epsilon_{i t}=0.6 \epsilon_{i t-1}+e_{i t}+0.2e_{i t-1}$ for $2\leq t\leq T$ and $1\leq i\leq N.$
\label{H02}
\end{enumerate}
To produce data under the null hypothesis, $e_{it}$ are independently generated from the following three distributions:
(1) $\mathcal{N}(0,1);$
(2) $t_{6}/\sqrt{6/4};$
(3) $(\chi_{5}^{2}-5)/\sqrt{10}$.
Then, we turn to produce data under the alternative hypothesis. The
data generating process is specified as
\begin{align}\label{simu model1}
y_{i t}=\alpha_{i}+\sum_{l=2}^{p}x_{li,t} \beta_{li}+\epsilon_{i t}^{\ast},\,\, 1\leq i\leq N,\,\,1\leq t\leq T,
\end{align}
where $\epsilon_{i t}^{\ast}$ are generated from the following two settings.
\begin{enumerate}[(I)]
\item Non-sparse case. Spatial moving average (SMA) model with order one, i.e. SMA(1): for all $1\leq t\leq T,$ $\epsilon_{1 t}^{*}=0.5\delta\epsilon_{2 t}+\epsilon_{1 t},$
$\epsilon_{N t}^{*}=0.5\delta\epsilon_{N-1 t}+\epsilon_{N t}$ and
$\epsilon_{i t}^{*}=\delta\left(0.5 \epsilon_{i-1 t}+0.5 \epsilon_{i+1 t}\right)
+\epsilon_{i t},$
where $2\leq i\leq N-1,$ $\delta=0.2$ and $\epsilon_{it}$ are
generated from settings (\ref{H01})-(\ref{H02}) with distributions
(1)-(3).\label{H11}
\item Sparse case. Let $(\epsilon_{1\cdot}^{\ast},\cdots,\epsilon_{N\cdot}^{\ast})^{\prime}
=\mathbf{\Psi}^{1/2}(\epsilon_{1\cdot},\cdots,\epsilon_{N\cdot})^{\prime},$
where $\epsilon_{i\cdot}^{\ast}=(\epsilon_{i1}^{\ast},\cdots,\epsilon_{iT}^{\ast})^{\prime}$ and $\epsilon_{i\cdot}=(\epsilon_{i1},\cdots,\epsilon_{iT})^{\prime},$ for all $1\leq i\leq N.$ Here, $\epsilon_{i\cdot}$ are generated from settings (\ref{H01})-(\ref{H02}) with distributions (1)-(3).
$\mathbf{\Psi}$ is constructed as follows.
Randomly select a subset $S \subset\{1, \cdots, N\}$ with cardinality
$\lceil N^{0.3}\rceil$, the greatest integer grater than or equal to $N^{0.3}.$ Let $(\mathbf{\Psi})_{i j}=1$
if $i=j$.
For $i<j$, define $(\mathbf{\Psi})_{i j}=0$ if $i \notin S$ or $j \notin S$,
and $(\mathbf{\Psi})_{i j}\mathop{\sim}\limits^{iid} U[\sqrt{4 \log N / T},
\sqrt{6\log N / T}]$ if $i, j \in S.$
\label{H12}
\end{enumerate}
In addition, for the choices of $p$, $N$ and $T$, we set $p\in \{3,5\},$ $N \in\{100, 200\}$, $T \in \{200, 300, 400,500\}.$
The results of the empirical size and power of the five tests in the
non-sparse and sparse cases of error correlation matrices are
summarized in Tables \ref{t1}-\ref{t3}. The power curves are
plotted in Figure \ref{fig1}. All results are based on 1,000
replications. We next analyze them in detail.
Table \ref{t1} indicates that in most cases $T_{C}$, $L_{N}$ and
$S_{N}$ have empirical sizes not much larger than $5\%$. Here, the
max based test $L_{N}$ and Fisher's combined probability test $T_{C}$
tend to have smaller empirical sizes than the sum based test $S_{N}$,
especially as $T$ is relatively small. This is not surprising
and is common for many max based tests, because the convergence rate
of the type I extreme distribution is typically slow
\citep{liu2008asymptotic}.
$\mathrm{CD_{P}}$ and $\mathrm{LM_{PUY}}$ fail to control the
empirical size because $\mE(\hat{\rho}_{ij}^2)$ may be seriously
affected by serial correlation.
This is also observed by \cite{baltagi2016}.
Tables \ref{t2} and \ref{t3} show the empirical powers in both
non-sparse and sparse cases of correlation matrices.
Since $\mathrm{CD_{P}}$ and $\mathrm{LM_{PUY}}$ fail to control the
empirical size, we exclude them from the empirical power results.
Table \ref{t2} and \ref{t3} indicate that $S_{N}$ and $T_{C}$
generally perform better than $L_{N}$ in non-sparse cases in terms of
empirical
powers, whereas $L_{N}$ and $T_{C}$ generally perform better than
$S_{N}$ in sparse cases. As expected, Fisher's combined probability
test $T_{C}$ has power
advantages regardless the local alternative is sparse or not.
\begin{table}[!ht]
\centering
\caption{The empirical size of the five tests under
settings (\ref{H01}) and (\ref{H02}) with distributions (1)-(3) at 5\% level. }\label{t1}
\vspace{0.35cm}
\renewcommand{\arraystretch}{0.98}
\tabcolsep 3pt
\scalebox{0.75}{
\begin{tabular}{cccccccccccccccccccccccc}\hline \hline \multicolumn{1}{l}{} & \multicolumn{1}{l}{} & \multicolumn{1}{l|}{} & \multicolumn{8}{c|}{Setting (\ref{H01})} & \multicolumn{8}{c}{Setting (\ref{H02})} \\\hline
& \multicolumn{2}{c|}{$p$} & \multicolumn{4}{c|}{3} & \multicolumn{4}{c|}{5} & \multicolumn{4}{c|}{3} & \multicolumn{4}{c}{5} \\\hline
& $N$ & \multicolumn{1}{c|}{$T$} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & 500 \\\hline
& \multicolumn{18}{c}{(1) Normal distribution} \\\hline
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 6.2 & 5.4 & 4.6 & \multicolumn{1}{c|}{5.9} & 5.7 & 6.4 & 5.2 & \multicolumn{1}{c|}{4.8} & 6.0 & 5.5 & 4.4 & \multicolumn{1}{c|}{5.6} & 5.6 & 6.5 & 5.2 & 4.6 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 3.4 & 3.2 & 3.9 & \multicolumn{1}{c|}{4.9} & 3.6 & 4.4 & 3.9 & \multicolumn{1}{c|}{3.9} & 2.2 & 2.5 & 3.7 & \multicolumn{1}{c|}{4.4} & 3.4 & 3.9 & 3.4 & 4.0 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 3.5 & 4.8 & 4.6 & \multicolumn{1}{c|}{4.9} & 3.9 & 5.4 & 4.3 & \multicolumn{1}{c|}{4.9} & 3.3 & 4.3 & 4.2 & \multicolumn{1}{c|}{4.2} & 3.3 & 5.1 & 4.1 & 3.6 \\
& & \multicolumn{1}{c|}{$\mathrm{CD_{P}}$} & 11.5 & 12.1 & 11.5 & \multicolumn{1}{c|}{13.6} & 12.5 & 14.3 & 12.2 & \multicolumn{1}{c|}{12.3} & 12.9 & 14.0 & 13.5 & \multicolumn{1}{c|}{15.5} & 14.3 & 16.3 & 14.3 & 14.0 \\
& & \multicolumn{1}{c|}{$\mathrm{LM_{PUY}}$} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 5.4 & 5.1 & 3.7 & \multicolumn{1}{c|}{6.4} & 5.6 & 5.7 & 5.7 & \multicolumn{1}{c|}{5.4} & 5.9 & 5.1 & 4.0 & \multicolumn{1}{c|}{6.6} & 5.9 & 5.7 & 6.3 & 5.3 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 2.1 & 2.9 & 3.6 & \multicolumn{1}{c|}{3.6} & 1.8 & 2.4 & 2.2 & \multicolumn{1}{c|}{3.1} & 1.5 & 2.7 & 2.6 & \multicolumn{1}{c|}{3.4} & 0.9 & 1.9 & 1.7 & 2.1 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 3.7 & 4.3 & 3.6 & \multicolumn{1}{c|}{5.1} & 3.4 & 4.3 & 3.5 & \multicolumn{1}{c|}{3.7} & 3.3 & 3.7 & 3.4 & \multicolumn{1}{c|}{4.3} & 2.9 & 3.6 & 3.0 & 3.5 \\
& & \multicolumn{1}{c|}{$\mathrm{CD_{P}}$} & 12.0 & 11.9 & 10.2 & \multicolumn{1}{c|}{13.9} & 12.3 & 12.6 & 13.0 & \multicolumn{1}{c|}{12.9} & 13.2 & 14.1 & 12.6 & \multicolumn{1}{c|}{15.3} & 13.7 & 14.8 & 15.2 & 14.9 \\
& & \multicolumn{1}{c|}{$\mathrm{LM_{PUY}}$} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & 100 \\ \hline
& \multicolumn{18}{c}{(2) $t_{6}$-distribution} \\ \cline{2-19}
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 4.9 & 5.7 & 5.2 & \multicolumn{1}{c|}{5.9} & 6.3 & 5.5 & 4.8 & \multicolumn{1}{c|}{5.2} & 4.7 & 5.8 & 5.1 & \multicolumn{1}{c|}{5.9} & 6.2 & 5.6 & 4.9 & 5.3 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 3.3 & 4.5 & 3.4 & \multicolumn{1}{c|}{5.2} & 2.0 & 2.9 & 3.8 & \multicolumn{1}{c|}{4.3} & 2.3 & 3.7 & 3.1 & \multicolumn{1}{c|}{5.1} & 1.5 & 2.6 & 3.7 & 3.9 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 3.7 & 5.6 & 4.8 & \multicolumn{1}{c|}{6.0} & 3.9 & 4.1 & 4.2 & \multicolumn{1}{c|}{4.6} & 3.5 & 5.2 & 4.7 & \multicolumn{1}{c|}{5.6} & 3.9 & 4.3 & 4.7 & 4.6 \\
& & \multicolumn{1}{c|}{$\mathrm{CD_{P}}$} & 11.9 & 11.9 & 11.9 & \multicolumn{1}{c|}{12.5} & 12.3 & 13.5 & 11.5 & \multicolumn{1}{c|}{11.7} & 14.0 & 13.9 & 13.8 & \multicolumn{1}{c|}{14.0} & 13.2 & 15.3 & 13.7 & 13.6 \\
& & \multicolumn{1}{c|}{$\mathrm{LM_{PUY}}$} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 5.1 & 4.7 & 4.6 & \multicolumn{1}{c|}{4.9} & 5.4 & 6.7 & 6.0 & \multicolumn{1}{c|}{4.5} & 4.9 & 5.0 & 4.7 & \multicolumn{1}{c|}{5.3} & 5.4 & 7.1 & 5.8 & 4.8 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 2.4 & 3.5 & 4.2 & \multicolumn{1}{c|}{4.2} & 3.0 & 2.5 & 3.2 & \multicolumn{1}{c|}{3.2} & 1.4 & 2.1 & 2.9 & \multicolumn{1}{c|}{3.9} & 2.0 & 2.4 & 3.0 & 2.8 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 4.2 & 4.7 & 4.4 & \multicolumn{1}{c|}{5.0} & 4.6 & 5.3 & 4.5 & \multicolumn{1}{c|}{4.2} & 3.8 & 4.0 & 3.8 & \multicolumn{1}{c|}{5.0} & 4.2 & 5.4 & 4.9 & 4.1 \\
& & \multicolumn{1}{c|}{$\mathrm{CD_{P}}$} & 10.5 & 11.8 & 9.3 & \multicolumn{1}{c|}{14.0} & 11.4 & 12.0 & 13.2 & \multicolumn{1}{c|}{12.9} & 12.4 & 13.9 & 11.3 & \multicolumn{1}{c|}{15.8} & 13.3 & 14.3 & 15.0 & 15.2 \\
& & \multicolumn{1}{c|}{$\mathrm{LM_{PUY}}$} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & 100 \\ \hline
& \multicolumn{18}{c}{(3) $\chi_{5}^2$-distribution} \\ \cline{2-19}
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 5.6 & 5.3 & 5.8 & \multicolumn{1}{c|}{6.4} & 6.7 & 5.2 & 4.8 & \multicolumn{1}{c|}{5.8} & 5.8 & 5.2 & 6.0 & \multicolumn{1}{c|}{6.6} & 6.9 & 5.2 & 4.6 & 5.6 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 3.4 & 3.4 & 3.4 & \multicolumn{1}{c|}{3.1} & 3.4 & 2.8 & 4.6 & \multicolumn{1}{c|}{3.7} & 2.5 & 2.3 & 2.9 & \multicolumn{1}{c|}{3.5} & 2.1 & 2.6 & 4.1 & 3.3 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 5.7 & 3.7 & 4.7 & \multicolumn{1}{c|}{5.3} & 5.0 & 4.0 & 3.9 & \multicolumn{1}{c|}{5.2} & 4.5 & 3.6 & 4.2 & \multicolumn{1}{c|}{4.8} & 4.3 & 3.3 & 3.9 & 4.8 \\
& & \multicolumn{1}{c|}{$\mathrm{CD_{P}}$} & 13.9 & 13.2 & 12.4 & \multicolumn{1}{c|}{13.7} & 12.4 & 11.7 & 11.4 & \multicolumn{1}{c|}{13.0} & 15.6 & 15.6 & 14.6 & \multicolumn{1}{c|}{14.8} & 13.7 & 12.9 & 13.6 & 15.1 \\
& & \multicolumn{1}{c|}{$\mathrm{LM_{PUY}}$} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 6.5 & 5.2 & 5.7 & \multicolumn{1}{c|}{6.5} & 5.4 & 5.9 & 6.3 & \multicolumn{1}{c|}{5.5} & 6.8 & 5.7 & 5.7 & \multicolumn{1}{c|}{6.6} & 5.7 & 5.8 & 6.7 & 5.1 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 1.9 & 3.3 & 3.9 & \multicolumn{1}{c|}{4.9} & 3.1 & 3.9 & 4.1 & \multicolumn{1}{c|}{3.7} & 1.2 & 2.7 & 3.2 & \multicolumn{1}{c|}{3.8} & 1.7 & 3.4 & 3.6 & 3.2 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 4.4 & 4.5 & 5.0 & \multicolumn{1}{c|}{5.1} & 4.7 & 4.6 & 5.2 & \multicolumn{1}{c|}{3.9} & 4.0 & 4.1 & 4.5 & \multicolumn{1}{c|}{5.0} & 3.8 & 4.1 & 4.4 & 3.8 \\
& & \multicolumn{1}{c|}{$\mathrm{CD_{P}}$} & 13.3 & 12.8 & 13.3 & \multicolumn{1}{c|}{14.5} & 13.4 & 13.0 & 13.5 & \multicolumn{1}{c|}{12.2} & 14.6 & 14.4 & 15.6 & \multicolumn{1}{c|}{16.6} & 16.3 & 14.4 & 15.4 & 13.3 \\
& & \multicolumn{1}{c|}{$\mathrm{LM_{PUY}}$} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & 100 \\ \hline
\hline
\end{tabular}}
\end{table}
\begin{table}[!ht]
\centering
\caption{The empirical power of the three tests at 5\% level under case (\ref{H11}). }\label{t2}
\vspace{0.35cm}
\renewcommand{\arraystretch}{1.1}
\tabcolsep 3.8pt
\scalebox{0.75}{
\begin{tabular}{cccccccccccccccccccccccc}\hline \hline \multicolumn{1}{l}{} & \multicolumn{1}{l}{} & \multicolumn{1}{l|}{} & \multicolumn{8}{c|}{Setting (\ref{H01})} & \multicolumn{8}{c}{Setting (\ref{H02})} \\\hline
& \multicolumn{2}{c|}{$p$} & \multicolumn{4}{c|}{3} & \multicolumn{4}{c|}{5} & \multicolumn{4}{c|}{3} & \multicolumn{4}{c}{5} \\\hline
& $N$ & \multicolumn{1}{c|}{$T$} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & 500 \\\hline
& \multicolumn{18}{c}{(1) Normal distribution} \\\hline
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 79.3 & 89.2 & 95.5 & \multicolumn{1}{c|}{98.5} & 76.3 & 88.6 & 95.4 & \multicolumn{1}{c|}{98.5} & 73.4 & 84.7 & 92.0 & \multicolumn{1}{c|}{96.6} & 70.3 & 83.8 & 91.9 & 95.8 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 40.9 & 83.3 & 98.7 & \multicolumn{1}{c|}{100} & 35.6 & 82.9 & 98.9 & \multicolumn{1}{c|}{100} & 23.5 & 59.2 & 90.4 & \multicolumn{1}{c|}{99.7} & 22.1 & 63.7 & 89.4 & 99.0 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 87.5 & 98.3 & 100 & \multicolumn{1}{c|}{100} & 84.6 & 98.5 & 99.9 & \multicolumn{1}{c|}{100} & 77.5 & 93.8 & 99.0 & \multicolumn{1}{c|}{99.9} & 74.6 & 94.9 & 99.6 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 77.2 & 90.3 & 95.1 & \multicolumn{1}{c|}{98.1} & 78.1 & 89.4 & 95.3 & \multicolumn{1}{c|}{97.9} & 70.9 & 85.4 & 91.1 & \multicolumn{1}{c|}{96.4} & 71.8 & 84.8 & 92.3 & 95.7 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 27.9 & 77.7 & 98.2 & \multicolumn{1}{c|}{100} & 25.2 & 77.0 & 97.3 & \multicolumn{1}{c|}{100} & 12.7 & 53.0 & 88.7 & \multicolumn{1}{c|}{98.8} & 12.6 & 52.9 & 86.5 & 99.2 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 82.8 & 98.8 & 99.9 & \multicolumn{1}{c|}{100} & 81.6 & 98.0 & 100 & \multicolumn{1}{c|}{100} & 70.3 & 94.3 & 99.1 & \multicolumn{1}{c|}{99.8} & 70.8 & 94.5 & 99.2 & 100 \\
\hline
& \multicolumn{18}{c}{(2) $t_{6}$-distribution} \\ \cline{2-19}
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 80.1 & 91.0 & 95.5 & \multicolumn{1}{c|}{98.3} & 78.0 & 87.8 & 94.4 & \multicolumn{1}{c|}{97.8} & 74.2 & 85.1 & 92.1 & \multicolumn{1}{c|}{95.9} & 72.5 & 83.6 & 91.0 & 95.9 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 35.9 & 79.3 & 98.4 & \multicolumn{1}{c|}{100} & 34.9 & 81.5 & 97.9 & \multicolumn{1}{c|}{100} & 21.9 & 57.8 & 90.3 & \multicolumn{1}{c|}{99.4} & 20.4 & 60.9 & 88.6 & 99.0 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 86.8 & 97.6 & 99.9 & \multicolumn{1}{c|}{100} & 85.0 & 98.3 & 100 & \multicolumn{1}{c|}{100} & 75.2 & 92.4 & 99.1 & \multicolumn{1}{c|}{99.9} & 75.5 & 93.9 & 99.1 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 77.9 & 90.7 & 94.9 & \multicolumn{1}{c|}{98.0} & 79.3 & 89.7 & 95.8 & \multicolumn{1}{c|}{98.3} & 72.2 & 86.4 & 92.3 & \multicolumn{1}{c|}{96.4} & 73.4 & 85.0 & 92.2 & 96.3 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 27.3 & 78.8 & 98.6 & \multicolumn{1}{c|}{100} & 24.3 & 75.4 & 98.4 & \multicolumn{1}{c|}{100} & 13.0 & 54.7 & 87.3 & \multicolumn{1}{c|}{98.1} & 12.9 & 51.2 & 86.5 & 99.2 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 83.4 & 98.5 & 99.9 & \multicolumn{1}{c|}{100} & 82.8 & 98.3 & 100 & \multicolumn{1}{c|}{100} & 72.8 & 93.4 & 98.9 & \multicolumn{1}{c|}{100} & 70.8 & 93.7 & 99.1 & 100 \\
\hline
& \multicolumn{18}{c}{(3) $\chi_{5}^2$-distribution} \\ \cline{2-19}
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 79.6 & 90.5 & 95.6 & \multicolumn{1}{c|}{98.0} & 78.8 & 90.2 & 94.8 & \multicolumn{1}{c|}{98.9} & 73.6 & 86.6 & 93.0 & \multicolumn{1}{c|}{96.1} & 74.3 & 85.9 & 91.8 & 97.3 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 41.4 & 82.5 & 98.2 & \multicolumn{1}{c|}{100} & 36.8 & 81.7 & 98.0 & \multicolumn{1}{c|}{100} & 24.7 & 62.9 & 90.3 & \multicolumn{1}{c|}{98.8} & 21.3 & 60.9 & 88.4 & 99.1 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 87.6 & 98.9 & 99.9 & \multicolumn{1}{c|}{100} & 86.9 & 98.1 & 99.7 & \multicolumn{1}{c|}{100} & 77.4 & 94.6 & 99.3 & \multicolumn{1}{c|}{100} & 76.9 & 93.0 & 99.1 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 77.7 & 90.0 & 96.6 & \multicolumn{1}{c|}{99.3} & 81.7 & 90.5 & 97.2 & \multicolumn{1}{c|}{97.8} & 71.7 & 85.6 & 93.6 & \multicolumn{1}{c|}{98.0} & 74.8 & 84.5 &94.4 & 96.1 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 31.4 & 78.6 & 98.5 & \multicolumn{1}{c|}{100} & 33.8 & 79.1 & 98.4 & \multicolumn{1}{c|}{100} & 16.5 & 54.1 & 90.2 & \multicolumn{1}{c|}{99.3} & 17.4 & 54.4 &88.6 & 99.0 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 84.0 & 98.5 & 100 & \multicolumn{1}{c|}{100} & 86.7 & 98.8 & 100 & \multicolumn{1}{c|}{100} & 74.0 & 93.7 & 99.6 & \multicolumn{1}{c|}{100} & 75.1 & 94.2 &99.9 & 100 \\
\hline\hline
\end{tabular}
}
\end{table}
\begin{table}[!ht]
\centering
\caption{The empirical power of the three tests at 5\% level under case (\ref{H12}). }\label{t3}
\vspace{0.35cm}
\renewcommand{\arraystretch}{1.1}
\tabcolsep 3.8pt
\scalebox{0.75}{
\begin{tabular}{cccccccccccccccccccccccc}\hline \hline \multicolumn{1}{l}{} & \multicolumn{1}{l}{} & \multicolumn{1}{l|}{} & \multicolumn{8}{c|}{Setting (\ref{H01})} & \multicolumn{8}{c}{Setting (\ref{H02})} \\\hline
& \multicolumn{2}{c|}{$p$} & \multicolumn{4}{c|}{3} & \multicolumn{4}{c|}{5} & \multicolumn{4}{c|}{3} & \multicolumn{4}{c}{5} \\\hline
& $N$ & \multicolumn{1}{c|}{$T$} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & \multicolumn{1}{c|}{500} & 200 & 300 & 400 & 500 \\\hline
& \multicolumn{18}{c}{(1) Normal distribution} \\ \cline{2-19}
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 11.8 & 14.4 & 13.6 & \multicolumn{1}{c|}{18.1} & 12.8 & 16.0 & 15.7 & \multicolumn{1}{c|}{16.4} & 10.5 & 12.4 & 13.8 & \multicolumn{1}{c|}{16.8} & 11.8 & 14.7 & 13.8 & 15.2 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 99.1 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 97.0 & 100 & 100 & \multicolumn{1}{c|}{100} & 97.1 & 100 & 100 & 100 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 98.5 & 100 & 100 & \multicolumn{1}{c|}{100} & 99.5 & 100 & 100 & \multicolumn{1}{c|}{100} & 94.9 & 100 & 100 & \multicolumn{1}{c|}{100} & 94.2 & 100 & 100 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 8.2 & 9.5 & 8.2 & \multicolumn{1}{c|}{12.6} & 8.9 & 9.9 & 11.7 & \multicolumn{1}{c|}{11.8} & 8.4 & 9.3 & 8.0 & \multicolumn{1}{c|}{12.0} & 8.0 & 9.8 & 11.9 & 11.1 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 58.3 & 97.2 & 99.6 & \multicolumn{1}{c|}{100} & 48.5 & 97.1 & 100 & \multicolumn{1}{c|}{100} & 35.6 & 88.7 & 99.2 & \multicolumn{1}{c|}{99.9} & 32.3 & 87.5 & 98.4 & 100 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 48.4 & 94.5 & 98.6 & \multicolumn{1}{c|}{100} & 41.4 & 94.1 & 100 & \multicolumn{1}{c|}{100} & 29.7 & 81.5 & 98.6 & \multicolumn{1}{c|}{100} & 27.6 & 81.4 & 97.2 & 100
\\
\hline
& \multicolumn{18}{c}{(2) $t_{6}$-distribution} \\ \cline{2-19}
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 12.2 & 13.5 & 15.0 & \multicolumn{1}{c|}{18.3} & 11.8 & 14.8 & 14.9 & \multicolumn{1}{c|}{16.5} & 10.6 & 12.7 & 14.5 & \multicolumn{1}{c|}{16.3} & 12.1 & 14.2 & 13.2 & 15.8 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 99.8 & 100 & 100 & \multicolumn{1}{c|}{100} & 99.6 & 100 & 100 & \multicolumn{1}{c|}{100} & 96.0 & 100 & 100 & \multicolumn{1}{c|}{100} & 97.2 & 100 & 100 & 100 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 99.8 & 100 & 100 & \multicolumn{1}{c|}{100} & 98.9 & 100 & 100 & \multicolumn{1}{c|}{100} & 93.9 & 100 & 100 & \multicolumn{1}{c|}{100} & 93.9 & 99.9 & 100 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 7.3 & 9.3 & 8.9 & \multicolumn{1}{c|}{11.9} & 8.3 & 11.0 & 11.3 & \multicolumn{1}{c|}{11.7} & 7.2 & 9.1 & 8.4 & \multicolumn{1}{c|}{11.7} & 7.7 & 9.9 & 11.3 & 10.6 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 65.3 & 98.2 & 100 & \multicolumn{1}{c|}{100} & 60.9 & 96.0 & 100 & \multicolumn{1}{c|}{100} & 35.8 & 89.1 & 99.3 & \multicolumn{1}{c|}{100} & 33.5 & 88.8 & 99.0 & 100 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 56.2 & 95.9 & 99.9 & \multicolumn{1}{c|}{100} & 51.5 & 92.8 & 99.8 & \multicolumn{1}{c|}{100} & 29.3 & 80.9 & 98.0 & \multicolumn{1}{c|}{100} & 27.7 & 83.0 & 97.7 & 99.8\\
\hline
& \multicolumn{18}{c}{(3) $\chi_{5}^2$-distribution} \\ \cline{2-19}
& 100 & \multicolumn{1}{c|}{$S_{N}$} & 13.6 & 14.6 & 15.5 & \multicolumn{1}{c|}{18.0} & 12.9 & 12.6 & 14.4 & \multicolumn{1}{c|}{17.5} & 12.6 & 13.9 & 14.2 & \multicolumn{1}{c|}{15.9} & 12.0 & 12.4 & 14.2 & 16.2 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 99.9 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 97.0 & 100 & 100 & \multicolumn{1}{c|}{100} & 97.7 & 100 & 100 & 100 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 99.7 & 100 & 100 & \multicolumn{1}{c|}{100} & 100 & 100 & 100 & \multicolumn{1}{c|}{100} & 93.9 & 100 & 100 & \multicolumn{1}{c|}{100} & 94.9 & 100 & 100 & 100 \\ \hline
& 200 & \multicolumn{1}{c|}{$S_{N}$} & 10.3 & 10.3 & 11.6 & \multicolumn{1}{c|}{13.2} & 9.4 & 10.1 & 12.1 & \multicolumn{1}{c|}{11.6} & 10.2 & 10.2 & 9.7 & \multicolumn{1}{c|}{12.7} & 9.8 & 9.8 & 11.9 & 11.1 \\
& & \multicolumn{1}{c|}{$L_{N}$} & 56.3 & 98.1 & 99.9 & \multicolumn{1}{c|}{100} & 59.1 & 97.9 & 100 & \multicolumn{1}{c|}{100} & 33.9 & 87.4 & 99.4 & \multicolumn{1}{c|}{100} & 35.4 & 85.5 & 99.1 & 100 \\
& & \multicolumn{1}{c|}{$T_{C}$} & 48.9 & 96.6 & 99.8 & \multicolumn{1}{c|}{100} & 50.9 & 95.4 & 99.9 & \multicolumn{1}{c|}{100} & 28.0 & 82.7 & 98.5 & \multicolumn{1}{c|}{100} & 31.4 & 79.9 & 97.7 & 99.9\\
\hline \hline
\end{tabular}
}
\end{table}
Figure \ref{fig1} shows how the empirical powers of the
three tests change as the degree of sparsity of the correlation matrix
changes, where
the $x$-axis is the level of density $k$ and the $y$-axis represents the empirical power.
To generate Figure \ref{fig1}, we designed the following simulation.
We set $N=100$, $T=300$, $p=3,$ $k=2, \cdots, 16$; a subset $S \subset\{1, \cdots, N\}$ is
randomly selected with cardinality $|S|=k$; $(\mathbf{\Psi})_{i j}=1$ if $i=j$; for
$i \neq j$, $(\mathbf{\Psi})_{i j}=0$ if $i \notin S$ or $j \notin S$, and $(\mathbf{\Psi})_{i j} \mathop{\sim}\limits^{iid} U[\sqrt{7 /k\log N / T}, \sqrt{9/k\log N / T}]$ if $i
\in S$ and $j \in S$. Hence, a larger $k$ means a higher level of density.
Figure \ref{fig1} indicates that the empirical power of Fisher's
combined probability test $T_{C}$ is always very close
to the maximum power of both tests for all $k$, which suggests that it
has robust empirical power performance regardless
the alternative is sparse or not. In contrast, the empirical power
curves of $S_N, L_N$ are both monotone, with the sum
based test $S_{N}$ gains more power
with the increase of the level of density, while the max based test
$L_{N}$ gains more power
with the decrease of the level of density.
\begin{figure}[http]
\centering
\includegraphics[width=6in]{figure.pdf}
\caption{The empirical power curves of the three tests at 5\% level under settings
(\ref{H01}) and (\ref{H02}).}\label{fig1}
\end{figure}
\section{Empirical application}\label{sec:data}
We now apply the proposed tests to analyze the securities in the S\&P 500 index. To account for the changes to the composition
of the index over time, we compiled weekly returns on all the securities that constitute the S\&P 500 index that have been listed over the
period from January 2005 to September 2018. Because the
securities that make up the index change over time,
we only consider $N=424$ securities that were included in the S\&P
500 index during the entire period. A total of $T=716$ consecutive
observations were obtained. The time series data on the safe rate of return, and the market
factors are obtained from Ken French's data library web page. The
one-week US treasury bill rate is chosen as the risk-free rate
($r_{ft}$). The value-weighted return on all NYSE, AMEX, and NASDAQ
stocks from CRSP is used as a proxy for the market return ($r_{mt}$).
The average return on the three small portfolios minus the average
return on the three big portfolios ($SMB_t$), and the average return
on two value portfolios minus the average return on two growth
portfolios ($HML_t$) are calculated based on the stocks listed on
the NYSE, AMEX and NASDAQ.
We use the Fama-French three-factor model proposed by \cite{Fama1993Common} to describe the above panel data, which is given by
\begin{align}\label{ff3}Y_{it}=r_{it}-r_{ft}=\alpha_i+\beta_{i1}
(r_{mt}-r_{ft})+\beta_{i2}SMB_t+\beta_{i3} HML_t+\epsilon_{it}, \end{align}
for each $1\leq i\leq N $ and $1\leq t\leq T$, where $r_{mt}-r_{ft}$ is referred
to as the market factor. Since cross-sectional dependence among the errors may lead to misspecified inference \citep{bernard1987cross},
we are interested in testing
\begin{align}\label{nulld}
H_{0}: \epsilon_{1\cdot}, \epsilon_{2\cdot}, \cdots, \epsilon_{N\cdot} \text { are independent random vectors.}
\end{align}
Before applying the proposed tests to the above panel data under the
Fama-French three-factor model, we need
to investigate whether there exists serial correlation in the residuals under such model.
To this end, we applied the Box-Pierce test, a traditional test for
autocorrelation, to the residual sequence of each security under the
Fama-French three-factor model. Figure \ref{Figp} is the histogram of
the $p$-values of the residual sequences, which suggests that for most
securities the Box-Pierce tests are rejected. Furthermore, we applied
a high-dimensional white noise test proposed by \cite{lz2019} to the
sequences of the residual vector and the resulting $p$-value is 0,
which also suggests rejecting. These results indicate that there
exists serial correlation in the residuals under the Fama-French
three-factor model. This may be because the model fails to take into
account all the useful factors with serial correlation. See
\cite{schwartz1977evidence} and \cite{rosenberg1982factor} for further
discussions on causes of serial correlation in the residuals. Hence,
it is reasonable and necessary for us to use a test that allows serial
correlation for the above panel data under the Fama-French
three-factor model.
\begin{figure}[ht]
\centering
\includegraphics[width=4in]{histp-eps-converted-to.pdf}
\caption{Histogram of the p-values of the Box-Pierce test for all residual sequences under the Fama-French three-factor model.}
\label{Figp}
\end{figure}
\begin{figure}[ht]
\centering
\includegraphics[width=4in]{histcor-eps-converted-to.pdf}
\caption{Histogram of the correlations between all pairwise residual sequences of the securities under the Fama-French three-factor model.}
\label{Figcor}
\end{figure}
To accomodate the residual serial correlation, we used
$S_N, L_N$ and $T_{C}$ to test the null hypothesis in (\ref{nulld}),
and obtained
the $p$-values $0$, $0.93$ and $0$
respectively. This seems to suggest a dense residual correlation
matrix with weak signals overall. To verify this,
we plotted the histogram of the correlations between all pairwise
residual sequences of the securities in Figure \ref{Figcor}, which
indicates that there are extensive non-zero correlations with small
absolute values. This explains why the sum based test $S_{N}$
and Fisher's combined probability test $T_{C}$ are able to detect the
deviation from $H_0$. In conclusion,
because both $S_N$ and $T_C$
reject the null hypothesis, we reject the null
hypothesis, and have the knowledge that the error correlation is
nondiagonal with off diagonal values dense with small magnitude.
\section{Conclusion}\label{sec:discuss}
In this paper, for testing cross-sectional independence under
heterogeneous panel data models with serial correlation, we proposed a
max based test based on the maximum of the squared cross-sectional
correlations of residuals to deal with sparse
alternatives. Furthermore, we proposed using Fisher's combined
probability test by combining the p-values of the proposed max based
test and the sum based test, which is applicable regardless
alternatives are sparse or not. In constructing the combined test, the
asymptotic null distribution is established strictly based on the
asymptotic independence of the max based and sum based test
statistics, which is an important contribution of this paper. In
addition, we relaxed the distribution assumption made in the existing
studies for testing independence with serial correlation from Gaussian
to sub-Gaussian. Finally, simulation studies and an empirical application
demonstrate the superiority of the proposed tests in comparison with
some of their competitors.
\newpage
\setcounter{table}{0}\renewcommand{\thetable}{S.\arabic{table}}
\setcounter{figure}{0}\renewcommand{\thefigure}{S.\arabic{figure}}
\setcounter{section}{0}\renewcommand{\thesection}{S.\arabic{section}}
%% Here are the title, author names and addresses
{\Large{Supplementary Material for "Fisher's combined probability test for cross-sectional independence in panel data models with serial correlation"}}
Supplementary Material is organized as follows.
In Section \ref{theorem in section 2}, we provide the proofs of all the theoretical results in Section \ref{sec:statistic}. In Section \ref{proof of lemmas}, we provide the proofs of some lemmas used in Section \ref{theorem in section 2}.
\textsc{Notations.}
For any square matrix $\mathbf{A},$
$(\mathbf{A})_{ij}$ denotes the $(i,j)$-th entry of $\mathbf{A}$,
$\tr(\mathbf{A})$ denotes the trace of $\mathbf{A}$,
$\|\mathbf{A}\|_{\mathrm{F}}$ denotes the Frobenius norm of matrix $\mathbf{A},$ $\|\mathbf{A}\|$ denotes the operator norm of $\mathbf{A}$,
$\lambda_{max}(\mathbf{A})$ denotes the largest eigenvalue of $\mathbf{A}.$
Let $\mathbf{A}$ and $\mathbf{B}$ be the two matrices, we define
$\mathbf{A}\otimes\mathbf{B}$ as the Kronecker product of $\mathbf{A}$ and $\mathbf{B};$
when $\mathbf{A}$ and $\mathbf{B}$ are two square matrices of the same order, define
$\mathbf{A}\circ\mathbf{B}$ as the Hadamard product of $\mathbf{A}$ and $\mathbf{B}$, that is, $(\mathbf{A}\circ\mathbf{B})_{ij}=(\mathbf{A})_{ij}(\mathbf{B})_{ij}.$
For $1\leq i \leq T,$ let $e_{i}$ be a $T$-dimensional vector, where the $i$-th element is 1 and the rest are 0.
Let ${\tau}_{T}=(1,1, \ldots, 1)^{\prime}\in \mathbb{R}^{T}.$ The notation $\mathbf{I}_{T}$ denotes the $T\times T$ identity matrix.
For any two real
numbers $x$ and $y$, let $ x \vee y=\max (x, y)$ and $x \wedge y=\min (x, y).$
For any vector $v\in\mathbb{R}^{T},$
$\|v\|$ denotes the Euclidean norm of $v;$
%$v$ follows multivariate distribution with mean vector $\bf{\mu}$
%and covariance matrix $\mathbf{\Sigma},$
%where we omit the subscript $T$ for $\mu$ and $\mathbf{\Sigma}$ throughout entire supplement for convenience.
Throughout the paper, $C$, $C^{\prime},c,c^{\prime},c_{1},c_{2},\cdots$, denote positive absolute constants.
\section{Proofs of the theorems}\label{theorem in section 2}
Note that we assume that the diagonal elements of $\mathbf{U}$ are all 1, i.e. $\mathbf{U}_{ii}=1,$ for all $1\leq i\leq N.$
Then, under $H_{0},$ we can obtain that $\mathbf{U}=\mathbf{I}_{N}.$
Let $m_{k}=\mE\left(z_{it}^{k}\right),$
where $z_{it}$ is the element at row $i$ and column $t$ of matrix $\mathbf{Z}.$
In order to analyze the moments of the quadratic form, we define some notations:
\begin{align*}
&m_{1}=0, \quad m_{2}=1, \quad m_{3}=\gamma_{1}, \quad m_{4}=\gamma_{2}+3, \quad m_{5}=\gamma_{3}+10 \gamma_{1}, \quad m_{6}=\gamma_{4}+15 \gamma_{2}+10 \gamma_{1}^{2}+15,\\
&m_{7}=\gamma_{5}+21 \gamma_{3}+35 \gamma_{2} \gamma_{1}+105 \gamma_{1}, \quad m_{8}=\gamma_{6}+28 \gamma_{4}+56 \gamma_{3} \gamma_{1}+35 \gamma_{2}^{2}+210 \gamma_{2}+280 \gamma_{1}^{2}+105.
\end{align*}
Note that by Assumption \ref{assum:E distribution}, we can conclude that $\gamma_{1}, \gamma_{2}, \gamma_{3}, \gamma_{4},\gamma_{5},\gamma_{6}$ are finite.
To facilitate the theoretical derivations, we recall some useful lemmas as follows.
\begin{lemma}\label{lem:laplace approxi}
\cite{lieberman1994,pesaran2012testing} Let $x=(x_{1}\cdots,x_{T})^{\prime}$ be a $T$-dimensional random vector. Let $\mathbf{A}$ and $\mathbf{B}$ be $T \times T$
nonstochastic matrices, where $\mathbf{A}$ is symmetric, $\mathbf{B}$ is semi-positive definite, and
suppose that $x_{1}\cdots,x_{T}$
are independently and identically distributed (iid) observations with zero mean and one variance.
Denote the $l$-th cumulant of $x^{\prime}\mathbf{B}x$ by $\kappa_{l}$ and the
$m+1$ order, $m+k$ degree generalised cumulant
of the product of $(x^{\prime}\mathbf{A}x)^k$ and $x^{\prime}\mathbf{B}x$ by $\kappa_{km}$. If the following three conditions are true:
(i) For $l=1,2, \ldots,$ we have $\kappa_{l}=O(T)$;
(ii) for $k=1,2, \ldots,$ we have $\kappa_{k 0}=\mE\big\{\left(x^{\prime} \mathbf{A} x\right)^{k}\big\}=O\left(T^{k}\right)$;
(iii) for $k, m=1,2, \ldots,$ we have $\kappa_{k m}=O\left(T^{l}\right)$, with $l \leqslant k$;
then, we have
\begin{align}\label{equation:laplace}
\mE\left\{\left(\frac{x^{\prime}\mathbf{A}x}{x^{\prime}\mathbf{B}x}\right)^{k}\right\}=\frac{\mE\big\{
\left(x^{\prime}\mathbf{A}x\right)^k\big\}}{\left\{\mE(x^{\prime}\mathbf{B}x)\right\}^{k}}+b_{k 1}+O\left(T^{-2}\right),
\end{align}
where
\begin{align*}
b_{k 1}=&\frac{k(k+1)}{2}\left[\frac{\mE\left\{\left(x^{\prime} \mathbf{A} x\right)^{k}\right\} \kappa_{2}}{\left\{\mE\left(x^{\prime} \mathbf{B} x\right)\right\}^{k+2}}\right]-k\left[\frac{\kappa_{k 1}}{\left\{\mE\left(x^{\prime} \mathbf{B} x\right)\right\}^{k+1}}\right]=O(T^{-1}),\\
\kappa_{k 1}=&\mE\left\{\left(x^{\prime} \mathbf{A} x\right)^{k}\left(x^{\prime} \mathbf{B} x\right)\right\}-\mE\left\{\left(x^{\prime} \mathbf{A} x\right)^{k}\right\} \mE\left(x^{\prime} \mathbf{B} x\right).
\end{align*}
\end{lemma}
\begin{remark}\label{laplace}
\cite{lieberman1994} proposed the Laplace approximation of moments of the ratio of quadratic forms, where requires $\mathbf{B}$ to be a positive definite matrix.
\cite{pesaran2012testing} relaxed this condition and allowed $\mathbf{B}$ to be a positive semi-definite matrix.
\end{remark}
\begin{lemma}\label{lem:sigma}
\cite{baltagi2016}
Recall that $\mathbf{M}_{i}=\mathbf{P}_{i}\mathbf{\Sigma } \mathbf{P}_{i}$ for all $1\leq i\leq N.$
For any fixed positive number $k$, we have
(1) $\frac{1}{T} \operatorname{tr}\left(\mathbf{\Sigma}^{k}\right)=O(1);$
(2) $\frac{1}{T} \operatorname{tr}\left(\mathbf{M}_{i}^{k}\right)=O(1);$
(3) $\frac{1}{T} \operatorname{tr}\left(\mathbf{M}_{i_{1}} \mathbf{M}_{i_{2}} \cdots \mathbf{M}_{i_{k}}\right)=O(1),$ for $1\leq i_{1}, i_{2} , \cdots , i_{k}\leq N.$
\end{lemma}
\begin{lemma}\label{lem:matrix inequality}
\cite{abadir2005matrix} Let $\mathbf{A}$ and $\mathbf{B}$ be the same order square matrices.
\begin{enumerate}[(1)]
\item $\tr(\mathbf{A}\circ\mathbf{B})\leq \sqrt{\tr(\mathbf{A}^2)\tr(\mathbf{B}^2)},$ $\tr(\mathbf{A}\circ\mathbf{A})\leq \tr(\mathbf{A}^{2}),$ when $\mathbf{A}$, $\mathbf{ B}$ are symmetric;
\item $\tr(\mathbf{A}\circ\mathbf{B})\leq C\tr(\mathbf{B})$ when $\mathbf{A},$ $\mathbf{B}$ are non-negative definite matrices and $\lambda_{max}(\mathbf{A})\leq C$ for some constant $C\geq 0;$
\item $\mathbf{A} \circ\mathbf{ B}$ is positive semidefinite when $\mathbf{A}$, $\mathbf{ B}$ are positive semidefinite (See Theorem 7.5.3 in \cite{bernstein2009matrix});
\item ${\tau}_{T}^{\prime}\mathbf{A}\circ\mathbf{B}{\tau}_{T}
= \tr(\mathbf{A}\mathbf{B}^{\prime})$ and $\tr^2(\mathbf{A}\mathbf{B}^{\prime})\leq \tr(\mathbf{A}^{\prime}\mathbf{A})\tr(\mathbf{B}^{\prime}\mathbf{B}) $ for any the same order square matrices $\mathbf{A}$ and $\mathbf{B}$ (See Exercise 12.5 and 12.32 in \cite{abadir2005matrix}).
\end{enumerate}
\end{lemma}
\begin{lemma}\label{le:moment of quadratic form}\cite{bao2010expectation} Suppose that $\boldsymbol{\xi}=\left(\xi_{1}, \xi_{2}, \ldots, \xi_{T}\right)^{\prime}$, and $\xi_{1}, \xi_{2}, \ldots, \xi_{T}$ are iid observations with zero mean, one variance, $\gamma_{1}=\mE\left(\xi_{t}^{3}\right), \gamma_{2}=\mE\left(\xi_{t}^{4}\right)-3, \gamma_{3}=\mE\left(\xi_{t}^{5}\right)-10 \gamma_{1}, \gamma_{4}=\mE\left(\xi_{t}^{6}\right)-15 \gamma_{2}-10 \gamma_{1}^{2}-15,
\gamma_{5}=\mE\left(\xi_{t}^{7}\right)-21\gamma_{3}-35\gamma_{2}\gamma_{1}-105\gamma_{1}$ and $\gamma_{6}=\mE\left(\xi_{t}^{8}\right)-28 \gamma_{4}-56 \gamma_{3} \gamma_{1}-35 \gamma_{2}^{2}-210 \gamma_{2}-280 \gamma_{1}^{2}-105,$ for all $t=1,2, \ldots, T.$ Suppose that $\mathbf{A}_{j}$, $j=1,2,3,4$ are $T \times T$ real symmetric matrices, and ${\tau}_{T}=(1,1,\cdots,1)^{\prime}$ is a $T$-dimensional vector. Then
\begin{align*}
&\mE\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{1} \boldsymbol{\xi}\right)=\tr\left(\mathbf{A}_{1}\right),\\
&\mE\left[\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{1} \boldsymbol{\xi}\right)\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{2} \boldsymbol{\xi}\right)\right]=\gamma_{2} \tr\left[\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right)\right]+\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2}\right)+2 \tr\left(\mathbf{A}_{1} \mathbf{A}_{2}\right),\\
&\mE\left[\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{1} \boldsymbol{\xi}\right)\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{2} \boldsymbol{\xi}\right)\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{3} \boldsymbol{\xi}\right)\right]=\gamma_{4} \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{2} \circ \mathbf{A}_{3}\right)+\gamma_{2} \tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2} \circ \mathbf{A}_{3}\right) \\
&+\gamma_{2} \tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{3}\right)+\gamma_{2} \tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right)+4 \gamma_{2} \tr\left[\mathbf{A}_{1} \circ\left(\mathbf{A}_{2} \mathbf{A}_{3}\right)\right] \\
&+4 \gamma_{2} \tr\left[\mathbf{A}_{2} \circ\left(\mathbf{A}_{1} \mathbf{A}_{3}\right)\right]+4 \gamma_{2} \tr\left[\mathbf{A}_{3} \circ\left(\mathbf{A}_{1} \mathbf{A}_{2}\right)\right]+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) {\tau}_{T}\right] \\
&+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) {\tau}_{T}\right]+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) {\tau}_{T}\right] \\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{2} \circ \mathbf{A}_{3}\right) {\tau}_{T}\right]+\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3}\right)+2 \tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2} \mathbf{A}_{3}\right) \\
&+2 \tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{1} \mathbf{A}_{3}\right)+2 \tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{1} \mathbf{A}_{2}\right)+8 \tr\left(\mathbf{A}_{1} \mathbf{A}_{2} \mathbf{A}_{3}\right),\\
&\mE\left[\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{1} \boldsymbol{\xi}\right)\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{2} \boldsymbol{\xi}\right)\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{3} \boldsymbol{\xi}\right)\left(\boldsymbol{\xi}^{\prime} \mathbf{A}_{4} \boldsymbol{\xi}\right)\right]=\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{4}\right) \\
&+2\left[\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3} \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{2} \mathbf{A}_{4}\right)\right.\\
&+\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{4}\right) \tr\left(\mathbf{A}_{2} \mathbf{A}_{3}\right)+\tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{1} \mathbf{A}_{4}\right) \\
&\left.+\tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{4}\right) \tr\left(\mathbf{A}_{1} \mathbf{A}_{3}\right)+\tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{4}\right)\tr\left(\mathbf{A}_{1} \mathbf{A}_{2}\right)\right] \\
&+4\left[\tr\left(\mathbf{A}_{1} \mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3} \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{1} \mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{2} \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{1} \mathbf{A}_{4}\right) \tr\left(\mathbf{A}_{2} \mathbf{A}_{3}\right)\right] \\
&+8\left[\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2} \mathbf{A}_{3} \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{1} \mathbf{A}_{3} \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{1} \mathbf{A}_{2} \mathbf{A}_{4}\right)\right.\\
&\left.+\tr\left(\mathbf{A}_{4}\right) \tr\left(\mathbf{A}_{1} \mathbf{A}_{2} \mathbf{A}_{3}\right)\right]+16\left[\tr\left(\mathbf{A}_{1} \mathbf{A}_{3} \mathbf{A}_{4} \mathbf{A}_{2}\right)+\tr\left(\mathbf{A}_{1} \mathbf{A}_{4} \mathbf{A}_{2} \mathbf{A}_{3}\right)\right. \\
&\left.+\tr\left(\mathbf{A}_{1} \mathbf{A}_{4} \mathbf{A}_{3} \mathbf{A}_{2}\right)\right]
+\gamma_{2} f_{\gamma_{2}}+\gamma_{4} f_{\gamma_{4}}+\gamma_{6} f_{\gamma_{6}}+\gamma_{1}^{2} f_{\gamma_{1}^{2}}+\gamma_{2}^{2} f_{\gamma_{2}^{2}}+\gamma_{1} \gamma_{3} f_{\gamma_{1} \gamma_{3}},
\end{align*}
where expressions for $f_{\gamma_{2}}, f_{\gamma_{4}}, f_{\gamma_{6}}, f_{\gamma_{1}^{2}}, f_{\gamma_{2}^{2}}$ and $f_{\gamma_{1} \gamma_{3}}$ are provided in \cite{bao2010expectation}:
\begin{align*}
f_{\gamma_{2}}=& \tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3} \circ \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{2} \circ\mathbf{A}_{4}\right)\\
&+\tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{4}\right)\tr\left(\mathbf{A}_{2} \circ\mathbf{A}_{3}\right)+\tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{4}\right)\\
&+\tr\left(\mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{4}\right) \tr\left(\mathbf{A}_{1} \circ\mathbf{A}_{3}\right)
+\tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{4}\right) \tr\left(\mathbf{A}_{1}\circ\mathbf{A}_{2}\right) \\
&+2\left[\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right) \tau_{T} \tr\left(\mathbf{A}_{3} \circ \mathbf{A}_{4}\right)+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{3}\right) \tau_{T} \tr\left(\mathbf{A}_{2} \circ \mathbf{A}_{4}\right)\right.\\
&+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{2} \circ \mathbf{A}_{3}\right)
+\tau_{T}^{\prime}\left(\mathbf{A}_{2} \circ \mathbf{A}_{3}\right)\tau_{T} \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{4}\right)
\\
&\left.+\tau_{T}^{\prime}\left(\mathbf{A}_{2} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{3}\right)+\tau_{T}^{\prime}\left(\mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right)\right]\\
&+4\left[\tr\left(\mathbf{A}_{1}\right) \tr\left\{\mathbf{A}_{2} \circ\left(\mathbf{A}_{3} \mathbf{A}_{4}\right)\right\}+\tr\left(\mathbf{A}_{1}\right) \tr\left\{\mathbf{A}_{3} \circ\left(\mathbf{A}_{2} \mathbf{A}_{4}\right)\right\}+\tr\left(\mathbf{A}_{1}\right) \tr\left\{\mathbf{A}_{4} \circ\left(\mathbf{A}_{2} \mathbf{A}_{3}\right)\right\}\right.\\
&+\tr\left(\mathbf{A}_{2}\right) \tr\left\{\mathbf{A}_{1} \circ\left(\mathbf{A}_{3} \mathbf{A}_{4}\right)\right\}+\tr\left(\mathbf{A}_{2}\right) \tr\left\{\mathbf{A}_{3} \circ\left(\mathbf{A}_{1} \mathbf{A}_{4}\right)\right\}+\tr\left(\mathbf{A}_{2}\right) \tr\left\{\mathbf{A}_{4} \circ\left(\mathbf{A}_{1} \mathbf{A}_{3}\right)\right\}\\
&+\tr\left(\mathbf{A}_{3}\right) \tr\left\{\mathbf{A}_{1} \circ\left(\mathbf{A}_{2} \mathbf{A}_{4}\right)\right\}+\tr\left(\mathbf{A}_{3}\right) \tr\left\{\mathbf{A}_{2} \circ\left(\mathbf{A}_{1} \mathbf{A}_{4}\right)\right\}+\tr\left(\mathbf{A}_{3}\right) \tr\left\}\mathbf{A}_{4} \circ\left(\mathbf{A}_{1} \mathbf{A}_{2}\right)\right\}\\
&\left.+\tr\left(\mathbf{A}_{4}\right)\tr\left\{\mathbf{A}_{1} \circ\left(\mathbf{A}_{2} \mathbf{A}_{3}\right)\right\}+\tr\left(\mathbf{A}_{4}\right) \tr\left\{\mathbf{A}_{2} \circ\left(\mathbf{A}_{1} \mathbf{A}_{3}\right)\right\}+\tr\left(\mathbf{A}_{4}\right) \tr\left\{\mathbf{A}_{3} \circ\left(\mathbf{A}_{1} \mathbf{A}_{2}\right)\right\}\right]
\\
&+8\left[\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{2}\mathbf{A}_{3}\mathbf{A}_{4}\right)+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right)\mathbf{A}_{2}\mathbf{A}_{4}\mathbf{A}_{3}\right)+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{3}\mathbf{A}_{2} \mathbf{A}_{4}\right)\right.\\
&+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{1} \mathbf{A}_{3} \mathbf{A}_{4}\right)+\tr\left(\left(I \circ \mathbf{A}_{2}\right) \mathbf{A}_{1}\mathbf{A}_{4} \mathbf{A}_{3}\right)+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{3}\mathbf{A}_{1} \mathbf{A}_{4}\right)\\
&+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{1}\mathbf{A}_{2} \mathbf{A}_{4}\right)+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{1} \mathbf{A}_{4} \mathbf{A}_{2}\right)+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{2}\mathbf{A}_{1} \mathbf{A}_{4}\right)\\
&\left.+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \mathbf{A}_{1}\mathbf{A}_{2} \mathbf{A}_{3}\right)+\tr\left(\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \mathbf{A}_{1} \mathbf{A}_{3} \mathbf{A}_{2}\right)+\tr\left(\left(\mathbf{I}_{T} \circ\mathbf{A}_{4}\right) \mathbf{A}_{2} \mathbf{A}_{1}\mathbf{A}_{3}\right)\right],
\\
&+16\left[\tau_{T}^{\prime}\left\{\mathbf{I}_{T}\circ\left(\mathbf{A}_{1} \mathbf{A}_{2}\right)\right\}\left\{\mathbf{I}_{T} \circ\left(\mathbf{A}_{3} \mathbf{A}_{4}\right)\right\} \tau_{T}+\tau_{T}^{\prime}\left\{\mathbf{I}_{T} \circ\left(\mathbf{A}_{1} \mathbf{A}_{3}\right)\right\}\left\{\mathbf{I}_{T} \circ\left(\mathbf{A}_{2} \mathbf{A}_{4}\right)\right\} \tau_{T}\right. \\
&\left.+\tau_{T}^{\prime}\left\{\mathbf{I}_{T} \circ\left(\mathbf{A}_{1} \mathbf{A}_{4}\right)\right\}\left\{\mathbf{I}_{T} \circ\left(\mathbf{A}_{2} \mathbf{A}_{3}\right)\right\} \tau_{T}\right],\\
f_{\gamma_{4}}=& \tr\left(\mathbf{A}_{1}\right) \tr\left(\mathbf{A}_{2} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{2}\right)\tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{2} \circ \mathbf{A}_{4}\right)\\
&+\tr\left(\mathbf{A}_{4}\right) \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{2} \circ \mathbf{A}_{3}\right)+4\left[\tr\left\{\mathbf{A}_{1} \circ \mathbf{A}_{2} \circ\left(\mathbf{A}_{3} \mathbf{A}_{4}\right)\right\}+\tr\left\{\mathbf{A}_{1} \circ \mathbf{A}_{3} \circ\left(\mathbf{A}_{2} \mathbf{A}_{4}\right)\right\}\right.\\
&+\tr\left\{\mathbf{A}_{1} \circ \mathbf{A}_{4} \circ\left(\mathbf{A}_{2} \mathbf{A}_{3}\right)\right\}+\tr\left\{\mathbf{A}_{2} \circ\mathbf{A}_{3} \circ\left(\mathbf{A}_{1} \mathbf{A}_{4}\right)\right\}+\tr\left\{\mathbf{A}_{2} \circ \mathbf{A}_{4} \circ\left(\mathbf{A}_{1} \mathbf{A}_{3}\right)\right\}\\
&\left.+\tr\left\{\mathbf{A}_{3} \circ\mathbf{A}_{4} \circ\left(\mathbf{A}_{1} \mathbf{A}_{2}\right)\right\}\right],\\
f_{\gamma_{6}}=&\tr\left(\mathbf{A}_{1}\circ\mathbf{A}_{2}\circ
\mathbf{A}_{3}\circ\mathbf{A}_{4}\right),\\
f_{\gamma_{1}^{2}}=& 2\left[\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{1}\right)+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \tau_{T} \tr\left(\mathbf{A}_{1}\right)\right.\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{1}\right)
+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{2}\right)\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \tau_{T} \tr\left(\mathbf{A}_{2}\right)+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{2}\right) \\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{3}\right)+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3}\right)\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{3}\right) +\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T}\circ \mathbf{A}_{3}\right) \tau_{T} \tr\left(\mathbf{A}_{4}\right)
\\
&\left.+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{4}\right)+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ\mathbf{A}_{2}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \tau_{T} \right]\\
&+4\left[\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{2} \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{2} \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \tau_{T}\right.\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{3} \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \tau_{T} +\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{1} \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T}\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{1} \mathbf{A}_{4}\left(\mathbf{I}_{T} \mathbf{A}_{3}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{3} \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \tau_{T} \\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{1} \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{1} \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \tau_{T}\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{2} \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \mathbf{A}_{1} \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \tau_{T}\\
&\left.+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \mathbf{A}_{1} \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \mathbf{A}_{2} \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \tau_{T}\right]\\
&+4\left[\tau_{T}^{\prime}\left(\mathbf{A}_{2} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{1}\right)+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{2}\right)
\right.\\
&\left.+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ\mathbf{A}_{2}\circ \mathbf{A}_{4}\right) \tau_{T} \tr\left(\mathbf{A}_{3}\right)+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{2} \circ\mathbf{A}_{3}\right) \tau_{T} \tr\left(\mathbf{A}_{4}\right)\right]\\
&+8\left[\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \tau_{T}\right.\\
&+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{3}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{3}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \tau_{T}\\
&+\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{4}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \tau_{T}+
\tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{4}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \tau_{T}\\
&+\tau_{T}^{\prime}\left(\mathbf{A}_{2} \circ \mathbf{A}_{3}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{A}_{2} \circ \mathbf{A}_{3}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \tau_{T}\\
&+\tau_{T}^{\prime}\left(\mathbf{A}_{2} \circ \mathbf{A}_{4}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{A}_{2} \circ \mathbf{A}_{4}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \tau_{T}\\
&\left.+\tau_{T}^{\prime}\left(\mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \tau_{T}\right]\\
&+16\left[\tr\left(\mathbf{A}_{1}\left(\mathbf{A}_{2} \circ \mathbf{A}_{3}\right) \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{1}\left(\mathbf{A}_{2} \circ \mathbf{A}_{4}\right) \mathbf{A}_{3}\right)+\tr\left(\mathbf{A}_{1}\left(\mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \mathbf{A}_{2}\right)\right. \\
&\left.+\tr\left(\mathbf{A}_{2}\left(\mathbf{A}_{1} \circ \mathbf{A}_{3}\right) \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{2}\left(\mathbf{A}_{1} \circ \mathbf{A}_{4}\right) \mathbf{A}_{3}\right)+\tr\left(\mathbf{A}_{3}\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right) \mathbf{A}_{4}\right)\right],\\
f_{\gamma_{2}^{2}}=& \tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right) \tr\left(\mathbf{A}_{3} \circ \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{3}\right) \tr\left(\mathbf{A}_{2} \circ \mathbf{A}_{4}\right)+\tr\left(\mathbf{A}_{1} \circ \mathbf{A}_{4}\right) \tr\left(\mathbf{A}_{2} \circ \mathbf{A}_{3}\right) \\
&+4\left[\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right)\left(\mathbf{A}_{2} \circ \mathbf{A}_{3}\right)\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right)\tau_{T}+
\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right)\left(\mathbf{A}_{2} \circ \mathbf{A}_{4}\right)\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right)\tau_{T}\right.\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right)\left(\mathbf{A}_{3} \circ \mathbf{A}_{4}\right)\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right)\tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right)\left(\mathbf{A}_{1} \circ \mathbf{A}_{3}\right)\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right)\tau_{T}\\
&\left.+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right)\left(\mathbf{A}_{1} \circ \mathbf{A}_{4}\right)\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right)\tau_{T}+
\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right)\left(\mathbf{A}_{1} \circ \mathbf{A}_{2}\right)\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right)\tau_{T}\right] \\
&+8 \tau_{T}^{\prime}\left(\mathbf{A}_{1} \circ \mathbf{A}_{2} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \tau_{T},\\
f_{\gamma_{1} \gamma_{3}} &=2\left[\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right)+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ
\mathbf{A}_{1}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2} \circ \mathbf{A}_{4}\right) \tau_{T}\right.\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ\mathbf{A}_{2}\circ \mathbf{A}_{3}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \tau_{T}\\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1} \circ \mathbf{A}_{4}\right) \tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T} \circ\mathbf{A}_{1} \circ \mathbf{A}_{3}\right) \tau_{T}
\\&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2} \circ \mathbf{A}_{4}\right) \tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1} \circ \mathbf{A}_{4}\right) \tau_{T} \\
&+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right) \mathbf{A}_{4}\left(\mathbf{I}_{T}\circ\mathbf{A}_{1} \circ \mathbf{A}_{2}\right) \tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \mathbf{A}_{1}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2} \circ \mathbf{A}_{3}\right) \tau_{T}\\
&\left.+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \mathbf{A}_{2}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1} \circ \mathbf{A}_{3}\right) \tau_{T}
+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right) \mathbf{A}_{3}\left(\mathbf{I}_{T} \circ\mathbf{A}_{1}\circ \mathbf{A}_{2}\right) \tau_{T}\right]\\
&+8\left[\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{1}\right)\left(\mathbf{A}_{2} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{2}\right)\left(\mathbf{A}_{1} \circ \mathbf{A}_{3} \circ \mathbf{A}_{4}\right) \tau_{T}\right.\\
&\left.+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{3}\right)\left(\mathbf{A}_{1} \circ\mathbf{A}_{2}\circ \mathbf{A}_{4}\right) \tau_{T}+\tau_{T}^{\prime}\left(\mathbf{I}_{T} \circ \mathbf{A}_{4}\right)\left(\mathbf{A}_{1} \circ \mathbf{A}_{2} \circ \mathbf{A}_{3}\right) \tau_{T}\right].
\end{align*}
\end{lemma}
Now, we are ready to present the proofs of the theorems in Section \ref{sec:statistic}.
\subsection{Proof of Theorem \ref{th:max null}}
Let
$$
\tilde{T}_{i j}\doteq\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}}
=\frac{z_{i}^{\prime}\mathbf{\Sigma}z_{j}}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}},\,\,\,\,\,\,\,\,\,
\rho_{ij}=\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}}{\|\epsilon_{i\cdot}\|\|\epsilon_{j\cdot}\|}.
$$
To simplify notation, we define $z_{i}\doteq(\mathbf{Z})_{i\cdot},$ where $(\mathbf{Z})_{i\cdot}$ represents the i-th row vector of $\mathbf{Z},$ for $1\leq i\leq N.$ Hence, we have
$\hat{\epsilon}_{i\cdot}=\mathbf{P}_{i}\epsilon_{i\cdot}=\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}z_{i}.$
\begin{lemma}\label{14Bi}\cite{feng2020}
Set $\mathbf{B}_{i}=\mathbf{x}_{i}\left(\mathbf{x}_{i}^{\prime} \mathbf{x}_{i}\right)^{-1} \mathbf{x}_{i}^{\prime},$ for $1 \leq i \leq N.$
Then,
$$
\max _{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}-{\rho}_{i j}\right| \leq 14 \cdot\left(\max _{1 \leq i \leq j \leq N} \frac{{\epsilon}_{j\cdot}^{\prime} \mathbf{B}_{i} {\epsilon}_{j\cdot}}{{\epsilon}_{j\cdot}^{\prime}{\epsilon}_{j\cdot}}\right).
$$
\end{lemma}
We now are ready to derive the asymptotic null distribution of $L_{N}.$
First, we will show that for $y\in\mathbb{R}$, under $H_{0}$,
\begin{align}\label{maxrhoijnull}
\mathbb{P}\left(\max _{1 \leq i<j \leq N} \tilde{T}_{i j}^{2}-4\log N+\log \log N\leq y\right) \rightarrow \exp \left\{-\frac{1}{\sqrt{8 \pi}} \exp \left(-\frac{y}{2}\right)\right\}.
\end{align}
By Theorem 1 in \cite{arratia1989}, we have
$$
\left|\mathbb{P}\left(\max _{1 \leq i<j \leq N} \tilde{T}_{i j}^{2} \leq t_{N}\right)-e^{-\tau_{N}}\right| \leq b_{1 N}+b_{2N}+b_{3N}
$$
where
$t_{N}=4 \log N-\log \log N+y, \tau_{N}=\underset{1\leq i<j\leq N}{\sum\sum} \mathbb{P}\left(\tilde{T}_{12}^{2}>t_{N}\right),$
\begin{align*}
&b_{1 N} =\underset{1\leq i<j\leq N}{\sum\sum}\sum_{(k,l)\in B_{ij}}\mathbb{P}\left(\tilde{T}_{ij}^{2}>t_{N}\right)\mathbb{P}\left(\tilde{T}_{kl}^{2}>t_{N}\right) \leq N^{3}\left[\mathbb{P}\left(\tilde{T}_{12}^{2}>t_{N}\right)\right]^{2}, \\
&b_{2 N} =\underset{1\leq i<j\leq N}{\sum\sum}\sum_{(k,l)\in B_{ij}\backslash\{(i,j)\}}\mathbb{P}\left(\tilde{T}_{ij}^{2}>t_{N},\tilde{T}_{kl}^{2}>t_{N}\right)\leq N^{3} \mathbb{P}\left(\tilde{T}_{12}^{2}>t_{N}, \tilde{T}_{13}^{2}>t_{N}\right),\\
&b_{3 N} =\underset{1\leq i<j\leq N}{\sum\sum}\mE\left|\mathbb{P}\left\{\tilde{T}_{ij}^{2}>t_{N}|\sigma(\tilde{T}_{kl}^{2}:(k,l)\notin B_{ij})\right\}-\mathbb{P}\left(\tilde{T}_{12}^{2}>t_{N}\right)\right|=0,
\end{align*}
for all $1\leq i<j\leq N$, $B_{ij}\doteq\{(k,l):1\leq k<l\leq N,\{k,l\}\cap\{i,j\}\neq\varnothing\},$
and the third term $b_{3N}$ on the right side of the inequality is equal to zero because for four different indices $i, j, k, l,$ $\tilde{T}_{i j}$ and $\tilde{T}_{k l}$ are independent.
Note that according to Theorem 1.1 in \cite{rudelson2013}, we have for any large $M> 0$, there exists some $C_{1} > 0$ such
that
$$
\mathbb{P}\left(\frac{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}{\tr(\mathbf{\Sigma}^{2})}>1+\varepsilon_{1}\right)
\leq2N^{-M},
$$
where $\varepsilon_{1}=C_{1}\sqrt{\log N/\tr(\mathbf{\Sigma}^2)}.$
By Corollary 3.1 in \cite{saulis1991limit} and $\tr(\mathbf{\Sigma}^{2})=\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}$, we have
\begin{align}\label{T12}
\mathbb{P}\left(\tilde{T}_{12}^{2}>t_{N}\right)=&
\mathbb{P}\left(\left|\tilde{T}_{12}\right|> \sqrt{t_{N}}\right)\n\\
=&\mathbb{P}\left\{\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{\tr(\mathbf{\Sigma}^{2})}}
>\sqrt{t_{N}}\right\}\n\\
\leq&\mathbb{P}\left\{\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{\tr(\mathbf{\Sigma}^{2})}}
>\sqrt{t_{N}},\frac{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}{\tr(\mathbf{\Sigma}^{2})}\leq1+\varepsilon_{1}\right\}
+\mathbb{P}\left(\frac{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}{\tr(\mathbf{\Sigma}^{2})}>1+\varepsilon_{1}\right)\n\\
\leq&\mathbb{P}\left\{\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}}
>\frac{\sqrt{t_{N}}}{1+\varepsilon_{1}}\right\}+2N^{-M}\n\\
\leq&\mE\left[\mE\left\{I\left(\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}}
>\frac{\sqrt{t_{N}}}{1+\varepsilon_{1}}\right)\Big|z_{2}\right\}\right]+2N^{-M}\n\\
\leq&\mE\left\{\mathbb{P}\left(\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}}
>\frac{\sqrt{t_{N}}}{1+\varepsilon_{1}}\Big|z_{2}\right)\right\}+2N^{-M}\n\\
=&\{1+o(1)\} \frac{2}{\sqrt{2 \pi t_{N}}} \exp^{-t_{N} / 2}+2N^{-M}\n\\
=&O(N^{-2}),
\end{align}
where last equality holds due to Proposition 2.1.2 in \cite{vershynin2018},
$I(\cdot)$ denotes indicative function, and $M$ is sufficiently large. Similarly, we have
\begin{align*}
\mathbb{P}\left(\tilde{T}_{12}^{2}\leq t_{N}\right)=&
\mathbb{P}\left(\left|\tilde{T}_{12}\right| \leq \sqrt{t_{N}}\right)\n\\
=&\mathbb{P}\left\{\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{\tr(\mathbf{\Sigma}^{2})}}
\leq\sqrt{t_{N}}\right\}\n\\
\leq&\mathbb{P}\left\{\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{\tr(\mathbf{\Sigma}^{2})}}
\leq\sqrt{t_{N}},\frac{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}{\tr(\mathbf{\Sigma}^{2})}\geq1-\varepsilon_{1}\right\}
+\mathbb{P}\left(\frac{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}{\tr(\mathbf{\Sigma}^{2})}<1-\varepsilon_{1}\right)\n\\
\leq&\mathbb{P}\left\{\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}}
\leq\frac{\sqrt{t_{N}}}{1-\varepsilon_{1}}\right\}+2N^{-M},
\end{align*}
then $1-\mathbb{P}\left(\tilde{T}_{12}^{2}<t_{N}\right)\geq
1-\mathbb{P}\left\{\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}}
<\frac{\sqrt{t_{N}}}{1-\varepsilon_{1}}\right\}-2N^{-M}.$
Hence, we have
\begin{align*}
&\mathbb{P}\left(\tilde{T}_{12}^{2}>t_{N}\right)\n\\
\geq&\mathbb{P}\left\{\frac{|z_{1}^{\prime}\mathbf{\Sigma}z_{2}|}{\sqrt{z_{2}^{\prime}\mathbf{\Sigma}^2z_{2}}}
>\frac{\sqrt{t_{N}}}{1-\varepsilon_{1}}\right\}-2N^{-M}\\
=&\{1-o(1)\} \frac{2}{\sqrt{2 \pi t_{N}}} \exp^{-t_{N} / 2}-2N^{-M}\\
=&O(N^{-2}).
\end{align*}
This shows that $\tau_{N} \sim \frac{1}{\sqrt{8 \pi}} e^{-y / 2}$ and $b_{1 N} \leq C N^{-1}$. For $b_{2 N}$, we have
$$
\mathbb{P}\left(\tilde{T}_{12}^{2}>t_{N}, \tilde{T}_{13}^{2}>t_{N}\right) \leq \mathbb{P}\left(\left|\tilde{T}_{12}-\tilde{T}_{13}\right| \geq 2 \sqrt{t_{N}}\right)+\mathbb{P}\left(\left|\tilde{T}_{12}+\tilde{T}_{13}\right| \geq 2 \sqrt{t_{N}}\right).
$$
Again, by Theorem 1.1 in \cite{rudelson2013}, for any large $M> 0$ , there exists some $C_{2} > 0$ such
that
$$
\mathbb{P}\left(\frac{(z_{2}-z_{3})^{\prime}\mathbf{\Sigma}^2(z_{2}-z_{3})}{\tr(\mathbf{\Sigma}^{2})}>2(1+\varepsilon_{2})\right)
\leq2N^{-M},
$$
where $\varepsilon_{2}=C_{2}\sqrt{\log N/\tr(\mathbf{\Sigma}^2)}.$
By Corollary 3.1 (Cram\'er type moderate deviation results) in \cite{saulis1991limit}, we have
\begin{align*}
&\mathbb{P}\left(\left|\tilde{T}_{12}-\tilde{T}_{13}\right| \geq 2 \sqrt{t_{N}}\right)\\
=&\mathbb{P}\left\{\frac{\left|z_{1}^{\prime}\mathbf{\Sigma}\left(z_{2}-z_{3}\right)\right|}
{\sqrt{\tr(\mathbf{\Sigma}^2)}} \geq 2 \sqrt{t_{N}}\right\}\\
\leq&\mathbb{P}\left\{\frac{\left|z_{1}^{\prime}\mathbf{\Sigma}\left(z_{2}-z_{3}\right)\right|}
{\sqrt{\tr(\mathbf{\Sigma}^2)}} \geq 2 \sqrt{t_{N}},\frac{(z_{2}-z_{3})^{\prime}\mathbf{\Sigma}^2(z_{2}-z_{3})}{\tr(\mathbf{\Sigma}^2)}<2(1+\varepsilon_{2})\right\}\\
&+
\mathbb{P}\left\{\frac{(z_{2}-z_{3})^{\prime}\mathbf{\Sigma}^2(z_{2}-z_{3})}{\tr(\mathbf{\Sigma}^2)}>2(1+\varepsilon_{2})\right\}\\
\leq&\mathbb{P}\left\{\frac{\left|z_{1}^{\prime}\mathbf{\Sigma}\left(z_{2}-z_{3}\right)\right|}
{\sqrt{(z_{2}-z_{3})^{\prime}\mathbf{\Sigma}^2(z_{2}-z_{3})}} \geq \frac{ \sqrt{2t_{N}}}{1+\varepsilon_{2}}\right\}
+2N^{-M}\\
\leq&\{1+o(1)\} \frac{\sqrt{\log N}}{2 \sqrt{\pi}} N^{-4} e^{-y} ,
\end{align*}
where $M$ is sufficiently large.
Similarly, $\mathbb{P}\left(\left|\tilde{T}_{12}+\tilde{T}_{13}\right| \geq 2 \sqrt{t_{n}}\right)=\{1+o(1)\} \frac{\sqrt{\log N}}{2 \sqrt{\pi}} N^{-4} e^{-y}$. Combining these inequalities, we have $b_{2 N} \leq C N^{-1} \sqrt{\log N}$ and
\begin{align}\label{eiej}
\mathbb{P}\left(\max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}-4\log N+\log \log N\leq y\right) \rightarrow G(y)=\exp \left\{-\frac{1}{\sqrt{8 \pi}} \exp \left(-\frac{y}{2}\right)\right\}.
\end{align}
Similarly, due to Theorem 1.1 in \cite{rudelson2013}, we have
\begin{align}\label{ineq-max}
\mathbb{P}\left(\max _{1 \leq i \leq N}\left|\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}-\tr\left(\mathbf{\Sigma}\right) \right| \geq C^{\prime} \sqrt{T\log N}\right)=O\left(N^{-M}\right),
\end{align}
where $M>0$ is sufficiently large, $C^{\prime}$ is a constant that depends on $M.$
Due to (\ref{eiej}), we have
\begin{align}\label{dis-rho}
\mathbb{P}\left(\frac{\tr^2(\mathbf{\Sigma})}{T^2\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\frac{\tr^2(\mathbf{\Sigma})}{T^2}}-4\log N+\log \log N\leq y\right) \rightarrow \exp \left\{-\frac{1}{\sqrt{8 \pi}} \exp \left(-\frac{y}{2}\right)\right\}.
\end{align}
Set
$$\Omega=\left\{\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\tr\left(\mathbf{\Sigma}\right)}{T^2}
+\frac{\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}\tr\left(\mathbf{\Sigma}\right)}{T^2}
-\frac{2\tr^2(\mathbf{\Sigma})}{T^2}\right|\leq2C_{1}\sqrt{\frac{\log N}{T}}\right\}.$$
Obviously, due to (\ref{ineq-max}),
we have \begin{align*}
\mathbb{P}\left(\Omega^{c}\right)\leq&\mathbb{P}\left\{\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}}{T}
+\frac{\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T}
-\frac{2\tr\left(\mathbf{\Sigma}\right)}{T}\right|>\frac{2C_{1}T}{\tr\left(\mathbf{\Sigma}\right)}\sqrt{\frac{\log N}{T}}\right\}\\
\leq&\mathbb{P}\left\{\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}}{T}
+\frac{\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T}
-\frac{2\tr\left(\mathbf{\Sigma}\right)}{T}\right|>\frac{2C_{1}}{\lambda_{max}(\mathbf{\Sigma})}\sqrt{\frac{\log N}{T}}\right\}\\
\leq&\mathbb{P}\left\{\max_{1\leq i\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}}{T}-\frac{\tr\left(\mathbf{\Sigma}\right)}{T}\right|
>\frac{C_{1}}{\lambda_{max}(\mathbf{\Sigma})}\sqrt{\frac{\log N}{T}}\right\}\\
&+
\mathbb{P}\left\{\max_{1\leq j\leq N}\left|\frac{\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T}-\frac{\tr\left(\mathbf{\Sigma}\right)}{T}\right|
>\frac{C_{1}}{\lambda_{max}(\mathbf{\Sigma})}\sqrt{\frac{\log N}{T}}\right\}=O\left(N^{-M}\right).
\end{align*}
We claim that for some constant $C_{3}>0$ such that
$$\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}-
\frac{\tr^2(\mathbf{\Sigma})}{T^2}\right|>C_{3}\sqrt{\frac{\log N}{T}}\right)\rightarrow 0$$
as $N,T$ are sufficiently large.
Notice that
\begin{align*}
&\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}-
\frac{\tr^2(\mathbf{\Sigma})}{T^2}\right|>C_{3}\sqrt{\frac{\log N}{T}}\right)\\
\leq&
\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}-
\frac{\tr^2(\mathbf{\Sigma})}{T^2}\right|>C_{3}\sqrt{\frac{\log N}{T}},\Omega\right)+\mathbb{P}\left(\Omega^{c}\right),
\end{align*}
we just need to prove that
$$
\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}-
\frac{\tr^2(\mathbf{\Sigma})}{T^2}\right|>C_{3}\sqrt{\frac{\log N}{T}},\Omega\right)\rightarrow 0.$$
In fact, due to (\ref{ineq-max}), for sufficiently large $C_{3}$ satisfying
$C_{3}-2C_{1}>C_{3}\sqrt{\frac{1}{2}}
$ and $\sqrt{\frac{C_{3}}{2}}>C_{1},$
we have
\begin{align*}
&\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}-
\frac{\tr^2(\mathbf{\Sigma})}{T^2}\right|>C_{3}\sqrt{\frac{\log N}{T}},\Omega\right)\\
\leq&\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}
-\tr^2(\mathbf{\Sigma})-\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\tr\left(\mathbf{\Sigma}\right)
-\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}\tr\left(\mathbf{\Sigma}\right)+2\tr^2(\mathbf{\Sigma})}{T^2}\right|
>C_{3}\sqrt{\frac{\log N}{2T}}\right)\\
\leq&\mathbb{P}\left\{\max_{1\leq i<j\leq N}\left|\left(\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}}{T}-
\frac{\tr\left(\mathbf{\Sigma}\right)}{T}\right)\left(\frac{\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T}-
\frac{\tr\left(\mathbf{\Sigma}\right)}{T}\right)\right|>C_{3}\sqrt{\frac{\log N}{2T}}\right\}\\
\leq&\mathbb{P}\left\{\max_{1\leq i\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}}{T}-
\frac{\tr\left(\mathbf{\Sigma}\right)}{T}\right|>\sqrt{C_{3}}\left(\frac{\log N}{2T}\right)^{1/4}\right\}\\
&+\mathbb{P}\left\{\max_{1\leq j\leq N}\left|\frac{\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T}-
\frac{\tr\left(\mathbf{\Sigma}\right)}{T}\right|>\sqrt{C_{3}}\left(\frac{\log N}{2T}\right)^{1/4}\right\}\\
\leq&\mathbb{P}\left\{\max_{1\leq i\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}}{T}-
\frac{\tr\left(\mathbf{\Sigma}\right)}{T}\right|>\sqrt{\frac{C_{3}\log N}{2T}}\right\}+\mathbb{P}\left\{\max_{1\leq j\leq N}\left|\frac{\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T}-
\frac{\tr\left(\mathbf{\Sigma}\right)}{T}\right|>\sqrt{\frac{C_{3}\log N}{2T}}\right\}\\
=&O\left(N^{-M}\right),
\end{align*}
where the last inequality holds due to (\ref{ineq-max}).
Then, we can conclude that
$$\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}-
\frac{\tr^2(\mathbf{\Sigma})}{T^2}\right|>C_{3}\sqrt{\frac{\log N}{T}}\right)
=O\left(N^{-M}\right)
$$
and
$$\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}/T^2}{\tr^2(\mathbf{\Sigma})/T^2}-
1\right|<\frac{C_{3}}{\lambda_{min}^2(\mathbf{\Sigma})}\sqrt{\frac{\log N}{T}}\right)\rightarrow 1.
$$
Thus, with probability tending to one, we have
$$
\begin{aligned}
\max _{1 \leq i <j\leq N}\left|\frac{\tr^2(\mathbf{\Sigma})/T^2}{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}/T^2}-1\right| & \leq \max _{1 \leq i <j\leq N} \frac{T^2}{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}} \max _{1 \leq i <j \leq N}\left|\frac{\tr^2(\mathbf{\Sigma})}{T^2}-\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}\right| \\
& \leq 2 \max _{1 \leq i <j\leq N} \frac{T^2}{\tr^2(\mathbf{\Sigma})} \max _{1 \leq i <j\leq N}\left|\frac{\tr^2(\mathbf{\Sigma})}{T^2}-\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}\right|
=O_{p}\left\{\sqrt{\frac{\log N}{T}}\right\}.
\end{aligned}
$$
This, together with (\ref{dis-rho}), we have
\begin{align}\label{dis-rho/}
&\mathbb{P}\left(\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}
{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}
}-4 \log N+\log \log N\leq y\right)\n\\
=&\mathbb{P}\left(\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\max _{1 \leq i<j \leq N} \rho^2_{ij}-4 \log N+\log \log N\leq y\right)
\rightarrow \exp \left\{-\frac{1}{\sqrt{8 \pi}} \exp \left(-\frac{y}{2}\right)\right\},
\end{align}
for any $y\in \mathbb{R}.$
Then, we want to prove that
$$\mathbb{P}\left(\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\max _{1 \leq i<j \leq N} \hat{\rho}^2_{ij}-4 \log N+\log \log N\leq y\right)
\rightarrow \exp \left\{-\frac{1}{\sqrt{8 \pi}} \exp \left(-\frac{y}{2}\right)\right\}.$$
So, we just need to show that for any $\epsilon>0$
$$\mathbb{P}\left(\max_{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}-{\rho}_{i j}\right|>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}{\sqrt{\log N}\tr\left(\mathbf{\Sigma}\right)}\right)\rightarrow 0.$$
This is because if it holds, we have
$$
\sqrt{\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \log N} \cdot\left(\max_{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}\right|-\max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right|\right) \rightarrow 0
$$
in probability. Set $\Delta=\max_{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}\right|-\max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right| .$ Then
\begin{align*}
&\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \max_{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}\right|^{2}\\
=&\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2}\left( \max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right|+\Delta\right)^{2}\\
=&\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right|^{2}+2 \frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right| \Delta+\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \Delta^{2}.
\end{align*}
The Slutsky lemma and (\ref{dis-rho/}) say that $(\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} / \log N)^{1 / 2} \max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right| \rightarrow 2$ in probability. Consequently,
$$
\begin{aligned}
&\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right| \Delta=\left(\frac{\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2}}{\log N}\right)^{1 / 2} \max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right| \cdot\Big(\sqrt{\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \log N} \Delta\Big) \rightarrow 0 \\
&\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \Delta^{2}=\Big[\sqrt{\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \log N} \Delta\Big]^{2} \cdot \frac{1}{\log N} \rightarrow 0
\end{aligned}
$$
in probability. Then, we can obtain that
$$
\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \max_{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}\right|^{2}=\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^2} \max_{1 \leq i<j \leq N}\left|{\rho}_{i j}\right|^{2}+o_{p}(1).
$$
So, we will prove
\begin{align}\label{pijhat-pij}
\mathbb{P}\left(\max_{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}-{\rho}_{i j}\right|>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}{\sqrt{\log N}\tr\left(\mathbf{\Sigma}\right)}\right)\rightarrow 0.
\end{align}
By Lemma \ref{14Bi},
\begin{align*}
&\mathbb{P}\left(\max_{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}-{\rho}_{i j}\right|>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}{\sqrt{\log N}\tr\left(\mathbf{\Sigma}\right)}\right)\\
\leq&N^2\max_{1 \leq i<j \leq N}\mathbb{P}\left( \frac{{\epsilon}_{j\cdot}^{\prime} \mathbf{B}_{i} {\epsilon}_{j\cdot}}{{\epsilon}_{j\cdot}^{\prime}{\epsilon}_{j\cdot}}>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}
{14\sqrt{\log N}\tr\left(\mathbf{\Sigma}\right)}\right)\\
\leq&N^2\max_{1 \leq i \leq N}\mathbb{P}\left( \frac{{\epsilon}_{1\cdot}^{\prime} \mathbf{B}_{i} {\epsilon}_{1\cdot}}{{\epsilon}_{1\cdot}^{\prime}{\epsilon}_{1\cdot}}>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}
{14\sqrt{\log N}\tr\left(\mathbf{\Sigma}\right)}\right)\\
\leq&N^2\max_{1 \leq i \leq N}\left[\mathbb{P}\left( \frac{{\epsilon}_{1\cdot}^{\prime} \mathbf{B}_{i} {\epsilon}_{1\cdot}}{{\epsilon}_{1\cdot}^{\prime}{\epsilon}_{1\cdot}}>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}
{14\sqrt{\log N}\tr\left(\mathbf{\Sigma}\right)},{\epsilon}_{1\cdot}^{\prime}{\epsilon}_{1\cdot}>\frac{\tr(\mathbf{\Sigma})}{2}\right)+
\mathbb{P}\left({\epsilon}_{1\cdot}^{\prime}{\epsilon}_{1\cdot}<\frac{\tr(\mathbf{\Sigma})}{2}\right)\right]\\
\leq&N^2\max_{1 \leq i \leq N}\left[\mathbb{P}\left( {\epsilon}_{1\cdot}^{\prime} \mathbf{B}_{i} {\epsilon}_{1\cdot}>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}
{28\sqrt{\log N}}\right)+\mathbb{P}\left(z_{1}^{\prime}\mathbf{\Sigma}z_{1}<\frac{\tr(\mathbf{\Sigma})}{2}\right)\right]\\
\leq&N^2\max_{1 \leq i \leq N}\left[\mathbb{P}\left( {\epsilon}_{1\cdot}^{\prime} \mathbf{B}_{i} {\epsilon}_{1\cdot}>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}
{28\sqrt{\log N}}\right)+2\exp\left(-\eta_{0} T\right)\right].
\end{align*}
Here, by Theorem 1.1 in \cite{rudelson2013}, there exists a constant $\eta_{0},\,C>0$ such
that
$$\mathbb{P}\left(z_{1}^{\prime}\mathbf{\Sigma}z_{1}<\frac{\tr(\mathbf{\Sigma})}{2}\right)\leq C\exp\left(-\eta_{0} T\right).$$
Note that $\tr(\mathbf{\Sigma}^{1/2}\mathbf{B}_{i}\mathbf{\Sigma}^{1/2})=\tr(\mathbf{B}_{i}\mathbf{\Sigma})\leq p\lambda_{max}(\mathbf{\Sigma})$
and $\tr(\mathbf{\Sigma}^{1/2}\mathbf{B}_{i}\mathbf{\Sigma}\mathbf{B}_{i}\mathbf{\Sigma}^{1/2})\leq\tr(\mathbf{B}_{i}\mathbf{\Sigma}^{2})\leq p\lambda_{max}(\mathbf{\Sigma})^{2}.$
Again, by Theorem 1.1 in \cite{rudelson2013} and Assumption \ref{assum:matrix}, for some constant $C>0,$ we have
\begin{align*}
&\mathbb{P}\left({\epsilon}_{1\cdot}^{\prime} \mathbf{B}_{i} {\epsilon}_{1\cdot}>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}
{28\sqrt{\log N}}\right)\\
\leq&\mathbb{P}\left({\epsilon}_{1\cdot}^{\prime} \mathbf{B}_{i} {\epsilon}_{1\cdot}>\tr(\mathbf{\Sigma}^{1/2}\mathbf{B}_{i}\mathbf{\Sigma}^{1/2})+\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}
{56\sqrt{\log N}}\right)\\
\leq&2\exp\left\{-C\sqrt{T/\log N}\right\},
\end{align*}
where the first inequality holds because that for sufficiently large $N,T$, $\tr(\mathbf{\Sigma}^{1/2}\mathbf{B}_{i}\mathbf{\Sigma}^{1/2})\leq p\lambda_{max}(\mathbf{\Sigma})
\leq\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}/
\{56\sqrt{\log N}\},$ for all $i=1,\cdots,N.$
As $p> 0$ is fixed, we have
\begin{align*}
&\mathbb{P}\left(\max_{1 \leq i<j \leq N}\left|\hat{\rho}_{i j}-{\rho}_{i j}\right|>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}{\sqrt{\log N}\tr\left(\mathbf{\Sigma}\right)}\right)\\
\leq&2N^2\exp\left(-C\sqrt{T/\log N}\right)+2N^2\exp\left(-\eta_{0} T\right)\rightarrow0.
\end{align*}
By the slutsky lemma again, we can obtain that
\begin{align}\label{tilde Tmax}
\mathbb{P}\left(\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\max _{1 \leq i<j \leq N} \hat{\rho}^2_{ij}-4 \log N+\log \log N\leq y\right)
\rightarrow \exp \left\{-\frac{1}{\sqrt{8 \pi}} \exp \left(-\frac{y}{2}\right)\right\}.
\end{align}
Then, we complete the proof. \hfill$\Box$
\subsection{Proof of Theorem \ref{th:max cons}}
\begin{lemma}\label{lemax1}
Under the same assumptions as in Theorem \ref{th:max cons}.
For any $\varepsilon\in(0,1)$ and sufficiently large $T,$
$$
\mathbb{P}\bigg(\left|\hat{\sigma}_{i j}-\sigma_{i j}\right| \geq x\sqrt{\frac{\sigma_{i j}^{2}+(1+|\gamma_{2}|)\sigma_{i i} \sigma_{j j}}{N}}\bigg) \leq C \exp \left(-\frac{x^2}{2}(1-\varepsilon)\right),
$$
uniformly for $x\in(0,N^{\frac{1}{2}\wedge(\frac{1}{\tau}-\frac{1}{2})}),$ where $C$ does not depend on $i, j.$
\end{lemma}
\begin{lemma}\label{lemax2}
Under the same assumptions as in Theorem \ref{th:max cons}. Let $\hat{\mathbf{\Gamma}}=\big(\frac{\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}{T}\big)_{1 \leq i, j \leq N}, \hat{\gamma}_{N}=\|\hat{\mathbf{\Gamma}}\|_{\mathrm{F}}^{2}-\frac{1}{T}\{\tr(\hat{\mathbf{\Gamma}})\}^{2}$ and $\gamma_{N}=\left(\frac{\tr(\mathbf{\Sigma})}{T}\right)^{2}\|\mathbf{U}\|_{\mathrm{F}}^{2} .$ We have
\begin{align}\label{gamman}
\frac{\hat{\gamma}_{N}}{N}=\frac{\gamma_{N}}{N} a_{N}+b_{N},
\end{align}
where $\left\{a_{N}\right\}$ are real numbers satisfying $1-c_{1}/T \leq a_{N} \leq 1+c_{2}N / T$ for some constant $c_{1},c_{2}>0,\left\{b_{N}\right\}$ are random variables satisfying
$$
\mathbb{E}\left(b_{N}^{2}\right)=O\left(\frac{1}{NT}\right).$$
\end{lemma}
With preparations earlier, we are now ready to prove Theorem \ref{th:max cons}.
According to (\ref{tilde Tmax}), we have $$\tr^2(\mathbf{\Sigma})
/\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}\max _{1 \leq i<j \leq N} \hat{\rho}^2_{ij}=O_{p}(\log N).$$
Notice that
\begin{align*}
& \Big|\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}\max _{1 \leq i<j \leq N} \hat{\rho}^2_{ij}-\frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\max _{1 \leq i<j \leq N} \hat{\rho}^2_{ij} \Big| \\
\leq & \frac{\tr^2(\mathbf{\Sigma})}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\max _{1 \leq i<j \leq N} \hat{\rho}^2_{ij} \Big|\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}
\frac{\left\|{\mathbf{\Sigma}}\right\|^2_{\mathrm{F}}}{\tr^2({\mathbf{\Sigma}})}
-1\Big|.
\end{align*}
To obtain the limit null distribution of $L_{N},$ we only need to show that
$$\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}
\frac{\left\|{\mathbf{\Sigma}}\right\|^2_{\mathrm{F}}}{\tr^2({\mathbf{\Sigma}})}
-1=o_{p}\{\log^{-1}N\}.$$
%Note that the lemma \ref{lemax1} and \ref{lemax2} are very similar to the lemma B.1 and B.2 in the Supplement to \cite{chen2018}.
So,
similar to the proof of Theorem 2.1 in \cite{chen2018}, we can obtain that
\begin{align}\label{max:cons}
\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}
\frac{\left\|{\mathbf{\Sigma}}\right\|^2_{\mathrm{F}}}{\tr^2({\mathbf{\Sigma}})}
=1+O_{P}\bigg\{\Big(\sqrt{\frac{\log T}{N}}\Big)^{\min (1,2-\tau)}\bigg\}
=1+o_{p}\{\log^{-1}N\},
\end{align}
due to Lemmas \ref{lemax1} and \ref{lemax2}. Then, we complete the proof and obtain that
$$\mathbb{P}\left(\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-4\log N+\log \log N\leq y\right) \rightarrow \exp \left\{-\frac{1}{\sqrt{8 \pi}} \exp \left(-\frac{y}{2}\right)\right\}.$$
\hfill$\Box$
%\begin{remark}
%It is worth noting that our proof idea is similar to that of \cite{chen2018}, but since we prove the limit distribution of max type statistic under sub-Gaussian distributions, the proof techniques used are not completely similar, such as Hanson-Wright inequality in \cite{rudelson2013}, and we also need to prove that the error caused by the residuals is negligible.
%\end{remark}
\subsection{Proof of Theorem \ref{th:maxpower}}
Under the assumptions in Theorem \ref{th:maxpower}, we establish two similar lemmas.
\begin{lemma}\label{H1:lemax1}
Under the same assumptions as in Theorem \ref{th:maxpower}. For any $\varepsilon\in(0,1)$ and sufficiently large $T,$
$$
\mathbb{P}\bigg(\left|\hat{\sigma}_{i j}-\sigma_{i j}\right| \geq x\sqrt{\frac{\tr(\mathbf{U}^2)(\sigma_{i i} \sigma_{j j}+\sigma_{i j}^{2})}{N^2}}\bigg) \leq C \exp \left(-C^{\prime}x^2(1-\varepsilon)\right),
$$
uniformly for $x\in(0,N^{\frac{1}{2}\wedge(\frac{1}{\tau}-\frac{1}{2})}),$ where $C$ and $C^{\prime} $do not depend on $i, j.$
\end{lemma}
\begin{lemma}\label{H1:lemax2}
Under the same assumptions as in Theorem \ref{th:maxpower}. Let $\hat{\mathbf{\Gamma}}=\big(\frac{\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}{T}\big)_{1 \leq i, j \leq N}, \hat{\gamma}_{N}=\|\hat{\mathbf{\Gamma}}\|_{\mathrm{F}}^{2}-\frac{1}{T}\{\tr(\hat{\mathbf{\Gamma}})\}^{2}$ and $\gamma_{N}=\left(\frac{\tr(\mathbf{\Sigma})}{T}\right)^{2}\|\mathbf{U}\|_{\mathrm{F}}^{2} .$ We have
\begin{align}\label{H1:gamman}
\frac{\hat{\gamma}_{N}}{N}=\frac{\gamma_{N}}{N} a_{N}+b_{N},
\end{align}
where $\left\{a_{N}\right\}$ are real numbers satisfying $1-c_{1}/T\leq a_{N} \leq 1+c_{2}N / T$ for some constant $c_{1},c_{2}>0,\left\{b_{N}\right\}$ are random variables satisfying
$$
\mathbb{E}\left(b_{N}^{2}\right)=O\left(\frac{1}{T^{\frac{1}{2}\vee \left(\frac{1}{\tau}-\frac{1}{2}\right)}}\right).$$
\end{lemma}
We are now ready to prove Theorem \ref{th:maxpower}.
Because for any $1\leq i\leq j\leq N,$ ${\epsilon}_{i\cdot}^{\prime}{\epsilon}_{j\cdot}={Z}^{\prime}\left[\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\otimes \mathbf{\Sigma}\right]{Z}$
and
${\epsilon}_{j\cdot}^{\prime}\mathbf{B}_{i}{\epsilon}_{j\cdot}={Z}^{\prime}\left[\{u_{j}^{1/2}(u_{j}^{1/2})^{\prime}\}\otimes (\mathbf{\Sigma}^{1/2}\mathbf{B}_{i}\mathbf{\Sigma}^{1/2})\right]{Z}.$
Similar to (\ref{pijhat-pij}), we can use the similar approach to prove that
$$\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\hat{\rho}_{ij}-{\rho}_{ij}\right|>\frac{\epsilon\sqrt{\tr\left(\mathbf{\Sigma}^2\right)}}{\sqrt{\log N}\tr\left(\mathbf{\Sigma}\right)}\right)\rightarrow 0,$$
and
$$\mathbb{P}\left(\max_{1\leq i<j\leq N}\left|\frac{\epsilon_{i\cdot}^{\prime}\epsilon_{i\cdot}\epsilon_{j\cdot}^{\prime}\epsilon_{j\cdot}}{T^2}-
\frac{\tr^2(\mathbf{\Sigma})}{T^2}\right|>C_{3}\sqrt{\frac{\log N}{T}}\right)
=O\left(N^{-M}\right).
$$
In addition, similar to the proof of Theorem 2.1 in \cite{chen2018}, we can also prove that
\begin{align}\label{max:cons}
\frac{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}{\tr^2(\tilde{\mathbf{\Sigma}})}
\frac{\tr^2({\mathbf{\Sigma}})}{\left\|{\mathbf{\Sigma}}\right\|^2_{\mathrm{F}}}
=1+O_{P}\bigg\{\Big(\sqrt{\frac{\log T}{N}}\Big)^{\min (1,2-\tau)}\bigg\},
\end{align}
due to Lemmas \ref{H1:lemax1} and \ref{H1:lemax2}.
Note that $$\sqrt{\var\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)}= \sqrt{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}\left(u_{i i} u_{j j}+u_{i j}^{2}\right)+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]\tr(\mathbf{\Sigma}\circ
\mathbf{\Sigma})},$$
where
$u_{l}^{1/2}\in \mathbb{R}^{N}$ denotes the $l$-th row vector of matrix $\mathbf{U}^{1/2}.$ Note that we assume that $(\mathbf{U})_{ii}=1,$ for all $1\leq i\leq N,$
so, we have $$\tr\left[\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]\tr(\mathbf{\Sigma}\circ
\mathbf{\Sigma})\leq \tr(\mathbf{\Sigma}^{2}).$$
Without loss of generality, we can assume that $u_{12} \geq \delta \sqrt{\frac{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}}{\tr^{2}(\mathbf{\Sigma})}\log N }$ for some constant $\delta>2$.
Note that for some constant $C>0,$
by Theorem 1.1 in \cite{rudelson2013} and Assumption \ref{assum:matrix},
we have
\begin{align*}
\mathbb{P}\left(\sqrt{\frac{\log N}{T}}\frac{\epsilon_{1\cdot}^{\prime}\epsilon_{2\cdot}-u_{12}\tr(\mathbf{\Sigma})}
{\sqrt{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}}}>\varepsilon\right)
\leq2\exp\left\{-C\varepsilon\log N\right\}\rightarrow 0,
\end{align*}
for all $\varepsilon>0.$
Then, there exist a constant $M>0$ satisfying
\begin{align*}
&\mathbb{P}\left(\sqrt{\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}}\max_{1\leq i< j\leq N}\hat{\rho}_{ij} > \sqrt{4 \log N-\log \log N+w_{\alpha}}\right)\\
\geq&
\mathbb{P}\left(\sqrt{\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}}\hat{\rho}_{12} > \sqrt{4 \log N-\log \log N+w_{\alpha}}\right)\\
\geq&\mathbb{P}\left(\sqrt{\frac{\tr^{2}(\mathbf{\Sigma})}{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}} }\hat{\rho}_{12} > \sqrt{4 \log N-\log \log N+w_{\alpha}}\Big(1+\frac{M\sqrt{\log N}}{\sqrt{N}}\Big)\right)+o(1)
\\
\geq&\mathbb{P}\left(\sqrt{\frac{\tr^{2}(\mathbf{\Sigma})}{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}} }{\rho}_{12}>\sqrt{4 \log N-\log \log N+w_{\alpha}}\Big(1+\frac{M\sqrt{\log N}}{\sqrt{N}}\Big)+o(1)\right)+o(1)\\
\geq&\mathbb{P}\left(\sqrt{\frac{\tr^{2}(\mathbf{\Sigma})}{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}} }{\rho}_{12} > \sqrt{4 \log N-\log \log N+w_{\alpha}}+o(1)\right)+o(1)\\
\geq&
\mathbb{P}\left(\sqrt{\frac{\tr^{2}(\mathbf{\Sigma})}{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}} } \frac{\epsilon_{1\cdot}^{\prime}\epsilon_{2\cdot}-\operatorname{tr}(\mathbf{\Sigma}) u_{12}}{\sqrt{\epsilon_{1\cdot}^{\prime}\epsilon_{1\cdot}\epsilon_{2\cdot}^{\prime}\epsilon_{2\cdot}}}
>-(\delta-2) \sqrt{\log N} / 2\right)+o(1),\\
\geq&
\mathbb{P}\left(\frac{\epsilon_{1\cdot}^{\prime}\epsilon_{2\cdot}-\operatorname{tr}(\mathbf{\Sigma}) u_{12}}{\sqrt{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}}}
>-(\delta-2) \sqrt{\log N} / 2\right)+o(1)\\
=&1-\mathbb{P}\left(\frac{\epsilon_{1\cdot}^{\prime}\epsilon_{2\cdot}-\operatorname{tr}(\mathbf{\Sigma}) u_{12}}{\sqrt{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}}}
\leq-(\delta-2) \sqrt{\log N} / 2\right)+o(1)\\
\geq&1-\mathbb{P}\left(\Big|\frac{\epsilon_{1\cdot}^{\prime}\epsilon_{2\cdot}-\operatorname{tr}(\mathbf{\Sigma}) u_{12}}{\sqrt{\|\mathbf{\Sigma}\|_{\mathrm{F}}^{2}}}\Big|
\geq(\delta-2) \sqrt{\log N} / 2\right)+o(1)\\
=&1-O(N^{-\varepsilon_{1}})+o(1)\rightarrow 1,
\end{align*}
where $\varepsilon_{1}=C\{(\delta-2)/2\}^{2}>0,$ for some constant $C>0.$
Consequently, we complete the proof of this theorem.\hfill$\Box$
\subsection{Proof of Theorem \ref{th:sum-max}}
Define
\begin{align}\label{signdef}
&\tilde{T}_{max}=
\max_{1\leq i<j\leq N}\frac{(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot})^2}
{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\n\\
&\Lambda_{N}=\{(i, j) ; 1 \leq i<j \leq N\} \nonumber\n\\
&A_{N}=\left\{S_{N}/\sigma_{S_{N}} \leq x\right\} \quad \text { and } \quad
B_{I}=\left\{\left|\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right|\geq l_{N}\right\}
\end{align}
for any $I=(i, j) \in \Lambda_{N},$ where $a_{N}=4\log N-\log \log N+y$
and
$$l_{N}=\sqrt{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}\left[4\log N-\log \log N+y\right]}
=\sqrt{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}a_{N}}.$$
To make a clear presentation, we impose a trivial ordering for elements in $\Lambda_{N} .$ For any $I_{1}=\left(i_{1}, j_{1}\right) \in \Lambda_{N}$ and $I_{2}=\left(i_{2}, j_{2}\right) \in$ $\Lambda_{N},$ we say $I_{1}<I_{2}$ if $i_{1}<i_{2}$ or $i_{1}=i_{2}$ but $j_{1}<j_{2}.$
Recall that under certain assumptions, \cite{baltagi2016} established the asymptotic property of $S_{N}$ that under the null hypothesis,
$S_{N}/{\hat{\sigma}}_{S_{N}}\to \mathcal{N}(0,1)$ in distribution when $\min(N,T)\rightarrow \infty.$
Since the assumptions in this paper are slightly different from those in \cite{baltagi2016},
we reconsider the asymptotic properties of $S_{N}$ under our
Assumptions \ref{assum:E distribution}-\ref{assum:matrix}. Similar to Theorems 2-3 in \cite{baltagi2016}, Lemma \ref{th:sum null} presents the asymptotic null distribution of $S_{N}$,
and Lemma \ref{th:sum cons} presents that $\hat{\sigma}_{S_{N}}$ is a ratio-consistent estimator of the variance of $S_{N}$.
\begin{lemma}\label{th:sum null}
Under Assumptions \ref{assum:E distribution}-\ref{assum:matrix} and the null hypothesis,
we have $S_{N}/{\sigma}_{S_{N}}\to \mathcal{N}(0,1)$ in distribution, as $\min(N,T)\rightarrow \infty$ with $\underset{\min(N,T) \rightarrow \infty}{\lim}N/T=\gamma\in (0,+\infty).$
Here,
$\mathbf{M}_{i}=\mathbf{P}_{i}\mathbf{\Sigma}\mathbf{P}_{i},$ for any $1\leq i\leq N,$
and $$\sigma^2_{S_{N}}=\frac{2}{N(N-1)}\underset{1\leq i<j\leq N}{\sum\sum}\tr\left(\mathbf{M}_{i}\mathbf{M}_{j}\right)
/\{\tr\left(\mathbf{M}_{i}\right)\tr\left(\mathbf{M}_{j}\right)\}.$$
\end{lemma}
%{\color{red}Since $\mathbf{\Sigma}$ is unobservable, $\hat{\sigma}^2_{S_{N}}$ can be estimated by}
Recall that
\begin{align}\label{sigmasumhat}
\hat{\sigma}^2_{S_{N}}=\frac{2}{N(N-1)}\underset{1\leq i<j\leq N}{\sum\sum} v_{j}^{\prime}\left(v_{i}-\bar{v}_{ij}\right) v_{i}^{\prime}\left(v_{j}-\bar{v}_{i j}\right),
\end{align}
$\bar{v}_{ij}=\sum_{1<k \neq i, j<N} v_{k}/(N-2)$ and $v_{k}=\hat{\epsilon}_{k\cdot}/\|\hat{\epsilon}_{k\cdot}\|$ for all $1\leq k\leq N.$
\begin{lemma}\label{th:sum cons}
Under Assumptions \ref{assum:E distribution}-\ref{assum:matrix} and the null hypothesis, as $\min(N,T)\rightarrow \infty$ with $\underset{\min(N,T) \rightarrow \infty}{\lim}N/T=\gamma\in (0,+\infty)$, we have $\hat{\sigma}^2_{S_{N}}/{\sigma}^2_{S_{N}}\to 1$ in probability.
\end{lemma}
Next, the following lemmas are provided for establishing the asymptotic independence between the two test statistics $S_{N}$
and $L_{N}.$
\begin{lemma}\label{indep hat}
Assume that Assumptions \ref{assum:E distribution}-\ref{assum:matrix} hold, under $H_{0},$ if $\tilde{T}_{max}-4 \log N+\log \log N$ and $S_{N}/\sigma_{S_{N}}$ are asymptotically independent,
then $L_{N}{\tr^2(\tilde{\mathbf{\Sigma}})}/{\|\tilde{\mathbf{\Sigma}}\|_{\mathrm{F}}^{2}}-4 \log N+\log \log N$ and $S_{N}/\hat{\sigma}_{S_{N}}$ are also asymptotically
independent.
\end{lemma}
\begin{lemma}\label{linear}
Under $H_0$ and same assumptions in Lemma \ref{indep hat}, let
$$
H(N, k)=\sum_{I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right).
$$
Then $\lim _{k \rightarrow \infty} \limsup _{\min(N,T) \rightarrow \infty} H(N, k)=0.$
\end{lemma}
\begin{lemma}\label{linear2}
Under the assumptions of Lemma \ref{indep hat},
$$
\sum_{I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N}}\left[\mathbb{P}\left(A_{N} B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)-\mathbb{P}\left(A_{N}\right) \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)\right] \rightarrow 0
$$
as $\min(N,T) \rightarrow \infty$ for each $k \geq 1.$
\end{lemma}
Next, we are ready to prove asymptotic independence stated in Theorem \ref{th:sum-max}.
By (\ref{eiej}) and Lemma \ref{th:sum null}, the following hold,
\begin{align}
&\max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}-4\log N+\log \log N\rightarrow G(y)\,\text { in distribution; }\label{demaxQ}\\
&\frac{S_{N}}{\sigma_{S_{N}}} \rightarrow N(0,1)\text { in distribution, }\label{desumQ}
\end{align}
where $G(y)=\exp \left\{- \exp \left(-y/2\right)/\sqrt{8\pi}\right\}.$
To show asymptotic independence, according to Lemma \ref{indep hat}, it is enough to show the limit of
\begin{align}
\lim _{\min(N,T) \rightarrow \infty} \mathbb{P}\left(\frac{S_{N}}{\sigma_{S_{N}}}\leq x, \max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}\leq a_{N}\right)=\Phi(x) \cdot G(y),
\end{align}
for any $x \in \mathbb{R}$ and $y \in \mathbb{R}$, where $\Phi(x)=(2 \pi)^{-1 / 2} \int_{-\infty}^{x} e^{-t^{2} / 2} d t$ and
$$a_{N}=4\log N-\log\log N+y,$$
which makes sense for large $N$. Because of (\ref{demaxQ}) and (\ref{desumQ}), the above is equivalent to that
\begin{align}\label{indedengjia1Q}
\lim _{\min(N,T) \rightarrow \infty} \mathbb{P}\left(\frac{S_{N}}{\sigma_{S_{N}}}\leq x, \max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}> a_{N}\right)=\Phi(x) \cdot\left\{1-G(y)\right\},
\end{align}
for any $x \in \mathbb{R}$ and $y \in \mathbb{R}$. Review notation $\Lambda_{N}, A_{N}$ and $B_{I}$ for any $I=$ $(i, j) \in \Lambda_{N}$ in (\ref{signdef}). Write
\begin{align}\label{unionQ}
\mathbb{P}\left(\frac{S_{N}}{\sigma_{S_{N}}}\leq x, \max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}> a_{N}\right)=\mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} A_{N} B_{I}\right).
\end{align}
Here the notation ${A}_{N} B_{I}$ stands for ${A}_{N} \cap B_{I}$. From the inclusion-exclusion principle,
\begin{align}\label{UAB2k+1Q}
\mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} A_{N} B_{I}\right) \leq & \sum_{I_{1} \in \Lambda_{N}} \mathbb{P}\left(A_{N} B_{I_{1}}\right)-\sum_{I_{1}<I_{2} \in \Lambda_{N}} \mathbb{P}\left(A_{N} B_{I_{1}} B_{I_{2}}\right)+\cdots+ \n\\
&\sum_{I_{1}<I_{2}<\cdots<I_{2 l+1} \in \Lambda_{N}}\mathbb{P}\left(A_{N} B_{I_{1}} B_{I_{2}} \cdots B_{I_{2 l+1}}\right)
\end{align}
and
\begin{align}\label{UAB2kQ}
\mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} A_{N} B_{I}\right) \geq& \sum_{I_{1} \in \Lambda_{N}} \mathbb{P}\left(A_{N} B_{I_{1}}\right)-\sum_{I_{1}<I_{2} \in \Lambda_{N}} \mathbb{P}\left(A_{N} B_{I_{1}} B_{I_{2}}\right)+\cdots-\\
&\sum_{I_{1}<I_{2}<\cdots<I_{2 l} \in \Lambda_{N}} \mathbb{P}\left(A_{N} B_{I_{1}} B_{I_{2}} \cdots B_{I_{2 l}}\right)
\end{align}
for any integer $l \geq 1$. Reviewing the definition
$$
H(N, k)=\sum_{I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right),
$$
for $k \geq 1$ in Lemma $\ref{linear},$ we have from the lemma that
\begin{align}\label{lHpkQ}
\lim _{k \rightarrow \infty} \limsup _{\min(N,T) \rightarrow \infty} H(N, k)=0.
\end{align}
Set
$$\zeta(p,k)
=\sum_{I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N}}\left\{\mathbb{P}\left(A_{N} B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)-\mathbb{P}\left(A_{N}\right) \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)\right\},
$$
for $T \geq 1 .$ By Lemma \ref{linear2},
\begin{align}\label{zetaQ}
\lim _{\min(N,T) \rightarrow \infty} \zeta(N, k)=0,
\end{align}
for each $k \geq 1$. The assertion (\ref{UAB2k+1Q}) implies that
\begin{align}\label{UzetaH2k+1Q}
& \mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} A_{N} B_{I}\right) \n\\
\leq & \mathbb{P}\left(A_{N}\right)\left\{\sum_{I_{1} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}}\right)-\sum_{I_{1}<I_{2} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}}\right)+\cdots-\right.\n\\
&\left.\sum_{I_{1}<I_{2}<\cdots<I_{2 l} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{2 l}}\right)\right\}+\left\{\sum_{k=1}^{2 l} \zeta(N, k)\right\}+H(N, 2 l+1) \n\\
\leq P &\left(A_{N}\right) \cdot \mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} B_{I}\right)+\left\{\sum_{k=1}^{2 l} \zeta(N, k)\right\}+H(N, 2 l+1),
\end{align}
where the inclusion-exclusion formula is used again in the last inequality, that is
$$
\begin{array}{r}
\mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} B_{I}\right) \geq\left\{\sum_{I_{1} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}}\right)-\sum_{I_{1}<I_{2} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}}\right)+\cdots-\right. \\
\left.\sum_{I_{1}<I_{2}<\cdots<I_{2 l} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{2 l}}\right)\right\},
\end{array}
$$
for all $l \geq 1$. By the definition of $a_{N}$ and (\ref{demaxQ}),
\begin{align*}
\mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} B_{I}\right)&=\mathbb{P}\left(\max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}>a_{N}\right)\n\\
&=\mathbb{P}\left\{\max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}
-4 \log N+\log \log N>y\right\} \n\\
&\rightarrow 1-G(y),
\end{align*}
as $\min(N,T) \rightarrow \infty$. By (\ref{desumQ}), $\mathbb{P}\left(A_{N}\right) \rightarrow \Phi(x)$ as $\min(N,T) \rightarrow \infty$. From (\ref{unionQ}), by fixing $l$ first and sending $\min(N,T) \rightarrow \infty$, we get from (\ref{zetaQ}) that
$$
\begin{aligned}
& \limsup _{\min(N,T) \rightarrow \infty} \mathbb{P}\left(\frac{S_{N}}{\sigma_{S_{N}}}\leq x, \max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}>a_{N}\right) \\
\leq & \Phi(x) \cdot\left\{1-G(y)\right\}+\limsup _{\min(N,T) \rightarrow \infty} H(N, 2 l+1).
\end{aligned}
$$
Now, let $l\rightarrow \infty$ and use (\ref{lHpkQ}) to see
\begin{align}\label{limitrightQ}
\limsup _{\min(N,T) \rightarrow \infty} \mathbb{P}\left(\frac{S_{N}}{\sigma_{S_{N}}}\leq x, \max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}>a_{N}\right) \leq \Phi(x) \cdot\left\{1-G(y)\right\}.
\end{align}
By applying the same argument to (\ref{UAB2kQ}), we see that the counterpart of
(\ref{UzetaH2k+1Q}) becomes
\begin{align*}
\mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} A_{N} B_{I}\right)
\geq & \mathbb{P}\left(A_{N}\right)\left\{\sum_{I_{1} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}}\right)-\sum_{I_{1}<I_{2} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}}\right)+\cdots+\right.\n\\
&\left.\sum_{I_{1}<I_{2}<\cdots<I_{2 l-1} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{2 l-1}}\right)\right\}+\n\\
&\left\{\sum_{k=1}^{2 l-1} \zeta(N, k)\right\}-H(N, 2 l) \n\\
\geq & \mathbb{P}\left(A_{N}\right) \cdot \mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} B_{I}\right)+\left\{\sum_{k=1}^{2 l-1} \zeta(N, k)\right\}-H(N, 2 l),
\end{align*}
where in the last step we use the inclusion-exclusion principle such that
\begin{align*}
\mathbb{P}\left(\bigcup_{I \in \Lambda_{N}} B_{I}\right) \leq\left\{\sum_{I_{1} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}}\right)-\sum_{I_{1}<I_{2} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}}\right)+\cdots+\right. \\
\left.\sum_{I_{1}<I_{2}<\cdots<I_{2 l-1} \in \Lambda_{N}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{2 l-1}}\right)\right\},
\end{align*}
for all $l \geq 1$. Review (\ref{unionQ}) and repeat the earlier procedure to see
$$
\limsup _{\min(N,T) \rightarrow \infty} \mathbb{P}\left(\frac{S_{N}}{\sigma_{S_{N}}}\leq x, \max _{1 \leq i<j \leq N} \frac{\left(\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot}\right)^2}{\left\|\mathbf{\Sigma}\right\|_{\mathrm{F}}^{2}}>a_{N}\right) \geq \Phi(x) \cdot\left\{1-G(y)\right\},
$$
by sending $\min(N,T) \rightarrow \infty$ and then sending $l \rightarrow \infty$. This and (\ref{limitrightQ}) yield (\ref{indedengjia1Q}). The proof is completed due to Lemma \ref{indep hat}.
\hfill$\Box$
\section{Proofs of the lemmas}\label{proof of lemmas}
In this section, we will prove the lemmas used in the previous section.
\subsection{Proof of Lemma \ref{lemax1}}
Recall that $\mathbf{M}_{i}=\mathbf{P}_{i}\mathbf{\Sigma}\mathbf{P}_{i},$
$\mathbf{B}_{i}= \mathbf{x}_{i}\left(\mathbf{x}_{i}^{\prime} \mathbf{x}_{i}\right)^{-1} \mathbf{x}_{i}^{\prime}.$
$(\mathbf{A})_{ij}$ is the element at row $i$ and column $j$ of matrix
$\mathbf{A},$ for any square matrix $\mathbf{A}.$ Hence, we define the element at row $i$ and column $j$ of matrix $\mathbf{M}_{l}=\mathbf{P}_{l}\mathbf{\Sigma}\mathbf{P}_{l}$ as $\sigma_{ij,l},$ $1\leq i \leq j \leq N.$ Let $\omega_{ij,l}=\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{l}e_{i}e_{j}^{\prime}\mathbf{P}_{l}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{l}e_{i}e_{j}^{\prime}\mathbf{P}_{l}\mathbf{\Sigma}^{1/2})\}\leq \sigma_{ii,l}\sigma_{jj,l},$
for $1\leq i <j\leq T;1\leq l\leq N.$
Then, we claim that
\begin{align}\label{sigmaij,l-sigmaij}
\Big|\Big(\mathbf{B}_{l}\mathbf{\Sigma}\Big)_{i j}\Big|=O(T^{-1 \vee-\frac{1}{\tau}}),\,\,\,\,\Big|\Big(\mathbf{\Sigma}\mathbf{B}_{l}\Big)_{i j}\Big|=O(T^{-1 \vee-\frac{1}{\tau}}),\,\,\,\,\,\Big|\Big(\mathbf{B}_{l}\mathbf{\Sigma}\mathbf{B}_{s}\Big)_{i j}\Big|=O(T^{-1\vee-\frac{1}{\tau}}),
\end{align}
for any $1\leq l,s\leq N,\,1\leq i,j\leq T.$
To simplify the expression, we let $b_{ij,l}\doteq\Big(\mathbf{B}_{l}\Big)_{i j} ,$
for any $1\leq l\leq N,\,1\leq i,j\leq T.$ Under Assumption \ref{assum:x_it}, we have $\max_{1\leq j\leq N}|b_{ij,l}|=O(T^{-1}).$
\begin{align*}
\Big|\Big(\mathbf{B}_{l}\mathbf{\Sigma}\Big)_{i j}\Big|=
\Big|\sum_{k=1}^{T}b_{ik,l}\sigma_{kj}\Big|
&\leq \sum_{k=1}^{T}|b_{ik,l}||\sigma_{kj}|\\
&\leq O(T^{-1})\sum_{k=1}^{T}|\sigma_{kj}|.
\end{align*}
Due to Assumption \ref{assum:matrix}, we have $\sum_{k=1}^{T}\left|\sigma_{j k}\right|^{\tau} \leq C$ for some $0<\tau<2$ and $1\leq j \leq T.$ So, if $0<\tau<1,$ we have $\sum_{k=1}^{T}|\sigma_{kj}|\leq\sum_{k=1}^{T}|\sigma_{kj}|^{\tau}|\sigma_{kj}|^{1-\tau}=O(1).$
If $1<\tau<2,$ from a convex inequality we have $$\left\{\big(\sum_{k=1}^{T}|\sigma_{kj}|\big)^{\tau}\right\}^{1/\tau}\leq \left\{T^{\tau-1}\sum_{k=1}^{T}|\sigma_{kj}|^{\tau}\right\}^{1/\tau}\leq O(T^{1-\frac{1}{\tau}}).$$
So, $
\Big|\Big(\mathbf{B}_{l}\mathbf{\Sigma}\Big)_{i j}\Big|=O(T^{-1\vee-\frac{1}{\tau}}).$
Similarly, we have $\Big|\Big(\mathbf{\Sigma}\mathbf{B}_{l}\Big)_{i j}\Big|=O(T^{-1\vee-\frac{1}{\tau}})$ and $\Big|\Big(\mathbf{B}_{l}\mathbf{\Sigma}\mathbf{B}_{s}\Big)_{i j}\Big|=O(T^{-1\vee-\frac{1}{\tau}}).$
Then, due to $\mathbf{P}_{l}\mathbf{\Sigma}\mathbf{P}_{l}=\mathbf{\Sigma}-
\mathbf{B}_{l}\mathbf{\Sigma}-\mathbf{\Sigma}\mathbf{B}_{l}
+\mathbf{B}_{l}\mathbf{\Sigma}\mathbf{B}_{l},$
we can conclude that
$$\left|\sum_{l=1}^{N}\frac{\sigma_{i j,l}-\sigma_{ij}}{N}\right|=O(T^{-1\vee-\frac{1}{\tau}})$$
and
$$\left|\sum_{l=1}^{N}\frac{\sigma_{i i,l} \sigma_{j j,l}+\sigma_{i j,l}^{2}-\sigma_{ii}\sigma_{jj}-\sigma^2_{ij}}{N}\right|=O(T^{-1\vee-\frac{1}{\tau}}).$$
Recall that $\bar{\hat{\epsilon}}_{\cdot j}=\sum_{l=1}^{N}\hat{\epsilon}_{lj}/N,$ $1\leq j\leq N,$
we have $\hat{\sigma}_{i j}=\frac{1}{N-1} \sum_{l=1}^{N} \hat{\epsilon}_{l i} \hat{\epsilon}_{l j} -\frac{N}{N-1} \bar{\hat{\epsilon}}_{\cdot i} \bar{\hat{\epsilon}}_{ \cdot j}$ Since $\cov\left(\hat{\epsilon}_{l i}, \hat{\epsilon}_{l j}\right)=\sigma_{i j,l}$, we obtain that $\var\left(\hat{\epsilon}_{l i} \hat{\epsilon}_{l j}\right)=\sigma_{i j,l}^{2}+\sigma_{i i,l} \sigma_{j j,l}+\gamma_{2}\omega_{ij,l}.$
By classical Cram\'er type large deviation results for independent random variables (see Corollary 3.1 in \cite{saulis1991limit}), we have for any $\varepsilon>0$,
\begin{align}\label{large deviation}
&\mathbb{P}\Big(\Big|\frac{\sum_{l=1}^{N}\left(\hat{\epsilon}_{l i} \hat{\epsilon}_{l j}-\sigma_{i j,l}\right)}{\sqrt{\sum_{l=1}^{N}\left(\sigma_{i j,l}^{2}+(1+|\gamma_{2}|)\sigma_{i i,l} \sigma_{j j,l}\right)}}\Big| \geq x\Big)\\
\leq&
\mathbb{P}\Big(\Big|\frac{\sum_{l=1}^{N}\left(\hat{\epsilon}_{l i} \hat{\epsilon}_{l j}-\sigma_{i j,l}\right)}{\sqrt{\sum_{l=1}^{N}\left(\sigma_{i i,l} \sigma_{j j,l}+\sigma_{i j,l}^{2}+\gamma_{2}\omega_{ij,l}\right)}}\Big| \geq x\Big) \\
\leq& C \exp \Big\{-\frac{x^{2}}{2}(1-\varepsilon)\Big\}
\end{align}
uniformly in $x \in\big[0, o(\sqrt{N})\big) .$ For $\bar{\hat{\epsilon}}_{\cdot j}$, we have $\var\left(\bar{\hat{\epsilon}}_{\cdot j}\right)=\sum_{l=1}^{ N} \sigma_{jj,l}/N^{2} .$ By (\ref{sigmaij,l-sigmaij}), $\var\left(\bar{\hat{\epsilon}}_{\cdot j}\right)\leq C N^{-1},$ uniformly in $1 \leq j \leq T .$ Again, by classical Cram\'er type large deviation results for independent random variables, we have for any $\varepsilon>0$,
$$
\mathbb{P}\left(\left|\bar{\hat{\epsilon}}_{\cdot j}\right| \geq x \sqrt{\var\left(\bar{\hat{\epsilon}}_{\cdot j}\right)}\right) \leq C \exp \Big\{-\frac{x^{2}}{2}(1-\varepsilon)\Big\}
$$
uniformly in $x \in\big[0, o(\sqrt{N})\big) .$ So
$$
\mathbb{P}\left(\left|\bar{\hat{\epsilon}}_{\cdot i}\bar{\hat{\epsilon}}_{\cdot j} \right| \geq x^{2} \sqrt{\var\left(\bar{\hat{\epsilon}}_{\cdot i}\right) \var\left(\bar{\hat{\epsilon}}_{\cdot j}\right)}\right) \leq 2C \exp \Big\{-\frac{x^{2}}{2}(1-\varepsilon)\Big\}
$$
uniformly in $x \in\big[0, o(\sqrt{N})\big) .$ We have, uniformly for $x \in\big[0, o(\sqrt{N})\big), x^{2} \sqrt{\var\left(\bar{\hat{\epsilon}}_{\cdot i}\right) \var\left(\bar{\hat{\epsilon}}_{\cdot i}\right)}=o(x / \sqrt{N})$. So for any $\delta>0$ and large $N$
\begin{align}\label{tail normal}
\mathbb{P}\left(\left|\bar{\hat{\epsilon}}_{\cdot i}\bar{\hat{\epsilon}}_{\cdot j}\right| \geq \delta \frac{x}{\sqrt{N}}\right) \leq 2C \exp \Big\{-\frac{x^{2}}{2}(1-\varepsilon)\Big\}
\end{align}
uniformly for $x \in\big[0, o(\sqrt{N})\big).$ Hence,
the lemma follows from (\ref{large deviation}) and (\ref{tail normal}).
%, for any $\varepsilon\in(0,1),$ we have
%\begin{align}\label{sigmaij,l}
%\mathbb{P}\bigg\{\Big|\hat{\sigma}_{ij}-\sum_{l=1}^{N}\frac{\sigma_{i j,l}}{N-1}\Big| \geq \frac{x}{N-1}\sqrt{\sum_{l=1}^{N}\sigma_{i j,l}^{2}+(1+|\gamma_{2}|)\sigma_{i i,l} \sigma_{j j,l}}\Big\} \leq C \exp \Big\{-\frac{x^{2}}{2}(1-\varepsilon)\bigg\}.
%\end{align}
%This
%proves the lemma.
\hfill$\Box$
\subsection{Proof of Lemma \ref{lemax2}}
So, according to the definitions in Lemma \ref{lemax2},
we have
\begin{align*}
\|\hat{\mathbf{\Gamma}}\|_{\mathrm{F}}^{2}&=\frac{1}{T^2}\sum_{i=1}^{N} \sum_{j=1}^{N}( \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2},\\
\{\tr(\hat{\mathbf{\Gamma}})\}^{2}&=\frac{1}{T^2}\left(\sum_{i=1}^{N}
\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2},\\
T^2a_{N}&=\frac{1}{\gamma_{N}}\left[\sum_{i=1}^{N} \sum_{j=1}^{N} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}-\frac{1}{T} \mathbb{E}\left(\sum_{i=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\right],\\
T^2b_{N}&=\frac{1}{N}\left[\sum_{i=1} \sum_{j=1} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}-\frac{1}{T}\left(\sum_{i=1} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}-\sum_{i=1} \sum_{j=1} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}+\frac{1}{T} \mathbb{E}\left(\sum_{i=1} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\right].
\end{align*}
It is easy to verify that $a_{N}$ and $b_{N}$ will make the equation (\ref{gamman}) true. In the following, we will prove that $a_{N}, b_{N}$ satisfy the properties in the lemma.
We first deal with the term $a_{N} .$ Recall that $\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot}$ and $\mathbf{P}_{i}\mathbf{\Sigma}\mathbf{P}_{i}=\mathbf{M}_{i},$ we have
\begin{equation*}
\mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2} =\left\{
\begin{aligned}
&\tr(\mathbf{M}_{i}\mathbf{M}_{j}) , & i\neq j, \\
&2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\} , & i=j.
\end{aligned}
\right.
\end{equation*}
Therefore,
$$\sum_{i=1}^{N}\sum_{j=1}^{N}\mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2} =\underset{i\neq j}{\sum^{N}\sum^{N}}\tr(\mathbf{M}_{i}\mathbf{M}_{j}) +\sum_{i=1}^{N}2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}.$$ Moreover, by Lemma \ref{le:moment of quadratic form},
\begin{align*}
\mathbb{E}\left(\sum_{i=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2} =&\sum_{i=1}^{N} \sum_{j=1}^{N} \mathbb{E}( \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot} \hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})\\
=&\sum_{i=1}^{N}\mathbb{E}( \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})+\underset{i\neq j}{\sum^{N}\sum^{N}}\mathbb{E}( \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot} \hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})\\
=&\sum_{i=1}^{N}2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&+\underset{i\neq j}{\sum^{N}\sum^{N}}\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j}).
\end{align*}
So we have
\begin{align*}
{T^2} \gamma_{N}a_{N}&=\left[\sum_{i=1}^{N} \sum_{j=1}^{N} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}-\frac{1}{T} \mathbb{E}\left(\sum_{i=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\right]\\
=&\underset{i\neq j}{\sum^{N}\sum^{N}}\tr(\mathbf{M}_{i}\mathbf{M}_{j}) +\sum_{i=1}^{N}2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&-\frac{1}{T}\Big[\sum_{i=1}^{N}2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})+
\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&\quad\quad\quad+\underset{i\neq j}{\sum^{N}\sum^{N}}\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\Big]\\
\geq&\sum_{i=1}^{N}\tr^2(\mathbf{M}_{i})\left(1-\frac{1}{T}\right)+\underset{i\neq j}{\sum^{N}\sum^{N}}\left\{\tr(\mathbf{M}_{i}\mathbf{M}_{j})-\frac{1}{T}\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\right\}\\
&-(1-\frac{1}{T})|\gamma_{2}|\sum_{i=1}^{N}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
\geq&\left(1-\frac{1}{T}\right)\sum_{i=1}^{N}\left\{\tr^2(\mathbf{\Sigma})-2\tr(\mathbf{\Sigma})
\tr(\mathbf{B}_{i}\mathbf{\Sigma})\right\}+\underset{i\neq j}{\sum^{N}\sum^{N}}
\left\{\tr(\mathbf{\Sigma}^2)-2\tr(\mathbf{\Sigma}^2\mathbf{B}_{j})\right.\\
&-\left.2\tr(\mathbf{\Sigma}^2\mathbf{B}_{i})-2\tr(\mathbf{\Sigma}\mathbf{B}_{i}
\mathbf{B}_{j}\mathbf{\Sigma}\mathbf{B}_{j})-2\tr(\mathbf{\Sigma}\mathbf{B}_{j}
\mathbf{B}_{i}\mathbf{\Sigma}\mathbf{B}_{i})+2\tr(\mathbf{B}_{i}\mathbf{\Sigma}^{2}\mathbf{B}_{j})\right\}\\
&-\frac{1}{T}\underset{i\neq j}{\sum^{N}\sum^{N}}\left\{\tr^2(\mathbf{\Sigma})
+\tr(\mathbf{B}_{i}\mathbf{\Sigma})\tr(\mathbf{B}_{j}\mathbf{\Sigma})\right\}
-|\gamma_{2}|\sum_{i=1}^{N}\{\tr(\mathbf{\Sigma}^2)+\tr(\mathbf{B}_{i}\mathbf{\Sigma}\mathbf{B}_{i}\mathbf{\Sigma})\}\\
\geq &N\tr^2(\mathbf{\Sigma})-2pN\tr(\mathbf{\Sigma})C-\frac{N\tr^2(\mathbf{\Sigma})}{T}
+\underset{i\neq j}{\sum^{N}\sum^{N}}\left\{\tr(\mathbf{\Sigma}^2)-\frac{\tr^2(\mathbf{\Sigma})}{T}\right\}\\
&-10pC^2N(N-1)-\frac{p^2C^2N(N-1)}{T}-|\gamma_{2}|C^2NT-|\gamma_{2}|NpC^2\\
\geq&T^2\gamma_{N}\left(1-\frac{c_{1}}{T}\right),
\end{align*}
for some constant $c_{1}>0.$
Moreover, we have
\begin{align*}
{T^2} \gamma_{N}a_{N}&=\left[\sum_{i=1}^{N} \sum_{j=1}^{N} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}-\frac{1}{T} \mathbb{E}\left(\sum_{i=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\right]\\
=&\underset{i\neq j}{\sum^{N}\sum^{N}}\tr(\mathbf{M}_{i}\mathbf{M}_{j}) +\sum_{i=1}^{N}2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})
+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&-\frac{1}{T}\Big[\sum_{i=1}^{N}2\tr(\mathbf{M}^2_{i})
+\tr^2(\mathbf{M}_{i})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&\quad\quad\quad+\underset{i\neq j}{\sum^{N}\sum^{N}}\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\Big]\\
\leq&\underset{i\neq j}{\sum^{N}\sum^{N}}\left\{\tr(\mathbf{\Sigma}^2)+2\tr(\mathbf{\Sigma}\mathbf{B}_{i}
\mathbf{\Sigma}\mathbf{B}_{j})+2\tr(\mathbf{B}_{i}\mathbf{\Sigma}^2\mathbf{B}_{j})
+\tr(\mathbf{\Sigma}\mathbf{B}_{j}\mathbf{\Sigma}\mathbf{B}_{j})
+\tr(\mathbf{\Sigma}\mathbf{B}_{i}\mathbf{\Sigma}\mathbf{B}_{i})\right.\\
&\left.+\tr(\mathbf{B}_{i}
\mathbf{\Sigma}\mathbf{B}_{i}\mathbf{B}_{j}
\mathbf{\Sigma}\mathbf{B}_{j})-2\tr(\mathbf{\Sigma}\mathbf{B}_{i}
\mathbf{B}_{j}\mathbf{\Sigma}\mathbf{B}_{j})-2\tr(\mathbf{\Sigma}\mathbf{B}_{j}
\mathbf{B}_{i}\mathbf{\Sigma}\mathbf{B}_{i})\right\}+2C^2NT\\
&+\sum_{i=1}^{N}\left\{
\tr^2(\mathbf{\Sigma})+\tr^2(\mathbf{\Sigma}\mathbf{B}_{i})\right\}+(1-\frac{1}{T})|\gamma_{2}|\sum_{i=1}^{N}\{\tr(\mathbf{\Sigma}^{2})
+\tr(\mathbf{B}_{i}\mathbf{\Sigma}\mathbf{B}_{i}\mathbf{\Sigma})\}\\
\leq&C^2N(N-1)T+11pC^2N(N-1)+2C^2NT+N\tr^2(\mathbf{\Sigma})+p^2C^2N\\
&+|\gamma_{2}|C^2NT+|\gamma_{2}|NpC^2\\
\leq&T^2\gamma_{N}\left(1+\frac{c_{2}N}{T}\right),
\end{align*}
for some constant $c_{2}>0.$ This proves that $a_{N}$ satisfies the inequality in the lemma.
It remains to calculate $b_{N}.$ We have
\begin{align*}
\var\bigg\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\bigg\}
&=\mE\Bigg[\bigg\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\bigg\}^{2}\Bigg]
-\mE^2\bigg\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\bigg\}.
\end{align*}
Note that $i\neq j\neq k$ means that $i,j,k$ are not equal, and
$i\neq j\neq k\neq l$ means that $i,j,k,l$ are not equal.
First, by Lemma \ref{le:moment of quadratic form}, we have
\begin{align*}
&\mE\Bigg[\bigg\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\bigg\}^{2}\Bigg]\\
=&\sum_{i=1}^{N}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{4}\right\}+2\underset{i\neq j}{\sum^{N}\sum^{N}}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{4}\right\}+\underset{i\neq j}{\sum^{N}\sum^{N}}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\right\}\\
&+4\underset{i\neq j}{\sum^{N}\sum^{N}}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}
(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}\right\}+4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}
(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{k\cdot})^{2}\right\}\\
&+2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}
(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{k\cdot})^{2}\right\}+\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}
(\hat{\epsilon}_{k\cdot}^{\prime}\hat{\epsilon}_{l\cdot})^{2}\right\}\\
=&\sum_{i=1}^{N}\Big\{48\tr(\mathbf{M}^4_{i})+32\tr(\mathbf{M}^3_{i})\tr(\mathbf{M}_{i})+12\tr^2(\mathbf{M}^2_{i})
+12\tr(\mathbf{M}^2_{i})\tr^2(\mathbf{M}_{i})+\tr^4(\mathbf{M}_{i})\\
&+\gamma_{2} f_{\gamma_{2}}+\gamma_{4} f_{\gamma_{4}}+\gamma_{6} f_{\gamma_{6}}+\gamma_{1}^{2} f_{\gamma_{1}^{2}}+\gamma_{2}^{2} f_{\gamma_{2}^{2}}+\gamma_{1} \gamma_{3} f_{\gamma_{1} \gamma_{3}}\Big\}\\
&+2\underset{i\neq j}{\sum^{N}\sum^{N}}\Big[6\tr(\mathbf{M}_{i}\mathbf{M}_{j})^2+3\tr^2(\mathbf{M}_{i}\mathbf{M}_{j})+
3\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{M}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{M}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}\\
&+\gamma_{2}
\mE\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{P}_{j}\epsilon_{j\cdot}\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{P}_{j}\epsilon_{j\cdot}\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}
\Big]\\
&+\underset{i\neq j}{\sum^{N}\sum^{N}}\Big(\left[2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})
+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\right]\\
&\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\times
\left[2\tr(\mathbf{M}^2_{j})+\tr^2(\mathbf{M}_{j})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}\right]\Big)\\
&+4\underset{i\neq j}{\sum^{N}\sum^{N}}\Big[\gamma_{4} \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\\
&+2\gamma_{2} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \\
&+\gamma_{2} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\\
&+8 \gamma_{2} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right] \\
&+4 \gamma_{2} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right]\\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right] \\
&+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right] \\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\\
&+\tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+4 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \\
&+2 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+8 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\Big]\\
&+4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
\Big[2\tr(\mathbf{M}_{j}\mathbf{M}_{i}\mathbf{M}_{k}\mathbf{M}_{i})+\tr(\mathbf{M}_{j}\mathbf{M}_{i})\tr(\mathbf{M}_{k}\mathbf{M}_{i})\\
&+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})
\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{M}_{k}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\Big]\\
&+2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
\mE\left[2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\right]\tr(\mathbf{M}_{j}\mathbf{M}_{k})\\
&+\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}\left\{\tr(\mathbf{M}_{i}\mathbf{M}_{j})\tr(\mathbf{M}_{k}\mathbf{M}_{l})\right\}.
\end{align*}
Here, because the diagonal elements of $(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})_{ii}$ are greater than 0 and less than $C,$ we can obtain that
\begin{align*}
f_{\gamma_{2}}=&6 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\\
&+12{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}{\tau}_{T}\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \\
&+48\tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right\}\\
&+96\tr\left[\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right]\\
&+48{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right\}\left\{\mathbf{I}_{T} \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right\} {\tau}_{T}\\
\leq&6 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \right)^{2}\right\}+12\tr^{2}\left\{\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)^{2}\right\}\\
&+48\tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)
\sqrt{\tr\left\{\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)^{2}\right\}
\tr\left\{\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)^{4}\right\}}\\
&+96C\tr\left\{\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)^{3}\right\}
+48\tr^{2}\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
=&O(T^3),\\
f_{\gamma_{4}}=& 4\tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \\
&+24\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right\}=O(T^2),\\
f_{\gamma_{6}}=& \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}=O(T),\\
f_{\gamma_{1}^{2}}=& 24{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}{\tau}_{T} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+48{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\tau_{T}\\
&+16{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}{\tau}_{T} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+96{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \tau_{T}\\
&+96\tr\left[\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right]\\
\leq&24C^2\sum_{s=1}^{T}\sum_{t=1}^{T}|\left(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)_{st}|
\tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+48C^{2}\sum_{s=1}^{T}\sum_{t=1}^{T}|(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})|_{st}\\
&+16\sum_{s=1}^{T}\sum_{t=1}^{T}|(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})|_{st}^{3} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+96\sum_{s=1}^{T}\sum_{t=1}^{T}\sum_{k=1}^{T}\left(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)^{2}_{sk}\left(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)_{kt}\left(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)_{tt}\\
&+96\sum_{s=1}^{T}\sum_{t=1}^{T}
\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \right)^{2}_{st}\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \right)_{ts}=O(T^3),\\
f_{\gamma_{2}^{2}}=& 3\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\\
&+24{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}{\tau}_{T}\\
&+8{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\\
\leq&3\tr^2\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+24\sum_{s=1}^{T}\sum_{t=1}^{T}\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2}\right)_{st}^{2}\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2}\right)_{ss}\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2}\right)_{tt}\\
&+8\sum_{s=1}^{T}\sum_{t=1}^{T}\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2}\right)_{st}^{4}\\
=&O(T^2),\\
f_{\gamma_{1} \gamma_{3}}=& 24{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}{\tau}_{T}\\
&+32{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}{\tau}_{T}\\
\leq&24\sum_{s=1}^{T}\sum_{t=1}^{T}
\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)_{st}
\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)_{ss}
\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)_{tt}^{2}\\
&+32\sum_{s=1}^{T}\sum_{t=1}^{T}\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)_{st}^{3}
\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)_{ss}\\
=&O(T^2),
\end{align*}
Then, by Lemma \ref{le:moment of quadratic form},
\begin{align*}
&\mE^2\Big\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\Big\}\\
=&
\left[\sum_{1\leq i\neq j\leq N}\tr(\mathbf{M}_{i}\mathbf{M}_{j}) +\sum_{i=1}^{N}2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\right]^2\\
=&\sum_{i=1}^{N}\Big[4\tr^2(\mathbf{M}^2_{i})+\tr^4(\mathbf{M}_{i})+4\tr(\mathbf{M}^2_{i})\tr^2(\mathbf{M}_{i})
+\gamma_{2}^2\tr^2\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&+4\gamma_{2}\tr(\mathbf{M}^2_{i})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}+2\gamma_{2}\tr^2(\mathbf{M}_{i})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\Big]\\
+&\underset{i\neq j}{\sum^{N}\sum^{N}}\Big[4\tr(\mathbf{M}^2_{i})\tr(\mathbf{M}^2_{j})+2\tr(\mathbf{M}^2_{i})\tr^2(\mathbf{M}_{j})
+2\tr(\mathbf{M}^2_{j})\tr^2(\mathbf{M}_{i})+\tr^2(\mathbf{M}_{i})\tr^2(\mathbf{M}_{j})\\
&+\gamma_{2}^2\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})
\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}\\
&+2\gamma_{2}\tr(\mathbf{M}^2_{i})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})
\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}
+2\gamma_{2}\tr\{(\mathbf{M}^2_{j})\tr(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})
\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&+\gamma_{2}\tr^2\{(\mathbf{M}_{i})\tr(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}+\gamma_{2}\tr^2(\mathbf{M}_{j})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})
\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\Big]\\
&+\underset{i\neq j}{\sum^{N}\sum^{N}}\Big[2\tr^2(\mathbf{M}_{i}\mathbf{M}_{j})+8\tr(\mathbf{M}^2_{i})\tr(\mathbf{M}_{i}\mathbf{M}_{j})
+4\tr(\mathbf{M}_{i}\mathbf{M}_{j})\tr^2(\mathbf{M}_{i})\\
&+4\gamma_{2}\tr(\mathbf{M}_{i}\mathbf{M}_{j})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\Big]\\
+&\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}\Big[4\tr(\mathbf{M}_{i}\mathbf{M}_{j})\tr(\mathbf{M}_{i}\mathbf{M}_{k})
+4\tr(\mathbf{M}_{i}\mathbf{M}_{j})\tr(\mathbf{M}^2_{k})+2\tr(\mathbf{M}_{i}\mathbf{M}_{j})\tr^2(\mathbf{M}_{k})\\
&+2\gamma_{2}\tr(\mathbf{M}_{i}\mathbf{M}_{j})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{k}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{k}\mathbf{\Sigma}^{1/2})\}\Big]\\
&+\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}\left\{\tr(\mathbf{M}_{i}\mathbf{M}_{j})\tr(\mathbf{M}_{k}\mathbf{M}_{l})\right\}.
\end{align*}
So, we have
\begin{align*}
\var\left\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\right\}
=&O\left(NT^3+N^2T^2+N^2T+N^3T\right).
\end{align*}
Similarly, we can obtain that
\begin{align*}
&\mE\Big(\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\Big)^{4}\\
=&\mE\Big(\sum_{i=1}^{N}\sum_{j=1}^{N}\sum_{k=1}^{N}\sum_{l=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}
\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{k\cdot}^{\prime}\hat{\epsilon}_{k\cdot}
\hat{\epsilon}_{l\cdot}^{\prime}\hat{\epsilon}_{l\cdot}\Big)\\
=&\sum_{i=1}^{N}\mE\left(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{4}
+3\underset{i \neq j}{\sum^{N}\sum^{N}}\mE(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^2(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^2
+4\underset{i \neq j}{\sum^{N}\sum^{N}}\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^3\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\}\\
&+6\underset{i \neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^2
\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{k\cdot}^{\prime}\hat{\epsilon}_{k\cdot}\}
+\underset{i \neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
\mE(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}
\hat{\epsilon}_{k\cdot}^{\prime}\hat{\epsilon}_{k\cdot}\hat{\epsilon}_{l\cdot}^{\prime}\hat{\epsilon}_{l\cdot})\\
=&\sum_{i=1}^{N}\Big\{48\tr(\mathbf{M}^4_{i})+32\tr(\mathbf{M}^3_{i})\tr(\mathbf{M}_{i})+12\tr^2(\mathbf{M}^2_{i})
+12\tr(\mathbf{M}^2_{i})\tr^2(\mathbf{M}_{i})
+\tr^4(\mathbf{M}_{i})\\
&+\gamma_{2} f_{\gamma_{2}}+\gamma_{4} f_{\gamma_{4}}+\gamma_{6} f_{\gamma_{6}}+\gamma_{1}^{2} f_{\gamma_{1}^{2}}+\gamma_{2}^{2} f_{\gamma_{2}^{2}}+\gamma_{1} \gamma_{3} f_{\gamma_{1} \gamma_{3}}\Big\}\\
&+3\underset{i \neq j}{\sum^{N}\sum^{N}}\Big(\left[2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})
+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(
\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\right]\\
&\,\,\,\,\,\,\,\,\,\,\,\,\times
\left[2\tr(\mathbf{M}^2_{j})+\tr^2(\mathbf{M}_{j})
+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ
(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}\right]\Big)
\\&+4\underset{i\neq j}{\sum^{N}\sum^{N}}\Big(
\gamma_{4} \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\\
&+3\gamma_{2} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \\
&+12\gamma_{2} \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right\} \\
&+6 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{(\mathbf{I}_{T} )\circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\\
& +4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\\
&+\tr^3\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)+6 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+8 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\Big)\tr(\mathbf{M}_{j})\\
&+6\underset{i \neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}[2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})
+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ
(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}]
\tr(\mathbf{M}_{j})\tr(\mathbf{M}_{k})\\&+\underset{i \neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})
\tr(\mathbf{M}_{k})\tr(\mathbf{M}_{l}),
\end{align*}
where $f_{\gamma_{2}},\,f_{\gamma_{4}},\, f_{\gamma_{6}},\, f_{\gamma_{1}^{2}},\, f_{\gamma_{2}^{2}},\,f_{\gamma_{1} \gamma_{3}}$ are the same as above,
and
\begin{align*}
&\Big\{\mE\Big(\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\Big)^{2}\Big\}^2\\
=&\Big[\sum_{i=1}^{N}2\tr(\mathbf{M}^2_{i})+\tr^2(\mathbf{M}_{i})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}+\underset{i\neq j}{\sum^{N}\sum^{N}}\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\Big]^2\\
=&\sum_{i=1}^{N}\Big[4\tr^2(\mathbf{M}^2_{i})+\tr^4(\mathbf{M}_{i})+4\tr(\mathbf{M}^2_{i})\tr^2(\mathbf{M}_{i})
+\gamma_{2}^2\tr^2\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&+4\gamma_{2}\tr(\mathbf{M}^2_{i})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}+2\gamma_{2}\tr^2(\mathbf{M}_{i})
\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})
\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\Big]\\
+&\underset{i\neq j}{\sum^{N}\sum^{N}}\Big[4\tr(\mathbf{M}^2_{i})\tr(\mathbf{M}^2_{j})+2\tr(\mathbf{M}^2_{i})\tr^2(\mathbf{M}_{j})
+2\tr(\mathbf{M}^2_{j})\tr^2(\mathbf{M}_{i})+\tr^2(\mathbf{M}_{i})\tr^2(\mathbf{M}_{j})\\
&+\gamma_{2}^2\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}
\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})
\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}\\
&+2\gamma_{2}\tr(\mathbf{M}^2_{i})
\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})
\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}
+2\gamma_{2}\tr(\mathbf{M}^2_{j})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\\
&+\gamma_{2}\tr^2\{(\mathbf{M}_{i})\tr(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}+\gamma_{2}\tr^2(\mathbf{M}_{j})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\Big]\\
&+\underset{i\neq j}{\sum^{N}\sum^{N}}\Big[2\tr^2(\mathbf{M}_{i})\tr^2(\mathbf{M}_{j})+8\tr(\mathbf{M}^2_{i})\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})
+4\tr^3(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\\
&+4\gamma_{2}\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\}\Big]\\
+&\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}\Big[4\tr^2(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\tr(\mathbf{M}_{k})
+4\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\tr(\mathbf{M}^2_{k})+2\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\tr^2(\mathbf{M}_{k})\\
&+2\gamma_{2}\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{k}\mathbf{\Sigma}^{1/2}
)\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{k}\mathbf{\Sigma}^{1/2})\}\Big]\\
&+\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}\left\{\tr(\mathbf{M}_{i})\tr(\mathbf{M}_{j})\tr(\mathbf{M}_{k})\tr(\mathbf{M}_{l})\right\}.
\end{align*}
So, we have
\begin{align*}
\var\bigg\{\Big(\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\Big)^{2}\bigg\}=O(NT^3+N^2T^3+N^3T^3).
\end{align*}
Finally, we have
\begin{align*}
\mE(b^2_{N})\leq&\frac{1}{N^2T^4}\Bigg[2\var\bigg\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\bigg\}+\frac{2}{T^2}
\var\bigg\{\Big(\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\Big)^{2}\bigg\}\Bigg]\\
=&O\left(\frac{1}{NT}\right).
\end{align*}
\hfill$\Box$
\subsection{Proof of Lemma \ref{H1:lemax1}}
Recall that $\mathbf{E}=\mathbf{U}^{1/2}\mathbf{Z}\mathbf{\Sigma}^{1/2},$
then we have $$\hat{\epsilon}_{li}\hat{\epsilon}_{lj}={Z}^{\prime}\left[\{u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\}\otimes \{m_{i}^{l}(m_{j}^{l})^{\prime}\}\right]{Z},$$ where
$Z=(z_{11},\cdots,z_{1T},z_{21},\cdots,z_{2T},\cdots,z_{N1},\cdots,z_{NT})^{\prime}\in
\mathbb{R}^{NT},$
$u_{l}^{1/2}\in \mathbb{R}^{N}$ denotes the $l$-th row vector of matrix $\mathbf{U}^{1/2},$ and $m_{i}^{l} \in \mathbb{R}^{T}$ represents the $i$-th row vector of matrix $\mathbf{P}_{l}\mathbf{\Sigma}^{1/2}.$
Hence, $\sum_{l=1}^{N}\hat{\epsilon}_{li}\hat{\epsilon}_{lj}={Z}^{\prime}\left[\sum_{l=1}^{N}\{u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\}\otimes \{m_{i}^{l}(m_{j}^{l})^{\prime}\}\right]{Z}.$
To simplify notation, we write $\mathbf{W}_{ij}$ as matrix $\sum_{l=1}^{N}\{u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\}\otimes \{m_{i}^{l}(m_{j}^{l})^{\prime}\}.$
Recall that $||\mathbf{A}||$ denotes the operator norm of matrix $\mathbf{A}.$
So, by the Hanson-Wright inequality in \cite{rudelson2013},
we can obtain that for every $t>0$,
$$
\mathbb{P}\left\{\left|{Z}^{\prime}\mathbf{W}_{ij}{Z}-\mathbb{E} \big({Z}^{\prime}\mathbf{W}_{ij}{Z}\big)\right|>t\right\}
\leq 2 \exp \left[-c \min \left(\frac{t^{2}}{K^{4}\|\mathbf{W}_{ij}\|_{\mathrm{F}}^{2}}, \frac{t}{K^{2}\|\mathbf{W}_{ij}\|}\right)\right].
$$
Obviously,
$\mathbb{E} \big({Z}^{\prime}\mathbf{W}_{ij}{Z}\big)=\tr(\mathbf{W}_{ij})
=\sum_{l=1}^{N}u_{ll}\sigma_{ij,l},$ where $u_{ll}=(\mathbf{U})_{ll},$ for $1\leq l\leq N.$
Recall that we assume that $u_{ll}=1,$
for $1\leq l\leq N.$
According to the properties of the Kronecker product,
we have
$$\tr\{\mathbf{W}_{ij}(\mathbf{W}_{ij})^{\prime}\}
=\underset{l,s=1}{\sum^{N}\sum^{N}}u_{ls}^2 (\mathbf{P}_{l}\mathbf{\Sigma}\mathbf{P}_{s})_{ii}
(\mathbf{P}_{l}\mathbf{\Sigma}\mathbf{P}_{s})_{jj}.$$
Due to $(\mathbf{P}_{l}\mathbf{\Sigma}\mathbf{P}_{s})_{jj}
=\sigma_{jj}-(\mathbf{B}_{l}\mathbf{\Sigma})_{jj}-(\mathbf{B}_{s}\mathbf{\Sigma})_{jj}-(\mathbf{B}_{l}
\mathbf{\Sigma}\mathbf{B}_{s})_{jj}$ and (\ref{sigmaij,l-sigmaij}),
we have $\tr\{\mathbf{W}_{ij}(\mathbf{W}_{ij})^{\prime}\}=\tr(\mathbf{U}^2)\sigma_{ii}\sigma_{jj}\{1+o(1)\}$
and $\tr\{\mathbf{W}_{ij}\mathbf{W}_{ij}\}=\tr(\mathbf{U}^2)\sigma_{ij}^{2}\{1+o(1)\}.$ Then, we focus on $\|\mathbf{W}_{ij}\|.$
According to the triangle inequality of norms, we have
\begin{align*}
\|\mathbf{W}_{ij}\|&\leq \|\sum_{l=1}^{N}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes \sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\|+\|\sum_{l=1}^{N}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes \big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\|\\\n
&\leq \sqrt{\lambda_{max}(\mathbf{U}^{2})\sigma_{ii}\sigma_{jj}}+\|\sum_{l=1}^{N}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes \big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\|,
\end{align*}
where $\sigma_{i}^{1/2}\in \mathbb{R}^{N}$ denotes the $i$-th row vector of matrix $\mathbf{\Sigma}^{1/2}.$
Then, using the definition of spectral norm, we can obtain that for any vector $x \in \mathbb{R}^{NT}$ with norm $\|x\|=1,$
\begin{align*}
&\|\sum_{l=1}^{N}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes \big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\|^2\\\n
%=&\max_{\|x\|=1}x^{\prime}\Big[\sum_{s=1}^{N}u_{s}^{1/2}(u_{s}^{1/2})^{\prime}\otimes \big\{m_{j}^{s}(m_{i}^{s})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}\Big]
%\Big[\sum_{l=1}^{N}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes \big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]x\\\n
=&\max_{\|x\|=1}x^{\prime}\Big[\sum_{s=1}^{N}\sum_{l=1}^{N}u_{sl}u_{s}^{1/2}(u_{l}^{1/2})^{\prime}\otimes \big\{m_{j}^{s}(m_{i}^{s})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]x\\\n
\leq&\max_{\|x\|=1}x^{\prime}\Big[\sum_{l=1}^{N}u_{ll}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]x\\\n
&+\max_{\|x\|=1}x^{\prime}\Big[\underset{l\neq s}{\sum^{N}\sum^{N}}u_{ls}u_{l}^{1/2}(u_{s}^{1/2})^{\prime}\otimes\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{s}(m_{j}^{s})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]x\\\n
=&\Delta_{1}+\Delta_{2}.
\end{align*}
Next, we first calculate $\Delta_{1},$
\begin{align*}
\Delta_{1}=&\max_{\|x\|=1}x^{\prime}\Big[\sum_{l=1}^{N}u_{ll}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]x\\\n
\leq & \tr\Big[\sum_{l=1}^{N}u_{ll}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]\\
\leq &\sum_{l=1}^{N}\tr\Big[u_{ll}u_{l}^{1/2}(u_{l}^{1/2})^{\prime}\otimes\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]\\
\leq&\sum_{l=1}^{N}u_{ll}^2\tr\Big[\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{l}(m_{j}^{l})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]\\
\leq &\sum_{l=1}^{N}u_{ll}^{2}\Big\{\sigma_{ii,l}\sigma_{jj,l}-2
(\mathbf{P}_{l}\mathbf{\Sigma})_{ii}(\mathbf{P}_{l}\mathbf{\Sigma})_{jj}+\sigma_{ii}\sigma_{jj}\Big\}\\
=&O(T^{0\vee(1-\frac{1}{\tau})})
\end{align*}
where the last inequality holds due to (\ref{sigmaij,l-sigmaij}). Similarly, we have
\begin{align*}
\Delta_{2}=&\max_{\|x\|=1}x^{\prime}\Big[\underset{l\neq s}{\sum^{N}\sum^{N}}u_{ls}u_{l}^{1/2}(u_{s}^{1/2})^{\prime}\otimes\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{s}(m_{j}^{s})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]x\\
\leq &\tr\Big[\underset{l\neq s}{\sum^{N}\sum^{N}}u_{ls}u_{l}^{1/2}(u_{s}^{1/2})^{\prime}\otimes\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{s}(m_{j}^{s})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]\\
\leq&\underset{l\neq s}{\sum^{N}\sum^{N}}\tr\Big[u_{ls}u_{l}^{1/2}(u_{s}^{1/2})^{\prime}\otimes\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{s}(m_{j}^{s})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]\\
\leq&\underset{l\neq s}{\sum^{N}\sum^{N}}u_{ls}^2\tr\Big[\big\{m_{j}^{l}(m_{i}^{l})^{\prime}-\sigma_{j}^{1/2}(\sigma_{i}^{1/2})^{\prime}\big\}
\big\{m_{i}^{s}(m_{j}^{s})^{\prime}-\sigma_{i}^{1/2}(\sigma_{j}^{1/2})^{\prime}\big\}\Big]\\
\leq&\underset{l\neq s}{\sum^{N}\sum^{N}}u_{ls}^2\Big\{(\mathbf{P}_{l}\mathbf{\Sigma}\mathbf{P}_{s})_{ii}
(\mathbf{P}_{l}\mathbf{\Sigma}\mathbf{P}_{s})_{jj}-(\mathbf{P}_{l}\mathbf{\Sigma})_{ii}
(\mathbf{P}_{l}\mathbf{\Sigma})_{jj}-(\mathbf{P}_{s}\mathbf{\Sigma})_{jj}
(\mathbf{P}_{s}\mathbf{\Sigma})_{jj}+\sigma_{ii}\sigma_{jj}\Big\}\\
=&O(T^{0\vee(1-\frac{1}{\tau})}).
\end{align*}
So, we can conclude that $\|\mathbf{W}_{ij}\|=O(T^{0\vee(\frac{1}{2}-\frac{1}{2\tau})}).$
Then, according to Theorem 1.1 in \cite{rudelson2013},
for every $t\in (0,o(N^{\frac{1}{2}\wedge \frac{1}{2\tau}})),$
there exist a constant $c>0,$
\begin{align}\label{H1:ehatlilj}
&\mathbb{P}\left(\left|\sum_{l=1}^{N}\hat{\epsilon}_{li}\hat{\epsilon}_{lj
}-\sum_{l=1}^{N}u_{ll}\sigma_{ij,l}\right|>t\sqrt{\tr(\mathbf{U}^{2})
(\sigma_{ii}\sigma_{jj}+\sigma_{ij}^{2})}\right)\\\n
\leq&2\exp\Big(-c\frac{t^2\tr(\mathbf{U}^{2})(\sigma_{ii}\sigma_{jj}+\sigma_{ij}^{2})}
{K^{4}\tr\{\mathbf{W}_{ij}(\mathbf{W}_{ij})^{\prime}\}}\Big)\\\n
\leq&2\exp\Big(-c^{\prime}\frac{t^2}
{K^{4}}\Big)
\end{align}
where $\tr\{\mathbf{W}_{ij}(\mathbf{W}_{ij})^{\prime}\}=\tr(\mathbf{U}^2)\sigma_{ii}\sigma_{jj}\{1+o(1)\},$
and $c^{\prime}>0$ is a positive constant independent of $N$ and $ T.$
In fact, $\bar{\hat{\epsilon}}_{\cdot j}=\sum_{l=1}^{N}\hat{\epsilon}_{li}/N$ can be seen as a linear combination of $z_{ks},$ $1\leq k,s\leq N,$ where $z_{ks}$ is the element at row $k$ and column $s$ of random matrix $\mathbf{Z}.$ Then, by classical Cram\'{e}r type large deviation results for independent random variables(see Corollary 3.1 in \cite{saulis1991limit}), we have for any $\varepsilon > 0,$
$$\mathbb{P}\left(|\bar{\hat{\epsilon}}_{\cdot j}|>x\sqrt{\var(\bar{\hat{\epsilon}}_{\cdot j})}\right)\leq C \exp\left\{-\frac{x^2}{2}(1-\varepsilon)\right\},$$
uniformly in $x \in[0, o(\sqrt{N})).$
For $\bar{\hat{\epsilon}}_{\cdot j},$ we have
$$\var(\bar{\hat{\epsilon}}_{\cdot j})=\frac{1}{N^2}\sum_{l=1}^{N}u_{ll}\sigma_{ii,l}
+\frac{1}{N^2}\underset{l\neq s}{\sum^{N}\sum^{N}}u_{ls}(\mathbf{P}_{l}\mathbf{\Sigma}\mathbf{P}_{s})_{ii}=O(N^{-1\vee-\frac{1}{\tau}}),$$
uniformly in $1 \leq j \leq N. $
So, for any $\varepsilon > 0,$
$$
\mathbb{P}\left(\left|\bar{\hat{\epsilon}}_{\cdot j}\bar{\hat{\epsilon}}_{\cdot i}\right| \geq x^{2} \sqrt{\var\left(\bar{\hat{\epsilon}}_{\cdot j}\right) \var\left(\bar{\hat{\epsilon}}_{\cdot i}\right)}\right) \leq 2C \exp\left\{-\frac{x^2}{2}(1-\varepsilon)\right\}
$$
uniformly in $x\in[0, o(\sqrt{N}))$. We have, uniformly for $x \in[0, o(N^{\frac{1}{2} \wedge\left(\frac{1}{\tau}-\frac{1}{2}\right)})),$ $x^{2} \sqrt{\var\left(\bar{\hat{\epsilon}}_{\cdot i}\right) \var\left(\bar{\hat{\epsilon}}_{\cdot j}\right)}=$ $o(x / \sqrt{N})$. So for any
$\varepsilon>0,$ large $N$, and any $\delta>0,$
\begin{align}\label{H1:ehatbarij}
\mathbb{P}\left(\left|\bar{\hat{\epsilon}}_{\cdot i} \bar{\hat{\epsilon}}_{\cdot j}\right| \geq \delta \frac{x}{\sqrt{N}}\right) \leq 2C \exp\left\{-\frac{x^2}{2}(1-\varepsilon)\right\},
\end{align}
uniformly and $x\in[0, o(N^{\frac{1}{2} \wedge\left(\frac{1}{\tau}-\frac{1}{2}\right)})).$
Then, the lemma follows from (\ref{H1:ehatlilj}) and (\ref{H1:ehatbarij}).\hfill$\Box$
\subsection{Proof of Lemma \ref{H1:lemax2}}
Recall that without loss of generality, we assume that $u_{ll}=1,$
for $1\leq l\leq N.$
Similarly, we have
\begin{align*}
\|\hat{\mathbf{\Gamma}}\|_{\mathrm{F}}^{2}&=\frac{1}{T^2}\sum_{i=1}^{N} \sum_{j=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot}^{2}\\
\{\tr(\hat{\mathbf{\Gamma}})\}^{2}&=\frac{1}{T^2}\left(\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\\
T^2a_{N}&=\frac{1}{\gamma_{N}}\left[\sum_{i=1}^{N} \sum_{j=1}^{N} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}-\frac{1}{T} \mathbb{E}\left(\sum_{i=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\right]\\
T^2b_{N}&=\frac{1}{N}\left[\sum_{i=1} \sum_{j=1} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}-\frac{1}{T}\left(\sum_{i=1} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}-\sum_{i=1} \sum_{j=1} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}+\frac{1}{T} \mathbb{E}\left(\sum_{i=1} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\right]
\end{align*}
it is easy to verify that $a_{N}$ and $b_{N}$ will make the equation (\ref{H1:gamman}) true.
In the following, we prove that $a_{N}, b_{N}$ satisfy the properties in the lemma.
We first deal with the term $a_{N} .$
Recall that $Z=(z_{11},\cdots,z_{1T},z_{21},\cdots,z_{2T},\cdots,z_{N1},\cdots,z_{NT})^{\prime}$
and $u_{l}^{1/2}\in \mathbb{R}^{N}$ denotes the $l$-th row vector of matrix $\mathbf{U}^{1/2},$
we have
\begin{equation*}
(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})={Z}^{\prime}
\left[\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\otimes(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\right]Z.
\end{equation*}
Therefore, due to Lemma \ref{le:moment of quadratic form},
\begin{align*}
\sum_{1 \leq i,j \leq N} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2} =&\sum_{1\leq i\neq j\leq N}u_{ij}^2\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})^2\}+u_{ij}^2\tr^2(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\\
&+ u_{ii}u_{jj}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})^{\prime}\}\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\right\}\\
&+\sum_{i=1}^{N}2u_{ii}^{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})^{2}\}+u_{ii}^{2}\tr^{2}(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\right\}.
\end{align*} Moreover,
\begin{align*}
\mathbb{E}\left(\sum_{i=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2} =&\sum_{i=1}^{N} \sum_{j=1}^{N} \mathbb{E}( \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot} \hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})\\
=&\sum_{i=1}^{N}\mathbb{E}( \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})+\underset{i\neq j}{\sum^{N}\sum^{N}}\mathbb{E}( \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot} \hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})\\
=&\sum_{i=1}^{N}2u_{ii}^{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})^2\}+u_{ii}^{2}\tr^2(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\right\}\\
&+\underset{i\neq j}{\sum^{N}\sum^{N}}2u_{ij}^{2}\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})+u_{ii}u_{jj}\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ
\{u_{j}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\right\}.
\end{align*}
Similarly, due to $\sum_{1\leq i\neq j\leq N}\tr\left[\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]=O(N),$ we have
\begin{align*}
{T^2} \gamma_{N}a_{N}&=\left[\sum_{i=1}^{N} \sum_{j=1}^{N} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}-\frac{1}{T} \mathbb{E}\left(\sum_{i=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\right]\\
=&\sum_{1\leq i\neq j\leq N}u_{ij}^2\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})^2\}+u_{ij}^2\tr^2(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\\
&+ u_{ii}u_{jj}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})^{\prime}\}\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\right\}\\
&+\sum_{i=1}^{N}2u_{ii}^{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})^{2}\}+u_{ii}^{2}\tr^{2}(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\right\}\\
&-\frac{1}{T}\Big[\sum_{i=1}^{N}2u_{ii}^{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})^2\}+u_{ii}^{2}\tr^2(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\right\}\\
&+\underset{i\neq j}{\sum^{N}\sum^{N}}2u_{ij}^{2}\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})+u_{ii}u_{jj}\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ
\{u_{j}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\right\}\Big]\\
\geq&\tr(\mathbf{U}^2)\tr^2(\mathbf{\Sigma})\{1-O(T^{-1})\}.
\end{align*}
Moreover, we have
\begin{align*}
{T^2} \gamma_{N}a_{N}&=\left[\sum_{i=1}^{N} \sum_{j=1}^{N} \mathbb{E} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}-\frac{1}{T} \mathbb{E}\left(\sum_{i=1}^{N} \hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^{2}\right]\\
=&\sum_{1\leq i\neq j\leq N}u_{ij}^2\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})^2\}+u_{ij}^2\tr^2(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\\
&+ u_{ii}u_{jj}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})^{\prime}\}\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\circ\{
u_{i}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\right\}\\
&+\sum_{i=1}^{N}2u_{ii}^{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})^{2}\}+u_{ii}^{2}\tr^{2}(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\right\}\\
&-\frac{1}{T}\Big[\sum_{i=1}^{N}2u_{ii}^{2}\tr\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})^2\}+u_{ii}^{2}\tr^2(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ
\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}
\mathbf{\Sigma}^{1/2})\right\}\\
&+\underset{i\neq j}{\sum^{N}\sum^{N}}2u_{ij}^{2}\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})+u_{ii}u_{jj}\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\tr(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\\
&+\gamma_{2}\tr\left[\{u_{i}^{1/2}(u_{i}^{1/2})^{\prime}\}\circ\{
u_{j}^{1/2}(u_{j}^{1/2})^{\prime}\}\right]\tr\left\{(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\right\}\Big]\\
\leq&\tr(\mathbf{U}^2)\tr^2(\mathbf{\Sigma})\{1+O(NT^{-1})\}.
\end{align*}
This proves that $a_{N}$ satisfies the inequality in the lemma.
It remains to calculate $b_{N}.$ We have
\begin{align*}
\var\left\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\right\}
&=\mE\left[\left\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\right\}^{2}\right]
-\mE^2\left\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\right\}.
\end{align*}
First,
\begin{align*}
&\var\left[\left\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\right\}^{2}\right]\\
=&\sum_{i=1}^{N}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{4}\right\}
-\mE^2\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}\}+2\underset{i\neq j}{\sum^{N}\sum^{N}}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{4}\right\}
-\mE^2\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\}\\
&+\underset{i\neq j}{\sum^{N}\sum^{N}}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}
(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\right\}
-\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}\}
\mE\{(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\}\\
&+4\underset{i\neq j}{\sum^{N}\sum^{N}}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}
(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}\right\}
-\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}\}
\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\}\\
&+2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}
(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{k\cdot})^{2}\right\}
-\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot})^{2}\}
\mE\{(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{k\cdot})^{2}\}\\
&+4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}
(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{k\cdot})^{2}\right\}
-\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\}
\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{k\cdot})^{2}\}\\
&+\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}\mE\left\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}
(\hat{\epsilon}_{k\cdot}^{\prime}\hat{\epsilon}_{l\cdot})^{2}\right\}
-\mE\{(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\}
\mE\{(\hat{\epsilon}_{k\cdot}^{\prime}\hat{\epsilon}_{l\cdot})^{2}\}\\
=&G_{1}+G_{2}+G_{3}+G_{4}+G_{5}+G_{6}+G_{7}.
\end{align*}
Recall that $\mathbf{Q}_{ij}=\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{P}_{j}
\mathbf{\Sigma}^{1/2}.$ Let $\mathbf{\Omega}_{ij}\doteq u_{i}^{1/2}(u_{j}^{1/2})^{\prime}$ for $1\leq i,j\leq N.$
First, we focus on $G_{1}.$
According to Lemma \ref{le:moment of quadratic form}, we have
$$
G_{1}=G_{1_{1}}+\gamma_{2}G_{1_{2}}+\gamma_{4}G_{1_{3}}+\gamma_{6}G_{1_{4}}
+\gamma_{1}^{2}G_{1_{5}}+\gamma_{2}^{2}G_{1_{6}}+\gamma_{1}\gamma_{3}G_{1_{7}},$$
where
\begin{align*}
G_{1_{1}}=&\sum_{i=1}^{N}8u_{ii}^{4}
\tr\left(\mathbf{Q}^{2}_{ii}\right)\tr^{2}\left(\mathbf{Q}_{ii}\right)
+8u_{ii}^{4}\tr^2\left(\mathbf{Q}_{ii}^{2}\right)
+32u_{ii}^{4}\tr\left(\mathbf{Q}_{ii}\right)\tr\left(\mathbf{Q}^{3}_{ii}\right)\\
&+48u_{ii}^{4}\tr\left(\mathbf{Q}_{ii}^{4}\right),\\
G_{1_{2}}=&\sum_{i=1}^{N}
4u_{ii}^{2}\tr^2\left(\mathbf{Q}_{ii}\right)
\tr\left\{\mathbf{\Omega}_{ii}\circ \mathbf{\Omega}_{ii}\right\}
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\\
&+8u_{ii}^{2}
\tr\left(\mathbf{Q}_{ii}^{2}\right)\tr\left\{\mathbf{\Omega}_{ii}\circ \mathbf{\Omega}_{ii}\right\}
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\\
&+48u_{ii}
\tr\left(\mathbf{Q}_{ii}\right)\tr\left\{\mathbf{\Omega}_{ii}\circ u_{ii}\mathbf{\Omega}_{ii}\right\}\tr\left(\mathbf{Q}_{ii}\circ \mathbf{Q}_{ii}^{2}\right)\\
&+96\tr\left\{\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)u_{ii}^{2}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}^{3}\right\}\\
&+48\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}^{2}\right)\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}^{2}\right)\right\}\tau_{T},\\
G_{1_{3}}=&\sum_{i=1}^{N}
4u_{ii}\tr\left(\mathbf{Q}_{ii}\right)\tr\left\{\mathbf{\Omega}_{ii}\circ \mathbf{\Omega}_{ii}\circ \mathbf{\Omega}_{ii}\right\}\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\\
&+24\tr\left\{\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\circ
u_{ii}\mathbf{\Omega}_{ii}\right\}\tr(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}^{2}),\\
G_{1_{4}}=&\sum_{i=1}^{N}
\tr\left\{\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right\}\tr\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right),\\
G_{1_{5}}=&\sum_{i=1}^{N}
24\tau_{N}^{\prime}\left[\left\{\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right\}\mathbf{\Omega}_{ii}
\left\{\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right\}\right]\tau_{N}
\tau_{T}^{\prime}\left[\left\{\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right\}\mathbf{Q}_{ii}
\left\{\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right\}\right]\tau_{T}u_{ii}\tr(\mathbf{Q}_{ii})\\
&+48\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ii}\mathbf{\Omega}_{ii}\left\{\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right\}\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}^{2}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\\
&+16\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}
\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}
\circ\mathbf{Q}_{ii}\right)\tau_{T}u_{ii}\tr\left(\mathbf{Q}_{ii}\right)\\
&+96\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}\prime\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&+96\tr\left\{\mathbf{\Omega}_{ii}\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ii}\right\}
\tr\left\{\mathbf{Q}_{ii}\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\right\},\\
G_{1_{6}}=&\sum_{i=1}^{N}
2\tr^{2}\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)\tr^{2}\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\\
&+24\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&+8\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}
\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}
\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T},\\
G_{1_{7}}=&\sum_{i=1}^{N}24\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&+32\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}.
\end{align*}
Obviously, $G_{1_{1}}=O(NT^{3}).$
Due to Lemma \ref{lem:matrix inequality} (1),
we have
\begin{align*}
G_{1_{2}}=&\sum_{i=1}^{N}
4u_{ii}^{2}\tr^2\left(\mathbf{Q}_{ii}\right)
\tr\left\{\mathbf{\Omega}_{ii}\circ \mathbf{\Omega}_{ii}\right\}
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\\
&+8u_{ii}^{2}
\tr\left(\mathbf{Q}_{ii}^{2}\right)\tr\left\{\mathbf{\Omega}_{ii}\circ \mathbf{\Omega}_{ii}\right\}
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\\
&+48u_{ii}
\tr\left(\mathbf{Q}_{ii}\right)\tr\left\{\mathbf{\Omega}_{ii}\circ u_{ii}\mathbf{\Omega}_{ii}\right\}\tr\left(\mathbf{Q}_{ii}\circ \mathbf{Q}_{ii}^{2}\right)\\
&+96\tr\left\{\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)u_{ii}^{2}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}^{3}\right\}\\
&+48\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}^{2}\right)\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}^{2}\right)\right\}\tau_{T}\\
=&O(NT^{3})+O(NT^{2})+\sum_{i=1}^{N}48u_{ii}
\tr\left(\mathbf{Q}_{ii}\right)\tr\left\{\mathbf{\Omega}_{ii}\circ u_{ii}\mathbf{\Omega}_{ii}\right\}\tr\left(\mathbf{Q}_{ii}\circ \mathbf{Q}_{ii}^{2}\right)\\
&+96\tr\left\{\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)u_{ii}^{2}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}^{3}\right\}\\
&+48\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}^{2}\right)\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}^{2}\right)\right\}\tau_{T}\\
=&O(NT^{3})+O(NT^{2})+\sum_{i=1}^{N}48u_{ii}u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\sum_{s=1}^{N}(\mathbf{\Omega}_{ii})_{ss}^{2}
\sum_{s=1}^{T}(\mathbf{Q}_{ii}^{2})_{ss}(\mathbf{Q}_{ii})_{ss}\\
&+96u_{ii}^{2}
\sum_{s=1}^{N}(\mathbf{\Omega}_{ii})_{ss}^{2}
\sum_{s=1}^{T}(\mathbf{Q}_{ii}^{3})_{ss}(\mathbf{Q}_{ii})_{ss}
+48u_{ii}^{2}
\sum_{s=1}^{N}(\mathbf{\Omega}_{ii})_{ss}^{2}
\sum_{s=1}^{T}(\mathbf{Q}_{ii}^{2})_{ss}^{2}\\
\leq&O(NT^{3})+O(NT^{2})+\sum_{i=1}^{N}48u_{ii}^{4}
|\tr\left(\mathbf{Q}_{ii}\right)|
|\sum_{s=1}^{T}(\mathbf{Q}_{ii}^{2})_{ss}(\mathbf{Q}_{ii})_{ss}|\\
&+96u_{ii}^{4}|
\sum_{s=1}^{T}(\mathbf{Q}_{ii}^{3})_{ss}(\mathbf{Q}_{ii})_{ss}|
+48u_{ii}^{4}|
\sum_{s=1}^{T}(\mathbf{Q}_{ii}^{2})_{ss}^{2}|\\
=&O(NT^{3})+O(NT^{2})+O(NT).
\end{align*}
Then,
\begin{align*}
G_{1_{3}}=&\sum_{i=1}^{N}
4u_{ii}\tr\left(\mathbf{Q}_{ii}\right)\tr\left\{\mathbf{\Omega}_{ii}\circ \mathbf{\Omega}_{ii}\circ \mathbf{\Omega}_{ii}\right\}\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\\
&+24\tr\left\{\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\circ
u_{ii}\mathbf{\Omega}_{ii}\right\}\tr(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}^{2})\\
=&\sum_{i=1}^{N}
4u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\sum_{s=1}^{N}(\mathbf{\Omega}_{ii})_{ss}^{3}
\sum_{s=1}^{T}(\mathbf{Q}_{ii})_{ss}^{3}+24u_{ii}
\sum_{s=1}^{N}(\mathbf{\Omega}_{ii})_{ss}^{3}
\sum_{s=1}^{T}(\mathbf{Q}_{ii})_{ss}^{2}(\mathbf{Q}_{ii}^{2})_{ss}\\
\leq&\sum_{i=1}^{N}
4u_{ii}^{4}
|\tr\left(\mathbf{Q}_{ii}\right)|
|\sum_{s=1}^{T}(\mathbf{Q}_{ii})_{ss}^{3}|+
24u_{ii}^{4}
|\sum_{s=1}^{T}(\mathbf{Q}_{ii})_{ss}^{2}(\mathbf{Q}_{ii}^{2})_{ss}|\\
=&O(NT^{2})+O(NT).
\end{align*}
Similarly, we can easily obtain that $G_{1_{4}}=O(NT).$
Note that for any square matrix $\mathbf{A}\in \mathbb{R}^{T\times T},$
$$
|\sum_{s=1}^{T}\sum_{t=1}^{T}(\mathbf{A})_{st}|
\leq\sqrt{T^{2}\tr\left(\mathbf{A}\mathbf{A}^{\prime}\right)}.$$
So,
we next have
\begin{align*}
G_{1_{5}}=&\sum_{i=1}^{N}
24\tau_{N}^{\prime}\left[\left\{\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right\}\mathbf{\Omega}_{ii}
\left\{\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right\}\right]\tau_{N}
\tau_{T}^{\prime}\left[\left\{\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right\}\mathbf{Q}_{ii}
\left\{\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right\}\right]\tau_{T}u_{ii}\tr(\mathbf{Q}_{ii})\\
&+48\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ii}\mathbf{\Omega}_{ii}\left\{\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right\}\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}^{2}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\\
&+16\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}
\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}
\circ\mathbf{Q}_{ii}\right)\tau_{T}u_{ii}\tr\left(\mathbf{Q}_{ii}\right)\\
&+96\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}\prime\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&+96\tr\left\{\mathbf{\Omega}_{ii}\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ii}\right\}
\tr\left\{\mathbf{Q}_{ii}\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\right\}\\
=&\sum_{i=1}^{N}
24\sum_{s=1}^{N}\sum_{t=1}^{N}(\mathbf{\Omega}_{ii})_{ss}
(\mathbf{\Omega}_{ii})_{st}
(\mathbf{\Omega}_{ii})_{tt}
\sum_{s=1}^{T}\sum_{t=1}^{T}(\mathbf{Q}_{ii})_{ss}
(\mathbf{Q}_{ii})_{st}
(\mathbf{Q}_{ii})_{tt}\\
&+48u_{ii}
\sum_{s=1}^{N}\sum_{t=1}^{N}(\mathbf{\Omega}_{ii})_{ss}
(\mathbf{\Omega}_{ii})_{st}
(\mathbf{\Omega}_{ii})_{tt}
\sum_{s=1}^{T}\sum_{t=1}^{T}(\mathbf{Q}_{ii})_{ss}
(\mathbf{Q}_{ii}^{2})_{st}
(\mathbf{Q}_{ii})_{tt}
\\
&+16u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\sum_{s=1}^{N}\sum_{t=1}^{N}(\mathbf{\Omega}_{ii})_{st}^{3}
\sum_{s=1}^{T}\sum_{t=1}^{T}(\mathbf{Q}_{ii})_{st}^{3}
\\
&+96
\sum_{s=1}^{N}\sum_{t=1}^{N}\sum_{q=1}^{N}(\mathbf{\Omega}_{ii})_{qt}
(\mathbf{\Omega}_{ii})_{tt}(\mathbf{\Omega}_{ii})_{sq}^{2}
\sum_{s=1}^{N}\sum_{t=1}^{N}\sum_{q=1}^{N}(\mathbf{Q}_{ii})_{qt}
(\mathbf{Q}_{ii})_{tt}(\mathbf{Q}_{ii})_{sq}^{2}
\\
&+96u_{ii}\sum_{s=1}^{N}\sum_{t=1}^{N}
(\mathbf{\Omega}_{ii})_{st}^{2}(\mathbf{\Omega}_{ii})_{ts}
\sum_{s=1}^{T}\sum_{t=1}^{T}
(\mathbf{Q}_{ii})_{st}^{2}(\mathbf{Q}_{ii}^{2})_{ts}\\
\leq&\sum_{i=1}^{N}
24u_{ii}^{2}
\sum_{s=1}^{T}\sum_{t=1}^{T}C|
(\mathbf{Q}_{ii})_{st}|+48u_{ii}^{3}
\sum_{s=1}^{T}\sum_{t=1}^{T}C|
(\mathbf{Q}_{ii}^{2})_{st}|
\\
&+16u_{ii}^{3}
\tr\left(\mathbf{Q}_{ii}\right)
\sum_{s=1}^{T}\sum_{t=1}^{T}C(\mathbf{Q}_{ii})_{st}^{2}
+96u_{ii}^{3}
\sum_{t=1}^{N}\sum_{q=1}^{N}|(\mathbf{Q}_{ii})_{qt}|
(\mathbf{Q}_{ii})_{tt}(\mathbf{Q}_{ii}^{2})_{qq}
\\
&+96u_{ii}^{3}
\sum_{s=1}^{T}\sum_{t=1}^{T}C
(\mathbf{Q}_{ii})_{st}^{2}\\=&O(NT^{3/2})+O(NT^{2})+O(NT).
\end{align*}
Similarly, for $G_{1_{6}},$
we have
\begin{align*}
G_{1_{6}}=&\sum_{i=1}^{N}
2\tr^{2}\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)\tr^{2}\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\\
&+24\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&+8\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}
\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}
\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
=&\sum_{i=1}^{N}
2\big\{\sum_{s=1}^{N}(\mathbf{\Omega}_{ii})_{ss}^{2}\big\}^{2}
\big\{\sum_{s=1}^{N}(\mathbf{Q}_{ii})_{ss}^{2}\big\}^{2}
+24\sum_{s=1}^{N}\sum_{t=1}^{N}(\mathbf{\Omega}_{ii})_{st}^{2}
(\mathbf{\Omega}_{ii})_{ss}(\mathbf{\Omega}_{ii})_{tt}
\sum_{s=1}^{T}\sum_{t=1}^{T}(\mathbf{Q}_{ii})_{st}^{2}
(\mathbf{Q}_{ii})_{ss}(\mathbf{Q}_{ii})_{tt}\\
&+8\sum_{s=1}^{N}\sum_{t=1}^{N}
(\mathbf{\Omega}_{ii})_{st}^{4}
\sum_{s=1}^{T}\sum_{t=1}^{T}
(\mathbf{Q}_{ii})_{st}^{4}\\
\leq&\sum_{i=1}^{N}
2u_{ii}^{4}\tr^{2}(\mathbf{Q}_{ii}^{2})
+24u_{ii}^{4}
\sum_{s=1}^{T}\sum_{t=1}^{T}C(\mathbf{Q}_{ii})_{st}^{2}+8u_{ii}^{4}
\sum_{s=1}^{T}\sum_{t=1}^{T}C
(\mathbf{Q}_{ii})_{st}^{2}\\
=&O(NT^{2})+O(NT).
\end{align*}
Furthermore, we have
\begin{align*}
G_{1_{7}}=&\sum_{i=1}^{N}24\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&+32\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
=&\sum_{i=1}^{N}24
\sum_{s=1}^{N}\sum_{t=1}^{N}(\mathbf{\Omega}_{ii})_{st}(\mathbf{\Omega}_{ii})_{ss}
(\mathbf{\Omega}_{ii})_{tt}^{2}
\sum_{s=1}^{T}\sum_{t=1}^{T}(\mathbf{Q}_{ii})_{st}(\mathbf{Q}_{ii})_{ss}
(\mathbf{Q}_{ii})_{tt}^{2}\\
&+32\sum_{s=1}^{N}\sum_{t=1}^{N}
(\mathbf{\Omega}_{ii})_{st}^{3}(\mathbf{\Omega}_{ii})_{ss}
\sum_{s=1}^{T}\sum_{t=1}^{T}
(\mathbf{Q}_{ii})_{st}^{3}(\mathbf{Q}_{ii})_{ss}\\
\leq&\sum_{i=1}^{N}24
u_{ii}^{3}
\sum_{s=1}^{T}\sum_{t=1}^{T}C|(\mathbf{Q}_{ii})_{st}|+32u_{ii}^{3}
\sum_{s=1}^{T}\sum_{t=1}^{T}C
(\mathbf{Q}_{ii})_{st}^{2}\\
=&O(NT^{3/2})+O(NT).
\end{align*}
Combining the above calculations, we can get $G_{1}=O(NT^{3}).$
We next deal with $G_{2}$. Similarly, due to Lemma \ref{le:moment of quadratic form}, we have
$$
G_{2}=G_{2_{1}}+\gamma_{2}G_{2_{2}}+\gamma_{4}G_{2_{3}}+\gamma_{6}G_{2_{4}}
+\gamma_{1}^{2}G_{2_{5}}+\gamma_{2}^{2}G_{2_{6}}+\gamma_{1}\gamma_{3}G_{2_{7}},$$
where
\begin{align*}
G_{2_{1}}=&2\underset{i\neq j}{\sum^{N}\sum^{N}}4u_{ij}^{4}\tr^{2}\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}^{2}\right)+4u_{ij}^{2}u_{ii}u_{jj}\tr^{2}\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right)\\
&+2u_{ij}^{4}\tr^{2}\left(\mathbf{Q}_{ij}^{2}\right)
+2u_{ii}^{2}u_{jj}^{2}\tr^{2}\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right)
+4u_{ij}^{2}u_{ii}u_{jj}\tr\left(\mathbf{Q}_{ij}^{2}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right)\\
&+8u_{ij}^{4}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}^{3}\right)
+24u_{ii}u_{jj}u_{ij}^{2}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{ji}\right)
+6u_{ij}^{4}\tr\left(\mathbf{Q}_{ij}^{4}\right)\\
&+24u_{ij}^{2}u_{ii}u_{jj}\tr\left(\mathbf{Q}_{ij}^{3}\mathbf{Q}_{ji}\right)
+12u_{ij}^{2}u_{ii}u_{jj}\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{ji}^{2}\right)
+6u_{ii}^{2}u_{jj}^{2}\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right),\\
G_{2_{2}}=&2\underset{i\neq j}{\sum^{N}\sum^{N}}
4u_{ij}^{2}\tr^{2}\left(\mathbf{Q}_{ij}\right)\tr\left\{\mathbf{\Omega}_{ij}\circ \mathbf{\Omega}_{ij}\right\}\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\\
&+8\Big\{\frac{1}{2}u_{ij}^{2}\tr\left(\mathbf{Q}_{ij}^{2}\right)
+\frac{1}{2}u_{ii}u_{jj}\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right)\Big\}
\tr\big(\mathbf{\Omega}_{ij}\circ \mathbf{\Omega}_{ij}\big)\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\\
&+48u_{ij}\tr\left(\mathbf{Q}_{ij}\right)\Big[\frac{1}{2}u_{ij}
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}^{2}\right)\\
&\quad+
\frac{1}{4}u_{jj}
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ij}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}\\
&\quad+
\frac{1}{4}u_{ii}
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{jj}\right)
\tr\left\{\mathbf{Q}_{ij}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}
\Big]\\
&+24\Big[u_{ii}u_{jj}
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ji}\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}
\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right\}\\
&\quad+u_{ij}^{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ij}\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}^{3}\right\}\\
&\quad+u_{ij}u_{jj}
tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}
\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right\}\\
&\quad+u_{ij}u_{ii}
tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{jj}\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}
\mathbf{Q}_{ji}\mathbf{Q}_{ij}\right\}\Big]\\
&+3\Big[4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}\right)^{2}\tau_{N}
\tau_{T}^{\prime}\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ij})\right\}^{2}\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}\right)^{2}\tau_{N}
\tau_{T}^{\prime}\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}^{2}\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}\right)^{2}\tau_{N}
\tau_{T}^{\prime}\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}^{2}\tau_{T}\\
&\quad+4\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}\right)\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ij}^{2}\right)
\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}\tau_{T}\\
&\quad+4\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}\right)\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ij}^{2}\right)
\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}\tau_{T}\\
&\quad+2\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}\right)\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}
\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}\tau_{T}\Big],\\
G_{2_{3}}=&2\underset{i\neq j}{\sum^{N}\sum^{N}}
4u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}
\circ\mathbf{\Omega}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}
\circ\mathbf{Q}_{ij}\right)\\
&+24\Big[\frac{1}{2}u_{ij}\tr\left(\mathbf{\Omega}_{ij}
\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}
\circ\mathbf{Q}_{ij}\circ\mathbf{Q}^{2}_{ij}\right)\\
&\quad+\frac{1}{4}u_{jj}\tr\left(\mathbf{\Omega}_{ij}
\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ\mathbf{Q}_{ij}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}
\\
&\quad+\frac{1}{4}u_{ii}\tr\left(\mathbf{\Omega}_{ij}
\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{jj}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ\mathbf{Q}_{ij}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}
\Big],\\
G_{2_{4}}=&2\underset{i\neq j}{\sum^{N}\sum^{N}}
\tr\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right),\\
G_{2_{5}}=&2\underset{i\neq j}{\sum^{N}\sum^{N}}
24u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&+12\Big\{2u_{ij}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}^{2}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+u_{jj}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ji}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
u_{ii}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{jj}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\Big\}\\
&+16u_{ij}\tr\left(\mathbf{Q}_{ij}\right)\frac{1}{4}
\Big[\tr\left\{\mathbf{\Omega}_{ij}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)\right\}\tr\left\{\mathbf{Q}_{ij}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\right\}\\
&\quad+2\tr\left\{\mathbf{\Omega}_{ij}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)\right\}\tr\left\{\mathbf{Q}_{ij}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\right\}\\
&\quad+\tr\left\{\mathbf{\Omega}_{ij}\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)\right\}\tr\left\{\mathbf{Q}_{ij}\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\right\}\Big]\\
&+12\Big\{\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\Big\}\\
&+12\Big[
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)u_{ij}
\mathbf{\Omega}_{ij}\right\}\tr\left\{\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}^{2}\right\}\\
&\quad+\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{ij}
\mathbf{\Omega}_{ji}\right\}\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}^{2}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{jj}
\mathbf{\Omega}_{ii}\right\}\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}
\mathbf{Q}_{ji}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{ii}
\mathbf{\Omega}_{jj}\right\}\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)u_{ij}
\mathbf{\Omega}_{ij}\right\}\tr\left\{\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ij}^{2}\right\}
\\
&\quad+\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{ij}
\mathbf{\Omega}_{ji}\right\}\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ji}^{2}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{jj}
\mathbf{\Omega}_{ii}\right\}\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ij}
\mathbf{Q}_{ji}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{ii}
\mathbf{\Omega}_{jj}\right\}\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right\}\Big],\\
G_{2_{6}}=&2\underset{i\neq j}{\sum^{N}\sum^{N}}
2\tr^{2}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)\tr^{2}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\\
&+12\Big\{\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\Big\}\\
&+\frac{1}{2}\big\{
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\tau_{T}
\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+6
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\tau_{T}
\big\},
\\
G_{2_{7}}=&2\underset{i\neq j}{\sum^{N}\sum^{N}}
12\Big\{\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\Big\}\\
&+4\Big\{\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)\tau_{T}
\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\Big\}.
\end{align*}
Because $\underset{i\neq j}{\sum\sum}u_{ij}^{4}<\underset{i\neq j}{\sum\sum}u_{ij}^{2}<\tr\left(\mathbf{U}^{2}\right)=O(N),$
we have $G_{2_{1}}=O(NT^{3})+O(N^{2}T^{2}).$
Similarly,
due to $\underset{i\neq j}{\sum\sum}|u_{ij}|=O(N^{1\vee(2-1/\tau)})$
and Lemma \ref{lem:sigma},
we have $G_{2_{2}}=O(NT^{3})+O(N^{2}T^{2}),$
$G_{2_{3}}=O(N^{3/2}T^{2})+O(N^{2}T),$
$G_{2_{4}}=O(N^{2}T),$
$G_{2_{5}}=O(N^{2}T^{2}),$
$G_{2_{6}}=O(N^{2}T^{2})$
and $G_{2_{7}}=O(N^{2}T^{3/2}).$
Combining the above calculations, we can get $G_{2}= O(NT^{3}+N^{2}T^{2}$. We further deal with $G_{3}$.
Similarly, due to Lemma \ref{le:moment of quadratic form}, we have
$$
G_{3}=G_{3_{1}}+\gamma_{2}G_{3_{2}}+\gamma_{4}G_{3_{3}}+\gamma_{6}G_{3_{4}}
+\gamma_{1}^{2}G_{3_{5}}+\gamma_{2}^{2}G_{3_{6}}+\gamma_{1}\gamma_{3}G_{3_{7}},$$
where
\begin{align*}
G_{3_{1}}=&\underset{i\neq j}{\sum^{N}\sum^{N}}8 u_{ii}u_{jj}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{Q}_{jj}\right)\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{jj}\right)
+8u_{ij}^{4}\tr^{2}\left(\mathbf{Q}_{ii}\mathbf{Q}_{jj}\right)\\
&+16u_{ii}u_{jj}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{jj}^{2}\right)
\tr\left(\mathbf{Q}_{ii}\right)+16u_{jj}u_{ii}u_{ij}^{2}
\tr\left(\mathbf{Q}_{ii}^{2}\mathbf{Q}_{jj}\right)
\tr\left(\mathbf{Q}_{jj}\right)\\
&+32u_{ii}u_{jj}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}^{2}\mathbf{Q}_{jj}^{2}\right)
+16u_{ij}^{4}\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{jj}\mathbf{Q}_{ii}
\mathbf{Q}_{jj}\right),\\
G_{3_{2}}=&\underset{i\neq j}{\sum^{N}\sum^{N}}
4
u_{ii}u_{jj}\tr\left(\mathbf{Q}_{ii}\right)\tr\left(\mathbf{Q}_{jj}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)\\
&+8\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)\tau_{T}
\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)\\
&+4\Big[
2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)\tr\left(\mathbf{\Omega}_{ii}\circ
u_{jj}\mathbf{\Omega}_{jj}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}^{2}\right)\\
&\quad+4u_{ii}\tr\left(\mathbf{Q}_{ii}\right)\tr\left(\mathbf{\Omega}_{jj}\circ
u_{ij}\mathbf{\Omega}_{ij}\right)
\tr\left\{\mathbf{Q}_{jj}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{jj})\right\}\\
&\quad+2u_{jj}\tr\left(\mathbf{Q}_{jj}\right)\tr\left(\mathbf{\Omega}_{jj}\circ
u_{ii}\mathbf{\Omega}_{ii}\right)
\tr\left(\mathbf{Q}_{jj}\circ\mathbf{Q}_{ii}^{2}\right)\\
&\quad+4u_{jj}\tr\left(\mathbf{Q}_{jj}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ
u_{ij}\mathbf{\Omega}_{ij}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{jj})\right\}\Big]\\
&+8\Big[4\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}u_{jj}\mathbf{\Omega}_{ij}\right\}\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ii}\mathbf{Q}_{jj}^{2}\right\}\\
&\quad+2
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}^{2}\mathbf{\Omega}_{jj}\right\}\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jj}\mathbf{Q}_{ii}\mathbf{Q}_{jj}\right\}\\
&\quad+4
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)
u_{ij}u_{ii}\mathbf{\Omega}_{ij}\right\}\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{jj}\right)
\mathbf{Q}_{ii}^{2}\mathbf{Q}_{jj}\right\}\\
&\quad+2
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)
u_{ij}^{2}\mathbf{\Omega}_{ii}\right\}\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{jj}\right)
\mathbf{Q}_{ii}\mathbf{Q}_{jj}\mathbf{Q}_{ii}\right\}\Big]\\
&+16\Big[\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}^{2}\right)
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jj}^{2}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}\right)
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ii}\mathbf{Q}_{jj})\right\}
\left\{\mathbf{I}_{T}\circ (\mathbf{Q}_{ii}\mathbf{Q}_{jj})\right\}\tau_{T}
\Big],\\
G_{3_{3}}=&\underset{i\neq j}{\sum^{N}\sum^{N}}
2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\circ\mathbf{\Omega}_{jj}
\right)\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj}
\right)\\
&+2u_{jj}\tr\left(\mathbf{Q}_{jj}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}
\right)\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}
\right)\\
&+4\Big[
\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ u_{jj}\mathbf{\Omega}_{jj}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}^{2}
\right)\\
&\quad+4\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\circ u_{ij}\mathbf{\Omega}_{ij}\right)
\tr\left\{\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{jj})
\right\}\\
&\quad+\tr\left(\mathbf{\Omega}_{jj}\circ\mathbf{\Omega}_{jj}\circ u_{ii}\mathbf{\Omega}_{ii}\right)
\tr\left(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj}\circ\mathbf{Q}_{ii}^{2}
\right)\Big],\\
G_{3_{4}}=&\underset{i\neq j}{\sum^{N}\sum^{N}}
\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{jj}\circ\mathbf{\Omega}_{jj}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ
\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj}\right),\\
G_{3_{5}}=&\underset{i\neq j}{\sum^{N}\sum^{N}}
2\Big\{4u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{jj}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\mathbf{Q}_{jj}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jj}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jj}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jj}\right)\tau_{T}
\\
&\quad+4u_{jj}\tr\left(\mathbf{Q}_{jj}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jj}\right)\tau_{T}
\\
&\quad+2u_{jj}\tr\left(\mathbf{Q}_{jj}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{jj}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\mathbf{Q}_{jj}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}\Big\}\\
&+4\Big\{4
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}\mathbf{\Omega}_{ij}
\left\{\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right\}\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ii}\mathbf{Q}_{jj}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jj}\right)\right\}\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{jj}\mathbf{\Omega}_{jj}
\left\{\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right\}\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jj}^{2}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)
u_{ii}\mathbf{\Omega}_{ii}
\left\{\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right\}\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jj}\right)
\mathbf{Q}_{ii}^{2}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jj}\right)\right\}\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)
u_{ij}\mathbf{\Omega}_{ij}
\left\{\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right\}\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jj}\right)
\mathbf{Q}_{ii}\mathbf{Q}_{jj}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\Big\}\\
&+4\Big\{2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}
\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}
\circ\mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+2u_{jj}\tr\left(\mathbf{Q}_{jj}\right)
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}
\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}
\circ\mathbf{Q}_{jj}\right)\tau_{T}\Big\}\\
&+8\Big\{2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{jj}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jj}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)
\mathbf{\Omega}_{jj}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)
\mathbf{Q}_{jj}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jj}\circ\mathbf{\Omega}_{jj}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\Big\}\\
&+16\Big[4\tr\left\{\mathbf{\Omega}_{ii}\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{jj}\right)\mathbf{\Omega}_{jj}\right\}
\tr\left\{\mathbf{Q}_{ii}\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{jj}\right)\mathbf{Q}_{jj}\right\}\\
&\quad+\tr\left\{\mathbf{\Omega}_{ii}\left(\mathbf{\Omega}_{jj}\circ
\mathbf{\Omega}_{jj}\right)\mathbf{\Omega}_{ii}\right\}
\tr\left\{\mathbf{Q}_{ii}\left(\mathbf{Q}_{jj}\circ
\mathbf{Q}_{jj}\right)\mathbf{Q}_{ii}\right\}\\
&\quad+\tr\left\{\mathbf{\Omega}_{jj}\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{jj}\right\}
\tr\left\{\mathbf{Q}_{jj}\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{jj}\right\}\Big],\\
G_{3_{6}}=&\underset{i\neq j}{\sum^{N}\sum^{N}}
2\tr^{2}\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{jj}\right)\tr^{2}\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{jj}\right)\\
&+4\Big\{4\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{jj}\circ\mathbf{\Omega}_{jj}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\right)\tau_{T}\Big\}\\
&+8\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}
\circ\mathbf{\Omega}_{jj}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}
\circ\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj}\right)\tau_{T},\\
G_{3_{7}}=&\underset{i\neq j}{\sum^{N}\sum^{N}}
2\Big\{2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+4\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{jj}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{jj}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+4\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)\mathbf{\Omega}_{jj}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\right)\mathbf{Q}_{jj}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\Big\}\\
&+8\Big\{2\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jj}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{jj}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jj}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{jj}\right)\tau_{T}\Big\}.
\end{align*}
Because $\underset{i\neq j}{\sum\sum}u_{ij}^{4}<\underset{i\neq j}{\sum\sum}u_{ij}^{2}<\tr\left(\mathbf{U}^{2}\right)=O(N),$
we have $G_{3_{1}}=O(NT^{3})+O(NT^{2}).$
Similarly,
due to $\underset{i\neq j}{\sum\sum}|u_{ij}|=O(N^{1\vee(2-1/\tau)})$
and Lemma \ref{lem:sigma},
we have $G_{3_{2}}=O(N^{2}T^{3})+O(N^{2}T^{2}),$
$G_{3_{3}}=O(N^{2}T),$
$G_{3_{4}}=O(N^{2}T),$
$G_{3_{5}}=O(N^{2}T^{5/2}),$
$G_{3_{6}}=O(N^{2}T^{2})$
and $G_{3_{7}}=O(N^{2}T^{3/2}).$
Combining the above calculations, we can get $G_{3}= O(N^{2}T^{3})$. We next deal with $G_{4}$.
Similarly,
$$
G_{4}=G_{4_{1}}+\gamma_{2}G_{4_{2}}+\gamma_{4}G_{4_{3}}+\gamma_{6}G_{4_{4}}
+\gamma_{1}^{2}G_{4_{5}}+\gamma_{2}^{2}G_{4_{6}}+\gamma_{1}\gamma_{3}G_{4_{7}},$$
where
\begin{align*}
G_{4_{1}}=&4\underset{i\neq j}{\sum^{N}\sum^{N}}8u_{ii}^{2}u_{ij}^{2}
\tr\left(\mathbf{Q}_{ii}\right)\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{ij}\right)+8u_{ii}^{2}u_{ij}^{2}
\tr^{2}\left(\mathbf{Q}_{ii}\mathbf{Q}_{ij}\right)\\
&+8u_{ii}^{2}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{ij}^{2}\right)
+4u_{ii}^{3}u_{jj}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right)\\
&+4u_{ii}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{ji}\mathbf{Q}_{ij}\right)
+16u_{ij}^{2}u_{ii}^{2}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ii}^{2}\mathbf{Q}_{ij}\right)\\
&+16u_{ii}^{2}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}^{2}\mathbf{Q}_{ij}^{2}\right)
+8u_{ii}^{3}u_{jj}\tr\left(\mathbf{Q}_{ii}^{2}\mathbf{Q}_{ij}
\mathbf{Q}_{ji}\right)\\
&+8u_{ii}^{2}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}^{2}\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right)+8u_{ii}^{2}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{ij}\mathbf{Q}_{ii}
\mathbf{Q}_{ij}
\right)\\
&+8u_{ii}^{2}u_{ij}^{2}\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{ij}\mathbf{Q}_{ii}
\mathbf{Q}_{ji}
\right),\\
G_{4_{2}}=&4\underset{i\neq j}{\sum^{N}\sum^{N}}4u_{ii}u_{ij}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{Q}_{ij}\right)\tr\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ii}\right)\tr\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ii}\right)\\
&+8u_{ii}u_{ij}\tr\left(
\mathbf{Q}_{ij}\mathbf{Q}_{ii}\right)\tr\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ii}\right)\tr\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ii}\right)\\
&+4\big[2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ u_{ii}\mathbf{\Omega}_{ii}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}^{2}\right)\\
&\quad+2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ u_{ij}\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ii})\right\}
\\
&\quad+2u_{ij}u_{ii}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ji}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ii})\right\}\\
&\quad+u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ u_{ij}\mathbf{\Omega}_{ij}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ij}^{2}\right)
\\
&\quad+\frac{1}{2}u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ u_{jj}\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}
\\
&\quad+\frac{1}{2}u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ u_{ii}\mathbf{\Omega}_{jj}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}\\
&\quad+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ u_{ij}\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ij}\circ( \mathbf{Q}_{ij}\mathbf{Q}_{ii})\right\}\\
&\quad+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ u_{ii}\mathbf{\Omega}_{ji}\right)
\tr\left\{\mathbf{Q}_{ij}\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ii})\right\}
\big]\\
&+8\big[2\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ij}u_{ii}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ii}^{2}\right\}\\
&\quad+2\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ii}^{2}\mathbf{\Omega}_{ji}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ii}^{2}\right\}\\
&\quad+2\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ii}u_{ij}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ii}\mathbf{Q}_{ij}\mathbf{Q}_{ii}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}^{2}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}^{2}\mathbf{Q}_{ii}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}u_{ii}\mathbf{\Omega}_{ji}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}^{2}\mathbf{Q}_{ii}\right\}
\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ii}u_{jj}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ji}\mathbf{Q}_{ii}\right\}
\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}u_{ii}\mathbf{\Omega}_{ji}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ij}\mathbf{Q}_{ii}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}u_{ii}\mathbf{\Omega}_{ij}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ii}\mathbf{Q}_{ij}\right\}
\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ii}^{2}\mathbf{\Omega}_{jj}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ii}\mathbf{Q}_{ij}\right\}
\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}^{2}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ii}\mathbf{Q}_{ji}\right\}
\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}u_{ii}\mathbf{\Omega}_{ji}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ii}\mathbf{Q}_{ji}\right\}
\big]\\
&+16\big[\frac{1}{4}\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ij})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{ii})\right\}\tau_{T}\\
&\quad+\frac{1}{4}\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ji})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{ii})\right\}\tau_{T}
\\
&\quad+\frac{1}{4}\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ u_{ii}
\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{ii})\right\}\tau_{T}
\\
&\quad+\frac{1}{4}\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}\right)
\left(\mathbf{I}_{N}\circ u_{ii} \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{ii})\right\}\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ji}\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ u_{ij} \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ji}\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ji}\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ji}\mathbf{Q}_{ii}\right)\tau_{T}\big],\\
G_{4_{3}}=&4\underset{i\neq j}{\sum^{N}\sum^{N}}
2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tr\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\\
&+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)\\
&+4\big[\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ u_{ii}\mathbf{\Omega}_{ii}\right)
\tr\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}^{2}\right)\\
&\quad+2
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\circ u_{ij}
\mathbf{\Omega}_{ii}\right)
\tr\left\{
\mathbf{Q}_{ij}\circ(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ij}
\mathbf{Q}_{ii})\right\}\\
&\quad+2
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\circ u_{ii}
\mathbf{\Omega}_{ji}\right)
\tr\left\{\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\circ(\mathbf{Q}_{ji}
\mathbf{Q}_{ii})\right\}\\
&\quad+\frac{1}{2}\tr\left(
\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ u_{ij}
\mathbf{\Omega}_{ij}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ij}^{2}\right)
\\
&\quad+\frac{1}{4}\tr\left(
\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ u_{jj}
\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji})
\right\}\\
&\quad+\frac{1}{4}\tr\left(
\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ u_{ii}
\mathbf{\Omega}_{jj}\right)
\tr\left\{\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij})
\right\}\big],\\
G_{4_{4}}=&4\underset{i\neq j}{\sum^{N}\sum^{N}}
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right),\\
G_{4_{5}}=&4\underset{i\neq j}{\sum^{N}\sum^{N}}
2\big\{4u_{ij}\tr\left(\mathbf{\Omega}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ij}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2u_{ij}\tr\left(\mathbf{\Omega}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2u_{ii}\tr\left(\mathbf{\Omega}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+2u_{ii}\tr\left(\mathbf{\Omega}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2u_{ii}\tr\left(\mathbf{\Omega}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ij}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ij}\right)\tau_{T}
\big\}\\
&+4\big\{2
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ij}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ii}\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ii}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ii}^{2}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\right\}\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}^{2}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{jj}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ii}\mathbf{\Omega}_{jj}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\right\}\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\right\}\tau_{T} \\
&\quad+2
\tau_{N}^{\prime}\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ii}\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\right\}\tau_{N}
\tau_{T}^{\prime}\left\{\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\right\}\tau_{T}
\big\}\\
&+4\big\{2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}
\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}
\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}
\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}
\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}
\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}
\circ\mathbf{Q}_{ii}\right)\tau_{T}
\big\}\\
&+8\big\{\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ji}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ji}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ji}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ji}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\big\}\\
&+16\big[
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ii}\right)u_{ii}\mathbf{\Omega}_{ij}\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\mathbf{Q}_{ij}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ii}\right)u_{ij}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\mathbf{Q}_{ji}\right\}
\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ii}\right)u_{ii}\mathbf{\Omega}_{ij}\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\mathbf{Q}_{ij}\right\}
\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ii}\right)u_{ij}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\mathbf{Q}_{ji}\right\}
\\
&\quad+
\frac{1}{2}
\tr\left\{\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)u_{ij}\mathbf{\Omega}_{ij}\right\}
\tr\left\{\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ij}^{2}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)u_{jj}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)u_{ii}\mathbf{\Omega}_{jj}\right\}
\tr\left\{\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}\right\}\\
&\quad+
\frac{1}{2}
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{ii}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ii}^{2}\right\}\\
&\quad+
\frac{1}{2}
\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{ii}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ii}^{2}\right\}\big],\\
G_{4_{6}}=&4\underset{i\neq j}{\sum^{N}\sum^{N}}
2\tr^{2}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)
\tr^{2}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)\\
&+4\big\{2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\big\}\\
&+4
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}
\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}
\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&+4
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}
\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}
\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T},\\
G_{4_{7}}=&4\underset{i\neq j}{\sum^{N}\sum^{N}}
2\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\big\}\\
&+8\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\big\}.
\end{align*}
Because $\underset{i\neq j}{\sum\sum}u_{ij}^{4}<\underset{i\neq j}{\sum\sum}u_{ij}^{2}<\tr\left(\mathbf{U}^{2}\right)=O(N),$
we have $G_{4_{1}}=O(NT^{3})+O(N^{2}T^{2}).$
Similarly,
due to $\underset{i\neq j}{\sum\sum}|u_{ij}|=O(N^{1\vee(2-1/\tau)})$
and Lemma \ref{lem:sigma},
we have $G_{4_{2}}=O(N^{1\vee(2-1/\tau)}T^{3})+O(N^{2}T^{2}),$
$G_{4_{3}}=O(N^{2}T^{2}),$
$G_{4_{4}}=O(N^{2}T),$
$G_{4_{5}}=O(N^{2}T^{5/2}),$
$G_{4_{6}}=O(N^{2}T^{2})$
and $G_{4_{7}}=O(N^{2}T^{2}).$
Combining the above calculations, we can get $G_{4}= O(N^{2}T^{5/2}+N^{1\vee(2-1/\tau)}T^{3})$. We next deal with $G_{5}$.
Similarly, due to Lemma \ref{le:moment of quadratic form}, we have
$$
G_{5}=G_{5_{1}}+\gamma_{2}G_{5_{2}}+\gamma_{4}G_{5_{3}}+\gamma_{6}G_{5_{4}}
+\gamma_{1}^{2}G_{5_{5}}+\gamma_{2}^{2}G_{5_{6}}+\gamma_{1}\gamma_{3}G_{5_{7}},$$
where
\begin{align*}
G_{5_{1}}=&2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}8u_{ii}u_{ij}u_{jk}u_{ik}
\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{Q}_{jk}\right)
\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{jk}\right)+8u_{ij}^{2}u_{ik}^{2}
\tr^{2}\left(\mathbf{Q}_{ii}\mathbf{Q}_{jk}\right)\\
&+8u_{ii}u_{ij}u_{jk}u_{ik}
\tr\left(\mathbf{Q}_{ii}\right)\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{jk}^{2}\right)
+4u_{ii}u_{ij}^{2}u_{kk}
\tr\left(\mathbf{Q}_{ii}\right)\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{jk}
\mathbf{Q}_{kj}\right)\\
&
+4u_{ij}^{2}u_{ii}u_{jj}
\tr\left(\mathbf{Q}_{ii}\right)\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{kj}
\mathbf{Q}_{jk}\right)+8u_{jk}u_{ik}u_{ij}u_{ii}
\tr\left(\mathbf{Q}_{jk}\right)\tr\left(\mathbf{Q}_{ii}^{2}
\mathbf{Q}_{jk}\right)
\\
&+16u_{ii}u_{ij}u_{kj}u_{ik}\tr\left(\mathbf{Q}_{ii}^{2}
\mathbf{Q}_{jk}^{2}\right)
+8u_{ii}u_{ij}^{2}u_{kk}\tr\left(\mathbf{Q}_{ii}^{2}
\mathbf{Q}_{jk}\mathbf{Q}_{kj}\right)
+8u_{ii}u_{jj}u_{ik}^{2}
\tr\left(\mathbf{Q}_{ii}^{2}
\mathbf{Q}_{kj}\mathbf{Q}_{jk}\right)
\\
&+8u_{ij}^{2}u_{ik}^{2}
\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{jk}\mathbf{Q}_{ii}
\mathbf{Q}_{jk}\right)+8
u_{ij}^{2}u_{ik}^{2}
\tr\left(\mathbf{Q}_{ii}\mathbf{Q}_{jk}\mathbf{Q}_{ii}
\mathbf{Q}_{kj}\right),\\
G_{5_{2}}=&2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
4u_{ii}u_{jk}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{Q}_{jk}\right)\tr\left(\mathbf{\Omega}_{jk}\circ
\mathbf{\Omega}_{ii}\right)\tr\left(\mathbf{Q}_{jk}\circ
\mathbf{Q}_{ii}\right)\\
&+8u_{ij}u_{ik}\tr\left(
\mathbf{Q}_{jk}\mathbf{Q}_{ii}\right)\tr\left(\mathbf{\Omega}_{jk}\circ
\mathbf{\Omega}_{ii}\right)\tr\left(\mathbf{Q}_{jk}\circ
\mathbf{Q}_{ii}\right)\\
&+4\big[2u_{jk}\tr\left(\mathbf{Q}_{jk}\right)
\tr\left(\mathbf{\Omega}_{jk}\circ u_{ii}\mathbf{\Omega}_{ii}\right)
\tr\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}^{2}\right)\\
&\quad+2u_{jk}\tr\left(\mathbf{Q}_{jk}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ u_{ik}\mathbf{\Omega}_{ji}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{ii})\right\}
\\
&\quad+2u_{jk}u_{ij}\tr\left(\mathbf{Q}_{jk}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ki}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{ii})\right\}\\
&\quad+u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ u_{jk}\mathbf{\Omega}_{jk}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{jk}^{2}\right)
\\
&\quad+\frac{1}{2}u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ u_{kk}\mathbf{\Omega}_{jj}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{kj})\right\}
\\
&\quad+\frac{1}{2}u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{ii}\circ u_{jj}\mathbf{\Omega}_{kk}\right)
\tr\left\{\mathbf{Q}_{ii}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{jk})\right\}\\
&\quad+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{jk}\circ u_{ik}\mathbf{\Omega}_{ji}\right)
\tr\left\{\mathbf{Q}_{jk}\circ( \mathbf{Q}_{jk}\mathbf{Q}_{ii})\right\}\\
&\quad+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tr\left(\mathbf{\Omega}_{jk}\circ u_{ij}\mathbf{\Omega}_{ki}\right)
\tr\left\{\mathbf{Q}_{jk}\circ (\mathbf{Q}_{kj}\mathbf{Q}_{ii})\right\}
\big]\\
&+8\big[2\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
u_{ik}u_{ii}\mathbf{\Omega}_{ji}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{jk}\right)
\mathbf{Q}_{jk}\mathbf{Q}_{ii}^{2}\right\}\\
&\quad+2\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
u_{ii}u_{ij}\mathbf{\Omega}_{ki}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{jk}\right)
\mathbf{Q}_{kj}\mathbf{Q}_{ii}^{2}\right\}\\
&\quad+2\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
u_{ij}u_{ik}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{jk}\right)
\mathbf{Q}_{ii}\mathbf{Q}_{jk}\mathbf{Q}_{ii}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{jk}u_{ik}\mathbf{\Omega}_{ji}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}^{2}\mathbf{Q}_{ii}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{jk}u_{ij}\mathbf{\Omega}_{ki}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{kj}^{2}\mathbf{Q}_{ii}\right\}
\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}u_{kk}\mathbf{\Omega}_{ji}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}\mathbf{Q}_{kj}\mathbf{Q}_{ii}\right\}
\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{jj}u_{ik}\mathbf{\Omega}_{ki}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{kj}\mathbf{Q}_{jk}\mathbf{Q}_{ii}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}u_{ik}\mathbf{\Omega}_{jk}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}\mathbf{Q}_{ii}\mathbf{Q}_{jk}\right\}
\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}^{2}\mathbf{\Omega}_{kk}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{kj}\mathbf{Q}_{ii}\mathbf{Q}_{jk}\right\}
\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ik}^{2}\mathbf{\Omega}_{jj}\right\}
\tr\left\{\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}\mathbf{Q}_{ii}\mathbf{Q}_{kj}\right\}
\big]\\
&+16\big[\frac{1}{4}\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{jk}\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{jk})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{ii})\right\}\tau_{T}\\
&\quad+\frac{1}{4}\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{kj}\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{kj})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{ii})\right\}\tau_{T}
\\
&\quad+\frac{1}{4}\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{kk}\mathbf{\Omega}_{jj}\right)
\left(\mathbf{I}_{N}\circ u_{ii}
\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{kj})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{ii})\right\}\tau_{T}
\\
&\quad+\frac{1}{4}\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{kk}\right)
\left(\mathbf{I}_{N}\circ u_{ii} \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{jk})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ii}\mathbf{Q}_{ii})\right\}\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{ii})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{ii})\right\}\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ki}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{ii})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{ii})\right\}\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ki}\right)
\left(\mathbf{I}_{N}\circ u_{ik} \mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{ii})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{ii})\right\}\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ki}\right)
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ki}\right)\tau_{N}
\tau_{T}^{\prime}\left\{
\mathbf{I}_{T}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{ii})\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{ii})\right\}\tau_{T}\big],\\
G_{5_{3}}=&2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
2u_{jk}\tr\left(\mathbf{Q}_{jk}\right)
\tr\left(
\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tr\left(
\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\\
&+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)\tr\left(
\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)
\tr\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)\\
&+4\big[\tr\left(
\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}\circ u_{ii}\mathbf{\Omega}_{ii}\right)
\tr\left(
\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}^{2}\right)\\
&\quad+2
\tr\left(
\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\circ u_{ik}
\mathbf{\Omega}_{ji}\right)
\tr\left\{
\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\circ(\mathbf{Q}_{jk}
\mathbf{Q}_{ii})\right\}\\
&\quad+2
\tr\left(
\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\circ u_{ij}
\mathbf{\Omega}_{ki}\right)
\tr\left\{\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\circ(\mathbf{Q}_{kj}
\mathbf{Q}_{ii})\right\}\\
&\quad+\frac{1}{2}\tr\left(
\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ u_{jk}
\mathbf{\Omega}_{jk}\right)
\tr\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{jk}^{2}\right)
\\
&\quad+\frac{1}{4}\tr\left(
\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ u_{kk}
\mathbf{\Omega}_{jj}\right)
\tr\left\{\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ(\mathbf{Q}_{jk}\mathbf{Q}_{kj}
)\right\}\\
&\quad+\frac{1}{4}\tr\left(
\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\circ u_{jj}
\mathbf{\Omega}_{kk}\right)
\tr\left\{\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\circ(\mathbf{Q}_{kj}\mathbf{Q}_{jk}
)\right\}\big],\\
G_{5_{4}}=&2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
\tr\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}\circ
\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\tr\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}\circ
\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right),\\
G_{5_{5}}=&2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
2\big\{4u_{jk}\tr\left(\mathbf{Q}_{jk}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jk}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jk}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2u_{jk}\tr\left(\mathbf{Q}_{jk}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\mathbf{Q}_{jk}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jk}\right)\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jk}\right)\mathbf{Q}_{jk}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jk}\right)\mathbf{\Omega}_{kj}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jk}\right)\mathbf{Q}_{kj}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jk}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ \mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jk}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ \mathbf{Q}_{jk}\right)\tau_{T}
\big\}\\
&+4\big\{2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
u_{ik}\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jk}\right)
\mathbf{Q}_{jk}\mathbf{Q}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
u_{ij}\mathbf{\Omega}_{ki}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jk}\right)
\mathbf{Q}_{kj}\mathbf{Q}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
u_{ii}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jk}\right)
\mathbf{Q}_{ii}^{2}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jk}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{jk}\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}^{2}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{kk}\mathbf{\Omega}_{jj}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}\mathbf{Q}_{kj}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{jj}\mathbf{\Omega}_{kk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{kj}\mathbf{Q}_{jk}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ik}\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}\mathbf{Q}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jk}\right)\tau_{T} \\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
u_{ij}\mathbf{\Omega}_{ki}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{kj}\mathbf{Q}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{Q}_{jk}\right)
\tau_{T}
\big\}\\
&+4\big\{2u_{jk}\tr\left(\mathbf{Q}_{jk}\right)
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}
\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}
\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}
\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}
\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
u_{ii}\tr\left(\mathbf{Q}_{ii}\right)
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{kj}
\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{kj}
\circ\mathbf{Q}_{ii}\right)\tau_{T}
\big\}\\
&+8\big\{\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{kj}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{kj}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{jk}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{jk}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{kj}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{kj}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{jk}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{kj}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{kj}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{jk}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{kj}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{kj}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{ii}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{ii}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{jk}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{jk}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\mathbf{\Omega}_{kj}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\mathbf{Q}_{kj}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)\tau_{T}
\big\}\\
&+16\big[
\tr\left\{\left(\mathbf{\Omega}_{jk}\circ
\mathbf{\Omega}_{ii}\right)u_{ij}\mathbf{\Omega}_{ik}\right\}
\tr\left\{\left(\mathbf{Q}_{jk}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\mathbf{Q}_{jk}\right\}\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{jk}\circ
\mathbf{\Omega}_{ii}\right)u_{ik}\mathbf{\Omega}_{ij}\right\}
\tr\left\{\left(\mathbf{Q}_{jk}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\mathbf{Q}_{kj}\right\}
\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{kj}\circ
\mathbf{\Omega}_{ii}\right)u_{ij}\mathbf{\Omega}_{ik}\right\}
\tr\left\{\left(\mathbf{Q}_{kj}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\mathbf{Q}_{jk}\right\}
\\
&\quad+
\tr\left\{\left(\mathbf{\Omega}_{kj}\circ
\mathbf{\Omega}_{ii}\right)u_{ik}\mathbf{\Omega}_{ij}\right\}
\tr\left\{\left(\mathbf{Q}_{kj}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}\mathbf{Q}_{kj}\right\}
\\
&\quad+
\frac{1}{2}
\tr\left\{\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)u_{jk}\mathbf{\Omega}_{jk}\right\}
\tr\left\{\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{jk}^{2}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)u_{kk}\mathbf{\Omega}_{jj}\right\}
\tr\left\{\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{jk}\mathbf{Q}_{kj}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{\Omega}_{ii}\circ
\mathbf{\Omega}_{ii}\right)u_{jj}\mathbf{\Omega}_{kk}\right\}
\tr\left\{\left(\mathbf{Q}_{ii}\circ
\mathbf{Q}_{ii}\right)\mathbf{Q}_{kj}\mathbf{Q}_{jk}\right\}\\
&\quad+
\frac{1}{2}
\tr\left\{\left(\mathbf{\Omega}_{jk}\circ
\mathbf{\Omega}_{jk}\right)u_{ii}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{Q}_{jk}\circ
\mathbf{Q}_{jk}\right)\mathbf{Q}_{ii}^{2}\right\}\\
&\quad+
\frac{1}{2}
\tr\left\{\left(\mathbf{\Omega}_{jk}\circ
\mathbf{\Omega}_{kj}\right)u_{ii}\mathbf{\Omega}_{ii}\right\}
\tr\left\{\left(\mathbf{Q}_{jk}\circ
\mathbf{Q}_{kj}\right)\mathbf{Q}_{ii}^{2}\right\}\big],\\
G_{5_{6}}=&2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
2\tr^{2}\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)
\tr^{2}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)\\
&+4\big\{2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)
\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)
\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
\left(\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)
\left(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{kj}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{kj}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\tau_{T}\big\}\\
&+4
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}
\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}
\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&+4
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{kj}
\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{kj}
\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T},\\
G_{5_{7}}=&2\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
2\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)\mathbf{Q}_{jk}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\mathbf{\Omega}_{kj}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)\mathbf{Q}_{kj}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{jk}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{kj}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{kj}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)\mathbf{Q}_{ii}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}\right)\tau_{T}
\big\}\\
&+8\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)
\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{jk}\right)
\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{ii}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{jk}\right)
\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{jk}\circ\mathbf{Q}_{kj}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{jk}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{jk}\circ\mathbf{Q}_{ii}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ii}\right)
\left(\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{kj}\circ\mathbf{\Omega}_{ii}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ii}\right)
\left(\mathbf{Q}_{kj}\circ\mathbf{Q}_{kj}\circ\mathbf{Q}_{ii}\right)\tau_{T}
\big\}.
\end{align*}
Because $\underset{i\neq j\neq k}{\sum\sum\sum}|u_{ij}u_{jk}|=O(N^{1\vee(3-2/\tau)})$
and
$\underset{i\neq j\neq k}{\sum\sum\sum}u_{ij}^{2}=O(N^{2}),$
we have $G_{5_{1}}=O(N^{1\vee(3-2/\tau)}T^{3})+O(N^{2}T^{2}).$
Similarly,
due to $\underset{i\neq j\neq k}{\sum\sum}|u_{ij}|=O(N^{2\vee (3-1/\tau)}),$
and Lemma \ref{lem:sigma},
we have $G_{5_{2}}=O(N^{2\vee (3-1/\tau)}T^{3})+O(N^{3}T^{2}),$
$G_{5_{3}}=O(N^{2\vee (3-1/\tau)}T^{2})+O(N^{3}T),$
$G_{5_{4}}=O(N^{3}T),$
$G_{5_{5}}=O(N^{3}T^{5/2}),$
$G_{5_{6}}=O(N^{3}T^{2})$
and $G_{5_{7}}=O(N^{3}T^{2}).$
Combining the above calculations, we can get $G_{5}= O(N^{1\vee(3-2/\tau)}T^{3})+ O(N^{3}T^{5/2})$. We next deal with $G_{6}$.
Similarly, due to Lemma \ref{le:moment of quadratic form}, we have
$$
G_{6}=G_{6_{1}}+\gamma_{2}G_{6_{2}}+\gamma_{4}G_{6_{3}}+\gamma_{6}G_{6_{4}}
+\gamma_{1}^{2}G_{6_{5}}+\gamma_{2}^{2}G_{6_{6}}+\gamma_{1}\gamma_{3}G_{6_{7}},$$
where
\begin{align*}
G_{6_{1}}=&4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
4u_{ij}^{2}u_{ik}^{2}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\right)
+4u_{ij}u_{ik}u_{jk}u_{ii}
\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ki}\right)\\
&+4u_{ij}u_{ik}u_{jk}u_{ii}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ki}\right)+
2u_{ij}^{2}u_{ik}^{2}
\tr^{2}\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\right)\\
&+2u_{jk}^{2}u_{ii}^{2}
\tr^{2}\left(\mathbf{Q}_{ij}\mathbf{Q}_{ki}\right)
+4u_{ij}^{2}u_{ik}^{2}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}^{2}\right)\\
&+4u_{jk}u_{ik}u_{ii}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ki}^{2}\right)+
4u_{ij}^{2}u_{ii}u_{kk}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\mathbf{Q}_{ki}\right)
\\
&+4u_{ik}u_{jk}u_{ij}u_{ii}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ki}\mathbf{Q}_{ik}\right)
+4u_{ik}^{2}u_{ij}^{2}
\tr\left(\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{Q}_{ik}\mathbf{Q}_{ij}\mathbf{Q}_{ij}\right)\\
&+4u_{ik}u_{jk}u_{ii}u_{ij}
\tr\left(\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{Q}_{ik}\mathbf{Q}_{ji}\mathbf{Q}_{ji}\right)
+4u_{ik}^{2}u_{ii}u_{jj}
\tr\left(\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{Q}_{ik}\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right)\\
&+4u_{jk}u_{ii}u_{ij}u_{ik}
\tr\left(\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{Q}_{ik}\mathbf{Q}_{ji}\mathbf{Q}_{ij}\right)
+4u_{ij}^{2}u_{ik}^{2}
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{ik}^{2}\right)\\
&+4u_{ij}^{2}u_{jk}u_{ik}u_{ii}
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{ki}^{2}\right)
+4u_{ij}^{2}u_{kk}u_{ii}
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{ik}\mathbf{Q}_{ki}\right)\\
&+4u_{ij}u_{jk}u_{ii}u_{ik}
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{ki}\mathbf{Q}_{ik}\right)
+4u_{jj}u_{ii}u_{ik}^{2}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\mathbf{Q}_{ik}^{2}\right)\\
&+4u_{ii}u_{ij}u_{ik}u_{jk}
\tr\left(\mathbf{Q}_{ji}\mathbf{Q}_{ij}\mathbf{Q}_{ik}^{2}\right)
+2u_{uu}u_{ij}^{2}u_{kk}
\tr\left(\mathbf{Q}_{ji}\mathbf{Q}_{ij}\mathbf{Q}_{ik}\mathbf{Q}_{ki}\right)\\
&+2u_{ii}^{2}u_{jk}^{2}
\tr\left(\mathbf{Q}_{ji}\mathbf{Q}_{ij}\mathbf{Q}_{ki}\mathbf{Q}_{ik}\right)
+2u_{ij}^{2}u_{ik}^{2}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\mathbf{Q}_{ij}\mathbf{Q}_{ik}\right)\\
&+4u_{ij}u_{ik}u_{jk}u_{ii}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\mathbf{Q}_{ij}\mathbf{Q}_{ki}\right)
+4u_{ij}u_{kj}u_{ii}u_{ik}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\mathbf{Q}_{ji}\mathbf{Q}_{ik}\right)\\
&+2u_{ij}u_{kj}u_{ik}u_{ii}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\mathbf{Q}_{ji}\mathbf{Q}_{ki}\right)
+2u_{jk}^{2}u_{ii}^{2}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ki}\mathbf{Q}_{ij}\mathbf{Q}_{ki}\right)\\
&+2u_{jk}u_{ij}u_{ii}u_{ik}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ki}\mathbf{Q}_{ji}\mathbf{Q}_{ik}\right)
+2u_{jj}u_{ii}^{2}u_{kk}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\mathbf{Q}_{ik}\mathbf{Q}_{ki}\right)
\\
&+2u_{jj}u_{ik}^{2}u_{ii}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\mathbf{Q}_{ki}\mathbf{Q}_{ik}\right),\\
G_{6_{2}}=&4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
4u_{ij}u_{ik}\tr\left(\mathbf{Q}_{ij}
\right)\tr\left(\mathbf{Q}_{ik}
\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}
\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}
\right)\\
&+2\big\{2u_{ij}u_{ik}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}
\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}
\right)\\
&\quad+2
u_{jk}u_{ii}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ki}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}
\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}
\right)
\big\}\\
&+4\big[
u_{ij}\tr\left(\mathbf{Q}_{ij}\right)\tr\left(\mathbf{\Omega}_{ij}
\circ u_{ik}\mathbf{\Omega}_{ik}\right)
\tr\left(\mathbf{Q}_{ij}
\circ \mathbf{Q}_{ik}^{2}\right)\\
&\quad+\frac{1}{2}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{kk}\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ik}\mathbf{Q}_{ki})\right\}\\
&\quad+\frac{1}{2}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{ii}\mathbf{\Omega}_{kk}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ki}\mathbf{Q}_{ik})\right\}\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ik}
\circ u_{ij}\mathbf{\Omega}_{ik}\right)
\tr\left\{\mathbf{Q}_{ik}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ik})\right\}
\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ik}
\circ u_{jk}\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ik}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ki})\right\}
\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ik}
\circ u_{ii}\mathbf{\Omega}_{jk}\right)
\tr\left\{\mathbf{Q}_{ik}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ik})\right\}
\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ik}
\circ u_{ik}\mathbf{\Omega}_{ji}\right)
\tr\left\{\mathbf{Q}_{ik}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ki})\right\}\\
&\quad+u_{ik}\tr\left(\mathbf{Q}_{ik}
\right)
\tr\left(\mathbf{\Omega}_{ik}
\circ u_{ij}\mathbf{\Omega}_{ij}\right)
\tr\left(\mathbf{Q}_{ik}
\circ \mathbf{Q}_{ij}^{2}\right)
\\
&\quad+\frac{1}{2}u_{ik}\tr\left(\mathbf{Q}_{ik}
\right)
\tr\left(\mathbf{\Omega}_{ik}
\circ u_{jj}\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ik}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}
\\
&\quad+\frac{1}{2}u_{ik}\tr\left(\mathbf{Q}_{ik}
\right)
\tr\left(\mathbf{\Omega}_{ik}
\circ u_{ii}\mathbf{\Omega}_{jj}\right)
\tr\left\{\mathbf{Q}_{ik}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}\\
&\quad+
u_{ik}\tr\left(\mathbf{Q}_{ik}
\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{ij}\mathbf{\Omega}_{ik}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ik})\right\}\\
&\quad+
u_{ik}\tr\left(\mathbf{Q}_{ik}
\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{jk}\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ki})\right\}\\
&\quad+
u_{ik}\tr\left(\mathbf{Q}_{ik}
\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{ii}\mathbf{\Omega}_{jk}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ik})\right\}\\
&\quad+
u_{ik}\tr\left(\mathbf{Q}_{ik}
\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{ik}\mathbf{\Omega}_{ji}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ki})\right\}
\big]\\
&+8\big\{\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ij}u_{ik}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ik}
\mathbf{Q}_{ik}\right\}\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{jk}u_{ik}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ki}
\mathbf{Q}_{ki}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ji}u_{kk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ik}
\mathbf{Q}_{ki}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{jk}u_{ii}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ki}
\mathbf{Q}_{ik}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ii}u_{ki}\mathbf{\Omega}_{jk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ik}
\mathbf{Q}_{ik}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}^{2}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ki}
\mathbf{Q}_{ki}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ii}u_{kk}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ik}
\mathbf{Q}_{ki}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{ii}\mathbf{\Omega}_{jk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ki}
\mathbf{Q}_{ik}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ij}^{2}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\mathbf{Q}_{ik}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ij}u_{ii}\mathbf{\Omega}_{jk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\mathbf{Q}_{ik}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{jj}u_{ii}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\mathbf{Q}_{ik}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ii}u_{ji}\mathbf{\Omega}_{jk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\mathbf{Q}_{ik}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ij}u_{jk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\mathbf{Q}_{ki}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ij}u_{ik}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\mathbf{Q}_{ki}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{jj}u_{ik}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\mathbf{Q}_{ki}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ii}u_{jk}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\mathbf{Q}_{ki}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{ji}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ij}
\mathbf{Q}_{ik}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{jk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ij}
\mathbf{Q}_{ki}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{jk}u_{ii}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ji}
\mathbf{Q}_{ik}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ii}u_{ji}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ij}
\mathbf{Q}_{ik}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ii}u_{jk}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ij}
\mathbf{Q}_{ki}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ii}u_{ji}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ji}
\mathbf{Q}_{ik}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{ji}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ji}
\mathbf{Q}_{ki}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ik}u_{ji}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ik}
\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ji}u_{kj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ik}
\mathbf{Q}_{ji}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{jk}u_{ii}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ki}
\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{jk}u_{ij}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ki}
\mathbf{Q}_{ji}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ii}u_{ik}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ik}
\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ii}u_{kj}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ik}
\mathbf{Q}_{ji}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ik}u_{ii}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ki}
\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ik}\right)u_{ik}u_{ji}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ki}
\mathbf{Q}_{ji}\right\}
\big\}\\
&+16\big[\frac{1}{4}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}
\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}^{2}
\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}^{2}
\right)\tau_{T}\\
&\quad+
\frac{1}{8}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}
\right)
\left(\mathbf{I}_{N}\circ u_{kk}\mathbf{\Omega}_{ii}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}^{2}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ik}\mathbf{Q}_{ki})
\right\}\tau_{T}\\
&\quad+
\frac{1}{8}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}
\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{kk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}^{2}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ki}\mathbf{Q}_{ik})
\right\}\tau_{T}\\
&\quad+
\frac{1}{16}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}
\right)
\left(\mathbf{I}_{N}\circ u_{kk}\mathbf{\Omega}_{ii}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ik}\mathbf{Q}_{ki})
\right\}\tau_{T}\\
&\quad+
\frac{1}{16}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}
\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{kk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ki}\mathbf{Q}_{ik}
)\right\}\tau_{T}\\
&\quad+
\frac{1}{16}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}
\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{kk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ki}\mathbf{Q}_{ik}
)\right\}\tau_{T}\\
&\quad+
\frac{1}{16}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}
\right)
\left(\mathbf{I}_{N}\circ u_{kk}\mathbf{\Omega}_{ii}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ik}\mathbf{Q}_{ki}
)\right\}\tau_{T}\\
&\quad+
\frac{1}{8}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}^{2}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji}
)\right\}\tau_{T}\\
&\quad+
\frac{1}{8}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}^{2}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij}
)\right\}\tau_{T}\\
&\quad+\frac{1}{8}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ji}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{ji}\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ik})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ik})
\right\}\tau_{T}
\\
&\quad+\frac{1}{8}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{ii}
\right)
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{ii}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ki})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ki})
\right\}\tau_{T}
\\
&\quad+\frac{1}{8}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jk}
\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ik})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ik})
\right\}\tau_{T}
\\
&\quad+\frac{1}{8}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ji}
\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ji}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ki})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ki})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ji}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{ii}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ik})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ki})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ji}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ik})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ik})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ji}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ji}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ik})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ki})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{ii}
\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ki})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ik})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{ii}
\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ji}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ki})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ki})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jk}
\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{ji}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ik})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ki})
\right\}\tau_{T}
\big],\\
G_{6_{3}}=&4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}
\circ\mathbf{\Omega}_{ik}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}
\circ\mathbf{Q}_{ik}\right)
\\
&+2u_{ik}\tr\left(\mathbf{Q}_{ik}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}
\circ\mathbf{\Omega}_{ik}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}
\circ\mathbf{Q}_{ik}\right)\\
&+4\big[
\frac{1}{2}
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
u_{ik}\mathbf{\Omega}_{ik}
\right)
\tr\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}^{2}
\right)\\
&\quad+\frac{1}{4}
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
u_{kk}\mathbf{\Omega}_{ii}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
(\mathbf{Q}_{ik}\mathbf{Q}_{ki})
\right\}\\
&\quad+\frac{1}{4}
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
u_{ii}\mathbf{\Omega}_{kk}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
(\mathbf{Q}_{ki}\mathbf{Q}_{ik})
\right\}\\
&\quad+\frac{1}{2}
\tr\left(
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\circ
u_{ij}\mathbf{\Omega}_{ij}
\right)
\tr\left(
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ij}^{2}
\right)\\
&\quad+\frac{1}{4}
\tr\left(
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\circ
u_{jj}\mathbf{\Omega}_{ii}
\right)
\tr\left\{
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\circ
(\mathbf{Q}_{ij}\mathbf{Q}_{ji})
\right\}\\
&\quad+\frac{1}{4}
\tr\left(
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\circ
u_{ii}\mathbf{\Omega}_{jj}
\right)
\tr\left\{
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\circ
(\mathbf{Q}_{ji}\mathbf{Q}_{ij})
\right\}\\
&\quad+
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\circ
u_{ji}\mathbf{\Omega}_{ik}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\circ
(\mathbf{Q}_{ij}\mathbf{Q}_{ik})
\right\}\\
&\quad+
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\circ
u_{jk}\mathbf{\Omega}_{ii}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\circ
(\mathbf{Q}_{ij}\mathbf{Q}_{ki})
\right\}\\
&\quad+
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\circ
u_{ii}\mathbf{\Omega}_{jk}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\circ
(\mathbf{Q}_{ji}\mathbf{Q}_{ik})
\right\}\\
&\quad+
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\circ
u_{ik}\mathbf{\Omega}_{ji}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\circ
(\mathbf{Q}_{ji}\mathbf{Q}_{ki})
\right\}
\big],\\
G_{6_{4}}=&4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\right),\\
G_{6_{5}}=&4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
2\big\{2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ik}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ik}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ki}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ki}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+2u_{ik}\tr\left(\mathbf{Q}_{ik}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+2u_{ik}\tr\left(\mathbf{Q}_{ik}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ji}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+2u_{ik}\tr\left(\mathbf{Q}_{ik}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ik}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ik}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\big\}\\
&+4\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ji}\mathbf{\Omega}_{ik}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ik}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{jk}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ki}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ii}\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ik}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ik}\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ki}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
u_{ji}\mathbf{\Omega}_{ik}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ik}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
u_{jk}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ki}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
u_{ii}\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ik}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
u_{ik}\mathbf{\Omega}_{ji}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{ki}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ik}\mathbf{\Omega}_{ik}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ik}\mathbf{Q}_{ik}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{kk}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ik}\mathbf{Q}_{ki}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
u_{ij}\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
u_{jj}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}
\big\}\\
&+4\big\{\frac{1}{2}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}
\right)\tau_{T}\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ki}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\circ\mathbf{Q}_{ki}
\right)\tau_{T}\\
&\quad+
\frac{1}{2}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ki}\circ\mathbf{\Omega}_{ki}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ki}\circ\mathbf{Q}_{ki}
\right)\tau_{T}\\
&\quad+
\frac{1}{2}u_{ik}\tr\left(\mathbf{Q}_{ik}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ik}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}
\right)\tau_{T}\\
&\quad+u_{ik}\tr\left(\mathbf{Q}_{ik}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ik}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}
\right)\tau_{T}\\
&\quad+
\frac{1}{2}u_{ik}\tr\left(\mathbf{Q}_{ik}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ik}\circ\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}
\right)\tau_{T}\big\}\\
&+8\big\{\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ki}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ki}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ki}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ki}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ik}\right)
\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ik}\right)
\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ki}\right)
\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ki}\right)
\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\big\}\\
&+16\gamma_{1}^{2}\big[
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}\right)u_{ik}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}\right)u_{jk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}\right)u_{ii}\mathbf{\Omega}_{kj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}\right)u_{ji}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ki}\right)u_{ik}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ki}\right)u_{jk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ki}\right)u_{ii}\mathbf{\Omega}_{kj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ki}\right)u_{ji}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ik}\right)u_{ik}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ik}\right)u_{jk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ik}\right)u_{ii}\mathbf{\Omega}_{kj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ik}\right)u_{ji}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ki}\right)u_{ik}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ki}\right)u_{jk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ki}\right)u_{ii}\mathbf{\Omega}_{kj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ki}\right)u_{ji}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ik}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{kk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ki}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{ii}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ik}
\right\}\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ki}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{ik}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ik}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{kk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ki}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{ii}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ik}
\right\}\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{ik}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ki}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)u_{ik}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ik}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)u_{kk}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ik}\mathbf{Q}_{ki}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)u_{ii}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ik}
\right\}\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)u_{ik}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{ki}\mathbf{Q}_{ki}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ik}\right)u_{ij}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ik}\right)u_{jj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ik}\right)u_{ii}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ik}\right)u_{ij}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ik}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ki}\right)u_{ij}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ki}\right)u_{jj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ki}\right)u_{ii}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ki}\right)u_{ij}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ki}\circ
\mathbf{\Omega}_{ki}\right)u_{ij}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{ki}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ki}\circ
\mathbf{\Omega}_{ki}\right)u_{jj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{ki}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ki}\circ
\mathbf{\Omega}_{ki}\right)u_{ii}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ki}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ki}\circ
\mathbf{\Omega}_{ki}\right)u_{ij}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{Q}_{ki}\circ
\mathbf{Q}_{ki}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\right\}
\big],\\
G_{6_{6}}=&4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
2\tr^{2}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\right)
\tr^{2}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\right)\\
&+4\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ki}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ki}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ik}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ik}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ki}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{ki}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ki}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ik}\circ\mathbf{Q}_{ki}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\right)\tau_{T}
\big\}\\
&+\big\{
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}
\circ\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}
\circ\mathbf{Q}_{ik}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ik}
\circ\mathbf{\Omega}_{ki}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ik}
\circ\mathbf{Q}_{ki}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ki}
\circ\mathbf{\Omega}_{ki}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ki}
\circ\mathbf{Q}_{ki}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ik}
\circ\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ik}
\circ\mathbf{Q}_{ik}
\right)\tau_{T}\\
&\quad+
2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ik}
\circ\mathbf{\Omega}_{ki}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ik}
\circ\mathbf{Q}_{ki}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ki}
\circ\mathbf{\Omega}_{ki}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ki}
\circ\mathbf{Q}_{ki}
\right)\tau_{T}
\big\},\\
G_{6_{7}}=&4\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}
2\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}
\right)\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}
\right)\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ik}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}
\right)\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}
\right)\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ik}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}
\right)\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}
\right)\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}
\right)\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}
\right)\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}
\right)\mathbf{\Omega}_{ik}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}
\right)\mathbf{Q}_{ik}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}
\right)\mathbf{\Omega}_{ki}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}
\right)\mathbf{Q}_{ki}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}
\right)\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}
\right)\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ik}
\right)\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ik}
\right)\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}
\right)\tau_{T}\big\}\\
&+\big\{2\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ki}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ki}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ki}\circ\mathbf{\Omega}_{ki}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ki}\circ\mathbf{Q}_{ki}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ik}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ik}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ik}\circ\mathbf{\Omega}_{ki}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ik}\circ\mathbf{Q}_{ki}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ki}\circ\mathbf{\Omega}_{ki}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ki}\circ\mathbf{Q}_{ki}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ik}\right)\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ik}\right)\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ik}\right)\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ik}\right)\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ik}\right)\left(\mathbf{\Omega}_{ik}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ik}\right)\left(\mathbf{Q}_{ik}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ik}\right)\left(\mathbf{\Omega}_{ki}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ik}\right)\left(\mathbf{Q}_{ki}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ik}\right)\left(\mathbf{\Omega}_{ki}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ik}\right)\left(\mathbf{Q}_{ki}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ik}\right)\left(\mathbf{\Omega}_{ki}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ik}\right)\left(\mathbf{Q}_{ki}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)\tau_{T}
\big\}.
\end{align*}
Due to $$\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}u_{ij}^{2}u_{ik}^{2}\leq\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}u_{ij}^{2}=O(N^{2})$$
and
$$\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}|u_{ij}u_{kj}u_{ik}|\leq\underset{i\neq j\neq k}{\sum^{N}\sum^{N}\sum^{N}}|u_{ij}u_{kj}|=O(N^{1\vee(3-2/\tau)}),$$
we have $G_{6_{1}}=O(N^{2}T^{3}+N^{1\vee(3-2/\tau)}T^{3}).$
Similarly,
due to $\underset{i\neq j\neq k}{\sum\sum}|u_{ij}|=O(N^{2\vee (3-1/\tau)})$
and Lemma \ref{lem:sigma},
we have $G_{6_{2}}=O(N^{1\vee(3-2/\tau)}T^{3})+O(N^{2\vee(3-1/\tau)}T^{2})+O(N^{3}T),$
$G_{6_{3}}=O(N^{2\vee(3-1/\tau)}T^{2})+O(N^{3}T),$
$G_{6_{4}}=O(N^{3}T),$
$G_{6_{5}}=O(N^{2\vee(3-1/\tau)}T^{5/2})+O(N^{3}T^{2}),$
$G_{6_{6}}=O(N^{3}T^{2})$
and $G_{6_{7}}=O(N^{3}T^{2}).$
Combining the above calculations, we can get $G_{6}= O(N^{1\vee(3-2/\tau)}T^{3})+ O(N^{3}T^{2}+O(N^{2\vee(3-1/\tau)}T^{5/2})$. We next deal with $G_{7}$.
Similarly, due to Lemma \ref{le:moment of quadratic form}, we have
$$
G_{7}=G_{7_{1}}+\gamma_{2}G_{7_{2}}+\gamma_{4}G_{7_{3}}+\gamma_{6}G_{7_{4}}
+\gamma_{1}^{2}G_{7_{5}}+\gamma_{2}^{2}G_{7_{6}}+\gamma_{1}\gamma_{3}G_{7_{7}},$$
where
\begin{align*}
G_{7_{1}}=&\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
4u_{ij}u_{kl}u_{kj}u_{il}\tr\left(\mathbf{Q}_{ij}
\right)\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{Q}_{ji}\mathbf{Q}_{lk}
\right)\\
&+4u_{ij}u_{kl}u_{ki}u_{lj}
\tr\left(\mathbf{Q}_{ij}
\right)\tr\left(\mathbf{Q}_{kl}
\right)\tr\left(\mathbf{Q}_{lk}\mathbf{Q}_{ij}
\right)+2u_{kj}^{2}u_{il}^{2}\tr^{2}\left(\mathbf{Q}_{ji}\mathbf{Q}_{lk}\right)\\
&
+2u_{ik}^{2}u_{lj}^{2}\tr^{2}\left(\mathbf{Q}_{ij}\mathbf{Q}_{lk}\right)+4u_{il}u_{ik}u_{lj}u_{kj}
\tr\left(\mathbf{Q}_{ji}\mathbf{Q}_{lk}\right)
\tr\left(\mathbf{Q}_{lk}\mathbf{Q}_{ij}\right)
\\
&+4u_{ij}u_{jk}u_{lk}u_{il}
\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{kl}^{2}\right)+4u_{ij}u_{jl}u_{kl}u_{ik}
\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{lk}^{2}\right)
\\
&
+4u_{jk}u_{ll}u_{ik}u_{ij}
\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{kl}\mathbf{Q}_{lk}\right)
+4u_{jl}u_{il}u_{kk}u_{ij}
\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{lk}\mathbf{Q}_{kl}\right)
\\
&+4u_{kl}u_{li}u_{ij}u_{kj}
\tr\left(\mathbf{Q}_{kl}\right)
\tr\left(\mathbf{Q}_{kl}\mathbf{Q}_{ij}^{2}\right)
+4u_{lj}u_{ij}u_{ik}u_{kl}
\tr\left(\mathbf{Q}_{kl}\right)
\tr\left(\mathbf{Q}_{kl}\mathbf{Q}_{ji}^{2}\right)
\\
&+4u_{li}u_{ik}u_{jj}u_{kl}
\tr\left(\mathbf{Q}_{kl}\right)
\tr\left(\mathbf{Q}_{kl}\mathbf{Q}_{ij}\mathbf{Q}_{ji}\right)
+4u_{ii}u_{kj}u_{lj}u_{kl}
\tr\left(\mathbf{Q}_{kl}\right)
\tr\left(\mathbf{Q}_{kl}\mathbf{Q}_{ji}\mathbf{Q}_{ij}\right)
\\
&+4u_{ij}u_{jk}u_{lk}u_{il}
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{kl}^{2}\right)
+4u_{ji}u_{jl}u_{kl}u_{ik}
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{lk}^{2}\right)
+4u_{ji}u_{jk}u_{ll}u_{ik}
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{kl}\mathbf{Q}_{lk}\right)\\
&+4u_{ij}u_{jl}u_{kk}u_{il}
\tr\left(\mathbf{Q}_{ij}^{2}\mathbf{Q}_{lk}\mathbf{Q}_{kl}\right)
+4u_{ii}u_{jk}u_{lk}u_{jl}
\tr\left(\mathbf{Q}_{ji}\mathbf{Q}_{ij}\mathbf{Q}_{kl}^{2}\right)\\
&+4u_{jj}u_{ik}u_{lk}u_{il}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\mathbf{Q}_{kl}^{2}\right)
+2u_{ii}u_{jk}^{2}u_{ll}
\tr\left(\mathbf{Q}_{ji}\mathbf{Q}_{ij}\mathbf{Q}_{kl}\mathbf{Q}_{lk}\right)\\
&+2u_{ii}u_{jl}^{2}u_{kk}
\tr\left(\mathbf{Q}_{ji}\mathbf{Q}_{ij}\mathbf{Q}_{lk}\mathbf{Q}_{kl}\right)
+2u_{jj}u_{ik}^{2}u_{ll}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\mathbf{Q}_{kl}\mathbf{Q}_{lk}\right)\\
&+2u_{jj}u_{il}^{2}u_{kk}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{ji}\mathbf{Q}_{lk}\mathbf{Q}_{kl}\right)
+2u_{jl}^{2}u_{ik}^{2}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{lk}\mathbf{Q}_{ij}\mathbf{Q}_{lk}\right)\\
&+2u_{jl}u_{kj}u_{ik}u_{il}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{lk}\mathbf{Q}_{ji}\mathbf{Q}_{kl}\right)
+4u_{jk}u_{lj}u_{ik}u_{il}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{kl}\mathbf{Q}_{ji}\mathbf{Q}_{kl}\right)
\\
&+2u_{jk}u_{lj}u_{il}u_{ik}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{kl}\mathbf{Q}_{ji}\mathbf{Q}_{lk}\right)
+4u_{jk}u_{li}u_{jl}u_{ik}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{kl}\mathbf{Q}_{ij}\mathbf{Q}_{lk}\right)\\
&+2u_{jk}^{2}u_{il}^{2}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{kl}\mathbf{Q}_{ij}\mathbf{Q}_{kl}\right),\\
G_{7_{2}}=&\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
4u_{ij}u_{kl}\tr\left(\mathbf{Q}_{ij}
\right)\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}
\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}
\right)\\
&+2\big\{2u_{jk}u_{il}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{kl}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}
\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}
\right)\\
&\quad+2
u_{jl}u_{ik}
\tr\left(\mathbf{Q}_{ij}\mathbf{Q}_{lk}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}
\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}
\right)
\big\}\\
&+4\big[
u_{ij}\tr\left(\mathbf{Q}_{ij}\right)\tr\left(\mathbf{\Omega}_{ij}
\circ u_{kl}\mathbf{\Omega}_{kl}\right)
\tr\left(\mathbf{Q}_{ij}
\circ \mathbf{Q}_{kl}^{2}\right)\\
&\quad+\frac{1}{2}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{ll}\mathbf{\Omega}_{kk}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{kl}\mathbf{Q}_{lk})\right\}\\
&\quad+\frac{1}{2}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{kk}\mathbf{\Omega}_{ll}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ( \mathbf{Q}_{lk}\mathbf{Q}_{kl})\right\}\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{kl}
\circ u_{jk}\mathbf{\Omega}_{il}\right)
\tr\left\{\mathbf{Q}_{kl}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{kl})\right\}
\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{kl}
\circ u_{jl}\mathbf{\Omega}_{ik}\right)
\tr\left\{\mathbf{Q}_{kl}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{lk})\right\}
\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{kl}
\circ u_{ik}\mathbf{\Omega}_{jl}\right)
\tr\left\{\mathbf{Q}_{kl}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{kl})\right\}
\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{kl}
\circ u_{il}\mathbf{\Omega}_{jk}\right)
\tr\left\{\mathbf{Q}_{kl}
\circ( \mathbf{Q}_{ji}\mathbf{Q}_{lk})\right\}\\
&\quad+u_{kl}\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{\Omega}_{kl}
\circ u_{ij}\mathbf{\Omega}_{ij}\right)
\tr\left(\mathbf{Q}_{kl}
\circ \mathbf{Q}_{ij}^{2}\right)
\\
&\quad+\frac{1}{2}u_{kl}\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{\Omega}_{kl}
\circ u_{jj}\mathbf{\Omega}_{ii}\right)
\tr\left\{\mathbf{Q}_{kl}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{ji})\right\}
\\
&\quad+\frac{1}{2}u_{kl}\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{\Omega}_{kl}
\circ u_{ii}\mathbf{\Omega}_{jj}\right)
\tr\left\{\mathbf{Q}_{kl}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{ij})\right\}\\
&\quad+
u_{kl}\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{jk}\mathbf{\Omega}_{il}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{kl})\right\}\\
&\quad+
u_{kl}\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{jl}\mathbf{\Omega}_{ik}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ij}\mathbf{Q}_{lk})\right\}\\
&\quad+
u_{kl}\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{ik}\mathbf{\Omega}_{jl}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{kl})\right\}\\
&\quad+
u_{kl}\tr\left(\mathbf{Q}_{kl}
\right)
\tr\left(\mathbf{\Omega}_{ij}
\circ u_{il}\mathbf{\Omega}_{jk}\right)
\tr\left\{\mathbf{Q}_{ij}
\circ (\mathbf{Q}_{ji}\mathbf{Q}_{lk})\right\}
\big]\\
&+8\big\{\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{jk}u_{lk}\mathbf{\Omega}_{il}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}\mathbf{Q}_{kl}
\mathbf{Q}_{kl}\right\}\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{jl}u_{kl}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}\mathbf{Q}_{lk}
\mathbf{Q}_{lk}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{jk}u_{ll}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}\mathbf{Q}_{kl}
\mathbf{Q}_{lk}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{jl}u_{kk}\mathbf{\Omega}_{il}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ij}\mathbf{Q}_{lk}
\mathbf{Q}_{kl}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{lk}\mathbf{\Omega}_{jl}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}\mathbf{Q}_{kl}
\mathbf{Q}_{kl}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{il}u_{kl}\mathbf{\Omega}_{jk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}\mathbf{Q}_{lk}
\mathbf{Q}_{lk}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{ll}\mathbf{\Omega}_{jk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}\mathbf{Q}_{kl}
\mathbf{Q}_{lk}\right\}
\\
&\quad+
\frac{1}{2}\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{il}u_{kk}\mathbf{\Omega}_{jl}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{ji}\mathbf{Q}_{lk}
\mathbf{Q}_{kl}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{ij}u_{jk}\mathbf{\Omega}_{il}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\mathbf{Q}_{kl}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{ij}u_{ik}\mathbf{\Omega}_{jl}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\mathbf{Q}_{kl}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{jj}u_{ik}\mathbf{\Omega}_{il}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\mathbf{Q}_{kl}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{ii}u_{jk}\mathbf{\Omega}_{jl}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\mathbf{Q}_{kl}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{ij}u_{jl}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\mathbf{Q}_{lk}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{ij}u_{il}\mathbf{\Omega}_{jk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\mathbf{Q}_{lk}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{jj}u_{il}\mathbf{\Omega}_{ik}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\mathbf{Q}_{lk}\right\}\\
&\quad+\frac{1}{2}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{ii}u_{jl}\mathbf{\Omega}_{jk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\mathbf{Q}_{lk}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{il}u_{jk}\mathbf{\Omega}_{kl}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ij}
\mathbf{Q}_{kl}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{il}u_{jl}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ij}
\mathbf{Q}_{lk}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{jl}u_{ik}\mathbf{\Omega}_{kl}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ji}
\mathbf{Q}_{kl}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{jk}\mathbf{\Omega}_{ll}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ij}
\mathbf{Q}_{kl}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{jl}\mathbf{\Omega}_{lk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ij}
\mathbf{Q}_{lk}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{ik}u_{jk}\mathbf{\Omega}_{ll}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ji}
\mathbf{Q}_{kl}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{ij}\right)u_{il}u_{jk}\mathbf{\Omega}_{lk}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ji}
\mathbf{Q}_{lk}\right\}\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{il}u_{jk}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{kl}
\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{jk}u_{lj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{kl}
\mathbf{Q}_{ji}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{jl}u_{ki}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{lk}
\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{jl}u_{kj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{lk}
\mathbf{Q}_{ji}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{ik}u_{il}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{kl}
\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{ik}u_{lj}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{kl}
\mathbf{Q}_{ji}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{il}u_{ik}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{lk}
\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}
\tr\left\{\left(\mathbf{I}_{N}\circ
\mathbf{\Omega}_{kl}\right)u_{il}u_{jk}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{I}_{T}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{lk}
\mathbf{Q}_{ji}\right\}
\big\}\\
&+16\big[\frac{1}{4}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}
\right)
\left(\mathbf{I}_{N}\circ u_{kl}\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}^{2}
\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}^{2}
\right)\tau_{T}\\
&\quad+
\frac{1}{8}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}
\right)
\left(\mathbf{I}_{N}\circ u_{ll}\mathbf{\Omega}_{kk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}^{2}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{kl}\mathbf{Q}_{lk}
)\right\}\tau_{T}\\
&\quad+
\frac{1}{8}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ij}\mathbf{\Omega}_{ij}
\right)
\left(\mathbf{I}_{N}\circ u_{kk}\mathbf{\Omega}_{ll}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}^{2}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{lk}\mathbf{Q}_{kl})
\right\}\tau_{T}\\
&\quad+
\frac{1}{16}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}
\right)
\left(\mathbf{I}_{N}\circ u_{ll}\mathbf{\Omega}_{kk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{kl}\mathbf{Q}_{lk})
\right\}\tau_{T}\\
&\quad+
\frac{1}{16}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}
\right)
\left(\mathbf{I}_{N}\circ u_{kk}\mathbf{\Omega}_{ll}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{lk}\mathbf{Q}_{kl})
\right\}\tau_{T}\\
&\quad+
\frac{1}{16}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}
\right)
\left(\mathbf{I}_{N}\circ u_{kk}\mathbf{\Omega}_{ll}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{lk}\mathbf{Q}_{kl}
)\right\}\tau_{T}\\
&\quad+
\frac{1}{16}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}
\right)
\left(\mathbf{I}_{N}\circ u_{ll}\mathbf{\Omega}_{kk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{kl}\mathbf{Q}_{lk}
)\right\}\tau_{T}\\
&\quad+
\frac{1}{8}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{kl}\mathbf{\Omega}_{kl}
\right)
\left(\mathbf{I}_{N}\circ u_{jj}\mathbf{\Omega}_{ii}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}^{2}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{ji})
\right\}\tau_{T}\\
&\quad+
\frac{1}{8}\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{kl}\mathbf{\Omega}_{kl}
\right)
\left(\mathbf{I}_{N}\circ u_{ii}\mathbf{\Omega}_{jj}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}^{2}
\right)
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{ij})
\right\}\tau_{T}\\
&\quad+\frac{1}{8}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{il}
\right)
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{il}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{kl})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{kl}
)\right\}\tau_{T}
\\
&\quad+\frac{1}{8}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jl}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{jl}\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{lk}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{lk}
)\right\}\tau_{T}
\\
&\quad+\frac{1}{8}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{jl}
\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{jl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{kl}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{kl}
)\right\}\tau_{T}
\\
&\quad+\frac{1}{8}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{il}\mathbf{\Omega}_{jk}
\right)
\left(\mathbf{I}_{N}\circ u_{il}\mathbf{\Omega}_{jk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{lk})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{lk})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{il}
\right)
\left(\mathbf{I}_{N}\circ u_{jl}\mathbf{\Omega}_{ik}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{kl})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{lk})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{il}
\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{jl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{kl})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{kl})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jk}\mathbf{\Omega}_{il}
\right)
\left(\mathbf{I}_{N}\circ u_{il}\mathbf{\Omega}_{jk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{kl})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{lk})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jl}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{jl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{lk})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{kl})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{jl}\mathbf{\Omega}_{ik}
\right)
\left(\mathbf{I}_{N}\circ u_{il}\mathbf{\Omega}_{jk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ij}\mathbf{Q}_{lk})
\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{lk})
\right\}\tau_{T}
\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ u_{ik}\mathbf{\Omega}_{jl}
\right)
\left(\mathbf{I}_{N}\circ u_{il}\mathbf{\Omega}_{jk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{kl}
)\right\}
\left\{\mathbf{I}_{T}\circ(\mathbf{Q}_{ji}\mathbf{Q}_{lk}
)\right\}\tau_{T}
\big],\\
G_{7_{3}}=&\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}
\circ\mathbf{\Omega}_{kl}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}
\circ\mathbf{Q}_{kl}\right)
\\
&+2u_{kl}\tr\left(\mathbf{Q}_{kl}\right)
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}
\circ\mathbf{\Omega}_{kl}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}
\circ\mathbf{Q}_{kl}\right)\\
&+4\big[
\frac{1}{2}
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
u_{kl}\mathbf{\Omega}_{kl}
\right)
\tr\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}^{2}
\right)\\
&\quad+\frac{1}{4}
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
u_{ll}\mathbf{\Omega}_{kk}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
(\mathbf{Q}_{kl}\mathbf{Q}_{lk})
\right\}\\
&\quad+\frac{1}{4}
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
u_{kk}\mathbf{\Omega}_{ll}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
(\mathbf{Q}_{lk}\mathbf{Q}_{kl})
\right\}\\
&\quad+\frac{1}{2}
\tr\left(
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\circ
u_{ij}\mathbf{\Omega}_{ij}
\right)
\tr\left(
\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\circ
\mathbf{Q}_{ij}^{2}
\right)\\
&\quad+\frac{1}{4}
\tr\left(
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\circ
u_{jj}\mathbf{\Omega}_{ii}
\right)
\tr\left\{
\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\circ
(\mathbf{Q}_{ij}\mathbf{Q}_{ji})
\right\}\\
&\quad+\frac{1}{4}
\tr\left(
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\circ
u_{ii}\mathbf{\Omega}_{jj}
\right)
\tr\left\{
\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\circ
(\mathbf{Q}_{ji}\mathbf{Q}_{ij})
\right\}\\
&\quad+
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\circ
u_{jk}\mathbf{\Omega}_{il}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\circ
(\mathbf{Q}_{ij}\mathbf{Q}_{kl})
\right\}\\
&\quad+
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\circ
u_{jl}\mathbf{\Omega}_{ik}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\circ
(\mathbf{Q}_{ij}\mathbf{Q}_{lk})
\right\}\\
&\quad+
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\circ
u_{ik}\mathbf{\Omega}_{jl}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\circ
(\mathbf{Q}_{ji}\mathbf{Q}_{kl})
\right\}\\
&\quad+
\tr\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\circ
u_{il}\mathbf{\Omega}_{jk}
\right)
\tr\left\{
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\circ
(\mathbf{Q}_{ji}\mathbf{Q}_{lk})
\right\}
\big],\\
G_{7_{4}}=&\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
\tr\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\right)
\tr\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\right),\\
G_{7_{5}}=&\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
2\big\{2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{kl}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{kl}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{lk}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{lk}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+2u_{kl}\tr\left(\mathbf{Q}_{kl}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+2u_{kl}\tr\left(\mathbf{Q}_{kl}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{ji}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+2u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{ij}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ij}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+2u_{kl}\tr\left(\mathbf{Q}_{kl}\right)
\tau_{N}^{\prime}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{kl}\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{kl}\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\big\}\\
&+4\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{jk}\mathbf{\Omega}_{il}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{kl}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{jl}\mathbf{\Omega}_{ik}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{lk}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ik}\mathbf{\Omega}_{jl}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{kl}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{il}\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{lk}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
u_{jk}\mathbf{\Omega}_{il}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{kl}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
u_{jl}\mathbf{\Omega}_{ik}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{lk}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
u_{ik}\mathbf{\Omega}_{jl}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{kl}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
u_{il}\mathbf{\Omega}_{jk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ji}\mathbf{Q}_{lk}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{kl}\mathbf{\Omega}_{kl}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{kl}\mathbf{Q}_{kl}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
u_{ll}\mathbf{\Omega}_{kk}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{kl}\mathbf{Q}_{lk}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
u_{ij}\mathbf{\Omega}_{ij}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
u_{jj}\mathbf{\Omega}_{ii}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}
\big\}\\
&+4\big\{\frac{1}{2}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}
\right)\tau_{T}\\
&\quad+u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{lk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\circ\mathbf{Q}_{lk}
\right)\tau_{T}\\
&\quad+
\frac{1}{2}u_{ij}\tr\left(\mathbf{Q}_{ij}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{lk}\circ\mathbf{\Omega}_{lk}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{lk}\circ\mathbf{Q}_{lk}
\right)\tau_{T}\\
&\quad+
\frac{1}{2}u_{kl}\tr\left(\mathbf{Q}_{kl}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{kl}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}
\right)\tau_{T}\\
&\quad+u_{kl}\tr\left(\mathbf{Q}_{kl}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{kl}\circ\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}
\right)\tau_{T}\\
&\quad+
\frac{1}{2}u_{kl}\tr\left(\mathbf{Q}_{kl}\right)
\tau_{N}^{\prime}
\left(\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{Q}_{kl}\circ\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}
\right)\tau_{T}\big\}\\
&+8\big\{\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{kl}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{lk}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{lk}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{lk}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{lk}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{4}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{kl}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{kl}\right)
\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{kl}\right)
\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}\left(
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{lk}\right)
\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(
\mathbf{Q}_{ji}\circ\mathbf{Q}_{lk}\right)
\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}
\big\}\\
&+16\big[
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}\right)u_{il}\mathbf{\Omega}_{kj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}\right)u_{jl}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}\right)u_{ik}\mathbf{\Omega}_{lj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}\right)u_{jk}\mathbf{\Omega}_{li}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{lk}\right)u_{il}\mathbf{\Omega}_{kj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{lk}\right)u_{jl}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{lk}\right)u_{ik}\mathbf{\Omega}_{lj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{lk}\right)u_{jk}\mathbf{\Omega}_{li}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{kl}\right)u_{il}\mathbf{\Omega}_{kj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{kl}\right)u_{jl}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{kl}\right)u_{ik}\mathbf{\Omega}_{lj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ij}\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{kl}\right)u_{jk}\mathbf{\Omega}_{li}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{lk}\right)u_{il}\mathbf{\Omega}_{kj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{lk}\right)u_{jl}\mathbf{\Omega}_{ki}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{kl}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{lk}\right)u_{ik}\mathbf{\Omega}_{lj}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{4}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{lk}\right)u_{jk}\mathbf{\Omega}_{li}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{lk}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{kl}\mathbf{\Omega}_{kl}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{kl}\mathbf{Q}_{kl}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{ll}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{kl}\mathbf{Q}_{lk}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{kk}\mathbf{\Omega}_{ll}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{lk}\mathbf{Q}_{kl}
\right\}\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\right)u_{kl}\mathbf{\Omega}_{lk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\right)\mathbf{Q}_{lk}\mathbf{Q}_{lk}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{kl}\mathbf{\Omega}_{kl}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{kl}\mathbf{Q}_{kl}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{ll}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{kl}\mathbf{Q}_{lk}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{kk}\mathbf{\Omega}_{ll}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{lk}\mathbf{Q}_{kl}
\right\}\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\right)u_{kl}\mathbf{\Omega}_{lk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{lk}\mathbf{Q}_{lk}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)u_{kl}\mathbf{\Omega}_{kl}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{kl}\mathbf{Q}_{kl}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)u_{ll}\mathbf{\Omega}_{kk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{kl}\mathbf{Q}_{lk}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)u_{kk}\mathbf{\Omega}_{ll}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{lk}\mathbf{Q}_{kl}
\right\}\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{ji}\right)u_{kl}\mathbf{\Omega}_{lk}
\right\}
\tr\left\{\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{ji}\right)\mathbf{Q}_{lk}\mathbf{Q}_{lk}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{kl}\right)u_{ij}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{kl}\right)u_{jj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{kl}\right)u_{ii}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{kl}\right)u_{ij}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{kl}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{lk}\right)u_{ij}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{lk}\right)u_{jj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{lk}\right)u_{ii}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\frac{1}{8}\tr\left\{\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{lk}\right)u_{ij}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{lk}\circ
\mathbf{\Omega}_{lk}\right)u_{ij}\mathbf{\Omega}_{ij}
\right\}
\tr\left\{\left(\mathbf{Q}_{lk}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ij}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{lk}\circ
\mathbf{\Omega}_{lk}\right)u_{jj}\mathbf{\Omega}_{ii}
\right\}
\tr\left\{\left(\mathbf{Q}_{lk}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{ij}\mathbf{Q}_{ji}
\right\}
\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{lk}\circ
\mathbf{\Omega}_{lk}\right)u_{ii}\mathbf{\Omega}_{jj}
\right\}
\tr\left\{\left(\mathbf{Q}_{lk}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ij}
\right\}\\
&\quad+
\frac{1}{16}\tr\left\{\left(\mathbf{\Omega}_{lk}\circ
\mathbf{\Omega}_{lk}\right)u_{ij}\mathbf{\Omega}_{ji}
\right\}
\tr\left\{\left(\mathbf{Q}_{lk}\circ
\mathbf{Q}_{lk}\right)\mathbf{Q}_{ji}\mathbf{Q}_{ji}
\right\}
\big],\\
G_{7_{6}}=&\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
2\tr^{2}\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\right)
\tr^{2}\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\right)\\
&+4\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{lk}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{lk}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{kl}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{kl}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{lk}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{ji}\circ\mathbf{Q}_{lk}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+
\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{lk}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{Q}_{kl}\circ\mathbf{Q}_{lk}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+\frac{1}{2}
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)
\left(\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)
\left(\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\right)\tau_{T}
\big\}\\
&+\big\{
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}
\circ\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}
\circ\mathbf{Q}_{kl}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}
\circ\mathbf{\Omega}_{lk}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}
\circ\mathbf{Q}_{lk}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{lk}
\circ\mathbf{\Omega}_{lk}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{lk}
\circ\mathbf{Q}_{lk}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{kl}
\circ\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{kl}
\circ\mathbf{Q}_{kl}
\right)\tau_{T}\\
&\quad+
2
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{kl}
\circ\mathbf{\Omega}_{lk}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{kl}
\circ\mathbf{Q}_{lk}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{lk}
\circ\mathbf{\Omega}_{lk}
\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{lk}
\circ\mathbf{Q}_{lk}
\right)\tau_{T}
\big\},\\
G_{7_{7}}=&\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
2\big\{
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}
\right)\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}
\right)\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\circ
\mathbf{Q}_{kl}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}
\right)\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}
\right)\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{kl}\circ
\mathbf{Q}_{kl}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}
\right)\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}
\right)\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}
\right)\tau_{T}\\
&\quad+
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}
\right)\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{ij}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}
\right)\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{ij}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}
\right)\mathbf{\Omega}_{kl}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}
\right)\mathbf{Q}_{kl}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}
\right)\mathbf{\Omega}_{lk}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{ij}
\right)\mathbf{Q}_{lk}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}
\right)\mathbf{\Omega}_{ij}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}
\right)\mathbf{Q}_{ij}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}
\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}
\left(\mathbf{I}_{N}\circ\mathbf{\Omega}_{kl}
\right)\mathbf{\Omega}_{ji}\left(
\mathbf{I}_{N}\circ\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}
\right)\tau_{N}
\tau_{T}^{\prime}
\left(\mathbf{I}_{T}\circ\mathbf{Q}_{kl}
\right)\mathbf{Q}_{ji}\left(
\mathbf{I}_{T}\circ\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}
\right)\tau_{T}\big\}\\
&+\big\{2\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{lk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{kl}\circ\mathbf{Q}_{lk}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ij}\circ
\mathbf{\Omega}_{lk}\circ\mathbf{\Omega}_{lk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ij}\circ
\mathbf{Q}_{lk}\circ\mathbf{Q}_{lk}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{kl}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{kl}\circ\mathbf{Q}_{kl}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{kl}\circ\mathbf{\Omega}_{lk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{kl}\circ\mathbf{Q}_{lk}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{ij}\right)\left(\mathbf{\Omega}_{ji}\circ
\mathbf{\Omega}_{lk}\circ\mathbf{\Omega}_{lk}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{ij}\right)\left(\mathbf{Q}_{ji}\circ
\mathbf{Q}_{lk}\circ\mathbf{Q}_{lk}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{kl}\right)\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{kl}\right)\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{kl}\right)\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{kl}\right)\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{kl}\right)\left(\mathbf{\Omega}_{kl}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{kl}\right)\left(\mathbf{Q}_{kl}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+2\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{kl}\right)\left(\mathbf{\Omega}_{lk}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ij}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{kl}\right)\left(\mathbf{Q}_{lk}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ij}\right)\tau_{T}\\
&\quad+4
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{kl}\right)\left(\mathbf{\Omega}_{lk}\circ
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{kl}\right)\left(\mathbf{Q}_{lk}\circ
\mathbf{Q}_{ij}\circ\mathbf{Q}_{ji}\right)\tau_{T}\\
&\quad+2
\tau_{N}^{\prime}\left(\mathbf{I}_{N}
\circ\mathbf{\Omega}_{kl}\right)\left(\mathbf{\Omega}_{lk}\circ
\mathbf{\Omega}_{ji}\circ\mathbf{\Omega}_{ji}\right)\tau_{N}
\tau_{T}^{\prime}\left(\mathbf{I}_{T}
\circ\mathbf{Q}_{kl}\right)\left(\mathbf{Q}_{lk}\circ
\mathbf{Q}_{ji}\circ\mathbf{Q}_{ji}\right)\tau_{T}
\big\}.
\end{align*}
Due to $$\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}u_{kj}^{2}u_{il}^{2}=O(N^{2}),$$
$$\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}|u_{ij}u_{jl}u_{kk}u_{li}|\leq
\sum_{i=1}^{N}\sum_{k=1}^{N}\sum_{j=1}^{N}|u_{ij}|\sum_{l=1}^{N}|u_{li}|
=O(N^{2\vee(4-2/\tau)})$$
and
$$\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}|u_{ij}u_{jk}u_{kl}u_{li}|\leq\underset{i\neq j}{\sum^{N}\sum^{N}}|u_{ij}|\underset{k\neq l}{\sum^{N}\sum^{N}}|u_{kl}|=O(N^{2\vee(4-2/\tau)}),$$
we have $G_{7_{1}}=O(N^{2\vee(4-2/\tau)}T^{3}).$
Similarly,
due to $$\underset{i\neq j\neq k\neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}|u_{ij}|=O(N^{3\vee(4-1/\tau)}),$$
and Lemma \ref{lem:sigma},
we have $G_{7_{2}}=O(N^{2\vee(4-2/\tau)}T^{3})+O(N^{3\vee(4-1/\tau)}T^{2}),$
$G_{7_{3}}=O(N^{3\vee(4-1/\tau)}T^{2})+O(N^{4}T),$
$G_{7_{4}}=O(N^{4}T),$
$G_{7_{5}}=O(N^{3\vee(4-1/\tau)}T^{5/2})+O(N^{4}T^{3/2}),$
and $G_{7_{7}}=O(N^{4}T^{3/2}).$
Here, we focus on calculating the first term of $G_{7_{6}}.$
Because
\begin{align*}
&\underset{i\neq j\neq k \neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}2\tr^{2}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\right)
\tr^{2}\left(
\mathbf{Q}_{ij}\circ\mathbf{Q}_{kl}\right)\\
=&O(T^{2})\underset{i\neq j\neq k \neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}\tr^{2}\left(
\mathbf{\Omega}_{ij}\circ\mathbf{\Omega}_{kl}\right)\\
=&O(T^{2})\underset{i\neq j\neq k \neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
\Big(\sum_{s=1}^{N}(\mathbf{\Omega}_{ij})_{ss}(\mathbf{\Omega}_{kl})_{ss}
\Big)^{2}\\
\leq&O(T^{2})\underset{i\neq j\neq k \neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
\sum_{s=1}^{N}(\mathbf{\Omega}_{ij})_{ss}^{2}(\mathbf{\Omega}_{kl})_{ss}^{2}\\
&+O(T^{2})\underset{i\neq j\neq k \neq l}{\sum^{N}\sum^{N}\sum^{N}\sum^{N}}
\underset{s_{1}\neq s_{2}}{\sum^{N}\sum^{N}}|(\mathbf{\Omega}_{ij})_{s_{1}s_{1}}(\mathbf{\Omega}_{kl})_{s_{1}s_{1}}
(\mathbf{\Omega}_{ij})_{s_{2}s_{2}}(\mathbf{\Omega}_{kl})_{s_{2}s_{2}}|\\
\leq&O(NT^{2}+N^{2}T^{2}),
\end{align*}
where the last inequality holds due to $\mathbf{\Omega}_{ij}=u_{i}^{1/2}(u_{j}^{1/2})^{\prime}$
and $(u_{i}^{1/2})^{\prime}u_{i}^{1/2}=1,$ for all $1\leq i,j\leq N.$
Hence, we have $G_{7_{6}}=O(N^{2}T^{2})+O(N^{4}T^{3/2}).$
and $G_{7}=O(N^{2\vee(4-2/\tau)}T^{3})+O(N^{3\vee(4-1/\tau)}T^{2})+
O(N^{3\vee(4-1/\tau)}T^{5/2})+O(N^{4}T^{3/2})$.
Combining $G_{1},\cdots,G_{7},$
we can conclude that
$$\var\Big[\big\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\big\}^{2}\Big]
=O(N^{4}T^{3/2}+N^{2}T^{3}+N^{3}T^{5/2}+N^{2\vee(4-2/\tau)}T^{3}+N^{3\vee(4-1/\tau)}T^{5/2}).$$
Similarly, for $$\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}
\hat{\epsilon}_{i\cdot}=\sum_{i=1}^{N}Z^{\prime}\left(\mathbf{\Omega}_{ii}\otimes
\mathbf{Q}_{ii}\right)Z=Z^{\prime}\left\{\sum_{i=1}^{N}\left(\mathbf{\Omega}_{ii}\otimes
\mathbf{Q}_{ii}\right)\right\}Z\doteq Z^{\prime}\mathbf{J}Z.$$
So, we have
$\var\bigg\{\Big(\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}
\hat{\epsilon}_{i\cdot}\Big)^{2}\bigg\}=\var\bigg\{\Big( Z^{\prime}\mathbf{J}Z\Big)^{2}\bigg\}.$
Again using Lemma \ref{le:moment of quadratic form},
we have
\begin{align*}
&\var\bigg\{\Big( Z^{\prime}\mathbf{J}Z\Big)^{2}\bigg\}\\
=&\mE\Big( Z^{\prime}\mathbf{J}Z\Big)^{4}-\mE^{2}\Big( Z^{\prime}\mathbf{J}Z\Big)^{2}\\
=&K_{1}+\gamma_{2}K_{2}+\gamma_{4}K_{3}+\gamma_{6}K_{4}+\gamma_{1}^{2}K_{5}
+\gamma_{2}^{2}K_{6}+\gamma_{1}\gamma_{3}K_{7},
\end{align*}
where
\begin{align*}
K_{1}=&8\tr^{2}\left(\mathbf{J}\right)\tr\left(\mathbf{J}^{2}\right)
+8\tr^{2}\left(\mathbf{J}^{2}\right)+32\tr\left(\mathbf{J}\right)\tr\left(\mathbf{J}^{3}\right)
+48\tr\left(\mathbf{J}^{4}\right),\\
K_{2}=&4\tr\left(\mathbf{J}\right) \tr\left(\mathbf{J}\right) \tr\left(\mathbf{J} \circ \mathbf{J}\right)+8{\tau}_{NT}^{\prime}\left(\mathbf{J} \circ \mathbf{J}\right){\tau}_{NT}\tr\left(\mathbf{J} \circ \mathbf{J}\right) +48\tr\left(\mathbf{J}\right) \tr\left(\mathbf{J} \circ\mathbf{J}^{2}\right)\\
&+96\tr\left\{\left(\mathbf{I}_{NT} \circ \mathbf{J}\right) \mathbf{J}^{3}\right\}+48{\tau}_{NT}^{\prime}\left(\mathbf{I}_{NT} \circ\mathbf{J}^{2}\right)\left(\mathbf{I}_{NT} \circ\mathbf{J}^{2}\right) {\tau}_{NT},\\
K_{3}=& 4\tr\left(\mathbf{J}\right) \tr\left(\mathbf{J} \circ \mathbf{J} \circ \mathbf{J}\right) +24\tr\left(\mathbf{J} \circ \mathbf{J} \circ\mathbf{J}^{2}\right),\\
K_{4}=& \tr\left(\mathbf{J} \circ \mathbf{J} \circ \mathbf{J} \circ \mathbf{J}\right),\\
K_{5}=& 24{\tau}_{NT}^{\prime}\left(\mathbf{I}_{NT} \circ \mathbf{J}\right) \mathbf{J}\left(\mathbf{I}_{NT} \circ \mathbf{J}\right){\tau}_{NT} \tr\left(\mathbf{J}\right)+48{\tau}_{NT}^{\prime}\left(\mathbf{I}_{NT} \circ \mathbf{J}\right) \mathbf{J} ^{2}\left(\mathbf{I}_{NT} \circ \mathbf{J}\right)\tau_{NT}\\
&+16{\tau}_{NT}^{\prime}\left(\mathbf{J} \circ \mathbf{J} \circ \mathbf{J}\right){\tau}_{NT} \tr\left(\mathbf{J}\right)+96{\tau}_{NT}^{\prime}\left(\mathbf{J} \circ \mathbf{J}\right) \mathbf{J}\left(\mathbf{I}_{NT} \circ \mathbf{J}\right) \tau_{NT}\\
&+96\tr\left(\mathbf{J}\left(\mathbf{J} \circ \mathbf{J}\right) \mathbf{J}\right),\\
K_{6}=& 3\tr\left(\mathbf{J} \circ \mathbf{J}\right) \tr\left(\mathbf{J} \circ \mathbf{J}\right)+24{\tau}_{NT}^{\prime}\left(\mathbf{I}_{NT} \circ \mathbf{J}\right)\left(\mathbf{J} \circ \mathbf{J}\right)\left(\mathbf{I}_{NT} \circ \mathbf{J}\right){\tau}_{NT}\\
&+8{\tau}_{NT}^{\prime}\left(\mathbf{J} \circ \mathbf{J} \circ \mathbf{J} \circ \mathbf{J}\right) {\tau}_{NT},\\
K_{7}=&24{\tau}_{NT}^{\prime}\left(\mathbf{I}_{NT} \circ \mathbf{J}\right) \mathbf{J}\left(\mathbf{I}_{NT} \circ \mathbf{J} \circ \mathbf{J}\right){\tau}_{NT}+32{\tau}_{NT}^{\prime}\left(\mathbf{I}_{NT} \circ \mathbf{J}\right)\left(\mathbf{J} \circ \mathbf{J} \circ \mathbf{J}\right){\tau}_{NT}.
\end{align*}
First, we have
\begin{align*}
\tr\left(\mathbf{J}\right)=&\sum_{i=1}^{N}u_{ii}\tr(\mathbf{Q}_{ii})=O(NT),\\
\tr\left(\mathbf{J}^{2}\right)=&\sum_{i=1}^{N}\sum_{j=1}^{N}u_{ij}^{2}
\tr(\mathbf{Q}_{ii}\mathbf{Q}_{jj})=O(NT).
\end{align*}
Then, due to Assumption \ref{assum:matrix} and
$$\sum_{i=1}^{N}\sum_{j=1}^{N}\sum_{k=1}^{N}|u_{ij}u_{jk}u_{ik}|
=O(N^{1\vee(3-2/\tau)}),$$
we have
\begin{align*}
\tr\left(\mathbf{J}^{3}\right)=&\sum_{i=1}^{N}\sum_{j=1}^{N}\sum_{k=1}^{N}
u_{ij}u_{jk}u_{ik}
\tr(\mathbf{Q}_{ii}\mathbf{Q}_{jj}\mathbf{Q}_{kk})\\
=&\tr\left(\mathbf{U}^{3}\right)
\tr\left(\mathbf{\Sigma}^{3}\right)+\sum_{i=1}^{N}\sum_{j=1}^{N}\sum_{k=1}^{N}u_{ij}u_{jk}u_{ik} \left\{\tr(\mathbf{Q}_{ii}\mathbf{Q}_{jj}\mathbf{Q}_{kk})-\tr\left(\mathbf{\Sigma}^{3}\right)\right\}\\
\leq&\tr\left(\mathbf{U}^{3}\right)
\tr\left(\mathbf{\Sigma}^{3}\right)+O(N^{1\vee(3-2/\tau)})=O(NT).
\end{align*}
Next, due to Assumption \ref{assum:matrix} and
\begin{align*}
&\sum_{i=1}^{N}\sum_{j=1}^{N}\sum_{k=1}^{N}\sum_{l=1}^{N}|u_{ij}u_{jk}u_{kl}u_{li}|\\
\leq&\sum_{i=1}^{N}\sum_{k=1}^{N}\left(\sum_{j=1}^{N}|u_{ij}u_{jk}|
\sum_{l=1}^{N}|u_{kl}u_{li}|\right)\\
\leq&\sum_{i=1}^{N}\sum_{k=1}^{N}\left(\sum_{j=1}^{N}u_{ij}^{2}\sum_{j=1}^{N}u_{jk}^{2}\right)\\
\leq&\sum_{i=1}^{N}\sum_{k=1}^{N}C
=O(N^{2}),
\end{align*}
we can get
\begin{align*}
\tr\left(\mathbf{J}^{4}\right)=&\sum_{i=1}^{N}\sum_{j=1}^{N}\sum_{k=1}^{N}
\sum_{l=1}^{N}
u_{ij}u_{jk}u_{kl}u_{li}
\tr(\mathbf{Q}_{ii}\mathbf{Q}_{jj}\mathbf{Q}_{kk}\mathbf{Q}_{ll})\\
=&\tr\left(\mathbf{U}^{4}\right)
\tr\left(\mathbf{\Sigma}^{4}\right)+\sum_{i=1}^{N}\sum_{j=1}^{N}
\sum_{k=1}^{N}\sum_{l=1}^{N}u_{ij}u_{jk}u_{kl}u_{li} \left\{\tr(\mathbf{Q}_{ii}\mathbf{Q}_{jj}\mathbf{Q}_{kk}\mathbf{Q}_{ll})
-\tr\left(\mathbf{\Sigma}^{4}\right)\right\}\\
\leq&\tr\left(\mathbf{U}^{4}\right)
\tr\left(\mathbf{\Sigma}^{4}\right)+O(N^{2})=O(NT).
\end{align*}
Similarly, according to the above formulas, we have
$K_{1}=O(N^{3}T^{3})=K_{2},$
$K_{3}=O(N^{2}T^{2}),$ $K_{4}=O(NT),$
$K_{5}=O(N^{3}T^{3}),$
$K_{6}=O(N^{2}T^{2})=K_{7}.$
Then, we can conclude that
$$\var\bigg\{\Big(\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}
\hat{\epsilon}_{i\cdot}\Big)^{2}\bigg\}=O(N^{3}T^{3}).$$
Finally, by Assumption \ref{assum:matrix}, we can obtain that
\begin{align*}
\mE(b^2_{N})\leq&\frac{1}{N^2T^4}\Bigg[2\var\bigg\{\sum_{i=1}^{N} \sum_{j=1}^{N} (\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}\bigg\}+\frac{2}{T^2}
\var\bigg\{\Big(\sum_{i=1}^{N}\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\Big)^{2}\bigg\}\Bigg]\\
=&O\left(\frac{1}{T^{\frac{1}{2}\vee \left(\frac{1}{\tau}-\frac{1}{2}\right)}}\right).
\end{align*}
\hfill$\Box$
\subsection{Proof of Lemma \ref{th:sum null}}
Recall that $\mathbf{M}_{i}=\mathbf{P}_{i}\mathbf{\Sigma}\mathbf{P}_{i}.$
Let
$\hat{\epsilon}_{i\cdot}=\mathbf{P}_{i}\epsilon_{i\cdot}=\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}z_{i},$
where $z_{i}=(z_{i1},\cdots,z_{iT})^{\prime}$ is the $i$-th row vector of $\mathbf{Z}.$ Define $\mathbf{Q}_{ij}\doteq\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2}
,$ then, we have $\mathbf{Q}_{ii}=\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}.$
Obviously, by Assumption \ref{assum:E distribution}, we have $\mE\left(S_{N}\right)=0.$ Then, we have
\begin{align*}
\mE\left(S^2_{N}\right)&=\frac{2}{N(N-1)}\underset{i<j}{\sum\sum}\underset{s<t}{\sum\sum}
\mE\left(\frac{\epsilon_{i\cdot}^{\prime} \mathbf{P}_{i} \mathbf{P}_{j} \epsilon_{j\cdot}}{\left\|\mathbf{P}_{i} \epsilon_{i\cdot}\right\| \cdot\left\|\mathbf{P}_{j} \epsilon_{j\cdot}\right\|}\frac{\epsilon_{s}^{\prime} \mathbf{P}_{s} \mathbf{P}_{t} \epsilon_{t}}{\left\|\mathbf{P}_{s} \epsilon_{s}\right\| \cdot\left\|\mathbf{P}_{t} \epsilon_{t}\right\|}\right)\\
&=\frac{2}{N(N-1)}\underset{i<j}{\sum\sum}
\mE\left\{\frac{\left(\epsilon_{i\cdot}^{\prime} \mathbf{P}_{i} \mathbf{P}_{j} \epsilon_{j\cdot}\right)^2}{\left\|\mathbf{P}_{i} \epsilon_{i\cdot}\right\|^2 \cdot\left\|\mathbf{P}_{j} \epsilon_{j\cdot}\right\|^2}\right\}\\
&=\frac{2}{N(N-1)}\underset{i<j}{\sum\sum}\mE\left[\frac{\epsilon_{j\cdot}^{\prime} \mathbf{P}_{j} \mathbf{M}_{i}\mathbf{P}_{j} \epsilon_{j\cdot}}{\tr(\mathbf{M}_{i})\left\|\mathbf{P}_{j} \epsilon_{j\cdot}\right\|^2}+\frac{(\epsilon_{j\cdot}^{\prime} \mathbf{P}_{j} \mathbf{M}_{i}\mathbf{P}_{j} \epsilon_{j\cdot})\cdot\{2\tr(\mathbf{M}^2_{i})+\gamma_{
2}\tr(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii})\}}{\tr^3(\mathbf{M}_{i})\left\|\mathbf{P}_{j} \epsilon_{j\cdot}\right\|^2}\right.\\
&\left.\,\,\,\,-\frac{[2(\epsilon_{j\cdot}^{\prime} \mathbf{P}_{j} \mathbf{M}^2_{i}\mathbf{P}_{j} \epsilon_{j\cdot})+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{P}_{j}
\epsilon_{j\cdot}\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})
\circ \mathbf{Q}_{ii}\}]}
{\tr^2(\mathbf{M}_{i})\left\|\mathbf{P}_{j} \epsilon_{j\cdot}\right\|^2}+O(T^{-2})\right]\\
&=\frac{2}{N(N-1)}\underset{i<j}{\sum\sum}\frac{\tr\left(\mathbf{M}_{i}
\mathbf{M}_{j}\right)}
{\tr\left(\mathbf{M}_{i}\right)\tr\left(\mathbf{M}_{j}\right)}
+O(T^{-2})\\
&=\sigma^2_{S_{N}}\left\{1+o(1)\right\},
\end{align*}
where the third equality holds due to Lemmas \ref{lem:laplace approxi} and \ref{le:moment of quadratic form}, and the last equality above holds because of Lemma \ref{lem:sigma}, and the last second equality above holds because of Lemmas \ref{lem:laplace approxi}, \ref{lem:sigma} and \ref{le:moment of quadratic form},
$\tr(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii})\leq \tr(\mathbf{Q}^2_{ii})=\tr(\mathbf{M}^2_{i})=O(T)$
and $$0\leq \tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{P}_{j}
\epsilon_{j\cdot}\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})
\circ \mathbf{Q}_{ii}\}/\left\|\mathbf{P}_{j} \epsilon_{j\cdot}\right\|\leq C \tr(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{P}_{j}
\epsilon_{j\cdot}\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})/\left\|\mathbf{P}_{j} \epsilon_{j\cdot}\right\|=O(1).$$
The normality of $S_{N}$ has yet to be proven. Define
$\sum_{j=2}^k Z_{j}$ where
$Z_{j}=\sum_{i=1}^{j-1}\sqrt{\frac{2}{N(N-1)}}\hat{\rho}_{i j},$
Let
$\mathcal{F}_{j}\doteq\sigma\{\epsilon_{1\cdot},\cdots,\epsilon_{j\cdot}\}$ be the $\sigma$-field
generated by $\{\epsilon_{i\cdot}, i\le j\}$, where $\epsilon_{i\cdot}=(\epsilon_{i1},\cdots,\epsilon_{iT})^{\prime}$.
Because $\epsilon_{1\cdot},\cdots,\epsilon_{N\cdot}$ are mutually independent under $H_{0}.$
Obviously, $\mE(Z_{j}\mid
\mathcal{F}_{j-1})=0$ and it follows that $\{\sum_{j=2}^k Z_{j},
\mathcal{F}_{k}; 2\le k\le N\}$ is a zero mean martingale.
According to %the Corollary 3.1 and Theorem 3.5 in
the Martingale
central limit theorem in \cite{hall2014}, we only need to show
\begin{align}\label{clt1}
\frac{\sum_{j=2}^{N}\mE\left(Z_{j}^2\mid\mathcal{F}_{j-1}\right)}{\sigma^2_{S_{N}}}\cp
1,
\end{align}
and
\begin{align}\label{clt2}
\mE\left\{\sum_{j=2}^{N}\mE\left(Z_{j}^4|\mathcal{F}_{j-1}\right)\right\}=o\{\sigma^4_{S_{N}}\}.
\end{align}
We first focus on (\ref{clt1}).
It can be shown that
\begin{align*}
\sum_{j=2}^{N}\mE(Z_{j}^2|\mathcal{F}_{j-1})
=&\sum_{j=2}^{N}\mE\left[\left.\left\{\sqrt{\frac{2}{N(N-1)}}\sum_{i=1}^{j-1}\hat{\rho}_{i j}\right\}^2\right|\mathcal{F}_{j-1}\right]\\
=&\sum_{j=2}^{N}\frac{2}{N(N-1)}\mE\left\{\left.\sum_{i_1=1}^{j-1}\sum_{i_2=1}^{j-1}\hat{\rho}_{i_{1} j}\hat{\rho}_{i_{2} j}\right|\mathcal{F}_{n,j-1}\right\}\\
=&\sum_{j=2}^{N}\frac{2}{N(N-1)}\sum_{i_1=1}^{j-1}\sum_{i_2=1}^{j-1}\mE\left(\left.\hat{\rho}_{i_{1} j}\hat{\rho}_{i_{2} j}\right|\mathcal{F}_{n,j-1}\right)\\
=& A+B,
\end{align*}
where
\bse
A=\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left\{\left.\left(\hat{\rho}_{ij}\right)^{2}\right|\mathcal{F}_{n,j-1}\right\}\,\,\mbox{and}\,\,
B=\frac{4}{N(N-1)}\sum_{j=2}^{N}\underset{i_1<i_2}{\sum^{j-1}\sum^{j-1}}\mE\left(\left.\hat{\rho}_{i_{1} j}\hat{\rho}_{i_{2} j}\right|\mathcal{F}_{n,j-1}\right).
\ese
First, we consider $A$. By Lemmas \ref{lem:laplace approxi}, \ref{lem:matrix inequality} and \ref{le:moment of quadratic form}, we have
\begin{align*}
A=&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}
\Bigg\{\frac{\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)}{\tr\left(\mathbf{M}_{j}\right)
\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}+\frac{\mE\left({\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{j}\hat{\epsilon}_{i\cdot}
\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{j}\epsilon_{j\cdot}^{\prime}\right)\var\left(\epsilon_{j\cdot}^{\prime}
\mathbf{P}_{j}\epsilon_{j\cdot}\right)}{\tr^3\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\\
&-\frac{\mE\left(\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\hat{\epsilon}_{i\cdot}\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{j}\epsilon_{j\cdot}
\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\epsilon_{j\cdot}\right)-
\mE\left(\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\hat{\epsilon}_{i\cdot}\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{j}\epsilon_{j\cdot}\right)
\mE\left(\epsilon_{j\cdot}^{\prime}\mathbf{P}_{j}\epsilon_{j\cdot}\right)}{\tr^2\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\Bigg\}+O\left(T^{-2}\right)\\
=&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}
\Bigg[\frac{\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)}{\tr\left(\mathbf{M}_{j}\right)
\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}+\frac{\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)
\{2\tr\left(\mathbf{M}_{j}^2
\right)+\gamma_{2}\tr(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj})\}}{\tr^3\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\\
&-\frac{2\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}^2_{j}\hat{\epsilon}_{i\cdot}\right)
}
{\tr^2\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}
-\frac{\gamma_{2}\tr[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{P}_{i}
\epsilon_{i\cdot}\epsilon_{i\cdot}^{\prime}\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ\mathbf{Q}_{jj}]}
{\tr^2\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\Bigg]+O\left(T^{-2}\right)\\
=&A_{1}+A_{2}-A_{3}-A_{4}+O\left(T^{-2}\right).
\end{align*}
By Lemmas \ref{lem:laplace approxi} and \ref{lem:sigma}, we have
\begin{align*}
\mE\left(A_{1}\right)=&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left(\frac{\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)}{\tr\left(\mathbf{M}_{j}\right)
\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right)\\
=&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\frac{\tr\left(\mathbf{M}_{i}\mathbf{M}_{j}\right)}{\tr\left(\mathbf{M}_{i}\right)\tr\left(\mathbf{M}_{j}\right)}+O\left\{\frac{1}{T\tr\left(\mathbf{M}_{j}\right)}\right\}\\
=&\sigma^2_{S_{N}}\left\{1+o(1)\right\}
\end{align*}
and
\begin{align*}
\var\left(A_{1}\right)=&\mE\left(A^2_{1}\right)-\mE^2\left(A_{1}\right)\\
\leq&
\frac{4}{N^2(N-1)^2}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left\{\frac{\tr^2\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)}
{\tr^2\left(\mathbf{M}_{j}\right)
\left(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^2}\right\}\\
%&+\frac{8}{N^2(N-1)^2}\sum_{j=2}^{N}\sum_{i_{1}<i_{2}}^{j-1}\mE\left\{\frac{\tr\left(\hat{\epsilon}_{i_{1}}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i_{1}}\right)
%\tr\left(\hat{\epsilon}_{i_{2}}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i_{2}}\right)}{\tr\left(\mathbf{M}_{j}\right)
%\hat{\epsilon}_{i_{1}}^{\prime}\hat{\epsilon}_{i_{1}}\tr\left(\mathbf{M}_{j}\right)
%\hat{\epsilon}_{i_{2}}^{\prime}\hat{\epsilon}_{i_{2}}}\right\}\\
&+\frac{8}{N^2(N-1)^2}\sum_{2\leq j_{1}<j_{2}\leq N}\sum_{1\leq i\leq j_{1}-1}\mE\left\{\frac{\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j_{1}}\hat{\epsilon}_{i\cdot}\right)
\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j_{2}}\hat{\epsilon}_{i\cdot}\right)}{\tr\left(\mathbf{M}_{j_{1}}\right)
\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\tr\left(\mathbf{M}_{j_{2}}\right)
\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right\}\\
%&+\frac{8}{N^2(N-1)^2}\sum_{2\leq j_{1}<j_{2}\leq N}\sum_{1\leq i\leq j_{1}-1}\mE\left\{\frac{\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j_{1}}\hat{\epsilon}_{i\cdot}\right)
%\tr\left(\hat{\epsilon}_{j_{1}}^{\prime}\mathbf{M}_{j_{2}}\hat{\epsilon}_{j_{1}}\right)}{\tr\left(\mathbf{M}_{j_{1}}\right)
%\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\tr\left(\mathbf{M}_{j_{2}}\right)
%\hat{\epsilon}_{j_{1}}^{\prime}\hat{\epsilon}_{j_{1}}}\right\}
\leq&\frac{4}{N^2(N-1)^2}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left\{\frac{\tr^2\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)}
{\tr^2\left(\mathbf{M}_{j}\right)
\left(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^2}\right\}\\
&+\frac{8}{N^2(N-1)^2}\sum_{2\leq j_{1}<j_{2}\leq N}\sum_{1\leq i\leq j_{1}-1}\frac{
\sqrt{\mE\left\{\frac{\tr^2\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j_{1}}\hat{\epsilon}_{i\cdot}\right)}
{\left(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^2}\right\}
\mE\left\{\frac{\tr^2\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j_{2}}\hat{\epsilon}_{i\cdot}\right)
}{\left(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^2}\right\}}}
{\tr\left(\mathbf{M}_{j_{1}}\right)\tr\left(\mathbf{M}_{j_{2}}\right)}\\
=&O\left(N^{-2}T^{-2}\right)+O\left(N^{-1}T^{-2}\right)\\
=&o\left(\sigma^4_{S_{N}}\right),
\end{align*}
where the last second equality holds because \begin{align*}
\mE\left\{\frac{\tr^2\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)}
{\tr^2\left(\mathbf{M}_{j}\right)
\left(\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}\right)^2}\right\}=&
\frac{\tr^2\left(\mathbf{M}_{i}\mathbf{M}_{j}
\right)+2\tr\left(\mathbf{M}_{i}\mathbf{M}_{j}\mathbf{M}_{i}\mathbf{M}_{j}\right)}{\tr^2\left(\mathbf{M}_{j}\right)\tr^2\left(\mathbf{M}_{i}\right)}\\
&+\frac{\gamma_{2}\tr[(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})]}{\tr^2\left(\mathbf{M}_{j}\right)\tr^2\left(\mathbf{M}_{i}\right)}
+O\left\{T^{-1}\tr^{-2}\left(\mathbf{M}_{j}\right)\right\}\\
\leq&\frac{\tr^2\left(\mathbf{M}_{i}\mathbf{M}_{j}
\right)+(2+|\gamma_{2}|)\tr\left(\mathbf{M}_{i}\mathbf{M}_{j}\mathbf{M}_{i}\mathbf{M}_{j}\right)}
{\tr^2\left(\mathbf{M}_{j}\right)\tr^2\left(\mathbf{M}_{i}\right)}+O\left\{T^{-1}\tr^{-2}\left(\mathbf{M}_{j}\right)\right\}\\
=&O(T^{-2}).
\end{align*}
Hence, we can conclude that $A_{1}=\sigma^2_{S_{N}}+o_{p}\left(\sigma^2_{S_{N}}\right).$ Then, we have
\begin{align*}
\mE\left(|A_{2}|\right)\leq&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left[\frac{
\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)
\{2\tr\left(\mathbf{M}_{j}^2\right)+|\gamma_{2}|\tr(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj})\}}{\tr^3\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right]\\
=&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\frac{2\tr\left(\mathbf{M}_{j}^2\right)+|\gamma_{2}|\tr(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj})}{\tr^3\left(\mathbf{M}_{j}\right)}
\mE\left\{\frac{\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i\cdot}\right)}{\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right\}\\
=&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\frac{2\tr\left(\mathbf{M}_{j}^2\right)+|\gamma_{2}|\tr(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj})}{\tr^3\left(\mathbf{M}_{j}\right)}
\left\{\frac{\tr\left(\mathbf{M}_{i}\mathbf{M}_{j}\right)}{\tr\left(\mathbf{M}_{i}\right)}+O\left(T^{-1}\right)
\right\}\\
=&O\left(T^{-2}\right)=o\left(\sigma^2_{S_{N}}\right),
\end{align*}
where the last two equality holds due to $\tr(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj})\leq \tr(\mathbf{M}_{j}\mathbf{M}_{j})=O(T).$
Hence, we have $A_{2}=o_{p}\left(\sigma^2_{S_{N}}\right)$ by Markov inequality.
Next, we have
\begin{align*}
\mE\left(A_{3}\right)=&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left\{\frac{
2\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}^2_{j}\hat{\epsilon}_{i\cdot}\right)}
{\tr^2\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right\}\\
=&\frac{4}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\frac{1}{\tr^2\left(\mathbf{M}_{j}\right)}
\mE\left\{\frac{\tr\left(\hat{\epsilon}_{i\cdot}^{\prime}\mathbf{M}^2_{j}\hat{\epsilon}_{i\cdot}\right)}{\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right\}\\
=&\frac{4}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\frac{1}{\tr^2\left(\mathbf{M}_{j}\right)}
\left\{\frac{\tr\left(\mathbf{M}_{i}\mathbf{M}^2_{j}\right)}{\tr\left(\mathbf{M}_{i}\right)}+O\left(T^{-1}\right)
\right\}\\
=&O\left(T^{-2}\right)=o\left(\sigma^2_{S_{N}}\right),
\end{align*}
and
\begin{align*}
\mE(A_{4}/\gamma_{2})=&\frac{2}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left\{\frac{\tr[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{P}_{i}
\epsilon_{i\cdot}\epsilon_{i\cdot}^{\prime}\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ\mathbf{Q}_{jj}]}
{\tr^2\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right\}\\
\leq&\frac{2C}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left\{\frac{\tr(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{P}_{i}
\epsilon_{i\cdot}\epsilon_{i\cdot}^{\prime}\mathbf{P}_{i}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})}
{\tr^2\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right\}\\
=&\frac{2C}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left\{\frac{\tr(\epsilon_{i\cdot}^{\prime}\mathbf{P}_{i}\mathbf{M}_{j}\mathbf{P}_{i}
\epsilon_{i\cdot})}
{\tr^2\left(\mathbf{M}_{j}\right)\hat{\epsilon}_{i\cdot}^{\prime}\hat{\epsilon}_{i\cdot}}\right\}\\
=&\frac{2C}{N(N-1)}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\left\{\frac{\tr(\mathbf{M}_{j}\mathbf{M}_{i})}
{\tr^2\left(\mathbf{M}_{j}\right)\tr(\mathbf{M}_{i})}+O\left(\frac{1}{T\tr^2\left(\mathbf{M}_{j}\right)}\right)\right\}\\
=&O\left(T^{-2}\right)=o\left(\sigma^2_{S_{N}}\right).
\end{align*}
Because $A_{3}$ and $A_{4}/\gamma_{2}$ are non-negative, we have $A_{3}=o_{p}\left(\sigma^2_{S_{N}}\right)$ and $A_{4}/\gamma_{2}=o_{p}\left(\sigma^2_{S_{N}}\right).$
Second, we focus on $B.$
\begin{align*}
B=&\frac{4}{N(N-1)}\sum_{j=2}^{N}\underset{i_1<i_2}{\sum^{j-1}\sum^{j-1}}\mE\left(\left.\hat{\rho}_{i_{1} j}\hat{\rho}_{i_{2} j}\right|\mathcal{F}_{n,j-1}\right)\\
=&\frac{4}{N(N-1)}\sum_{j=2}^{N}\left\{\underset{i_1<i_2}{\sum^{j-1}\sum^{j-1}}\mE\left(\left.\hat{\rho}_{i_{1} j}\hat{\rho}_{i_{2} j}\right|\mathcal{F}_{n,j-1}\right)\right\}\\
=&\sum_{j=2}^{N}\left\{\frac{4}{N(N-1)}\underset{i_1<i_2}{\sum^{j-1}\sum^{j-1}}\delta_{i_{1}i_{2}}\right\},
\end{align*}
where we denote that $\delta_{i_{1}i_{2}}\doteq\mE\left(\left.\hat{\rho}_{i_{1} j}\hat{\rho}_{i_{2} j}\right|\mathcal{F}_{n,j-1}\right).$
Obviously, we have $\mE\left(B\right)=0 $ due to the law of total expectation.
By Lemmas \ref{lem:laplace approxi} and \ref{lem:matrix inequality}, we have
\begin{align*}
\delta_{i_{1}i_{2}}
=&\frac{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\mathbf{M}_{j}
\hat{\epsilon}_{i_{2}\cdot}}{\sqrt{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\hat{\epsilon}_{i_{2}\cdot}}\tr\left(\mathbf{M}_{j}\right)}
+\frac{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\mathbf{M}_{j}
\hat{\epsilon}_{i_{2}\cdot}
\{2\tr(\mathbf{M}^2_{j})+\gamma_{2}\tr(\mathbf{Q}_{jj}\circ\mathbf{Q}_{jj})\}}
{\sqrt{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\hat{\epsilon}_{i_{2}\cdot}}\tr^3\left(\mathbf{M}_{j}\right)}\\
&-\frac{2\hat{\epsilon}_{i_{1}\cdot}^{\prime}\mathbf{M}^2_{j}
\hat{\epsilon}_{i_{2}\cdot}+\gamma_{2}\tr[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ\mathbf{Q}_{jj}]}{\sqrt{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\hat{\epsilon}_{i_{2}\cdot}}\tr^2\left(\mathbf{M}_{j}\right)}+O\left(T^{-2}\right)
\end{align*}
for any $1\leq i_{1}<i_{2}<j\leq N.$
Moreover, we have
\begin{align*}
\mE\left(\delta^2_{i_{1}i_{2}}\right)\leq&5\mE\left\{\frac{\left(\hat{\epsilon}_{i_{1}\cdot}^{\prime}\mathbf{M}_{j}
\hat{\epsilon}_{i_{2}\cdot}\right)^2}{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\hat{\epsilon}_{i_{2}\cdot}\tr^2\left(\mathbf{M}_{j}\right)}\right\}
+5\mE\left\{\frac{(2+|\gamma_{2}|)^2\left(
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\mathbf{M}_{j}\hat{\epsilon}_{i_{1}\cdot}\right)^2
\tr^2\left(\mathbf{M}^2_{j}\right)}
{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\hat{\epsilon}_{i_{2}\cdot}\tr^6\left(\mathbf{M}_{j}\right)}\right\}\\
&+5\mE\left\{\frac{4\left(\hat{\epsilon}_{i_{2}\cdot}^{\prime}\mathbf{M}^2_{j}\hat{\epsilon}_{i_{1}\cdot}\right)^2}
{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\hat{\epsilon}_{i_{2}\cdot}\tr^4\left(\mathbf{M}_{j}\right)}\right\}+
5\mE\left\{\frac{\gamma_{2}^2\tr^2[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ\mathbf{Q}_{jj}]}
{\hat{\epsilon}_{i_{1}\cdot}^{\prime}\hat{\epsilon}_{i_{1}\cdot}
\hat{\epsilon}_{i_{2}\cdot}^{\prime}\hat{\epsilon}_{i_{2}\cdot}\tr^4\left(\mathbf{M}_{j}\right)}\right\}
+O\left(T^{-4}\right)\\
\leq&\frac{5\tr\left(\mathbf{M}_{i_{1}}\mathbf{M}_{j}\mathbf{M}_{i_{2}}\mathbf{M}_{j}\right)}
{\tr^2\left(\mathbf{M}_{j}\right)\tr\left(\mathbf{M}_{i_{1}}\right)\tr\left(\mathbf{M}_{i_{2}}\right)}
+\frac{5(2+|\gamma_{2}|)^2\tr\left(\mathbf{M}_{i_{1}}\mathbf{M}_{j}\mathbf{M}_{i_{2}}\mathbf{M}_{j}\right)\tr^2\left(\mathbf{M}^2_{j}\right)}
{\tr\left(\mathbf{M}_{i_{1}}\right)\tr\left(\mathbf{M}_{i_{2}}\right)\tr^6\left(\mathbf{M}_{j}\right)}\\
&+\frac{20\tr\left(\mathbf{M}_{i_{1}}\mathbf{M}^2_{j}\mathbf{M}_{i_{2}}\mathbf{M}^2_{j}\right)}
{\tr\left(\mathbf{M}_{i_{1}}\right)\tr\left(\mathbf{M}_{i_{2}}\right)\tr^4\left(\mathbf{M}_{j}\right)}
+\frac{5C^2\gamma^2_{2}T\tr\left(\mathbf{M}_{i_{1}}\mathbf{M}_{j}\right)\tr\left(\mathbf{M}_{i_{2}}\mathbf{M}_{j}\right)}
{\tr\left(\mathbf{M}_{i_{1}}\right)\tr\left(\mathbf{M}_{i_{2}}\right)\tr^4\left(\mathbf{M}_{j}\right)}
+O\left(T^{-3}\right)\\
=&O\left(T^{-3}\right),
\end{align*}
where the last two inequality holds due to $$0\leq \tr^{2}[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}
\epsilon_{i_{1}}\epsilon_{i_{2}}^{\prime}\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})
\circ \mathbf{Q}_{ii}]\leq TC^{2} (\epsilon_{i_{2}}^{\prime}\mathbf{M}_{j}\epsilon_{i_{2}})(\epsilon_{i_{1}}^{\prime}\mathbf{M}_{j}\epsilon_{i_{1}}),$$
by Lemma \ref{lem:matrix inequality}, for some constant $C\geq0.$
So, we can calculate $\mE(B^2),$
\begin{align*}
\mE\left(B^2\right)\leq & N\sum_{j=2}^{N}\frac{16}{N^2(N-2)^2}\mE\left(\underset{i_1<i_2}{\sum^{j-1}\sum^{j-1}}
\underset{i_3<i_4}{\sum^{j-1}\sum^{j-1}}\delta_{i_{1}i_{2}}\delta_{i_{3}i_{4}}\right)\\
\leq&\frac{16}{N(N-2)^2}\sum_{j=2}^{N}\underset{i_1<i_2}{\sum^{j-1}\sum^{j-1}}
\mE\left(\delta^2_{i_{1}i_{2}}\right)\\
=&O\left(T^{-3}\right)=o\left(\sigma^4_{S_{N}}\right),
\end{align*}
then, we obtain that $B=o_{p}\left(\sigma^2_{S_{N}}\right).$
Next, we will focus on (\ref{clt2}) and prove that
\begin{align*}
\sum_{j=2}^{N}\mE\left(Z_{j}^4\right)=o\{\sigma^4_{S_{N}}\}
\end{align*}
due to $\mE\left\{\sum_{j=2}^{N}\mE\left(Z_{j}^4|\mathcal{F}_{j-1}\right)\right\}=\sum_{j=2}^{N}\mE\left(Z_{j}^4\right).$
We divide $\sum_{j=2}^{N}\mE\left(Z_{j}^4\right)$ into two parts:
\begin{align*}
\sum_{j=2}^{N}\mE\left(Z_{j}^4\right)=&\frac{4}{N^2(N-1)^2}\sum_{j=2}^{N}
\sum_{i_{1}=1}^{j-1}\sum_{i_{2}=1}^{j-1}\sum_{i_{3}=1}^{j-1}\sum_{i_{4}=1}^{j-1}
\mE\left(\hat{\rho}_{i_{1} j}\hat{\rho}_{i_{2} j}\hat{\rho}_{i_{3}j}\hat{\rho}_{i_{4}j}\right)\\
=&\frac{4}{N^2(N-1)^2}\sum_{j=2}^{N}\sum_{i=1}^{j-1}\mE\left(\hat{\rho}^4_{i j}\right)
+\frac{12}{N^2(N-1)^2}\sum_{j=2}^{N}\underset{i_{1}\neq i_{2}}{\sum^{j-1}\sum^{j-1}}
\mE\left(\hat{\rho}^2_{i_{1}j}\hat{\rho}^2_{i_{2}j}\right)\\
=&C+D.
\end{align*}
By Lemma \ref{lem:laplace approxi}, we have
\begin{align*}
\mE\left(\hat{\rho}^4_{i j}\right)
=&\mE\left\{\mE\left(\left.\hat{\rho}^4_{i j}\right|\epsilon_{j\cdot}\right)\right\}\\
=&\mE\Bigg[\frac{2\tr\left(\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot}
\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot}\right)+
\tr^2\left(\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot}\right)
+\gamma_{2}\tr[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})]}
{\left(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\right)^2\tr^2\left(\mathbf{M}_{i}\right)}\\
&+
\frac{3\mE\left\{\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}{\epsilon}_{i\cdot}\right)^2|\epsilon_{j\cdot}\right\}
\{2\tr\left(\mathbf{M}^2_{i}\right)+\gamma_{2}\tr(\mathbf{Q}_{ii}\circ\mathbf{Q}_{ii})\}}
{\left(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\right)^2\tr^4\left(\mathbf{M}_{i}\right)}\\
&-\frac{2\mE\left\{\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}{\epsilon}_{i\cdot}\right)^2
\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
{\epsilon}_{i\cdot}\right)|\epsilon_{j\cdot}\right\}-2\mE\left\{\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}{\epsilon}_{i\cdot}\right)^2|\epsilon_{j\cdot}\right\}
\mE\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
{\epsilon}_{i\cdot}\right)}
{\left(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\right)^2\tr^3\left(\mathbf{M}_{i}\right)}+O\left(T^{-2}\right)
\Bigg].
\end{align*}
Then, by Lemma \ref{le:moment of quadratic form},
we have \begin{align*}
&\mE\left\{\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}{\epsilon}_{i\cdot}\right)^2
\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
{\epsilon}_{i\cdot}\right)|\epsilon_{j\cdot}\right\}-\mE\left\{\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}{\epsilon}_{i\cdot}\right)^2|\epsilon_{j\cdot}\right\}
\mE\left({\epsilon}_{i\cdot}^{\prime}\mathbf{P}_{i}
{\epsilon}_{i\cdot}\right)\\
=&\gamma_{4} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}) \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right]\\&+2\gamma_{2} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right] \\
&+8 \gamma_{2} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right] \\
&+4 \gamma_{2} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right]\\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right] \\
&+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right] \\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\\
&+4 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \\
&+8 \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
\geq&\gamma_{4} \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\\&+2\gamma_{2} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \\
&+8 \gamma_{2} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right] \\
&+4 \gamma_{2} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} ) \circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right]\\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\\&+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right] \\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right],
\end{align*}
where the last inequality holds because the last four terms on the right side of the first equation above are all greater than or equal to zero, according to Lemma \ref{lem:matrix inequality}.
Then, we have
\begin{align*}
\mE\left(\hat{\rho}^4_{i j}\right)
\leq&\mE\Bigg(\frac{(3+|\gamma_{2}|)\left(\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot}\right)^2}
{\left(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\right)^2\tr^2\left(\mathbf{M}_{i}\right)}
+\frac{3(3+|\gamma_{2}|)(2+|\gamma_{2}|)\left(\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot}\right)^2\tr\left(\mathbf{M}^2_{i}\right)}
{\left(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\right)^2\tr^4\left(\mathbf{M}_{i}\right)}\\
&+2\tr^{-3}(\mathbf{M}_{i})(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{-2}
\Big\{\gamma_{4} \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\}\\&+2\gamma_{2} \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \\
&+8 \gamma_{2} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right] \\
&+4 \gamma_{2} \tr\left[(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\right]\\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\\
&+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right] \\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\Big\}+O\left(T^{-2}\right)
\Bigg)\\
\leq&\mE\Bigg(\frac{(3+|\gamma_{2}|)\left(\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot}\right)^2}
{\left(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\right)^2\tr^2\left(\mathbf{M}_{i}\right)}
+\frac{3(3+|\gamma_{2}|)(2+|\gamma_{2}|)\left(\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot}\right)^2\tr\left(\mathbf{M}^2_{i}\right)}
{\left(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\right)^2\tr^4\left(\mathbf{M}_{i}\right)}
\\
&+2\tr^{-3}(\mathbf{M}_{i})(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{-2}
\Big\{C|\gamma_{4}| \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+2C|\gamma_{2}| \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right) \\
&+8 |\gamma_{2} | (\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot})^{3/2} (\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}^3_{i}\hat{\epsilon}_{j\cdot})^{1/2}\\
&+4C |\gamma_{2} | \tr\left(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\right)\\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\\
&+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right] \\
&+4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\Big\}+O\left(T^{-2}\right)
\Bigg).
\end{align*}
Because $8 |\gamma_{2} | (\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot})^{3/2} (\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}^3_{i}\hat{\epsilon}_{j\cdot})^{1/2}/(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}=O(1),$
\begin{align*}
&\frac{4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]}{(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}}\\
\leq&4 \gamma_{1}^{2}\sum_{s=1}^{T}\sum_{k=1}^{T}\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)_{sk}\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)_{ss}
\big(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\big)_{kk}\\
\leq&4\gamma_{1}^{2}O(1)\sqrt{T^{2}\tr\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)^{2}}=O(T)
\end{align*}
and \begin{align*}
&4 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ ( \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2} )\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\Big\}\\
\leq&4\gamma_{1}^{2}\sum_{s=1}^{T}\sum_{k=1}^{T}\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)_{sk}^{2}\big(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\big)_{sk}\\
\leq&4\gamma_{1}^{2}O(1)\tr\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)^{2}\\
=&O(1),
\end{align*} we have
\begin{align*}
\mE\left(\hat{\rho}^4_{i j}\right)\leq&\left\{\frac{(3+|\gamma_{2}|)}{\tr^2\left(\mathbf{M}_{i}\right)}+
\frac{3(3+|\gamma_{2}|)(2+|\gamma_{2}|)\tr\left(\mathbf{M}^2_{i}\right)}{\tr^4\left(\mathbf{M}_{i}\right)}\right\}
\mE\left\{\frac{\left(\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot}\right)^2}
{\left(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}\right)^2}\right\}\\
&+\mE\Big[\tr^{-3}(\mathbf{M}_{i})(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{-2}
\Big\{(C|\gamma_{4}|+2C|\gamma_{2}|+4C |\gamma_{2} | ) (\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{M}_{i}\hat{\epsilon}_{j\cdot})^2\\
&+2 \gamma_{1}^{2}\left[{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}\right]\Big\}\Big]\\
&+O\left(T^{-2}\right)\\
\leq&\left\{\frac{(3+|\gamma_{2}|)}{\tr^2\left(\mathbf{M}_{i}\right)}+
\frac{3(3+|\gamma_{2}|)(2+|\gamma_{2}|)\tr\left(\mathbf{M}^2_{i}\right)}{\tr^4\left(\mathbf{M}_{i}\right)}
+\frac{C|\gamma_{4}|+2C|\gamma_{2}|+4C |\gamma_{2} |}{\tr^3(\mathbf{M}_{i})}\right\}\\
\,\,\,\,&\times\bigg[\frac{2\tr\left\{\left(\mathbf{M}_{i}\mathbf{M}_{j}\right)^2\right\}
+\tr^2\left(\mathbf{M}_{i}\mathbf{M}_{j}\right)+\gamma_{2}\tr\{(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{M}_{i}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{j}\mathbf{M}_{i}
\mathbf{P}_{j}\mathbf{\Sigma}^{1/2})\}}{\tr^2\left(\mathbf{M}_{j}\right)}\\
&+O\left(T^{-1}\right)\bigg]\\
&+2 \gamma_{1}^{2}\mE\Big\{\frac{{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}}{\tr^{3}(\mathbf{M}_{i})(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}
}\Big\}+O\left(T^{-2}\right)\\
=&O\left(T^{-2}\right),
\end{align*}
where the last equality holds because of the fact
\begin{align}\label{fact}
&\frac{2\gamma_{1}^{2}{\tau}_{T}^{\prime}\left\{\mathbf{I}_{T} \circ (\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} \mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\left\{\mathbf{I}_{T} \circ(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2})\right\} {\tau}_{T}}{(\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot})^{2}
}\n\\
\leq&2\gamma_{1}^{2}\sum_{s=1}^{T}\sum_{k=1}^{T}\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)_{ss}\big(\mathbf{\Sigma}^{1/2}\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}\big)_{sk}
\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)_{kk}\n\\
\leq&2\gamma_{1}^{2}O(1)\sum_{s=1}^{T}\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)_{ss}\sum_{k=1}^{T}\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)_{kk}\n\\
\leq&2\gamma_{1}^{2}O(1)\tr^{2}\Big(\frac{\mathbf{\Sigma}^{1/2}
\mathbf{P}_{i}\hat{\epsilon}_{j\cdot}\hat{\epsilon}_{j\cdot}^{\prime}
\mathbf{P}_{i}\mathbf{\Sigma}^{1/2}}
{\hat{\epsilon}_{j\cdot}^{\prime}\hat{\epsilon}_{j\cdot}}\Big)=O(1).
\end{align}
So, $C=O\left(N^{-2}T^{-2}\right)=o\left(\sigma^4_{S_{N}}\right).$ Next,
we consider $D.$
\begin{align*}
\mE\left(\hat{\rho}^2_{i_{1}j}\hat{\rho}^2_{i_{2}j}\right)
\leq\sqrt{\mE\left(\hat{\rho}^4_{i_{1}j}\right)\mE\left(\hat{\rho}^4_{i_{2}j}\right)}=O\left(T^{-2}\right).
\end{align*}
Then, $D=O\left(N^{-1}T^{-2}\right)=o\left(\sigma^4_{S_{N}}\right)$ and $\sum_{j=2}^{N}\mE\left(Z_{j}^4\right)=o\{\sigma^4_{S_{N}}\}.$
Then, we complete the proof. \hfill$\Box$
\subsection{Proof of Lemma \ref{th:sum cons}}
Recall that
\begin{align*}
\hat{\sigma}^2_{S_{N}}=\frac{2}{N(N-1)}\sum_{j=2}^{N} \sum_{i=1}^{j-1} v_{j}^{\prime}\left(v_{i}-\bar{v}_{ij}\right) v_{i}^{\prime}\left(v_{j}-\bar{v}_{i j}\right),
\end{align*}
where $\bar{v}_{ij}=\frac{1}{N-2} \sum_{1<\tau \neq i, j<N} v_{\tau}$ and $v_{\tau}=\hat{\epsilon}_{\tau}/\|\hat{\epsilon}_{\tau}\|.$
We have
\begin{align*}
\hat{\sigma}^2_{S_{N}}=&\frac{2}{N(N-1)}\sum_{j=2}^{N} \sum_{i=1}^{j-1} v_{i}^{\prime}\left(v_{j}-\bar{v}_{ij}\right) v_{j}^{\prime}\left(v_{i}-\bar{v}_{i j}\right)\\
=&\frac{2}{N(N-1)}\sum_{j=2}^{N} \sum_{i=1}^{j-1} \left(v_{i}^{\prime}v_{j}\right)^2
-\frac{2}{N(N-1)}\sum_{j=2}^{N} \sum_{i=1}^{j-1}v_{i}^{\prime}v_{j}v_{j}^{\prime}\bar{v}_{i j}\\
&-\frac{2}{N(N-1)}\sum_{j=2}^{N} \sum_{i=1}^{j-1}v_{i}^{\prime}\bar{v}_{i j}v_{j}^{\prime}v_{i}
+\frac{2}{N(N-1)}\sum_{j=2}^{N} \sum_{i=1}^{j-1}v_{i}^{\prime}\bar{v}_{i j}v_{j}^{\prime}\bar{v}_{i j}\\
=&F_{1}-F_{2}-F_{3}+F_{4}.
\end{align*}
Obviously, $\mE\left(F_{1}\right)=\sigma^2_{S_{N}}\left\{1+o(1)\right\}$ and
$\mE\left(F_{i}\right)=0$ for $i=2,3,4.$ Then, we just need to prove that $\var\left(F_{1}\right)=o\left(\sigma^4_{S_{N}}\right)$ and
$$\mE\left(F^2_{i}\right)=o\left(\sigma^4_{S_{N}}\right)$$
for $i=2,\cdots,4.$
\begin{align*}
\var\left(F^2_{1}\right)\leq&\frac{4}{N^2(N-1)^2}\sum_{j=2}^{N}\sum_{i=1}^{j-1}
\mE\left\{\left(v_{i}^{\prime}v_{j}\right)^4\right\}
+\frac{8}{N^2(N-1)^2}\underset{i_{1}<i_{2}<i_{3}}{\sum\sum\sum}\mE\left\{\left(v_{i_{1}}^{\prime}v_{i_{2}}\right)^2
\left(v_{i_{1}}^{\prime}v_{i_{3}}\right)^2\right\}\\
&+\frac{8}{N^2(N-1)^2}\underset{i_{1}<i_{2}<i_{3}}{\sum\sum\sum}\mE\left\{\left(v_{i_{1}}^{\prime}v_{i_{3}}\right)^2
\left(v_{i_{2}}^{\prime}v_{i_{3}}\right)^2\right\}\\&+
\frac{8}{N^2(N-1)^2}\underset{i_{1}<i_{2}<i_{3}}{\sum\sum\sum}\mE\left\{\left(v_{i_{1}}^{\prime}v_{i_{2}}\right)^2
\left(v_{i_{2}}^{\prime}v_{i_{3}}\right)^2\right\}.
\end{align*}
Because $\mE\left\{\left(v_{i_{1}}^{\prime}v_{i_{2}}\right)^2
\left(v_{i_{1}}^{\prime}v_{i_{3}}\right)^2\right\}\leq\sqrt{\mE\left\{\left(v_{i_{1}}^{\prime}v_{i_{2}}\right)^4\right\}
\mE\left\{\left(v_{i_{1}}^{\prime}v_{i_{3}}\right)^4\right\}}$ and $\mE\left\{\left(v_{i_{1}}^{\prime}v_{i_{2}}\right)^4\right\}=\mE\left(\hat{\rho}^4_{i_{1}i_{2}}\right)=O\left(T^{-2}\right),$
we have $\var\left(F^2_{1}\right)=O\left(N^{-1}T^{-2}\right)=o\left(\sigma^4_{S_{N}}\right).$
We next deal with $F_{2}.$
$$F_{2}=\frac{2}{N(N-1)}\sum_{j=2}^{N} \sum_{i=1}^{j-1}v_{i}^{\prime}v_{j}v_{j}^{\prime}\bar{v}_{i j}=\frac{2}{N(N-1)(N-2)}\sum_{j=2}^{N} \sum_{i=1}^{j-1}\sum_{\tau\neq i,j}v_{i}^{\prime}v_{j}v_{j}^{\prime}{v}_{\tau}.$$
So,
\begin{align*}
\mE\left(F^2_{2}\right)=&\frac{4}{N^2(N-1)^2(N-2)^2}\sum_{j_{1}=2}^{N} \sum_{i_{1}=1}^{j_{1}-1}\sum_{\tau_{1}\neq i_{1},j_{1}}\sum_{j_{2}=2}^{N} \sum_{i_{2}=1}^{j_{2}-1}\sum_{\tau_{2}\neq i_{2},j_{2}}\mE\left(v_{i_{1}}^{\prime}v_{j_{1}}v_{j_{1}}^{\prime}v_{\tau_{1}}
v_{i_{2}}^{\prime}v_{j_{2}}v_{j_{2}}^{\prime}v_{\tau_{2}}\right)\\
=&\frac{4}{N^2(N-1)^2(N-2)^2}\sum_{j=2}^{N} \overset{j-1}{\sum_{i=1}} \sum_{\tau\neq i,j}\mE\left(v_{i}^{\prime}v_{j}v_{j}^{\prime}v_{\tau}
v_{i}^{\prime}v_{j}v_{j}^{\prime}v_{\tau}\right)\\
+&\frac{8}{N^2(N-1)^2(N-2)^2}\underset{j_{1}<j_{2}}{\sum^{N}\sum^{N}} \overset{j_{1}-1}{\sum_{i=1}} \sum_{\tau\neq i,j_{1},j_{2}}\mE\left(v_{i}^{\prime}v_{j_{1}}v_{j_{1}}^{\prime}v_{\tau}
v_{i}^{\prime}v_{j_{2}}v_{j_{2}}^{\prime}v_{\tau}\right).
\end{align*}
Because
\begin{align*}
\mE\left(v_{i}^{\prime}v_{j}v_{j}^{\prime}v_{\tau}
v_{i}^{\prime}v_{j}v_{j}^{\prime}v_{\tau}\right)=&
\mE\left\{\left(v_{i}^{\prime}v_{j}v_{j}^{\prime}v_{\tau}\right)^2\right\}\\
=&\mE\left(\hat{\rho}^2_{ij}\hat{\rho}^2_{j\tau}\right)\\
\leq&\sqrt{\mE\left(\hat{\rho}^4_{ij}\right)\mE\left(\hat{\rho}^4_{j\tau}\right)}\\
=&O\left(T^{-2}\right).
\end{align*}
and
\begin{align*}
&\mE\left(v_{i}^{\prime}v_{j_{1}}v_{j_{1}}^{\prime}v_{\tau}
v_{i}^{\prime}v_{j_{2}}v_{j_{2}}^{\prime}v_{\tau}\right)\\
\leq&\sqrt{\mE\left\{\left(v_{i}^{\prime}v_{j_{1}}v_{j_{1}}^{\prime}v_{\tau}\right)^2\right\}
\mE\left\{\left(v_{i}^{\prime}v_{j_{2}}v_{j_{2}}^{\prime}v_{\tau}\right)^2\right\}}\\
=&O\left(T^{-2}\right),
\end{align*}
we have $\mE\left(F^2_{2}\right)=O\left(N^{-2}T^{-2}\right)=o\left(\sigma^4_{S_{N}}\right).$
Similarly, we have $F_3=o_{p}\left(\sigma^2_{S_{N}}\right).$ Now, we consider $F_{4},$
\begin{align*}
F_{4}=&\frac{2}{N(N-1)}\sum_{j=2}^{N} \sum_{i=1}^{j-1}v_{i}^{\prime}\bar{v}_{i j}v_{j}^{\prime}\bar{v}_{i j}\\
=&\frac{2}{N(N-1)(N-2)^2}\sum_{j=2}^{N} \sum_{i=1}^{j-1}\sum_{\tau_{1}\neq i,j}\sum_{\tau_{2}\neq i,j}v_{i}^{\prime}{v}_{\tau_{1}}v_{j}^{\prime}{v}_{\tau_{2}}\\
=&\frac{2}{N(N-1)(N-2)^2}\sum_{j=2}^{N} \sum_{i=1}^{j-1}\sum_{\tau\neq i,j}v_{i}^{\prime}{v}_{\tau}v_{j}^{\prime}{v}_{\tau}+\frac{2}{N(N-1)(N-2)^2}\sum_{j=2}^{N} \sum_{i=1}^{j-1}\sum_{\tau_{1}\neq\tau_{2}\neq i,j}v_{i}^{\prime}{v}_{\tau_{1}}v_{j}^{\prime}{v}_{\tau_{2}}\\
=&F_{41}+F_{42}.
\end{align*}
So,
\begin{align*}
\mE\left(F^2_{41}\right)=&\frac{4}{N^2(N-1)^2(N-2)^4}\sum_{j=2}^{N} \sum_{i=1}^{j-1}\sum_{\tau_{1},\tau_{2}\neq i,j}\mE\left(v_{i}^{\prime}{v}_{\tau_{1}}v_{j}^{\prime}{v}_{\tau_{1}}v_{i}^{\prime}{v}_{\tau_{2}}v_{j}^{\prime}{v}_{\tau_{2}}\right)\\
=&O\left(N^{-4}\right)\sqrt{\mE\left(v_{i}^{\prime}{v}_{\tau_{1}}v_{j}^{\prime}{v}_{\tau_{1}}\right)^2
\mE\left(v_{i}^{\prime}{v}_{\tau_{2}}v_{j}^{\prime}{v}_{\tau_{2}}\right)^2}\\
=&O\left(N^{-4}T^{-2}\right).
\end{align*}
Similarly, we have $\mE\left(F^2_{42}\right)=O\left(N^{-4}T^{-2}\right).$
%\begin{align*}
%\mE\left(F^2_{42}\right)=&\frac{4}{N^2(N-1)^2(N-2)^4}\sum_{j=2}^{N} \sum_{i=1}^{j-1}\sum_{\tau_{1}\neq\tau_{2}\neq i,j}\mE\left(v_{i}^{\prime}{v}_{\tau_{1}}v_{j}^{\prime}{v}_{\tau_{1}}v_{i}^{\prime}{v}_{\tau_{2}}v_{j}^{\prime}{v}_{\tau_{2}}\right)\\
%=&O\left(N^{-4}\right)\mE\left(v_{i}^{\prime}{v}_{\tau_{1}}v_{j}^{\prime}{v}_{\tau_{1}}v_{i}^{\prime}{v}_{\tau_{2}}v_{j}^{\prime}{v}_{\tau_{2}}\right)\\
%=&O\left(N^{-4}\right)\sqrt{\mE\left(v_{i}^{\prime}{v}_{\tau_{1}}v_{j}^{\prime}{v}_{\tau_{1}}\right)^2
%\mE\left(v_{i}^{\prime}{v}_{\tau_{2}}v_{j}^{\prime}{v}_{\tau_{2}}\right)^2}\\
%=&O\left(N^{-4}T^{-2}\right)
%\end{align*}
Next, we can conclude that $\mE\left(F^2_{4}\right)=o\left(\sigma^4_{S_{N}}\right)$ due to
$\mE\left(F^2_{4}\right)\leq2\mE\left(F^2_{41}+F^2_{42}\right),$
which leads to
$$\hat{\sigma}^2_{S_{N}}/{\sigma}^2_{S_{N}} \rightarrow1 ,$$ in probability.\hfill$\Box$
\subsection{Proof of Lemma \ref{indep hat}}
By Theorem \ref{th:max null} and Lemma \ref{th:sum null}, the following hold,
\begin{align}
&\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-4 \log N+\log \log N\rightarrow G(y) \,\,\text {in distribution; }\label{Tmax} \\
&S_{N}/\hat{\sigma}_{S_{N}} \rightarrow \mathcal{N}(0,1) \text { in distribution. }\label{Tsum}
\end{align}
First, we will prove that if $\tilde{T}_{max}-\log N+\log \log N$ and $S_{N}/\sigma_{S_{N}}$ are asymptotically independent,
then $\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-\log N+\log \log N$ and $S_{N}/{\sigma}_{S_{N}}$ are also asymptotically
independent.
Due to (\ref{eiej}), (\ref{dis-rho/}) and (\ref{max:cons}), we have
\begin{align}\label{diff of Lv and LVhat}
\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-\tilde{T}_{max}\rightarrow0,
\end{align}
in probability.
Then, to show asymptotic independence, it is enough to show
\begin{align}\label{indepe}
\lim _{\min(N,T) \rightarrow \infty} P\left(S_{N}/\sigma_{S_{N}} \leq x,\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-4 \log N+\log \log N\leq y\right)=\Phi(x) \cdot G(y),
\end{align}
for any $x \in \mathbb{R}$ and $y \in \mathbb{R},$ where $\Phi(x)=(2 \pi)^{-1 / 2} \int_{-\infty}^{x} e^{-t^{2} / 2} d t$ and $G(y)=\exp \left\{-\frac{1}{\sqrt{8 \pi}} \exp \left(-\frac{y}{2}\right)\right\}.$ Let $$a_{N}\doteq4\log N-\log \log N+y.$$
Due to (\ref{Tmax}) and (\ref{Tsum}), we know (\ref{indepe}) is equivalent to that
\begin{align}\label{indepLV}
\lim _{\min(N,T) \rightarrow \infty} P\left(S_{N}/\sigma_{S_{N}}\leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right)=\Phi(x) \cdot\left\{1-G(y)\right\},
\end{align}
for any $x \in \mathbb{R}$ and $y \in \mathbb{R}$. By the assumption, we know that
\begin{align}\label{indepLVhat}
\lim _{\min(N,T) \rightarrow \infty} P\left(S_{N}/\sigma_{S_{N}} \leq x, \tilde{T}_{max}>a_{N}\right)=\Phi(x) \cdot\left\{1-G(y)\right\},
\end{align}
for any $x \in \mathbb{R}$ and $y \in \mathbb{R}$. We show next that (\ref{indepLVhat}) implies (\ref{indepLV}). By (\ref{diff of Lv and LVhat}),
$$
\left|\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-\tilde{T}_{max}\right| \rightarrow 0
$$
in probability. Given $\epsilon \in(0,1) .$ Set
$$
\Omega=\left\{\left|\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-\tilde{T}_{max}\right|<\epsilon\right\}.
$$
Then
\begin{align}\label{OmegaV=1}
\lim _{\min(N,T) \rightarrow \infty} P\left(\Omega\right)=1.
\end{align}
Now,
\begin{align}\label{OmgaV condition}
& P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right) \n\\
\leq & P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}, \Omega\right)+P\left(\Omega^{c}\right).
\end{align}
$\operatorname{On} \Omega$, if $\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}$ then
\begin{align}\label{LVhatqeg}
\tilde{T}_{max} \geq \frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-\left|\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-\tilde{T}_{max}\right|>a_{N}-\epsilon.
\end{align}
Define
\begin{align*}
\tilde{a}_{N}\doteq4\log N-\log\log N+y-2\epsilon,
\end{align*}
which makes sense for large $T$. Thus,
$$
a_{N}-\tilde{a}_{N}>\epsilon,
$$
and (\ref{LVhatqeg}) conclude that
$$
\tilde{T}_{max} \geq \tilde{a}_{N},
$$
as $p$ is sufficiently large. Review (\ref{OmgaV condition}). We have
$$
\begin{aligned}
& P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right) \\
\leq & P\left(S_{N}/\sigma_{S_{N}} \leq x, \tilde{T}_{max} \geq \tilde{a}_{N}\right)+P\left(\Omega^{c}\right).
\end{aligned}
$$
Immediately from (\ref{indepLVhat}) and (\ref{OmegaV=1}) we get
$$
\limsup _{\min(N,T) \rightarrow \infty} P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right) \leq \Phi(x) \cdot\left\{1-G(y-2 \epsilon)\right\},
$$
for any $\epsilon \in(0,1)$. Inspect that the left-hand side of the above does not depend on $\epsilon$. Letting $\epsilon \downarrow 0$, we obtain
\begin{align}\label{96}
\limsup _{\min(N,T) \rightarrow \infty} P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right) \leq \Phi(x) \cdot\left\{1-G(y)\right\},
\end{align}
for any $x \in \mathbb{R}$ and $y \in \mathbb{R}$. In the following we will show the lower limit. Evidently,
\begin{align}\label{97}
& P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right) \n\\
\geq & P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}, \Omega\right).
\end{align}
Set
$$
\tilde{a}_{N}^{\prime}\doteq4\log N-\log \log N+y+2\epsilon,
$$
Therefore,
$
\tilde{a}_{N}^{\prime}>a_{N}+\epsilon.
$
It is straightforward to verify that
\begin{align}
\left\{\tilde{T}_{max}>\tilde{a}_{N}^{\prime}, \Omega\right\} \subset\left\{\tilde{T}_{max}>a_{N}+\epsilon, \Omega\right\} \subset\left\{\frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}, \Omega\right\},
\end{align}
as $N$ is sufficiently large, where the last inclusion follows from the definition of $\Omega .$ By (\ref{97}),
\begin{align*}
& P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right) \\
\geq & P\left(S_{N}/\sigma_{S_{N}} \leq x, \tilde{T}_{max}>\tilde{a}_{N}^{\prime}, \Omega\right).
\end{align*}
Thus, from (\ref{indepLVhat}) and (\ref{OmegaV=1}) we get
$$
\liminf _{\min(N,T) \rightarrow \infty} P\left(S_{N}/\sigma_{S_{N}} \leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right) \geq \Phi(x) \cdot\left\{1-G(y+2 \epsilon)\right\},
$$
for any $\epsilon \in(0,1)$. Sending $\epsilon \downarrow 0$ we see
$$
\liminf _{\min(N,T) \rightarrow \infty} P\left(S_{N}/\sigma_{S_{N}}\leq x, \frac{\tr^2(\tilde{\mathbf{\Sigma}})}{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}>a_{N}\right) \geq \Phi(x) \cdot\left\{1-G(y)\right\},
$$
for any $x \in \mathbb{R}$ and $y \in \mathbb{R}$.
We have proved that $\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-4 \log N+\log \log N$ and $S_{N}/\sigma_{S_{N}}$ are asymptotically independent.
Similarly, it is easy to prove that $\frac{\tr^2(\tilde{\mathbf{\Sigma}})}
{\|\tilde{\mathbf{\Sigma}}\|^2_{\mathrm{F}}}L_{N}-4 \log N+\log N$ and $S_{N}/\hat{\sigma}_{S_{N}}$ are also asymptotically
independent. Consequently, we complete the proof.
\hfill$\Box$
\subsection{Proof of Lemma \ref{linear}}
For $I_{l}$ appeared in $H(N, k),$ write $I_{l}=\left(i_{l}, j_{l}\right)$ for $l=1, \cdots, k .$ Now we classify the indices $I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N}$ in the definition of $H(N, k)$ into three cases. Let $\Gamma_{N, 1}$ be the set of indices $\left(I_{1}, \cdots, I_{k}\right)$ such that no two of the $2 k$ indices $\left\{i_{l}, j_{l} ; l=1, \cdots, k\right\}$ are identical. Let $\Gamma_{N, 2}$ be the set of indices $\left(I_{1}, \cdots, I_{k}\right)$ such that either $i_{1}=$ $\cdots=i_{k}$ or $j_{1}=\cdots=j_{k} .$ Let $\Gamma_{N, 3}$ be the set of indices $I_{1}<I_{2}<\cdots<I_{k} \in$
$\Lambda_{N}$ excluding $\Gamma_{N, 1} \cup \Gamma_{N, 2} .$ In the following we will estimate
$$
D_{j}:=\sum_{I_{1}<I_{2}<\cdots<I_{k} \in \Gamma_{N, j}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)
$$
for $j=1,2,3$ one by one. We will see $D_{1}$ contributes essentially the sum in the expression of $H(N, k)$ by an easy argument; the term $D_{2}$ is negligible
and its computation is trivial; the term $D_{3}$ is also negligible but its estimate is most involved.
Step 1: the estimate of $D_{1}$. Recall $B_{I}=\left\{\left|\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot} \right|\geq l_{N}\right\}$ if $I=(i, j) \in$ $\Lambda_{N}.$ By the definition of $\Gamma_{N, 1},$ we know that $B_{I_{1}}, B_{I_{2}} \cdots, B_{I_{k}}$ are independent. For large $N$,
$$
\max _{I \in \Lambda_{N}} \mathbb{P}\left(B_{I}\right)=\mathbb{P}\left\{\left|\epsilon_{i\cdot}^{\prime}\epsilon_{j\cdot} \right|\geq l_{N}\right\} \leq \frac{C}{N^{2}}
$$
where the inequality holds due to (\ref{T12}). Then,
\begin{align}\label{O1}
D_{1} \leq \frac{C}{N^{2 k}} \cdot\binom{\frac{N(N-1)}{2}}{k}\leq \frac{C}{k !}
\end{align}
Step
2: the estimate of $D_{2}$. Evidently, the size of $\Gamma_{N, 2}$ is no more than $\binom{N}{1}\cdot\binom{N}{k}\cdot2 \leq 2 N^{k+1} .$
Recall that $z_{i}$ is the $i$-th row vector of matrix $\mathbf{Z}.$
Similarly, there exists some $C_{3} > 0$ such that for fixed $k>0$ and $a_{l}\in\{-1,1\},$ $1\leq l\leq k, $
that
$$
\mathbb{P}\left(\frac{\Big(\sum_{l=1}^{k}a_{l}z_{{l}}\Big)^{\prime}\mathbf{\Sigma}^2\Big(\sum_{l=1}^{k}a_{l}z_{{l}}\Big)}
{\tr(\mathbf{\Sigma}^{2})}>k(1+\varepsilon_{3})\right)
\leq2N^{-3k},
$$
where $\varepsilon_{3}=C_{3}\sqrt{\log N/\tr(\mathbf{\Sigma}^2)}.$
Again, by Cram\'er type moderate deviation results in \cite{statulevivcius1966} and Theorem 1.1 in \cite{rudelson2013}, for $j_{1}\neq j_{2}\neq \cdots\neq j_{k},$
\begin{align}\label{step2}
&\mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)\n\\
=&\mathbb{P}\left(|\epsilon_{i_{1}\cdot}^{\prime}\epsilon_{j_{1}\cdot}|>l_{N},\cdots, |\epsilon_{i_{1}\cdot}^{\prime}\epsilon_{j_{k}\cdot}|>l_{N}\right)\n\\
\leq&\sum_{a_{l}\in\{-1,1\}}\mathbb{P}\bigg\{\Big|\epsilon_{i_{1}\cdot}^{\prime}\Big(\sum_{l=1}^{k}a_{l}\epsilon_{j_{l}\cdot}
\Big)\Big|>k\cdot l_{N}\bigg\}\n\\
\leq&\sum_{a_{l}\in\{-1,1\}}\mathbb{P}\bigg\{\Big|z_{i1}^{\prime}\mathbf{\Sigma}\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)
\Big|>k\cdot \sqrt{a_{N}\tr(\mathbf{\Sigma}^2)}\bigg\}\n\\
\leq&\sum_{a_{l}\in\{-1,1\}}\mathbb{P}\bigg\{\frac{\Big|z_{i1}^{\prime}\mathbf{\Sigma}\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)
\Big|}{\sqrt{\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)^{\prime}\mathbf{\Sigma}^2\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)}}>\frac{\sqrt{ka_{N}}}{\sqrt{1+\varepsilon_{3}}}\bigg\}\n\\
&+\sum_{a_{l}\in\{-1,1\}}
\mathbb{P}\bigg\{\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)^{\prime}\mathbf{\Sigma}^2\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)
>k(1+\varepsilon_{3})\tr(\mathbf{\Sigma}^2)\bigg\}\n\\
\leq&\sum_{a_{l}\in\{-1,1\}}\mE\Bigg[\mE\bigg\{\frac{\Big|z_{i1}^{\prime}\mathbf{\Sigma}\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)
\Big|}{\sqrt{\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)^{\prime}\mathbf{\Sigma}^2\Big(\sum_{l=1}^{k}a_{l}z_{j_{l}}\Big)}}>\frac{ \sqrt{ka_{N}}}{\sqrt{1+\varepsilon_{3}}}|z_{j_{1}},\cdots,z_{j_{k}}\bigg\}\Bigg]+2^{k+1}N^{-3k}\n\\
\leq&\sum_{a_{l}\in\{-1,1\}}\left\{1+o(1)\right\}\frac{2}{\sqrt{ka_{N}2\pi}}\exp\left(-ka_{N}/2\right)+2^{k+1}N^{-3k}\n\\
\leq&\frac{2^{k+1}(2\pi)^{-1/2}}{\sqrt{4k\log N}}\exp\left[-2k\log N+k\log \log N /2-ky/2\right]+2^{k+1}N^{-3k}\n\\
\leq&2^{k+1}\left(\log N\right)^{k/2-1/2}N^{-2k}\exp(-ky/2)+2^{k+1}N^{-3k},
\end{align}
where $ka_{N}=4k\log N-k\log \log N +ky.$ So, for $k\geq 2,$ we have
\begin{align}\label{O2}
D_{2} & \leq 2 N^{k+1} \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)\n\\
&=2 N^{k+1} \cdot 2^{k+1}\left(\log N\right)^{k/2-1/2}N^{-2k}\exp(-ky/2) \rightarrow 0.
\end{align}
Step 3 : the estimate of $D_{3}$. Fix a tuple $\left(I_{1}, I_{2}, \cdots, I_{l}\right) \in \Gamma_{N, 3}$. By the ordering imposed on $\Lambda_{N},$ we see that $i_{1} \leq i_{2} \leq \cdots \leq i_{k} .$ Let's assume that
$$i_{1}=i_{2}=\cdots=i_{n_{1}},$$
$$i_{n_{1}+1}=i_{n_{1}+2}=\cdots=i_{n_{1}+n_{2}},$$
$$\cdots\cdots\cdots\cdots\cdots\cdots\cdots\cdots\cdots$$
$$i_{n_{1}+\cdots+n_{r-1}+1}=i_{n_{1}+\cdots+n_{r-1}+2}=\cdots=i_{n_{1}+\cdots+n_{r}},$$
where $n_{1}+\cdots+n_{r}=k.$
Let $\mathcal{F}_{1}$ be the set of random vectors $\left\{\epsilon_{j_{1}\cdot}, \epsilon_{i_{l}\cdot}, \epsilon_{j_{l}\cdot} ; 2 \leq l \leq k\right\}$.
Similar to (\ref{step2}),
for sufficiently large constant $C>0,$ we have
$$
\begin{aligned}
&\P\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right) \\ =&\mE\left\{\P\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}} \mid \mathcal{F}_{1}\right)\right\} \\
=&\mE\left[\P\left\{\min_{1\leq l\leq n_{1}}|\epsilon_{i_{1}\cdot}^{\prime}\epsilon_{j_{l}\cdot}|\geq l_{N} \mid \mathcal{F}_{1}\right\} \cdot \prod_{l=n_{1}+1}^{k} I\left(B_{I_{l}}\right)\right]\\
\leq &\mE\left[\sum_{a_{l}\in\{-1,1\}}\P\left\{\Big|\epsilon_{i_{1}\cdot}^{\prime}(\sum_{l=1}^{n_{1}}a_{l}\epsilon_{j_{l}\cdot})\Big|\geq l_{N} \mid \mathcal{F}_{1}\right\} \cdot \prod_{l=n_{1}+1}^{k} I\left(B_{I_{l}}\right)\right]\\
\leq&\mE\left[\sum_{a_{l}\in\{-1,1\}}\P\left\{\Big|\epsilon_{i_{1}\cdot}^{\prime}(\sum_{l=1}^{n_{1}}a_{l}\epsilon_{j_{l}\cdot})\Big|\geq l_{N} \mid \mathcal{F}_{1}\right\}\right. \\
&\left.\quad\quad\cdot I\Big\{\frac{\big(\sum_{l=1}^{n_{1}}a_{l}z_{j_{l}}\big)^{\prime}\mathbf{\Sigma}^{2}\big(\sum_{l=1}^{n_{1}}a_{l}z_{j_{l}}\big)}
{n_{1}\tr(\mathbf{\Sigma}^{2})}>1+C\sqrt{\frac{\log N}{\tr(\mathbf{\Sigma}^{2})}}\Big\}\cdot \prod_{l=n_{1}+1}^{k} I\left(B_{I_{l}}\right)\right]\\
&+\mE\left[I\Big\{\frac{\big(\sum_{l=1}^{n_{1}}a_{l}z_{j_{l}}\big)^{\prime}\mathbf{\Sigma}^{2}\big(\sum_{l=1}^{n_{1}}a_{l}z_{j_{l}}\big)}
{n_{1}\tr(\mathbf{\Sigma}^{2})}>1+C\sqrt{\frac{\log N}{\tr(\mathbf{\Sigma}^{2})}}\Big\}\cdot \prod_{l=n_{1}+1}^{k} I\left(B_{I_{l}}\right)\right]\\
\leq&O(N^{-2n_{1}})\mE\left[\prod_{l=n_{1}+1}^{k} I\left(B_{I_{l}}\right)\right]+2N^{-3k}\\
\leq &O(N^{-2n_{1}})\P\left( B_{I_{n_{1}+1}} \cdots B_{I_{k}}\right)+2N^{-3k}.
\end{aligned}
$$
In the same way, we finally obtain
\begin{align*}
\P\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)=&\P\Big(\prod_{l=1}^{n_{1}}B_{I_{l}}\Big)\cdots\P\Big(\prod_{l=n_{r-1}+1}^{n_{r}}B_{I_{l}}\Big)\\
\leq& \frac{C}{N^{2n_{1}+\cdots+2n_{r}}}\\
\leq&\frac{C}{N^{2k}},
\end{align*}
for some constant $C$ and sufficiently large $N$.
Recall $I_{l}=\left(i_{l}, j_{l}\right)$ for each $1 \leq l \leq k .$ In view of the definition of $\Gamma_{N, 3},$ there are at least two of the $2 k$ indices from $\left\{\left(i_{l}, j_{l}\right) ; 1 \leq l \leq k\right\}$ are identical for any $\left(I_{1}, \cdots, I_{k}\right) \in \Gamma_{N, 3} .$ Let $\kappa=\left|\left\{i_{l}, j_{l} ; 1 \leq l \leq k\right\}\right|$ for such $\left(I_{1}, I_{2}, \cdots, I_{k}\right).$
Easily, $k+1 \leq \kappa \leq 2 k-1 .$ To see how many such $\left(I_{1}, \cdots, I_{k}\right)$ with $\mid\left\{i_{l}, j_{l} ; 1 \leq\right.$ $l \leq k\} \mid=\kappa,$ first pick $\kappa$ many indices from $\{1,2, \cdots, N\},$ which has the total number of ways $\binom{N}{\kappa}\leq N^{\kappa},$ then use the $\kappa$ many indices to make a $\left(I_{1}, \cdots, I_{k}\right) \in \Gamma_{N, 3} .$ The total number of ways to do so is no more than $\kappa^{2 k}$. Therefore,
$$
\left|\Gamma_{N, 3}\right| \leq \sum_{\kappa=k+1}^{2 k-1} N^{\kappa} \cdot \kappa^{2 k} \leq(2 k)^{2 k} \cdot N^{2 k-1}.
$$
Similar to the (\ref{step2}), for fixed $k$,
we can have $\mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)=O(N^{-2k})$ and
\begin{align*}
D_{3} \leq&(2 k)^{2 k} \cdot2^k\left(\log N\right)^{k/2-1/2}N^{-2k}\exp(-ky/2)\\
\leq&(2 k)^{2 k} \cdot2^k\exp(-ky/2)\frac{\left(\log N\right)^{k/2-1/2}}{N}\rightarrow 0
\end{align*}
as $N$ is sufficiently large. \hfill$\Box$
\subsection{Proof of Lemma \ref{linear2}}
For $I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N},$ write $I_{l}=\left(i_{l}, j_{l}\right)$ for $l=1,2, \cdots, k .$ Set
$$
\Lambda_{N, k}=\left\{\left(i_{l}, j\right) ; i_{l}<j \leq N, 1 \leq l \leq k\right\} \bigcup\left\{\left(i, j_{l}\right) ; 1 \leq i<j_{l}, 1 \leq l \leq k\right\}
$$
for $k \geq 1 .$ It is easy to check that $\left|\Lambda_{N, k}\right|=\sum_{l=1}^{k}\left(N-i_{l}+j_{l}-2\right) .$ since $i_{l}<j_{l}$ for each $l,$ we see that
$$
k(N-1) \leq\left|\Lambda_{N, k}\right| \leq \sum_{l=1}^{k}\left(N+j_{l}\right) \leq 2 k N.
$$
Recall
$$
A_{N}=\left\{S_{N}/\sqrt{\sigma^2_{S_{N}}} \leq x\right\} \quad x \in \mathbb{R},
$$
for $N\geq 3$ and
$$
S_{N}^{k}=\sqrt{\frac{2}{N(N-1)}}\sum_{(i, j) \in \Lambda_{N, k}}\hat{\rho}_{ij},
$$
for $N \geq k \geq 1$.
Observe that $B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}$ is an event generated by random vectors $\left\{\hat{\rho}_{ij};(i, j) \in \Lambda_{N, k}\right\} .$ A crucial observation is that $S_{N}-S_{N}^{k}$ is independent of $B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}} .$ It is easy to see that
\begin{align*}
S_{N}^{k} =&\sqrt{\frac{2}{N(N-1)}}{\sum\sum}_{\left\{\left(i_{l}, j\right) ; i_{l}<j \leq N, 1 \leq l \leq k\right\}}\hat{\rho}_{i_{l}j}\\
&+\sqrt{\frac{2}{N(N-1)}}{\sum\sum}_{\left\{\left(i, j_{l}\right) ; 1 \leq i<j_{l}, 1 \leq l \leq k\right\}} \hat{\rho}_{ij_{l}}\\
&-\sqrt{\frac{2}{N(N-1)}}\sum_{s=1}^{k}\sum_{t=1}^{k} \hat{\rho}_{i_{s} j_{t}}\\
\doteq&D_{N, 1}+D_{N, 2}-D_{N, 3}.
\end{align*}
Fix $\epsilon \in(0,1),$ for even $\tau$ and $\tau>2$,
we have
\begin{align*}
\mE\left|D_{N, 1}\right|^{\tau}\leq k^{\tau-1}\frac{2^{\tau/2}}{N^{\tau/2}(N-1)^{\tau/2}}\sum_{l=1}^{k}\mE\left(\left|\sum_{j=i_{l}+1}^{N}\hat{\rho}_{i_{l}j}\right|^{\tau}\right),
\end{align*}
where
$$
\mE\left\{\left(\sum_{j=i_{l}+1}^{N}\hat{\rho}_{i_{l}j}\right)^{\tau}\right\}
=\mE\left[\mE\left\{\left.\left(\sum_{j=i_{l}+1}^{N}\hat{\rho}_{i_{l}j}\right)^{\tau}\right|\epsilon_{i_{l}\cdot}\right\}\right].
$$
By Lemma 2 in \cite{feng2020}, there exists a constant $K_{\tau}> 0$ depending on $\tau$ only such that
$$\mE\left\{\left.\left(\sum_{j=i_{l}+1}^{N}\hat{\rho}_{i_{l}j}\right)^{\tau}\right|\epsilon_{i_{l}\cdot}\right\}
\leq K_{\tau}(N-i_{l})^{\tau/2-1}\sum_{j=i_{l}+1}^{N}\mE\left\{\left.\left(\hat{\rho}_{i_{l}j}\right)^{\tau}\right|\epsilon_{i_{l}\cdot}\right\}.$$
So, we have
$$\mE\left|D_{N, 1}\right|^{\tau}\leq k^{\tau-1}\frac{2^{\tau/2}}{N^{\tau/2}(N-1)^{\tau/2}}\sum_{l=1}^{k}
K_{\tau}(N-i_{l})^{\tau/2-1}\sum_{j=i_{l}+1}^{N}\mE\left[\mE\left\{\left.\left(
\hat{\rho}_{i_{l}j}\right)^{\tau}\right|\epsilon_{i_{l}\cdot}\right\}\right].
$$
By Khintchine's inequality (Exercise 2.6.5 in \cite{vershynin2018}) and Lemma \ref{lem:laplace approxi}, it's easy to obtain that
$\mE\left[\mE\left\{\left.\left(
\hat{\rho}_{i_{l}j}\right)^{\tau}\right|\epsilon_{i_{l}\cdot}\right\}\right]\leq (\tau/2)^{\tau/4}O\left(T^{-\tau/2}\right).$
Hence, together with $\sigma_{S_{N}}=O\left(T^{-1/2}\right),$ we obtain that
\begin{align*}
\mathbb{P}\left(\left|D_{N,1}\right|>\sigma_{S_{N}}\epsilon\right) & \leq \frac{\mE\left|D_{N,1}\right|^{\tau}}{\sigma^{\tau}_{S_{N}}\epsilon^{\tau}}\\
&=\frac{K_{\tau}k^{\tau}\tau^{\tau/4}2^{\tau/4}}{\epsilon^{\tau}}\cdot O\left(N^{-\tau/2}\right),
\end{align*}
where the last equality holds due to Assumption \ref{assum:x_it}.
Similarly,
$$
\mathbb{P}\left(\left|D_{N,2}\right|>\sigma_{S_{N}}\epsilon\right) \leq \frac{k^{\tau}\tau^{\tau/4}2^{\tau/4}}{\epsilon^{\tau}}\cdot O\left(N^{-\tau/2}\right).
$$
Lastly, for even $\tau$ and $\tau>2$,
\begin{align*}
\mE\left(\left|D_{N, 3}\right|^{\tau}\right) & \leq \frac{2^{\tau/2}}{N^{\tau/2}(N-1)^{\tau/2}} \cdot k^{2(\tau-1)} \cdot \sum_{s=1}^{k} \sum_{t=1}^{k} \mE\left(\left|\hat{\rho}_{i_{s} j_{t}}\right|^{\tau}\right) \\
& \leq \frac{2^{\tau/2}}{N^{\tau/2}(N-1)^{\tau/2}} \cdot k^{2\tau}\mE\left(\left|\hat{\rho}_{i_{s} j_{t}}\right|^{\tau}\right).
\end{align*}
So, we have
$$
\begin{aligned}
\mathbb{P}\left\{\frac{|S_{N}^k|}{\sigma_{S_{N}}} \geq \epsilon\right\} &\leq C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}},
\end{aligned}
$$
for large $N$, where $C^{\prime}$ is a constant depending on $\epsilon$ but free of $N.$
Fix $I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N} .$ By the definition of $A_{N}$,
$$
\begin{aligned}
& \mathbb{P}\left\{A_{N}(x) B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right\} \\
\leq & \mathbb{P}\left\{A_{N}(x) B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}, \frac{\left|S_{N}^{k} \right|}{\sigma_{S_{N}}} <\epsilon\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}} \\
\leq & \mathbb{P}\left\{\frac{S_{N}-S_{N}^{k}}{\sigma_{S_{N}}} \leq x+\epsilon, B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}} \\
=& \mathbb{P}\left\{\frac{S_{N}-S_{N}^{k}}{\sigma_{S_{N}}} \leq x+\epsilon\right\} \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}},
\end{aligned}
$$
by the independence between $S_{N}-S_{N}^{k}$ and $B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}} .$ Now
$$
\begin{aligned}
& \mathbb{P}\left\{\frac{S_{N}-S_{N}^{k}}{\sigma_{S_{N}}} \leq x+\epsilon\right\} \\
\leq & \mathbb{P}\left\{\frac{S_{N}-S_{N}^{k}}{\sigma_{S_{N}}} \leq x+\epsilon,\frac{\left|S_{N}^{k}\right|}{\sigma_{S_{N}}} <\epsilon\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}} \\
\leq & \mathbb{P}\left\{\frac{S_{N}}{\sigma_{S_{N}}} \leq x+2 \epsilon\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}} \\
\leq & \mathbb{P}\left\{A_{N}(x+2 \epsilon)\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}}.
\end{aligned}
$$
Combing the two inequalities to get
\begin{align}\label{leq1}
& \mathbb{P}\left\{A_{N}(x) B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right\}\n \\
\leq & \mathbb{P}\left\{A_{N}(x+2 \epsilon)\right\} \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)+2 C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}}.
\end{align}
Similarly,
$$
\begin{aligned}
& \mathbb{P}\left\{\frac{S_{N}-S_{N}^k}{\sigma_{S_{N}}} \leq x-\epsilon, B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right\} \\
\leq & \mathbb{P}\left\{\frac{S_{N}-S_{N}^k}{\sigma_{S_{N}}} \leq x-\epsilon, B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}},\right.\\
&\left.\frac{\left|S_{N}^k\right|}{\sigma_{S_{N}}} <\epsilon\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}} \\
\leq & \mathbb{P}\left\{\frac{S_{N}}{\sigma_{S_{N}}} \leq x, B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}}.
\end{aligned}
$$
In other words, by independence,
\begin{align*}
&\mathbb{P}\left\{A_{N}(x) B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right\} \\
\geq&\mathbb{P}\left\{\frac{S_{N}-S_{N}^k}{\sigma_{S_{N}}} \leq x-\epsilon\right\}\cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)-C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}}.
\end{align*}
Furthermore,
\begin{align}
&\mathbb{P}\left\{\frac{S_{N}}{\sigma_{S_{N}}} \leq x-2 \epsilon\right\}\n\\
\leq &\mathbb{P}\left\{\frac{S_{N}}{\sigma_{S_{N}}} \leq x-2 \epsilon, \frac{\left|S_{N}^k\right|}{\sigma_{S_{N}}} <\epsilon\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}} \n\\
\leq& \mathbb{P}\left\{\frac{S_{N}-S_{N}^k}{\sigma_{S_{N}}} \leq x-\epsilon\right\}+C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}}.
\end{align}
The above two strings of inequalities imply
$$
\begin{aligned}
& \mathbb{P}\left\{A_{N}(x) B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right\} \\
\geq & \mathbb{P}\left\{\frac{S_{N}}{\sigma_{S_{N}}} \leq x-2 \epsilon\right\} \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)-2 C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}},
\end{aligned}
$$
which joining with (\ref{leq1}) yields
$$
\begin{aligned}
&\left|\mathbb{P}\left\{A_{N}(x) B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right\}-\mathbb{P}\left\{A_{N}(x)\right\} \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)\right| \\
\leq & \Delta_{N, \epsilon} \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)+4 C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}},
\end{aligned}
$$
where
$$
\Delta_{N, \epsilon}\doteq\left|\mathbb{P}\left\{A_{N}(x)\right\}-\mathbb{P}\left\{A_{N}(x+2 \epsilon)\right\}\right|+\left| \mathbb{P}\left\{A_{N}(x)\right\}-\mathbb{P}\left\{A_{N}(x-2 \epsilon) \right\}\right|.
$$
In particular,
\begin{align}\label{diff of phi}
\Delta_{N, \epsilon} \rightarrow|\Phi(x+2 \epsilon)-\Phi(x)|+|\Phi(x-2 \epsilon)-\Phi(x)|,
\end{align}
as $\min(N,T) \rightarrow \infty$ by Theorem \ref{th:max null}. As a consequence,
$$
\begin{aligned}
\zeta(N, k)\doteq& \sum_{I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N}}\left[\mathbb{P}\left(A_{N}(x) B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)-\right.\\
&\left.\mathbb{P}\left\{A_{N}(x)\right\} \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)\right] \\
\leq & \sum_{I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{N}}\left\{\Delta_{N, \epsilon} \cdot \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)+4 C^{\prime} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}}\right\} \\
\leq & \Delta_{N, \epsilon} \cdot H(N, k)+\left(4 C^{\prime}\right) \cdot\binom{\frac{1}{2}N(N-1)}{k} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2}},
\end{aligned}
$$
where
$$
H(N, k)=\sum_{I_{1}<I_{2}<\cdots<I_{k} \in \Lambda_{k}} \mathbb{P}\left(B_{I_{1}} B_{I_{2}} \cdots B_{I_{k}}\right)
$$
as defined in Lemmas \ref{linear} and \ref{linear2}, we know $\lim \sup _{\min(N,T) \rightarrow \infty} H(N, k) \leq C / k !$, where $C$ is a universal constant. Picking $\tau=6k,$ and using the trivial fact $\binom{r}{i} \leq r^{i}$ for any integers $1 \leq i \leq r,$ we have that
$$
\dbinom{\frac{1}{2}N(N-1)}{k}\cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2 }} \leq N^{2 k} \cdot \frac{k^{2\tau}\tau^{\tau/4}}{N^{\tau/2 }} \leq \frac{k^{2\tau}\tau^{\tau/4}}{N^{k}}\rightarrow 0.
$$
Hence, from (\ref{diff of phi})
$$
\begin{aligned}
\limsup _{\min(N,T) \rightarrow \infty} \zeta(N, k) & \leq \frac{C}{k !} \cdot \limsup _{\min(N,T) \rightarrow \infty} \Delta_{N, \epsilon} \\
&=\frac{C}{k !} \cdot\left\{|\Phi(x+2 \epsilon)-\Phi(x)|+|\Phi(x-2 \epsilon)-\Phi(x)|\right\},
\end{aligned}
$$
for any $\epsilon>0 .$ The desired result follows by sending $\epsilon \downarrow 0$.
\hfill$\Box$
%\bibliographystyle{asa}
%\begingroup
%\baselineskip=16.5pt
%\bibliography{reference}
%\endgroup
\begin{thebibliography}{39}
\newcommand{\enquote}[1]{``#1''}
\expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
\bibitem[{Abadir and Magnus(2005)}]{abadir2005matrix}
Abadir, K.~M. and Magnus, J.~R. (2005), \textit{Matrix algebra}, vol.~1,
Cambridge University Press.
\bibitem[{Anselin and Bera(1998)}]{anselin1998}
Anselin, L. and Bera, A.~K. (1998), \enquote{Spatial dependence in linear
regression models with an introduction to spatial econometrics,}
\textit{Statistics textbooks and monographs}, 155, 237--290.
\bibitem[{Arratia et~al.(1989)Arratia, Goldstein, and Gordon}]{arratia1989}
Arratia, R., Goldstein, L. and Gordon, L. (1989), \enquote{Two moments suffice
for Poisson approximations: the Chen-Stein method,} \textit{The Annals of
Probability}, 9--25.
\bibitem[{Baltagi et~al.(2012)Baltagi, Feng, and Kao}]{baltagi2012}
Baltagi, B.~H., Feng, Q. and Kao, C. (2012), \enquote{A Lagrange Multiplier
test for cross-sectional dependence in a fixed effects panel data model,}
\textit{Journal of Econometrics}, 170, 164--177.
\bibitem[{Baltagi et~al.(2016)Baltagi, Kao, and Peng}]{baltagi2016}
Baltagi, B.~H., Kao, C. and Peng, B. (2016), \enquote{Testing cross-sectional
correlation in large panel data models with serial correlation,}
\textit{Econometrics}, 4, 44.
\bibitem[{Bao and Ullah(2010)}]{bao2010expectation}
Bao, Y. and Ullah, A. (2010), \enquote{Expectation of quadratic forms in normal
and nonnormal variables with applications,} \textit{Journal of Statistical
Planning and Inference}, 140, 1193--1205.
\bibitem[{Bernard(1987)}]{bernard1987cross}
Bernard, V.~L. (1987), \enquote{Cross-sectional dependence and problems in
inference in market-based accounting research,} \textit{Journal of Accounting
Research}, 1--48.
\bibitem[{Bernstein(2009)}]{bernstein2009matrix}
Bernstein, D.~S. (2009), \textit{Matrix mathematics}, Princeton university
press.
\bibitem[{Box et~al.(2015)Box, Jenkins, Reinsel, and Ljung}]{box2015}
Box, G.~E., Jenkins, G.~M., Reinsel, G.~C., and Ljung, G.~M. (2015),
\textit{Time series analysis: forecasting and control}, John Wiley \& Sons.
\bibitem[{Breusch and Pagan(1980)}]{breusch1980l}
Breusch, T.~S. and Pagan, A.~R. (1980), \enquote{The Lagrange multiplier test
and its applications to model specification in econometrics,} \textit{The
review of economic studies}, 47, 239--253.
\bibitem[{Cai et~al.(2016)Cai, Ren, and Zhou}]{crz2016}
Cai, T.~T., Ren, Z. and Zhou, H.~H. (2016), \enquote{Estimating structured
high-dimensional covariance and precision matrices: optimal rates and
adaptive estimation,} \textit{Electronic Journal of Statistics}, 10, 1--59.
\bibitem[{Chen and Liu(2018)}]{chen2018}
Chen, X. and Liu, W. (2018), \enquote{Testing independence with
high-dimensional correlated samples,} \textit{The Annals of Statistics}, 46,
866--894.
\bibitem[{Fama and French(1993)}]{Fama1993Common}
Fama, E. and French, K. (1993), \enquote{Common risk factors in the returns on
stocks and bonds,} \textit{Journal of Financial Economics}, 33, 3--56.
\bibitem[{Feng et~al.(2020)Feng, Ding, and Liu}]{feng2020fixed}
Feng, L., Ding, Y. and Liu, B. (2020), \enquote{Rank-based Tests for
Cross-sectional Dependence in Large (N, T) Fixed Effects Panel Data Models,}
\textit{Oxford Bulletin of Economics and Statistics}, 82, 1198--1216.
\bibitem[{Feng et~al.(2022)Feng, Jiang, Liu, and Xiong}]{feng2020}
Feng, L., Jiang, T., Liu, B., and Xiong, W. (2022), \enquote{Max-sum tests for
cross-sectional dependence of high-demensional panel data,} \textit{The
Annals of Statistics}, 50, 1124--1143.
\bibitem[{Gao et~al.(2017)Gao, Han, Pan, and Yang}]{gao2017high}
Gao, J., Han, X., Pan, G., and Yang, Y. (2017), \enquote{High dimensional
correlation matrices: The central limit theorem and its applications,}
\textit{Journal of the Royal Statistical Society: Series B (Statistical
Methodology)}, 79, 677--693.
\bibitem[{Hall and Heyde(2014)}]{hall2014}
Hall, P. and Heyde, C.~C. (2014), \textit{Martingale limit theory and its
application}, Academic press.
\bibitem[{Hong(2010)}]{Hong2010}
Hong, Y. (2010), \textit{Macroeconometrics and Time Series Analysis}, Palgrave
Macmillan UK.
\bibitem[{Hornstein et~al.(2018)Hornstein, Fan, Shedden, and
Zhou}]{hornstein2018joint}
Hornstein, M., Fan, R., Shedden, K., and Zhou, S. (2018), \enquote{Joint mean
and covariance estimation with unreplicated matrix-variate data,}
\textit{Journal of the American Statistical Association}.
\bibitem[{Kapoor et~al.(2007)Kapoor, Kelejian, and Prucha}]{kapoor2007}
Kapoor, M., Kelejian, H.~H. and Prucha, I.~R. (2007), \enquote{Panel data
models with spatially correlated error components,} \textit{Journal of
econometrics}, 140, 97--130.
\bibitem[{Kelejian and Prucha(1999)}]{kelejian1999}
Kelejian, H.~H. and Prucha, I.~R. (1999), \enquote{A generalized moments
estimator for the autoregressive parameter in a spatial model,}
\textit{International economic review}, 40, 509--533.
\bibitem[{Lan et~al.(2017)Lan, Pan, Luo, and Cheng}]{lan2017}
Lan, W., Pan, R., Luo, R., and Cheng, Y. (2017), \enquote{High dimensional
cross-sectional dependence test under arbitrary serial correlation,}
\textit{Science China Mathematics}, 60, 345--360.
\bibitem[{Lee(2007)}]{lee2007}
Lee, L.-f. (2007), \enquote{GMM and 2SLS estimation of mixed regressive,
spatial autoregressive models,} \textit{Journal of Econometrics}, 137,
489--514.
\bibitem[{Lee and Yu(2010)}]{lee2010}
Lee, L.-f. and Yu, J. (2010), \enquote{Estimation of spatial autoregressive
panel data models with fixed effects,} \textit{Journal of econometrics}, 154,
165--185.
\bibitem[{Li and Xue(2015)}]{lx2015}
Li, D. and Xue, L. (2015), \enquote{Joint limiting laws for high-dimensional
independence tests,} \textit{arXiv:1512.08819}.
\bibitem[{Li et~al.(2019)Li, Lam, Yao, and Yao}]{lz2019}
Li, Z., Lam, C., Yao, J., and Yao, Q. (2019), \enquote{On testing for
high-dimensional white noise,} \textit{The Annals of Statistics}, 47,
3382--3412.
\bibitem[{Lieberman(1994)}]{lieberman1994}
Lieberman, O. (1994), \enquote{A Laplace approximation to the moments of a
ratio of quadratic forms,} \textit{Biometrika}, 81, 681--690.
\bibitem[{Liu et~al.(2008)Liu, Lin, and Shao}]{liu2008asymptotic}
Liu, W.-D., Lin, Z. and Shao, Q.-M. (2008), \enquote{The asymptotic
distribution and Berry--Esseen bound of a new test for independence in high
dimension with an application to stochastic optimization,} \textit{The Annals
of Applied Probability}, 18, 2337--2366.
\bibitem[{Pesaran(2004)}]{pesaran2004}
Pesaran, M.~H. (2004), \enquote{General diagnostic tests for cross section
dependence in panels (IZA Discussion Paper No. 1240),} \textit{Institute for
the Study of Labor (IZA)}.
\bibitem[{Pesaran(2015)}]{pesaran2015}
--- (2015), \enquote{Testing weak cross-sectional dependence in large panels,}
\textit{Econometric reviews}, 34, 1089--1117.
\bibitem[{Pesaran et~al.(2008)Pesaran, Ullah, and Yamagata}]{pesaran2008}
Pesaran, M.~H., Ullah, A. and Yamagata, T. (2008), \enquote{A bias-adjusted LM
test of error cross-section independence,} \textit{The Econometrics Journal},
11, 105--127.
\bibitem[{Pesaran and Yamagata(2012)}]{pesaran2012testing}
Pesaran, M.~H. and Yamagata, T. (2012), \enquote{Testing CAPM with a large
number of assets,} in \textit{AFA 2013 San Diego Meetings Paper}.
\bibitem[{Rosenberg and Rudd(1982)}]{rosenberg1982factor}
Rosenberg, B. and Rudd, A. (1982), \enquote{Factor-related and specific returns
of common stocks: Serial correlation and market inefficiency,} \textit{The
Journal of Finance}, 37, 543--554.
\bibitem[{Rudelson and Vershynin(2013)}]{rudelson2013}
Rudelson, M. and Vershynin, R. (2013), \enquote{Hanson-wright inequality and
sub-gaussian concentration,} \textit{Electronic Communications in
Probability}, 18, 1--9.
\bibitem[{Saulis and Statulevicius(1991)}]{saulis1991limit}
Saulis, L. and Statulevicius, V. (1991), \textit{Limit theorems for large
deviations}, vol.~73, Springer Science \& Business Media.
\bibitem[{Schwartz and Whitcomb(1977)}]{schwartz1977evidence}
Schwartz, R.~A. and Whitcomb, D.~K. (1977), \enquote{Evidence on the presence
and causes of serial correlation in market model residuals,} \textit{Journal
of financial and quantitative analysis}, 12, 291--313.
\bibitem[{Statulevi{\v{c}}ius(1966)}]{statulevivcius1966}
Statulevi{\v{c}}ius, V. (1966), \enquote{On large deviations,}
\textit{Zeitschrift f{\"u}r Wahrscheinlichkeitstheorie und verwandte
Gebiete}, 6, 133--144.
\bibitem[{Vershynin(2018)}]{vershynin2018}
Vershynin, R. (2018), \textit{High-Dimensional Probability: An Introduction
with Applications in Data Science}, Cambridge University Press.
\bibitem[{Wei(2006)}]{wei2006}
Wei, W. (2006), \textit{Time Series Analysis: Univariate and Multivariate
Methods}, Pearson Addison Wesley.
\end{thebibliography}
\end{document}