syn-pdfQA/01.1_Input_Files_Non_PDF/research articles/2510.23061v1.tex

\documentclass[aps,twocolumn,pra,tightenlines,floatfix,superscriptaddress]{revtex4-2}

\usepackage[breaklinks=true]{hyperref}
\usepackage{graphicx}
\usepackage[english]{babel}
\usepackage{amsmath}
\usepackage{amssymb}
\usepackage{times}
\usepackage{lipsum}

\begin{document}

\title{Effects of  particle-hole fluctuations on the  superfluid transition in two-dimensional atomic Fermi gases}


 \author{Junru Wu}
 \affiliation{Hefei National Research Center for Physical Sciences at the Microscale and School of Physical Sciences, University 
   of Science and Technology of China, Hefei, Anhui 230026, China}
 \affiliation{Shanghai Research Center for Quantum Science and CAS Center for Excellence in Quantum Information and Quantum Physics, 
 University of Science and Technology of China, Shanghai 201315, China}
 \affiliation{Hefei National Laboratory, University of
  Science and Technology of China, Hefei 230088, China}
 \author{Zongpu Wang}
 \affiliation{Hefei National Research Center for Physical Sciences at the Microscale and School of Physical Sciences, University 
   of Science and Technology of China, Hefei, Anhui 230026, China}
 \affiliation{Department of Physics, University of Hong Kong, Pokfulam Road, Hong Kong}
% 王宗璞
 \author{Lin Sun}
 \affiliation{Hefei National Laboratory, University of
  Science and Technology of China, Hefei 230088, China}
 \affiliation{Shanghai Research Center for Quantum Science and CAS Center for Excellence in Quantum Information and Quantum Physics, 
 University of Science and Technology of China, Shanghai 201315, China}
 \author{Kaichao Zhang}
 \affiliation{Hefei National Research Center for Physical Sciences at the Microscale and School of Physical Sciences, University 
   of Science and Technology of China, Hefei, Anhui 230026, China}
 \affiliation{Shanghai Research Center for Quantum Science and CAS Center for Excellence in Quantum Information and Quantum Physics, 
 University of Science and Technology of China, Shanghai 201315, China}
 \affiliation{Hefei National Laboratory, University of
  Science and Technology of China, Hefei 230088, China}
 \author{Chuping Li}
 \affiliation{Hefei National Research Center for Physical Sciences at the Microscale and School of Physical Sciences, University 
   of Science and Technology of China, Hefei, Anhui 230026, China}
 \affiliation{Shanghai Research Center for Quantum Science and CAS Center for Excellence in Quantum Information and Quantum Physics, 
 University of Science and Technology of China, Shanghai 201315, China}
 \affiliation{Hefei National Laboratory, University of
  Science and Technology of China, Hefei 230088, China}
 \author{Yuxuan Wu}
 \affiliation{Hefei National Research Center for Physical Sciences at the Microscale and School of Physical Sciences, University 
   of Science and Technology of China, Hefei, Anhui 230026, China}
 \affiliation{Shanghai Research Center for Quantum Science and CAS Center for Excellence in Quantum Information and Quantum Physics, 
 University of Science and Technology of China, Shanghai 201315, China}
 \affiliation{Hefei National Laboratory, University of
  Science and Technology of China, Hefei 230088, China}
 \author{Pengyi Chen}
 \affiliation{Hefei National Research Center for Physical Sciences at the Microscale and School of Physical Sciences, University 
   of Science and Technology of China, Hefei, Anhui 230026, China}
 \affiliation{Shanghai Research Center for Quantum Science and CAS Center for Excellence in Quantum Information and Quantum Physics, 
 University of Science and Technology of China, Shanghai 201315, China}
 \affiliation{Hefei National Laboratory, University of
  Science and Technology of China, Hefei 230088, China}
 \author{Dingli Yuan}
 \affiliation{Hefei National Research Center for Physical Sciences at the Microscale and School of Physical Sciences, University 
   of Science and Technology of China, Hefei, Anhui 230026, China}
 \affiliation{Shanghai Research Center for Quantum Science and CAS Center for Excellence in Quantum Information and Quantum Physics, 
 University of Science and Technology of China, Shanghai 201315, China}
 \affiliation{Hefei National Laboratory, University of
  Science and Technology of China, Hefei 230088, China}
 \author{Qijin Chen}
 \email[Corresponding author: ]{qjc@ustc.edu.cn}
 \affiliation{Hefei National Research Center for Physical Sciences at the Microscale and School of Physical Sciences, University 
   of Science and Technology of China, Hefei, Anhui 230026, China}
 \affiliation{Shanghai Research Center for Quantum Science and CAS Center for Excellence in Quantum Information and Quantum Physics, 
 University of Science and Technology of China, Shanghai 201315, China}
 \affiliation{Hefei National Laboratory, University of
  Science and Technology of China, Hefei 230088, China}


\date{\today}

\begin{abstract}
  Proper treatment of the many-body interactions is of paramount importance in our understanding of strongly correlated systems. 
  Here we investigate the effects of particle-hole fluctuations on the
  Berezinskii-Kosterlitz-Thouless (BKT) transition in two-dimensional
  Fermi gases throughout the entire BCS-BEC crossover.  We include
  self-consistently in the self energy treatment the entire
  particle-hole $T$ matrix, which constitutes a renormalization of the
  bare interaction that appears in the particle-particle scattering
  $T$ matrix, leading to a screening of the pairing interaction and hence a dramatic reduction of the pairing gap and the transition temperature. The BKT
  transition temperature $T_\text{BKT}$ is determined by the critical
  phase space density, for which the pair density and pair mass are
  determined using a pairing fluctuation theory, which accommodates
  self-consistently the important self-energy feedback in the
  treatment of finite-momentum pairing fluctuations.  The screening
  strength varies continuously from its maximum in the BCS limit to
  essentially zero in BEC limit. In the unitary regime, it leads to
  an interaction-dependent shift of $T_\text{BKT}$ towards the BEC
  regime. This shift is crucial in an attempt to explain experimental
  data quantitatively, which often depends on the interaction
  strength.  Our findings are consistent with available experimental
  results in the unitary and BEC regimes and with quantum Monte Carlo
  simulations in the BCS and unitary regimes.
\end{abstract}

\maketitle


\section{Introduction}

Strongly correlated systems constitute the main challenge and are thus
at the heart and frontier of in condensed matter physics. Due to the
low dimensionality, fluctuations in two dimensions (2D) are usually
strong, positioning the system in the strongly correlated regime, for
which a proper treatment of interaction effects is of paramount
importance, in order to understand various experiments and physical
phenomena at a \emph{quantitative} level.

In the absence of long range order a la the Mermin-Wagner theorem
\cite{Mermin1966PRL}, phase transition in 2D is
in general of the Berezinskii \cite{berezinskii1972}, Kosterlitz and
Thouless (BKT) \cite{kosterlitz1973} type. Indeed, 
superfluid transitions in ultracold atomic Bose gases have been
experimentally found to exhibit a strong BKT nature in 2D
\cite{Hadzibabic2006N,Clade2009PRL,Tung2010PRL}.
%
BKT transitions have been of wide interest since the most important
high $T_c$ superconductors are quasi-2D layered materials
\cite{Kosterlitz2}, for which the low dimensionality is favorable for
the formation of the wide-spread pseudogap phenomena, which are
arguably attributable to strong fluctuations
\cite{Loktev2001PR,chen2024RMP}. It is also the model commonly used
for describing the superconducting transition in thin films
\cite{Hetel2007,xue2013SSC}. Recent exciting discoveries of (quasi-)2D
superconductors
\cite{Bovzovic2016N,Agterberg2017PRL,Hsu2017NC,Cao2018N} have added to
the interest in 2D phase transitions.  The BKT nature of the
superfluid transition in 2D atomic Fermi gases has also been
experimentally confirmed \cite{Ries2015PRL,Murthy2015PRL}. There are
also, albeit not many, theoretical studies of the BKT superfluid
transition in a 2D Fermi system for both weak and strong pairing
interactions, e.g., Ref.~\cite{Wang2020NJP,Shi2024}.

In this paper, we report our study on the important yet often
neglected effect of the particle-hole fluctuations on 2D Fermi gas
superfluidity, to which similar studies has not been reported in the
literature \cite{Shi2024}. Similar to its 3D counterpart, we find that
particle-hole fluctuations lead to a screening to the pairing
interaction, causing a shift of the superfluid transition curve toward
the BEC regime as a function of the pairing strength. This
substantially suppresses $T_c$ and the gap in the BCS and unitary
regimes, and has an important physical significance when comparing
between theory and experiment.
 
It is important to note that there has been a historical controversy
surrounding 2D fermionic superfluids, dating back to Kosterlitz and
Thouless \cite{kosterlitz1973}, which primarily concerns observable
signatures and the applicability of BKT physics. It is not
straightforward to apply the BKT theory based on the XY model to
fermionic superfluids. In the XY model, in which the superfluid
density $n_s/m$ provides the phase stiffness, it is the phase
fluctuations of the vortex type that dominate the destruction of the
quasi-long range superfluid order and drive the system into a
disordered normal state. Unlike the superfluids of true bosons,
however, the superfluid density is also suppressed by the
pair-breaking Bogoliubov quasiparticle excitations as the temperature
$T$ increases. In addition, both the density and the mass of the
bosons (i.e., fermion pair) now depend on the temperature and the
pairing interaction strength, except in the deep BEC regime.
%
Therefore, computing the superfluid density using a mean-field
approximation would yield a constant value $n_s/m=n/m$ at $T=0$ (or $n/4m$ from the boson point of view, with $n_B = n/2$ and $m_B = 2m$),
independent of the interaction strength. (Following standard notations, here $n$, $m$, $n_B$,  $m_B$ and $n_s$ denote total fermion number density, fermion mass, Cooper pair number density, Cooper pair mass, and superfluid number density, respectively.) At finite $T$, the pair-breaking effect due to quasiparticle excitations can be partly accounted for via solving the mean-field BCS gap equation, leading to a mean-field result of $(n_s/m)^\text{BCS}$. However, not being able to properly take care of the effective pair mass and pair number density  is thus expected to give
rise to an overestimate of the superfluid transition temperature
$T_\text{BKT}$. Indeed, the fact $T_\text{BKT}$, as measured experimentally in
2D atomic Fermi gases, varies with interaction strength, reflects that
both $n_\text{B}$ and $M_\text{B}$ vary with the interaction.

The pair number density at given temperature and interaction is normally governed by the pair dispersion. The latter, or equivalently pair mass, can, in principle, be extracted from the pair propagator.
%
However, there are not many calculations of the fermionic $T_\text{BKT}$ in the literature \cite{botelho2006vortex,Bauer2014PRL,Bighin2016PRB,Mulkerin2017PRA}, partly due to the inapplicability of the simple XY model directly to the Fermi system. The fact that  $(n_s/m)^\text{BCS}$ decreases with temperature below $T_\text{BKT}$ in the BCS regime manifests the fact that amplitude fluctuations play an important role. Therefore, the superfluid transition in a 2D Fermi gas  has a much richer physics than the simple XY model can capture.
%
To account  for the finite temperature and interaction effects, Wu et al \cite{Wu2015PRL} and Wang et al
Ref.~\cite{Wang2020NJP} determined $T_\text{BKT}$ using the critical
phase space density criterion,  based on a quantum Monte Carlo simulation \cite{Prokofev2001PRL} along with experimental support \cite{Murthy2015PRL}. A pairing fluctuation theory
\cite{chen1998PRL} that was developed to address the pseudogap
phenomena in 3D superconductors was used to determine the effective pair number
density $n_\text{B}$ and pair mass $M_\text{B}$, which reflect the
effects of both amplitude and phase fluctuations, governed by temperature and interaction strength. Nevertheless, this earlier work considered only the particle-particle channel of the $T$-matrix.


It has been known that particle-hole fluctuations may play an
important role in the 3D superfluid behavior, despite that they are often
neglected in theoretical treatments of superconductivity
\cite{SchriefferBook}.  Gor'kov and Melik-Barkhudarov (GMB) \cite{GMB} first
found that the lowest order particle-hole fluctuations may reduce both
$T_c$ and the zero temperature gap $\Delta_0$ by a factor of 0.44 in a
BCS superconductor.
%
The study of the GMB effect at the lowest order were extended
to atomic Fermi gases in the continuum \cite {Heiselberg2000PRL} or an
optical lattice \cite{Kim2009PRL}. The effect beyond the lowest order,
by including the full particle-hole $T$-matrix in a ladder
approximation has been studied without \cite{Yu2009PRA} and with
\cite{Chen2016SR} the self-energy feedback.


In 2D, there have been no similar studies in the literature,
however. It is the purpose of the present work to investigate the
effect of particle-hole fluctuations on the
atomic Fermi gas superfluidity in 2D. The low dimensionality
further enhances the strong pairing fluctuations, which are already
present when the interaction is strong. The very BKT nature of the
transition requires that there is already a sizable pairing amplitude,
i.e., (pseudo)gap, at $T_\text{BKT}$. Therefore, this necessitates the
self-consistent inclusion of the self-energy feedback in (both the
particle-particle and) the particle-hole $T$-matrices.
%
Here we
%investigate the particle-hole channel effect on the
%2D Fermi gas superfluidity by adding
incorporate the particle-hole channel
contributions to the pairing fluctuation theory
\cite{chen1998PRL,Chen1999PRB,Chen2016SR}, and thus go beyond
previous studies \cite{Wu2015PRL,Wang2020NJP}.  This pairing
fluctuation theory includes self-consistently the finite-momentum
pairing fluctuations in the single-fermion self energy, and has
successfully addressed multiple high $T_c$
\cite{chen1998PRL,Chen1999PRB,Chen2001PRB,ReviewLTP-Full} and atomic
Fermi gas experiments
\cite{chen2005PR,Kinast2005S,Chen2009PRL,FrontPhys}. In particular, it
features naturally a pseudogap in 3D unitary Fermi gases, which has been
unequivocally corroborated by a recent experiment \cite{li2024nature}.
%
Following the previous work in 3D \cite{Chen2016SR}, we show that the
particle-hole $T$-matrix serves as a renormalized pairing interaction,
which is to appear in the $T$-matrix in the usual particle-particle
channel.  We find that the particle-hole fluctuations lead to a
screening to the pairing interaction, and the inverse pairing
interaction is effectively shifted by the temperature-dependent,
angular-averaged (at two different levels) particle-hole
susceptibility $\langle\chi_\text{ph}\rangle$, which approaches a
negative constant, $-m/2\pi$, in the BCS limit and increases gradually
in the crossover regime toward zero in the BEC limit.  In result, the
particle-hole channel shifts the $T_\text{BKT}$ curve towards the BEC
regime as a function of pairing strength.  Furthermore, comparison
shows that the inclusion of particle-hole channel leads to a better
overall agreement between our calculated $T_\text{BKT}$ with the results from
experiment and quantum Monte Carlo simulations.


\section{Theoretical Formalism}

\subsection{Overview of the Pair Fluctuation Theory without the Particle-Hole Channel}

To be self-contained, we first recapitulate the pairing fluctuation
theory \cite{chen1998PRL,Chen1999PRB,Wu2015PRL} without including the
particle-hole channel. This serves as a basis, on top of which the
particle-hole channel effect is built. Note that here we will consider
only the formalism without the superfluid order parameter,
%$\Delta_\text{sc}$,
tailored for 2D in the absence of a true long
range order.

We consider a 2D Fermi gas with a short-range $s$-wave attractive
interaction $V_{\mathbf{k},\mathbf{k}^{\prime}} = U <0$, as described
by a generic grand canonical Hamiltonian \cite{Chen1999PRB}, which
includes pairing with a finite momentum $\mathbf{q}$.  The free
fermion dispersion is given by $\xi_{\mathbf{k}} =
\epsilon_{\mathbf{k}} - \mu \equiv \mathbf{k}^2/2m - \mu$, with
chemical potential $\mu$. The Fermi momentum is $\hbar k_\text{F} =
\sqrt{2\pi n}\hbar $, and Fermi energy $E_\text{F} \equiv
k_\text{B}T_\text{F} = \hbar^2 k_\text{F}^2/2m$. As usual, we shall
use the natural units, and set $\hbar = k_\text{B}=1$ and the volume
to unity \cite{chen1998PRL}.

The fermions acquire self energy via pair binding and unbinding. Our
approximated equations, derived via an equations of motion approach \cite{Kadanoff1961PR,ChenPhD}, can be
cast into a $T$ matrix formalism, with a mix of bare Green's function
$G_0(K)=(\mathrm{i}\omega_{l}-\xi_{\mathbf{k}})^{-1}$ and full Green's function $G(K)$ in the pair susceptibility
$\chi(Q) = \sum_K G(K)G_0(Q-K)$. Here we use the four-momentum
notation, $K \equiv(\mathrm{i} \omega_l, \mathbf{k})$ and $Q
\equiv(\mathrm{i} \Omega_n, \mathbf{q})$, $\sum_K \equiv T \sum_{l,
  \mathbf{k}}$ and $\sum_Q \equiv T \sum_{n, \mathbf{q}}$, with
$\omega_l$ and $\Omega_n$ being Matsubara frequencies for fermions and
bosons, respectively \cite{fetter}. 
%
The $T$-matrix $t(Q)$ is given by $t(Q)=U/[1+U \chi(Q)]$, with the self
energy \[ \Sigma(K)=\sum_{Q}t(Q)G_0(Q-K). \] 

The Thouless criterion (for superfluid transition in 3D) requires %that the condition for pairs to generate macroscopic occupation at zero momentum 
%
$t^{-1}(Q = 0)=U^{-1}+\chi(0)=0$. In order to accommodate the absence
of long range order in 2D, we generalize the Thouless criterion to
allow for a very small but finite pair chemical potential $
\mu_\text{p} $, $ U^{-1}+\chi(0) = t^{-1}(0)  \propto\mu_\text{p}$.
%, where $a_0$ is
%the coefficient of the linear frequency term in the inverse $T$ matrix
%expansion (see below).
In this way, $t(Q)$ is highly peaked
around $Q=0$.  Thus the main contribution to $\Sigma(K)$ comes from
the vicinity of zero momentum for pairs, leading to the approximation
of $\Sigma(K) \approx -\Delta^2 G_0(-K)$, where we define the
pseudogap as $\Delta^2 =-\sum_{Q}t(Q)$.
%Therefore, the total self energy is given by $\Sigma(K) = -\Delta^2 G_0(-K)$, 
%and the total energy gap is $\Delta^2=\Delta_\text{sc}^2+\Delta_\text{pg}^2$. 
%Here $\Delta_\text{pg}$ comes from the contribution of nonzero momentum pairs, 
%while $\Delta_\text{sc}$ comes from the contribution of zero momentum pairs. 
With this approximation, the full Green's function takes a simple BCS-like form and is given by
\begin{equation*}
    G(K)=\frac{u_{\mathbf{k}}^{2}}{\mathrm{i}\omega_{n}-E_{\mathbf{k}}}+\frac{v_{\mathbf{k}}^{2}}{\mathrm{i}\omega_{n}+E_{\mathbf{k}}}\,,
\end{equation*}
where $E_{\mathbf{k}}=\sqrt{\xi_{\mathbf{k}}^2+\Delta^2}$, 
$u_\mathbf{k}^2 = (1+\xi_\mathbf{k}/E_{\mathbf{k}})/2$, 
and $v_\mathbf{k}^2 = (1-\xi_\mathbf{k}/E_{\mathbf{k}})/2$.
%Note that in a 2D continuum, according to the Mermin-Wagner theorem \cite{Mermin1966PRL}, 
%all fermion pairs are inherently non-condensed at finite temperatures, 
%leading to $T_\text{c} = 0$ and $\Delta=\Delta_\text{pg}$ for $T \ge 0$. 
%
Then one obtains a generalized BCS-like gap equation,
\begin{equation}
  \label{eq:gap}
  a_0 \mu_\text{p} = \sum_{\mathbf{k}}\left[\frac{1-2 f(E_{\mathbf{k}})}{2 E_{\mathbf{k}}} - \frac{1}{2\epsilon_\mathbf{k} + \epsilon_\text{B}}\right]\,,
\end{equation}
where 
$a_0$ is the coefficient of the linear $\Omega$ term in the Taylor expansion of the inverse $T$-matrix. 
Here the gap equation has been regularized via $U^{-1} = - \sum_{\mathbf{k}}  1/(2 \epsilon_{\mathbf{k}}+\epsilon_{\mathrm{B}})$, 
where the two-body binding energy $\epsilon^{}_\text{B} = 1/ma^2_\text{2D}$ with the 2D scattering length $a^{}_\text{2D}$ \cite{Levinsen2015}. 
%
The fermion number constraint $n = 2\sum_K G(K)$ yields 
\begin{equation}
  \label{eq:eqn}
  n = \sum_{\mathbf{k}}\left[1-\frac{\xi_{\mathbf{k}}}{E_{\mathbf{k}}}+2 \frac{\xi_{\mathbf{k}}}{E_{\mathbf{k}}} f(E_{\mathbf{k}})\right]\,,
\end{equation}
where $f(x)$ is the Fermi distribution function. 

To extract the pair dispersion, one can Taylor expand $t^{-1}$ near $Q=0$ and analytically continue to the real frequency axis, with
 $(\mathrm{i}\Omega_l \rightarrow \Omega+\mathrm{i}0^+)$, so that
\begin{equation} t^{-1}(\Omega,\mathbf{q}) \approx a_1\Omega^2+a_0(\Omega-\Omega_\mathbf{q}^0+\mu_\text{p}),\end{equation}
where $\Omega_{\mathbf{q}}^0 = \mathbf{q}^2 / 2M_\text{B}$.
Consequently, the definition of the pseudogap
yields
%
\begin{equation}
   \label{eq:pg}
   a_0 \Delta^2 = \sum_{\mathbf{q}}\left[1+4\frac{a_1}{a_0}(\Omega_{\mathbf{q}}^0-\mu_\text{p})\right]^{-1/2} b(\Omega_{\mathbf{q}}),
\end{equation}
where $b(x)$ is the Bose distribution function 
and $\Omega_{\mathbf{q}}=\left[\sqrt{a_0^2+4a_0a_1(\Omega_{\mathbf{q}}^0-\mu_\text{p})}-a_0\right]/2a_1$ represents the pair dispersion. 
The coefficients $a_0$, $a_1$, $M_\text{B}$ are determined via the expansion process. 
Except in the weak coupling BCS regime, the $a_1$ term serves as a small quantitative correction and can often be neglected, so that $\Omega_{\mathbf{q}}\approx \Omega_{\mathbf{q}}^0-\mu_\text{p}$. 

\subsection{BKT Criterion}

At $T=0$ where $\mu_\text{p}=0$, Eqs.~(\ref{eq:gap}) and (\ref{eq:eqn}) give $\mu = E_\text{F}-\epsilon^{}_\text{B}/2$ 
and $\Delta = \sqrt{2E_\text{F}\epsilon_\text{B}}$ \cite{Randeria1990PRB}. 
At finite temperature, the summation in Eq.~(\ref{eq:pg}) analytically leads to 
$\frac{a_0}{a_1}-\sqrt{\frac{a_0}{a_1}(\frac{a_0}{a_1}-4\mu_\text{p})} = 2T \ln(1-e^{-\mathcal{D}_{\text{B}}})$, 
which reduces to $\mu_\text{p} = T \ln(1-e^{-\mathcal{D}_{\text{B}}})$ in the BEC regime. 
Here $n^{}_\text{B} = a_0 \Delta^2$ represents the pair density, 
and $\mathcal{D}_{\text{B}}$ %= 2 \pi n_{\text{B}} / M_\text{B} T $
is the bosonic phase space density. 
Thus $\mu_\text{p}$ is primarily determined by $\mathcal{D}_{\text{B}}$ at given temperature, 
indicating that the BKT transition may occur when $\mathcal{D}_{\text{B}}$ is 
large enough so that $\mu_\text{p}$ becomes sufficiently close to zero \cite{Prokofev2002PRA,Murthy2015PRL,Wu2015PRL,Wang2020NJP}. 
Moreover, in the BEC regime, a large $\mathcal{D}_{\text{B}}$ provides a sharp peak distribution of the bosonic number density 
at zero momentum via $b(-\mu_\text{p})=e^{\mathcal{D}_{\text{B}}}-1$, signaling a quasi-condensation \cite{Wu2015PRL}. Alternatively, when the phase space density becomes large enough, the wave functions of neighboring pairs start to overlap with each other, and hence help to establish quasi-long-range phase coherence.
%Therefore, a critical value of $\mathcal{D}_{\text{B}}$ is required to determine the occurrence of the BKT transition. 

When approached from the high-temperature region
\cite{kosterlitz2013,Hadzibabic2006N}, the bosonic BKT transition
occurs when $\mathcal{D}_{\text{B}}(T)$ reaches a critical value
$\mathcal{D}_{\text{B}}(T_\text{BKT})$ \cite{Prokofev2002PRA}.
Estimates of $\mathcal{D}_{\text{B}}(T_\text{BKT})$ for fermionic
superfluids are provided in Refs.~\cite{Ries2015PRL,Murthy2015PRL},
with values ranging from approximately 4.9 to 6.45.  In comparison,
the analogous systems in atomic Bose gases typically hover around 8
\cite{Tung2010PRL}.  The lowest value,
$\mathcal{D}_{\text{B}}(T_\text{BKT}) = 4.9$, which is the closest to
the factor of 4 in the usual BKT relation, was reported to offer the
best fit for the experimental results on Fermi gases
\cite{Murthy2015PRL}.  Thus we choose
$\mathcal{D}_{\text{B}}(T_\text{BKT}) = 4.9$, and the BKT transition
temperature $T_\text{BKT}$ is determined by
\begin{equation}
  \label{eq:BKT}
  \frac{n^{}_{\text{B}}}{M_{\text{B}}}=\frac{4.9}{2 \pi} T_{\text{BKT}}\,.
\end{equation}

\subsection{Contributions of the Particle-Hole Channel}

Following Ref.~\cite{Chen2016SR}, we introduce the contribution of the particle-hole channel, 
by renormalizing the pairing strength in $t(Q)$, 
which leads to a new full $T$-matrix $t_\text{2}(Q)$ that includes both particle-particle and particle-hole contributions. 
The expression for $t_\text{2}(Q)$ is given by
\begin{equation*}
  t_\text{2}(Q) = \frac{1}{t^{-1}_\text{ph}(K + K' - Q) + \chi(Q)}\,,
\end{equation*}
which self-consistently includes the self-energy feedback, with $K,K'$ being the external fermion momentum.
Here the particle-hole channel $T$-matrix $t^{-1}_\text{ph}(Q') = U^{-1} + \chi_{\text{ph}}(Q')$ describes the particle-hole scattering, 
and the particle-hole susceptibility $\chi_{\text{ph}}(Q') = \sum_{K} G(K) G_0(K-Q')$, where the particle-hole momentum $Q'=K + K' - Q$.
Moreover, assuming that  the fermions near the Fermi surface  dominate  the particle-hole channel contributions, 
we replace the particle-hole susceptibility with an average $\langle{\chi_\text{ph}\rangle}$ on or near the Fermi surface, 
where the frequency part of the particle-hole susceptibility is set to 0 with $\mathrm{i}\Omega'_n = 0$ \cite{gor1961}, 
leading to a zero imaginary part of $\chi_{\text{ph}}(Q')$ and therefore a purely real $\langle\chi_{\text{ph}}\rangle$ \cite{Chen2016SR}. 

We have proposed two methods for averaging $\chi_{\text{ph}}(Q')$, referred to as level 1 and level 2, respectively. 
The level 1 average involves an on-shell and elastic scattering on the Fermi surface, with $|\mathbf{k}|=|\mathbf{k}'|=k_\mu = \sqrt{2m\max(\mu,0)}$, 
where the momentum part of $\chi_{\text{ph}}(Q')$ is determined by $|{\mathbf{q}'}|=\left|\mathbf{k}+\mathbf{k}^{\prime}\right|= k_\mu\sqrt{2(1+\cos \theta)}$. 
Here  $\chi_{\text{ph}}(Q')$ is averaged over scattering angles $\theta$ (between $\mathbf{k}$ and $\mathbf{k}^{\prime}$). 
This averaging process, focusing solely on the Fermi surface, 
is commonly used in the literature on the studies of induced interactions. 
In contrast, the level 2 average considers that the states within the energy range $\xi_\mathbf{k} \in [-\min(\Delta, \mu),\Delta]$ of a typical s-wave superconductor are most significantly affected by pairing (for $\mu > 0$). 
Hence, for the level 2 average, while keeping the on-shell and elastic scattering, 
the average is performed over a range of $|\mathbf{k}|$ such that the quasi-particle energy $E_{\mathbf{k}} \in\left[\min (E_{\mathbf{k}}), \min (\sqrt{E^2_{\mathbf{k}}+\Delta^2})\right]$, 
where $\min (E_{\mathbf{k}})=\Delta$ if $\mu>0$, or $\min (E_{\mathbf{k}})=\sqrt{\mu^2+\Delta^2}$ if $\mu<0$.

Then with this frequency and momentum independent $\langle{\chi_\text{ph}\rangle}$, 
the new full $T$-matrix $t_\text{eff}(Q)$ reads
\begin{equation}
  \label{teff}
  t_\text{eff}(Q) = \frac{1}{U^{-1} + \langle\chi_{\text{ph}}\rangle + \chi(Q)}\,.
\end{equation}
The gap equation with the particle-hole channel effect is modified into 
\begin{equation}
  \label{eq:gapph}
  a^{}_0 \mu_\text{p} =  \langle\chi_{\text{ph}}\rangle + \sum_{\mathbf{k}}\left[\frac{1-2 f(E_{\mathbf{k}})}{2 E_{\mathbf{k}}} - \frac{1}{2\epsilon^{}_\mathbf{k} + \epsilon^{}_\text{B}}\right]\,,
\end{equation}
while the other equations remain unchanged.

Equations (\ref{eq:eqn}), (\ref{eq:pg}), and (\ref{eq:gapph}) form a closed set of self-consistent equations, 
which can be used to solve for $(\mu, \Delta, \mu_\text{p})$, along with $T_\text{BKT}$ via the BKT criterion given by Eq.~(\ref{eq:BKT}). 

Throughout the BCS-BEC crossover,  $\langle\chi_{\text{ph}}\rangle$ remains negative as a function of the coupling strength.
From Eq.~(\ref{eq:gapph}), the particle-hole channel constitutes a renormalization of the pairing interaction, with a net effect given by replacing $1/U$ with $ 1/U_\text{eff}\equiv 1/U + \langle\chi_{\text{ph}}\rangle$. Diagrammatically, this amounts to replacing the bare interaction $U$ with the full particle-hole $T$-matrix $t_\text{ph}$ in the particle-particle scattering diagrams \cite{Chen2016SR}. Given the negative sign of $\langle\chi_{\text{ph}}\rangle$, one can see immediately that $|U_\text{eff}|< |U|$. Thus  $U_\text{eff}$ represents a weaker, screened pairing interaction.


\section{Numerical Results and Discussions}

\subsection{Behaviors of the particle-hole susceptibility}

\begin{figure}
% \centerline{\includegraphics[clip,width=3.4in]{Fig1_Tc&T0_chi.vs.T.pdf}} 
\centerline{\includegraphics[clip,width=3.4in]{Fig1.pdf}} 
\caption{
  Angular average of  the on-shell particle-hole susceptibility $\langle\chi_{\text{ph}}(0,|\mathbf{k}+\mathbf{k}'|)\rangle/2m$ with $k=k'$ 
  as a function of momentum $k/k_\text{F}$ at unitarity $\ln(k_\text{F} a_\text{2D}) = 0$ for $T=0$ and $T=T_\text{BKT}$, 
  where $T_\text{BKT}/T_\text{F} = 0.079 $ and the corresponding $\Delta$, $\mu$ and $\mu_\text{p}$ are calculated without the particle-hole channel effect.}
\label{fig:chiph}
\end{figure}

First, we present in Fig.~\ref{fig:chiph} the (level 1) angular
average of the particle-hole susceptibility at zero frequency as a
function of momentum $k$, under the on-shell condition
$|\mathbf{k}|=|\mathbf{k}'|=k$.  Here we focus on the unitary case at
$T=T_\text{BKT}$ (black solid line) and zero-temperature (red dashed line), and
$\langle\chi_{\text{ph}}(0,|\mathbf{k}+\mathbf{k}'|)\rangle$ is
calculated using the corresponding solution of $(\Delta,\mu, 
\mu_\text{p})$ at $T_\text{BKT}/T_\text{F} = 0.079$ and
$\ln(k_\text{F} a_\text{2D}) = 0$,  solved in the
absence of particle-hole fluctuations. The slight difference between these two curves reveals 
%$\langle\chi_{\text{ph}}(0,|\mathbf{k}+\mathbf{k}'|)\rangle$ exhibits
a weak temperature dependence. Importantly, both curves show a strong momentum dependency, with the amplitude decreasing
monotonically as the momentum rises.
%Given this monotonicity, the fact that the zero $T$ curve lies
%below its counterpart at $T_\text{BKT}$ indicates $\mu(T_\text{BKT}) >
%\mu(T=0)$.
Note that the momentum dependencies in 2D appear distinct from those
in 3D \cite{Chen2016SR}, where
$\langle\chi_{\text{ph}}(0,|\mathbf{k}+\mathbf{k}'|)\rangle$ exhibits
a nonmonotonic momentum dependence at low $T$ for the unitary case.

\begin{figure}
% \centerline{\includegraphics[clip,width=3.4in]{Fig2_Tc_chi.vs.lnkFa.pdf}} 
\centerline{\includegraphics[clip,width=3.4in]{Fig2.pdf}} 
\caption{ $-\langle\chi_{\text{ph}}\rangle / 2m$ at $T=T_\text{BKT}$,
  averaged at both level 1 (red dashed) and level 2 (black solid line),
  as a function of $\ln(k_\text{F}a_\text{2D})$ throughout BCS-BEC
  crossover.  The magenta dotted line indicates the BCS limit value,
  $1/4\pi$, given by $-\langle\chi_{\text{ph}}\rangle = {m}/{2\pi}$.}
\label{fig:chiphTc}
\end{figure}

Shown in Fig.~\ref{fig:chiphTc} is $-\langle\chi_{\text{ph}}\rangle$
at $T=T_\text{BKT}$ as a function of $\ln(k_\text{F}a_\text{2D})$
throughout the BCS-BEC crossover, averaged at both level 1 (red dashed) and
level 2 (black solid curve), along with its BCS limit $1/4\pi\approx
0.0796$, given by $\langle\chi_{\text{ph}}\rangle=-{m}/{2\pi}$.  The
value of $-\langle\chi_{\text{ph}}\rangle / 2m$ decreases
monotonically with increasing interaction strength and exhibits an
exponential behavior in the deep BEC regime for $\ln(k_\text{F}
a_\text{2D}) < -2$. As the gap diminishes, the average at both levels
converges in the BCS limit.  However, in the crossover regime, the two
levels differ significantly, with a smaller absolute value for the
level 2 average. This can be understood from Fig.~\ref{fig:chiph};
compared with level 1 averaging, level 2 averaging involves a range of
$k$'s so that the larger $k$ contributions dominates due to its larger
phase space area, leading to a  smaller absolute value of the
average. In other words, level 1 averaging on the Fermi surface only leads to a significant over-estimate of the particle-hole contributions in the unitary regime.

The asymptotic behavior of $\langle\chi_{\text{ph}}\rangle$ can be readily
solved analytically in both the BCS and the BEC limits.  In the weak
coupling limit, where $\ln(k_\text{F} a_\text{2D}) \rightarrow
\infty$, $\Delta \rightarrow 0$, and $T \leq T_\text{BKT} \rightarrow
0$, the two levels of average converge to $\chi_{\text{ph}}(Q')
\approx \sum_{K} G_0(K) G_0(K-Q')$. Under the on-shell condition $\xi_{\mathbf{k}}=\xi_{\mathbf{k}-\mathbf{q}'}$, the integrand becomes the derivative of the Fermi function and one readily obtains
\begin{equation}
  \chi_{\text{ph}}(0,\mathbf{q}') = \sum_\mathbf{k}\frac{f(\xi_{\mathbf{k}})-f(\xi_{\mathbf{k}-\mathbf{q}'})}{\xi_{\mathbf{k}}-\xi_{\mathbf{k}-\mathbf{q}'}} \approx -\frac{m}{2\pi} = \langle\chi_{\text{ph}} \rangle\,.
  \label{eq:chiph}
\end{equation} 
%Here  the integral is carried out since the Fermi function becomes a step function for $T \leq T_\text{BKT}$ in the deep BCS limit. 
We emphasize that this result is independent of the density, due to the constant density of states in 2D. This is to be contrasted with its 3D counterpart, which is proportional to the on-shell momentum $k$. In the strong coupling limit, where $\ln(k_\text{F} a_\text{2D}) \rightarrow -\infty$, 
we have $|\mu| \gg \Delta \gg \epsilon_\text{F}$, which indicates that $E_\mathbf{k} \approx \xi_\mathbf{k} \approx |\mu|$. Thus particle-hole fluctuations are exponentially suppressed, so that $\langle\chi_{\text{ph}}\rangle$ approaches zero. 



\subsection{Effect of particle-hole channel on the BKT transition}

\begin{figure}
% \centerline{\includegraphics[clip,width=3.4in]{Fig3_T0_gap&chi.vs.lnkFa.pdf}} 
\centerline{\includegraphics[clip,width=3.4in]{Fig3.pdf}} 
\caption{
  (a) Effect of the particle-hole channel contributions on $\Delta$, along with (b) the corresponding $-\langle\chi_{\text{ph}}\rangle / 2m$,  at $T=0$, 
  as the function of $\ln(k_\text{F}a_\text{2D})$ throughout the BCS-BEC crossover. Shown are results without (black solid curve) and with particle-hole channel contributions averaged at level 1 (red dashed) and level 2 (blue dot-dashed curve).
  }
\label{fig:chiphT0}
\end{figure}

In this section, we explore the impact of the particle-hole channel on
the pairing phenomena and the BKT transition of a 2D Fermi gas.  We
begin by examining the behavior of the pairing gap $\Delta$ at zero
temperature with and without the particle-hole channel contributions.
Plotted in Fig.~\ref{fig:chiphT0}(a) is $\Delta$ as a function of
interaction. For comparison, we plot the results both without (black
solid line) and with the particle-hole channel effect, with the
particle-hole susceptibility $\langle{\chi_\text{ph}\rangle}$ averaged
at level 1 (red dashed curve) and level 2 (blue dot-dashed curve),
respectively.  The corresponding $\langle{\chi_\text{ph}\rangle}$ is
shown in Fig.~\ref{fig:chiphT0}(b).  For all cases, $\Delta$ decreases
monotonically from BEC to BCS, as expected.  A substantial reduction
of $\Delta$ by the particle-hole fluctuations occurs in the crossover
and BCS regimes.  In the BEC regime, $\langle{\chi_\text{ph}\rangle}$
approaches zero, rendering the particle-hole channel effect
negligible.  Consistent with Fig.~\ref{fig:chiphT0}(b), level 2
averaging showed a weaker particle-hole effect, and thus a smaller
reduction of $\Delta(T=0)$.  Note that in Fig.~\ref{fig:chiphT0}(b),
there exists a kink around $\mu=0$ in $\langle{\chi_\text{ph}\rangle}$
for both levels of averaging, mainly because the Fermi surface shrinks
to zero abruptly with a finite slope as a function of increasing
pairing strength.  The Fermi function in the integrand of
$\langle{\chi_\text{ph}\rangle}$ becomes a step function at zero $T$,
resulting in a delta function in the derivative of the integrand of
$\langle{\chi_\text{ph}\rangle}$ and hence a discontinuity when
crossing $\mu=0$ (See Appendix \ref{sec:AppA} for details). This slope discontinuity becomes more prominent in the level 2 average, since the range for $k < k_\mu$  in the average shrinks to zero abruptly when $\mu= 0$ \footnote{It should be noted that $\mu=0$ occurs at $\ln (k_Fa_\text{2D}) \approx -0.25 $ and -0.175 for level 1 and 2 averaging, respectively, as they are calculated with the self-consistent solutions of $T_\text{BKT}$.}.

\begin{figure*}
% \centerline{\includegraphics[clip,height=2.7in]{Fig4_Tc&gap&mu&mup&np&mB.vs.lnkFa.pdf}} 
\centerline{\includegraphics[clip,height=2.7in]{Fig4.pdf}}
\caption{ Effect of the particle-hole channel contributions on (a)
  $T_\text{BKT}$, (b) $\Delta$, (c) $\mu$, (d) $\mu_\text{p}$, (e)
  $n_\text{B}$ and (f) $M_\text{B}$ versus
  $\ln(k_\text{F}a_\text{2D})$, with the average
  $\langle\chi_{\text{ph}}\rangle / 2m$ calculated at level 1 (red dashed)
  and level 2 (blue dot-dashed curves), respectively.  They should be compared
  with the results without the particle-hole effect (black solid curves).  }
\label{fig:Tc}
\end{figure*}

For the effect of the particle-hole channel on the behavior of the BKT
transition temperature $T_\text{BKT}$, we first analytically estimate
the ratio between the two BKT transition temperatures in the BCS limit with and
without the particle-hole channel at the same coupling strength, where
the latter is denoted as $T_\text{BKT}^\text{BCS}$.  Here $\Delta
\rightarrow 0$, so that the following summation can be performed
analytically,
\begin{equation*}
    \sum_{\mathbf{k}}\left[\frac{1-2 f(\xi_{\mathbf{k}})}{2 \xi_{\mathbf{k}}}-\frac{1}{2 \epsilon_{\mathbf{k}}+\epsilon_\text{B}}\right] = \frac{m}{4\pi} \ln \left( \frac{2e^{2\gamma} \epsilon_\text{B} E_\text{F}}{\pi^2 T^2}\right),
\end{equation*}
where $\gamma \approx 0.5772157$ is Euler’s constant. (A detailed
derivation can be found in Appendix \ref{sec:AppB}.)  Substituting the
above relation for the corresponding term in Eqs.~(\ref{eq:gap}) and
(\ref{eq:gapph}), we obtain
\begin{equation*}
  \frac{T_\text{BKT}}{T_\text{BKT}^\text{BCS}} = e^{ 2 \pi \langle\chi_{\text{ph}}\rangle / m} = e^{-1} \approx 0.37 \,,
\end{equation*}
where we have taken into account that $t^{-1}(0)$ and
$t_\text{2}^{-1}(0)$ are sufficiently small so that
$|t^{-1}(0)|,|t_\text{2}^{-1}(0)| \ll
|\langle\chi_{\text{ph}}\rangle|$ in the weak coupling limit.

Next, we present in Fig.~\ref{fig:Tc} the effect of the particle-hole
channel on the evolution of (a) $T_\text{BKT}$, along with (b)
$\Delta$, (c) $\mu$, (d) $\mu_\text{p}$, (e) $n_\text{B}$, and (f)
$M_\text{B}$ at $T_\text{BKT}$ as a function of
$\ln(k_\text{F}a_\text{2D})$. The results without particle-hole fluctuations are presented as the black solid curves. Starting from the weak coupling BCS limit (black curve), as the interaction strength increases, $T_\text{BKT}$ increases, and then reaches a maximum %near $\ln (k_\text{F}a_\text{2D})=0$
in the intermediate regime, where 
$\mu_\text{p}$ reaches a minimum simultaneously.  As the interaction strength increases further past the maximum, $T_\text{BKT}$
decreases and reaches a minimum near unitarity  $\ln (k_\text{F}a_\text{2D})=0$, where $\mu=0$.  Beyond
this point, the system enters  the BEC regime, where all
fermions pair up with $2n_\text{B}/n \approx 1$.  The behavior of
$T_\text{BKT}$ is then influenced by the shrinking pair size, as
indicated by a decrease in the pair mass $M_\text{B}$ toward $2m$,
reaching a BEC asymptote  $T_\text{BKT}/T_\text{F}\approx 0.109$.  The gap $\Delta$ consistently increases with pairing
strength. In comparison, the particle-hole channel contributions cause a \emph{non-uniform shift} of all curves toward the BEC regime on the right. This shift is the largest in the BCS limit, and vanishes in the BEC regime, as indicated in Fig.~\ref{fig:chiphTc}. It exhibits a strong dependence on $\ln (k_\text{F}a_\text{2D})$ in the crossover regime. Now the maximum of $T_\text{BKT}$ occurs  closer to  unitarity, and the locate for $\mu=0$ is shifted into the BEC regime.  All $T_\text{BKT}$ curves with and without the
particle-hole effect converge to the same BEC asymptote.
From Fig.~\ref{fig:Tc}(c), we have a tiny $|\mu_\text{p}| \le 1.2
\times 10^{-3} E_\text{F}$ throughout the BCS-BEC crossover for all
cases. This ensures that $t(Q)$ remains highly peaked at $Q=0$ and thus validates our pseudogap approximation for the self-energy,
$\Sigma(K) \approx -\Delta^2 G_0(-K)$.

\begin{figure*}
% \centerline{\includegraphics[clip,height=2.7in]{Fig5b_BT_gap&mup.vs.T.pdf}} 
\centerline{\includegraphics[clip,width=6.6in]{Fig5.pdf}}
\caption{ Effect of the particle-hole channel contributions on
  behaviors of $\Delta$ (top row) and $\mu_\text{p}$ (bottom row), as a
  function of $T/T_\text{BKT}$, with
  $\ln(k_\text{F}a_\text{2D})=-1,0,2$ from left to right for the BCS,
  unitary, and BEC regimes, respectively.  The black solid curve
  represents calculations without the particle-hole channel,
  while the red dashed and green dot-dashed curves include the
  particle-hole channel effect, using $\langle\chi_{\text{ph}}\rangle
  / 2m$ under level 1 and level 2 averaging, respectively.  }
\label{fig:BT}
\end{figure*}

Now we investigate in Fig.~\ref{fig:BT} the evolution of $\Delta$ 
(top row) and $\mu_\text{p}$ (bottom row) as a function of reduced
temperature $T/T_\text{BKT}$ in the BCS, unitary, and BEC regimes from
left to right without (black) and with (red and green) the
particle-hole channel effect.  The gap $\Delta$  remains nearly constant except in the BCS case (f),  where a significant decrease can be discerned near $T_\text{BKT}$.  The value of $\Delta$
with the particle-hole channel effect is reduced from the counterpart
without the particle-hole channel effect by a factor of $1/e$ in the BCS limit. This reduction factor becomes smaller in the unitary and BEC regimes,  consistent with the zero $T$ and $T_\text{BKT}$
gap behaviors as shown in Fig.~\ref{fig:chiphT0}(a) and Fig.~\ref{fig:Tc}(b).  At the same time, for all cases, $\mu_\text{p}$
decreases continuously to zero as $T$ decreases, consistent with
$\mu_\text{p}=0$ at $T=0$ for a true long-range-order ground state. Note that below  $0.6T_\text{BKT}$, $\mu_\text{p}$ comes essentially zero.

\subsection{Comparison with different results}

\begin{figure}
\centerline{\includegraphics[width=3.3in]{Fig6.pdf}}
  \caption{ Comparison of theoretical calculations for
    $T_\text{BKT}/T_\text{F}$ with experiment and QMC results as a
    function of $\ln(k_\text{F}a_\text{2D})$ throughout the BCS-BEC
    crossover.  (a) Overlay of theoretical $T_\text{BKT}$ without
    (black solid line) and with the particle-hole channel calculated
    with the level 1 average (light cyan solid line), on top of the contour
    plot of experimentally measured quasi-condensate fractions
    \cite{Ries2015PRL}. Adapted from Ref.~\cite{Wang2020NJP}.  (b)
    Overlay of our $T_\text{BKT}$ on top of a collection of
    experimental data \cite{Ries2015PRL,Lennart2021} and various
    theoretical results
    \cite{Petrov2003PRA,Bighin2016PRB,Bauer2014PRL,Mulkerin2017PRA,He2022PRL}.
    The black solid and dashed lines represent $T_\text{BKT}$
    calculated using our pairing fluctuation theory with and without
    the particle-hole channel contributions, respectively. Here the
    particle-hole channel effect was calculated with level 1 averaging
    of $\langle \chi_\text{ph}^{}\rangle$.  Adapted from
    Ref.~\cite{He2022PRL}.  }
    \label{fig:comp}
\end{figure}

Finally, in Fig.~\ref{fig:comp}, we compare our theoretical  $T_\text{BKT}$   with available experimental data \cite{Ries2015PRL} and QMC results on the BKT transition in 2D Fermi gases, as a function of $\ln(k_\text{F}a_\text{2D})$ throughout the BCS-BEC crossover. 
Fig.~\ref{fig:comp}(a) presents the 
measured quasi-condensate fractions $N_q/N$ \cite{Ries2015PRL}, overlaid on top of which are the theory curves calculated using our pairing fluctuation theory without (black solid) and with (light cyan solid curve) the particle-hole  contributions (level 1). The black solid line was previously presented in Ref.~\cite{Wang2020NJP}. The  experimental data are not dense enough to enable a successful detection of the minimum in  $T_\text{BKT}$ near  $\ln(k_\text{F}a_\text{2D}) = 0$, nevertheless, it does seem to suggest there is a minimum in the contour of the quasi-condensate fraction. Thus, our theory is consistent with the data, including the presence of the minimum.
In Fig.~\ref{fig:comp}(b), we compare our theoretical result of $T_\text{BKT}$ with a collection of other theories \cite{Petrov2003PRA,Bighin2016PRB,Bauer2014PRL,Mulkerin2017PRA}, 
and results obtained from Quantum Monte Carlo (QMC) using a 2D lattice model \cite{He2022PRL}. 
Also plotted are the experimental results from Refs.~\cite{Ries2015PRL} and \cite{Lennart2021}. 
%
In the BEC regime, our results both with (black solid) and without (black dashed line) are in quantitative agreement with both experiments. Here the particle-hole susceptibility is averaged at level 1. On the BCS side, the curves of Mulkerin et al \cite{Mulkerin2017PRA} and the BCS mean-field treatment are close to our results without particle-hole, suggesting that particle-hole contributions are not included in Mulkerin et al's calculations \cite{Mulkerin2017PRA}. Our result with particle-hole contributions (black solid) are in good agreement with the QMC result in the infinite $L$ limit, except that QMC does not show a minimum in the unitary regime \cite{He2022PRL}. The results of Bauer et al \cite{Bauer2014PRL} and Petrov et al \cite{Petrov2003PRA} are in good agreement with QMC only in the very weak BCS regime, $\ln (k_\text{F}a_\text{2D}) > 2$.
We also notice there is a large difference between the finite ($L=45$) and infinite lattice QMC results. It is unusual that the result of Bighin et al \cite{Bighin2016PRB} is far above the BCS mean-field treatment in the BCS limit. It should be noted that the experimental data in the BCS regime are significantly above all theory curves, except the mean-field result. This suggests that some other factors, e.g., nonequilibrium, finite size effect, needs to be considered. Therefore, our result should not be compared with the experimental data in the BCS regime.
%
In short, our results with the particle-hole contributions are
consistent with experiment in the unitary and BEC regimes and in good
agreement with QMC results in the unitary and BCS regimes. This calls
for more data from future experiment.

Note that since the particle-hole susceptibility in Eq.~(\ref{eq:chiph}) is density independent in 2D, the particle-hole contributions in the weak coupling limit would have already been automatically included, if the 2D scattering length $a_\text{2D}$ were measured directly through experiment, e.g., by measuring the two-body binding energy in the dilute limit. Instead,  $a_\text{2D}$ is usually calculated from the 3D scattering length as in a deeply confined pancake-shaped trap \cite{petrov2001PRA}, thus the effect of particle-hole fluctuations should be seriously taken into account when comparing experiment and theory. In Fig.~\ref{fig:comp}, the same definition of  $a_\text{2D}$ was used in both QMC and our present work. This makes it possible to have a good agreement in the BCS regime between QMC and our result with particle-hole fluctuations included. Conversely, one can experimentally determine the particle-hole susceptibility in the BCS limit by measuring the 2D scattering length and comparing with that calculated from $a_\text{3D}$, and obtain
%
$$\langle \chi_\text{ph}\rangle = \frac{m}{2\pi}\ln \frac{a^\text{exp}_\text{2D}}{a^{}_\text{2D}}\,,$$
%
where $a^\text{exp}_\text{2D}$ denotes the experimentally measured 2D scattering length. Nevertheless, away from the weak coupling limit, a nontrivial density dependence should emerge as a nonzero pairing gap develops with increasing pairing strength at a finite density. 


\section{Conclusions}

In summary, we have studied the impact of the particle-hole channel on
BKT physics in Fermi gases within the context of the BCS-BEC
crossover.  We introduce the particle-hole channel effect by
incorporating an average particle-hole susceptibility, which
self-consistently includes the self-energy feedback, leading to an
effective renormalization of the pairing strength.  The dynamic
structure of the angular-averaged particle-hole susceptibility
exhibits strong dependencies on momentum and temperature, showing
distinct momentum dependencies at low temperatures compared to its 3D
counterpart.  Furthermore, we perform averaging of the particle-hole
susceptibility at two different levels, revealing important physical
consequences in the crossover and BCS regimes.  The particle-hole
channel provides a screening of the pairing interaction and shifts the
BKT transition temperature and the pairing gap curves towards the BEC regime as a function of $\ln (k^{}_Fa^{}_\text{2D})$.
Additionally, a comparison
shows that the BKT transition temperature calculated with the
particle-hole channel provides a better fit for the experimental data
and QMC results.  Future experiments with more elaborate measurements
are called for to resolve various discrepancies.

Finally, it should be mentioned that, in addition to the simple
averaged particle-hole susceptibility, there are a series of higher
order corrections, including a higher order $T$-matrix
\cite{Chen2016SR}, which may contribute significant modifications to
the present results. In addition, the BKT criterion
\cite{Wu2015PRL,Wang2020NJP} is merely based on numerical simulations
\cite{Prokofev2001PRL} to provide a value of the phase space
density. Despite the experimental support \cite{Murthy2015PRL}, an
analytical derivation of such a criterion would be highly desirable.


\section{Acknowledgments}
This work was supported by the Innovation Program 
for Quantum Science and Technology (Grant No. 2021ZD0301904). 

\appendix

\section{Slope discontinuity in $ \chi_{\text{ph}}$ across $\mu=0$ at zero $T$}
\label{sec:AppA}

The expression for the particle-hole susceptibility $\chi_{\text{ph}}(Q')$ is given by 
\begin{eqnarray*}
  &&\chi_{\text{ph}}(Q') =\\
  &&\sum_{\mathbf{k}}\left[\frac{f(E_{\mathbf{k}})-f(\xi_{\mathbf{k}-{\mathbf{q}'}})}{E_{\mathbf{k}}-\xi_{\mathbf{k}-{\mathbf{q}'}}-i \Omega'_n} u_{\mathbf{k}}^2
  -\frac{1-f(E_{\mathbf{k}})-f(\xi_{\mathbf{k}-{\mathbf{q}'}})}{E_{\mathbf{k}}+\xi_{\mathbf{k}-{\mathbf{q}'}}+i \Omega'_n} v_{\mathbf{k}}^2\right].
\end{eqnarray*}
%
Upon analytical continuation, $\mathrm{i}\Omega'_n \rightarrow \Omega' + \mathrm{i}0^+$, 
we separate the retarded $\chi^R_{\text{ph}}(\Omega',{\mathbf{q}'})$ into real and imaginary parts, 
$\chi^R_{\text{ph}}(\Omega',{\mathbf{q}'})=\chi^\prime_{\text{ph}}(\Omega',{\mathbf{q}'})+\mathrm{i}\chi^{\prime\prime}_{\text{ph}}(\Omega',{\mathbf{q}'})$.
Furthermore, we set $\mathrm{i}\Omega'_n = 0$, which leads to $\chi^{\prime\prime}_{\text{ph}}(0,{\mathbf{q}'})=0$, 
and the real part is expressed as
\begin{eqnarray*}
  &&\chi^\prime_{\text{ph}}(0,{\mathbf{q}'}) = \\
  &&\sum_{\mathbf{k}} \left[\frac{f(E_{\mathbf{k}})-f(\xi_{\mathbf{k}-{\mathbf{q}'}})}{ E_{\mathbf{k}}-\xi_{\mathbf{k}-{\mathbf{q}'}}}u^2_{\mathbf{k}} - 
    \frac{1 -f( E_{\mathbf{k}})-f(\xi_{\mathbf{k}-{\mathbf{q}'}})}{ E_{\mathbf{k}}+\xi_{\mathbf{k}-{\mathbf{q}'}}}v^2_{\mathbf{k}}\right].
\end{eqnarray*}
%
At $T=0$, we have $f(x) = 1-\Theta(x)$. 
Thus $\chi^\prime_{\text{ph}}(0,{\mathbf{q}'})$ is given by 
\begin{equation*}
    \chi^\prime_{\text{ph}}(0,{\mathbf{q}'}) = 
    \sum_{\mathbf{k}} \left[\frac{\Theta(\xi_{\mathbf{k}-{\mathbf{q}'}})-1}{ E_{\mathbf{k}}-\xi_{\mathbf{k}-{\mathbf{q}'}}}u^2_{\mathbf{k}} - 
    \frac{\Theta(\xi_{\mathbf{k}-{\mathbf{q}'}})}{ E_{\mathbf{k}}+\xi_{\mathbf{k}-{\mathbf{q}'}}}v^2_{\mathbf{k}}\right].
\end{equation*}
%
Then the derivative of $\chi^\prime_{\text{ph}}(0,{\mathbf{q}'})$ with respect to $\mu$ is given by 
\begin{eqnarray*}
  &&\frac{\partial}{\partial \mu} \chi^\prime_{\text{ph}}(0,{\mathbf{q}'}) = 
  -\sum_{\mathbf{k}} \left[\frac{1+ \delta(\xi_{\mathbf{k}-{\mathbf{q}'}})}{ E_{\mathbf{k}}+\xi_{\mathbf{k}-{\mathbf{q}'}}}v^2_{\mathbf{k}} -  \frac{\delta(\xi_{\mathbf{k}-{\mathbf{q}'}})}{ E_{\mathbf{k}}-\xi_{\mathbf{k}-{\mathbf{q}'}}}u^2_{\mathbf{k}}\right] \\
%    + \frac{\delta(\xi_{\mathbf{k}-{\mathbf{q}'}})}{ E_{\mathbf{k}}+\xi_{\mathbf{k}-{\mathbf{q}'}}}v^2_{\mathbf{k}}\right] \\
&& = 
\left\{
\begin{aligned}
	& -\sum_{\mathbf{k}} \frac{v^2_{\mathbf{k}}}{ E_{\mathbf{k}}+\xi_{\mathbf{k}-{\mathbf{q}'}}} \,, & \mu < 0\,, \\
	& -\sum_{\mathbf{k}} \frac{v^2_{\mathbf{k}}}{ E_{\mathbf{k}}+\xi_{\mathbf{k}-{\mathbf{q}'}}} + \sum_{\phi \in A} \frac{\xi_{\mathbf{k}'}}{ E_{\mathbf{k}'}^2}\,,
	& \mu\geq 0\,,
\end{aligned}
\right.
\end{eqnarray*}
where $A = \left\{\phi: p^2(\cos(\phi)^2-1)/2+\mu \right\}$ with $\mathbf{k}' = p\cos(\phi) \pm \sqrt{p^2(\cos(\phi)^2-1)/2+\mu}$ given by $\xi_{\mathbf{k}-{\mathbf{q}'}} = 0$. 
Thus, an additional term in the derivative of $\chi^\prime_{\text{ph}}(0,{\mathbf{q}'})$ 
emerges when the system changes from $\mu<0$ to $\mu \ge 0$, leading to a slope discontinuity of $\chi_{\text{ph}}$.

\section{Analytical result of 2D  gap equation in the BCS limit}
\label{sec:AppB}
To obtain an analytical result of the BCS gap equation for the 2D case, 
we define $\epsilon = \epsilon_\mathbf{k} / E_\text{F}$, 
$t = T / T_\text{F}$, and $a = \epsilon_\text{B}/E_\text{F}$. 
Since $\mu \approx E_F$, we have
\begin{align*}
    & \sum_{\mathbf{k}}\left[\frac{1-2 f(\xi_{\mathbf{k}})}{2 \xi_{\mathbf{k}}}-\frac{1}{2 \epsilon_{\mathbf{k}}+\epsilon_\text{B}}\right]
 \\ = & \frac{m}{2\pi} \int_{0}^\infty d\epsilon \left[ \frac{\tanh \frac{\epsilon-1}{2t}}{2(\epsilon-1)} - \frac{1}{2\epsilon + a} \right]
 \\ = & \frac{m}{4\pi} \int_{-\frac{1}{2t}}^\infty dx \left[ \frac{\tanh x}{x} - \frac{1}{x + \frac{1}{2t} + \frac{a}{4t}} \right]
% \\ = & \frac{m}{4\pi} \left[ \ln x \tanh x \big|_0^\infty + \ln x \tanh x \Big|_0^{\frac{1}{2t}} \right.
% + 2\ln\frac{4e^\gamma}{\pi} 
% \\ & \left. - \left.\ln \left( x+\frac{1}{2t}+\frac{a}{4t} \right) \right|_{-\frac{1}{2t}}^\infty \right]
  =  \frac{m}{4\pi} \ln \left( \frac{2e^{2\gamma}a}{\pi^2 t^2}\right) .
\end{align*}
Here we have utilized the fact that ${1}/{2t} \gg 1$ to justify the approximation $\tanh({1}/{2t}) = 1$.

\bibliography{2DBKTph.bib}

\end{document}