Bell Nonlocality Test on Two-Mode Squeezed Output Generated in Double-Cavity Optomechanical
Extracted source text · arXiv 2604.12050
% ****** Start of file apssamp.tex ******
%
% This file is part of the APS files in the REVTeX 4.2 distribution.
% Version 4.2a of REVTeX, December 2014
%
% Copyright (c) 2014 The American Physical Society.
%
% See the REVTeX 4 README file for restrictions and more information.
%
% TeX'ing this file requires that you have AMS-LaTeX 2.0 installed
% as well as the rest of the prerequisites for REVTeX 4.2
%
% See the REVTeX 4 README file
% It also requires running BibTeX. The commands are as follows:
%
% 1) latex apssamp.tex
% 2) bibtex apssamp
% 3) latex apssamp.tex
% 4) latex apssamp.tex
%
\documentclass[%
reprint, %onecolumn,
%superscriptaddress,
%groupedaddress,
%unsortedaddress,
%runinaddress,
%frontmatterverbose,
%preprint,
%preprintnumbers,
%nofootinbib,
%nobibnotes,
%bibnotes,
amsmath,amssymb,
aps,
%pra,
%prb,
%rmp,
%prstab,
%prstper,
%floatfix,
]{revtex4-2}
\usepackage{graphicx}% Include figure files
\usepackage{dcolumn}% Align table columns on decimal point
\usepackage{bm}% bold math
%\usepackage{hyperref}% add hypertext capabilities
%\usepackage[mathlines]{lineno}% Enable numbering of text and display math
%\linenumbers\relax % Commence numbering lines
%\usepackage[showframe,%Uncomment any one of the following lines to test
%%scale=0.7, marginratio={1:1, 2:3}, ignoreall,% default settings
%%text={7in,10in},centering,
%%margin=1.5in,
%%total={6.5in,8.75in}, top=1.2in, left=0.9in, includefoot,
%%height=10in,a5paper,hmargin={3cm,0.8in},
%]{geometry}
\usepackage{times}
\usepackage{hyperref}
\usepackage{comment}
\begin{document}
\preprint{APS/123-QED}
\title{ %Generation of Two-Mode Squeezed Outputs using Double-Cavity Optomechanical Systems and Bell Measurement on it
Bell Nonlocality Test on Two-Mode Squeezed Output Generated in Double-Cavity Optomechanical Systems
}
%\thanks{A footnote to the article title}%
\author{Souvik Agasti}
\email{souvik.agasti@uhasselt.be}
\affiliation{
IMOMEC division, IMEC, Wetenschapspark 1, B-3590 Diepenbeek, Belgium
}%
\affiliation{
Institute for Materials Research (IMO), Hasselt University, Wetenschapspark 1, B-3590 Diepenbeek, Belgium
}%
%\date{\today}% It is always \today, today,
% but any date may be explicitly specified
\begin{abstract}
We explore here how to generate a two-mode squeezed output using reservoir engineering in a double-cavity optomechanical system coupled to a common mechanical resonator. %Such hybrid systems can experimentally be accessible in an electro/optomechanical interface that can be used for high-fidelity state transfer, quantum communication, and meteorological purposes. By testing the violation of CHSH Bell's inequality, we show that maximal squeezing does not necessarily lead to nonlocality; it may arise in states with lower squeezing. Exploring the CHSH inequality for different finesse, we show that the region of nonlocality can even be wider, even though the region of squeezing reduces. In all the cases we investigated, the phenomenon highlights the crucial role of mixedness of the states.
% We investigate the generation of two-mode squeezed output via reservoir engineering in a double-cavity optomechanical system coupled to a common mechanical resonator.
Such hybrid platforms are experimentally accessible in electro-optomechanical interfaces and are relevant for high-fidelity state transfer, quantum communication, and metrological applications. By examining violations of the CHSH Bell inequality, we demonstrate that maximal squeezing does not necessarily imply nonlocality; instead, nonlocal correlations can emerge in states with lower squeezing. Furthermore, by analyzing the CHSH inequality across different cavity finesse values, we find that the parameter region supporting nonlocality can broaden even as the squeezing region shrinks. Across all regimes considered, %our results highlight the crucial role played by mixedness of the states.
our results emphasize the crucial influence of the mixedness of the state in determining the relationship between squeezing and nonlocality.
\end{abstract}
%\keywords{Suggested keywords}%Use showkeys class option if keyword
%display desired
\maketitle
%\tableofcontents
\section{Introduction}\label{Introduction}
Continuous-variable (CV) entangled Gaussian states have attracted considerable attention due to their applications in precision measurement, quantum communication, and high-fidelity state transfer \cite{quantum_teleportation, Clerk_Fidelity, vitali_TMS_generator}. The most popular example of such a bipartite Gaussian CV state is the two-mode squeezed vacuum (TMSV) state. The entanglement shared between its constituent modes is of fundamental and practical interest, with applications ranging from quantum memories \cite{Quantum_memory} to gravitational-wave metrology \cite{study4roadmap, my_BAE}. entanglement motivates the test of a hidden variable in such systems that puts a limitation on local realism, which has been determined through Bell’s inequality \cite{mypaper_TMSV_filter, mypaper_TMSV_filter_STD}. Quantum nonlocality for spatially separated systems was formalized by Clauser, Horne, Shimony, and Holt (CHSH) \cite{bell_CHSH, bell_CH}, and later reformulated in phase space by Banaszek and W\'odkiewicz using Wigner quasiprobability distributions \cite{Banaszek_TMSV_bell, Banaszek_TMSV_bell_PRL}. These approaches provide criteria for nonlocality in bipartite Gaussian CV states and highlight the important role of state mixedness \cite{agasti_nonlocally_CV}, consistent with the fact that \textit{all nonlocally realizable bipartite Gaussian states must be entangled}.
Squeezed light is most commonly generated via parametric down-conversion in Kerr nonlinear media or through four-wave mixing in atomic vapors, with substantial experimental progress in both directions \cite{squeeze_light_sourse, TMS_generation, Squeezing_generation}. However, Kerr-based schemes typically involve non-Gaussian dynamics, leading to phase-space distortions that limit the achievable squeezing \cite{Agasti_Kerr_PhyScr, Agasti_Kerr_JOSAB}. Even at strong nonlinearities, phase-space twisting and non-Gaussian features become increasingly pronounced, and optimal squeezing is attained only over a restricted interaction-time window. Nevertheless, nonlinear processes tend to amplify input noise, thereby degrading the quality of squeezing.
% \textit{wrap around}
As an alternative, optomechanical systems can exhibit an effective Kerr-type nonlinearity \cite{Equivalence_optomechanics_Kerr}, motivating their use as platforms for generating nonclassical states. In particular, double-cavity optomechanical systems with reservoir engineering have emerged as promising candidates for generating two-mode squeezing (TMS) \cite{Clerk_Fidelity, Reservoir_Engineering_two_mode, vitali_TMS_generator}. Such architectures can be realized, for instance, by coupling optical cavities to microwave superconducting circuits \cite{Teufel_om_experiment}, or by incorporating high-Q membranes within optical cavity setups \cite{double_cav_alternate}. Systems of this kind—featuring two cavities coupled to a common mechanical resonator—have also been employed in back-action-evading measurements relevant to gravitational-wave detection \cite{my_BAE} and in precision force sensing \cite{force_sensing_taylor}.
In this work, we employ reservoir engineering in a double-cavity optomechanical system to generate strongly entangled output states that serve as sources of TMS. While entanglement in such systems has been studied previously \cite{Reservoir_Engineering_two_mode, vitali_TMS_generator}, our approach combines sideband cooling in one cavity—effectively enabling state exchange between optical and mechanical modes—with amplification in the other cavity, which drives the generation of entanglement.
Given that optomechanically generated TMS states are Gaussian and highly entangled, it is natural to investigate their Bell nonlocality. Beyond foundational interest, such tests are relevant for certifying security in quantum cryptographic protocols \cite{cryptographic_Bell}. A few ideas for testing the violation of CHSH inequalities in an electro- and optomechanical setting have been proposed before \cite{Proposal_Bell_test_Electromechanics, Proposal_Bell_test_optomechanics}; however, they are based on different configurations than the one discussed here. Here, we instead focus on Bell tests applied to specific spectral components of the output Bogoliubov modes, selected via filtering. This enables us to directly probe the relationship between entanglement (or TMS) and nonlocality, with particular emphasis on the role of mixedness of the state.
The article is organized as follows: in the next section, we show how we can generate TMS Bogoliubov modes in the output of a double-cavity optomechanical system using a blue and a red detuned pump. Furthermore, we exploit quantum correlation between the specific spectral components of their outputs to study the discrepancy between TMS and the violation of the CHSH inequality. By varying the finesses of the cavities, we show that the parameter regime supporting nonlocality can broaden even as the region exhibiting TMS shrinks. The article ends up investigating the role of the quality-factor of the mechanical oscillator and the temperature of its reservoir on the TMS and nonlocality between outputs, emphasizing throughout the critical role of state mixedness in governing their interplay.
%--------------------------------------
\begin{figure}
\includegraphics[width= 0.7 \linewidth]{Block_diagram}
\caption{ Block diagram of (a) an optomechanical system where two cavities coupled to a common mechanical oscillator and the filtering of modes at output, (b) homo/heterodyne measurement of filtered output and the bipartite quantum correlation (LO refers to the local oscillator), and (c) illustration of the ranges of frequencies at which the red and blue detuned cavities and their filter frequencies operate on.
}\label{Block_diagram}
\end{figure}
\begin{figure}
\includegraphics[width= 1 \linewidth]{SQZFilter_nonloc}
\caption{ (a) TMS and (b) maximal value of Bell function vs normalized output filter frequency for different bandwidths of filters $\epsilon = \tau \times \omega_m$, at a fixed central frequency of the filter at red detuned port $\Omega_- = \omega_m$. The cavity detuning frequencies are $-\Delta_+ = \Delta_- = \omega_{m} = 2\pi \times 10 $ MHz. Other parameters are, for example cavity linedidth $\kappa_+ = \kappa_- = \kappa =0.02 \, \omega_m$, effective optomechanical couplings
$ G_- = 0.15 \,\omega_m , G_+ = 0.2 \,G_- $, the quality factor $Q \equiv \omega_m/\gamma_m = 1.5 \times 10^5,$ and the temperature is around $25 mK$ which has corresponding thermal population $
n_m \approxeq 500$.
This parameter set remains consistent with the parameters used in experiments conducted in
\cite{Teufel_om_experiment}, which has also been followed in \cite{vitali_TMS_generator}. The only exception considered here is a lower mechanical quality factor than \cite{Teufel_om_experiment}, so that
A mirror can have a highly reflective coating.
}\label{SQZFilter_nonloc}
\end{figure}
\section{Model: Double-Cavity Optomechanical System}
\subsection{The Hamiltonian}
The Hamiltonian of a double-cavity optomechanical system coupled to a common mechanical oscillator is
\begin{equation} \label{system_Hamiltonian}
H_{OM} = \omega_m b^\dagger b + \sum_{k=\pm} \omega_{k} a_k^\dagger a_k + g_k , a_k^\dagger a_k (b^\dagger + b),
\end{equation}
where $a_k$ ($a_k^\dagger$) denote the annihilation (creation) operators of the cavity modes with $k \in [+,-]$, and $b$ ($b^\dagger$) corresponds to the mechanical mode. The resonance frequencies of the cavity and mechanical modes are $\omega_k$ and $\omega_m$, respectively, with dissipation rates $\kappa_k$ and $\gamma_m$. The single-photon optomechanical coupling strengths are given by $g_k = (\omega_k/L_k)\sqrt{\hbar/(m\omega_m)}$, where $L_k$ is the effective cavity length and $m$ is the effective mass of the mechanical resonator. In the regime where the mechanical frequency is much smaller than the cavity free spectral range ($\omega_m \ll \mathrm{FSR} \sim c/L_k$), each cavity can be treated as single-mode, as the external drives predominantly address a single resonance. The cavities are driven by lasers with frequencies $\omega_k^L$, yielding the driving Hamiltonian
\begin{equation}
H_{\mathrm{drive}} = i \sum_{k=\pm} \left(a_k^\dagger E_k e^{-i\omega_k^L t} - \mathrm{h.c.}\right),
\end{equation}
where the drive amplitudes are $|E_k| = \sqrt{2P_k \kappa_k/(\hbar \omega_k^L)}$ and depend on the input powers $P_k$.
Entanglement between the two cavity modes can be generated by driving one cavity (labeled $+$) with a blue-detuned pump and the other ($-$) with a red-detuned pump. The corresponding detunings are defined as $\Delta_k = \omega_k - \omega_k^L$. %To achieve two-mode squeezing, the detunings are chosen to be opposite and approximately resonant with the mechanical frequency, i.e., $\Delta_- = -\Delta_+ \approx \omega_m$.
% ($\omega^L_+ \approx \omega_+ + \omega_{m}$ and $\omega^L_- \approx \omega_- - \omega_{m}$)
The basic block diagram is presented in Fig. \ref{Block_diagram}, where two different drives of the cavities fix the cavity detunings $\Delta_k = \omega_{k} - \omega^L_k $. %\approx -k \omega_m. %The requirement of the
To achieve TMS, these opposite detunings are required to be opposite: $\Delta_- = - \Delta_+ \approx \omega_m$.
Transforming to a rotating frame at the pump frequencies and neglecting nonresonant terms, the linearized Hamiltonian in the sideband-resolved regime ($\omega_m \gg \kappa_k$) takes the form
\begin{align}\label{linearized_Hamiltonian}
H_{OM} = &\Delta_+ a_+^\dagger a_+ + \Delta_- a_-^\dagger a_- + \omega_m b^\dagger b \\
&+ G_+ (a_+^\dagger b^\dagger + a_+b) + G_- (a_- b^\dagger + a_-^\dagger b) \nonumber
\end{align}
where the effective optomechanical coupling rates are $G_k = g_k \alpha_k$, with $\alpha_k = E_k/(\kappa_k - i\Delta_k)$ (taken to be real without loss of generality). The stability of the system is analyzed in Appendix~\ref{Stability_System} using the Routh–Hurwitz criterion.
The blue-detuned drive ($+$) induces optomechanical parametric amplification and generates entanglement between the cavity and mechanical modes, while the red-detuned drive ($-$) enables state transfer between them. Such a configuration can be realized experimentally, for example, by coupling optical cavities to microwave superconducting circuits \cite{Teufel_om_experiment}, or by employing high-$Q$ mechanical resonators embedded in optical cavity setups \cite{double_cav_alternate}.
\subsection{Modes of Output}
The dynamics of the cavity fields are obtained from the quantum Langevin equations (see Appendix~\ref{Langevin_EOM}). To generate hybrid Bogoliubov modes, the cavity detunings must be appropriately matched. In particular, frequency locking requires $\Delta_- = -\Delta_+ = \omega_m$ \cite{vitali_TMS_generator}, which ensures that the relevant output modes are resonant, by locking the frequencies.
Within the input–output formalism for open quantum systems, the output field operators are related to the intracavity and input noise operators via
\begin{equation}
a_k^{\mathrm{out}} = \sqrt{\kappa_k}, a_k - a_k^{\mathrm{in}},
\end{equation}
where $a_k^{\mathrm{in}}$ and $a_k^{\mathrm{out}}$ denote the input and output fields of cavity $k$. In the high-cooperativity regime, $C_- = 4G_-^2/(\kappa_- \gamma_m) \gg 1$, the output modes evaluated at frequencies $\Delta_\pm$ take the form
\begin{subequations} \label{output_bogoliubov}
\begin{align}
a_+^{\mathrm{out}} &= \eta_+ b_{\mathrm{in}}^\dagger + \sinh r\, {a_-^{\mathrm{in}}}^\dagger + \cosh r\, a_+^{\mathrm{in}}, \\
a_-^{\mathrm{out}} &= \eta_- b_{\mathrm{in}} - \sinh r\, {a_+^{\mathrm{in}}}^\dagger - \cosh r\, a_-^{\mathrm{in}}
\end{align}
\end{subequations}
where
$\eta_+ = \frac{ -2i \sqrt{\gamma_m \kappa_+ } G_+ \kappa_- }{ G_-^2 \kappa_+ - G_+^2 \kappa_- }$ % \frac{2i \sqrt{\gamma_m \kappa_+ } G_+ \kappa_- }{ G_+^2 \kappa_- - \kappa_+ \left( \gamma_m \kappa_- + G_-^2 \right)}
,
$\eta_- = \frac{ -2i \sqrt{\gamma_m \kappa_-} G_- \kappa_+}{ G_-^2 \kappa_+ - G_+^2 \kappa_- }$ %\frac{ -2i \sqrt{\gamma_m \kappa_-} G_- \kappa_+}{ \kappa_+ \left( \gamma_- \kappa_-+ G_-^2\right)- G_+^2 \kappa_- }
and
\begin{align}\label{squeezing_factor}
\sinh r &= \frac{2 G_- G_+ \sqrt{\kappa_- \kappa_+}}{\kappa_+ G_-^2 - \kappa_- G_+^2}, \
\cosh r &= \frac{G_-^2 \kappa_+ + G_+^2 \kappa_-}{G_-^2 \kappa_+ - G_+^2 \kappa_-}.
\end{align}
%\frac{2 G_- G_+ \sqrt{ \kappa_- \kappa_+ } }{ \kappa_+ \left(\gamma_m \kappa _- + G_-^2\right)-G_+^2 \kappa_-}
% - \frac{- \gamma_m \kappa_- \kappa_+ + G_-^2 \kappa_+ + G_+^2 \kappa_-}{-\gamma_m \kappa_- \kappa_+ - G_-^2 \kappa_+ + G_+^2 \kappa_-}
where $r$ is the squeezing factor, highlighting its dependence solely on the effective optomechanical couplings. The last two terms of each equation in Eq. \eqref{output_bogoliubov} represent TMSV Bogoliubov modes, and the first element is the thermal noise contributed by the reservoir of the mechanical oscillator.
% $\eta_+ = (2i\sqrt{\kappa_+\gamma_m}/S) ( G_- \Gamma - G_+ B) , \eta_- = (-2 i\sqrt{\kappa_-\gamma_m}/S)( G_+ \Gamma + G_- A ) $ and $\cosh r = (2\kappa_+ B\gamma_{m}/S -1 ) = ( 2\kappa_- A \gamma_{m}/S - 1) , \, \sinh r = (\Gamma \gamma_{m} 2 \sqrt{\kappa_+\kappa_-}/S)$
\begin{comment}
\begin{widetext}
\begin{subequations}
\begin{align}
a_+^{out} &= (\sqrt{2\kappa_+}/S) ( + i G_- \Gamma - i G_+ B) \sqrt{2\gamma_m} b_{in}^\dagger + (\Gamma \gamma_{m} 2\sqrt{\kappa_+ \kappa_-}/S) {a_-^{in}}^\dagger + (2\kappa_+ B\gamma_{m}/S -1 ) a_+^{in} \\
a_-^{out} &= (\sqrt{2\kappa_-}/S)( - i G_+ \Gamma- i G_- A ) \sqrt{2\gamma_m} b_{in} - (\Gamma \gamma_{m} 2 \sqrt{\kappa_+\kappa_-}/S) {a_+^{in}}^\dagger + ( 2\kappa_- A \gamma_{m}/S - 1) a_-^{in}
\end{align}
\end{subequations}
\end{widetext}
$\eta_+ = (2i\sqrt{\kappa_+\gamma_m}/S) ( G_- \Gamma - G_+ B) , \eta_- = (-2 i\sqrt{\kappa_-\gamma_m}/S)( G_+ \Gamma + G_- A ) $ and $\cosh r = (2\kappa_+ B\gamma_{m}/S -1 ) = ( 2\kappa_- A \gamma_{m}/S - 1) , \, \sinh r = (\Gamma \gamma_{m} 2 \sqrt{\kappa_+\kappa_-}/S)$
where
$\Gamma = G_+G_-, \, A = ( \gamma_{m} \kappa_+ - |G_+|^2 ) , \, B = ( \gamma_{m} \kappa_- + |G_-|^2 ) $ and
$ S=( AB + |\Gamma|^2 ) $, which becomes $A = - |G_+|^2 $,
$B = |G_-|^2 $,
$ S= \kappa_+\kappa_-\gamma_m^2+ \gamma_{m} \kappa_1|G_-|^2 - |G_+|^2\gamma_{m} \kappa_2 $ for in
the limit of large cooperativity $ C_k = 4G_k^2/ \kappa_k\gamma_{m}>> 1 )$. To be hybrid-Bogoliubov modes with a squeezing factor $r$, it gives
\begin{equation}
\tanh r = \frac{(\Gamma \gamma_{m} 2\sqrt{\kappa_1 \kappa_2})}{|2\kappa_1 |G_-|^2 \gamma_{m} -S |} = \frac{(\Gamma \gamma_{m} 2 \sqrt{\kappa_1\kappa_2})}{| 2\kappa_2 |G_+|^2 \gamma_{m} + S|} \approx \frac{2G_+G_-}{G_+^2 + G_-^2}
\end{equation}
\begin{align}
S &= ( AB + |\Gamma|^2 )= |G_-|^2 \kappa_1 \gamma_{m} + |G_+|^2 \kappa_2 \gamma_{m} \\
& ( \gamma_{m} \kappa_1 - |G_+|^2 ) ( \gamma_{m} \kappa_2 + |G_-|^2 ) + |G_+G_-|^2 \\
&= \gamma_{m} \kappa_1 \gamma_{m} \kappa_2 - |G_+|^2\gamma_{m} \kappa_2 + \gamma_{m} \kappa_1|G_-|^2 \\
& \approx |G_-|^2 \kappa_1 \gamma_{m} - |G_+|^2 \kappa_2 \gamma_{m}
\end{align}
\begin{subequations} \label{Bogoliubov_mode}
\begin{align}
\beta_1 &= a_1 \cosh r + a_2^\dagger \sinh r \\
\beta_2 &= a_2 \cosh r + a_1^\dagger \sinh r,
\end{align}
\end{subequations}
\end{comment}
We focus on the steady-state regime and consider filtered output modes centered around the cavity resonances. Independent output modes can be defined by temporal filtering,
\begin{equation}
f_k^{\mathrm{out}}(t) = \int_{-\infty}^{t} h_k(t - t'), a_k^{\mathrm{out}}(t'), \mathrm{d}t',
\end{equation}
where $h_k(t)$ are causal filter functions satisfying the normalization condition $\int |h_k(t)|^2 \mathrm{d}t = 1$. We adopt exponentially decaying filters of the form
\begin{equation} \label{filterfunction}
h_k(t) = \sqrt{\frac{2}{\tau}}, \Theta(t), e^{-(1/\tau + i \Omega_k)t},
\end{equation}
where $\Theta(t)$ is the Heaviside step function, $1/\tau$ defines the bandwidth (taken identical for both modes), and $\Omega_k$ is the central frequency measured relative to the corresponding drive.
Since the output modes are expected to be entangled around $\pm\Delta$, we fix the central frequencies around $\Omega_k \approx \Delta_k$, while testing squeezing and nonlocality of the output signals. In the following section, we primarily test how resilient the correlation remains with filter parameters, and furthermore with the parameters of the cavities.
%------------------------------------------------
\begin{figure}
\includegraphics[width= 1 \linewidth]{SQZ_nonLocal}
\caption{ (a-I) $\log[S_q]$ representing TMS and (b-I) the maximal value of Bell function ($B_{max}$) as a function of $G_-$ and the ratio $G_+/G_-$. (a-II) squeezing factor ($r$) vs $G_-$ and (a-III) $r$ vs $G_+/G_-$, and (b-II) purity of the state ($\mu$) vs $G_-$ and (b-III) $\mu$ vs $G_+/G_-$. The black-dashed line in (a-I) represents SQL ($S_q = 1$) and the black-solid line in (b-I) stands for the nonlocality limit $B_{max} = 2$. All other parameters remain the same with Fig. \ref{SQZFilter_nonloc}.
}\label{SQZnonlocal}
\end{figure}
\begin{figure}
\includegraphics[width= 1 \linewidth]{nonlocal_SQZ_border_k}
\caption{ (a) SQL ($S_q = 1$) and (b) nonlocality limit $(B_{max} = 2)$ as a function of $G_-$ and the ratio $G_+/G_-$, for the variation of (I) $\kappa_+$ and (II) $\kappa_-$. As shown in Fig. \ref{SQZnonlocal}, the regions at the right side of the boundaries exhibit TMS ($S_q \leq 1$) and nonlocality ($B_{max} \geq 2$). Insets in (a) $r$ vs $G_+/G_-$ and (b) $\mu$ vs $G_+/G_-$ for $G_- = 0.15 \, \omega_{m}$. All other parameters remain the same with Fig. \ref{SQZFilter_nonloc}.
}\label{nonlocal_SQZ_border_k}
\end{figure}
\begin{figure}[t!]
\includegraphics[width= 1 \linewidth]{nonlocal_border_gamma}
\caption{ (a) SQL ($S_q = 1$) and (b) nonlocality limit $(B_{max} = 2)$ as a function of $G_-$ and the ratio $G_+/G_-$, for the variation of (I) linewidth of mechanical oscillator and (II) the temperature of its reservoir. As shown in Fig. \ref{SQZnonlocal}, the regions at the right side of the boundaries exhibit TMS ($S_q \leq 1$) and nonlocality ($B_{max} \geq 2$). Inset in (b-I) $\mu$ vs $G_-$ for $G_+/G_- = 0.2 $. All other parameters remain the same with Fig. \ref{SQZFilter_nonloc}.
}\label{nonlocal_border_gamma}
\end{figure}
\section{Results: Charecterization of Two mode Squeezing} \label{Results}
\subsection{Impact of Filter frequency and bandwidth}
The entanglement between the two filtered output modes is maximized when the filters select narrowband components centered at the red- and blue-detuned sidebands. This corresponds to choosing the filter frequencies to match the cavity detunings, i.e., $\Omega_- = \Delta_- = -\Delta_+ = -\Omega_+ = \omega_m$. In Fig.~\ref{SQZFilter_nonloc}, we illustrate how the two-mode squeezing (TMS) of the filtered outputs depends on the filter central frequencies and linewidths.
As a quantitative measure of entanglement, we consider the fluctuations of an optimized hybrid quadrature, as $
S_q(X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)})|_{min} = \frac{1}{2} \langle \{X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)}, X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)}\} \rangle_{min},$
where $X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)}$ denotes a weighted hybrid quadrature, and $\mu_+$ and $\mu_-$ are the weight factors and $\phi_+$ and $\phi_-$ are the arbitery phase angles. A detailed definition of the hybrid quadrature and the procedure for obtaining the optimally squeezed quadrature are provided in Appendix~\ref{Maximally_Optimized_Squeezed_Quadrature}. The standard quantum limit (SQL) sets the boundary between separable and entangled states \cite{Duan_Inseparability_entanglement, Vitali_Zippilli_NJP}.
In our analysis, we vary the central filter frequency of the blue-detuned cavity for different linewidths and compute the corresponding optimized TMS [Fig.~\ref{SQZFilter_nonloc}(a)]. In parallel, we evaluate the maximal Bell parameter $B_{\max}$ [Fig.~\ref{SQZFilter_nonloc}(b)] by exploiting correlations between the filtered modes (see Appendix~\ref{nonlocality_bell} for details). The bound $B_{\max} \leq 2$ defines the regime of local realism, and its violation signals nonlocal correlations.
Consistent with expectations, both TMS and Bell violation are maximized for homodyne-like conditions, i.e., when the filters are symmetric ($\Omega_+ = -\Omega_-$), and decrease rapidly as the system moves toward heterodyne-like configurations. This reduction is more pronounced for narrower filter bandwidths (larger $\epsilon$), in agreement with earlier observations of entanglement in the output spectra of similar double-cavity optomechanical systems \cite{vitali_TMS_generator}.
Physically, a narrow bandwidth suppresses frequency mismatch under symmetric filtering, thereby enhancing quantum correlations and leading to stronger TMS and nonlocality. In contrast, for asymmetric (mismatched) filtering, a broader bandwidth increases the effective spectral overlap, facilitating coherent transfer and partially restoring correlations. Similar behavior has been reported in the context of coherent transfer of TMSV states \cite{mypaper_TMSV_filter, mypaper_TMSV_filter_STD}.
\subsection{TMS vs nonLocality}
Since both TMS and nonlocality are maximized under symmetric filtering, $\Omega_- = -\Omega_+ = \omega_m$, we impose this condition to identify experimentally accessible parameter regimes. As shown in Fig.~\ref{SQZnonlocal}(a-I,b-I), the violation of the CHSH inequality occurs only within a finite region of parameter space, whereas TMS persists over a much broader range. Although the double-cavity optomechanical setup provides an efficient source of entangled states, Bell nonlocality emerges predominantly in the large-cooperativity regime, $C_- = 4G_-^2/(\kappa_- \gamma_m) \gg 1$.
Increasing the ratio $G_+/G_-$ enhances the degree of squeezing [cf. Eq.~\eqref{squeezing_factor}], but the corresponding Bell parameter exhibits a nonmonotonic, bell-shaped dependence. In particular, maximal squeezing does not guarantee a violation of the CHSH inequality; instead, nonlocality can be observed for states with lower squeezing. This behavior is governed by the mixedness of the state, which plays a central role in limiting Bell violations, which can also be visualized analytically from the Eq. \eqref{nonlocality_formula} given in Appendix~\ref{nonlocality_bell}.
To elucidate this interplay, we compare the squeezing parameter $r$ with the purity of the state $\mu = \mathrm{Tr}(\rho^2)$ (see Eq. \eqref{purity}) as a function of the effective optomechanical coupling. As shown in Fig.~\ref{SQZnonlocal}(a-II,b-II), increasing cooperativity initially enhances squeezing before reaching saturation, while the purity undergoes a transient reduction followed by recovery. This behavior explains the simultaneous increase in both TMS and Bell violation observed in Fig.~\ref{SQZnonlocal}(a-I,b-I). A different trend emerges when varying the ratio $G_+/G_-$. Following Eq. \eqref{squeezing_factor}, although the squeezing parameter continues to increase [Fig.~\ref{SQZnonlocal}(a-III)], the Bell parameter decreases due to the growing mixedness of the output state [Fig.~\ref{SQZnonlocal}(b-III)]. Consequently, CHSH violation is restricted to states with moderate squeezing and sufficiently high purity.
These results reinforce the well-known fact that, for bipartite Gaussian CV states, entanglement is necessary but not sufficient for nonlocality \cite{agasti_nonlocally_CV}. Moreover, the Gaussian nature of the states imposes an upper bound on the CHSH parameter ($\lesssim 2.19$), even though the degree of squeezing itself is unbounded.
\subsection{ TMS and nonLocality vs Cavity Finesses}
The most interesting phenomenon is observed when we examine how the boundaries of TMS and nonlocality evolve with the cavity linewidths (or, equivalently, finesse). As shown in Fig.~\ref{nonlocal_SQZ_border_k}(a-I,a-II), increasing $\kappa_+$ while decreasing $\kappa_-$ shifts the SQL boundary toward higher cooperativity ($C_-$), thereby reducing the parameter region supporting TMS. In contrast, the corresponding nonlocality region, surprisingly, expands under the same conditions [Fig.~\ref{nonlocal_SQZ_border_k}(b-I,b-II)].
This contrasting behavior can be understood by analyzing the the degree of squeezing and purity of the states (see insets of Fig.~\ref{nonlocal_SQZ_border_k}). From Eq.~\eqref{squeezing_factor}, increasing $\kappa_+$ and decreasing $\kappa_-$ necessarily reduce the achievable squeezing, leading to a contraction of the TMS region. Conversely, at the same time, these changes help preserve the purity of the state, thereby enlarging the parameter space over which Bell inequality violations can occur.
Despite this expansion, the nonlocality region remains strictly contained within the entangled region, reaffirming that entanglement extends beyond the subset of states that exhibit Bell nonlocality.
%While working on it, we keep in mind that the change in the rate of dissipation $\kappa_k$ changes the internal cavity field $\alpha_k$, which modifies the effective optomechanical cou
\subsection{ TMS and nonLocality vs Q factor and temperature }
We further investigate how the squeezing and entanglement boundaries evolve with the mechanical quality factor and the temperature of the optomechanical system. As shown in Fig.~\ref{nonlocal_border_gamma}(a), increasing the mechanical dissipation rate $\gamma_m$ has only a weak effect on TMS, and the SQL boundary remains largely unchanged. This behavior is consistent with Eq.~\eqref{squeezing_factor}, where the squeezing parameter $r$ hardly depends on $\gamma_m$. The small residual shifts in the boundary primarily originate from the coefficients $\eta_k$ in Eq.~\eqref{output_bogoliubov}.
%is slightly pushed back for higher $G_+$, In contrast at lower $G_+$, the SQL boundery reduces
In contrast, the nonlocality boundary decreases monotonically with increasing mechanical linewidth. This trend can be understood in terms of the purity of the state, which degrades as mechanical dissipation increases, leading to more mixed output states and consequently weaker Bell violations.
Finally, increasing the thermal occupation of the mechanical reservoir further degrades quantum correlations. The additional thermal noise suppresses both the TMS and the Bell parameter, thereby reducing the extent of the SQL boundary as well as the region of nonlocality.
%***contribution of $\eta_\pm$
\section{CONCLUSION}
We have demonstrated how two-mode-squeezed (TMS) output fields can be generated in a double-cavity optomechanical system. Such hybrid quantum interfaces can be prepared experimentally by integrating optical cavities with microwave superconducting circuits \cite{Teufel_om_experiment}, or by employing high-$Q$ mechanical membranes in double-cavity configurations \cite{double_cav_alternate}. This approach provides an alternative route to TMS generation beyond conventional schemes based on parametric down-conversion in Kerr media or four-wave mixing in atomic vapors \cite{squeeze_light_sourse, TMS_generation, Squeezing_generation}, and is well suited for applications in precision measurements and quantum communication.
By exploiting correlations between spectrally matched output modes, we have investigated the violation of the CHSH Bell inequality. Our results show that the parameter region exhibiting Bell nonlocality is significantly smaller than the region where squeezing falls below the standard quantum limit (SQL). In particular, maximal squeezing does not necessarily imply Bell violation; instead, nonlocality can arise for states with comparatively lower squeezing.
We further find that the boundaries of TMS and nonlocality respond differently to variations in system parameters. Notably, changing the cavity linewidths (finesse) causes the TMS and nonlocality regions to shift in opposite directions, a behavior that can be understood in terms of the purity of the state. Additionally, increasing the mechanical dissipation rate reduces the extent of Bell violation, while leaving the TMS region largely unaffected. In contrast, thermal noise in the mechanical reservoir suppresses both squeezing and nonlocality by degrading the underlying quantum correlations.
Moreover, the generation of strongly entangled TMS states in optomechanical systems remains of broad interest, both for practical applications and for fundamental studies. In particular, the investigation of Bell nonlocality in such platforms is relevant for quantum communication protocols, constraints imposed by correlated measurements, and foundational tests of quantum mechanics \cite{Proposal_Bell_test_Electromechanics, Proposal_Bell_test_optomechanics}.
\begin{acknowledgments}
The author is grateful to A. Shukla for his insightful comments while developing the idea and validating the work. The work has been supported by the European Commission, MSCA GA no 101065991 (SingletSQL).
\end{acknowledgments}
\section*{ Disclosures }
The author declares no conflicts of interest.
\appendix
% \onecolumn
\section{Equation of Motion and Correlation Matrix }\label{Correlation_Matrix_Elements}
\subsection{Equation of Motion}\label{Langevin_EOM}
The Heisenberg-Langevin equation of motion of the system is given by
\begin{equation}
\dot{\mathbf{u} } = A \mathbf{u} +\mathbf{u}^{in}
\end{equation}
where
\begin{align}\label{A_mat}
A = \left(
\begin{array}{cccccc}
-\gamma_m & 0 & 0 & G_+ & 0 & -G_- \\
0 & -\gamma_m & G_+ & 0 & G_- & 0 \\
0 & G_+ & -\kappa_+ & 0 & 0 & 0 \\
G_+ & 0 & 0 & -\kappa_+ & 0 & 0 \\
0 & -G_- & 0 & 0 & -\kappa_- & 0 \\
G_- & 0 & 0 & 0 & 0 & -\kappa_- \\
\end{array}
\right)
\end{align}
where $\mathbf{u}=[X_+,Y_+,X_-,Y_-,x_m,p_m]$ where $X_k = (a_k + a_k^\dagger)/\sqrt{2}, Y_k = -i(a_k - a_k^\dagger)/\sqrt{2}$ are the amplitude and phase quadratures of cavity $k$ and $x_m = (b + b^\dagger)/\sqrt{2}, p_m = -i(b - b^\dagger)/\sqrt{2}$ are for mechanical oscillator, respectively. Similearly, $\mathbf{u}^{in}=[\sqrt{2\kappa_+}X_{+}^{in}, \sqrt{2\kappa_+}Y_+^{in}, \sqrt{2\kappa_-}X_-^{in}, \sqrt{2\kappa_-}Y_-^{in}, \sqrt{2\gamma_m}x_m^{in},\sqrt{2\gamma_m}p_m^{in}]$ where $X_{k}^{in} = (a_k^{in} + {a^{in}_k}^\dagger)/\sqrt{2}, Y_{k}^{in} = -i(a_k^{in} - {a_k^{in}}^\dagger)/\sqrt{2}$ and $x^{in}_m = (b^{in} + { b^{in}}^\dagger)/\sqrt{2}, p^{in}_m = -i(b^{in} - {b^{in}}^\dagger)/\sqrt{2}$ are the amplitude and phase quadratures and $ a_k^{in}({a_k^{in}}^\dagger), b^{in}({b^{in}}^\dagger)$ are the anihilation (creation) field operators of the inputs of cavity $k$ and mechanical oscillator, respectively. The thermal occupation of the mechanical reservior is $\langle\ b^\dagger b \rangle= n_{m} = (e^{\hbar \omega_m/k_BT}-1)^{-1}$. The correlation function of the input noise to the cavity is given by
\begin{equation}
\langle a_k^{in}(t) {a_k^{in}}^\dagger (t') \rangle = [N(\omega_{k}) +1] \delta(t-t')
\end{equation}
where $N(\omega_{k}) = (e^{\hbar \omega_{k}/k_BT}-1)^{-1}$ are the mean thermal occupation numbers of the cavity reservoirs. As the optical frequency is very high $\hbar \omega_{k}/k_BT >>1$, the thermal bath moreovetr behaves as a vacuum $(N(\omega_c) \approx 0)$.
\begin{figure}
\includegraphics[width= 1 \linewidth]{stability}
\caption{ Stability of the system for $\kappa_+ = 2\kappa$ at left, $\kappa_+ = \kappa$ at center and $\kappa_+ = \kappa/2 $ at right. Green is the stable region, and the blue region is unstable. All other parameters remain the same with Fig. \ref{SQZFilter_nonloc}
}\label{stability}
\end{figure}
\subsection{Stability of the System} \label{Stability_System}
The stability of the system can be determined by following Routh-Hurwitz stability criteria, yielding a nontrivial constraint which says the real parts of all the eigenvalues of the matrix $A$, given in Eq. Eq. \eqref{A_mat} must be negative. In Fig. \ref{stability}, we distinguish the regions where the system can reach a stable state from the region where it does not, and realize that in the case of $\kappa_+ = \kappa_-$, the entire region of $G_+ \leq G_-$ gives a stable solution of the system. However, in the case of $\kappa_+ \neq \kappa_-$, the region where one can end up obtaining a stable solution reduces significantly, especially for high cooperativity $(C_-)$.
\subsection{Filtered Output}
To determine the output correlation, we need to determine the filtered output quadratures. In the case of TMS output, the filtered quadratures can be determined as
\begin{equation}\label{filtered_quadrature}
\begin{bmatrix}
X_k^{f,out} (r,t) \\
Y_k^{f,out} (-r,t)
\end{bmatrix} = \int_{-\infty}^{t} \mathrm{d}t' \, T_k(t-t') \begin{bmatrix}
X_k^{out} (r,t) \\
Y_k^{out} (-r,t)
\end{bmatrix}
\end{equation}
where $X_{k}^{out} = (a_k^{out} + {a^{out}_k}^\dagger)/\sqrt{2}, Y_{k}^{out} = -i(a_k^{out} - {a_k^{out}}^\dagger)/\sqrt{2}$, $X_{k}^{f,out} = (f_k^{out} + {f^{out}_k}^\dagger)/\sqrt{2}, Y_{k}^{f,out} = -i(f_k^{out} - {f_k^{out}}^\dagger)/\sqrt{2}$, and
\begin{equation}
T_k(t) = \begin{bmatrix}
\mathcal{R} (h_k^{f,out} ) & -\mathcal{I} (h_k^{f,out} ) \\
\mathcal{I} (h_k^{f,out} ) & \mathcal{R} (h_k^{f,out} )
\end{bmatrix}
\end{equation}
where $\mathcal{R}, \mathcal{I}$ represent the real and imaginary parts of the function, respectively. The spectral frequency distribution of the filter function given in Eq. \eqref{filterfunction} in Fourier space is
\begin{equation}
\tilde{h}_k(\omega) = \frac{\sqrt{\tau/\pi}}{1-i\tau(\omega-\Omega_k)}
\end{equation}
Following Eq. \eqref{filtered_quadrature}, in the frequency domain, the filtered quadratures are given by
\begin{equation}
\begin{bmatrix}
\tilde{X}_k^{f,out} (r,\omega) \\
\tilde{Y}_k^{f,out} (-r,\omega)
\end{bmatrix} = \sqrt{2\pi} \, \tilde{T}_k(\omega) \begin{bmatrix}
X_k^{out} (r,\omega) \\
Y_k^{out} (-r,\omega)
\end{bmatrix}
\end{equation}
where $\tilde{X}_k^{f,out}(r,\omega),\tilde{Y}_k^{f,out}(r,\omega),\tilde{T}_k(t)$ is the Fourier transformed version of ${X}_k^{f,out}(r,t),{Y}_k^{f,out}(r,t)$ and ${T}_k(t)$.
The elements of the associated covariance matrix of the outputs are
\begin{equation}\label{correlation_mat}
V_{ij}^{f,out}
(\infty) = \frac{1}{2} \lim_{t \to \infty} \langle \{u_{i}^{f,out}, u_{i}^{f,out} \} \rangle
\end{equation}
where $\mathbf{u}^{f,out} =[X_+^{f,out},Y_+^{f,out},X_-^{f,out},Y_-^{f,out}]$ where $X_k^{f,out} = (f_k^{out} + {f_k^{out} }^\dagger)/\sqrt{2}, Y_k^{f,out} = -i(f_k^{out} - {f_k^{out} }^\dagger)/\sqrt{2}$ are the amplitude and phase quadratures of respective cavities.
\section{ Squeezing and non-Locality}
\subsection{Maximally Optimized Two-Mode Squeezed Quadrature} \label{Maximally_Optimized_Squeezed_Quadrature}
The weighted quadrature is
\begin{multline}\label{weighted_quadrature}
X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)} = \frac{1}{\sqrt{\mu_+^2+\mu_-^2}} \bigg[\mu_+ e^{-i\phi_+} f_+^{out} + \mu_+ e^{i\phi_+} {f_+^{out}}^\dagger \\
+ \mu_- e^{-i\phi_-} f_-^{out} + \mu_- e^{i\phi_-} {f_-^{out}}^\dagger \bigg]
\end{multline}
where $\mu_+,\mu_- \geq 0$ are the weight parameters introduced as the scaling of two systems, and $\phi_+, \phi_-$ are the arbitrary phase angles of the filtered composite quadrature.
The hybrid quadrature variance, defined by $S_q(X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)}) = \frac{1}{2} \langle \{X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)}, X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)}\} \rangle$ becomes
\begin{widetext}
\begin{align}
S_q(X^{(\phi_+\phi_-)}_{(\mu_+\mu_-)}) = \frac{2}{\mu_+^2+\mu_-^2} \bigg[ &\mu_+^2 (\cos^2 (\phi_+) V_{11}^{f,out} + \sin^2 (\phi_+) V_{22}^{f,out} + \sin (2\phi_+) V_{12}^{f,out}) \nonumber\\
& + \mu_-^2 (\cos^2 (\phi_-) V_{33}^{f,out} + \sin^2 (\phi_-) V_{44}^{f,out} + \sin (2\phi_-) V_{34}^{f,out}) \nonumber\\
& + 2\mu_+\mu_- \cos(\phi_+) \cos(\phi_-) V_{31}^{f,out} + 2\mu_+\mu_- \sin(\phi_+) \sin(\phi_-) V_{24}^{f,out} \nonumber\\
& + 2\mu_+\mu_- \cos(\phi_+) \sin(\phi_-) V_{41}^{f,out} + 2\mu_+\mu_- \sin(\phi_+) \cos(\phi_-) V_{23}^{f,out}
\bigg]
\end{align}
\end{widetext}
The maximally squeezed quadrature is optimized by varrying $\mu_+,\mu_-$ and $( \phi_+, \phi_-)$ to minimize $S_q$, which is furthermore used in Sec. \ref{Results} as the measure of TMS ($S_q|_{min}$). The measurement of maximally optimized TMS is a direct measurement of the entanglement between specific spectral modes \cite{Vitali_Zippilli_NJP}. The squeezing variance can be used to construct entanglement criteria. The maximal squeezing, i.e., minimal TMS variance $(S_q|_{min})$ going below SQL (which is 1) ensures the state to be entangled \cite{Vitali_Zippilli_NJP, Duan_Inseparability_entanglement}.
\subsection{nonlocality-Maximized Bell's function} \label{nonlocality_bell}
The generalized condition for CHSH nonlocality for a CV bipartite Gaussian system is done using the Banaszek-Wodkiewicz phase-space Wigner representation. In this process, one has to transform the correlation matrix $\textbf{V}^{f,out}$ given in Eq. \eqref{correlation_mat} to a standard form using a local linear unitary Bogoliubov operator, and afterwards expressing
the Bell’s function in terms of Wigner functions through Banaszek-W\'odkiewicz representation \cite{agasti_nonlocally_CV}. We find the maximal value of the Bell's function, through investigating over the infinite range of the phase space, as
well as
\begin{equation}\label{nonlocality_formula} B_{max} = \mu \left[1 + \left(\frac{ \sqrt{nm} }{ \sqrt{ nm } +\tilde{c} }\right)^{\frac{ \sqrt{ nm } }{ \sqrt{ nm } +2\tilde{c} }} \left( \frac{ \sqrt{ n m } +2\tilde{c} }{ \sqrt{ n m } +\tilde{c} }\right) \right]
\end{equation}
where
$n^2 = \det(V_+^{f,out}), m^2 = \det(V_-^{f,out}), c_1c_2 = \det(V_{\pm}^{f,out})$ and $\det(\textbf{V}^{f,out}) = (nm-c_1^2)(nm-c_2^2) $, where $\textbf{V}^{f,out} = \begin{bmatrix}
\textbf{V}^{f,out}_+ & \textbf{V}^{f,out}_\pm \\
\textbf{V}^{f,out'}_\pm &\textbf{V}^{f,out}_-
\end{bmatrix}$
and $\tilde{c} = \max[|c_1|,|c_2|]$.
The quantum mechanical criteria of the nonlocal realism of a bipartite measurement are the violation of the condition $|B|_{max}\leq 2$. Larger $|B|_{max}$ indicates non-locality to be stronger. The maximum value of $|B|_{max}$ can reach in case of infinitely entangled pure TMS Gaussian states is 2.19 \cite{agasti_nonlocally_CV}. The first term of Eq. \eqref{nonlocality_formula}, given by the mixedness of the state
\begin{equation}\label{purity}
\mu = \text{Tr}[\rho^2]= \frac{1}{4 \sqrt{\det(\textbf{V}^{f,out} )} }
\end{equation}
By comparing the resulting nonlocality bound with Simon’s separability criterion \cite{Simon}, one can also demonstrate that entanglement is necessary but not sufficient for Bell nonlocality, whereas Bell nonlocality always implies entanglement within the class of bipartite Gaussian states. This provides a unified analytical framework clarifying the boundary between these fundamental quantum correlations \cite{agasti_nonlocally_CV}.
\nocite{*}
\bibliography{apssamp}% Produces the bibliography via BibTeX.
\bibliographystyle{apsrev4-2}
\end{document}
%
% ****** End of file apssamp.tex ******