---
title: Two-Mode Squeezed Thermal States
url: https://www.emergentmind.com/topics/two-mode-squeezed-thermal-states
type: topic
---

# Two-Mode Squeezed Thermal States

Searching arXiv for the specified paper and closely related work on two-mode squeezed thermal states to ground the article in cited literature.
arxiv_search query="Two-Mode Squeezed Thermal States 1203.0595 photon-added two-mode squeezed thermal state teleportation covariance matrix" max_results=10
Two-mode squeezed thermal states (TMSTSs), also denoted TMST or two-mode squeezed thermal states in the literature, are bipartite continuous-variable Gaussian states obtained by applying a two-mode squeezing transformation to a product of single-mode thermal states. In the symmetric case, both input modes carry the same mean thermal population, while the squeezing parameter controls the intermode correlations; the pure limit recovers the two-mode squeezed vacuum, and the limit \(r=0\) gives two independent thermal modes [1203.0595]. Because they incorporate both EPR-type correlations and input thermal noise, TMSTSs serve as a realistic resource model for continuous-variable teleportation, Gaussian-state metrology, optomechanical reservoir engineering, and spectral-mode quantum state reconstruction [2403.03204].

## 1. Definition, notation, and limiting cases

The general two-mode squeezed thermal state can be written as
\[
\hat\rho_{\rm ST}(n_{1},n_{2},r,\phi)
=\hat S_{12}(r,\phi)\,
\bigl[\hat\rho_{T}(n_{1})\otimes\hat\rho_{T}(n_{2})\bigr]\,
\hat S_{12}^\dagger(r,\phi),
\]
where the input single-mode thermal state is
\[
\hat\rho_T(n)=\frac{1}{n+1}\exp\!\Bigl[-\ln\!\Bigl(\frac{n+1}{n}\Bigr)\hat a^\dagger\hat a\Bigr],\qquad n\ge 0,
\]
and \(r,\phi\) are the squeezing strength and phase, respectively [1605.06524]. In the symmetric case used extensively in teleportation and non-Gaussian-state studies,
\[
\rho_{\rm TMSTS}(r,\bar n)=S(r)\,\rho_{th,1}\otimes\rho_{th,2}\,S^\dagger(r),
\]
with
\[
S(r)=\exp\!\bigl[r\,(a^\dagger b^\dagger-a\,b)\bigr],
\qquad
\rho_{th}
=\sum_{k=0}^\infty
\frac{\bar n^k}{(\bar n+1)^{k+1}}\,|k\rangle\langle k|.
\]
The literature also uses the phase-convention
\[
S_2(r)=\exp\!\bigl[r\,(a_1a_2-a_1^\dagger a_2^\dagger)\bigr]
\]
for the same class of states [2403.03204].

Several limits organize the subject. When \(n_{1}=n_{2}=0\), the state reduces to the two-mode squeezed vacuum [2406.09134]. When \(r=0\), it reduces to the input product of thermal states [2409.07349]. In the special symmetric thermal case \(n_{1}=n_{2}=\bar n\), the state is zero-mean and fully specified by a \(4\times 4\) covariance matrix, which makes it part of the standard two-mode Gaussian-state manifold studied in metrology and information geometry [1605.06524].

A recurrent notational distinction is between the broad class of squeezed thermal states with possibly unequal thermal inputs \((n_1,n_2)\) and the symmetric TMSTS/TMST family with \(n_1=n_2=\bar n\). This distinction matters for covariance asymmetry, entanglement thresholds, and the angle of maximally squeezed hybrid quadratures in filtered settings [2406.09134].

## 2. Gaussian phase-space structure

In quadrature form, a zero-mean two-mode squeezed thermal state has covariance matrix
\[
V_{\rm ST}=
\begin{pmatrix}
b_{1}\,I_{2}&c\,R(\phi)\\
c\,R^{T}(\phi)&b_{2}\,I_{2}
\end{pmatrix},
\qquad
R(\phi)=
\begin{pmatrix}
\cos\phi&-\sin\phi\\
\sin\phi&\cos\phi
\end{pmatrix},
\]
with
\[
b_{1}=(n_{1}+\tfrac12)\cosh^{2}r+(n_{2}+\tfrac12)\sinh^{2}r,
\]
\[
b_{2}=(n_{1}+\tfrac12)\sinh^{2}r+(n_{2}+\tfrac12)\cosh^{2}r,
\]
\[
c=(n_{1}+n_{2}+1)\sinh r\,\cosh r.
\]
For \(\phi=0\), this becomes the standard-form matrix with opposite signs in the \(p\)-quadrature correlations [1605.06524].

For the symmetric TMST parameterization used in teleportation studies, defining \(\kappa\equiv \mu+\tfrac12\), the covariance matrix in the ordering \((q_1,p_1,q_2,p_2)\) is
\[
V_{\rm TMST}(r,\mu)
=\kappa
\begin{pmatrix}
\cosh 2r&0&\sinh 2r&0\\
0&\cosh 2r&0&-\sinh 2r\\
\sinh 2r&0&\cosh 2r&0\\
0&-\sinh 2r&0&\cosh 2r
\end{pmatrix},
\]
and the Wigner function is
\[
W(\xi)=\frac{1}{\pi^2\sqrt{\det V}}
\exp\!\Bigl[-\frac12\,\xi^{T}V^{-1}\xi\Bigr],
\qquad
\xi=(q_1,p_1,q_2,p_2)^T.
\]
The corresponding characteristic function is
\[
\chi_{\rm TMST}(\tau_1,\sigma_1,\tau_2,\sigma_2)
=
\exp\!\Bigl[
-\kappa(\tau_1^2+\sigma_1^2+\tau_2^2+\sigma_2^2)\frac{\cosh 2r}{2}
+\kappa(\tau_1\tau_2-\sigma_1\sigma_2)\sinh 2r
\Bigr].
\]
These formulas make explicit that the Gaussian TMST is fully encoded by second moments [2403.03204].

An alternative derivation starts from the Glauber-Sudarshan representation of the thermal input,
\[
\rho_{th}
=\frac{1}{\bar n}\int\frac{d^2\alpha}{\pi}\,
e^{-|\alpha|^2/\bar n}\,
|\alpha\rangle\langle\alpha|,
\]
which leads, after two-mode squeezing, to a two-mode \(P\)-function
\[
P(\alpha,\beta)
=\tilde A_1
\exp\!\Bigl[
\tilde A_2(\alpha^*\beta^*+\alpha\beta)
-\tilde A_3(|\alpha|^2+|\beta|^2)
\Bigr].
\]
This form is especially useful in normal-order calculations and in deriving closed expressions for non-Gaussian deformations of the state [1203.0595].

The Gaussian structure also fixes collective-quadrature squeezing. For symmetric mechanical realizations, the collective quadratures
\[
X_+=\frac{X_1+Y_2}{\sqrt2},
\qquad
X_-=\frac{Y_1+X_2}{\sqrt2}
\]
have stationary variances
\[
\sigma_+^2=\frac{\sigma_0^2}{1+g},
\qquad
\sigma_-^2=\frac{\sigma_0^2}{1-g},
\]
showing the familiar coexistence of squeezed and anti-squeezed directions [1509.02723].

## 3. Non-Gaussian deformations: photon addition, subtraction, and catalysis

A central non-Gaussian extension is the photon-added TMSTS,
\[
\rho_{m,n}^{\rm add}
=
N_{m,n}^{-1}\;
a^{\dagger m}b^{\dagger n}\,
\rho_{\rm TMSTS}(r,\bar n)\,
a^m b^n,
\]
where
\[
N_{m,n}
=
\mathrm{Tr}\!\left(
a^{\dagger m}b^{\dagger n}\rho_{\rm TMSTS}a^m b^n
\right)
\]
is the normalization factor [1203.0595]. Defining
\[
B_1=\cosh^2r+\bar n\cosh 2r,
\qquad
B_2=(2\bar n+1)\sinh r\cosh r,
\]
and, for \(m\le n\),
\[
\omega=\bar n^2+(2\bar n+1)\cosh^2r,
\qquad
\upsilon=\bar n(\bar n+1)\cosh 4r+(\bar n+\cosh^2r)\cosh 2r,
\]
one obtains the closed form
\[
N_{m,n}
=
m!\,n!\;B_1^{\,n-m}\,\omega^m\;
P_m^{(0,n-m)}\!\Bigl(\frac{\upsilon}{\omega}\Bigr),
\]
so the normalization is expressed through a Jacobi polynomial [1203.0595].

The Wigner function of the photon-added state factorizes as
\[
W_{m,n}(\alpha,\beta)=W_0(\alpha,\beta)\,F_{m,n}(\alpha,\beta),
\]
where \(W_0\) is the Gaussian background of the underlying TMSTS and \(F_{m,n}\) is a non-Gaussian multiplier written in terms of two-variable Hermite polynomials. For single-photon addition \((m=0,n=1)\),
\[
W_{0,1}(0,0)
=
-\frac{\bar n+\cosh^2r}{(2\bar n+1)^3\pi^2(\cosh^2r+\bar n\cosh 2r)}<0
\qquad (r>0),
\]
which gives an analytic certificate of Wigner negativity [1203.0595]. This sharply separates the Gaussian TMSTS, whose Wigner function is Gaussian, from its photon-added descendants, which are non-Gaussian and can be Wigner-negative.

A broader experimentally realistic scheme applies three operations to a TMST resource: photon subtraction, photon addition, and photon catalysis. Each TMST mode \(A_i\) is mixed with an ancillary Fock state \(|m_i\rangle\) at a beam splitter of transmissivity \(T_i\), and one conditions on detecting \(n_i\) photons in the auxiliary output. In the idealized limit \(T_i\to 1\), photon subtraction corresponds to
\[
\rho\to a_1a_2\,\rho\,a_2^\dagger a_1^\dagger,
\]
photon addition to
\[
\rho\to a_1^\dagger a_2^\dagger\,\rho\,a_2 a_1,
\]
and photon catalysis tends to an operator proportional to \(a_i a_i^\dagger\) [2403.03204].

These operations are not interchangeable. Photon addition and subtraction both increase non-Gaussianity, but their entanglement and teleportation effects differ quantitatively; photon catalysis introduces a transmissivity-dependent mixture of identity and photon-number-dependent damping rather than a simple ladder-operator action [2403.03204].

## 4. Entanglement, EPR squeezing, and non-locality

For Gaussian TMSTSs, entanglement is naturally formulated through partial transpose at the covariance-matrix level. In the symmetric unfiltered case,
\[
\nu_{\pm}=(n_{\rm in}+\tfrac12)e^{\pm 2r},
\]
where \(\nu_-\) is the smallest symplectic eigenvalue of the partially transposed covariance matrix, and the logarithmic negativity is
\[
E_N=\max\!\Bigl[0,-\ln\!\bigl(2\nu_-\bigr)\Bigr]
=
\max\!\Bigl[0,-\ln\!\bigl(2(n_{\rm in}+\tfrac12)e^{-2r}\bigr)\Bigr].
\]
Thus \(E_N>0\) is necessary and sufficient for two-mode Gaussian entanglement in this setting [2409.07349].

Equivalent criteria appear in different normalizations. In the Gaussian optomechanical treatment, the Duan condition reads
\[
\langle X_+^2\rangle+\langle P_-^2\rangle<1,
\]
while in spectral homodyne reconstruction the Simon PPT test is implemented through the smallest partially transposed symplectic eigenvalue \(\tilde \nu_-\), with entanglement present iff \(\tilde \nu_-<1\) in that quadrature convention [1404.2672]. In the electromechanical parametrically driven setting, the analytic entanglement condition is
\[
\sinh^2 r>n_S n_A,
\]
and for equal thermal populations \(n_S=n_A=n_{\rm th}\) it is reported as
\[
r>\operatorname{arcsinh}(n_{\rm th}),
\qquad
e^{2r}>2n_{\rm th}+1
\]
[1405.5270].

For non-Gaussian photon-added TMSTSs, one convenient sufficient inseparability test is the Shchukin-Vogel quantity
\[
SV_{m,n}
=
\Bigl\langle a^\dagger a-\tfrac12\Bigr\rangle
\Bigl\langle b^\dagger b-\tfrac12\Bigr\rangle
-\langle a^\dagger b^\dagger\rangle\,\langle a\,b\rangle,
\]
with inseparability certified when \(SV_{m,n}<0\). Numerical plots show that for realistic \(\bar n\), this threshold is reached at lower squeezing \(r\) for photon addition than for photon subtraction; for \(\bar n=1\), the quoted thresholds are \(r_a\approx 0.31\) for single-photon addition and \(r_c\approx 0.55\) for photon subtraction [1203.0595].

Two points qualify the common expectation that “more squeezing always helps.” First, Bell-CHSH non-locality is stricter than entanglement. Using displaced parity operators, the phase-space CHSH parameter satisfies \(|B|_{\max}\le 2\) under local realism, and for a pure two-mode squeezed vacuum one has
\[
B_{\max}=2\sqrt{1+\tanh^2(2r)},
\qquad
B_{\max}\to 2\sqrt2 \ \text{as}\ r\to\infty.
\]
For mixed or filtered states, the Bell-violating region lies strictly inside the entangled region [2406.09134]. Second, with non-identical filters the filtered logarithmic negativity becomes bell-shaped in \(r\): it rises, reaches a maximum near
\[
r_{\max}\approx \frac12\operatorname{arctanh}(K_f),
\]
and vanishes at
\[
r_{EN}^{\rm cutoff}=\operatorname{arctanh}(K_f),
\]
where \(K_f<1\) is the filter-overlap factor. This means that arbitrarily large input squeezing can be counterproductive once spectral overlap is reduced [2406.09134].

## 5. Teleportation resource and statistical geometry

In the Braunstein-Kimble protocol, a TMST or its non-Gaussian variants may be used as the shared entangled resource. For a coherent-state input, the Gaussian benchmark fidelity is
\[
F_{\rm TMST}(r,\mu)
=
\bigl[1+(2\bar n+1)e^{-2r}\bigr]^{-1},
\]
which makes the degradation from thermal noise explicit [2403.03204]. In the photon-added analysis, the coherent-state \(|0\rangle\) fidelity is written as \(\mathcal F_{m,n}^{(\bar n,r)}\), with special cases
\[
\mathcal F_{0,0}
=
\frac{1}{(2\bar n+1)e^{-2r}+1},
\]
\[
\mathcal F_{1,1}
=
\frac{(1+\tanh r)^3}{4(1+\tanh^2 r)},
\qquad
\mathcal F_{0,1}
=
\frac{1+\tanh r}{4(1-\tanh r)}\operatorname{sech}^2 r
\quad (\bar n=0).
\]
The numerical comparison shows that symmetric photon addition \((m=n)\) raises fidelity for \(r\) above a modest threshold, whereas non-symmetric addition \((m\neq n)\) typically degrades fidelity [1203.0595].

When success probability is incorporated, the comparison among non-Gaussian operations changes. Defining the fidelity enhancement \(\Delta F^{NG}=F^{NG}-F^{TMST}\), the success probability \(P^{NG}\), and the figure of merit
\[
\mathcal R^{NG}=\Delta F^{NG}\times P^{NG},
\]
numerical optimization for teleporting a coherent state with \(\mu=0.51\) yields the following quoted optima: single-photon catalysis is optimal at low squeezing with \(r_{\rm opt}\approx 0.24\), \(T_{\rm opt}\approx 0.18\), \(F^{1-PC}\approx 0.66\), \(\Delta F\approx 0.055\), \(P\approx 4\times 10^{-2}\), and \(\mathcal R\approx 2.2\times 10^{-3}\); single-photon subtraction is optimal at intermediate squeezing with \(r_{\rm opt}\approx 0.64\), \(T_{\rm opt}\approx 0.78\), \(F^{1-PS}\approx 0.81\), \(\Delta F\approx 0.033\), \(P\approx 2.5\times 10^{-2}\), and \(\mathcal R\approx 8.2\times 10^{-4}\) [2403.03204].

Beyond communication, TMSTSs define a four-parameter Gaussian-state manifold with parameters \((n_1,n_2,r,\phi)\). Marian and Marian showed that the quantum Fisher information metric on this manifold is diagonal. Writing \(\theta=2r\),
\[
(ds_{\rm F})^{2}
=
H_{n_1}\,dn_1^2
+
H_{n_2}\,dn_2^2
+
H_{\theta}\,d\theta^2
+
H_{\phi}\,d\phi^2,
\]
with
\[
H_{n_1}=\frac{1}{n_1(n_1+1)},
\qquad
H_{n_2}=\frac{1}{n_2(n_2+1)},
\]
\[
H_{\theta}
=
\frac{(n_1+n_2+1)^2}{2n_1n_2+n_1+n_2+1},
\qquad
H_{\phi}
=
\frac{(n_1+n_2+1)^2\sinh^2\theta}{2n_1n_2+n_1+n_2+1}.
\]
The diagonal form implies orthogonality of the parameters in the local metric sense, and the quantum Cramér-Rao bound becomes
\[
\mathrm{Var}(\hat\xi_\alpha)\ge \frac{1}{N\,H_{\xi_\alpha}}
\]
for \(N\) independent copies [1605.06524].

## 6. Preparation, reconstruction, filtering, and dynamical settings

TMSTSs appear both as directly generated optical states and as effective steady states in engineered dissipative systems. In optical experiments, symmetric two-mode squeezed thermal states of spectral sideband modes \(\omega_0\pm \Omega\) can be reconstructed with a single homodyne detector and active cavity stabilization. The measured covariance matrix is assembled in the symmetric/antisymmetric spectral-mode basis and then transformed back to the original sideband basis without iterative maximum-likelihood reconstruction, because Gaussianity reduces the problem to moment inversion [1505.03903]. The same experiment reports successful tests on states ranging from uncorrelated coherent states to entangled states, with reconstructed examples including a pure squeezed state with \(\tilde\nu_-=0.50\pm 0.02\) and \(E_N\approx 1\) bit [1505.03903].

In optomechanics, TMSTS-like states arise as targets of reservoir engineering. A three-mode setup with two mechanical oscillators coupled to a single cavity mode admits an effective linearized Hamiltonian
\[
\hat H_{\rm lin}
=
\Omega(\hat a^\dagger\hat a-\hat b^\dagger\hat b)
+
G_+\bigl[(\hat a+\hat b)\hat c+\text{h.c.}\bigr]
+
G_-\bigl[(\hat a+\hat b)\hat c^\dagger+\text{h.c.}\bigr],
\]
which is rewritten in terms of Bogoliubov operators
\[
\hat\beta_1=\hat a\cosh r+\hat b^\dagger\sinh r,
\qquad
\hat\beta_2=\hat b\cosh r+\hat a^\dagger\sinh r,
\]
with \(\tanh r=G_+/G_-\). Cooling the sum mode \(\hat\beta_1+\hat\beta_2\) together with coherent coupling from \(\Omega\) yields a highly pure two-mode squeezed mechanical steady state; in the resolved-sideband regime, cooperativities \(C_-\gtrsim 10^3\) are quoted as sufficient for \(>10\) dB of two-mode squeezing, \(E_{\mathcal N}\gtrsim 2\), purity \(\mu\approx 0.9\), and teleportation fidelity \(\mathcal F>0.9\) [1404.2672].

Closely related dissipation-driven proposals show that finite-temperature steady states can still be written formally as
\[
\rho_{\rm th}^{(2)}(r)
=
\hat S_{12}(r)\,
\bigl[\rho_T(n_{T1})\otimes\rho_T(n_{T2})\bigr]\,
\hat S_{12}^\dagger(r),
\]
and that ground-state precooling is not necessary for two-mode squeezing. In the large-cooperativity limit, the maximal tolerable thermal occupation \(\bar n_{\rm th,max}\) can become very large even at modest \(r\); one worked example quotes \(\bar n_{\rm th,max}\sim 70\) for \(\kappa=0.05\,\omega_m\), \(\gamma_m=10^{-4}\omega_m\), \(\chi_1=0.01\omega_m\), and \(\chi_2=0.03\omega_m\) [1301.5698].

Parametric modulation provides a complementary route. In a cavity opto-mechanical system, modulation of the optical spring at \(\omega_1+\omega_2\) generates the two-mode squeezing Hamiltonian
\[
H_I=\hbar G(a_1a_2+a_1^\dagger a_2^\dagger),
\]
and pulsed excitation can transiently surpass the stationary \(3\) dB limit. The reported experiment reaches a minimal normalized spectral density \(S_{\min}\approx 0.28\) at \(g\approx 0.9\), corresponding to \(S_{dB}\approx -5.5\) dB, and in the pulsed regime the minimal variance falls transiently to \(\sigma_{\min}^2\approx 0.28\) at \(t\approx 1\text{--}2\) ms [1509.02723]. In an electromechanical resonator implementing the analogous parametric-down-conversion Hamiltonian
\[
H_{\rm int}=\hbar g(a_S a_A+a_S^\dagger a_A^\dagger),
\]
the measured suppression of joint quadrature noise reaches \(-4.76\) dB below the bare thermal level at \(T=300\) K, with correlation coefficients approaching unity as the pump is increased [1405.5270].

Filtered and open-system variants further refine the picture. For filtered two-mode squeezed mixed states, entanglement and non-locality are maximized when the two filters are identical, and the filter overlap factor \(K_f\) rescales the intermode correlations [2406.09134]. Under thermalization dynamics in a Markovian bath,
\[
\Sigma(t)=e^{-\Gamma t}\Sigma(0)+\bigl(2n_{\rm env}+1\bigr)\frac12\bigl[1-e^{-\Gamma t}\bigr]I_4,
\]
so the smallest partially transposed symplectic eigenvalue evolves as
\[
\nu_-(t)
=
(n_{\rm env}+\tfrac12)\bigl[1-e^{-\Gamma t}\bigr]
+
(n_{\rm in}+\tfrac12)e^{-\Gamma t}e^{-2r}.
\]
The reported conclusions are that increasing the thermal population of the environment enhances the rate of dissipation, stronger interaction slows dissipation in a normalized dimensionless time scale, and identical filters keep entanglement and non-locality at their peak [2409.07349].

Taken together, these results place TMSTSs at the intersection of Gaussian-state theory, non-Gaussian state engineering, CV teleportation, and engineered dissipation. Their defining feature is not merely the coexistence of squeezing and thermal noise, but the fact that this coexistence remains analytically tractable across covariance-matrix methods, entanglement criteria, homodyne tomography, and several experimentally relevant non-Gaussian and filtered generalizations.

Source: https://www.emergentmind.com/topics/two-mode-squeezed-thermal-states