Gaussian states and operations – a quick reference

Jonatan Bohr Brask1, ∗

1
Department of Physics, Technical University of Denmark, Fysikvej, Kongens Lyngby 2800, Denmark
(Dated: February 12, 2021)
Gaussian quantum states of bosonic systems are an important class of states. In particular,
they play a key role in quantum optics as all processes generated by Hamiltonians up to second
order in the field operators (i.e. linear optics and quadrature squeezing) preserve Gaussianity. A
powerful approach to calculations and analysis of Gaussian states is using phase-space variables and
symplectic transformations. The purpose of this note is to serve as a concise reference for performing
phase-space calculations on Gaussian states. In particular, we list symplectic transformations for
commonly used optical operations (displacements, beam splitters, squeezing), and formulae for
tracing out modes, treating homodyne measurements, and computing fidelities.
arXiv:2102.05748v1 [quant-ph] 10 Feb 2021

I. SOURCES Note that different conventions for the normalisation in

Eq. (3) can be found in the √ literature (for instance,
There are many excellent sources on the analysis of Ref. [2] uses 1 rather than 1/ 2, and Ref. [5] uses 1/2).
Gaussian states and transformation and their use in We will stick to the canoncial convention above through-
quantum optics and quantum information science, which out. Following the convention in quantum optics, we also
provide much more thorough treatments of the subject refer to x̂k and p̂k as quadrature operators, and we define
than presented here. The present note should be thought the general, rotated quadrature
of only as a quick-reference collection of fomulae to aid
1
in calculations. As such, I will not generally provide ci- q̂k (φ) = √ (âk e−iφ + â†k eiφ ) (5)
tations to the original work deriving each of the results 2
presented. For thorough references to primary sources = cos(φ)x̂k + sin(φ)p̂k , (6)
and historical overview, readers may refer to the reviews
[1, 2]. In addition to these reviews, I draw on the ex- such that q̂k (0) = x̂k and q̂k ( π2 ) = p̂k . For the commuta-
cellent lecture notes of Ferraro, Olivares, and Paris [3], tor, we have
as well as the PhD thesis of Zhang [4], the books by
Gerry and Knight [5] and Serafini [6], and on the work of [q̂k (φ), q̂l (ϕ)] = iδkl sin(ϕ − φ). (7)
Banchi, Braunstein, and Pirandola on computing fidelity
It will be convenient to collect the 2n quadrature opera-
for arbitrary Gaussian states [7].
tors for the n modes into a vector. We define the column
vector
II. BASIC CONCEPTS AND NOTATION x̂1
 
 p̂1 
1. Mode and quadrature operators  . 
 ..  .
r̂ =  (8)

x̂ 
n
We will consider bosonic systems consisting of n har-
p̂n
monic oscillators, with corresponding annihilation and
creation operators âk , â†k with k = 1, . . . , n. We also re- The canonical commutation relations can then be written
fer to âk as mode operators. Throughout the note, we set
~ = 1 and adopt the canonical commutation relations [r̂k , r̂l ] = iΩkl , (9)
[âk , âl ] = [â†k , â†l ] = 0, (1) where Ω is the symplectic matrix defined by
[âk , â†l ] = δkl . (2) 2n    
M 0 1 0 1
We define Hermitian position- and momentum-like oper- Ω= = ⊗ 11n , (10)
−1 0 −1 0
ators for each mode k=1

1 1 where 11n is the n × n identity matrix.

x̂k = √ (âk + â†k ), p̂k = √ (âk − â†k ), (3)
2 i 2
with the canonical commutator
2. Displacement operators
[x̂k , p̂l ] = iδkl . (4)
The displacement operator on a single mode k is
∗ † ∗
[email protected] D̂k (α) = eαâk −α âk
, (11)
 2

where α is a complex number. It transforms the mode the xk and pk are phase-space variables, and the Wigner
operator as function is a phase-space representation of the state ρ.
All the information contained in ρ is also contained in W
D̂k† (α)âk D̂k (α) = âk + α. (12) and vice versa, i.e. the Wigner function is an equivalent
representation of the quantum state of the system.
When applied to the vacuum state (ground state), it gen- Another equivalent representation of ρ is the charac-
erates the coherent states |αi = D̂(α)|vaci. teristic function of n complex variables α1 , . . . , αn
The joint displacement operator on n modes is
χ(α) = Tr[ρD̂(α)], (18)
D̂(α) = D̂1 (α1 ) ⊗ · · · ⊗ D̂n (αn ), (13)
or equivalently of 2n real variables χ(r) = Tr[ρD̂(r)].
where α = (α1 , . . . , αn ). Using the vector of quadra- The Wigner function can also be expressed as the Fourier
ture operators and the symplectic matrix above, we can transform of the characteristic function
equivalently express the joint displacement operator as Z
a function of 2n real variables, collected in a vector 1
exp −ir T Ωs χ(s)ds.


W (r) = (19)
r = (x1 , p1 , . . . , xn , pn )T , as (2π)2n R2n

D̂(r) = exp ir̂ T Ωr ,

(14)
5. Gaussian states
xk√
+ipk
with the relation αk = 2
to the complex displace-
ments. Gaussian states are those states for which the Wigner
function is a (multivariate) Gaussian function.
Gaussian states are completely determined by the first
3. Covariance matrices and second moments of the quadrature operators. That
is, by the vector of expectation values r̄ = hr̂i and the
Given a quantum state ρ of an n-mode system, the covariance matrix σ. We refer to r̄ as the displacement
corresponding covariance matrix σ has elements vector of the state. In terms of the displacement vector
and covariance matrix, the Wigner function of a Gaussian
Covariance matrix state can be expressed

1 Wigner function (Gaussian state)

σkl = h{r̂k , r̂l }i − hr̂k ihr̂l i (15)
2
1 1 T
σ −1 (r−r̄)
W (r) = e− 2 (r−r̄) (20)
where {·, ·} denotes the anticommutator and hÂi =
p
(2π)n det(σ)
Tr[Âρ] is the expectation value.
The covariance matrix (for valid density matrices ρ) is where det(σ) and σ −1 denote the determinant and in-
real, symmetric, positive definite σ > 0, and fulfills verse of σ, respectively. Note that the Wigner function
i is normalised
σ + Ω ≥ 0. (16)
2
Z
W (r)dr = 1. (21)
Note that the diagonal elements of σ are the variances R2n
of the quadrature operators, while non-zero off-diagonal
elements correspond to correlations between quadratures.
6. Gaussian operations

4. Wigner functions and characteristic functions Gaussian operations are those which take Gaussian
states to Gaussian states, i.e. those which preserve Gaus-
We let x and p denote n-dimensional real vectors. Fur- sianity.
thermore, let |xk i denote an eigenstate of x̂k with eigen- Since Gaussian states are completely determined by
value xk , and |xi = |x1 , x2 , . . . , xn i = |x1 i|x2 i · · · |xn i. r̄ and σ, Gaussian operations are then determined by
The Wigner function corresponding to state ρ is then how they transform r̄ and σ. Unitary Gaussian oper-
defined by ations correspond to those which can be generated by
Hamiltonians which are (at most) quadratic in the mode
operators âk and â†k . These correspond exactly to sym-
Z
1 1 1
W (x, p) = hx + q|ρ|x − qieip·q dq. (17)
(2π)n Rn 2 2 plectic transformations of the displacement vector and
covariance matrix. They map
We may also write W (r) for r = (x1 , p1 , . . . , xn , pn ),
where it is understood that W (r) = W (x, p). Here, σ → F σF T , and r̄ → F r̄ + d, (22)
 3

where d is a 2n-dimensional real vector of displacements, 2. Phase shifts

and the matrix F fulfills
A phase shift on a single mode by an angle φ transforms
F ΩF T = Ω. (23)
the mode operators as
Gaussian unitaries are thus determined by a symplectic
matrix F and a displacement vector d. Every Gaussian â → e−iφ â, ↠→ eiφ ↠. (30)
unitary corresponds to a symplectic transformation and,
conversely, for every symplectic transformation there ex- The quadrature operators hence transform as
ist a quadratic Hamiltonian and corresponding Gaussian
unitary which generates it. x̂ → cos(φ)x̂ + sin(φ)p̂ = x̂(φ), (31)
It follows from the Euler decomposition that every π
p̂ → cos(φ)p̂ − sin(φ)x̂ = x̂(φ + ). (32)
Gaussian unitary on n modes can be realised by a passive, 2
linear transformation followed by single-mode squeezing
in each mode, followed by another passive, linear trans- As a symplectic transformation on a single mode there-
formation. That is, in the language of quantum optics, by fore, the phase shift corresponds to a rotation
combining beam splitters, phase shifts, and single-mode
squeezing. Phase shift

 
III. GAUSSIAN UNITARIES cos(φ) sin(φ)
F = ≡ R(φ), (33)
− sin(φ) cos(φ)
In this section, we list symplectic transformations cor- d = 0. (34)
responding to some commonly encountered Gaussian uni-
tary operations.
When phase shifts φ1 , . . . , φn are applied to each mode
of an n-mode system, we have
1. Displacements n
M
F = R(φk ), (35)
From (12), a displacement on a single mode k trans- k=1
forms the quadrature operators as d = 0. (36)
α+α ∗ √
D̂† (α)x̂D̂(α) = x̂ + √ = x̂k + 2 Re(α), (24)
2
α − α∗ √ 3. Beam splitters
D̂† (α)p̂D̂(α) = p̂ + √ = x̂k + 2 Im(α). (25)
i 2
We consider a beam splitter with transmittivity η,
For an n-mode state and complex displacements α = which transforms the mode operators â1 , â2 of two modes
(α1 , . . . , αn ), it follows that as

σ→σ (the covariance matrix is unchanged), (26) √ p

√ â1 → η â1 + 1 − η â2 , (37)
r̄ → r̄ + 2(Re(α1 ), Im(α1 ), . . . , Re(αn ), Im(αn )T .
p √
â2 → − 1 − η â1 + η â2 . (38)
(27)
That is
As a symplectic transformation therefore, a displacement
defined by an n-dimensional complex vector α corre-    √ √  
â1 √ η 1−η â1
sponds to → √ . (39)
â2 − 1−η η â2

Displacement The quadrature operators then transform as

 √ √
x̂1 η 0 1−η √ 0 x̂1
   
F = 112n , (28) √
 p̂1   √0 η 0 1 − η   p̂1 
x̂  → − 1 − η √ .
Re(α1 ) 0  x̂2 
√0 η
 
2
√
 Im(α1 )  p̂2 0 − 1−η 0 η p̂2
√  . 
(40)
 ..  .
d = 2 (29)

Re(α ) The beam splitter thus does not contribute any displace-
n
ment but it does mix the mode operators. As a symplec-
Im(αn )
tic transformation on two modes, we have
 4

Beam splitter 5. Two-mode squeezing

 √ √  The two-mode squeezing operator acting on modes 1,2

√ η 112 1 − η 112 is
F = √ ≡ Fη , (41)
− 1 − η 112 η 112 ∗
â1 â2 −ξâ†1 â†2
Ŝ2 (ξ) = eξ , (55)
d = 0. (42)
iθ
with ξ = re . It transforms the mode operators accord-
When acting on a pair of modes (say k, l) of an n- ing to
mode system, the symplectic matrix should be equal to
Ŝ2 (ξ)† â1 Ŝ2 (ξ) = cosh(r)â1 − eiθ sinh(r)â†2 , (56)
Fη in the kl-subspace and identity on the rest of the
space. Let P be a permutation matrix shifting modes k Ŝ2 (ξ)† â†1 Ŝ2 (ξ) = cosh(r)â†1 −e −iθ
sinh(r)â2 , (57)
and l to the first two modes, i.e. P (x1 , p1 , . . . , xn , pn )T =
Ŝ2 (ξ)† â2 Ŝ2 (ξ) = cosh(r)â2 − e sinh(r)â†1 ,
iθ
(58)
(xk , pk , xl , pl , . . .)T . Then
  Ŝ2 (ξ)† â†2 Ŝ2 (ξ) = cosh(r)â†2 − e−iθ sinh(r)â1 . (59)
−1 Fη 0
F =P P, (43) Setting cr = cosh(r), sr = sinh(r), the quadratures then
0 112(n−2)
transform as
d = 0. (44)
Ŝ2 (ξ)† x̂1 Ŝ2 (ξ) = cr x̂1 − cos(θ)sr x̂2 − sin(θ)sr p̂2 , (60)

4. Single-mode squeezing Ŝ2 (ξ)† p̂1 Ŝ2 (ξ) = cr p̂1 + cos(θ)sr p̂2 − sin(θ)sr x̂2 , (61)
†
Ŝ2 (ξ) x̂2 Ŝ2 (ξ) = cr x̂2 − cos(θ)sr x̂1 − sin(θ)sr p̂1 , (62)
The single-mode squeezing operator is Ŝ2 (ξ)† p̂2 Ŝ2 (ξ) = cr p̂2 + cos(θ)sr p̂1 − sin(θ)sr x̂1 . (63)
1 ∗ 2 † 2
Ŝ(ξ) = e 2 (ξ â −ξ(â ) ) , (45) Note that for θ = 0, this squeezes both x̂1 + x̂2 and p̂1 − p̂2
iθ since x̂1 + x̂2 → e−r (x̂1 + x̂2 ) and p̂1 − p̂2 → e−r (p̂1 + p̂2 ).
with ξ = re . The mode operators then transform as
As a symplectic transformation on two modes, we
Ŝ(ξ)† âŜ(ξ) = cosh(r)â − eiθ sinh(r)↠, (46) have
† † † −iθ
Ŝ(ξ) â Ŝ(ξ) = cosh(r)â − e sinh(r)â. (47)
Two-mode squeezing
Hence, setting cr = cosh(r), sr = sinh(r), the quadrature
operators transform as  
cosh(r)112 − sinh(r)Sθ
Ŝ(ξ)† x̂Ŝ(ξ) = [cr − cos(θ)sr ]x̂ − sin(θ)sr p̂, (48) F = ≡ F2,ξ (64)
− sinh(r)Sθ cosh(r)112
Ŝ(ξ)† p̂Ŝ(ξ) = [cr + cos(θ)sr ]p̂ − sin(θ)sr x̂. (49) d = 0, (65)
Note that this squeezes x̂( θ2 ) and anti-squeezes x̂( θ2 + π2 ). where again
For example, for θ = 0 we get x̂ → e−r x̂ and p̂ → er p̂.  
As a symplectic transformation on a single mode, we cos(θ) sin(θ)
Sθ = . (66)
have sin(θ) − cos(θ)

Single-mode squeezing When acting on a pair of modes (say k, l) of an n-

mode system, the symplectic matrix should be equal to
FBS in the kl-subspace and identity on the rest of the
F = cosh(r)112 − sinh(r)Sθ ≡ Fξ , (50) space. Let P be a permutation matrix shifting modes k
d = 0, (51) and l to the first two modes, i.e. P (x1 , p1 , . . . , xn , pn )T =
(xk , pk , xl , pl , . . .)T . Then
where  
−1 F2,ξ 0
  F =P P, (67)
cos(θ) sin(θ) 0 112(n−2)
Sθ = . (52)
sin(θ) − cos(θ) d = 0. (68)

When single-mode squeezing with parameters

ξ1 , . . . , ξn are applied to each mode of an n-mode IV. OTHER GAUSSIAN OPERATIONS
system, we have
n
M 1. Tracing out
F = Fξk , (53)
k=1 Consider a bidivision of an n-mode system into sub-
d = 0. (54) systems A and B and order the modes such that r̂ =
 5

(r̂A , r̂b )T , where r̂A and r̂B are vectors of quadratures We would like to express the conditional state of A upon
for A and B, respectively. obtaining a particular measurement outcome as well as
For any state ρ of the joint system and any operator the probability distribution over outcomes in terms of
ÔA acting only in A, we have elements of σ and r̄.
The covariance matrix of A after the measurement is
hÔA iρ = Tr[(ÔA ⊗ 11B )ρ] = TrA [ÔA ρA ], (69) independent of the measurement outcome and is given by
where ρA is the reduced state of A. Similarly for B. It
σA = A − C(ΠBΠ)−1 C T , (74)
follows that the covariance matrix of the joint system
can be written
  where
A C
σ= , (70)
 
CT B 1 0
Π= , (75)
0 0
where A and B are the covariance matrices correspond-
ing to ρA and ρB , respectively, and C encodes correla- and (ΠBΠ)−1 denotes the (Moore-Penrose) pseudoin-
tions between A and B. The displacement vector of the verse. Specifically, (ΠBΠ)−1 = B11 Π, with B11 the
joint system is top-left entry of B. Hence,
   
hr̂A iρA r̄A
r̄ = hr̂iρ = = . (71) x̂-measurement – post-meas. covariance
hr̂B iρB r̄B
When the state of the joint system is Gaussian, the 1
reduced states ρA and ρB are also Gaussian and are sim- σA = A − CΠC T (76)
B11
ply defined by σA = A and r̄A and by σB = B and r̄B ,
respectively.
The displacement vector of A after the measurement
Tracing out does depend on the outcome. We let u denote the mea-
surement outcome and define the vector
For a joint system AB in a Gaussian state ρAB with  
displacement vector r̄ = (r̄A , r̄b )T and covariance u
u= . (77)
matrix 0
 
A C The displacement vector of the conditional state is then
σ= , (72)
CT B
r̄A = a − C(ΠBΠ)−1 (b − u), (78)
the reduced states ρA , ρB of subsystems A and B, re-
spectively, are also Gaussian with displacement vec- or equivalently
tors r̄A and r̄B and covariance matrices A and B.
x̂-measurement – post-meas. displacement
Starting from an arbitrary ordering of modes, σ and
r̄ can always be brought on the form above by applying
1
an appropriate permutation reordering the modes with r̄A = a − CΠ(b − u) (79)
those of A first followed by those of B. B11

The probability distribution over outcomes only de-

2. Homodyne measurements pends on the reduced state of the subsystem being mea-
sured. In the case of a single mode considered here
Ideal homodyne measurements correspond to projec- that means it depends only on B and b. For an x̂-
tive measurements of quadrature operators. measurement, the distribution is a Gaussian with vari-
ance B11 centered at b1 , i.e.
Measurements of x̂
x̂-measurement – outcome distribution
We consider an ideal homodyne measurement of x̂n on
the last mode of an n-mode system in a Gaussian state.
1 1 2
Denoting the measured mode subsystem B and the first Px̂ (u) = √ e− 2B11 (u−b1 ) (80)
n − 1 modes A, we can write the covariance matrix and 2πB11
displacement vector of the state
    which is also the marginal of the Wigner function for
A C a the reduced state of B.
σ= , r̄ = . (73)
CT B b
 6

Measurements in other directions Gaussian integral - n dimensions

Note that, while we have given the expressions here for an For A symmetric, real, positive definite and b ∈ Rn
x̂-measurement, a measurement of an arbitrary quadra- s
(2π)n 1 bT A−1 b
Z
ture x̂(φ) can be modelled as first applying a phase shift − 12 r T Ar+bT r
e dr = e2 . (90)
of −φ, rotating x̂(φ) to x̂(0) = x̂, and then measuring Rn det(A)
x̂. Hence, in the above expressions, one only needs to
replace In particular, with σ a covariance matrix
Z
B → R(−φ)BRT (−φ), (81) 1 T −1 p
e− 2 (r−r̄) σ (r−r̄) dr = (2π)2n det(σ), (91)
T R2n
C → CR (−φ), (82)
b → R(−φ)b, (83) from which it can be seen that the Wigner function (20)
of a Gaussian state is normalised.
where
2. Trace rule in phase space
 
cos(φ) − sin(φ)
R(−φ) = . (84)
sin(φ) cos(φ)
The trace of the product of two states is the integral
In particular, for a measurement of p̂ = x̂( π2 ) with out- of the product of their Wigner functions
come u, we have Z
n
Tr[ρ1 ρ2 ] = (2π) W1 (r)W2 (r)dr. (92)
1 R2n
σA = A − CΠ0 C T , (85)
B22 A similar expression holds for the characteristic func-
tions. For Gaussian states, this leads to the following
and expression.
1
r̄A = a − CΠ0 (b + u0 ), (86) Overlap of Gaussian states
B22
For ρ1 , ρ2 Gaussian states with covariance matrices
where
σ1 , σ2 and displacement vectors r1 , r2
   
0 0 0
Π0 = , u0 = . (87) 1 1 T
Σ−1 (r1 −r2 )
0 1 u Tr[ρ1 ρ2 ] = p e− 2 (r1 −r2 ) , (93)
det(Σ)
The distribution over outcomes is obtained from Eq. (80)
where Σ = σ1 + σ2 .
by replacing B11 → B22 and b1 → b2 .

V. GAUSSIAN INTEGRALS AND FIDELITY 3. Fidelity

1. Integrals of multivariate Gaussian functions For two arbitrary mixed states ρ1 and ρ2 of a given
system, we define the fidelity as (note that the fidelity is
sometimes defined as the square of this expression)
Here, we list some useful expressions for integrals of q 
Gaussian functions. √ √
F (ρ1 , ρ2 ) = Tr ρ1 ρ2 ρ1 . (94)
Gaussian integral - one dimension
When both states are Gaussian, following Ref. [7] the
For a, b, c ∈ R and a > 0 fidelity can be expressed in terms of their covariance ma-
Z ∞ r trices σ1 , σ2 and displacement vectors r̄1 , r̄2 as follows
2 π b2 +c
e−ax +bx+c dx = e 4a . (88) 1
F (ρ1 , ρ2 ) = F0 (ρ1 , ρ2 )e− 4 (r̄2 −r̄1 )
T
(σ1 +σ2 )−1 (r̄2 −r̄1 )
(95)
−∞ a
where the factor F0 (ρ1 , ρ2 ) (the fidelity for equal displace-
In particular, for σ > 0 and x̄ ∈ R ments) can be expressed
Z ∞
  q  1/4
√ 1 −2
 det 2( 11 + 4 (V Ω) + 11)V 
1 2
e− 2σ (x−x̄) dx = 2πσ. (89) F0 (ρ1 , ρ2 ) =   ,
−∞ det(σ1 + σ2 )
In higher dimensions we have (96)
 7

with composite system ρAB = ρA L ⊗ ρB , the covariance matrix

1 is the direct sum σAB = σA σB and the displacement
V = ΩT (σ1 + σ2 )−1 ( Ω + σ2 Ωσ1 ). (97) T
vector is the concatenation r̄AB T
= (r̄A T
, r̄B ).
4
Or, equivalently
1. Vacuum state
√
!1/4
det ( 11 − W −2 + 11)W iΩ
F0 (ρ1 , ρ2 ) = , (98)
det(σ1 + σ2 ) For a single-mode system, the vacuum state |vaci is
the ground state of the quantum harmonic oscillator cor-
with responding, in quantum optics, to a zero-photon Fock
state. For a single mode, we thus have
W = −2V iΩ. (99)
In some important cases, these general expressions Vacuum state
simplify significantly.
1 
2 0
One state pure σ= 1 , (105)
0 2

When (at least) one of the states is pure, the above ex- r̄ = 0. (106)
pressions reduce to
For multimode systems, |vaci is also used to refer to the
Fidelity with pure state joint ground state, i.e. a product of zero-photon states
for each mode.

1 1 T −1
F (ρ1 , ρ2 ) = p
4
e− 4 (r̄2 −r̄1 ) Σ (r̄2 −r̄1 ) , 2. Coherent state
det(Σ)
(100)
where Σ = σ1 + σ2 . Coherent states |αi, α ∈ C are displaced vacuum
p states, i.e. |αi = D̂(α)|vaci. The displacement operation
That is, F0 (ρ1 , ρ2 ) = 1/ 4 det(σ1 + σ2 ) in this case. preserves the variances while changing the displacement
vector. Thus
Single mode
Coherent state
When there is only a single mode (but both states may
be mixed) we have 1 
2 0
σ= , (107)
0 12
Fidelity for single-mode mixed states √ Re(α)
 
r̄ = 2 . (108)
Im(α)
1 1 T −1
F (ρ1 , ρ2 ) = √ e− 4 (r̄2 −r̄1 ) (σ1 +σ2 ) (r̄2 −r̄1 ) ,
Γ
(101)
3. Thermal state
where
√ √
Γ = ∆ + Λ − Λ, (102) A canonical thermal state of a single bosonic mode
∆ = det(σ1 + σ2 ), (103) with Hamiltonian Ĥ = ω↠â+ 21 and inverse temperature
β = 1/T is given by (setting kB = 1)
1 1
Λ = 4(det(σ1 ) − )(det(σ2 ) − ). (104) †
4 4 e−βωâ â
τ (β) = . (109)
Tr[e−βω↠â ]
The mean number of excitations (mean number of pho-
VI. GAUSSIAN STATES
tons) n̄ and the temperature are related by
n̄
In this section, we list the covariance matrices and dis- e−βω = . (110)
placement vectors for some commonly encountered Gaus- 1 + n̄
sian states. In terms of the mean photon number, in the Fock basis
We only give the expressions for a single or two modes ∞  n
below. They can easily be extended to multimode sys- 1 X n̄
τ (β) = |nihn|. (111)
tems by exploiting the fact that for a product state of a 1 + n̄ n=0 1 + n̄
 8

The covariance matrix and displacement vector are 5. Two-mode squeezed vacuum state

Thermal state
Two-mode squeezed vacuum is obtained by applying
the two-mode squeezing operator to the joint vacuum
1
 
n̄ + 2 0 state of a two-mode system, i.e. |S2 (ξ)i = Ŝ2 (ξ)|vaci
σ= 1 , (112)
0 n̄ + 2 with ξ = reiθ . Using the symplectic transformation for
r̄ = 0. (113) two-mode squeezing one finds
Two-mode squeezed vacuum
Note that the zero-temperature thermal state is the vac-
uum state.  1
cosh(2r)112 − 12 sinh(2r)Sθ

σ= 2 , (119)
− 12 sinh(2r)Sθ 21 cosh(2r)112
4. Single-mode squeezed vacuum state r̄ = 0, (120)

where
Single-mode squeezed vacuum is obtained by apply-
ing the single-mode squeezing operator to the vacuum,
 
cos(θ) sin(θ)
i.e. |S(ξ)i = Ŝ(ξ)|vaci. Using the symplectic transfor- Sθ = . (121)
sin(θ) − cos(θ)
mation for single-mode squeezing with ξ = reiθ ,

Single-mode squeezed vacuum

In particular, for θ = 0 we have
1 1
σ= cosh(2r)112 − sinh(2r)Sθ , (114)
2 2 
1 0

r̄ = 0, (115) Sθ = , (122)
0 −1
where
and Var(x̂1 + x̂2 ) = Var(p̂1 − p̂2 ) = e−2r .
 
cos(θ) sin(θ)
Sθ = . (116)
sin(θ) − cos(θ)

In particular, for squeezing along x̂ (θ = 0) we get

 1 −2r 
e 0
σ= 2 1 2r , (117) VII. COMMENTS AND SUGGESTIONS
0 2e

and for squeezing along p̂ (θ = π) we have Any comments and corrections are welcome, as are sug-
 1 2r  gestions for further relevant results and expressions which
e 0
σ= 2 . (118) might be included.
0 12 e−2r

[1] Samuel L. Braunstein and Peter van Loock, “Quantum Astronomy, Aarhus University (2018).
information with continuous variables,” Rev. Mod. Phys. [5] Christopher Gerry and Peter Knight, Introductory Quan-
77, 513–577 (2005). tum Optics (Cambridge University Press, Cambridge, UK,
[2] Christian Weedbrook, Stefano Pirandola, Raúl García- 2005).
Patrón, Nicolas J. Cerf, Timothy C. Ralph, Jeffrey H. [6] Alessio Serafini, Quantum Continuous Variables - A
Shapiro, and Seth Lloyd, “Gaussian quantum informa- Primer of Theoretical Methods (CRC Press, 2017).
tion,” Rev. Mod. Phys. 84, 621–669 (2012). [7] Leonardo Banchi, Samuel L. Braunstein, and Stefano Pi-
[3] Alessandro Ferraro, Stefano Olivares, and Matteo G. A. randola, “Quantum fidelity for arbitrary Gaussian states,”
Paris, Lecture Notes: Gaussian states in continuous vari- Phys. Rev. Lett. 115, 260501 (2015).
able quantum information (Bibliopolis, Napoli, 2005).
[4] Jinglei Zhang, Quantum measurement and preparation of
Gaussian states, Ph.D. thesis, Department of Physics and