Open Quantum Systems

Roberto S. Vitancol

Theoretical Physics Group, National Institute of Physics,

University of the Philippines, Diliman Quezon City, 1101 Philippines
(Dated: August 29, 2013)
PACS numbers: 03.65.Db
INTRODUCTION
Quantum mechanical systems are open quantum sys-
tems. Strictly speaking, the only closed system is the
entire universe and any realistic system interacts with its
environment.
In contrast to the case of a closed systems, the quan-
tum dynamics of an open system cannot, in general, be
represented in terms of a unitary time evolution. In many
cases it turns out to be useful to formulate, instead, the
dynamics of an open system by means of an appropri-
ate equation of motion for its density matrix, a quantum
master equation.
MARKOVIAN MASTER EQUATION:
MICROSCOPIC DERIVATIONS
A system whose unpurturbed Hamltonian is H
S
inter-
acts with the environment whose Hamitonian is H
E
so
that the total Hamiltonian of the system and environ-
ment is
H
tot
= H
S
I
E
+ I
S
H
E
+ H
int
(1)
where H
int
is the Hamiltonian of the interaction between
the system and the environment.
The reduced density matrix of the system, (t), is ob-
tained by taking the partial trace over the environment
of the total density matrix of the system and environ-
ment:
(t) = tr
E
_

tot
(t)
_
(2)
Assume that at t=0, the system and the environment are
uncorrelated:

tot
(t) = (0)
E
(3)
where
E
is the initial state of the environment.
Assume that the environment is a large resevoir that it
is unaected by the system. With this assumption, the
total density matrix remains factorized in time. This is
known as the Born Approximation:

tot
(t) (t)
E
(4)
such that
E
is a stationary state of the environment,
[H
E
,
E
] = 0 (5)
In interaction representation,

t

tot
=
1
i
_

H
int
,
tot
(t)
_
(6)
Integrating equation (6),

tot
(t) =
tot
(0) +
1
i
_
t
0
ds
_

H
int
(s),
tot
(s)
_
(7)
and substituting (7) back into (6), we have,

t

tot
=
1
i
_

H
int
(t),
tot
(0)
_
+
1
(i)
2
_
t
0
ds
_

H
int
(t),
_

H
int
(s),
tot
(s)
__
(8)
To obtain the evolution of the system in interaction rep-
resentation, we trace-out the environment in (8) knowing
that equation-(2) is preserved in the interaction represen-
tation: (t) = tr
E
_

tot
(t)
_
and that the trace commutes
with
t
: tr
E
(
t

tot
) =
t
tr
E
(
tot
). Hence,

t
=
1
i
tr
E
_
_

H
int
(t),
tot
(0)
_
_
+
1
(i)
2
_
t
0
ds tr
E
_
_

H
int
(t),
_

H
int
(s),
tot
(s)
__
_
(9)
To evaluate (9), we must rst calculate
tr
E
_
_

H
int
(t),
tot
(0)
_
_
= tr
E
_
_

H
int
(t), (0)
E
_
_
(10)

H
int
(t) can be written in the form

H
int
(t) =

A
k
(t)

B
k
(t) (11)
where

A
k
(t) =

A

k
(t) are eigenoperators of the system
Hamiltonian and

B
k
(t) =

B

k
(t) are eigenoperators of
the environment Hamiltonian. Thus,
tr
E
_
_

H
int
(t),
tot
(0)
_
_
=

k
tr
E
_
_

A
k
(t)

B
k
(t)
__
(0)
E
_

_
(0)
E
__

A
k
(t)

B
k
(t)
_
_
=

k
tr
E
_

A
k
(t) (0)

B
k
(t)
E
(0)

A
k
(t)
E

B
k
(t)
_
=

k
_

A
k
(t) (0)tr
E
_

B
k
(t)
E
_
(0)

A
k
(t)tr
E
_

E

B
k
(t)
_
_
=

k
tr
E
_

E

B
k
(t)
__

A
k
(t) (0) (0)

A
k
(t)
_
=

k
_

B
k
(t)
_
E
_

A
k
(t), (0)
_
(12)
For the second term of (9) we compute
_

H
int
(t),
_

H
int
(s), (s)
E
__
=

k,l
_

A
k
(t)

B
k
(t),
_

A
l
(s)

B
l
(s), (s)
E
__
=

k,l
_

A
k
(t)

B
k
(t),
_

A
l
(s)

B
l
(s)
__
(s)
E
_

_
(s)
E
__

A
l
(s)

B
l
(s)
_
=

k,l
_
_

A
k
(t)

A
l
(s) (s)
_

B
k
(t)

B
l
(s)
E
_

_

A
k
(t) (s)

A
l
(s)
_

B
k
(t)
E

B
l
(s)
_
+
_
(s)

A
l
(s)

A
k
(t)
_

E

B
l
(s)

B
k
(t)
_

_

A
l
(s) (s)

A
k
(t)
_

B
l
(s)
E

B
k
(t)
_
_
(13)
tr
E
_
_

H
int
(t),
_

H
int
(s), (s)
E
__
_
=

k,l
_
tr
E
_

E

B
k
(t)

B
l
(s)
_

_

A
k
(t)

A
l
(s) (s)

A
l
(s) (s)

A
k
(t)
_
+ Hermitian conjugate terms
_
(14)
Let
C
kl
(t, s) = tr
E
_

E

B
k
(t)

B
l
(s)
_
= tr
E
_

E
e
iHEt/
B
k
e
iHEt/
e
iHEs/
B
l
e
iHEs/
_ (15)
Using the cyclic property of the trace and the commuta-
tion of
E
with H
E
in (5),
C
kl
(t, s) = tr
E
_

E
e
iHE(ts)/
B
k
e
iHE(ts)/
B
l
_
(16)
Then
C
kl
(t, s) = C
kl
(t s) (17)
C
kl
() = tr
E
_

E
e
iHE/
B
k
e
iHE/
B
l
_
(18)
Combining (17) and (15) into (14) and substituting this
along with (12) into (9),

t
=
1
i

k
_

B
k
(t)
_
E
_

A
k
(t), (0)
_
+
1
(i)
2

k,l
__
t
0
ds C
kl
(t s)
_

A
k
(t)

A
l
(s) (s)

A
l
(s) (s)

A
k
(t) + h.c.
_
_
(19)
The environment Hamiltonian can always be reformu-
lated such that the environmental energy
_

B
k
(t)
_
E
= 0
for all k.
The function C
kl
() is determined by the environmen-
tal state alone and is appreciable only for a short range
about = 0. The remaining second term of (19) can be
written in super-operator form:

t
(t) =
_
t
0
ds /(t s) (s) (20)
where
/(t s) (s) =
1
(i)
2

k,l
_
C
kl
(t s)
_

A
k
(t)

A
l
(s) (s)

A
l
(s) (s)

A
k
(t)
_
+ h.c.
_
(21)
In order to simplify (20) further we perform the Markov
approximation: Since the integrand (21) is appreciable
only when t s, we can replace (s) by (t) in the inte-
gral,

t
(t) =
_
t
0
ds /(t s) (t) (22)
where we have obtained an equation of motion for the
reduced systems density matrix in which the time devel-
opment of the state of the system at time t only depends
on the present state (t). This is known as the Redeld
equation.
The Redeld equation is local in time, but it is not yet
a Markovian master equation because the time evolution
of the reduced density matrix still depends upon an ex-
plicit choice for the initial preparation at time t = 0. But
since the kernel /(t s) is appreciable only at the origin,
= t s = 0, we can replace the limit of integration by
innity:
_
t
0
ds /(t s) =
_
t
0
d /()
=
_

0
d /()
(23)
The Markov approximation is justied if the time scale
over which the state of the system varies appreciably is
large compared to the time scale over which the reser-
voir correlation functions C
kl
() would decay. Thus, we
obtain a Markovian quantum master equation

(t) =
_

0
d /() (t) (24)
Let H
S
[E = E[E where the eigenvalues E are non-
degenerate and eigenvectors [E form a basis for the sys-
tem Hilbert space. Then we can dene eigenoperators of
H
S
for frequncies :
A
k
() =

E=
E[A
k
[E

[EE

[
= A

k
()
(25)
where

A
k
() =

E=
[EE[A
k
[E

[
=

,E
[EE[A
k
[E

[
= A
k
(26)
Then

A
k
(t) = e
iHSt/
A
k
e
iHSt/
=

e
iHSt/
A
k
()e
iHSt/
=

e
iHSt/

E=
E[A
k
[E

[EE

[e
iHSt/
=

E=
E[A
k
[E

e
iHSt/
[EE

[e
iHSt/
=

E=
E[A
k
[E

[EE

[e
i(E

E)t/
=

e
it

E=
E[A
k
[E

[EE

A
k
(t) =

e
it
A
k
()
=

e
it
A

k
()
(27)
Then (22) becomes

t
=
1
(i)
2

k,l
__
t
0
ds C
kl
(t s)
_

e
it
A

k
()e
i

s
A
l
(

) (t)

e
i

s
A
l
(

) (t)e
it
A

k
()
+ Hermitian conjugates
_
_

t
=
1
(i)
2

k,l,

__
t
0
ds C
kl
(t s)e
i

s+it
_
A

k
()A
l
(

) (t) A
l
(

) (t)A

k
()
+ Hermitian conjugates
_
_
(28)
If we change varibles to = t s or s = t and
change the limits of integration to innity,
_
t
0
ds C
kl
(t s)e
i

s+it
=
_
0
t
(1)d C
kl
()e
i

(t)+it
= e
i(

)t
_
t
0
d C
kl
()e
i

e
i(

)t
_

0
d C
kl
()e
i

(29)
Dening the Fourier transform of the resevoir correla-
tion functions C
kl
() as

kl
() =
1

2
_

0
d e
i
C
kl
() (30)
Equation (28) then becomes

t
=

k,l,

e
i(

)t

kl
(

)
_
A
l
(

) (t)A

k
() A

k
()A
l
(

) (t)
_
+ h.c.
(31)
For times t which are large compared to the time scale
given by the smallest enegy spacing, it is reasonable to ex-
pect that only equal pairs of frequencies

= will con-
tribute appreciably to the sum and the terms for which

,= may be neglected since they oscillate very rapidly

compared to the relaxation time of the system density
matrix. In quantum optical systems, this is known as
the rotating wave approximation(RWA). Then

t
=

k,l,

kl
()
_
A
l
() (t)A

k
()
A

k
()A
l
() (t)
_
+ h.c. (32)
Separating
kl
into its real and imaginary components,

kl
() =
1
2

kl
() + i S
kl
() (33)
where

kl
() =
kl
() +

lk
() =
_

d e
i
C
kl
() (34)
S
kl
() =
1
2i
_

kl
()

lk
()
_
(35)
Then

t
=

k,l,

_
1
2

kl
() + i S
kl
()
_
_
A
l
() (t)A

k
() A

k
()A
l
() (t)
_
+

k,l,

_
1
2

kl
() i S
kl
()
_
_
A
k
() (t)A

l
() (t)A

l
()A
k
()
_
=

k,l,

1
2

kl
()
_
A
l
() A

k
() A

k
()A
l
()
+ A
k
() A

l
() A

l
()A
k
()
_
+ i

k,l,

S
kl
()
_
A
l
() A

k
() A

k
()A
l
()
A
k
() A

l
() + A

l
()A
k
()
_
(36)
k and l are just dummy indices so we can interchange k
and l in the summation of the Hermitian conjugate terms
resulting to

t
=
1
i

k,l,

S
kl
()
_
A

k
()A
l
() A

k
()A
l
()
_
+

k,l,

kl
()
_
A
l
() A

k
()
1
2
A

k
()A
l
()

1
2
A

k
()A
l
()
_

t
=
1
i
_
H
LS
, (t)

+T
_
(t)
_
(37)
where
H
LS
=

k,l,

S
kl
()A

k
()A
l
() (38)
is known as the Lamb shift Hamiltonian and
T
_

_
=

k,l,

kl
()
_
A
l
() A

k
()
1
2
_
A

k
()A
l
(),

_
(39)
is the dissipator of the master equation.
The density matrix in the interaction representation is
related to the density matrix in Schrodinger representa-
tion by
(t) = e
iHSt/
(t)e
iHSt/
(40)
Dierentiating,

t
(t) =
i

H
S
e
iHSt/
(t)e
iHSt/
+ e
iHSt/

t
(t)e
iHSt/

e
iHSt/
(t)e
iHSt/
H
S

t
(t) =
1
i
e
iHSt/
_
H
S
, (t)

e
iHSt/
+ e
iHSt/

t
(t)e
iHSt/
(41)
From (25) one can derive the follwing commutation prop-
erties of A
k
():
_
H
S
, A
k
()

= A
k
() (42)
_
H
S
, A

k
()

= +A

k
() (43)
_
H
S
, A

k
(), A
l
()

=
_
H
S
, H
LS

= 0 (44)
With (40) and these commutation relations, we can ex-
press (37) as

t
(t) = e
iHSt/
1
i
_
H
LS
, (t)

e
iHSt/
+ e
iHSt/
T
_
(t)
_
e
iHSt/
(45)
After equating (45) and (41), we nally obtain the
Markovian master equation in Schrodinger representa-
tion:

t
(t) =
1
i
_
H
S
+ H
LS
, (t)

k,l,

kl
()
_
A
l
()A

k
()
1
2
A

k
()A
l
()

1
2
A

k
()A
l
()
_
(46)
Interaction of the system with the environment leads to
a shift of the unpurturbed energy levels of the system
(H
S
H
S
+ H
LS
) induced by the exchange of energy
with the environment.
NON-MARKOVIAN QUANTUM DYNAMICS
Strong environment coupling, correlation and entan-
glement at initial state, and a nite resrvoir can lead to
long memory times and to a failure of the Markovian
approximation. To deal with non-Markovian quantum
dynamics, we can use projection operator techniques.
Projection Operator Techniques
The basic idea of projection operator techniques is to
reagrd the tracing over of the environment as a formal
projection T in the state space of the total system.
The superoperator T has the property of a projection
operator, T
2
= T and the density matrix T is the rele-
vant part of the density matrix of the total system.
Correspondingly, one denes a projection Q
onto the irrelevant part the irrelevant part Q where
T + Q = J, the identity map. The aim is to derive a
closed equation of motion for the relevant part T.
In the case of the Nakajima-Zwanzig method this
is an integro-dierential equation involving a retarded
time integration over the history of the reduced system.
The time-convolutionless projection operator technique
(TCL) leads to a time-local expansion of the equation
of motion with respect to the strength of the system
and environment coupling. The convolution in integro-
dierential models is usually an undesirable characteris-
tic because, even when the equation can be formulated,
the methods to nd its solution are complicated. The
idea of the time-convolutionless (TCL) forms consist of
removing this convolution from the evolution equation to
end up with an ordinary dierential equation.
Projection Operators
In order to derive an an exact equation of motion for
the reduced density matrix
S
= tr
E
of the open system,
we dene the super-operator T according to
T = tr
E

E
=
S

E
(47)
Dening the corresponding projection super-operator Q
Q = T (48)
Projection operators have the following properties:
T
2
= T (49)
Q
2
= Q (50)
T +Q = J (51)
TQ = QT = 0 (52)
Nakajima-Zwanzig Projection Operator Technique
Consider an open system whose total Hamiltonian is
H = H
0
+ H
I
(53)
where H
0
generates the uncoupled time evolution of the
system and environment, H
I
is the interaction Hamilto-
nian, and is a dimensionless expansion parameter.
In the interaction representation

I
(t)
t
= i
_
H
I
,
I
(t)

/(t)
I
(t) (54)
where = 1 and /(t) is the Liouville super-operator.
Applying the projection super-operators,

t
T
I
(t) = T

I
(t)
t
= T/(t)
I
(t) (55)

t
Q
I
(t) = Q

I
(t)
t
= Q/(t)
I
(t) (56)
Upon inserting the identity J = T +Q,

t
T
I
(t) = T/(t)T
I
(t) + T/(t)Q
I
(t) (57)

t
Q
I
(t) = Q/(t)T
I
(t) + Q/(t)Q
I
(t) (58)
To obtain a closed equation of motion for the relevant
part of the density matrix, we solve (58) and insert the
solution to (57).
The solution of (58) may be expressed as
Q
I
(t) = G(t, t
0
)Q
I
(t
0
) +
_
t
t0
ds G(t, s)Q/(s)T
I
(s)
(59)
where
G(t, s) T

exp
_

_
t
t0
ds

Q/(s

(60)
Inserting (59) into (57)

t
T
I
(t) = T/(t)G(t, t
0
)Q
I
(t
0
) + T/(t)T
I
(t)
+
2
_
t
t0
dsT/(t)G(t, s)Q/(s)T
I
(s)
(61)
This equation is known as the Nakajima-Zwanzig equa-
tion. It is an exact equation for the relevant degrees of
freedom of the reduced system. The rst term is an inho-
mogenous term which depends upon the initial condition
at time t
0
. The integral is taken over the past history of
the system in the time interval [t
0
, t]. Thus it describes
comletely non-Markovian eects of the reduced density
matrix dynamics.

rvitancol@yahoo.com
[1] H.P. Breuer, F. Petruccione The Theory of Open Quantum
Systems, Oxford University Press, New York (2002)