F. Fagnola, Quantum Markov Semigroups and Quantum Flows

QUANTUM MARKOV SEMIGROUPS AND
QUANTUM FLOWS
Franco Fagnola
Appeared in: Proyecciones 18, n.3 (1999) 1144.
Preface
A.N. Kolmogorov and J. von Neumann proposed in the 30s two sets of axioms for the
mathematical modelling of random phenomena. In the classical (Kolmogorov) models one
introduces a triple (, F, IP ), where is the set of all possible results of the random phenomenon, F is a -algebra of subsets of called events, IP a probability measure on F.
In the quantum (von Neumann) one starts with a pair (A, ) where A is a von Neumann
algebra (of operators) whose projections are called events and is a state (probability) on
A. The original scheme of von Neumann, however, covered only some fundamental ideas of
what could be called a non-commutative measure theory, while more probabilistic aspects
such as the notions of random variable and stochastic process, probability and conditional
expectation, Markov chains and Markov processes, statistical independence, were absent
from this scheme.
Starting from the second half of the 70s this program was developed systematically.
The notions of random variable and stochastic process were stated in a purely algebraic way; a non-trivial notion of quantum conditional expectation allowing the construction of the first examples of quantum Markov processes was introduced by L. Accardi and
extended to arbitrary von Neumann algebras by L. Accardi and C. Cecchini in [3]; the study
of limit theorems, law of large numbers and central limit theorem, was started by W. von
Waldenfels.
At the beginning of the 80s the notion of quantum stochastic process was formulated
in the form nowadays generally accepted by L. Accardi, A. Frigerio and J.T. Lewis (see [5]).
A quantum stochastic process on a *-algebra A with values in a *-algebra B is a family j =
(jt)t0 of *-homomorphism jt : B A. This generalises the notion of the flow of a classical
stochastic process (X(t))t0 on a probability space (, F, IP ) with values in a measurable
space (E, E). Indeed a classical process can be described through the homomorphisms
jt : L(E, E) L (, F, IP ),
jt (f) = f(X(t));
for this reason quantum stochastic processes are also called quantum flows.
Quantum stochastic processes also generalise evolutions of a closed quantum system
represented by a Hilbert space h with Hamiltonian H. In this case a random variable X
is a self-adjoint operator and the evolution of X is given by the homomorphisms
jt (X) = U (t)XU (t)
jt : B(h) B(h),
where B(h) is the von Neumann algebra of all bounded operators in h and (U (t))t0 is
the unitary group generated by H. This evolution is considered reversible because j is a
one-parameter group.
The developement, almost at the same time, of the quantum stochastic calculus by
R.L. Hudson and K.R. Parthasarathy [58] gave a strong impulse to the study of quantum
stochastic processes (especially quantum Markov processes in the sense of [1]) both in the
direction of applications to physics, especially to quantum optics, and to the theory of
I
II
continual measurements (A. Barchielli, E.B. Davies, G. Lupieri), and in the applications to
classical probability (P.A. Meyer [69], K.R. Parthasarathy [74]).
The homomorphisms of the most studied class of quantum Markov processes are of the
form
jt (X) = V (t)XV (t)
(1)
jt : B(h) B(H),
where H is the tensor product of h and a Fock space representing the external noise, and
(V (t))t0 is a family of unitary operators on H satisfying a quantum stochastic differential
equation of Hudson-Parthasarathy type

dV (t) = V (t) L1 dA+ (t) + L2 d(t) + L3dA(t) + L4 dt , V (0) = 1l.
(2)
The noises A, A+ , are the annihilation, creation and numbers process of Boson Fock
quantum stochastic calculus (we wrote a quantum stochastic differential equation only for
one-dimensional noise for simplicity), L1 , . . . , L4 are operators in the system space h and
satisfy appropriated conditions for the operators V (t) to be unitary.
The family (V (t))t0 is not a semigroup but it is a cocycle with respect to time shift. This
property plays a key role when showing that j enjoys a quantum analogue of the Markov
property.
From the probabilistic point of view it is interesting to note that the family of operators
(A+ (t) + A(t))t0 , (i(A+ (t) A(t)))t0 can be interpreted as a pair of non-commuting
Brownian motions and ((t) + zA(t) + zA+ (t) + |z|2t)t0 (z C)
I as a Poisson process with
intensity |z|2.
L. Accardi, A. Frigerio and Y.G. Lu showed that these quantum Brownian motions can
be obtained as a weak coupling limit of quantum electrodymanics under the dipole and
rotating wave approximation and the above quantum analogue of the Poisson process arises
in a stochastic limit of a system interacting with a low density gas (see [6], [8]). These results
showed rigorously under what conditions a quantum Markov semigroup and a quantum
stochastic differential equation gives a good markovian approximation of an Hamiltonian
evolution (see [54] for further references). Moreover, Barchielli [15] applied for the first
time these quantum stochastic differential equations to the study of physical phenomena in
quantum optics as, for instance, the electron shelving effect.
At the same time quantum stochastic differential equations of the form (2) (and their
natural generalisations driven by multidimensional or infinite dimensional noises) were studied systematically by many authors (see e.g. [37], [38], [39], [40], [41], [52], [71]) together
with their applications to quantum Markov processes and semigroups in particular to the
realisation of flows of classical stochastic processes as quantum flows in Fock space.
The study of quantum stochastic differential equations as (2) is not simple when the
operators L1, . . . , L4 are unbounded. In this case an associated semigroup T of operators
(Tt)t0 on B(h) which is, roughly speaking, the infinitesimal generator of the process (ut )t0
plays a fundamental role. The operators Tt of this semigroup are normal and enjoy a
fundamental property called complete positivity. Unboundedness of L1 , . . . , L4 makes it
difficult to characterise the infinitesimal generator of T and to determine whether it is
identity preserving or Markov. This is a crucial step for proving that solutions to (2) are
family of unitary operators and almost the whole Chapter 3 will be dedicated to the study
of this problem.
In this work we give a self-contained exposition of several basic results scattered in
the literature that has grown very rapidly in the last years and develop a general framework for constructing quantum Markov processes as quantum (Evans-Hudson) flows in Fock
space. Moreover we collect several new results on quantum Markov semigroups and quantum stochastic differential equations that are necessary tools in the applications to the
construction of special quantum flows.
III
The main application is the realization of diffusion processes with smooth (C 4 , say)
covariance and drift as quantum flows in Fock space. This was first done in our paper [41]
for strongly elliptic diffusions. As P.-A. Meyer writes ([69] p.186), this is much better than
all previous results, though not yet completely satisfactory since ellipticity plays no role in
Itos theory. Here we remove this hypothesis by applying a new result on quantum Markov
semigroups (see [28]) and some new results on quantum flows.
In Chapters 1 and 2 we recall some definitions of quantum probability theory. The main
goal is to show that the fundamental definitions and properties of classical probability and
Markov processes are easily formulated in an algebraic language suitable for the study of
Markov processes appearing in quantum theory. In particular the notion of complete positivity which is the appropriate generalisation of positivity in the non-commutative framework
is discussed in detail.
In Chapter 3 we give a self-contained exposition of the basic results on uniformly continuous completely positive semigroups on the Von Neumann algebra B(h) of all bounded
operators in a Hilbert space h proving the well-known Lindblads theorem. Moreover we
describe the construction of completely positive semigroups of normal operators on B(h)
continuous for the -weak topology (the construction of the so-called minimal quantum dynalmical semigroup) and we give conditions - necessary and sufficient or simply sufficient in order for the semigroup to be identity preserving (i.e. Markov).
In Chapter 4 first we study the following problem: given a classical markovian semigroup
is it a non-trivial restriction to an abelian subalgebra of a quantum Markov semigroup? We
show that the problem can be attacked in two steps: 1) represent the infinitesimal generator
of the classical Markov process in a Lindblad form so that we can find immediately an
infinitesimal generator for the candidate quantum Markov semigroup, 2) show that the minimal quantum dynamical semigroup associated with this infinitesimal generator is Markov.
We shall see that the answer to the above extension problem is in the affirmative in
several cases and prove a new result on quantum extension of classical semigroups of diffusion
processes.
Then we apply our results on quantum Markov semigroups to construct the solution to
a class of quantum master equations.
In Chapter 5 we discuss a general scheme for constructing quantum Markov processes
through Markov operator cocycles (see Chapter 2, Section 3). In particular we show how to
construct a quantum flow with unbounded structure maps satisfying a quantum stochastic
differential equation and give a rather general sufficient condition in order for the restriction
to an abelian subalgebra to be a commutative flow. As an application we show that diffusion
processes with smooth coefficients can be realised as quantum flows in Fock space.
Acknowledgements
I would like to thank L. Accardi, K.R. Parthasarathy, R. Rebolledo and K.B. Sinha for
many fruitful discussions. I would also like to thank H. Comman, R. Monte, K.B. Sinha and
S. Wills who kindly readed and corrected portions of the manuscript.
It is a great pleasure to thank also R. Soto and the Editorial Board of Proyecciones
for publication of this unusually lengthy article.
Contents
1 Algebraic Probability Spaces
1.1 Fundamental definitions . . .
1.2 Classical probability spaces as
1.3 Quantum probability spaces .
1.4 Conditional expectation . . .
1.5 Topologies on algebras . . . .
.
.
.
.
.
3
3
4
4
8
9
2 Algebraic Markov Processes

2.1 Fundamental definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Completely positive linear maps . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 A quantum Feynman-Kac formula . . . . . . . . . . . . . . . . . . . . . . . .
13
13
16
24
3 Quantum Markov Semigroups

3.1 Fundamental definitions and examples .
3.2 Uniformly continuous QDS . . . . . . .
3.3 Minimal quantum dynamical semigroup
3.4 The resolvent of the minimal semigroup
3.5 Conservativity . . . . . . . . . . . . . .
3.6 Sufficient conditions for conservativity .
.
.
.
.
.
.
27
27
33
39
48
50
58
4 Classical and Quantum Semigroups

4.1 Preliminary definitions and results . . . . . . . . . . . . . . . . . . . . . . . .
4.2 Diffusion processes on IRd . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 A quantum master equation . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
67
71
77
. . . . . .
algebraic
. . . . . .
. . . . . .
. . . . . .
.
.
.
.
.
.
.
.
.
.
.
5 Quantum Flows
5.1 Quantum stochastic calculus . . . . . . .
5.2 Time reversal and dual cocycles . . . . . .
5.3 Quantum stochastic differential equations
5.4 Unitary solutions . . . . . . . . . . . . . .
5.5 Inner quantum flows . . . . . . . . . . . .
5.6 Quantum diffusions . . . . . . . . . . . . .
5.7 Conclusion . . . . . . . . . . . . . . . . .
A Results on semigroups
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
83
83
90
93
101
106
110
115
117
CONTENTS
Algebraic Probability Spaces

In this chapter we recall some definitions of quantum probability theory in the general
framework developed by Accardi and several co-workers. Several expository papers in the
series Quantum Probability and Related Topics and two monographes [69], [74] are already
available in the literature. We will only illustrate the definitions by two examples which are
included only with the aim to give a flavour of the relation with classical probability.
The reader is supposed to be familiar with the basic language and facts on C -algebras
and von Neumann algebras. In Section 5 we recall only some fundamental definitions and
results.
1.1
Fundamental definitions
Definition 1.1 A -algebra is a complex algebra A with an involution, denoted by , with

the following properties:
(a + b)
(a )
=
=
+
a
b
a
b a
(ab)
for all a, b A, , C.
I An element a of A is called positive if there exists b A such
that a = b b.
We shall consider often -algebras with a unit denoted by 1l.
Definition 1.2 Let A be a -algebra with a unit 1l. A state on A is a linear map
: A C
I
with the properties:
1. (positivity) (a a) 0, for all a A,
2. (normalisation) (1l) = 1.
We can now introduce the notion of algebraic probability space and algebraic random
variable according to Accardi, Frigerio and Lewis [5].
Definition 1.3 An algebraic probability space is a pair (A, ) where A is a -algebra with
unit and is a state on A.
3
1. ALGEBRAIC PROBABILITY SPACES
Definition 1.4 Let (A, ) be an algebraic probability space and let B be a -algebra. An
algebraic random variable on A with values in B is a -homomorphism
j : B A.
The above quite general definitions will be always applied in concrete cases when A
and B are at least C -algebras. In this case it is well known (see [21] Th. 2.1.10 p.60) that
A and B are isomorphic to a norm-closed -algebra of bounded operators in a Hilbert space.
1.2
Classical probability spaces as algebraic
A measurable space (, F) clearly determines uniquely the commutative -algebra A =

L(, F;C)
I of F-measurable, bounded, complex-valued functions on . A probability
measure IP on F induces a state on A by
Z
(f) =
f()dIP ().
Therefore the classical probability space (, F, IP ) can be considered also as the algebraic
probability space (A, ).
Let (E, E) be a measurable space. Classical random variables on with values in E
can be also interpreted as algebraic random variables. In fact, consider the -algebra B =
L(E, E;C).
I A classical random variable x can be described as an algebraic random variable
by the -homomorphism
j : B 7 A,
j(f) = f x.
It is worth noticing here that each event can be represented by a projection in the
-algebra A through the identification with its indicator function.
1.3
Quantum probability spaces
Algebraic probability spaces appear as the basic structure in mathematical models for quantum mechanics. The mathematical structure however is (or at least can be assumed to be)
richer with some more analytical properties on the algebra A and state . In order to
distinguish an important class of algebraic probability spaces we give the following:
Definition 1.5 A quantum probability space is a pair (A, ) where A is a von Neumann
algebra and a -weakly continuous state on A.
An event in the quantum probability space (A, ) is a projection operator in A.
A quantum random variable in (A, ) with values in a von Neumann algebra B is a
-weakly continuous homomorphism
j : B A.
Although one could define events in the same way also when A is only a C -algebra the
set of events in this case might be very poor. Indeed, if A is the C -algebra of complexvalued continuous functions on IRd , then the set of events is trivial. On the other hand a
von Neumann algebra A is generated by projections in A.
Notice that, contrary to the classical case, the intersection (product) of two events is no
longer an event if the corresponding orthogonal projections do not commute.
We will show now in which sense a self-adjoint operator X affiliated with the von Neumann algebra A of a quantum probability space (A, ) can be considered both as a classical
and a quantum random variable.
1.3. QUANTUM PROBABILITY SPACES
As we already noted we can assume that A is a sub von Neumann algebra of the von
Neumann algebra B(h) of bounded operators on a Hilbert space h. Therefore X is a selfadjoint operator on h.
Let B be the von Neumann algebra L (IR;C).
I
By the spectral theorem in functional
calculus form ([78] Th. VIII.5 p. 262), for all f B, we can define the element f(X) of A.
The map
j : B 7 A,
j(f) = f(X)
is a *-homomorphism. This is clearly -weakly continuous because, for every increasing net
(f ) of positive elements of B with least upper bound f in B we have
sup j(f ) = sup f (X) = f(X) = j(f).
Therefore X defines a quantum random variable.

Let = IR and let F be the Borel -field on . Since both j and are -weakly
continuous we can define a probability measure on F by putting
IP (B) = (1B (X))
(1.1)
where 1B denotes the indicator function of a B F. Thus we have constructed a classical

probability space (, F, IP ).
We want to show now that X can be represented as a classical real random variable on
(, F, IP ). Fix a unit vector e in h and consider the closed subspace h0 of h generated by e
and vectors of the form
f1 (X)f2 (X) . . . fn (X)e
with n 1, f1, . . . , fn B. Let U be the unique unitary operator
U : h0 L2(, F, IP )
such that
U e = 1,
U f1 (X)f2 (X) . . . fn (X)e = f1 f2 . . . fn .
The operator U is unitary because of the relation

Z
g()dIP () = (g(X))
IR
for g integrable with respect to IP which follows immediately from (1.1).

It is easily checked that the following diagram commutes:
h0
U y
h0
yU
L2 (, F, IP ) L2 (, F, IP )
fX
(where fX (x) = x). Precisely we can show that
v D(X) if and only if fX ()(U v)() L2 (, F, IP ),
if w U (D(X)), then (U XU w)() = fX ()w().
Therefore U XU acts on L2 (, F, IP ) as the multiplication operator by a real function fX
and the self-adjoint operator X defines a classical real random variable.
Two non-commuting self-adjoint operators, however, cannot be represented as multiplication operators on the same Hilbert space L2 (, F, IP ). This, roughly speaking, can
be summarized by saying that (A, ) is a quantum probability space containing infinitely

many classical probability spaces.
The following are concrete example in which the above fact occurs. We will use these
models to illustrate also how classical probabilistic notions appear naturally in the quantum
context.
Example 1.1 Spin matrices
Let h = C
I 2, and let A be the von Neumann algebra of 2 2 complex valued matrices and
be any state on A. The pair (A, ) is clearly a quantum probability space. The self-adjoint
operators on h (also called spin matrices or Pauli matrices).

0 1
0 i
1 0
1 =
,
2 =
,
3 =
,
1 0
i 0
0 1
represent three non-commuting quantum random variables. The above discussion shows
immediately that 3 can be represented as a classical real random variable x taking values
{1, +1} with probabilities

1 0
0 0
IP {X = 1} =
,
IP {X = 1} =
.
0 0
0 1
Notice that 1 and 1 are the eigenvalues of 3 and the events {3 = 1} and {3 = 1} are
represented respectively by the orthogonal projections

1 0
0 0
,
.
0 0
0 1
In a similar way the laws of 1 and 2 can be computed using the spectral decompositions

1/2 1/2
1/2 1/2
1 =
1/2 1/2
1/2 1/2

1/2 i/2
1/2 i/2
2 =
i/2 1/2
i/2 1/2
Thus 1, 2, 3 are three non-commutative random variables with values in {1, +1}.
Example 1.2 The harmonic oscillator
Let h be the Hilbert space
h = l2 (IN ),
IN = { 0, 1, 2, . . .}
with the canonical orthonormal basis (en ). We consider then the following operators:
1. annihilation operator
D(a)
aen
{x = (xn)n h |
n|xn|2 < }
n en1
if n > 0,
ae0 = 0,
2. creation operator
D(a )
a en
= {x = (xn )n h |
n|xn|2 < }
n + 1 en+1
for all n 0,
1.3. QUANTUM PROBABILITY SPACES
3. number operator
D(N )
{x = (xn )n h |
N en
a aen = nen
for all n 0.
n2|xn|2 < }
4. momentum (or electric field) operator

i
p = (a a)
2
D(p) = D(a) = D(a ),

5. position (or magnetic field) operator
1
q = (a + a) .
2
D(q) = D(a) = D(a ),
It is well-known that the operators N, p, q are self-adjoint. Therefore they can be considered as algebraic random variables. The canonical commutation relations (domains of the
operators involved will be made precise later)
[a, a] = aa a a = 1l,
(N + 1)a = aN,
a (N + 1) = N a
([, ] denoting the commutator) yield

[q, p] = i1l,
[N, p] = iq,
[N, q] = ip.
Therefore N, p, q can not be represented as random variables on the same classical probability
space.
The above model is called the Heisenberg representation of the canonical commutation
relations over C.
I It is well-known that it is unitarily equivalent to the Schr
odinger representation. Indeed, letting (Hn )n0 be the orthonormal sequence of the Hermite polynomials in
L2(IR; 1/4 exp(x2 /2)dx) the unitary operator
U : l2 (IN ) L2(IR; 1/4 exp(x2 /2)dx),
Uen = Hn
allows us to move from one representation to the other.

p
l2 (IN )
U y
l2 (IN )
yU
L2 (IR; 1/4 exp(x2 /2)dx) L2 (IR; 1/4 exp(x2 /2)dx)

d
i dx
We refer to Meyers book [69] for more details on this subject. The following is a conversion
table
Heisenberg representation
p
q
N
Schroedinger representation
d
i
dx
x

1
d2
2
2 +x 1
2
dx
Let be the state on B(L2 (IR; 1/4 exp(x2 /2)dx)) defined by the unit vector H0
(coinciding with the constant function 1)
(a) = hH0, aH0i .
Well-known facts allow us to compute easily the law of the random variables p, q. Notice
first that the unitary groups generated by q and p are
(exp(itq)v)(x) = exp(itx)v(x),
(exp(itp))v(x) = v(x + t).
Therefore the characteristic functions of p and q in the state are given by

(exp(itq))
(exp(itp))
=
=
hH0, exp(itq)H0 i = exp(t2 /4),

hH0, exp(itp)H0 i = exp(t2 /4).
Thus p and q are two non-commuting random variables with gaussian distribution with
mean 0 and variance 1/2.
Moreover we have
(N ) = hH0, N H0i = he0 , a ae0 i = 0.
Therefore, since N is a non negative random variable, it is almost surely 0 with respect to
the probability law determined by the state .
1.4
Conditional expectation
Examples given in the previous section motivate the following

Definition 1.6 Let (A, ) be an algebraic probability space. For all element a of A we call
the expectation, or mean value, of a in the state the number (a).
For all a, b A we define the covariance
(a (a)) (b (b))
and the variance
(a (a)) (a (a)) .
Notice that here we defined an algebraic analogue of expectation, covariance and variance
essentially for bounded random variables. In fact a is an element of A.
We are now in a position to introduce the notion of conditional expectation.
Definition 1.7 Let (A, ) be an algebraic probability space and let A0 be a sub -algebra of
A with unit 1l. A conditional expectation is a linear map
IE [ | A0 ] : A 7 A0
with the following properties:
1. for all a A such that a 0, we have IE [ a | A0 ] 0,
2. IE [ 1l | A0 ] = 1l,
3. for all a0 A0 and all a A we have
IE [ a0 a | A0 ] = a0IE [ a | A0 ] ,
1.5. TOPOLOGIES ON ALGEBRAS
4. for all a A we have IE [ a | A0 ] = (IE [ a | A0 ]) ,

5. for all a A0 we have
(a) = (IE [ a | A0 ]) .
Example 1.3 The conditional expectations we shall use are of the following form. Let h, h1
be complex separable Hilbert spaces. Consider the Hilbert space k = h h1 and consider
the -algebras
A = B(k),
A0 = B(h) 1l.
For all states on B(h) and 1 on B(h1 ), consider the state on A defined by
(a a1 ) = (a)1 (a1 )
for all a B(h), a1 B(h1 ). Then the linear map
IE [ | A0 ] : A 7 A0 ,
IE [ a a1 | A0 ] = 1 (a1) (a 1l)
is a conditional expectation.
Although our definition of conditional expectation in the -algebraic language seems to
be quite natural, it is too restrictive in quantum probability. In fact, contrary to the classical
case, given a quantum probability space (A, ) and a sub -algebra let A0 of A with identity
1l, a conditional expectation IE [ | A0 ] : A A0 may not exists. A detailed discussion on
conditional expectations in von Neumann algebras can be found in [73] and [77] by D. Petz.

Definition 1.8 Let (A, ) be an algebraic probability space. A filtration is a family (At] t0
of sub--algebras of A such that
As] At]
for all s t.

Definition 1.9 Let (A, ) be a quantum probability space and let (At] t0 be a filtration.
Suppose that, for all t 0, there exists conditional expectations

IE | At] : A At] .

The family IE | At] t0 is called projective if, for all s t, we have

IE IE | As] | At] = IE | As] .
We refer to the recent books of Meyer [69], Parthasarathy [74] and the references therein
for a more detailed introduction to the language of quantum probability and other interesting
examples.
1.5
Topologies on algebras
A von Neumann algebra (or W -algebra) could be defined intrinsically as a special C algebra. However we will consider only concrete von Neumann algebras, that is sub-algebras of operators on some Hilbert space h closed under the weak or -weak or strong
operator topology (see [21] pp.71-72).
These topologies can be defined through the convergence of nets.
Definition 1.10 Let (x) be a net in B(h) and let x B(h). We say that:
10

1. (x ) converges weakly to x if hv, xui converges to hv, xui for every v, u h.
P
P
2. (x ) converges -weakly to x if the sum nhvn , xuni converges to the sum P nhvn , xuni
2
for every
P pair2of sequences (vn )n, (un)n of elements of h such that the series n kunk
and n kvn k converge.
3. (x ) converges strongly to x if xu converges to xu for every u h.
P
4. (x ) converges -strongly to x if the sum n k(x Px)unk2 converges to 0 for every
sequence (un )n of elements of h such that the series n kunk2 converges.
Clearly (x) converges -weakly to x if and only if, for every trace class operator in
h, tr(x ) converges to tr(x).
The following facts will be frequently used:
(a) the -weak topology is stronger than the weak topology and not comparable to the
strong one,
(b) the weak and -weak topology coincide on bounded subsets of B(h).
Definition 1.11 A von Neumann algebra is a *-subalgebra of B(h) containing 1l closed
in the weak (or by the bicommutant theorem [21] Th. 2.4.11 p.72 - -weak or strong or
-strong) topology.
We recall the following property of the cone of positive elements A+ of a von Neumann
algebra A (see [21] Lemma 2.4.19 p.76).
Proposition 1.12 Let A be a von Neumann algebra of operators acting on a Hilbert space
h and let (x ) be an increasing net in A+ with an upper bound in A+ . Then (x) has a
least upper bound x = sup x in A+ and the net converges -strongly to x.
We define a useful class of positive functionals on a von Neumann algebra.
Definition 1.13 Let A be a von Neumann algebra of operators acting on the Hilbert space
h and let be a positive linear functional on A. We say that is normal if
sup (x ) = (sup x ).
The following fundamental properties of states on a von Neumann algebra are well-known
(see [21] Theorem 2.4.21 p.76).
Theorem 1.14 Let A be a von Neumann algebra of operators acting on a Hilbert space h
and let be a state on A. The following conditions are equivalent:
1. is normal,
2. is -weakly continuous,
3. there exists a density matrix (i.e. a positive trace-class operator on h with tr() = 1)
such that (x) = tr(x).
We shall use often a consequence of the equivalence of 1 and 2. Let us recall that a subset
E of a Hilbert space h is called total in h if the linear manifold generated by E is dense in h.
Proposition 1.15 Let A and B be von Neumann algebras, B acting on a Hilbert space h,
and let T : A B be a positive linear map. The following conditions are equivalent:
1.5. TOPOLOGIES ON ALGEBRAS
11
1. T is -weakly continuous (i.e. continuous with respect to the -weak topologies on A

and B),
2. for every increasing net (x) in A+ with least upper bound x in A+ the increasing
net (T x ) in B+ converges -weakly to T x,
3. for every increasing net (x ) in A+ with least upper bound x in A+ we have
lim hu, (T x)ui = sup hu, (T x)ui = hu, (T x)ui
for each u in a linear submanifold of h which is norm-dense in h,

4. for every increasing net (x ) in A+ with least upper bound x in A+ we have
lim hv, (T x )ui = hv, (T x)ui
(1.2)
for each v, u in a total subset of h.

Proof. 1 implies 2. Indeed it suffices to note that the net (x ) converges weakly to x by
Proposition 1.12.
2 implies 3. Obvious since the linear functionals on B y hu, yui with u h are
-weakly continuous.
3 implies 4. We show first that 3 implies that the net (hu, (T x)ui) converges to
hu, (T x)ui for each u h. Indeed, for every > 0, there exists u in the dense subset such
that ku uk < . The inequality T x T x clearly implies kT xk kT xk. We have then
|hu, (T x)ui hu, (T x)ui|
|hu u , (T x)ui hu u , (T x)ui|
+
+
|hu, (T x)(u u )i hu, (T x)(u u)i|

|hu, (T x)u i hu , (T x)ui|
ku uk (kT xk + kT xk) (kuk + kuk)
|hu, (T x)u i hu , (T x)ui| .
Therefore we have
lim |hu, (T x)ui hu, (T x)ui| 2kT xk(2kuk + ).
Since is arbitrary, the net (hu, (T x)ui) converges to hu, (T x)ui for each u h.
By the polarisation identity

1 X k
v + ik u, (T x)(v + ik u)
i
4
3
hv, (T x )ui =
k=0
it follows then that (1.2) holds for each v, u h.

4 implies 3. In fact 4 implies that (1.2) holds for each v, u in the dense subset of h
linearly spanned by the total set. This linear span is obviously dense.
3 implies 2. Let (vn )n0, (un)n0 be two sequences of vectors in h such that the sequences
(kvnk)n0, (kunk)n0 are square-summable. We must show that
X
X
lim
hvn , (T x)uni =
hvn , (T x)uni .
n0
n0
To this end, for every > 0, take an integer such that

X
X
kunk2 < ,
kvn k2 < .
n>
n>
12
We have then

X
X

hv
,
(T
x
)u
i
hv
,
(T
x)u
i
n
n
n
n

n0

n0
X
(kT xk + kT xk)
kun k kvn k
n>
X

X

+
hvn , (T x)uni
hvn, (T x)uni .

n=0
n=0
The first term, since kT xk kT xk, can be estimated by

!
X
X
2
2
kT xk
< 2kT xk.
kunk +
kvn k
n>
n>
Moreover, as we have shown in the proof that 3 implies 4, the property 3 implies the
convergence of the net (hv, (T x)ui) to hv, (T x)ui for each v, u h. We have then

X
X

lim
hvn, (T x)un i
hvn, (T x)uni 2kT xk.
n0

n0
Since is arbitrary this shows that (1.2) holds.
2 implies 1. Let (x ) be a net in A converging -weakly to x. For every pair (vn )n0 ,
(un)n0 of sequences of vectors in h such that (kvn k)n0, (kun k)n0 are square-summable
let be the -weakly continuous functional on B
X
(y) =
hvn , yuni .
n0
By the complex polarisation identity can be written as a linear combination of four positive
linear functionals k , k = 0, 1, 2, 3, with
X

k (y) =
(v + ik u), y(v + ik u) .
n0
Therefore, in order to show that 2 implies 1, it suffices to prove that the net ((T x ))
converges to (T x) for every positive linear functional of the above form.
If, for such an , we have (T 1l) = 0, then we have also (T x) = 0 for each self-adjoint
element x of A because kxk1l x kxk1l and T is positive. Therefore
(T x) = (T (x + x ))/2 + i(T (x x) /i)/2 = 0
for each x A. Thus there is nothing to prove.
If (T 1l) > 0, then, letting
e (y) = (T y)/(T 1l),
we define a state
e on A. Condition 2 implies then that
e is normal. Then, by virtue of
Theorem 1.14, it is -weakly continuous so that
lim (T x) = lim (T 1l)e
(T x ) = (T 1l)e
(T x) = (T x).
This completes the proof.

A map enjoying the property 3 is also called normal. We will often use the above
equivalence to show that a positive map is continuous for the -weak topology. Moreover
we will often use normal with the same meaning of -weak when no confusion can arise.
We refer to the book [21] for more detailed results on von Neumann algebras.
Algebraic Markov Processes

In this chapter we describe the abstract definition and the basic facts on algebraic Markov
processes (see [5]). The main goal is to show that the fundamental definitions and properties
of Markov processes are easily formulated in an algebraic language suitable for the study of
Markov processes appearing in quantum theory. Moreover, we discuss in detail the notion
of complete positivity which turns out to be the natural generalisation of positivity for
commutative (classical) case and a non-commutative version of the Feynman-Kac formula
which is the basic ingredient in the construction of Markov cocycles and processes.
2.1
Fundamental definitions
Definition 2.1 Let (A, ) be an algebraic probability space and let B be a *-algebra. An
algebraic stochastic process on A with values in B is a family u = (ut )t0 of algebraic
random variables on A with
values in B. The quantum stochastic process u is adapted with
respect to a filtration At] t0 if
ut(B) At]
for all t 0.
In the remaining part of this chapter we fix the following framework:

an algebraic probability space (A, ) with a filtration At] t0 and 1l A0] ,

a projective family of conditional expectations IE | At] t0,
an adapted process u with values in A0] ,

a family of conditional expectations IE | ut A0]
, such that
t0

IE IE[a | At] ] | ut A0] = IE a | ut A0] for every a in the -algebra generated by us A0]
with s t.
Definition 2.2
An adapted algebraic
stochastic process u is an algebraic Markov process,

with respect to the filtration At] t0, if, for all s, t 0 and all X A0] , we have

IE ut+s (X) | As] = IE ut+s (X) | us A0] .
(2.1)
An algebraic Markov process is covariant or homogeneous if, for all s, t 0 and all a A0] ,
we have

ut IE us (a) | A0] = IE ut+s (a) | At] .
(2.2)
13
14
2. ALGEBRAIC MARKOV PROCESSES
Example 2.1 To illustrate the above definition we show that an homogeneous classical
Markov process can be considered as a covariant quantum Markov process.
Let x = (xt)t0 be a classical (adapted) Markov process with values in a measurable
space (E, E), initial law and transition probability function
P : D E E IR,
where D = { (s, t) [0, +) | 0 s t }. Suppose that, for every (s, t) D, x E and
every A E such that (A) = 0 we have P (s, t, x, A) = 0. We consider the canonical
realization on the classical probability space (, F, IP ) where
Y
E,
F = t0E,
=
t0
and IP is the probability measure on F defined by

Z
Z
Z
IE [f(xt1 , . . . , xtn )] =
d(z)
P (0, t1, z, dz1)
P (t1, t2, z1 , dz2)...
E
E
E
Z
...
P (tn1, tn, zn1, dzn)f(z1 , . . . , zn )
E
where 0 t1 < t2 < . . . < tn . Consider the filtration Ft]

Y
Ft] =
E.
t0
given by
0st
Consider a quantum probability space (A, ) with a filtration At]

A
t0
where

L (, F;C)
I ,
At] = L , Ft];C
I ,
(f(xt1 , . . . , xtn )) = IE [f(xt1 , . . ., xtn )] .
The classical process x defines a family of *-homomorphisms

ut : A0] At] ,
ut (f) = f(xt ).
Therefore the classical Markov property

IE f(xt+s ) | Fs] = IE [ f(xt+s ) | xs ]
for all s, t 0 and all f A0] is immediately translated, in the algebraic language, into the
identity (2.1).
The Markov process is time homogeneous if and only if
Z
Z
f(y)P (s, t + s, x, dy) =
f(y)P (0, t, x, dy)
E
for all s, t 0, and all f A0] . Now (2.2) can be easily understood in view of the following
correspondence table between the quantum and classical case
quantum
IE ut+s(f) | At]

IE us(f) | A0]

ut IE us(f) | A0]
classical
ZE
ZE
f(y)P (t, t + s, xt, dy)
f(y)P (0, s, x, dy)

f(y)P (0, s, xt , dy).
2.1. FUNDAMENTAL DEFINITIONS
15
Proposition 2.3 The following conditions are equivalent:

1. u is an homogeneous algebraic Markov process,
2. for all n 1 and all a1 , . . ., an A0] , 0 < t1 < . . . < tn, s 0 we have

IE ut1+s (a1 ) . . . utn+s (an ) | As]

= IE ut1 +s (a1 ) . . . utn +s (an ) | us(A0] ) ,

ut1 IE us(a1 ) | A0]
= IE ut1 +s (a1 ) | At1] .
Proof. Clearly it suffices to prove that, if u is an homogeneous quantum Markov process,
then the first identity of condition 2 holds. We will consider the case n = 2 for simplicity.
Using the Markov, covariance and *-homomorphism property of u we can show that the
conditional expectation

IE ut1+s (a1 )ut2 +s (a2 ) | As]
is equal to
=
=
=
=
=
=

IE ut1 +s (a1)IE ut2 +s (a2 ) | At1+s] | As]

IE ut1 +s (a1)ut1 +s IE ut2 t1 (a2 ) | A0] | As]

IE ut1 +s a1IE ut2 t1 (a2 ) | A0] | As]

IE ut1 +s a1IE ut2 t1 (a2 ) | A0] | us A0]

IE ut1 +s (a1)ut1 +s IE ut2 t1 (a2 ) | A0] | us(A0] )

IE ut1 +s (a1)IE ut2 +s (a2 ) | At1+s] | us(A0] )

IE ut1 +s (a1 )ut2+s (a2 ) | us(A0] ) .

The following proposition shows that, as in classical probability, one can associate a
semigroup to an algebraic Markov process.
Proposition 2.4 Let u be a covariant algebraic Markov process. For all t 0 define the
map

Tt : A0] A0] ,
(2.3)
Tt(a) = IE ut (a) | A0]
Then T = (Tt )t0 is a semigroup of operators in A0] with the following properties
1. for every integer n 1 and every family a1 , . . . , an, b1, . . . , bn of elements of A0] we
have
n
X
bp Tt(ap aq )bq 0
(2.4)
p,q=1
for every t 0,
2. Tt (1l) = 1l for every t 0.
Proof. T is a semigroup in fact, for every t, s 0 and a A0] ,we have,

Tt+s(a)
IE ut+s(a) | A0]

(projectivity)
= IE IE ut+s(a) | At] | A0]

(covariance)
= IE ut IE us(a) | A0] | A0]
= Tt (Ts (a))
16
Clearly property 2 holds since 1l belongs to A0] and both ut and the conditional expectation
preserve 1l.
Finally for every integer n 1 and every family a1 , . . . , an, b1, . . . , bn of elements of A0]
the left-hand side of (2.4) is equal to
n
X

bp IE ut(ap aq ) | A0] bq
p,q=1
n
X

IE bp ut (ap )ut (aq )bq | A0]
p,q=1
= IE
"
n
X
ut(ap )bp
p=1
n
X
ut (aq )bq
q=1

A0]
which is clearly a positive operator because of property 1 of conditional expectation. This

completes the proof.
Property 1 of algebraic Markov semigroup is called complete positivity and plays a key role
in our exposition. As we have seen in the proof of Proposition 2.4, it follows from complete
positivity of conditional expectation. More general conditional expectations in quantum
probability (see [3], [77]) are also completely positive. Therefore complete positivity is the
proper generalization of positivity in the non-commutative framework.
The next section will be devoted to a detailed study of this property and its connections
with positivity.
2.2
Completely positive linear maps
Let A and B be two -algebras with unit. We denote by 1l the unit of both since each time
it will be clear from the context to which algebra it belongs.
The obvious generalization of the notion of positivity for classical (sub) Markov operators
is too weak when A and B are non commutative. Proposition 2.4 motivates the introduction
of the following stronger notion of positivity.
Definition 2.5 The linear map T : A B is called:
1. n-positive if for every family a1 , . . . , an of elements of A and every family b1, . . . , bn
of elements of B we have
n
X
bp T (ap aq )bq 0,
p,q=1
2. completely positive if it is n positive for every integer n 1.

The following fact is obvious.
Proposition 2.6 Let T : A B be a -homomorphism. Then T is completely positive.
In the remaining part of this section A and B will be C -algebras with unit 1l.
Note that, for every integer n 1, the algebraic tensor product -algebras A Mn and
B Mn can be represented as the -algebras of n n matrices with entries in A and B
respectively. Every element x of A Mn can be written in the form
X
x=
xij Eij .
(2.5)
1i,jn
2.2. COMPLETELY POSITIVE LINEAR MAPS
17
where Eij denotes the n n matrix with all the entries equal to 0 except the ij-th which is
equal to 1.
Given a linear map T : A B we can define linear maps Tn : A Mn B Mn by
Tn (a Eij ) = T (a) Eij
(2.6)
In order to give a useful condition equivalent to complete positivity by means of the

maps Tn we prove a simple fact on positive elements of A Mn in the case when A is a
C -algebra.
Proposition 2.7 Let A be a C -algebra and let x be an element of A Mn . The following
conditions are equivalent:
1. x is positive,
2. x is a finite sum of matrices of the form
X
ai aj Eij
1i,jn
with a1, . . . , an A,
3. for all a1 , . . ., an A we have
X
ai xij aj 0.
1i,j,n
Proof. 1 implies 2. In fact, since x is positive, it can be written in the form y y with
y A Mn (see, for example, [21], Th. 2.2.10). Writing y in the form (2.5) we have
x=
n
X
X
yì
y`j Eij .
`=1 1i,jn
2 obviously implies 3.
3 implies 1. By representing A as a sub-C -algebra of the algebra of all bounded operators on a Hilbert space H (see [21] Th. 2.1.10 p. 60) and decomposing H into cyclic
orthogonal subspaces we may suppose that there exists a cyclic vector u for the representation. Condition 3 then implies then the inequality
X
hai u, xij aj ui 0.
i,j
Therefore, since the vector u is cyclic, we have

X
hvi , xij vj i 0
i,j
for all v1 , . . ., vn H. This completes the proof.

We are now in a position to prove the following
Proposition 2.8 Let A, B be C -algebras and let T : A B be a linear map. The following
1. T is completely positive,
18

2. for every integer n 1 the map Tn defined by (2.6) is positive.
Proof. The P
second condition implies the first by parts 2 and 3 of Proposition
P 2.7. Since the
operator x = i,j ai aj Eij in AMn is positive, we have that Tn (x) = i,j T (ai aj )Eij
is positive.
P
Conversely the first condition implies that i,j T (ai aj ) Eij is positive. Therefore Tn
is positive because of the equivalence of conditions 1 and 2 of Proposition 2.7.
Proposition 2.9 Let T : A B be a linear map where B is the C -algebra B(K) of all
bounded operators on a Hilbert space K. Then T is completely positive if and only if for
every integer n 1 and every a1 , . . ., an A, u1, . . . , un K
X
hui , T (ai aj )uj i 0.
1i,jn
Proof. Notice that the C -algebra B Mn can be represented as the C -algebra of all
bounded operators on the n-fold direct sum K . . . K. Therefore the above condition is
clearly equivalent to positivity of the map Tn on A Mn for every integer n 1.
The following counterexample, essentially due to W.B. Arveson [12], shows the existence
of positive maps that are not completely positive. Let n 2 be an integer and let both A
and B be the -algebra Mn. Consider the positive linear map
T : M n Mn ,
T (a) = at
(2.7)
(at denoting the transpose matrix). We will prove that T is not 2-positive. Let b be the
2 2 matrix with entries in Mn having the matrix E11 as 11-th entry, E1n as 12-th entry,
En1 as 21-th entry, Enn as 22-th entry. For example, when n = 2, we have
1 0 0 1
1 0 0 0
0 0 0 0
0 0 1 0
b=
, T2 (b) =
.
0 0 0 0
0 1 0 0
1 0 0 1
0 0 0 1
It is easy to check that the matrix b/2 is a self-adjoint projection. Thus it is positive.
However the 2n 2n real matrix Tn (b) is not positive because its elements do not satisfy
the inequality |xij |2 xiixjj for i = n, j = n + 1.
Other counterexamples of maps which are n-positive but not (n + 1)-positive for an
arbitrary integer n can be found in [29].
We deduce now two useful properties of 2-positive maps.
Proposition 2.10 Let A, B (B B(K)) be C -algebras with unit and let T : A B with
be a 2-positive map. Then:
1. if T (1l) is invertible in B then for all a A we have the Schwarz inequality
T (a )(T (1l))1 T (a) T (a a),
(2.8)
2. for all a A we have the inequality

T (a )T (a) kT (1l)k T (a a),
3. T is continuous and
kT k = kT (1l)k .
(2.9)
19
Proof. The operator in B(K) M2

T (a a)
a a a
T (a )
0 0
= T2
+
T (a) T (1l) + 1l
a
1l
0 1l
is positive for every > 0. Hence, for every u, v K we have
hu, T (aa)ui + hu, T (a)vi + hv, T (a)ui + hv, (T (1l) + 1l)vi 0.
The operator T (1l) + 1l has a bounded inverse. Taking v = (T (1l) + 1l)1 T (a)u we have
the inequality

u, T (a)(T (1l) + 1l)1T (a)u hu, T (a a)ui
for all u K. If T (1l) is invertible in B then, by letting tend to 0 we obtain the inequality
(2.8).
In any case, since 1l kT (1l) + 1lk (T (1l) + 1l)1 , we find the inequality
T (a )T (a) kT (1l) + 1lk T (a )(T (1l) + 1l)1 T (a) kT (1l) + 1lk T (a a).
Therefore, letting tend to 0, we obtain (2.9).
2
The inequality (2.9) and the property kxxk = kxk of C norms immediately yield

2
2
2
kT (a)k kT (1l)k kT (a a)k kT (1l)k T (kak2 1l) = kT (1l)k kak .
The following results show, in particular, that a positive linear map T : A B is in fact
completely positive when at least one of the C -algebras A and B is commutative. We refer
to [12], [82] for the proofs.
Theorem 2.11 (Arveson) Let B be a commutative C -algebra. Then every positive linear
map T : A B is completely positive.
Theorem 2.12 (Stinespring) Let A be a commutative C -algebra. Then every positive
linear map T : A B is completely positive.
The following simple properties of completely positive maps turn out to be useful.
Proposition 2.13 Let T : A B and S : A B be two completely positive linear maps.
Then the map S + T completely positive.
Proof. Obvious.
Proposition 2.14 Let A1, A2 , A3 be C -algebras and T : A1 A2 , S : A2 A3 be two
completely positive linear maps. Then the linear map S T : A1 A3 is completely positive.
Proof. It suffices to notice that, for every integer n 1, the map (S T )n : A1 Mn
A3 Mn coincides with the composition Sn Tn .
Proposition 2.15 Let K be a Hilbert space and let (Tm )m1 be a sequence of completely positive linear maps Tm : A B(K). Suppose that, for every a A, the sequence (Tm (a))m1
converges weakly. Then the linear map T : A B(K) defined by
T (a) = lim Tm (a)
m
is completely positive.
20
Proof. By Proposition 2.9 it suffices to note that

X
X
hui , T (ai aj )uj i = lim
hui, Tm (ai aj )uj i 0,
m
1i,jn
1i,jn
for every integer n 1, every u1 , . . ., un K and every a1 , . . . , an A.

W.F. Stinespring proved in [82] the following characterization of completely positive
maps.
Theorem 2.16 (Stinespring) Let B be a sub C -algebra of the algebra of all bounded operators on a Hilbert space H and let A be a C -algebra with unit. A linear map T : A B is
completely positive if and only if it has the form
T (a) = V (a)V
(2.10)
where (, K) is a representation of A on K for some Hilbert space K, and V is a bounded

operator from H to K.
Proof. Let T be a linear map of the form (2.10) and let (aij )1i,jn be a positive matrix
in A Mn. For all vectors (uj )1jn in H we have then
X
hui, T (aij )uj i =
i,j
hV ui , (aij )V uj i 0
i,j
because is completely positive by Proposition 2.6.

Conversely suppose that T is completely positive and consider the vector space A H,
the algebraic tensor product of A and H. On this space we define the bilinear form (, )
X
(x, y) =
hui , T (ai bj )vj i
i,j
for x =
ai ui and y =
j bj
vj in A H. Since T is completely positive we have
(x, x) =
hui , T (ai aj )uj i 0
i,j
for all x A H. Hence the bilinear form (, ) is positive. Consider the algebra homomorphism 0 defined on A with values in linear transformations in A H
!
X
X
0(a)
ai ui =
(aai ) ui.
i
Notice that, for all x, y as above, we have

(x, 0(a)y) = (0 (a )x, y) .
It follows that, for every x A H, the linear map
: A C,
I
(a) = (x, 0(a)x)
is a positive linear functional on A. Therefore we have (see [21] Prop. 2.3.11)

2
k0(a)xk = (x, 0(a a)x) ka ak (1l) = kak kxk .
(2.11)
21
Let N be the linear subspace of vectors x in AH such that (x, x) = 0. Since N is invariant
under 0(a) for every a A because of (2.11), we can consider the quotient pre-Hilbert space
A H/N defining the pre-scalar product on A H/N by
(x + N , x + N ) = (x, x) .
Let K be the Hilbert space obtained by completion. By the above construction the homomorphism 0 extends to a representation of A into B(K) such that
(a)(x + N ) = 0(a)x + N
for a A and x A H. Consider the linear operator V : H K
V u = 1l u + N .
This operator is bounded because of the inequality
2
kV uk = hu, T (1l)ui kT (1l)k kuk .

A straightforward computation yields (2.10).
Definition 2.17 A pair (, V ) satisfying (2.10) is called a Stinespring representation of the
completely positive map T . It is called a minimal Stinespring representation if the Hilbert
space K coincides with the closure of the vector subspace generated by
{(a)V u | a A, u H} .
(2.12)
In other words a pair (, V ) is a minimal Stinespring representation if the set (2.12) is

total in K.
Every completely positive map admits a minimal Stinespring representation. In fact,
with the notation of the proof of Theorem 2.16, it suffices to consider as Hilbert space K
the closure K1 of the vector space generated by (2.12). The restriction 1 of to K1 also
satisfies (2.10).
The minimal Stinespring representation is unique in the following sense
Proposition 2.18 Let 1 and 2 be two representations of A on Hilbert spaces K1 and K2
and let Vi : H Ki be two bounded operators such that
{i (a)Vi u | a A, u H},
is total Ki for i = 1, 2 and such that
T (a) = Vi (a)Vi
for i = 1, 2. Then there exists a unitary map U : K1 K2 such that
U V1 = V2 ,
U 1(a) = 2(a)U
for all a A.
Proof. Let U : K1 K2 be the densely defined linear map defined by
n
n
X
X
U
1 (aj )V1 uj =
2(aj )V2 uj
j=1
j=1
(2.13)
22
for every integer n 1 and a1 , . . . , an A, u1, . . . , un H. A straightforward computation

shows that
(U 1(b)V1 v, U 1(a)V1 u)2 = hv, T (b a)ui = (V1 v, 1(b a)V1 u)1 ,
where (, )j denotes the scalar product in Kj , for all a, b A and u, v H. Therefore U is
an isometry and can be extended to K1 by an obvious density argument. In a similar way
one can prove that also U : K2 K1 is an isometry. Thus U is unitary. Finally, since
U V1 u = U 1(1l)V1 u = 2(1l)V2 u = V2 u,
U 1 (a)V1 u = 2(a)V2 u
for every u H, (2.13) follows.

We finish this section by proving K. Kraus characterisation [64] of -weakly continuous
(i.e. normal) completely positive maps.
Lemma 2.19 Let A and B be two von Neumann algebras of operators on Hilbert spaces H
and K. A normal completely positive map T : A B can be written in the form
T (a) = V (a)V
where V is a bounded operator from K to a Hilbert space K1 and is a normal representation
of A in B(K1 ).
Proof. Let (, V ) be a minimal Stinespring representation of T with V : K K1. We
check that is normal.
Let (x ) be a non-decreasing net of elements of A converging to x (x A) in the
-weak topology. For all vectors u, v K and operators a, b A we have
lim h(b)V v, (x)(a)V ui
= lim hV v, (b xa)V ui
= lim hv, T (b x a)ui
= hv, T (b xa)ui
= h(b)V v, (x)(a)ui
because T is normal. Thus is normal by Proposition 1.15 4.
Theorem 2.20 (Kraus) Let A be a von Neumann algebra of operators on a Hilbert space
H and let K be another Hilbert space. A linear map T : A B(K) is normal and completely
positive if and only if it can be represented in the form
T (a) =
Vj aVj
(2.14)
j=1
where (Vj )
j=1 are bounded operators from K to H such that the series
strongly.
Vj aVj converge
Proof. Clearly a completely positive map of the form (2.14) is normal. Indeed, for every
non-decreasing net (x ) of positive elements of A converging strongly to its least upper
bound x we have
X
sup hu, T (x)ui =
sup hVj u, xVj ui
hVj u, xVj ui
hu, T (x)ui
23
for every u K.
Conversely we can represent the normal completely positive map T in the form T (a) =
V (a)V with normal as in Lemma 2.19. Therefore it suffices to establish (2.14) for the
representation . By decomposing K1 into cyclic orthogonal subspaces we can suppose that
there exists a cyclic vector w for (A) in K1. The state on A
a hw, (a)wi
is normal because is normal. Hence (see, for example, [21] Th. 2.4.21 p.76) there exists a
sequence (un)n1 of vectors in H such that
X
kunk = 1,
hw, (a)wi =
n1
hun , auni
n1
for all a A. Moreover, for every x A and n 1, we have

2
kxunk = hun , (xx)uni hw, (xx)wi = k(x)wk .

Therefore there exist contractions Vn : K1 K such that
Vn (x)w = xun.
For all x A we have also
h(x)w, (a)(x)wi
hw, (xax)wi
X
huj , xaxuj i
hxuj , axuj i
j=1
hVj (x)w, aVj (x)wi
j=1

(x)w, Vj aVj (x)w .
j=1
j=1
This completes the proof because w is cyclic for A.

Remark. It is worth noticing here that Kraus representation (2.14) can also be written in
the form
T (a) = V (a 1l)V
where 1l denotes the identity operator in another Hilbert space H1 and V : K H H1 is a
bounded operator. In fact it suffices to consider an orthonormal basis (ej ) in H1 and define
V u = (Vj u) ej .
The notion of minimality for Kraus representations is obviously a special case of the
analogue for Stinespring representation.
24
2.3
A quantum Feynman-Kac formula
In this section we outline a perturbation technique similar to the Feynman-Kac perturbation

in classical probability. The abstract algebraic generalisation described by Accardi in [1]
shows the importance of the notion of cocycle in the construction of algebraic Markov
processes.
We shall use the notation of Section 2.1. (A,
) is an algebraic probability space with a
filtration (At] )t0; conditional expectations IE | At] will be denoted also by IEt] .
Definition 2.21 A family (t )t0 of -homomorphisms of A is called a covariant shift if:
1. (semigroup property) 0 (a) = a and t (s (a)) = t+s (a) for all s, t 0 and all a A,
2. (left inverse) for all t 0 the map t has a left inverse denoted t , i.e. for all a A,
we have
t (t (a)) = a,
3. (covariance) for all s, t 0 and all a A we have

t IE s (a) | A0] = IE t+s (a) | At] .
It is easy to check that the standard time shift of classical homogeneous Markov process
is a covariant shift according to the above definition. In fact, with the notation of Example
2.1, is defined by
t (f(xt1 , . . . , xtn )) = f(xt+t1 , . . . , xt+tn ).
Consider a family (jt)t0 of -homomorphisms on A which are adapted in the sense that

jt IE[ a | At] ] = IE[ jt (a) | At] ]
(2.15)
for every a A. We now try to find conditions in order that the algebraic process u on
(A, ) with values in A0] defined by
ut () = jt (t ())
(2.16)
is a covariant algebraic Markov process.

The following proposition gives a necessary condition.
Proposition 2.22 Let (jt )t0 be a family of -homomorphisms on A satisfying condition
(2.15). Suppose that the algebraic process u defined by (2.16) is a covariant algebraic Markov
process. Then, for all s, t 0 and all a t+s (A0] ) we have
IE0] [jt+s (a)] = IE0] [jt (t (js (t (a))))] .
(2.17)
Proof. If u is a covariant quantum Markov process then, by Proposition 2.4, the family T
of linear maps defined by (2.3) is a semigroup on A0] . Now, for all s, t 0 and all a A0] ,
denoting the conditional expectation with respect to As] by IEs] , we have

Tt (Ts(a)) = IE0] jt t IE0] [ js (s (a))]

covariance
= IE0] jt IEt] [ t (js (s (a)))]

j is adapted
= IE0] IEt] [ jt (t (js (s (a))))]
projectivity of IE
= IE0] [ jt (t (js (s (a)))) ]
inverse & semigroup
= IE0] [ jt (t js t ) (t+s (a))]
On the other hand Tt+s is defined by
Tt+s (a) = IE0] [jt+s (t+s (a))]
for all a A0] . Therefore, since T is a semigroup, 2.17 holds.
2.3. A QUANTUM FEYNMAN-KAC FORMULA
25
Definition 2.23 A family (jt )t0 of -homomorphisms on A satisfying (2.15) is called a

Markov cocycle with respect to the covariant shift if, for all s, t 0 and all a A0] , we
have
jt+s (t+s (a)) = jt (t (js (s (a)))) .
(2.18)
We prove now the fundamental result of [1].
Theorem 2.24 Let be a covariant shift and let (jt )t0 be a Markov cocycle with respect
to . Then the algebraic process u defined by (2.16) is a covariant algebraic Markov process
on (A, ) with values in A0] .
Proof. The proof of Proposition 2.22 shows that the maps (Tt)t0 are a semigroup. Let
us show first that u is a quantum Markov process. For all a A0] and all s, t 0 we have
IEs] [ut+s(a)]
cocycle property
projectivity of IE
covariance of and IE
= IEs] [jt+s (t+s (a))]

= IEs] [js ((s jt s ) t+s (a))]

= IEs] js IEs] [(s jt s ) t+s (a)]

= IEs] (js s ) IE0] [jt (t (a)))]
= IEs] [(js s ) (Tt(a))]]
= us (Tt(a))
Let us show now that u is covariant; for all a A0] and all s, t 0, we have

ut IE0] [us(a)] = ut (Ts(a)) = IEt] [ut+s(a)]
Remark. The analogy with the classical Feynman-Kac formula becomes clear if we take
a standard Brownian motion (wt)t0 with natural filtration (Ft )t0
Ft = {ws | 0 s t}
as the classical stochastic process x in Example 2.1. Let Mt be the multiplicative functionals
Z t

Mt = exp
c(ws )ds
0
and let
jt(a) = Mt a.
Clearly jt : A A is not an identity preserving homomorphism but it enjoys the cocycle
property (2.18) with respect to the standard (classical) shift (see [2]).
Proposition 2.25 Let H be a complex separable
Hilbert space, let A be the *-algebra B(H)

of all bounded operators on H and let At] t0 a filtration of sub-*-algebras of A. Consider
a family (Vt )t0 of unitary operators on H such that Vt At] and, for all t 0, define the
map
jt : A A, jt (a) = Vt aVt
(2.19)
Suppose that, for all s, t 0, we have
Vt+s = Vt t (Vs )
Then j is a cocycle with respect to .
(2.20)
26
Proof. For all s, t 0 and all a A0] we have

jt t js s (a)
=
=
=
Vt (t (Vs s (a)Vs )) Vt
[Vt t (Vs )] t+s (a) [Vt t (Vs )]

jt+s (t+s (a))
as required.
The family of operators (Vt )t0 defining the cocycle j is called an operator cocycle or,
when no confusion can arise, simply a cocycle.
Quantum Markov Semigroups

Semigroups of completely positive operators are a fundamental tool in the theory of quantum Markov processes. Several results on this class of semigroups - mostly in the uniformly
continuous case - are scattered in the literature. In this chapter we give a self-contained exposition of the basic results on uniformly continuous completely positive semigroups. Moreover
we describe the construction of completely positive semigroups on B(h) that are continuous
in the -weak topology and we give conditions - necessary and sufficient or simply sufficient
- for the semigroup to be identity preserving (i.e. Markov).
3.1
Fundamental definitions and examples
In this section A will denote a W -algebra of operators acting on a Hilbert space H and A
will denote its predual.
Definition 3.1 A quantum dynamical semigroup on A is a family T = (Tt)t0 of bounded
operators on A with the following properties:
1. T0 (a) = a, for all a A,
2. Tt+s (a) = Tt (Ts(a)), for all s, t 0 and all a A,
3. Tt is completely positive for all t 0,
4. Tt is a -weakly continuous operator in A for all t 0,
5. for each a A, the map t Tt(a) is continuous with respect to the -weak topology
on A.
Notice that, when 3 holds, then 4 is equivalent to normality of the maps Tt by Proposition
1.15.
We introduce the usual definition of the infinitesimal generator of T .
Definition 3.2 The infinitesimal generator the quantum dynamical semigroup T is the operator L whose domain D(L) is the vector space of elements a in A for which there exists
an element b of A such that
Tt (a) a
b = lim
t0
t
in the -weak topology, and L(a) = b.
27
28
3. QUANTUM MARKOV SEMIGROUPS

We now give some examples.
Example 3.1 Let A = B(H) and let (Pt)t0 be a strongly continuous semigroup on H.
The family of linear operators Tt : A A defined by
Tt(a) = Pt aPt
is a quantum dynamical semigroup. In fact all the continuity properties of T follow from
Kraus theorem and the strong continuity of P . Let G denote the infinitesimal generator of
P . The infinitesimal generator L of T is given (formally if G is unbounded) by
L(a) = G a + aG.
Notice that, if G is unbounded, then T is not necessarily strongly continuous.
The following example shows that classical Markov semigroups can be viewed as a special
class of quantum dynamical semigroups. However some care must be taken dealing with
topologies.
Example 3.2 Let (X, X ) be a measurable space and let be a -finite measure on X .
Then L (X,C;
I d) is a commutative W -algebra of multiplication operators acting on the
Hilbert space L2(X,C;
I d). The predual is the Banach space L1 (X,C;
I d).
Consider a family (P (t, ; ))t0 of transition probabilities on X X with the following
properties:
1. for all x X, P (t, x; ) is a probability measure on X , absolutely continuous with
respect to for t > 0, and coinciding with the Dirac measure x for t = 0,
2. for all measurable sets A X , P (t, ; A) is a measurable essentially bounded function
on X and the map t P (t, ; A) defined on [0, +[ with values in L (X, X ; d) is
-weakly continuous,
3. (Chapman-Kolmogorov equation) for all t, s 0, x X and A X we have
Z
P (t + s, x; A) =
P (t, y; A)P (s, x; dy).
X
Let (Tt)t0 be the family of operators on L (X,C;

I d)
Z
(Ttf)(x) =
f(y)P (t, x; dy).
X
Then T is a quantum dynamical semigroup in the W -algebra L (X,C;

I d). In fact property
1 of a quantum dynamical semigroup follows from P (0, x; dy) = x (dy) and property 2 follows
from the Chapman-Kolmogorov equation. Complete positivity follows from the fact that
the W -algebra L (X,C;
I d) is commutative and that the linear operators Tt are positive.
The continuity property 5 follows from the -weakly continuity of the map t P (t, ; A)
for every A X . Indeed, writing f as the sum four positive functions and remembering
that the -weak and weak topology coincide on bounded subsets of L (X,C;
I d), we can
easily see that it suffices to show that, for every positive element f of L (X,C;
I d) and
every g L2 (X,C;
I d), the function

Z
Z
Z
2
t g(),
f(y)P (t, ; dy)g() =
|g(x)| d(x)
f(y)P (t, x; dy)
(3.1)
X
3.1. FUNDAMENTAL DEFINITIONS AND EXAMPLES
29
is continuous. This is clearly the case whenever f is a simple function because of the weakly continuity of the maps t P (t, ; A) with A X . Approximating f by an increasing
(resp. decreasing) sequence of simple functions converging to f almost everywhere we see
that (3.1) is lower (resp. upper) semicontinuous. Therefore the function (3.1) is continuous.
In order to prove the continuity property 4, since Tt is positive, by Proposition 1.15
it suffices to check that it is normal. Thus we must show that, for every increasing net
(f ) in L1 (X,C;
I d) converging -weakly to f in L1 (X,C;
I d) and every g L1 (X,C;
I d)
non-negative we have
Z
Z
Z
Z
sup
g(x)d(x)
f (y)P (t, x; dy) =
g(x)d(x)
f(y)P (t, x; dy).
(3.2)
By [21] Lemma 2.4.19 p.76, the net (f ) converges -strongly to f. Therefore, denoting the
Radon-Nikodym derivative of P (t, x; dy) with respect to by p(t, x, ), we have
Z
D
E
sup
f (y)P (t, x; dy) = sup (p(t, x, ))1/2, f()(p(t, x, ))1/2
X
D
E
=
(p(t, x, ))1/2, f()(p(t, x, ))1/2
Z
=
f(y)P (t, x; dy)
X
for every x X. It follows that the right-hand side is the least upper bound of the integrals
of f with respect to P (t, x; dy) and the same argument yields (3.2). This shows that (Tt)t0
is a quantum dynamical semigroup.
It is worth noticing that a large class of classical Markov semigroups enjoys the so-called
Feller property. This means, roughly speaking, that the semigroup can be restricted to a
space of continuous functions and studied as a strongly continuous contractive semigroup.
The original semigroup on L (X, X ; ) is then the unique -weakly continuous extension of
this restriction. Clearly the quantum analogue of the Feller property is the following: there
exists a C -algebra A0 that is dense in A in the -weak operator topology, invariant under
the action of the quantum dynamical semigroup T , and such that the restriction of T to A0
is strongly continuous. We will not be concerned however with the problem of establishing
whether some quantum dynamical semigroup enjoys this property.
The following fact will be useful in the construction of another example of a quantum
dynamical semigroup through a generalization of Example 3.1
Proposition 3.3 Let (X, X ) be a measurable space and let a finite measure on X . Let
(U (t, x))t0,xX be a family of bounded operators on a Hilbert space H such that
1. for all x X the map t U (t, x) is strongly continuous,
2. for all t 0 the map x U (t, x) is strongly measurable,
3. for all t 0 there exists a positive function gt on X, integrable with respect to , such
that
sup kU (s, x)k gt(x)
0st
Then the map : [0, +[B(H) B(H) defined by the integral (in the -weak topology)
Z
(t, a) =
U (t, x) aU (t, x)d(x)
X
is -weakly continuous in both arguments and completely positive in the second.
30
Proof. Let n 1 and a1 , . . . , an, b1, . . . , bn be elements of B(H). Since

2

Z X
n

bj (t, aj ai )bi =
aj U (t, x)bj d(x) 0

X j=1

i,j=1
n
X
the map (t, ) is completely positive by Proposition 2.8.

Note that is uniformly bounded on sets of the form [0, t]B(H) with t fixed. Therefore,
since the -weak and weak topology coincide on bounded sets of B(H), in order to show
that is -weakly continuous in a it suffices to prove that, for every u H, the positive
linear functional on B(H)
Z
a
hU (t, x)u, aU (t, x)ui d(x)
X
is -weakly continuous. To this end, by [21] Th. 2.4.21 p.76, it suffices to show that, for
every increasing net (a ) in B(H) with least upper bound a we have
Z
Z
sup
hU (t, x)u, aU (t, x)ui d(x) =
hU (t, x)u, aU (t, x)ui d(x).
Since (a ) converges strongly to a (see [21] Lemma 2.4.19) for all x X, we have
sup hU (t, x)u, aU (t, x)ui = hU (t, x)u, aU (t, x)ui .
The conclusion then follows since the map on L1 (X, X , )

Z
g
g(x)d(x)
X
is a -weakly continuous functional. (This is in practice a monotone convergence theorem

for increasing nets in L1 (X, X , )).
Finally to show that is -weakly continuous in t it suffices to use the inequality
Z
|hu, ((t, a) (s, a))ui| 2c(r) kak kuk
k(U (t, x) U (s, x))uk d(x)
X
for u H and t, s [0, r] (r > 0 fixed) where c(r) is a positive constant depending only on
r and apply Lebesgues theorem.
The following example is a generalization of Example 3.1 due to Parthasarathy ([74] Example 30.1 p. 258). However the semigroup he constructs there is not a quantum dynamical
semigroup in the sense of his definition (Sect. 30 p.257) since it is not necessarily strongly
continuous. We shall exhibit a counterexample.
Example 3.3 Let (Bt )t0 be a classical Brownian motion on a filtered probability space
(, F, IP ) and let H be a complex separable Hilbert space. For every bounded self-adjoint
operator L on H let us define the map Tt : B(H) B(H)
Z
Tt(a) =
exp(iLBt ())a exp(iLBt ())dIP ()
(3.3)
2
Z
1
x
=
exp(ix tL)a exp(ix tL) exp
dx
2
2 IR
3.1. FUNDAMENTAL DEFINITIONS AND EXAMPLES
31
Since (Bt )t0 has stationary independent increments it can be shown as in [74] Example30.1
p. 258 that (Tt )t0 is a semigroup. Applying Proposition 3.3 with U (t, x) = exp(ix tL)
and equal to the standard gaussian measure on the Borel -algebra of IR it is easy to see
that also the properties 3, 4 and 5 of Definition 3.1 are fulfilled. Hence T is a quantum
dynamical semigroup.
The infinitesimal generator L of T is given (at least formally if L is unbounded) by
L(X) =

1 2
L X 2LXL + XL2 .
2
(3.4)
The quantum dynamical semigroup T admits restrictions to abelian subalgebras of B(H)

which are classical Markov semigroups. Our argument here is only formal since the operator
L is unbounded; we will rigorously deal with this example later.
Let H be the Hilbert space L2(IR;C)
I and let L be the first derivative with its natural
domain. Denote by M (f) the multiplication operator by a bounded smooth function f
M (f) : H H,
(M (f)u)(x) = f(x)u(x).
A straightforward computation shows that

L(M (f)) = M

1 00
.
f
2
This means that the restriction of the semigroup T to the abelian C -algebra of bounded
complex-valued continuous functions on the real line coincides with the classical Markov
semigroup of Brownian motion.
We give now the counterexample showing that T is not necessarily strongly continuous.
Let us fix H = l2 (N) with canonical orthonormal basis (en )n0 and let N and S be the
number operator and right shift operator
X

D(N ) =
uH
n2 |un|2 < + ,
Nu =
nun en
n0
n0
X
D(S) = H,
Su =
unen+1 ,
n0
where u =
n0 un en .
Clearly N 2 is the self-adjoint operator

D(N 2 ) = u H
n4 |un|2 < + ,
n0
N u2 =
n2un en .
(3.5)
n0
Let L = N 2. We now prove that the quantum dynamical semigroup T is not strongly
continuous on B(H).
By virtue of (3.3), for every u H and x B(H) we have
=
=
hu, (Tt(x) x)ui

2
Z

1
y
u, exp(iyN 2 )x exp(iyN 2 ) x u exp

dy
2t
2t IR
2
Z D
E
1
y
u, exp(iy tN 2)x exp(iy tN 2) x u exp

dy
2
2 IR
32
Now let x be the right shift operator S and, for every n 1, let vn be the vector
1
vn =
n
ek .
n<k2n
We have then, for every k 1,
exp(iy tN 2)S exp(iy tN 2)ek = exp(iy t(2k + 1))ek+1 .

Hence we have
=
=
hvn , (Tt(S) S)vn i

Z
E

X D
2
1
ej , exp(iy t(2k + 1) 1 ek+1 ey /2 dy

n 2 IR n<j,k2n
2
Z

X
1
1
y
exp(iy t(2k + 1) 1 exp

dy
n
2
2 IR
n+1<k2n
1
n

exp(t(2k + 1)2/2) 1
n+1<k2n
Therefore we obtain the inequalities

X
|hvn, (Tt(S) S)vn i| =
1
n
1
n
n1
(1 exp(t(2n + 5)2 /2))
n
(1 exp(t(2k + 1)2 /2))
n+1<k2n
(1 exp(t(2n + 5)2 /2))
n+1<k2n
This implies that

kTt(S) Sk lim sup |hvn, (Tt(S) S)vn i| = 1
n
for every t > 0 and shows that t is not strongly continuous.

The above examples and the analogy with classical probability motivate the following
Definition 3.4 A quantum dynamical semigroup is called Markov or conservative or identity preserving (resp. sub-Markov) if
Tt(1l) = 1l,
(resp.Tt(1l) 1l)
for every t 0.
Quantum dynamical semigroup can be viewed as dual semigroups of strongly continuous
semigroups on the predual Banach space A .
Definition 3.5 The predual semigroup of a quantum dynamical semigroup T on A is the
semigroup S of operators in A defined by
(St ()) (a) = (Tt (a))
for every a A and every A .
3.2. UNIFORMLY CONTINUOUS QDS
33
Since T is continuous with respect to the weak topology on A the semigroup S is also
continuous with respect to the weak topology on the Banach space A . Therefore, by a wellknown fact (see, for instance [21] Cor. 3.1.8 p.168), S is a strongly continuous semigroup in
the Banach space A . Moreover the following are well-known results in semigroup theory:
Proposition 3.6 Let T be a quantum dynamical semigroup on A then:
1. there exists two real numbers M, such that
kTtk M exp(t),
t 0,
2. the infinitesimal generator L is densely defined and closed for the -weak topology,
3. if <e > then the range of 1l L coincides with A and we have the inequality

(1l L)1(a)
M
kak ,
<e
4. the resolvent operator (1l L)1 is given by the Laplace transform

Z
(1l L)1 (a) =
dt exp(t)Tt (a)
0
for every a A and every complex number with <e > .

Proof. We refer to Prop. 3.1.6 p. 166 in [21].
3.2
Uniformly continuous QDS
Uniformly continuous quantum dynamical semigroups already form an interesting class, in

contrast to the case of classical uniformly continuous semigroups. We refer to the book [10]
for several physical examples.
Definition 3.7 A quantum dynamical semigroup is called uniformly continuous if
lim kTt T0 k = 0.
t0
The following proposition summarizes the well-known results on uniformly continuous

semigroups
Proposition 3.8 Let T be a semigroup of bounded operators on a Banach space E. The
following conditions are equivalent:
1. the map t Tt is uniformly continuous,
2. the map t Tt is uniformly differentiable,
3. the infinitesimal generator L is a bounded operator and
Tt =
X tn
Ln
n!
n0
the series being uniformly convergent for every real number t.
34
If these conditions are fulfilled then T can be extended to a uniformly continuous group of
operators on E satisfying
kTtk exp (|t| kLk) .
We refer to [21] Prop. 3.1.1 p. 161 for the proof. Notice that, if the infinitesimal
generator is bounded, then T can be extended as a group of bounded operators in E. The
operator Tt(1l) is invertible for t in a neighbourhood of 0. Indeed we have the inequality
kTt (1l) 1lk |t| kLk exp (|t| kLk) .
The right-hand side is smaller than one for t in a neighbourhood of 0. Therefore the inverse
of Tt(1l) can be defined by von Neumanns series
(Tt(1l))
(1l Tt(1l)) .
n=0
1
It then follows that the map t (Tt(1l)) is uniformly differentiable in a neighbourhood

of 0 and
d
(3.6)
(Tt(1l))1 = (Tt(1l))1 Tt(L(1l)) (Tt (1l))1 .
dt
Proposition 3.9 Let T be a uniformly continuous semigroup of bounded operators Tt in
a von Neumann algebra A with infinitesimal generator L. The following conditions are
equivalent:
1. Tt is -weakly continuous for every t 0,
2. L is -weakly continuous.
Proof. The statement clearly follows from Proposition 3.8 and the fact that the set of
-weakly continuous operators on A is norm closed.
To show this notice first that a linear map T on A is -weakly continuous if and only if,
for every -weakly continuous linear functional on A, the linear functional a (T a) is
-weakly continuous. The set of -weakly continuous linear functionals on A (the predual
A of A), however, is a norm closed subspace of the dual Banach space of A by Proposition
2.4.18 p.75 [21]. Moreover, for every sequence (Tn )n0 of -weakly continuous operators on
A converging in norm to T , we have
sup k(Tn a) (T a)k kk kTn T k ,
kak1
i.e. the sequence ((Tn ())n0 of -weakly continuous linear functionals A is convergent in
norm. This completes the proof.
The following property of uniformly continuous groups in A turns out to be useful often.
It was proved in [35] (Cor.3, p. 210). Here we give a slightly different proof.
Proposition 3.10 Let L be a bounded operator on a von Neumann algebra A such that
L(a) = (L(a)) for every a A. The following conditions are equivalent:
1. for all a A and t 0 we have
exp(tL)(a ) exp(tL)(a) exp(tL)(a a),
(3.7)
a L(a) + L(a )a L(a a).
(3.8)
2. for all a A we have
35
Proof. Condition 2. clearly follows from 1. by differentiating at t = 0.

We prove the converse. As a first step notice that for every a, b A such that ab = 0,
from (3.8), we have
b L(a a)b b a L(a)b + bL(a )ab = (ab) L(a)b + b L(a )(ab) = 0.
1
Then we prove that the operator ( L) is non-negative for every > kLk. To this end
it is enough to show that if a is a self-adjoint element of A such that ( L) (a) is nonnegative, then a is non-negative. Let a = x y be the decomposition of a into its positive
and negative parts. Clearly xy = 0. Thus, by the above remark, we have yL(x)y 0.
Therefore we have

0 y 1 1 L (a) y

= yay y 1 L (a)y
=
y3 1yL(x)y + 1 yL(y)y
y3 + 1yL(y)y.
Hence 0 y3 1 yL(y)y. Thus we obtain the inequality kyk3 1kLk kyk3 . This
implies y = 0 since 1 kLk < 1 and a = x is non-negative.
It follows that exp(tL) is non-negative for every t > 0 because
n
exp(tL) = lim (1l (t/n)L)

n
We can show now that the inequality (3.7) holds. Let t 0 fixed. Using condition 2.
and positivity of exp((t s)L) we have
d
exp((t s)L) (exp(sL)(a ) exp((sL)(a))
ds

= exp((t s)L) L (exp(sL)(a ) exp(sL)(a))
+ L(exp(sL)(a )) exp(sL)(a) + exp(sL)(a )L(exp(sL)(a))
0.
Integrating this inequality on [0, t] (t 0) we obtain the inequality (3.7).

The remaining part of this section will be devoted to the study of the structure of the
infinitesimal generator of a quantum dynamical semigroup which is already revealed by the
examples in the previous section.
We introduce first a notion related to complete positivity.
Definition 3.11 A bounded linear operator L on a von Neumann algebra A is called conditionally completely positive if for every integer n 1, the linear map L(n) on A Mn
defined by L(n)(a Eij ) = L(a) Eij ( 1 i, j n) satisfies the inequality
L(n)(x x) x L(n)(x) L(n)(x )x + x L(n)(1l)x 0
(3.9)
for every x A Mn.

The infinitesimal generator of a quantum dynamical semigroup is characterised by conditional complete positivity.
Proposition 3.12 Let T be a uniformly continuous semigroup on a von Neumann algebra
A with infinitesimal generator L. Then Tt is completely positive for every t 0 if and only
if L is conditionally completely positive and L(a ) = (L(a)) for every a A.
36
Proof. Let T be a quantum dynamical semigroup. With the notation introduced in Section
(n)
2.2 the semigroup (Tt )t0 is a uniformly continuous semigroup in A Mn . Therefore the
Schwarz inequality (2.8) in Proposition 2.10

1
(n)
(n)
(n)
(n)
Tt (x ) Tt (1l)
Tt (x) Tt (x x)
holds for every x AMn and every t 0. Differentiating at t = 0 (using (3.6)) we see that
L(n) satisfies (3.9). Hence L is conditionally completely positive. Moreover L(a) = (L(a))
because Tt(a ) = (Tt(a)) for t 0.
In order to prove the converse we may assume first that L(1l) 0. Indeed, if this is not
the case, it suffices to consider the operator L1 = L c where c = kL(1l)k. Note that the
left-hand side of (3.9) is the same for both L and L1 and that
Tt = ect exp(tL1).
Therefore Tt is completely positive if and only if exp(tL1 ) is.
So assume L(1l) 0, then for every n 1, the inequality (3.9) yields
xL(n) (x) + L(n)(x )x L(n)(x x).
(n)
Therefore Proposition 3.10 implies that the operators Tt

are completely positive.
are positive i.e. the operators Tt
Lemma 3.13 Let L be a conditionally completely positive operator on a von Neumann

algebra A. Then for every integer n 1, every family a1 , . . . , an of elements of A and every
family u1, . . . , un of vectors in H such that
n
X
aj uj = 0
j=1
we have
n
X
hui , L(ai aj )uj i 0.
i,j=1
P
Proof. Considering the element x = j aj Ej in A Mn and the vector u = (u1, . . . , un)
in the direct sum of n copies of H we have
D
E
0
u, L(n)(x x) x L(n)(x) L(n)(x )x + x L(n)(1l)x u
=
n
X
i,j=1
n
X
hui , (L(ai aj ) ai L(aj ) L(ai )aj + ai L(1l)aj ) uj i

hui , L(ai aj )uj i .
i,j=1
This proves the lemma.

Theorem 3.14 Suppose that A = B(H). A bounded linear map L on A such that L(a ) =
(L(a)) for every a A is conditionally completely positive if and only if there exists a
completely positive map on A and an element G of A such that
L(a) = (a) + G a + aG
for every a A. Moreover the operator G satisfies the inequality
G + G L(1l).
(3.10)
37
Proof. Let us show first that an operator of the form (3.10) is conditionally completely
positive. Fix an integer n and denote by Gn the operator G 1l in A Mn . For every
x A Mn we have
L(n) (x) = (n)(x) + Gnx + xGn.
A straightforward computation shows that the left-hand side of (3.9) is equal to
(n)(x x) (n)(x )x x (n)(x) + x (n)(1l)x.
Since is completely positive, then (n) is positive for every n 1. Thus, for every t 0,
exp(t(n) ) is positive i.e. exp(t) is completely positive. It follows then from Proposition
3.12 that is conditionally completely positive. This and the above identity imply that L
is conditionally completely positive.
Clearly L(a ) = (L(a)) because is positive.
Conversely given a conditionally completely positive L such that L(a ) = (L(a)) for
a A fix a unit vector e in H and consider the operator G in H with adjoint operator
defined by
1
G u = L (|uihe|) e he, L (|eihe|) ei u
(3.11)
2
for u H. For every n 1 and every a1 , . . . , an A, u1, . . . , un H let
un+1 = e,
v=
n
X
aj uj ,
an+1 = |vihe|.
j=1
Clearly we have
Pn+1
j=1
aj uj = 0. Applying Lemma 3.13, we obtain the inequality

n
X
hui, L(ai aj )uj i +
i,j=1
n
X
n
X
hui , L(|ai vihe|)ei
i=1

2
e, L(|eihaj v|)uj + he, L (|eihe|) ei kvk .
j=1
By virtue of (3.11) the sum of the last three terms can be written in the form
n
X
hui, Gai vi +
i=1
n
X

G aj v, uj
j=1
n
X
hui, Gai aj uj i
i,j=1
n
X
hui , ai aj Guj i .
i,j=1
Therefore we obtain the inequality

n
X
hui , (L(ai aj ) Gai aj ai aj G) uj i 0.
i,j=1
Proposition 2.9 then implies that the map defined by

(a) = L(a) G a aG
is completely positive.
Finally, the operator G in (3.10) satisfies the inequality
G + G = L(1l) (1l) L(1l).
38

Remark. The choice of the operator G is clearly not unique. In fact, for every c > 0, taking
G0 = G c1l we can write L in the form
L(a) = ( + 2c1l)(a) + G0a + aG0.
It is worth mentioning here that Christensen and Evans [30] proved Theorem 3.14 for
conditionally completely positive maps on an arbitrary von Neumann algebra. However the
proof is much more difficult.
We are now in a position to prove the characterisation of the infinitesimal generator of
a quantum dynamical semigroup due to Lindblad [65] in the case of an arbitrary Hilbert
space H and to Gorini, Kossakowski and Sudarshan [55] in the case of a finite dimensional
Hilbert space.
Theorem 3.15 (Lindblad) A bounded operator L on B(H) is the infinitesimal generator of
a uniformly continuous quantum dynamical semigroup if and only if there exists a complex
separable Hilbert space K, a bounded operator L : H H K and an operator G in H such
that
L(a) = L (a 1l)L + Ga + aG
(3.12)
for all a B(H). The operator L can be chosen so that the set
{(a 1l)Lu | a B(H), u H}
is total in H K.
Proof. If T is a quantum dynamical semigroup then the infinitesimal generator L is conditionally completely positive by Proposition 3.12. Hence it can be represented in the form
(3.10) by Theorem 3.14. Moreover L is -weakly continuous by Proposition 3.9. Since the
map a aG + G a is obviously -weakly continuous, applying Kraus Theorem 2.20 to the
map we obtain the desired representation of L.
Conversely, if the operator L can be represented in the form (3.12), then it is -weakly
continuous and conditionally completely positive by Theorem 3.14. Therefore it is the
infinitesimal generator of a uniformly continuous quantum dynamical semigroup by Proposition 3.9 and Proposition 3.12.
Remark. Notice that, when the operators Tt are not continuous in the -weak topology, we
can obtain a similar characterization of L applying Stinesprings theorem instead of Kraus
theorem.
Lindblads theorem also allows us to characterize the infinitesimal generators of contractive quantum dynamical semigroups. These can be represented in the form (3.12) with
operators G and L satisfying the inequality
L(1l) = G + G + L L 0.
This remark is the starting point in the construction of quantum dynamical semigroup which
are not necessarily norm continuous.
3.3. MINIMAL QUANTUM DYNAMICAL SEMIGROUP
3.3
39
Minimal quantum dynamical semigroup
The class of uniformly continuous quantum dynamical semigroups is too small for the applications in quantum probability and mathematical physics (see, for example, [10], [32], [46],
[69]). The problem of constructing quantum dynamical semigroups with unbounded generator, in principle, could be treated with the Hille-Yosida theorem at least in the case when
the domain of the infinitesimal generator is an algebra so that conditional complete positivity makes sense. However in all the applications the infinitesimal generator L is not given
explicitly but it is given formally in a generalised Lindblad form (3.12) with unbounded
operators G and L.
E.B. Davies, following Katos paper [62], constructed in [32] the predual semigroup of a
quantum dynamical semigroup on the von Neumann algebra B(h) from given operators G
and L in H.
A.M. Chebotarev (see [22], [23], [24]) constructed directly a quantum dynamical semigroup from the operators G and L in H generalising K.L. Chungs construction of the
minimal solution of Feller-Kolmogorov equations for countable state Markov chains (see [31]
Th.1 p.231 and also [49]).
Here we follow Chebotarevs construction of the minimal solution. The only assumption
we make on the operators G and L in this section is the following:
Hypothesis A
the operator G is the infinitesimal generator of a strongly continuous contraction semigroup (Pt )t0 in h,
the domain of the operators (L` )
`=1 contains the domain of G and, for every u D(G)
we have
X
hu, Gui + hGu, ui +
hL` u, Lùi 0.
(3.13)
`=1
Notice that the operators L` are not assumed to be closable.

Lemma 3.16 Suppose that the hypothesis A holds. Then for every ` 1, the following
conditions hold
1. the maps t 7 L` P (t)u, are norm continuous for every u D(G),
2. the maps t 7 L` P (t)u are norm differentiable for every u D(G2 ) and
d
kL` P (t)uk2 = L` P (t)Gu,
dt
3. the map t 7
`=1
kyL` P (t)uk2 is differentiable for every u D(G2 ) and y B(h)
`=1
`=1
X
d X
kyL` P (t)uk2 = 2<e
hyL` P (t)u, yL` P (t)Gui .
dt
Proof. For each s, t 0 and each u D(G), the inequality (3.13) yields
kL` P (t)u L` P (s)uk 2<e h(P (t) P (s))u, G(P (t) P (s))ui
`=1
Therefore 1 follows from continuity of the map t P (t)Gu.
40

For each s, t and each u D(G2 ) we have

1
s L` (P (t + s) P (t))u L` P (t)Gu 2

2<e s1 (P (t + s) P (t))u P (t)Gu, G(s1(P (t + s) P (t))u P (t)Gu)

= 2<e P (t)(s1 (P (s) 1l)u Gu), P (t)(s1(P (s) 1l)Gu G2u) .
Since u D(G2 ) we can let s tend to 0 to get the desired identity.

For each s, t and each u D(G2 ) we have
kyL` P (t + s)uk
`=1
kyL` P (t)uk
`=1
kyL` P (t + s)u yL` P (t)uk
`=1
2<e
hyL` P (t)u, yL`(P (t + s) P (t))Gui
`=1
kyL` P (t + s)u yL` P (t))uk
`=1
+
+

Z t+s

X
yL` P (t)u, yL` (P (r) P (t))Gu dr
2<e
t
`=1
2s<e
`=1
The first term is an infinitesimum of order bigger than s. Indeed, by the hypothesis A, it
can be majorized by
2kyk2 <e h(P (t + s) P (t))u, (P (t + s) P (t))Gui
Z t+s
Z t+s

2
P (r)G2u dr
kyk
kP (r)Guk dr
t
The second term is also an infinitesimum of order bigger than s because its modulus can be
majorized by
2
2kyk
kL` P (t)uk
!1/2
`=1
Z

X

L`

`=1
t+s
t
2 !1/2

(P (r) P (t)) Gu dr
.

The inequality (3.13) allows then to estimate the square of the only factor depending on s
by
Z t+s

Z t+s
2<e
(P (r) P (t))Gu dr,
(P (r) P (t))G2u dr
t
t+s
k(P (r) P (t))Guk dr

t
t+s
t

(P (r) P (t))G2u dr.
Therefore we have
1
lim
s0 s
kyL` P (t + s)uk
`=1
This proves 3.
X
`=1
kyL` P (t)uk
= 2<e
X
`=1
41
Lemma 3.17 Let E = { (t, s) IR2 | 0 s t } and let f : E IR be a continuous

function such that its partial derivative with respect to the first variable is also continuous.
Then
Z t
Z t
f(t, s)
f(t, s)ds = f(t, t) +
ds.
t 0
t
0
The proof is elementary. We omit it.
We shall denote by k k the norm in B(h).
For all x B(h) consider the sesquilinear form
L(x) in h with domain D(G) D(G)
given by
X
hv,
L(x)ui = hv, xGui + hGv, xui +
hL` v, xL` ui
(3.14)
`=1
We want to construct a quantum dynamical semigroup satisfying the equation

Z t
hv, Tt (x)ui = hv, xui +
hv,
L (Ts (x)) ui ds
(3.15)
Proposition 3.18 Suppose that the hypothesis A holds and, for all x B(h), let (Tt(x))t0
be a -weakly continuous family of elements of B(h) such that
kTt(x)k kxk .
The following conditions are equivalent:
1. equation (3.15) holds for all v, u D(G),
2. for all v, u D(G) we have
hv, Tt(x)ui
=
+
hP (t)v, xP (t)ui
Z t
X
hL` P (t s)v, Ts (x)L` P (t s)ui ds.
`=1
(3.16)
Proof. We remind first that D(G2 ) is a core for G.

In order to show that 1 implies 2 we fix t and compute the derivative
X
d
hL` P (t s)v, Ts (x)L` P (t s)ui
hP (t s)v, Ts (x)P (t s)ui =
ds
`=1
using equation (3.15). Clearly (3.16) follows integrating the above identity on the interval
[0, t].
We prove now that condition 2 implies condition 1. For each v, u D(G2 ), by using
Lemma 3.16 3, Lemma 3.17 and the complex polarisation identity, we compute the derivative
d
hv, Tt(x)ui =
dt
hP (t)v, xP (t)Gui + hP (t)Gv, xP (t)ui +
hL` v, Tt (x)L` ui
`=1
Z tX
hL` P (t s)v, Ts (x)L` P (t s)Gui ds
0 `=1
Z tX
hL` P (t s)Gv, Ts(x)L` P (t s)ui ds.
0 `=1
42
The sum of the first and fourth (resp. second and fifth) term equals
hv, Tt(x)Gui ,
(resp. hGv, Tt(x)ui)
because of (3.16). The right-hand side of the above equation thus coincides with hv,
L(Tt(x))ui.
Therefore (3.15), for v, u D(G2 ), follows by integrating on [0, t]. Since D(G2 ) is a core for
G the proof is complete.
We construct now a solution of (3.16) by iteration.
Proposition 3.19 Suppose that the hypothesis A holds. Then there exists a sequence
(n)
(Tt )n0 of linear contractions on B(h) which satisfy
D
E
(0)
v, Tt (x)u
= hP (t)v, xP (t)ui
D
E
(n+1)
v, Tt
(x)u
= hP (t)v, xP (t)ui
(3.17)
Z
E
X tD
+
L` P (t s)v, Ts(n)(x)L` P (t s)u ds
`=1
for t 0, x B(h), u, v D(G). Furthermore:

(n)
: B(h) B(h) is completely positive for every t 0 and every n 0,
(n)
: B(h) B(h) is normal for every t 0 and every n 0,
1. the map Tt
2. the map Tt
(n)
3. the map t Tt
(x) is -weakly continuous for every x B(h),
(n)
4. the sequence (Tt

(n)
5. Tt
(x))n0 is increasing for every positive x B(h),
(1l) 1l for every n 0, t 0.

(0)
Proof. Clearly the linear maps Tt are well defined and enjoy all the above properties (see
Example 3.1).
(n)
Suppose that the linear maps Tt are contractions on B(h) satisfying 1 2 and 3 for a
(n+1)
fixed integer n. Let us establish that the maps Tt
enjoy the same properties.
For every x B(h) and v, u D(G) the integral in the right hand side of (3.17) is well
(n)
defined because of the properties of maps Tt . Moreover, by using the Schwarz inequality,
we have

Z tD
X
E

L` P (t s)v, Ts(n)(x)L` P (t s)u ds

`=1 0
Z t
X
kxk
kL` P (t s)vk kL` P (t s)uk ds
`=1
kxk
Z
X
`=1
! 12
kL` P (t s)vk ds
0
Z
X
`=1
! 12
kL` P (t s)uk ds
By virtue of the inequality (3.13) in hypothesis A, the square of the third factor in the
right-hand side can be estimated by the square root of
Z t
Z t
d
2
2<e hP (t s)u, GP (t s)ui ds =
kP (t s)k ds
dt
0
0
=
kuk kP (t)uk .
(3.18)
43
Hence, due to the elementary inequality

(a2 b2 )1/2(c2 d2 )1/2 ac bd,
for every non negative real numbers such that a b and c d we have the estimate
Z

X t D
E

(n)
L` P (t s)v, Ts (x)L` P (t s)u ds

0
`=1
kxk (kvk kuk kP (t)vk kP (t)uk)
Since the scalar product hP (t)v, xP (t)ui can be obviously estimated by kxk kP (t)vk
kP (t)uk, the right-hand side of (3.17) is majorized by
kxk kvk kuk.
(n+1)
Therefore it defines a contractive operator Tt

in B(h).
(n+1)
The operator Tt
is completely positive. In fact, for every integer m 1, vectors
u1, . . . , um D(G) and operators x1, . . . , xn B(h) we have
m D
X
(n+1)
u i , Tt
i,j=1
m X
Z tD
X
i,j=1 `=1
m
E
X
(xi xj )uj =
hP (t)ui, xi xj P (t)uj i
i,j=1
E
L` P (t s)ui , Ts(n)(xi xj )L` P (t s)uj ds.
The right-hand side is positive due to the induction hypothesis. Therefore the operator
(n+1)
Tt
is completely positive by Proposition 2.9.
We prove now, again by induction, that the property 2. Clearly 2 holds for n = 0.
Suppose that it has been established for an integer n. In order to prove it for the integer
n + 1 it suffices to show that, for every u D(G) and every increasing net (x) of positive
elements of B(h) converging to an element x of B(h) the supremum over of
Z tD
X
`=1
coincides with
E
L` P (t s)u, Ts(n) (x)L` P (t s)u ds
Z tD
X
`=1
E
L` P (t s)u, Ts(n) (x)L` P (t s)u ds.
Since all the terms in the above sums are positive and the series is convergent we can
exchange the supremum and summation over `. Thus it suffices to show the above property
for each term of the sum.
The family of positive continuous functions on [0, t] (with parameter )
D
E
s L` P (t s)u, Ts(n) (x)L` P (t s)u
is increasing and converges pointwise to the continuous function
D
E
s L` P (t s)u, Ts(n)(x)L` P (t s)u .
Dinis lemma implies then that it converges uniformly. Thus we can exchange the supremum
over and the integral on [0, t] to obtain the desired property for the integer n + 1.
44

In a similar way, using the induction hypothesis, it is easy to see that the operators
enjoy properties the property 3.
Property 4 follows by induction from the identity
D
E
(n+1)
(n)
u, Tt
(x) Tt (x) u
Z tD

E
X
=
L` P (t s)u, Ts(n)(x) Ts(n1)(x) L` P (t s)u ds
(n+1)
Tt
`=1
for every n 1.
Finally, notice that the operators Tt are contractive so that kTt(1l)k 1. Thus, since
they are also positive, we have 0 Tt(1l) 1l This completes the proof.
Lemma 3.20 Suppose that the hypothesis A holds. Then there exists a family (Tt)0 of
contractive linear maps on B(h) such that:
1. Tt is completely positive for every t 0,
2. Tt is normal for every t 0,
3. the family (Tt (x))t0 of linear operators on B(h) solves equations (3.15) and (3.16)
for all x B(h),
4. for every x B(h) the map t Tt(x) is continuous with respect to the -weak topology
on B(h).
(n)
Proof. Fix t 0 and let (Tt )n0 be the sequence of positive linear operators on B(h)
constructed in Proposition 3.19.
(n)
For every positive operator x B(h) and every t 0 the sequence (Tt (x))n0 is
increasing and bounded from above by kxk 1l. Therefore the limit
D
E
(n)
lim u, Tt (x)u
n
exists for every u h.

By the complex polarisation formula,
4 hv, Tt(x)ui =
3
X

ik v + ik u, Tt(x)(v + ik u)
k=0
it follows that the limit

lim
(n)
v, Tt
(x)u
(3.19)
exists for every v, u h.

Writing an arbitrary element x of B(h) first as the sum
x=
x + x
x x
+i
2
2i
of two self-adjoint operators in B(h) and then writing each one of these self-adjoint operators
as the difference of their positive and negative part we can easily see that the limit (3.19)
(n)
exists for every v, u h and every x B(h). Moreover, since the operators Tt
are
contractions, we have

D
E

(n)
lim v, Tt (x)u kxk kvk kuk.
n
45
It follows that there exist contractions Tt in B(h) such that

D
hv, Tt(x)ui = lim
(n)
v, Tt
E
(x)u .
(3.20)
Clearly the maps Tt and property 1 then follows from Proposition 2.15.
In order to prove 2 consider an increasing net (x ) of positive operators in B(h) with
(n)
least upper bound x. Since the maps Tt (n 0) are -weakly continuous, for every u h
we have
D
E
(n)
sup hu, Tt(x)ui = sup sup u, Tt (x)u
n
D
E
(n)
= sup sup u, Tt (x)u
n
D
E
(n)
= sup u, Tt (x)u
n
= hu, Tt(x)ui .
This shows that the maps Tt are normal.
Letting n tend to infinity in (3.17) it follows that equation (3.16) holds. Equation (3.15)
also holds because of Proposition 3.18.
Finally, for every u, v h and x B(h), the map
t hv, Tt(x)ui
is continuous because of equation (3.16). Therefore the uniformly bounded map t Tt(x)
is continuous with respect to the weak and -weak topology on B(h), completing the proof.
The following Lemma, which can be considered as a quantum version of a classical
formula (see [31] Ch.2, Sect. 18, p. 230), is a fundamental tool in the proof of the semigroup
property for the operators (Tt )t0 constructed in Lemma 3.20.
(n)
Lemma 3.21 Suppose that the hypothesis A holds. Let (Wt )n0 be the sequence of positive linear maps on B(h) defined inductively by
Wt
(0)
Tt
(n+1)
Tt
Wt
(0)
(n+1)
(n)
Tt
Then, for each n 0, the following identity holds

(n)
Wt+s (x) =
n
X
(k)
Wt

Ws(nk)(x) .
(3.21)
k=0
Proof. Clearly from (3.17), we have

D
(n+1)
v, Wt
(x)u
Z tD
X
`=1
for v, u D(G), x B(h) and n 0.
E
L` P (t s)v, Ws(n) (x)L` P (t s)u ds
(3.22)
46
Identity (3.21) clearly holds for n = 0. Assume that it has been established for a given
n, then, using (3.22), for every v, u D(G), x B(h), we have
n+1
XD
(k)
v, Wt
k=0
Z tD
n X
X
(0)
v, Wt
E
Ws(n+1k)(x) u
E
Ws(n+1) (x) u

E
L` P (t r)v, Wr(k) Ws(nk) (x) L` P (t r)u dr
k=0 `=1
Identity (3.21) for the given n and the semigroup property of (P (t))t0 yield then
n+1
XD
(k)
v, Wt
E
Ws(n+1k)(x) u
k=0

E
(0)
v, Wt
Ws(n+1) (x) u
Z tD
E
X
(n)
L` P (t r)v, Ws+r (x)L` P (t r)u dr
`=1
`=1
E
P (t)v, Ws(n+1) (x)P (t)u
Z t+s D
E
X
L` P (t + s r)v, Wr(n) (x)L` P (t + s r)u dr
E
P (t)v, Ws(n+1) (x)P (t)u
Z sD
E
X
L` P (s r)P (t)v, Wr(n) (x)L` P (s r)P (t)u dr
`=1 0
Z t+s
X
`=1
E
L` P (t + s r)v, Wr(n) (x)L` P (t + s r)u dr
The
sum of the
D
E first two terms vanishes and the third one is equal to the scalar product
(n)
v, Wt+s (x)u by virtue of (3.22). This proves identity (3.21).
Theorem 3.22 Suppose that the hypothesis A holds. Then there exists a quantum dynam(min)
ical semigroup (Tt
)t0 solving equations (3.16), (3.15) with the following properties:
(min)
1. Tt
(1l) 1l for every t 0,
2. for every -weakly continuous family (Tt)t0 of positive maps on B(h) satisfying (3.16)
and (3.15) and every positive operator x B(h), we have
(min)
Tt
(x) Tt (x)
(3.23)
for all t 0.
(min)
)t0 be the family of linear maps constructed in Lemma 3.20. In order

Proof. Let (Tt
to show that it is a quantum dynamical semigroup clearly it suffices to prove the semigroup
property
(min)
(min)
Tt+s (x) = Tt
(Ts(min) (x))
(3.24)
47
for x B(h) and t, s 0. By Lemma 3.21 we have the identities

D
(n)
u, Tt+s (x)u
n D
X
(k)
u, Wt+s(x)u
k=0
k D
n X
X
(j)
u, Wt

E
Ws(kj) (x) u
k=0 j=0
for every u h and every positive operator x in B(h). Exchanging summations over k and
j we obtain
D
(n)
u, Tt+s (x)u
n D
X
(j)
u, Wt
E
Ts(nj)(x) u .
j=0
(j)
Since the maps Wt
are -weakly continuous, we can let n tend to infinity to get
(min)
u, Tt+s (x)u
D
X
(j)
u, Wt
j=0
(min)
u, Tt
E
Ts(min) (x) u
E
Ts(min) (x) u .
By complex polarization we have also

D

E D
E
(min)
(min)
v, Tt+s (x)u = v, Tt
Ts(min) (x) u
for every v, u h. Therefore, writing an arbitrary x B(h) as a linear combination of
positive self-adjoint operators in B(h) we obtain the semigroup law (3.24).
This proves that T (min) is a quantum dynamical semigroup. Property 1. clearly follows
from condition 5 in Proposition 3.19.
Finally, for every -weakly continuous family (Tt)t0 of positive maps on B(h) satisfying
(3.16) and (3.15) and every positive operator x B(h), we have
(0)
Tt (x) P (t) xP (t) = Tt (x)

and, for n 0, u h,
D
=

E
(n+1)
u, Tt(x) Tt
(x) u
Z tD

E
X
L` P (t s)u, Ts(x) Ts(n)(x) L` P (t s)u ds.
`=1
Therefore a straightforward induction argument yields the inequality

(n)
Tt
(x) Tt (x)
for every n 0. This implies (3.23) by construction of T (min) .

The following corollary (see [40] Corollary 4.4 p. 159, [19] Remark 2.1 p.91) gives a
simple and useful property of Markov minimal quantum dynamical semigroups.
Corollary 3.23 Suppose that the hypothesis A holds and that the minimal quantum dynamical semigroup T (min) is Markov. Then it is the unique -weakly continuous family (Tt)t0
of positive maps on B(h) satisfying (3.15).
48
Proof. Let T be -weakly continuous family of positive maps satisfying (3.15). Applying
Theorem 3.22, for every x B(h) such that 0 x 1l and every t 0, we have
(min)
Tt
(x)
Tt (x)
= Tt (1l) Tt(1l x)
1l Tt(1l x)
(min)
1l Tt
=
(1l x)
(min)
Tt
(x).
Since every y B(h) can be written as a linear combination of four such operators x, it
(min)
follows that Tt
(y) = Tt (y) for every y B(h).
3.4
The resolvent of the minimal semigroup
In the previous section we constructed the minimal quantum dynamical semigroup associated
with operators G, L` satisfying the hypothesis A. This semigroup satisfies the equations
(3.15) and (3.16) but, in general, it is not characterised by this property. Moreover we
would like identify the domain of the infinitesimal generator of T (min) knowing only the
operators G, L` .
As a first step in the study of these problems we give a useful explicit representation of
the resolvent of the minimal quantum dynamical semigroup obtained by A.M. Chebotarev
(see [25], Lemma 4).
We follow the article [28] Sect. 3 (see also [26] Sect. 2)
Proposition 3.24 Suppose that the hypothesis A holds. The linear positive maps P :
B(h) B(h) and Q : B(h) B(h) defined by
Z
hv, P(x)ui =
exp(s) hP (s)v, xP (s)ui ds
0
hv, Q(x)ui
Z
X
`=1
exp(s) hL` P (s)v, xL` P (s)ui ds
(3.25)
for > 0 and x B(h), v, u D(G) are normal and completely positive. Moreover we have
kP k 1 ,
kQ k 1.
Proof. The map P is well defined, normal and completely positive by Proposition 3.3.
Moreover, for every u, v h and x B(h), we have the estimates
Z
|hv, P(x)ui| kxk
exp(s)kP (s)vk kP (s)ukds
0
kxk
Z
1/2 Z
kvk ds
1/2
kuk ds
2
1 kxk kvk kuk.
The linear map Q is also well defined and contractive. In fact the modulus of the
right-hand side of (3.25) can be estimated by kxk times
Z
X
eskL` P (s)vk kL` P (s)uk ds
`=1
3.4. THE RESOLVENT OF THE MINIMAL SEMIGROUP
Z
X
`=1
!1/2
kL` P (s)vk ds
Z
X
`=1
49
!1/2
kL` P (s)uk ds
Estimating each integral as in (3.18) we obtain

|hv, Q (x)ui| kxk kvk kuk.
This proves that Q is well defined and contractive. The argument of the proof of Proposition
3.19 shows that it is also normal and completely positive.
(min)
The resolvent (R
)>0 of the minimal quantum dynamical semigroup, which is characterised by the equation
D
E Z
D
E
(min)
v, R
(x)u =
exp(s) v, Ts(min) (x)u ds,
0
with x B(h) and v, u h, admits the following representation (see [28] Th. 3.1):
Theorem 3.25 For every > 0 and x B(h) we have
(min)
(x) =
Qk (P(x))
(3.26)
k=0
the series being convergent for the strong operator topology.

(n)
(n)
Proof. Let (R )n0 be the sequence of linear monotone maps R : B(h) B(h) given
by
D
E Z
D
E
(n)
v, R (x)u =
exp(s) v, Ts(n)(x)u ds
0
where the maps
(n)
Ts
(n)
are defined by (3.17). Since the maps Ts
(n)
are contractions

R
(n)
well defined. Moreover, for all positive element x of B(h), the sequence R (x)
is
is
n0
non-decreasing. Therefore, by the definition of minimal quantum dynamical semigroup, for

all u h we have
D
E D
E Z
D
E
(n)
(min)
sup u, R (x)u = u, R
(x)u =
exp(s) u, Ts(min) (x)u ds.
n0
Equation (3.17) yields

D
E Z
(n+1)
u, R
(x)u =
+
Z
X
`=1
etdt
Z tD
et hP (t)v, xP (t)ui dt
0
E
L` P (t s)u, Ts(n) (x)L` P (t s)u ds
for all u, v D(G). By the change of variables (r, s) = (t s, s) in the above double integral
we have
D
E
(n+1)
u, R
(x)u = hu, P(x)ui
(3.27)
Z
Z
D
E
X
+
er dr
es L` P (r)u, Ts(n)(x)L` P (r)u ds
`=1
= hu, P(x)ui +
XZ
`1
D
E
(n)
er L` P (r)u, R (x)L` P (r)u dr.
50
Therefore we obtain the recursion formula

(n+1)
(n)
(x) = P(x) + Q (R (x)).
Iterating n times this equation we have

(n+1)
(x) =
n+1
X
Qk (P (x))
k=0
and (3.26) follows letting n tend to +. Formula (3.26) also holds for an arbitrary element
of B(h) since each bounded operator can be written as a linear combination of positive
self-adjoint operators.
The domain of the infinitesimal generator of the minimal quantum dynamical semigroup
(min)
T (min) coincides with the range of the resolvent R
by well-known results on semigroups.
Therefore the above result also characterises the domain of the infinitesimal generator of
T (min) .
We will give a simpler characterisation of this domain for Markov quantum dynamical
semigroups later in Proposition 3.33.
3.5
Conservativity
The minimal quantum dynamical semigroup T (min) we constructed in Section 3.3 is a quantum subMarkov semigroup.
Let S (min) be the predual semigroup on the Banach space of trace class operators in h
(see Definition 3.5) of T (min) . The following proposition is immediately proved.
Proposition 3.26 Suppose that the hypothesis A holds. Then, for every t 0, the following
(min)
1. Tt
(1l) = 1l,

(min)
2. tr St
() = tr() for every trace class operator in h.
Proof. Notice that

(min)
(min)
(min)
tr St
() = tr St
()1l = tr Tt
(1l) .
The equivalence of 1 and 2 is now clear.
In order T (min) to be Markov, i.e. conservative or identity preserving, it is necessary
that
E
d D

(min)
hv,
L(1l)ui =
(1l)u
=0
v, Tt
dt
t=0
for every v, u D(G). In other words the left-hand side of (3.13) must vanish. We state
this property as a stronger form of hypothesis A.
Hypothesis AA
the operator G is the infinitesimal generator of a strongly continuous contraction semigroup (Pt)t0 in H,
3.5. CONSERVATIVITY
51
the domain of the operators (L` )

`=1 contains the domain of G and, for every u D(G)
we have
X
hu, Gui + hGu, ui +
hL` u, Lùi = 0.
(3.28)
`=1
This hypothesis unfortunately is not sufficient for the minimal quantum dynamical semigroup to be identity preserving. We shall give some examples later.
In order to study conditions equivalent to conservativity we prove first a useful identity
Proposition 3.27 Suppose that hypothesis AA holds and fix > 0. For all n 1 we have
n
X
Qk (P (1l)) + Qn+1
(1l) = 1l.
(3.29)
k=0
Proof. For all u D(G) a standard computation yields

Z
X
`=1
kL` P (t)uk dt = 2<e
et hP (t)u, GP (t)ui dt
0
d
2
et kP (t)uk dt
dt
0
Z
2
2
= kuk
et kP (t)uk dt
Therefore we have
P (1l) + Q (1l) = 1l.
This proves (3.29) for n = 0. Suppose it has been established for an integer n. Applying
the map Q to both sides of (3.29) yields
n+1
X
Qk(P (1l)) + Qn+2

(1l) = Q (1l) = 1l P (1l).
k=1
This proves (3.29) for the integer n + 1 and completes the proof.
The representation formula (3.26) for the resolvent of the minimal quantum dynamical
semigroup allows us to prove a necessary and sufficient condition for conservativity obtained
by A.M. Chebotarev in [22] Th. 3.2 (see also [23], [39] Prop. 3.5, 3.6, [26], Prop. 2.7).
Theorem 3.28 Suppose that the hypothesis AA holds and let > 0 fixed. Then the following conditions are equivalent:
1. the minimal quantum dynamical semigroup T (min) is Markov,
2. s limn Qn
(1l) = 0,
3. there exists no non-zero x B(h) such that Q (x) = x.
Proof. The sequence of positive operators (Qn
(1l))n0 is decreasing. In fact (3.29) yields
n+1
Qn
(1l) = Q (P (1l)).
(1l) Q
Therefore it is strongly convergent to a positive operator x. Letting n tend to + in (3.29),

we have
(min)
R
(1l) + x = 1l.
52

(min)
(1l) coincides with 1l. Therefore the

Now condition 1 can be alternatively stated as: R
equivalence of conditions 1 and 2 follows.
Condition 2 implies condition 3. In fact, for every x B(h), which is a fixed point for
the map Q we have
Q (x ) = (Q (x)) = x .
Therefore both x + x and i(x x) are fixed point for the map Q . Applying Qn
to the
inequalities
2kxk 1l x + x 2kxk 1l,
2kxk1l i(x x) 2kxk1l,
we have
n
2kxkQn
(1l) x + x 2kxk Q (1l),
n
2kxkQn
(1l) i(x x ) 2kxk Q (1l).
Letting n tend to infinity we obtain x = 0.

Finally condition 3 implies condition 2 since, if the decreasing sequence of operators
(Qn
(1l))n0 converges strongly to a non-zero operator x, then we have

n
n
x = s lim Qn
(1l) = w lim Q (Q (1l)) = Q w lim Q (1l) = Q (x).
n
Therefore condition 3 does not hold.

(min)
We give now an example in which the inequality Tt

(1l) < 1l holds for some t > 0.
This establishes a relationship with a problem in classical probability: escape at infinity in
finite time of trajectories of a continuous time Markov process. This example, which is a
generalization of Example 3.3 p. 174 in [32], was essentially considered in [38] Sect. 2.
Example 3.4 Let h be the Hilbert space l2 (IN ) of square-summable complex-valued sequences and let ((n))n0 be a complex sequence such that (n) 6= 0 for every n 0. Let
S be the right shift operator on h defined on the canonical orthonormal basis (en )n0 by
Sen = en+1 , for n 0. Consider the operator (N ) in h defined by
X

D((N )) = u h
|(n)|2|un|2 < ,
(N )u =
(n)unen
n0
n0
Let G and L1 be the operators with domain D(|(N )|2 )

1
G = |(N )|2,
2
L1 = S(N )
and let L` = 0 for every ` 2. Since S is an isometry the hypothesis AA holds.

A straightforward computation shows that the form
L(x) given by (3.14) applied to a
multiplication operator M (f) by a function f on IN yields
L(M (f)) = M (Af)

where
(Af)(n) = |(n)|2(f(n + 1) f(n)).
Thus A coincides (on good functions) with the infinitesimal generator of a classical purebirth process.
Let T (min) be the minimal quantum dynamical semigroup associated with G and L.
3.5. CONSERVATIVITY
53
In this case we can compute explicitly the operators Qn

1 (1l). In fact, for every n, m 0
and every k 1 we have
k1
Y

em , Qk1 (1l)en = hem , en i
j=0
|(n + j)|2
.
1 + |(n + j)|2
(3.30)
This can be proved by induction. Formula (3.30) is true for k = 1 because

Z
hem , Q1(1l)en i = hem , en i
es|(n)|2 exp(|(n)|2s)ds
0
hem , en i
|(n)|2
.
1 + |(n)|2
A similar explicit computation allows us to complete the induction argument.

We can now give a necessary and sufficient condition on the sequence ((n))n0 in order
the minimal quantum dynamical semigroup T (min) to be identity preserving.
Proposition 3.29 The following conditions are equivalent:
(min)
1. Tt
(1l) = 1l for every t 0,

P
2. the series n0 |(n)|2 diverges.
Proof. By virtue of Theorem 3.28 it suffices to show that condition 2 is equivalent to
lim
k1
Y
j=0
|(n + j)|2
=0
1 + |(n + j)|2
(3.31)
for every n 0.
Due to the inequality
k
Y
j=0
k
k
Y
X
|(n + j)|2
1
1
,
=
1 +
1 + |(n + j)|2
1 + |(n + j)|2
|(n + j)|2
j=0
j=0
the limit (3.31) vanishes if condition

2 holds.
P
Conversely, if the series n0 |(n)|2 converges, then using the elementary inequality
log(1 + x) x (x 0), we have
k
k
2
Y
X

|(n + j)|
log
=
log 1 + |(n + j)|2

2
1 + |(n + j)|
j=0
j=0
k
X
|(n + j)|2
j=0
Therefore
lim log
and so (3.31) does not hold.
k
Y
|(n + j)|
> .
1
+
|(n + j)|2
j=0
54
P
2
It is worth noticing here that divergence of the series
is equivalent to
n0 |(n)|
non-explosion (i.e. non-escape at infinity in finite time) of a classical pure-birth process
with birth intensities (|(n)|2 )n0 (see [51]).
Here the minimal quantum dynamical semigroup T (min) preserves the identity operator 1l
if and only if this associated (in a sense that will be made precise later) classical stochastic
process enjoys the above regularity property.
In order to give another necessary and sufficient condition for the minimal quantum
dynamical semigroup to be Markov we study the linear manifold
A = {x B(h) | Q(x) = x} .
Clearly A is closed for the -weak topology. Moreover we have the following characterisation (see [19], Th. 2.4 p. 92)
Proposition 3.30 Suppose that the hypothesis AA holds and let > 0 fixed. Then, for all
x B(h), we have
L(x) = x if and only if Q (x) = x. In particular
A = { x B(h) |
L(x) = x }
Proof. Let x be an element of B(h) such that
L(x) = x. Then, for every u, v D(G) and
every t 0, we have
hL` P (t)v, xL` P (t)ui =
hP (t)v, xP (t)ui
`=1
hGP (t)v, xP (t)ui hP (t)v, xGP (t)ui .
Taking Laplace transform of both sides in this identity we obtain

Z
d
hv, Q(x)ui =
exp(t) hP (t)v, xP (t)ui dt = hv, xui
dt
0
for every u, v D(G) and hence, by density of D(G) in h, for every u, v h. Thus
Q(x) = x.
Conversely we now show that the relation
L(x) = x holds in the form sense if Q(x) = x.
Let R(/2; G) be the resolvent operator (/2 G)1. Notice that, for all ` 1, the operator L` R(/2; G) has a bounded extension because of (3.28) and the fact that GR(/2; G)
can be extended to the bounded operator (/2)R(/2; G) 1l. Hence, using the well-known
properties of resolvent operators, we can easily compute the derivative
d
L` P (t)R(/2; G)u =
dt
=
(L` R(/2; G))

L` P (t)u +
d
P (t)u
dt
L` P (t)R(/2; G)u
2
for u D(G). For every u, v D(G), letting v0 = R(/2; G)v, u0 = R(/2; G)u, since
GR(/2; G) = (/2)R(/2; G) 1l, we then have
hL` v0 , xLù0 i
`=1
Z
X
`=1
Z
X
`=1
d
(exp(t) hL` P (t)v0 , xL`P (t)u0 i) dt
dt
et (hL` P (t)v, xL` P (t)u0i + hL` P (t)v0 , xL`P (t)ui) dt

0
3.5. CONSERVATIVITY
55
Therefore, since Q (x) = x, we obtain
hL` v0 , xLù0 i
= hv, Q (x)u0 i + hv0 , Q(x)ui
`=1
= + h((/2)1l G)v0 , xu0i + hv0 , x((/2)1l G)u0 i .

This shows that
L(x) = x in the form sense because u0, v0 can be chosen arbitrarily in
D(G).
The following necessary and sufficient condition for conservativity ([22] Th. 3.3) is now
obvious.
Proposition 3.31 Suppose that the hypothesis AA holds and let > 0 fixed. The following
1. the minimal quantum dynamical semigroup T (min) is Markov,
2. there exists no non-zero x B(h) such that
L(x) = x.
Proof. Indeed conditions 1 and 2 are equivalent to A = {0} by Proposition 3.30 and
Theorem 3.28.
We finish this section with two properties of Markov minimal quantum dynamical semigroups.
The first one is a result due to E.B. Davies (see [32] Th. 3.2 p.174) giving another
necessary and sufficient condition for the minimal quantum dynamical semigroup T (min) to
be Markov.
Proposition 3.32 Suppose that the hypothesis AA holds. The linear manifold V generated
by the rank-one operators
|uihv|,
u, v D(G)
is contained in the domain of the infinitesimal generator L of the predual semigroup S (min)
of T (min) and
X
L (|uihv|) = |Guihv| +
|LùihL` v| + |uihGv|
(3.32)
`=1
(where the series converges in trace norm). Moreover the following conditions are equivalent:
1. the linear manifold V is a core for L,
2. the minimal quantum dynamical semigroup T (min) is Markov.
Proof. For every u, v D(G) and every x B(h) equation (3.15) can be written as

(min)
tr xSt
(|uihv|)
=
tr (x|uihv|) +
Z tD

E
v,
L Ts(min) (x) u ds.
Therefore we have
1
1 (min)
(|uihv|) |uihv| =
tr x St
t
t
+
D
X
`=1
Gv, Ts(min) (x)u
(3.33)
L` v, Ts(min) (x)L` u + v, Ts(min) (x)Gu
ds
56
The functions
D
E
s Gv, Ts(min) (x)u ,
D
E
s v, Ts(min) (x)Gu ,
D
E
s L` v, Ts(min) (x)L` u ,
` 1,
are clearly continuous. Moreover, by virtue of (3.28), we have the estimate

D
E

L` v, Ts(min) (x)L` u kxk kL` vk kL` uk
for every ` 1. Notice that the series
kL` vk kL` uk
`=1
kL` vk
!1/2
`=1
(2<ehv, Gvi)
kL` uk
!1/2
`=1
1/2
(2<ehu, Gui)
1/2
converge. Hence the function

s
D
X
L` v, Ts(min) (x)L` u
`=1
is continuous by Lebesgues theorem (applied to the counting measure on IN ). Letting t

tend to 0 in (3.33) we have then

1 (min)
lim tr x St
(|uihv|) |uihv| = hv,
L(x)ui.
t0+ t
Therefore, since the strong generator and the weak generator of S (min) coincide (see, for
example, [76] Th.1.3 p.43), the rank-one operator |uihv| (u, v D(G)) belongs to the domain
of L and (3.32) holds (the series is convergent in trace norm).
We show now the equivalence of conditions 1 and 2.
Suppose that T (min) is Markov. Since V is trace-norm dense in the Banach space of trace
class operators, V is a core for L if and only if the orthogonal in B(h) of the linear manifold
( L )(V) is trivial for some > 0 (see, for example, [33], Prop. 3.1 p. 17 or [63] Problem
5.19 p. 166). If x is an element of this orthogonal complement we have
tr ((( L)(|uihv|)) x) = 0
i.e.
L(x) = x in the form sense. Hence x = 0 by Proposition 3.31 because T (min) is Markov.
Conversely notice that for every V we have
tr (L ()) = 0.
Therefore, since V is a core for L this identity holds for all in the domain of L . For all
such we have then

d (min)
(min)
() = tr L St
() = 0.
tr St
dt
(min)
()) = tr() for every t 0 and every trace class operator because
This shows that tr(St
V is dense with respect to the trace norm. Thus T (min) is Markov by Proposition 3.26.
3.5. CONSERVATIVITY
57
The second one is a useful characterisation of the domain obtained in [45] Lemma 1.1.
This is the counterpart at the level of the infinitesimal generator of a well-known property
of weak*-continuous semigroups (see [21] Prop. 3.1.23 p.182). Namely, in the present case,
an operator x B(h) belongs to the domain of the infinitesimal generator L of a quantum
dynamical semigroup T if and only if
sup t1kTt(x) xk < .
t0
Proposition 3.33 Suppose that the hypothesis AA holds and that the minimal quantum
dynamical semigroup is Markov. Then the domain of the infinitesimal generator L(min) of
T (min) is given by all elements x B(h) such that the sesquilinear form
L(x) on D(G)D(G)
(v, u) hv,
L(x)ui
(3.34)
is norm-continuous.
(min)
Proof. Let L
be the infinitesimal generator of the predual semigroup S (min) . Since
(min)
L(min) = (L
) , an operator x B(h) belongs to the domain of L if and only if the linear
(min)
form on D(L
)

(min)
tr L
()x
(3.35)
is continuous for the trace norm kk1. Therefore the proof consists essentially in establishing
the equivalence of this property with the norm continuity of (3.34).
Clearly the continuity is a necessary condition for x being an element of D(L(min) )
because
D
E

|hv,
L(x)ui| = v, L(min) (x)u
kL(min) (x)k kvk kuk

kL(min) (x)k k |uihv| k1.
We prove now that it is also sufficient. Notice first that, if the sesquilinear form (3.34)
is norm continuous, then there exists an operator y B(h) such that
hv,
L(x)ui = hv, yui
for each v, u D(G). Any V can be written as a finite sum
d
X
|uj ihvj |
j=1
where (uj )1jd, (vj )1jd are elements of D(G). We have then

d

d

X
X

(min)

L
()x
=
hv
,
L
(x)u
i
=
hv
,
yu
i
tr

j
j
j
j = |tr(y)| kyk kk1 ,

j=1
j=1

whenever x B(h) is chosen so that the form (3.34) is continuous. This inequality also
(min)
holds for D(L
) because the linear manifold V generated by operators |uihv| with
(min)
u, v D(G) is a core for L
by Proposition 3.32. Therefore x belongs to the domain of
(min)
L
.
58
3.6
Sufficient conditions for conservativity
The necessary and sufficient conditions for conservativity that were proved in the previous
section turn out to be applicable only in some simple cases when, due to the structure of the
operators G and L and to a symmetry or an invariance of the form
L, the minimal quantum
dynamical semigroup T (min) leaves invariant an abelian subalgebra of B(h) containing the
unit 1l. In these cases (see, for example, [19], [27], [56]) it suffices to study the problem for
the classical subMarkov semigroup on the invariant abelian subalgebra.
In this section we give a simple general sufficient condition for the minimal quantum
dynamical semigroup to be Markov obtained by A.M. Chebotarev and the author [28].
The starting point of our analysis is a simple remark which is an immediate consequence
of Theorem 3.28. The minimal quantum dynamical semigroup T (min) is Markov whenever,
for a fixed > 0, the series

u, Qk(1l)u
(3.36)
k=1
is convergent for all u in a dense subspace of h. In fact, in this case, condition 2 of Theorem
3.28 holds. Indeed convergence of the series (3.36) implies that the sequence (hu, Qk(1l)ui)k1
vanishes as k goes to infinity.
In order to find easily verifiable conditions on the operators G, L` that guarantee convergence of (3.36), we start by proving an easy estimate.
Let R(n; G) denote the resolvent operator (n1l G)1 . We can define bounded operators
in h formally written as
(nL` R(n; G)) (nL` R(n; G))

`=1
by (3.28). In fact we have

X

hnL` R(n; G)u, nL`R(n; G)ui

2 |hnR(n; G)u, nGR(n; G)ui|
`=1
2 knR(n; G)uk knGR(n; G)uk
4n kuk
for every u D(G) by well-known properties of resolvent operators. Thus we denote by Fn

the unique bounded positive operator satisfying
hnL` R(n; G)u, nL`R(n; G)ui = hu, Fnui .
`=1
We can prove the following

Proposition 3.34 Suppose that the hypothesis AA holds. For every u h we have
D
E
X

(min)
u, Qk(1l)u lim inf u, R (Fn )u .
n
k=1
Proof. For u D(G), n 1, we have

hu, P(Fn)ui
Z
X
`=1 0
Z
X
`=1
et knL` R(n; G)P (t)uk dt

2
et kL` P (t)(nR(n; G)u)k dt
hnR(n; G)u, nQ(1l)R(n; G)ui .
3.6. SUFFICIENT CONDITIONS FOR CONSERVATIVITY
59
Therefore we have
P (Fn ) = n2 R(n; G)Q(1l)R(n; G).
It follows that the sequence (P (Fn))n1 is uniformly bounded and converges strongly to
Q(1l) by well-known properties of resolvent operators.
Since the maps Qk are normal, we can use the representation (3.26) of the resolvent
(min)
R
and Fatous Lemma to get
X
X

u, Qk(1l)u lim inf
u, Qk(P (Fn))u
n
k=1
k=0
D
E
(min)
= lim inf u, R (Fn )u
n
for every u h. This completes the proof.

In view of the above proposition it is now clear that the series (3.36) converges whenever
(min)
we can control R
(Fn ). This can be done, for example, whenever we can find a positive
self-adjoint operator C such that
D
D
E
E
(min)
(min)
Fn C ,
and
sup u, R (Fn)u sup u, R (C )u <
>0
n1
for u in a dense subset in h, where C are the bounded approximations

C = C(1l + C)1 .
D
E
(min)
In order to estimate u, R (C )u we introduce our main assumption:
Hypothesis C - Let G, L` (` 1) be operators satisfying the hypothesis AA. The
operators satisfy the hypothesis C if there exist a positive self-adjoint operator C and a
linear manifold D with the following properties
D is contained in the domain of G,
D is a core for C 1/2,
D is an invariant domain for the operators P (t) (t 0) and the linear manifold
R(; G)(D) is contained in the domain of C 1/2 for each > 0,
the linear manifolds L` (R(; G)(D)), ` 1, > 0 are contained in the domain of
C 1/2,
there exists a positive constant b such that
D

2
D
E X
E

2<e C 1/2u, C 1/2Gu +
C 1/2L` u, C 1/2L` u b C 1/2u
(3.37)
`=1
for every u R(; G)(D) for some > 0.

It is worth noticing here that, disregarding the technical assumptions, the hypothesis C
heuristically means that there exist a selfadjoint operator C such that
L(C) bC.
Remark. Note that the well-known identity GR(; G) = R(; G) 1l implies immediately that the linear manifold GR(; G)(D) is contained in the domain of C 1/2 therefore
C 1/2GR(1; G)u in (3.37) makes sense.
60
Remark. The above hypothesis C is a straightforward generalization of hypothesis C in

[28] where D was assumed to be equal to D(G). The results obtained under this modified
assumption are the same and the proofs are almost identical.
In our main application here G (resp. L` ) will be a second (resp. first) order differenI with regular coefficients and C will be the Laplace operator in
tial operator in L2 (IRd ;C)
L2(IRd ;C).
I In this case it will be convenient to take D = H 2 (IRd ;C).
I
We prove now some facts following from the hypothesis C. Let k kC denote the graph
norm of the domain of C 1/2

2

2
2
kukC = kuk + C 1/2u .
and let k kC, denote the corresponding operator norm.
Remark. Note that the hypothesis C implies that R(; G)(D) is dense both in h (with the
norm k k) and D(C 1/2) (with the norm k kC ) for each > 0. Indeed, D is a core for G
since it is invariant under the operators P (t) (see e.g. [21] Cor. 3.1.7 p. 167). Therefore
R(; G)(D) is dense in h by [33], Prop. 3.1 p. 17 or [63] Problem 5.19 p. 166). Moreover,
if v belongs to the orthogonal of R(; G)(D) in D(C 1/2), then, for each u D we have
D
E
0 = hv, R(; G)uiC = hv, R(; G)ui + C 1/2v, C 1/2R(; G)u .
Thus
D
E
1/2

C v, C 1/2R(; G)u kvk kR(; G)uk .
This shows that C 1/2v belongs to the domain of C 1/2 and

hR(; G)(1l + C)v, ui = 0.
Since D is dense in h, we find R(; G )(1l+C)v = 0 i.e. (1l+C)v = (G )R(; G)(1l+C)v =
0. Therefore v = 0 because 1l + C has a bounded inverse.
Similar (indeed simpler) arguments show that R(; G)(D) is dense also in h for the norm
k k.
Lemma 3.35 Suppose that the hypothesis C holds. Then, for each > b, the domain of
C 1/2 is invariant under R(; G) and we have
kR(; G)kC, ( b)
(3.38)
Proof. Let u be an element of D and let v = R(; G)u. Then v C 1/2 and the inequality
(3.37) yields

2
1/2 2

C u = C 1/2(1l 1 G)v

2
2
D
E

= C 1/2v 21<e C 1/2v, C 1/2Gv + 2 C 1/2Gv
2

1 1 b C 1/2v

2

= ( b) C 1/2R(; G)u .
The domain D being a core for C 1/2 for each u D(C 1/2) we have then

1/2
1/2

1/2
1
C u ( b) C 1/2u .
C R(; G)u 1/2 1 1 b
This proves the invariance of D(C 1/2) under R(; G) and the inequality (3.38).
61
Proposition 3.36 Suppose that the hypothesis C holds. Then D(C 1/2) is invariant under
P (t) for each t 0, and the restrictions of the operators P (t) to D(C 1/2) form a strongly
continuous semigroup on the Hilbert space D(C 1/2 ) with norm k kC . Moreover we have
kP (t)kC, exp(bt).
Proof. The restriction to D(C 1/2 ) of the resolvents R(; G) obviously satisfies the resolvent
identity
R(; G) R(; G) = ( )R(; G)R(; G)
for each , > b. Therefore it is a resolvent family in the Hilbert space D(C 1/2) with the
norm k kC .
Moreover ker (R(; G)) = {0} since R(; G)u = 0 implies u = (1l G)R(; G)u = 0 and
R(; G)(D(C 1/2)) is dense in D(C 1/2) as remarked before. It follows then from [76] Th.9.3
p.39 that the restriction to D(C 1/2 ) of the operators R(; G) yields the resolvent family of
a unique densely defined closed operator on D(C 1/2 ).
The inequality (3.38) implies (see [76] Cor.3.8 p.12) that the above densely defined closed
operator is the infinitesimal generator of a strongly continuous contraction semigroup on the
Hilbert space D(C 1/2).
Since this semigroup is given by the strong limit in D(C 1/2)
n n n
lim
R
;G
n t
t
(see [76] Th.8.3 p.33) it follows that D(C 1/2 ) is invariant under the operators P (t) and the
semigroup on D(C 1/2) coincides with the semigroup obtained by restricting the operators
P (t) to D(C 1/2). Moreover we have
kP (t)kC,
lim k((n/t)R(n/t; G))nkC,

n
lim 1 n1bt
= exp(bt).
n
n

Corollary 3.37 Suppose that the hypothesis C holds. Then, for each u R(; G)(D) with

2
> 0, the function t C 1/2P (t)u is differentiable and
2
D
E
d

1/2
C P (t)u = 2<e C 1/2P (t)u, C 1/2GP (t)u .
dt
Proof. For every > 0 we have
C1/2P (t +
s)u =
C1/2P (t)u
t+s
C1/2P (r)Gu dr
t
for every s, t 0 and u R(; G)(D). Notice that, since u = R(; G)v with v D,
then Gu = R(; G)v v belongs to D(C 1/2 ). Moreover, by Proposition 3.36, the function
t C 1/2P (t)Gu is continuous on [0, +[ for the norm k k on h and satisfies the inequality
kC 1/2P (t)Guk exp(bt)kC 1/2Guk. This allows us to let tend to 0 to get
Z t+s
C 1/2P (t + s)u = C 1/2P (t)u +
C 1/2P (r)Gu dr.
t
A straightforward computation shows that the difference

2
2
1/2

C P (t + s)u C 1/2P (t)u
62
for s, t 0, can be written in the form

2
D
E
1/2

C (P (t + s) P (t)) u + 2<e C 1/2P (t)u, C 1/2 (P (t + s) P (t)) u
Z t+s
2
Z t+s D
E

1/2

C 1/2P (t)u, C 1/2P (r)Gu dr +
= 2<e
C
P
(r)Gu
dr

t
Therefore the function t kC 1/2P (t)uk2 is absolutely continuous. Moreover the last term
is an infinitesimum of order bigger than s as s tends to 0 since
Z

t+s
t
2
Z

C 1/2P (r)Gu dr
t+s
t
2
1/2

C P (r)Gu dr.
The proof is now immediate.

(min)
Under the hypothesis C we can prove a useful estimate of R
(C).
Proposition 3.38 Suppose that the hypothesis C holds. Then, for all > b and all u
D(C 1/2), we have
2
D
E

(min)
( b) sup u, R (C )u C 1/2u .
(3.39)
>0
(n)
Proof. Let R
be the sequence of positive linear maps considered in the proof of

n>0
Theorem 3.25.
(n)
It suffices to show that, for all n 0, > b and u D(C 1/2 ), the operator R (C )
satisfies
2
D
E

(n)
( b) sup u, R (C)u C 1/2u .
(3.40)
>0
The above inequality holds for n = 0 and u R(; G)(D). Indeed, since D is invariant under
P (t), then P (t)R(; G)(D) = R(; G)(P (t)(D)) R(; G)(D) is contained in D(C 1/2 ) and,
integrating by parts, we have
D
E
(0)
u, R (C)u
Z
=
et hP (t)u, CP (t)ui dt
Z0

2

et C 1/2P (t)u dt
0
Z

2
D
E

= C 1/2u + 2<e
et C 1/2P (t)u, C 1/2GP (t)u dt.
0
The inequality (3.37) yields

D
E
(0)
u, R (C )u
Z

2
2

C 1/2u + b
et C 1/2P (t)u dt
0

2
D
E

(0)
= C 1/2u + b sup u, R (C)u .
>0
This proves (3.40) for n = 0 and u R(; G)(D). Since R(; G)(D) is dense in D(C 1/2) for
the norm k kC , (3.40) holds also for u D(C 1/2 ).
63
Suppose that (3.40) has been established for an integer n and every u D(C 1/2). Notice
that for every u R(; G)(D) the vectors L` P (t)u belong to D(C 1/2 ). Thus, from the
(n)
second equation (3.17) (or from (3.27)) and the definition of R , we have
D
E
(n+1)
u, R
(C)u
Z
D
E
X
(n)
= hu, P(C)ui +
et L` P (t)u, R (C )L` P (t)u dt
`=1
1 X
b
hu, P(C)ui +
`=1

2

et C 1/2L` P (t)u dt.
Using the inequality (3.37) and integrating by parts we obtain

Z

2
X

et C 1/2L` P (t)u dt
`=1

2
d

1/2
C P (t)u dt
e
dt
Z0

2

+b
et C 1/2P (t)u dt
0
Z

2
1/2 2

= C u ( b)
et C 1/2P (t)u dt
0

1/2 2
C u ( b) hu, P(C )ui .
Z
Therefore (3.40) for n+1 and u R(; G)(D)follows. Since R(; G)(D) is dense in D(C 1/2),
this completes the proof.
We can now prove the main results of this section.
Theorem 3.39 Suppose that there exists an operator C satisfying hypothesis C such that
hu, Fnui hu, Cui
for each u D(C), n 1. Then the minimal quantum dynamical semigroup is Markov.
Proof. Fix > b. Under the hypotheses of the theorem, for > 0, the bounded operators
(Fn) and C satisfy the inequality (Fn) C (see, for example, [78] Chap. 8, Ex. 51,
p.317). Applying Proposition 3.38, we obtain the estimate

u, Qk(1l)u
D
E
(min)
lim inf sup u, R ((Fn ))u
k=1
D
E
(min)
sup u, R (C )u
n >0
>0

2

( b)1 C 1/2u < +
for every u D(C 1/2). Therefore the minimal quantum dynamical semigroup is Markov
because condition 2 of Theorem 3.28 is fulfilled.
Remark. Notice that, in the above theorem, we did not assume that the quadratic form
u 2<e hu, Gui
with domain D(G) is closable. This phenomenon happens in quite simple and interesting
cases as, for example, in Example 5.3 in [28].
64
Theorem 3.40 Suppose that the hypothesis AA holds and there exists a positive self-adjoint
operator in h such that:
1. the domain of the positive square root 1/2, contains the domain of G and, for every
u D(G), we have
2<e hu, Gui =
D
E
hL` u, L` ui = 1/2u, 1/2u ,
`=1
2. there exists a positive self-adjoint operator C satisfying the hypothesis C such that the
domain of C is contained in the domain of and, for every u D(C), we have
D
E D
E
1/2u, 1/2u C 1/2u, C 1/2u .
Then the minimal quantum dynamical semigroup is Markov.
Proof. Let > b and u D fixed. For > 0, the bounded operators and C satisfy
the inequality C (see [78] Chap. 8, Ex. 51, p.317). Moreover, for u D(G), we have
Z

2

sup hu, P( )ui =
et 1/2P (t)u dt
>0
Z
X
`=1
et kL` P (t)uk dt
0
= hu, Q(1l)ui .
Therefore the increasing family of operators (P ( ))>0 is uniformly bounded and, since
D(G) is dense in h, it follows that it converges strongly to Q(1l) as goes to 0. The maps
Qk being -weakly continuous, we have
X
X

u, Qk+1
u, Qk(P ( ))u
(1
l)u
=
sup
>0
k=0
k=0
D
E
(min)
= sup u, R
( )u
>0
by Theorem 3.25. From Proposition 3.38 we obtain the estimate

2
D
E
X

(min)
u, Qk(1l)u sup u, R
(C)u ( b)1 C 1/2u
k=1
>0
for every u D(C 1/2). Therefore the minimal quantum dynamical semigroup is Markov
because condition 2 of Theorem 3.28 holds.
It has been shown in [53] that condition C can be interpreted as an a priori estimate on
the minimal quantum subMarkov semigroup associated with the operators G, L` .
The following corollary gives an easier (but weaker) sufficient condition for conservativity.
Corollary 3.41 Suppose that the hypothesis AA holds and there exists a self-adjoint operator C with domain coinciding with the domain of G and a core D for C with the following
properties:
a) L` (D) D(C 1/2) for all ` 1,
65
b) there exists a self-adjoint operator such that

2<e hu, Gui = hu, ui hu, Cui
for all u D,
c) there exists a positive constant b such that the inequality
2<e hCu, Gui +
D
X
E
C 1/2L` u, C 1/2L` u b hu, Cui
(3.41)
`=1
holds for every u D.

Then the minimal quantum dynamical semigroup is Markov.
Proof. If we consider D(G) and D(C) as Hilbert spaces equipped with their respective
graph norms, then it follows readily that the inclusion map form D(G) to D(C) is closed
and everywhere defined. An application of the Closed Graph Theorem implies that G is
relatively bounded with respect to C. Exchanging G and C the same argument shows that
also C is relatively bounded with respect to G. Thus there exist constants c1 , c2, c3, c4 such
that
kGuk c1kCuk + c2 kuk,
kCuk c3kGuk + c4kuk.
Therefore D is a core for both C and G and, for each u D(G) = D(C), there exists a
sequence (un )n1 in D such that the sequences (Gun)n1 and (Cun)n1 converge strongly
to Gu and Cu respectively.
This remark, together with the hypothesis b), allows to check immediately the hypothesis
1 of Theorem 3.40 and the inequality in the hypothesis 2 of the same theorem.
The Corollary follows if we show that the operator C satisfies the hypothesis C taking
there as domain D the domain of G. Indeed this is obviously a core for C 1/2 and it is
obviously invariant for the operators P (t) for t 0 and R(; G) for > 0.
Taking sequences (un )n1 in D such that (Gun)n1 and (Cun )n1 converge strongly to
Gu and Cu we can show that the linear manifolds L` (D(G)) are contained in the domain
of C 1/2 and that the inequality (3.41) holds for all u D(G). If, moreover, u belongs to
R(; G)(D(G)) = D(G2 ), then the inequality (3.37) follows.
This shows that C satisfies the hypothesis C.
66
Classical and Quantum

Semigroups
Several quantum Markov semigroups admit an invariant abelian subalgebra which can be
looked at as an algebra of bounded functions on some measurable space. The restriction to
this subalgebra is a semigroup of positive and identity preserving operators i.e. a classical
Markov semigroup (see Examples 3.3, 3.4).
We first study the following problem: given a classical Markov semigroup is it the restriction to an abelian subalgebra of a quantum Markov semigroup? The answer to this
question would be a step towards the understanding which classical processes can appear in
quantum stochastics. Here shall see that the answer is in the affirmative in for homogeneous
diffusions on IRd with smooth coefficients (see also the partial result of [41]).
It can be shown that the answer is in the affirmative also for:
countable state Markov chains (see [71]),
some diffusion with non-smooth drift (e.g. the transient Bessel processes [44]),
the Azema martingales for parameters smaller than a critical value (see [27]).
A complete characterisation of classical semigroups which can arise as restrictions of
quantum Markov semigroups however is not known. We point out the analytical difficulties
related to the quantum interpretation of classical boundary conditions.
This problem, however, is also interesting from a probabilistic point of view because it is
a natural generalisation of the theory of classical stochastic processes. In fact B.V.R. Bhat
and K.R. Parthasarathy showed in [20] that every quantum Markov semigroup admits a
canonical dilation to a quantum Markov process (in the appropriated sense).
In the second part of this chapter we apply our results to construct a class of quantum
Markov semigroup arising in Quantum Optics in the weak coupling or singular coupling limit
of a multi-mode Boson field interacting with a reservoir in the so-called squeezed vacuum.
These semigroups appear in several contexts in physical applications (see e.g. [10], [15],
[47]).
4.1
Preliminary definitions and results
Let E be a closed subset of IRd and let E be the Borel -field of E. Let L (E;C)
I (resp.
Cl0(E;C))
I be the Banach space of complex-valued measurable (resp. continuous) functions
on E (resp. having a limit as |x| if E is unbounded) endowed with the norm
kfk = sup |f(x)|.
xE
67
68
4. CLASSICAL AND QUANTUM SEMIGROUPS
Moreover, for all integer k 1, let Clk (E;C)

I be the vector space of complex-valued functions
on E with all the partial derivatives of the first k orders in Cl0(E;C).
I Let
P : [0, +) E E [0, 1]
be a transition probability. Consider the identity preserving semigroup T = (Tt )t0 on
L (E;C)
I defined by
Z
Tt f(x) =
f(y)P (t, x; dy).
E
Suppose that the semigroup T enjoys the Feller property i.e.

1. for all f Cl0(E;C)
I and all t 0 the function Tt f belongs to Cl0(E;C),
I
2. the restriction of T to Cl0(E;C)
I is a strongly continuous contraction semigroup.
Clearly, in this case, the semigroup T is uniquely determined by its restriction to Cl0 (E;C).
I
Therefore we shall work with this restriction. Denote by A the infinitesimal generator, the
operator on Cl0(E;C)
I defined by

D(A) = f Cl0(E;C)
I
|
lim (Tt f f)/t exists in norm ,
t0+
Af
lim (Tt f f)/t.
t0+
Let h = L2 (E;C)
I and let B(h) be the algebra of all bounded operators on h. Clearly we
have the embeddings
Cl0(E;C)
I
, L (E;C)
I , B(h)
f
f
M (f)
where M (f) denotes the multiplication operator by f.
In order to extend the semigroup T to a quantum Markov semigroup on B(h) as a first
step we find a suitable representation for the infinitesimal generator A allowing to find the
operators G and L` related to forms
L(M (f)) for a multiplication operator M (f) by a
function f.
Definition 4.1 We say that the infinitesimal generator A can be represented in Lindblad
form if there exists operators G, (L` )`=1 on h satisfying the hypothesis AA such that
hv, (Af)ui = hv, fGui +
hL` v, fL` ui + hGv, fui
`=1
for all v, u D(G) and all f D(A).

The following proposition gives necessary conditions for an infinitesimal generator A to
be representable in Lindblad form.
Proposition 4.2 Suppose that A can be represented in Lindblad form. Let u, v D(G) and
let f D(A).
1. We have the inequality
| hv, (Af)ui | (kuk + kvk) (kGuk + kGvk) kfk .
(4.1)
4.1. PRELIMINARY DEFINITIONS AND RESULTS
69
2. The non-negative function

g = |v(Gu)| + |u(Gv)| +
|(L` v)(L` u)|
(4.2)
`=1
belongs to L1 (E;C)
I and we have the inequality
Z
| hv, (Af)ui |
g(x)|f(x)|dx.
(4.3)
Proof. By Definition 4.1 and the Schwarz inequality we can estimate the modulus of the
scalar product hv, (Af)ui by
X
kGvk kuk +
kLj vk kLj uk + kvk kGuk kfk
j=1
1 X
kGvk kuk +
kLj vk + kLj uk
j=1
+ kvk kGuk kfk
Using the hypothesis AA we obtain the inequalities
kLj uk
j=1
kLj vk
2<e hu, Gui 2 kuk kGuk ,
2<e hv, Gvi 2 kvk kGvk .
j=1
Then a simple computation yields (4.1).

By the Schwarz inequality and hypothesis AA, for all u, v D(G), we have
Z
X
0
|(Lj v)(x)(Lj u)(x)|dx =
j=1
Z
X
|(Lj v)(x)| |(Lj u)(x)|dx
j=1 0
kLj vk kLj uk
j=1
<e hv, Gvi <e hu, Gui
kuk kGuk + kvk kGvk .
This shows that the function g given by (4.2) belongs to L1 (E;C).

I Therefore it is easy to
obtain the inequality (4.3).
When A can be represented in Lindblad form we can construct the minimal quantum
dynamical semigroup on B(h) associated with A.
The following result can be applied to check whether it is an extension of the corresponding classical Markov semigroup generated by A.
Theorem 4.3 Suppose that A can be represented in Lindblad form through operators G,
L` and let T be the minimal quantum dynamical semigroup solving (3.15) with the given
operators G, L` . Then T is an extension of the classical semigroup T generated by A if and
only if it is Markov.
70
Proof. If T is an extension of T then, denoting by 1 the constant function equal to 1 on

E, and identifying bounded functions with the corresponding multiplication operators, we
have
Tt(1l) = Tt (1) = 1 = 1l.
Therefore T is Markov.
Conversely, notice that, for every f D(A), and v, u D(G) we have
Z t
hv, (A(Ts f))ui ds
hv, (Tt f)ui = hv, fui +
0
Z t
= hv, fui +
hv,
L(Ts f)ui ds
0
Hence, if T is identity preserving, then by Corollary 3.23 we have

Tt (M (f)) = M (Tt (f))
for every t 0. The above identity yields also for f Cl0(E;C)
I because the domain D(A)
is dense for the uniform norm.
Remark. Proposition 4.2 and an analogue of Theorem 4.3 also hold when the semigroup T
is w -continuous, i.e. continuous with respect to the topology (L , L1). Here we deal with
classical Markov semigroups enjoying the Feller property because they are widely studied in
the literature.
The above discussion suggests to study quantum extensions of classical Markov semigroups in two steps:
1. represent the infinitesimal generator of the classical semigroup in Lindblad form,
2. show that the minimal quantum dynamical semigroup arising is identity preserving.
The analytical tools for the second step were already discussed.
The following heuristic remarks suggest a recipe for the first one.
Suppose that
X
L(x) = G x +
L` xL` + xG
`=1
where the operators G, L` satisfy the hypothesis AA. Then

L(1l) = 0 i.e.
G + G =
L` L` .
`=1
Letting
iH =
1X
L` L` G
2
`=1
the operator H turns out to be symmetric since (formally)

1X
L` L` G
2
iH
`=1
`=1
`=1
X
1X
=
L` L` +
L` L` + G
2
= iH.
4.2. DIFFUSION PROCESSES ON IRD
71
Therefore we can write G in the form

1X
L` L` iH.
2
G=
(4.4)
`=1
Then the form

L can be written as
L(x) = i[H, x]
1X
(L` L` x 2L` xL` + xL` L` ) .
2
`=1
A formal computation yields

(xy) xL
(y)
hv, (L
L(x)y) ui =
h[L` , x]v, [L`, y]ui .
`=1
The following is our recipe: given the classical infinitesimal generator then we try to find
operators L` such that
hv, (A(fg) fA(g) A(g)f) ui =
[L` , M (f)]v,
[L` , M (g)]u .
(4.5)
`=1
Then we try to find a symmetric operator H such that

1X
(L` L` M (f) 2L` M (f)L` + M (f)L` L` )
2
i[H, M (f)] =
L(M (f)) +
(4.6)
`=1
and consider as operator G an operator defined formally by the right-hand side of (4.4).
Domain problems will be considered later.
4.2
Diffusion processes on IRd
In this section we will show how to construct a quantum Markov semigroup extending the
semigroup of homogeneous diffusions (y(t))t0 on IRd satisfying the stochastic differential
equation
d
X
dyj (t) =
jk (y(t))dwk (t) + bj (y(t))dt,
1jd
(4.7)
k=1
(where (wk )1kn is a d dimensional Wiener process) under the following regularity condition on the matrix (a square root of the covariance matrix) and on the drift vector
b:
Hypothesis D
1. The functions jk : IRd IR (1 j, k d) are bounded and four times differentiable
with bounded partial derivatives of the first four orders,
2. the functions bj : IRd IR (1 j d) are bounded and three times differentiable
with bounded partial derivatives of the first three orders.
It is well known that, if and b are Lipschitz, then, for every fixed initial condition
y(0) = y0 , there exists a unique Markov process y on IRd satisfying (4.7). We refer to the
72
books [83] by D.W. Stroock and S.R.S. Varadhan, [33] by S.N. Ethier and T.G. Kurtz, [59]
by K. It
o, H.P. McKean for detailed results on these processes. Here we use the above
stronger assumption simplifying the analytical problems with the unbounded infinitesimal
generator.
Let a = and let j denote the partial derivative with respect to the j-th coordinate.
The infinitesimal generator is the operator
Af =
d
d
X
1 X
ajk j k f +
bj j f
2
j=1
(4.8)
j,k=1
characterized by Corollary
A.2.

Fix h = L2 IRd ;C
I . We compute now formally the operators G, L` .
For every f, g Cl2(IRd ;C)
I we have
A(fg) fA(g) A(f)g
d
X
ajk(j f)(k g)
j,k=1
d
X
`=1
d
X
j=1
`j (j f)
(4.9)
d
X
`k (k g)
k=1
Formula (4.5) suggests to find operators L` such that
[M (f), L` ] [L`, M (g)]
`=1
coincides with the multiplication operator by the right-hand side of (4.9). This is the case
if we consider the operators
d
X
L` =
`k k + `
(4.10)
k=1
where (` )1`d is a d-dimensional vector with entries in C 2(IR; IR) and L` = 0 for ` d.
Indeed the commutators [L` , M (g)], [M (f), L` ] coincide with the multiplication operators
by
d
d
X
X
`k (k g),
`j (j f).
k=1
j=1
With this choice of the operators L` a straightforward computation shows that (with
the convention of summation over repeated indices and identification of functions with the
corresponding multiplication operator)
L` L` M (f) 2L` M (f)L` + M (f)L` L`
= L` [L` , M (f)] [L` , M (f)]L`
= (k `k + ` )`j M (j f) + M (j f)`j (`k k + ` )
= `k `j M (k j f) + (2(` `j ) (k akj ))M (j f)
+(`j `k `k `j )M (j f)k
= `k `j M (k j f) + (2(` `j ) (k akj ))M (j f)
where we have used
(`j `k `k `j ) = ajk akj = 0.
73
Therefore the right-hand side of (4.6) coincides with the multiplication by

1
bj (k akj ) + ` `j M (j f).
2
Consider a d-dimensional vector (j )1jd with entries in Cb1 (IR; IR). Letting
iX
H=
(j j + j j ) ,
2
d
(4.11)
j=1
L(M (f)) coincides with the multiplication operator by Af if and only if

j = bj
k=1
`=1
X
1X
(k akj ) +
` `j .
2
(4.12)
Consider the closure of the operator G with domain Cc(IRd ;C)

I given by (4.4) where L` ,
H are defined on Cc (IRd ;C)
I by (4.10) and (4.11) respectively. Straightforward computations
yield
G=
1
1
1
akj k j + ((k akj ) 2j ) j + ((k `k ` ) (j j ) ` ` )
2
2
2
(4.13)
and
1
1
1
(4.14)
akj k j + ((k akj ) + 2j ) j + ((k `k ` ) + (j j ) ` ` )
2
2
2
We shall construct the minimal quantum dynamical semigroup associated with G and
L` . As first step let us check hypothesis AA.
G =
Proposition 4.4 Suppose that hypothesis D holds and ` , ` Cb3(IRd ; IR) for ` = 1, . . ., d.
Then the operators G and L` (` = 1, . . ., d) obtained as closure of the operators with domain
Cc(IRd ;C)
I defined by (4.13), (4.10) satisfy hypothesis AA.
Proof. By virtue of Theorem A.3 in the Appendix, the operator G is the infinitesimal
generator of a strongly continuous contraction semigroup in h and Cc(IRd ;C)
I is a core for
G. Moreover, for every u, v Cc (IRd ;C),
I we have
hv, Gui +
hL` v, L` ui + hGv, ui = 0.
`=1
For every u D(G), let (un )n1 be a sequence of elements of Cc (IRd ;C)
I such that
lim un = u,
lim Gun = Gu.
The above identity yields

d
X
kL` (un um )k = 2<e h(un um ), G(un um )i .
`=1
Therefore the sequence (L` un)n1 also converges to L` u for ` = 1, . . . , d and

hu, Gui +
hL` u, Lùi + hGu, ui = 0.
`=1
The identity (3.28) follows then by complex polarisation.

The operators G, L` , allow to represent the infinitesimal generator (4.8) of the homogeneous classical diffusion (4.7) in Lindablad form.
74
Proposition 4.5 Suppose that hypothesis D holds and let G, L` (` = 1, . . . , d) be as in

Proposition 4.4. For every f D(A) and every v, u D(G) we have
hv, (Af)ui = hv, fGui +
hL` v, fL` ui + hGv, fui .
`=1
Proof. Indeed the above identity holds for v, u Cc (IRd ;C)

I and f Cc (IRd ;C)
I or
constant. The conclusion follows easily since these domains are a core for both G and A
and the fact (shown in the proof of Proposition 4.4) that convergence of a series (Gun)n1
with u D(G) to Gu implies the convergence of (L` un )n1 to Lu.
The following is the key step in the construction of the semigroup of a quantum diffusion.
Theorem 4.6 Suppose that hypothesis D holds and let ` , ` Cb3 (IRd ; IR) for ` = 1, . . ., d.
Then the minimal quantum dynamical constructed from G (4.13) and L` (4.10) is conservative.
Proof. We shall apply Theorem 3.40. Let C and be the operators given by
d
X
D(C) = H 2(IRd ;C),
I
Cu = c(, )
j j u + u ,
j=1
D() = H 2(IRd ;C),

I
u =
1
2
d
X
L` L` u.
`=1
where c(, ) is the constant

c(, ) = 2
d
X

max k`1k2 , . . . , k`dk2 , k`k2 .
`=1
Clearly C and are self-adjoint by von Neumanns theorem (see [63] Th.3.24 p.275). The
linear manifold Cc (IRd ;C)
I is a core for both (for example by Theorem A.3). Moreover it is
easy to check that, for every u Cc(IRd ;C),
I we have
2<e hu, Gui =
D
D
E
hL` u, L` ui = 1/2u, 1/2u ,
`=1
1/2
E D
E
u, 1/2u C 1/2u, C 1/2u .
Since Cc(IRd ;C)

I is also core for G, the above identities hold for u D(G) and u D(C).
Thus the hypothesis 1 of Theorem 3.40 holds.
In order to apply this result it suffices then to check that the operator C satisfies the
hypothesis C.
Notice that, for every u Cc (IRd ;C),
I integrating by parts we can write the right-hand
side of (3.37) in the form
2<e hCu, Gui +
hL` u, CLùi
`=1
= i (hHu, Cui hCu, Hui)

d
1X
(hu, L` L` Cui 2 hu, L` CL` ui + hu, CL` L` ui)
2
`=1
75
= i (hHu, Cui hCu, Hui)

1X
(hL` u, [C, L`]ui + h[C, L`]u, Lùi)
2
d
`=1
Cc (IRd ;C),
I
Now, for every u

still integrating by parts we have (with the convention
of summation over repeated indices)
|hHu, Cui hCu, Hui|
1
= |h(j j + j j )u, k k ui hk k u, (j j + j j )ui|
2
1
= h(j j + j j )k u, k ui h((k j )j + j (k j ))u, k ui
2

+hk u, (j j + j j )k ui + hk u, ((k j )j + j (k j ))ui
Since (j j + j j ) is antisymmetric on Cc (IRd ;C)
I the sum of the first and third term in
the right-hand side vanishes and we remain with
2 |hHu, Cui hCu, Hui|

= h((k j )j + j (k j ))u, k ui + hk u, ((k j )j + j (k j ))ui

= 2hj u, (k j )k ui + 2hk u, (k j )j ui + 2<ehu, (j k j )k ui
Elementary inequalities lead us to the estimates
2 |hHu, Cui hCu, Hui| 2 sup kj k j k
k,j
kuk kk uk
k=1
d
X
+ 4 sup kk j k
k,j
kj uk kk uk
j,k=1
sup kj k j k
k,j
d
X
kuk2 + kk uk2
k=1
d
X
+ 4 sup kk j k
k,j
k,j
!2
kk uk
d
X
kuk2 + kk uk2
k=1
+ 4d sup kk j k
k,j
Since
hu, Cui = kuk +
d
X
kk uk2.
k=1
d
X
kk uk2
k=1
for every u Cc (IRd ;C)

I we have then
|hHu, Cui hCu, Hui| b1 hu, Cui
where
b1 = d sup kj k j k + 2 sup kk j k
k,j
k=1
sup kj k j k
d
X
k,j
76
Similar estimates can be done to show that, for every second order differential operator
X with coefficients in Cb2(IRd ;C)
I and every u Cc (IRd ;C)
I we have
|hu, Xui| b2 hu, Cui
(4.15)
where b2 is a constant depending only on the coefficients of X.

Let us estimate now
d
X
(hL` u, [C, L`]ui + h[C, L`]u, Lùi)
(4.16)
`=1
The coefficients of the operators L` (4.10) and L` belong to Cb3(IRd ; IR). Therefore, computing explicitely the action of commutators [C, L`] on u Cc(IRd ;C),
I we can write the
above sum as
X
(h`m m u, k (k `j )j ui + hk (k `j )j u, `mm ui)

`,j,m,k
2<e
hj (j ` )u, `m m ui 2<e
`,j,m
h` u, k (k `j )j ui .
`,j,k
The regularity assumption on , allows us to establish an estimate like (4.15) of the last
two sums.
Integrating by parts we can write the first sum as
X
X
`m m u, (k2`m )m u,
h`m m u, (k `j )k j ui
`,j,m,k
`,j,m,k
h(k `j )j u, `m k m ui +
`,j,m,k
h(k `j )j u, (k `m )m ui .
`,j,m,k
The second and fourth term satisfy an estimate like (4.15) since they involve only second
order partial derivatives. Exchanging j and m in the first sum we can write the sum of the
first and third term as
X
(hj u, (k `j )`m k m ui hj u, `j (k `m )k m ui) = 0.
`,j,m,k
This shows that (4.16) can be estimated by b3 hu, Cui where b3 is a constant depending only
on and .
Summing up we have shown that there exists a positive constant b depending only on ,
and such that, for every u Cc (IRd ;C),
I we have
2<e hCu, Gui +
hL` u, CL` ui b hu, Cui .
(4.17)
`=1
2
The above inequality obviously holds for every u C
(IRd ;C)
I with > d/2 by a standard
approximation argument with mollifyiers.
2
The domain C
(IRd ;C)
I ( > d/2) is invariant under the semigroup P (t) generated
by G and under the resolvents R(; G) by Theorem A.1. Therefore it satisfies all the
2
assumptions on the domain D in the hypothesis C. Thus we take D = C
(IRd ;C)
I ( > d/2)
The inequality (3.37) for u R(; G)(D) now follows from (4.17) because, if u belongs to
R(; G)(D), then by the identity GR(; G) = R(; G) 1l, Gu belongs also to the domain
of C 1/2.
By Theorem 4.3 we proved then the following
4.3. A QUANTUM MASTER EQUATION
77
Proposition 4.7 Suppose that hypothesis D holds and let ` , ` Cb3(IR; IR) for ` =
1, . . ., d. Then the minimal quantum Markov semigroup constructed from the above operators G and L` is an extension to B(h) of the semigroup of a classical d dimensional
diffusion.
When dealing with diffusions on domains with boundaries difficult analytical problems
arise to carry on the same program. We refer to the papers [18], [28] and [42] for discussions
of some important examples.
4.3
A quantum master equation
In this section we construct a quantum Markov semigroup arising as the solution to a class
of master equations in Quantum Optics (see e.g. [15], [47]).
We start by introducing the Heisenberg representation of the d-dimensional canonical
commutation relations.
Let h be the tensor product Hilbert space of d copies of l2 (IN ) and denote by (e )IN d
the canonical othonormal basis. Here is a multindex and the vector e is the tensor
product of unit vectors e(1), . . . , e(d) taken from the canonical orthonormal basis of each
copy of l2 (IN ) (see the example (1.2)).
We denote by ak , ak , Nk the standard ampliation to h of the annihilation, creation and
number operator acting on the k-th copy of l2 (IN ).
The master equation we shall study is the following
d
(t) = L ((t))
dt
where L is the predual operator on density matrices of the Lindblad operator on observables
1 X
k (ak ak x 2ak xak + xak ak ) + k (ak ak x 2ak xak + xak ak )
2
k=1

2
k (a2 x 2ak xak + xa2 )
+k (a2
x
2a
xa
+
xa
)
+
k
k
k
k
k
k
d
L(x) =
+i
d
X

2
k ak ak + k a2
k ak , x .
k + k ak + k ak +
(4.18)
k=1
Here k is a real number, k , k , k are complex numbers, k 0, k 0 and

|k |2 k k
(4.19)
for all k {1, . . ., d}. These inequalities are necessary and sufficient conditions for L to be
formally (i.e. algebraically) conditional completely positive as it will be clear later.
Solving the above quantum master equation means showing that the operator L, interpreted in the form sense as in Chapter 3, can be viewed as the infinitesimal generator of a
quantum Markov semigrup.
As a first step we identify the operators G, L` allowing to represent in Lindblad form
the above operator L (denoted
L as in Sect.1 to make clear that it should be understood in
the form sense). We proceed as in Sect.1 noting that, for suitable x, y B(h),
(y)
L(xy) xL
L(x)y =
X
`
[L` , x][y, L`]
78
must be positive. If
L is given by (4.18) and x = y , then a straightforward computation
shows that
d
X
k [y, ak ] [y, ak ] + k [y, ak ] [y, ak ] + k [y, ak ] [y, ak ] + k [y, ak ] [y, ak ]
k=1
is positive. Taking operators y obtained by standard ampliation of operators acting on the

k-th copy of l2 (IN ) we can see that this is the case if and only if each term of the above sum
must be positive.
Thus we concentrate on k = 1 and drop the index to reduce the clutter of the notation.
Letting L = za + wa with z, w complex numbers to be determined we have
[y, L][y, L] = |z|2[y, a][y, a] + zw[y, a] [y, a ] + z w[y,
a ][y, a] + |w|2[y, a ][y, a ].
Hence, choosing two numbers 0 , 0 such that 0 0 , 0 0 and 0 0 = ||2, we
can find a pair z, w of complex numbers (unique up to a phase factor) such that
|w|2 = 0.
|z|2 = 0 , zw = , z w
= ,
Therefore can write the decomposition
a ][y, a] + [y, a ] [y, a ]
[y, a][y, a] + [y, a] [y, a] + [y,
=
[y, L][y, L] + ( 0)[y, a] [y, a] + ( 0 )[y, a ][y, a ].
To reduce the number of operators L` in the representation in Lindblad form (to the dimension of the minimal representation space associated with the completely positive part of L
we shall choose 0 = if < and 0 = , if .
Summing up, in order to represent L in the Lindblad form, we define (domains will be
made precise later), for k = 1, . . . , d,
L2k1
zk ak + wk ak ,
L2k
L2k
=
=
(k 0k )1/2ak ,
(k 0k )1/2 ak ,
if
if
k > k ,
k k .
With this choice of L` , a natural choice for the operator G is the following
2d
G=
1X
L` L` iH
2
`=1
where
H=
d
X
2
k ak ak + k a2
k ak .
k + k ak + k ak +
k=1
The above operators are obviously defined on the linear manifold D spanned by the
elements e of the canonical orthonormal basis of h. Now we shall construct rigorously the
quantum Markov semigroup associated with an appropriated extension of these operators.
As a first step we shall prove that the closure of the above operator G with domain D
generates a strongly continuous contraction semigroup on h applying the following result
due to Palle E.T. Jorgensen (see [60] Th.2 p.398)
Theorem 4.8 Let G be a dissipative linear operator on a Hilbert space h. Let (Dn )n1 be
an increasing family of closed subspaces of h whose union is dense in h and contained in
the domain of G and let PDn be the orthogonal projection of h onto Dn . Suppose that there
79
generates a
exists an integer n0 such that GDn Dn+n0 for all n 1. Then the closure G
if there exists
strongly continuous contraction semigroup on h and n1Dn is a core for G,
a sequence (cn )n1 in IR+ such that kGPDn PDn GPDn k cn for all n and
c1
n = .
n=1
With the above notation let Dn be the linear manifold spanned by vectors e with
|| n. Clearly D = n1Dn . The operator G is obviously densely defined and dissipative.
is
Therefore it is closable (see e.g. [21] Lemma 3.1.14 p.175) and its closure, denoted G
dissipative.
is the infinitesimal generator of a strongly continuous
Proposition 4.9 The operator G
contraction semigroup on h.
Proof. Clearly, by the explicit form of the action of creation and annihilation operators on
vectors e , the operator G maps Dn into Dn+2 for all n 0. In order to apply Theorem
4.8 we shall show
P that the norm kGPDn PDn GPDn k grows at most linearly. To this end
let us fix u = k|n z e a vector in Dn . A straightforward computation yields
(GPDn PDn GPDn ) u
= i
d
X
k=1
d
X
k
k=1
d
X
k
k=1
((k) + 1)
1/2
z e+1k
||=n
+ k
(((k) + 1)((k) + 2))
1/2
z e+2k
||=n1
+ k
X
(((k) + 1)((k) + 2))
1/2
z e+2k .
||=n
The squared norm of the third term can written

d
X
X k0
k
+ k0
+ k
2
2
0
0
k,k =1 ||,| |=n
((0(k0 ) + 1)(0(k0 ) + 2)((k) + 1)((k) + 2))

0
1/2
z0 z he0 +2k0 , e+2k i
Note that, for each k, k {1, . . ., d} and IN with || = n, there exists exactly a
multindex 0 = (, k, k0) such that the scalar product he0 +2k0 , e+2k i does not vanish.
Therefore the squared norm of the third term can be estimated by

d
X
X k0
k

2

(n + 2)
2 + k0 2 + k z(,k,k0) |z |
0
k,k =1 ||=n

d

X

(n + 2)2 X k0
k0 k + k
z(,k,k0) 2 + |z |2
+

2
2
2
0
k,k =1
(n + 2)2 kuk2
||=n
!2
k

+ k .
2

d
X
k=1
In a similar way we can majorize the squared norms of the first and second term respectively
by
!2
!2
d
d
X
X

k
+ k .
(n + 1)2 kuk2
|k | ,
(n + 2)2 kuk2
2

k=1
k=1
80
By the elementary |r + s + t|2 3(|r|2 + |s|2 + |t|2) we have then
kGPDn PDn GPDn uk
!2
d
X
k

+ k +
3 2
2

k=1
Since the series
n (n
d
X
k=1
!2 1/2
|k | (n + 2)kuk.
+ 2)1 is divergent the proof is complete.
A similar argument (or an application of an extension of Nelsons theorem on analytic

vectors, [79] Th.X.40 p.206) allows us to prove the following
Proposition 4.10 The closure of the operator on h with domain D defined by
2d
X
L` L` =
`=1
d
X
2
k ak ak + k a2
k + k ak + k ak ak
(4.20)
k=1
is essentially self-adjoint.
Let us denote by G both the operator G and its closure. The operators L` can be
extended to the domain of G and further extended to the domain
(
)

X
X

1/2
2
u hu =
z e ,
|| |z | <
by standard arguments. Moreover, it follows readily that the hypothesis AA holds.

In order to solve the quantum master equation (4.18) it suffices now to prove that the
minimal quantum dynamical semigroup associated with the operators G, L` ( ` = 1, . . ., 2d)
is Markov. To this end we shall choose a suitable operator C and apply the results of Section
3. The choice of the operator C, however, depends upon some non-degeneracy conditions
on the parameters k , k , k , k , k , k . We shall study only the following case and make
some comments on the solution of our problem in the other cases at the end.
Hypothesis ND - For each k { 1, . . ., d } at least one among k , k , |k |, |k | is nonzero.
Proposition 4.11 Suppose that the hypothesis ND holds. Let C be the operator defined by
(
)

X
X
X

2
2
D(C) = u h u =
ze ,
(|| + 1)|z | < , Cu = c
||ze .
where c > 0 is a constant. Then C is self-adjoint and D is a core for C. Moreover:

(1) the domain of G coincides with the domain of C,
(2) if c > max1kd(k + k + 2|k |), for all u D we have
hu, ui hu, Cui.
Proof. For every IN d and every n 1 we have C ne = ||ne . Therefore C is selfadjoint and D is a core for C by Nelsons theorem on analytic vectors (see e.g. [79] Th.X.39
p.202). (This can be shown alternatively applying [60] Corollary
P 1 p.397 with Dn as before).
A direct computation allows to show
that
for
each
u
=
z e in the domain of C
P
the sequence (un )n1 defined by un = ||n z e for n 1 converges strongly to u and
(Gun)n1 converges strongly to in h. Since G is closed, it follows that u belongs to the
domain of G. Thus the domain of G is contained in the domain of C.
81
On the other hand the restriction of G to D is relatively bounded with respect to C.

Indeed, for all u D and all k { 1, . . ., d } we have
2
2 2
2
ka2
k uk = hu, ak ak ui = hak ak u, (ak ak + 1l)ui k(ak ak + 1l)uk .
So that
ka2
k uk kCuk + kuk.
Moreover, similar estimates can be easily obtained for the other terms in the sum defining
G. Thus G is relatively bounded with respect to C.
It follows that the domain of G contains that of C and (1) follows.
Finally, in order to prove (2) it suffices to note that, for all u D and all k { 1, . . ., d },
we have
2
hu, (k a2
k + k ak )ui =
2<ek hu, a2
k ui
2|k | kak uk kak uk
2|k | (hu, ak ak ui hu, ak ak ui)
2|k |hu, akak ui.
1/2
Then (2) follows easily.

Proposition 4.12 Suppose that the hypothesis ND holds. Then the minimal quantum dynamical semigroup associated with the operators G, L` ( ` = 1, . . ., 2d) is Markov.
Proof. We apply Corollary (3.41). The domains of G and C coincide by the above Proposition. We check now the hypotheses a), b), c) of the Corollary.
The hypothesis a) clearly holds because D is invariant under the L` s.
The hypothesis b) also holds by Proposition 4.11 if we choose c big enough.
Finally we check condition c). Since D is an invariant domain for the operators G, L` ,
C it suffices to compute algebraically
L(C). Indeed the left-hand side of (3.41) can be
written in the form hu,
L(C)ui. Although there are at times a product of eight operators
in the working below it should be remembered that these expressions should strictly be
interpreted as forms, and so everything is well-defined by the invariance of D under the
action of creation and annihilation operators.
Note that, for every k, m { 1, . . ., d }, by the canonical commutation relations [ak , am ] =
k,m 1l (on the domain D), we have
ak ak (am am ) 2ak (am am )ak + (am am )ak ak

ak ak (am am ) 2ak (am am )ak + (am am )ak ak
2
a2
k (am am ) 2ak (am am )ak + (am am )ak
=
=
=
2k,m ,
2k,m,
0,
a2k (am am ) 2ak (am am )ak + (am am )a2k

2
+ k a2
k ak , amam
k + k ak + k ak +
=
=
0,

k,m 2k a2k 2k a2
k ak k ak
k +
k ak ak
Therefore, for u D, we obtain

hu,
L(C)ui =
d
X
hu, (2k a2k 2k a2

k ak k ak )ui + 2
k +
k=1
d
X
(k k )kuk2.
k=1
The same arguments of the proof of Proposition 4.11 lead to the inequality
hu, (k a2k k a2
k )ui 2|k |hu, ak ak ui,
82
and

hu, (
kak k ak )ui = 2|k| kuk kak uk |k | kuk2 + kak uk2 = |k |hu, akak ui.
Summing up we find
hu,
L(C)ui c1 max { 4|k| + |k| } hu, Cui + 2
1kd
max
d
X
(k k )kuk2
k=1
max { 4|k | + |k | } , 2
1kd
d
X
(k k )
hu, Cui.
k=1
This shows that condition c) of Corollary (3.41) holds. Then the proof is complete.
Remark. If the hypothesis ND does not hold let

n
o

K1 =
k { 1, . . ., d } k + k + |k | + |k | > 0 ,

n
o

K2 =
k { 1, . . ., d } K1 |k| > 0 .
Then we can show that the minimal quantum dynamical semigroup associated with the
operators G, L` ( ` = 1, . . ., 2d) is Markov by the same arguments with the choice of the
operator C given, up to a multiplicative constant c, by
X
X
ak ak +
(ak ak )1/2.
kK1
kK2
Quantum Flows
We develop a rather general framework for constructing quantum Markov processes through
Markov operator cocycles (see Chapter 2, Section 3) that satisfy a quantum stochastic
differential equation.
In order to achieve this goal we first recall the basic facts of Boson Fock quantum
stochastic calculus and then give the fundamental results in the theory of quantum stochastic differential equations concerning existence, uniqueness, time reversal, isometricity and
coisometricity of solutions. Next we construct the quantum flow associated with a Markov
operator cocycle, and give a condition that guarantees that the restriction to a commutative
subalgebra is a commutative flow.
As an application we show that every classical multidimensional diffusion process (with
covariance and drift as in Chapter 4, Section 2) can be realised as a restriction to a commutative subalgebra of B(h) of a quantum flow.
Our processes are, in particular, Evans-Hudson flows [36].
5.1
Quantum stochastic calculus
In this section we give a short outline of the quantum stochastic calculus of Hudson and
Parthasarathy [58] in a Boson Fock space which is the main tool for the construction of
Markov operator cocycles. We refer to the monographs [69] and [74] for a complete exposition
with proofs.
A quantum stochastic calculus can be developed in several frameworks: Fermion Fock
space, Full Fock space, finite difference Fock space ...; we refer to [4] for a unified approach.
There exists several approaches also to quantum stochastic calculus in Boson-Fock space
(see [13], [14], [58], [66], [67], [84] ...): we shall use here a notation introduced by Belavkin
[16] as described by Parthasarathy (see, for example, [70], Sect. 2).
Let h, k be two complex separable Hilbert spaces and let { e` | ` S } be an orthonormal
basis in k (S is to be thought of as a subset of IN ). Let MS denote the dense linear
submanifold of L2 (IR+ ) k of elements f satisfying he` , f(t)ik = 0 for all t 0 for all but
a finite number of indices ` and let M0S be the submanifold of continuous functions in MS .
Let S = S {, +} and consider the Hilbert spaces
= Ce
k
I k Ce
I + ,

b = hk
L2 (IR+ ) k
H = h L (IR+ ) k ,
H
2
where (K) denotes the Boson Fock space over a Hilbert space K.
83
(5.1)
84
5. QUANTUM FLOWS
b will be denoted by kk . The exponential vector

The uniform operator norm in B(H)
in the Boson Fock space (K) with test function f K will be denoted by e(f) where
e(f) =
X f n
.
n!
n0
For all t > 0, we have the tensor product decomposition

H = h (L2 (0, t) k) (L2 (t, ) k).
Let e(0t] ) and e(0[t ) denote respectively the Fock vacuum in the Boson Fock spaces (L2 (0, t)
k) and (L2 (t, ) k). Both h (L2 (0, t) k) and h (L2 (t, ) k) can be identified
with two subspaces of H
h (L2 (0, t) k) e(0[t ),
h e(0t] ) (L2 (t, ) k),
(5.2)
via the unitary isomorphisms

u e(f1(0,t) ) u e(f1(0,t) ) e(0[t ),
u e(f1(t,) ) u e(f1(t,) ) e(0t] ).
Let us consider the *-algebra B = B(H) of all bounded operators on H. Let 0 be
the state on the *-algebra of all bounded operators on (L2 (IR+ ) k) associated with the
vacuum vector e(0)

0 : B (L2 (IR+ ) k) C,
I
0 (B) = he(0), Be(0)i .
For any state on B(h) the map
: B C,
I
= 0
is a state on B. The pair (B, ) is a quantum probability space.

Because of the embeddings (5.2), if we set
B0] = B(h),
Bt] = B h (L2 (0, t) k)
for all t 0, then we obtain filtration of B.

For all t 0 we can define a conditional expectation
IEt] : B Bt]
as follows. Let B be an element of B. The sesquilinear form on h (L2 (0, t) k)

(v, u) (v e(0[t )), B(u e(0[t ))
(v, u h (L2 (0, t) k)) is bounded. The conditional expectation of B with respect to
Bt] , denoted IEt] [B], is the element of Bt] representing this sesquilinear form. Denoting by

t] the orthogonal projection of H onto the closed subspace h L2(0, t) k) e(0[t ) we
have

IEt] [L] = t] Lt]
1l(L2 (t,)k).
2
h(L (0,t)k)
The properties of conditional

expectations (see Chapter 1, Section 4) are easily checked.

Clearly, the family IEt] t0 is also projective.
5.1. QUANTUM STOCHASTIC CALCULUS
85
For each t 0 let t be the right shift on the interval [0, t] defined on L2 (IR+ ) k by

f(x t) if x > t
.
(t f)(x) =
0
if x t

Let (t) be the operators in L2 (IR+ ) k defined by second quantization of t,
(t)e(f) = e(t f)
(5.3)
for all f L2 (IR+ ) k. The operators t and (t ) are isometries for every t 0. Notice
that, for all s, t 0 we have
(s) (t+s) = (t),
(s )(t) = (s+t).
For each s 0 and each bounded operator x B(H) the operator (s )x(s) maps
h (L2 (s, ) k) into itself. Indeed we have the diagram
(s ) x (s)
h (L2 (s, ) k) H H h (L2 (s, ) k)
The canonical extension of (s )x(s) to H via ampliation will be denoted by s (x).
Clearly (s )s0 is a family of covariant shifts on B(H). Moreover, for all x B(h) and all
s 0, we have
s (x) = x.
This makes an important and relevant difference with the usual shift of classical Markov
processes (see [2], [68]).
It is easy to see that is a covariant shift with respect to the family (IEt] )t0 of conditional
expectations.
We now introduce the operator cocycles that we shall construct by means of quantum
stochastic calculus.
Definition 5.1 A family (X(t))t0 of bounded operators in H is called a left cocycle (resp.
right cocycle) if for every t, s 0 we have
X(t + s) = X(s)s (X(t)),
(resp. X(t + s) = s (X(t))X(s) )
(5.4)
defined by
Let F be the unique unitary operator in k
F e` = e`
if ` S,
F e+ = e ,
F e = e+ ,
onto Ce
and let E , E, E+ be the orthogonal projections of k
I , k, Ce
I + respectively.
We denote by IB the Belavkin algebra associated with (h, k)
n
o
| Lue = L ue+ = 0 for all u h
IB = L B(h k)
The map L 7 Lb where Lb = F L F is an involution of IB.
b {` | `, m S}
defined by
Let us introduce the quantum noises in H,
m
+
m (t)
= A+ (1(0,t) |em i),
if m S, ` = +
`m (t)
` (t)
+
(t)
`m (t)
= (1(0,t) |em ihe` |),
if m, ` S
= A(1(0,t) he` |),
if ` S, m =
= t 1l
if ` = +, m =
= 0
otherwise
86
5. QUANTUM FLOWS
where A+ , A, denote respectively the creation, annihilation and gauge operators in (L2 (IR+ )
k) defined, for each u h and each exponential vector e(f) by

d

ue(f + 1(0,t)em )
d
=0

d
i1(0,t) |em ihe` |
= i ue(e
f)
=0
d

= e` 1(0,t), f ue(f)
A+ (1(0,t) |em i)ue(f)
(1(0,t) |em ihe` |)ue(f)

A(1(0,t) he` |)ue(f)
where |em ihe` | denotes the operator on k
|em ihe` |x = he` , xi em .

Notice that quantum noises `m with either ` S or m S are martingales with respect to
the family of conditional expectations (IEt] )t0.
bS be the dense linear
Let D be a dense linear submanifold of h and let DS , DS and D
b generated by { ue` | u D, ` S }, { ue(f) | u
submanifolds
of h k, H and H

f MS respectively.
D, f MS }, ue` e(f) | u D, ` S,
b is called (DS , MS )-adapted if it satDefinition 5.2 A family (L(t))t0 of operators in H
isfies the following conditions:
bS for all t 0,
1. the domain of L(t) contains D
2. for each DS , f MS , t 0 we have the tensor product factorisation

L(t)e(f) = L(t)e(1(0,t) f) e(1(t,+) f).
A (DS , MS )-adapted family of operators L is called (DS , MS )-regular if the map
t 7 L(t)e(f)
is continuous for all DS , f MS .
In a similar way one can define (D, M0S ) and (D, MS )-adapted regular families of operators.
A (D, MS )-adapted family of operators X = (X(t))t0 in H is called unitary, isometric, coisometric, contractive if, for all t 0, the operators X(t) are unitary, isometric,
coisometric, contractive.
Let I(DS , MS ) be the vector space of regular (DS , MS )-adapted families of operators
b satisfying
L in H
hve+ e(g), L(t)ue` e(f)i = 0,
L(t)ue e(f) = 0
With every L I(DS , MS ) it is possible to associate

for all u, v D, f, g MS , ` S.
`

b defined by
the set Lm | `, m S of (D, MS )-adapted regular families of operators in H

ve(g), L`m (t)ue(f) = hve` e(g), L(t)uem e(f)i .
In the new notation one can write the stochastic integral
Z t X
L (t) =
L`m (t)dm
` (t)
0
`,mS
87
on the domain DS as a regular (D, MS )-adapted family of operators. This is very convenient
in the computations with the Ito formula; indeed, consider quantum stochastic integrals
Z t

X(t) = X0 +
L1 (s)dA+ (s) + L2 (s)d(s) + L3 (s)dA(s) + L4 (s)ds
0
Z t

Y (t) = Y0 +
M1 (s)dA+ (s) + M2 (s)d(s) + M3 (s)dA(s) + M4 (s)ds
0
where Lj , Mj are (hS , MS )-adapted families of bounded processes for j = 1, . . . , 4 and take
= Ce
is unitarily isomorphic
k =C
I so that h k
I Ce
I 0 Ce
I + . The Hilbert space h k
to h h h. The operators L,M corresponding to the above stochastic integrals and the
operator F can be represented by the 3 3 matrices
0 L3 L4
0 M3 M4
0 0 1l
L = 0 L2 L1 , M = 0 M2 M1 , F = 0 1l 0 .
(5.5)
0 0
0
0 0
0
1l 0 0
We can then write
X(t) = X0 + L (t),
Y (t) = Y0 + M (t),
Y (t) = Y0 + M b (t).
In the new notation the Ito formula can be written in the form
dY (t)dX(t) = dM b (t)dL(t) = dM bL (t).
This reduces the computation of a product of two stochastic differentials to the computation
of a product of two matrices.
The following are fundamental formulae of Boson Fock space quantum stochastic calculus; the proof can be found in [69], and in [74].
Proposition 5.3 Let L, M I(DS , MS ) and let X0 , Y0 be operators on h with domain
containing D. Consider the families (X(t))t0 , (Y (t))t0 of operators
X(t) = X0 + L (t),
Y (t) = Y0 + M (t).
For each v, u D, f, g MS the following formulae hold:

hve(g), X(t)ue(f)i = hve(g), X0 ue(f)i
Z t
+
hv(e + g(s))e(g), L(s)u(f(s) + e+ )e(f)i ds
(5.6)
hY (t)ve(g), X(t)ue(f)i = hY0ve(g), X0 ue(f)i

Z t
+
(Y (s) + F M (s)F )v(e + g(s))e(g),
0

(X(s) + L(s))u(f(s) + e+ )e(f)
hY (s)v(e + g(s))e(g), X(s)u(f(s) + e+ )e(f)i
(5.7)
ds
The above formulae can be written also in coordinate notation. The first, for example,
becomes
hve(g), X(t)ue(f)i = hve(g), X0 ue(f)i
88
5. QUANTUM FLOWS
X Z t

+
ve(g), L`m (s)ue(f) g`(s)fm (s)ds
0
`,mS
XZ
`S

ve(g), L`+ (s)ue(f) g` (s)ds
X Z t

+
ve(g), L
m (s)ue(f) fm (s)ds
+
mS 0
Z t

ve(g), L
+ (s)ue(f) ds
(5.8)
Coherent vectors in the above formulae are used as test vectors. Using the n-chaos
vectors instead the above formula can be written as follows
0
hvgn , X(t)uf n i = hve(g), X0 ue(f)i

E
X Z tD
0
+
vg(n 1), L`m (s)uf (n1) g` (s)fm (s)ds
`,mS
XZ tD
`S
E
0
vg(n 1) , L`+ (s)uf n g` (s)ds
X Z tD
mS 0
Z tD
E
0
(n1)
vgn , L
fm (s)ds
(s)uf
m
E
0
n
vgn , L
ds
(s)uf
+
(5.9)
0
0
for every pair n, n0 of non-negative integers with the conventions gn = f n = 0 if n < 0

and g0 = f 0 = e(0).
The fundamental formulae of quantum stochastic calculus can be used to prove the
following useful estimate of quantum stochastic integrals.
Corollary 5.4 Under the assumptions of Proposition 5.3 we have

Z t

2
2
k(X(t) X(s))ue(f)k
1 + kf(r)k2 dr
2ekf k exp
s
t
kL(r)u(e+ + f(r))e(f)k dr
for each t, s ]0, +[ with 0 s t.

We finish this preliminary section by recalling the basic results on quantum stochastic
differential equations. Since the solution cannot commute with the coefficient L IB we
shall distinguish between left (resp. right) quantum stochastic differential equations
dX(t) = X(t)dL (t),
(resp. dY (t) = dL (t)Y (t) )
where the solution acts by left (resp. right) multiplication on the stochastic differential. The
above equations have to be interpreted as
Z t
Z t
X(t) = X0 +
X(s)dL (s),
(resp. Y (t) = Y0 +
dL(s)Y (s) )
0
where X0 , Y0 are bounded operators on h (acting on H by standard ampliation) and (Y (t))t0,

(X(t))t0 are (h, MS )-regular adapted families of operators.
We refer to [69], and [74] for the proof of the following result
89
Proposition 5.5 For each M, L IB and U0 , V0 B(h) there exists (h, MS )-regular
adapted families of operators (U (t))t0 , (V (t))t0 satisfying the quantum stochastic differential equations
dU (t) = dM (t)U (t),
U (0) = U0 ,
(5.10)
dV (t) = V (t)dL (t),
V (0) = V0 .
(5.11)
Moreover:
1. (U (t))t0 is the unique (h, MS )-regular adapted family of operators satisfying (5.10),
2. (V (t))t0 is the unique (h, MS )-regular adapted family of operators such that
sup sup kV (s)ue(f)k <
0st kuk1
for each t > 0 and each f k satisfying (5.11).

Existence can be shown by the Picards iteration method. Uniqueness for (U (t))t0
follows immediately from Corollary 5.4 and Gronwalls lemma. Uniqueness for (V (t))t0
can be proved as in [74] (Prop. 26.1). The additional property needed will be called initial
space boundedness.
The following proposition (see, for example, [40] Prop. 3.1 p.150) gives a necessary and
sufficient condition for (U (t))t0 to be a contractive right cocycle.
Proposition 5.6 Let M be an element of IB and let (U (t))t0 be the unique (h, MS )-regular
adapted family of operators satisfying the right quantum stochastic differential equation
dU (t) = dM (t)U (t),
U (0) = 1l.
(5.12)
The operators U (t) (t 0) are contractions if and only if

F (M + M b + M bM ) 0.
(5.13)
In this case (U (t))t0 is a right cocycle.

Pm
Proof. Assume that (5.13) holds. Consider vectors in H of the form j=1 uj e(fj ) with
m
(uj )m
o
j=1 in h and functions (fj )j=1 in MS continuous at the point 0. Applying the It
formula (5.7) we have
Z t
X
2
2
kU (t)k = kk +
U (r)
uj (e + fj (r))e(fj ),
0
j
b
F (M + M + M b M )U (r)

uj (e + fj (r))e(fj ) dr
Then the inequality (5.13) implies that the operators U (t) are contractions for all t 0.
Conversely, if the operators U (t) are contractions, for all (fj )m
j=1 such that 0 is a continuity point for every function fj , we have
X

X
b
b
uj (e + fj (0))e(fj ), F (M + M + M M )
uj (e + fj (0))e(fj )
j
This proves that the inequality (5.13) holds.

1
2
2
= lim+
kU (t)k kk 0
t0 t
90
5. QUANTUM FLOWS
In this case the right cocycle identity follows by a uniqueness argument as in the paper
[58]. Indeed, if the operators U (t) (t 0) are contractions then for every fixed s 0 the
(h, MS )-regular adapted family of operators
X(t) = U (t + s),
Y (t) = s (U (t))U (s)
both satisfy the quantum stochastic differential equation (5.10) with the same initial condition, and therefore they coincide.
Taking the adjoint of the operators U (t) we immediately prove the following fact.
Corollary 5.7 Let L be an element of IB and let (V (t))t0 be the unique (h, MS )-regular
adapted initial space bounded family of operators satisfying the left quantum stochastic differential equation
dV (t) = V (t)dL(t),
V (0) = 1l.
(5.14)
Then the operators V (t) (t 0) are contractions if and only if
F (L + Lb + LLb ) 0.
(5.15)
In this case (V (t))t0 is a left cocycle.

Proof. If the operators V (t) are contractions then the adjoint operators U (t) = V (t)
are also contractions and satisfy the right quantum stochastic differential equation dU (t) =
dLb (t)U (t). Therefore, by Proposition 5.6, the inequality (5.15) holds.
Conversely, if (5.15) holds, then the unique (h, MS )-regular adapted initial space bounded
family of operators satisfying the right quantum stochastic differential equation dU (t) =
dLb (t)U (t) with U (0) = 1l is contractive. The adjoint family (U (t) )t0 is also (h, MS )regular adapted, contractive and satisfies (5.15). Since the (h, MS )-regular adapted family
of operators solving this equation is unique, it follows that the operators V (t) are also
contractions.
Clearly (V (t))t0 is a left cocycle because it is the adjoint of a right cocycle.
5.2
Time reversal and dual cocycles
In this section we define a time reversal on Fock space that transforms a right (resp. left)
cocycle into right (resp. left) cocycle. Moreover we give the relation between the quantum
stochastic differential equations satisfied by a cocycle and its time reversed counterpart.
Time reversal plays a fundamental role in the study of quantum stochastic differential
equations with an unbounded operator L when it is too hard to make sense of a right
differential equation and one needs to write a differential equation for the adjoint of a left
cocycle. In fact, if a left cocycle V satisfies (5.15), then the time reversed of the adjoint
right cocycle is again a left cocycle which can be shown to satisfy the left equation (5.15) roughly speaking - with L replaced by Lb .
Let t be the unitary time reversal on the interval [0, t] defined on L2 (IR+ ) k by

f(t s) if s t
(t f)(s) =
.
f(s)
if s > t
Let (t ) be the operator on (L2 (IR+ ) k) defined by second quantization of t
(t )e(f) = e(t f).
5.2. TIME REVERSAL AND DUAL COCYCLES
91
The operators are self-adjoint and satisfy

t t = 1l,
(t )(t ) = 1l.
Let Rt be the operator on B defined by

Rt : B B,
Rt (x) = (t )x(t ) .
The following lemma gives a useful relationship between time reversal and shift. Recall
that a bounded operator x is called t-adapted if it is factorised as in condition 2 of Definition
5.2.
Lemma 5.8 For every t, s 0 and every t-adapted bounded operator V (t) in H we have
Rt+s (Rt (V (t)) = s (V (t)).
(5.16)
Proof. For every v, u h and g, f L2 (IR+ ) k we have

hve(g), Rt+s (Rt(V (t)) ue(f)i = hve(t t+s g), V (t)ue(t t+s f)i .
(5.17)
A simple computation yields
f(s + r)
(t (t+s f))(r) = f(t + s r)
f(r)
if r < t,
if t < r < t + s,
if r > t + s.
Therefore we have the identity

t (t+s f) = (sf)1]0,t[ + (t+s f)1]t,t+s[ + f1]t+s,[ .
By the tensor product factorization of Fock space and t-adaptedness of V (t) we write the
right-hand side of (5.17) as

ve((s g)1]0,t[ ), V (t)ue((s f)1]0,t[ ) e(g1]t+s,[ ), e(f1]t+s,[ )

e((t+s g)1]t,t+s[ ), e((t+s f)1]t,t+s[ ) .
Computing the second and third scalar product we write the right-hand side of (5.17) in the
form

ve(g1]s,t+s[ ), (s)V (t)(s )ue(f1]s,t+s[ )
Z s

Z
exp
(
g f)(r)drf +
(
gf)(r)dr
0
t+s
hve(g), s (V (t))ue(f)i .

We recall now the notion of dual cocycle due to Journe (see [61] p. 294 and also [69] p.
174).
Proposition 5.9 Let (V (t))t0 be a left (resp. right) cocycle. The family of operators
(Ve (t))t0 defined by
Ve (t) = Rt (V (t) )
(5.18)
is a left (resp. right) cocycle.
92
5. QUANTUM FLOWS
Proof. By virtue of the cocycle property, for every s, t 0 we have

Ve (t + s)
= Rt+s (V (t + s) )
= Rt+s (s (V (t) ) V (s) )
= Rt+s (s (V (t) )) Rt+s (V (s) ) .
Applying Rt+s to both sides of (5.16) we have

Rt (V (t) ) = Rt+s (s (V (t) )) .
The same identity yields also
Rt+s (V (s) ) = Rt+s (Rs (Rs (V (s) ))) = t (Rs (V (s) )) .
Definition 5.10 The cocycle (Ve (t))t0 defined by (5.18) is called dual cocycle of the cocycle
(V (t))t0 .
We now study the relationship between a cocycle satisfying a left quantum stochastic
differential equation and its dual.
Lemma 5.11 Let t, s be two non-negative real numbers and let X(s) (resp. Y (t)) be an
s-adapted (resp. t-adapted) bounded operator. For every u, v h and every f, g MS we
have
hY (t)ve(g), t (Rs (X(s)))ue(f)i = hY (t)ve(t,t+s g), t (X(s))ue(t,t+s f)i
(5.19)
where t,t+s is the time reversal operator on the interval [t, t + s] defined by

f(2t + s r) if t < r < t + s,
(t,t+s f)(r) =
f(r)
otherwise.
Proof. Since the homomorphisms t and Rs are normal it suffices to prove the lemma for
the factorised operators
X(s) = X0 Xs ,
Y (t) = Y0 Yt
where X0 , Y0 are bounded operators in the initial space and Xs (resp. Yt ) is a bounded
operator in the factor (L2 (0, s)k) (resp. (L2 (0, t)k)) of the Fock space (L2 (IR+ )k).
In this case
the left-hand side of (5.19) can be written as the product of the factors
hY0v, X0 ui, Yt e(g1[0,t] ), e(f1[0,t] ) and

e(g1[t,[ ), (t)(s )Xs (s )(t )e(f1[t,[ )

= e(s t(g1[t,[ )), Xs e(s t(f1[t,[ ))
A simple computation shows that
s t(g1[t,[ ) = t t,t+sg,
s t(f1[t,[ ) = tt,t+s f.
Therefore we have

e(s t (g1[t,[ )), Xs e(s t(f1[t,[ ))
= he(t t,t+sg)), Xs e(t t,t+s f)i

= e((t,t+s g)1[t,[ ), t (Xs )e((t,t+s f)1[t,[ ) .

The product of this with Yt e(g1[0,t] ), e(f1[0,t] ) is equal to
hYt ve(t,t+s g), t (Xs )ue(t,t+s f)i .
Multiplication by the initial space factor hY0 v, X0 ui then yields (5.19).
5.3. QUANTUM STOCHASTIC DIFFERENTIAL EQUATIONS
93
Proposition 5.12 Let L be an element of IB satisfying the inequality (5.15) and let (V (t))t0
be the unique (h, MS )-regular adapted contractive left cocycle (V (t))t0 satisfying (5.14).
Then the dual cocycle (Ve (t))t0 satisfies the left quantum stochastic differential equation
dVe (t) = Ve (t)dLb (t),
Ve (0) = 1l.
Proof. By virtue of Proposition 5.9, for every t, s 0 we have

Ve (t + s) Ve (t) = Ve (t)t Ve (s) 1l .
Therefore, applying (5.19), we obtain
D

E
ve(g), Ve (t + s) Ve (t) ue(f)
D
E
=
Ve (t) ve(g), t (Rs (V (s) 1l)) ue(f)
D
E
=
Ve (t) ve(t,t+s g), t (V (s) 1l) ue(t,t+s f)
The action of t on a stochastic integral can be computed explicitly as follows
Z s

t (V (s) 1l) = t
dLb (r)V (r)
=
0
t+s
dLb (r)t (V (r t) ) .
t
Therefore, by the first fundamental formula of quantum stochastic calculus (5.6), we have
D

E
ve(g), Ve (t + s) Ve (t) ue(f)

Z t+s
e
=
V (t) ve(t,t+s g),
dLb (r)t (V (r t) ) ue(t,t+s f)
=
t+s
t
Ve (t) v(e + g(2t + s r))e(t,t+s g),
E
Lb t (V (r t) ) u(f(2t + s r) + e+ )e(t,t+s f) dr.
Dividing by s and letting s tend to 0, for every t which is a continuity point for both g and
f we have
E
E D
d D
ve(g), Ve (t)ue(f) = v(e + g(t))e(g), Ve (t)Lb u(f(t) + e+ )e(f) .
dt
5.3
Quantum stochastic differential equations
In this section we recall the main results on left quantum stochastic differential equations
of the form (5.14) with a possibly unbounded operator L giving in particular necessary and
sufficient conditions in order the operators V (t) to be isometries, coisometries or unitaries.
These extend the results of [37] [50], [52] for special classes of operators L.
Several results show that good operator cocycles satisfy a quantum stochastic differential equation (see, for example [7], [40], [57]). A full characterization (a quantum analogue
94
5. QUANTUM FLOWS
of the classical Stones theorem on strongly continuous unitary groups), however, is not
available.
As a first step we complete the study of the case when L is bounded.
Proposition 5.13 Let L IB and let (V (t))t0 be the unique (h, MS )-regular adapted
initial space bounded process solving the quantum stochastic differential equation
V (0) = 1l.
(5.20)
Then:
1. the operators (V (t))t0 are contractions if and only if one of the following equivalent
inequalities hold

L + Lb + LLb F 0,
F L + Lb + Lb L 0,
(5.21)
2. the operators (V (t))t0 are isometries if and only if
L + Lb + Lb L = 0,
(5.22)
3. the operators (V (t))t0 are coisometries if and only if

L + Lb + LLb = 0,
(5.23)
4. the operators (V (t))t0 are unitary if and only if

L + Lb + Lb L = 0,
L + Lb + LLb = 0.
(5.24)
Proof. Clearly the operators V (t) (t 0) are contractions if and only if the first inequality
(5.21) holds by Corollary 5.7.
The equivalence of the second inequality (5.21) and contractivity of the operators V (t)
can be shown in the same way considering the dual cocycle Ve . In fact contractivity of Ve is
equivalent to that of V and the dual cocycle satisfies a left quantum stochastic differential
equation with Lb instead of L by Proposition 5.12.
We prove then 2. For every v, u h and every f, g MS formula (5.6) yields
D
E
Ve (t) ve(g), Ve (t) ue(f) = hve(g), ue(f)i
Z t

+
Ve (s) v(e + g(s))e(g), F (L + Lb + Lb L)Ve (s) u(e + f(s))e(f) ds
0
Since Ve (t) = Rt (V (t)) this shows that the operators (V (t))t0 are isometries if and only if
(5.22) holds.
The third statement can be proved in the same way considering the right cocycle (V (t) )t0.
The fourth statement follows immediately from 2 and 3.
Remark. When L is the operator (5.5) we have
0 L3 + L1 + L1 L2
b
b
L + L + L L = 0 L2 + L2 + L2 L2
0
0
L4 + L4 + L1 L1
L1 + L3 + L2 L1
0
In this case, condition (5.22), (resp. (5.23)) is equivalent to the well-known condition for
the solution of a quantum stochastic differential equation to be isometric (resp. coisometric)
found in the seminal paper [58] by R.L. Hudson and K.R. Parthasarathy.
95
Note that, if the operators (Lk )4k=1 are bounded (for simplicity), the identity L4 + L4 +
L1 L1 = 0 implies that
1
L4 + L1 L1
2
has the form iH where H is bounded self-adjoint operator. Therefore, the choice
L1 = L2 = L3 = 0,
L4 = iH,
yields a group of unitary operators (V (t))t0 satisfying the Schroedinger equation

dV (t) = iHV (t)dt.
In this sense (5.20) is a generalization of the Schroedinger equation.
The following propositions motivate partially the assumptions under which we shall study
the quantum stochastic differential equation (5.20) with an unbounded operator L.
Let us recall first the following (see, for example, [61] or [69] Ch. VI, Sect. 12).
Proposition 5.14 Let (V (t))t0 be a contractive left cocycle in H. Then the family (P (t))t0
of operators on h defined by
P (t) = IE0] [ V (t) ]
(5.25)
is a contraction semigroup in h. If the cocycle is strongly continuous then the semigroup
(P (t))t0 is also.
Proof. Since (t )t0 is a covariant shift and V is a cocycle for every t, s 0 we have
P (t + s)
=
=
=
IE0] [V (t + s)]
IE0] [V (s)s (V (t))]
IE0] [V (s)IEs] [s(V (t))]]
IE0] [V (s)s (IE0] [V (t)])].
Notice that s (P (t)) = P (t) because the shift (t )t0 leaves invariant operators on the initial
space. Hence we have
P (t + s) = IE0] [V (s)]P (t) = P (s)P (t).
Therefore (P (t))t0 is a semigroup. Moreover, for every u h, we have
k(P (t + s) P (t))uk =
sup
|hv, (P (t + s) P (t))ui|
vh, kvk=1
sup
|hve(0), (V (t + s) V (t))ue(0)i|
vh, kvk=1
k(V (t + s) V (t))ue(0)k .
Strong continuity of V implies then the strong continuity of (P (t))t0. A similar argument
shows that the semigroup is also contractive.
From now on we shall always suppose that (P (t))t0 is strongly continuous. The infinitesimal generator G of P is defined as the set of u h such that the limit
lim t1 (P (t)u u)
t0+
exists in the strong (or equivalently weak) topology of h.

The following natural hypothesis on the operator L that will be in force throughout the
rest of this chapter.
Hypothesis L
96
5. QUANTUM FLOWS
1. There exists an operator G which is the infinitesimal generator of a strongly continuous
contraction semigroup in h and a core D for G such that the domain of the operator
L contains the domain DS,G
DS,G = { we + uf + ve+ | w h, u, v D, f k } ,
2. for all u, v D, w h, we have
Lwe = 0,
E+ L(ve+ + uf) = 0
hw, Gvi = hwe , Lve+ i
(5.26)
3. for all x DS,G we have

hF x, Lxi + hLx, F xi + hF Lx, Lxi 0.
(5.27)
Remark. The above hypothesis implies that the operators G and L`+ satisfy the hypothesis A introduced in Chapter 3. Indeed, for every u D, taking x = ue+ , (5.27) reads
as
`

hGu, ui +
L+ u, L`+ u + hu, Gui 0.
`=1
Since D is a core for G, the operators L`+ can be extended to D(G) so that the above in
equality holds also for u D(G).
Remark. Since the coefficient L is unbounded but the solutions we shall construct are
contractive the left quantum stochastic differential equation will be interpreted through
quantum stochastic integrals defined on vectors of the form ue(f) with u D. We shall
stress this fact by speaking of quantum stochastic differential equations in DS .
can be approximated by a sequence of
Under the hypothesis L an operator L in h k
bounded operators (Ln )n1 which are the infinitesimal generators of contractive cocycles.
For every integer n 1, let R(n; G) be the bounded operator (n1l G)1. Note that
the adjoint operator G is the infinitesimal generator of the dual contraction semigroup and
we have R(n; G ) = R(n; G) . Moreover the well-known properties of resolvent operators
relations for all w h, v D(G) yield
lim nR(n; G)w = w,
lim nGR(n; G)v = Gv.
satisfying the hypothesis L. For each n 1

Proposition 5.15 Let L be an operator in h k
with domain DS defined by
let In , Ln be the operators in h k
In = nR(n; G)E + E + nR(n; G)E+ ,
Ln = In LIn .
(5.28)
Then the operator Ln has a bounded extension which is an element of IB satisfying the
inequalities (5.21) and its uniform norm can estimated by
kLn k 2 n + 3 n + 1 .
Moreover, for all DS , we have
lim Ln = L.
(5.29)
97
Proof. The operator Ln satisfies the inequality (5.27). Indeed it suffices to write (5.27)
for vectors of the form In x with x DS,G (the vector In x belongs to the domain of L by
the well-known properties of the operators R(n; G) and the remark after the hypothesis L)
and use the commutation of the operator F with the operators In and In .
Moreover, because of (5.26), in order to prove that Ln is bounded, it suffices to estimate
of the form PJ uj (fj + e+ ) with uj D, fj MS
the norm of Ln for vectors in h k
j=1
for all j. In this case we have
kLn k kELn Ek + kE Ln Ek + kELn E+ k + kELn E+ k
(5.30)
The norm of E Ln E+ can be estimated using the third identity (5.26). We have in fact,
for each v, u D,
hve , ELn E+ ue+ i = n2 hR(n; G)v, GR(n; G)ui
Then, using the identity GR(n; G) = nR(n; G) 1l and the contractivity of nR(n; G), we
can estimate the norm of GR(n; G) by 2. Therefore we obtain the inequality
|hve , ELn E+ ue+ i| 2n kvk kuk
which implies
kE Ln E+ k 2n.
(5.31)
Taking a vector x DS in the range of E from the inequality (5.27) for Ln we obtain
immediately the estimates
kE + ELn Ek 1,
kELn Ek 2.
(5.32)
Consider now a vector x = ue+ with u D. The inequality (5.27) for Ln yields
2
2n2 <e hR(n; G)u, GR(n; G)ui + kELn E+ ue+ k 0

Hence, using again the equality GR(n; G) = nR(n; G) 1l, and the contractivity of nR(n; G)
we get the inequality
2
2
kELn E+ ue+ k 4n kue+ k
which implies
kELn E+ k 2 n.
(5.33)
To estimate the norm of ELn E we apply inequality (5.27) for Ln with vectors x of
the form x0 + ve+ with x0 finite linear combination of vectors in DS in the range of E,
v D and IR. It turns out that the number
hx0, Lx0i + hLx0, x0i + hLx0 , ELx0i
+2 (hve , Ln ve+ i + hLn ve+ , ve i + hLn ve+ , F Lnve+ i)
+2<e hLx0, F Lnve+ i + 2<e (hLx0 , ve i + hLn ve+ , x0i)

2
2
= 2<e hx0 , Lx0i + kLx0 k + 2 2<e hve , Lnve+ i + kELn ve+ k
+2<e (hLx0 , F Lnve+ i + hLx0 , ve i + hLn ve+ , x0i)
is real negative for every IR. Let
2
a =
2<e hve , Lnve+ i kELn E+ ve+ k
c =
b =
2<e hx0 , Lx0i kELx0 k = k(E + EL)x0 k + kx0k2

<e {hve , ELn Ex0 i + h(E + ELnE)x0 , ELnE+ ve+ i}
98
5. QUANTUM FLOWS
Therefore we have the inequality

a2 2b + c 0.
Applying (5.27) for Ln with x = ve+ and x = x0 separately we find that a and c are
nonnegative. Moreover the estimates (5.31) and (5.32) yield
a 4nkvk2,
Therefore |b|
c kx0k2 .
(5.34)
ac and we obtain the estimate
|<e hve , ELn Ex0i| 2 ac + |h(E + ELn E)x0, ELn E+ ve+ i|

Multiplying v by a complex number of modulus 1 such that the scalar product in the righthand side becomes real, from (5.32), (5.33), (5.34) we obtain the inequality
|hve , E LnEx0 i| 4 n kvk kx0k

for all v D which implies
kELn Ek 4 n
Using (5.32), (5.33), (5.34) and the above inequality from (5.30) we obtain the claimed
estimate for the norm of Ln. Clearly Ln IB because of (5.26).
Now (5.29) can be proved by decomposing Ln as we did to obtain (5.30) and using the
properties of resolvent operators (R(n; G ))n1. In fact, for every u D and f k, we have
2
k(E Ln E+ E LE+ )ue+ k
2

(n R(n; G )GR(n; G) G)u 2
2
k(ELn E+ ELE+ )ue+ k = kEL(nR(n; G)u u)e+ k

2<e hnR(n; G)u u, G(nR(n; G)u u)i
(E Ln E ELE)uf = E(nR(n; G )L L)uf
(ELn E ELE)uf
0.

satisfying hypothesis L and let (Ln )n1
Proposition 5.16 Let L be an operator on h k
be a sequence of elements of IB satisfying (5.27) such that, for all DS , we have
lim Ln = L.
For all integer n let Vn be the unique (h, MS )-adapted regular contractive solution of the
quantum stochastic differential equation in h
dVn (t) = Vn (t)dLn (t),
Vn (0) = 1l.
There exist a subsequence (nm )m1 and operators V (t) defined by

V (t) = w lim Vnm (t)
m
(5.35)
satisfying the quantum stochastic differential equation in DS

V (0) = 1l.
(5.36)
99
Proof. For each n 1 and each u D, f MS by the Ito formula (5.7), Corollary 5.4
and contractivity of Vn (t) we have
Z t
2
2
k(Vn (t) Vn (s)) ue(f)k 2 exp(1 + 2 kfk )
kLn u(f(r) + e+ )k2 dr.
(5.37)
s
Let N (f) be the number of nonzero fj with j S and let c0(u, f) be the constant

n
o
2
2
2
max
max
kLn uej k , kLn ue+ k , kuk , 1 .
nN
j{ kS|fk 60 }
A straightforward computation yields

kLn u(f(r) + e+ )k2 (1 + N (f))c0 (u, f) 1 + kf(r)k2 .
Therefore, for each y H, u D, f MS , the sequence of functions on [0, +[

t 7 hy, Vn (t)ue(f)i
is equicontinuous and equibounded. Hence, by the Ascoli-Arzel`
a theorem, there exists a
subsequence uniformly convergent on every bounded interval of [0, +[. The diagonalisation
argument and the separability of H allow to find a subsequence (Vnm )m1 weakly convergent
to an (h, MS )-adapted contractive process V .
Lower semicontinuity of the norm with respect to the weak convergence and (5.37) imply
that V is strongly continuous. Moreover, for all IN , by (5.6) we have

ve(g), Vnm (t)ue(f)
= hve(g), ue(f)i
Z t
+
hv(e + g(r))e(g), Vnm (r)Lnm u(f(r) + e+ )e(f)i dr
0
Letting m tend to infinity, contractions Vnm (s) converge weakly to V (s) and vectors Lnm u(f(r)+
e+ ) converge strongly to Lu(f(r) + e+ ). Therefore the desired conclusion follows.
The above propositions allow us to prove immediately the following
satisfying hypothesis L. Then there exists a
Theorem 5.17 Let L be an operator on h k
(D, MS )-adapted regular contractive solution of the quantum stochastic differential equation
(5.36).
Arguing by induction on the number n of the n-th chaos of the Fock space as in [39] can
prove the following uniqueness result due to A. Mohari ([70]).
satisfying hypothesis L. Then the (h, MS )Theorem 5.18 Let L be an operator on h k
adapted regular contractive solution of the quantum stochastic differential equation (5.36) is
unique.
Proof. Let (X(t))t0 be the difference of two (h, MS )-adapted regular contractive solution
of (5.36). Applying the first fundamental formula (5.9) we have
Z tD
E
0
n0
n
hvg , X(t)uf i =
vgn , X(s)Guf n ds
+
X Z tD
XZ tD
`S
E
0
vg(n 1), X(s)L`+ uf n g` (s)ds
X Z tD
mS
E
0
vg(n 1) , X(s)L`m uf (n1) g` (s)fm (s)ds
`,mS
E
0
(n1)
vgn , X(s)L
fm (s)ds
m uf
100
5. QUANTUM FLOWS
0
for every u, v D and every pair n, n0 of integer numbers with the convention gn = f n =
0 if n < 0 and g0 = f 0 = e(0).
We prove that the left-hand side vanishes by induction on n + n0.
For every > 0 the bilinear form on h
Z
exp(t) hve(0), X(t)ue(0)i dt
(v, u)
0
is bounded because kX(s)k 2 and

Z

exp(t) hve(0), X(t)ue(0)i dt 21kvk kuk.

0
Hence there exists a bounded operator R in h such that

Z
hv, Rui =
exp(t) hve(0), X(t)ue(0)i dt.
0
The first fundamental formula (5.9) for n = n0 = 0, u, v D, yields

Z
Z t
hv, Rui =
exp(t)dt
hve(0), X(s)Gue(0)i ds
0
0
Z
Z
=
hve(0), X(s)Gue(0)i ds
exp(t)dt
0
= hv, R Gui
We have then
R(1l G)u = 0
for every u D. Since D is a core for G, the linear manifold (1l G)(D) is dense in h.
Thus R vanishes. Therefore hve(0), X(t)ue(0)i also vanishes for every t 0.
This establishes our claim for n + n0 = 0. Suppose it has been established for every n, n0
such that n + n0 m. Then, for every n, n0 with n + n0 = m + 1, the induction hypothesis
allows us to write formula (5.9) as
Z tD
E
0
0
hvgn , X(t)uf n i =
vgn , X(s)Guf n ds.
0
0
The same argument we used in the case when n = n0 = 0 shows then that hvgn , X(t)uf n i
vanishes for every t 0.
This completes the induction argument and the proof.
Uniqueness allows us to deduce the cocycle property as in the paper [58].
satisfying hypothesis L. Then the unique
Corollary 5.19 Let L be an operator on h k
(h, MS )-adapted regular contractive solution of the quantum stochastic differential equation
(5.36) is a left cocycle.
Proof. In fact, since the operators V (t) (t 0) are contractions then for every fixed s 0
the (h, MS )-regular adapted family of operators
X(t) = V (t + s),
Y (t) = V (s)s (V (t))
satisfy both the quantum stochastic differential equation (5.36) with the same initial condition. Therefore they coincide.
5.4. UNITARY SOLUTIONS
5.4
101
Unitary solutions
In this section we will study conditions on the operator L (satisfying the hypothesis L) for
the solution (V (t))t0 of (5.36) to be a family of isometries (resp. coisometries, unitaries).
Proposition 5.13 suggests necessary condition: the left-hand side of (5.27) vanishes for
x DS . This condition, as AA for quantum dynamical semigroups, unfortunately is not
sufficient. However we will show that, under the hypothesis L, the operators (V (t))t0 are
isometries if and only if an associated quantum dynamical semigroup is identity preserving.
This reduces our problem to the problem studied in Chapter 3, Sections 5 and 6.
With a given contractive cocycle V , by Theorem 2.24, we can associate quantum dynamical semigroup T = (Tt )t0 on B(h) defined by
hv, Tt(x)ui
= hV (t)ve(0), xV (t)ue(0)i
D
E
=
Ve (t) ve(0), xVe (t)ue(0)
(5.38)
for all t 0, x B(h), v, u D. Here the second identity follows from the properties of
time reversal operators Rt and the fact that Rt e(0) = e(0).
In this section the following hypothesis on the operator L will be in force
Hypothesis LL
satisfies the hypothesis L and, for all x DS,G we have
The operator L in h k
hF x, Lxi + hLx, F xi + hF Lx, Lxi = 0.
(5.39)
satisfying the hypothesis LL. For every [0, 1[

Lemma 5.20 Let L be an operator in h k
consider the operators in h k

1l() = E + E + E+ ,
L() = 1l() L1l() + (1 2)E LE+
(5.40)
The operator L() satisfies the hypothesis L.

Proof. The operators 1l() and F commute and DS,G is invariant under 1l() . Therefore
L() obviously satisfies 1 and 2 of hypothesis L.
Moreover, for every x DS,G , we have
D
E D
E D
E
F x, L()x + L() x, F x + F L() x, L() x
D
E D
E D
E
=
F 1l() x, L1l() x + L1l() x, F 1l() x + F L1l() x, L1l() x
+ (1 2 ) (hF x, ELE+ xi + hE LE+ x, F xi)
= (1 2 ) (hF E+ x, LE+ xi + hLE+ x, F E+xi)
= (1 2 )hLE+ x, ELE+ xi 0.
The above lemma allows us to state the following
satisfying the hypothesis LL. For every
Proposition 5.21 Let L be an operator in h k
()
[0, 1[ let V
be the unique regular (h, MS )-adapted contractive process satisfying the
quantum stochastic differential equation
dV () (t) = V () (t)dL() (t),
V () (0) = 1l
in h and let T () be the associated quantum dynamical semigroup. The following facts hold:
102
5. QUANTUM FLOWS
1. for every t 0
w lim V () (t) = V (t)
1
where (V (t))t0 is the unique solution of (5.36),

()
2. for every t 0 and every positive operator x B(h) the operators (Tt
increasing in .
(x))[0,1[ are
Proof. As in the proof of Proposition 5.16, applying Corollary 5.4 we can find the estimate
Z t

2

()

()
V
1 + kf(r)k2 dr
(t)
V
(s)
ue(f)
c(u,
f)

s
for each u D and f M where c(u, f) is a constant independent of . The equicontinuity

and diagonalization argument in the proof of Proposition 5.16 can be used also here to show
that for every sequence (n )n0 converging to 1 there exists a subsequence (nm )m0 such
that the limit
w lim V (nm ) (t)
m
exists for every t 0.

Moreover, it can be shown as in the proof of the same proposition, that the limit satisfies
(5.36). Therefore it coincides with the unique solution of (5.36) by Theorem 5.18.
Summing up: every sequence (V (n ) )n0 admits a subsequence which converges weakly
to the unique solution of (5.36). This proves the claim 1.
Let T () be the quantum dynamical semigroup associated with V () via (5.38) and
let (P (t))t0 be the strongly continuous contraction semigroup on h generated by G. A
straightforward computation yields
E
d D ()
V (s)P (t s)ue(0), xV () (s)P (t s)ue(0)
ds
E
XD
2
V () (s)L`+ P (t s)ue(0), xV () (s)L`+ P (t s)ue(0)
`S
for every u D and x B(h). Therefore, integrating on [0, t], we have (dropping the index
+ of L`+ )
D
E
()
u, Tt (x)u
= hP (t)u, xP (t)ui
(5.41)
Z
E
X tD
+ 2
L` P (t s)u, Ts() (x)L` P (t s)u ds.
`S
For every 2 1 < 1 and every positive operator x B(h) we have then
D
E
( )
( )
u, Tt 1 (x) Tt 2 (x) u
Z
E
X tD `
= 12 22
L P (t s)u, Ts(1 ) (x)L` P (t s)u ds
`S
22
XZ t D
`S
22

E
L` P (t s)u, Ts(1 ) (x) Ts(2 ) (x) L` P (t s)u ds
XZ t D
`S

E
L` P (t s)u, Ts(1 ) (x) Ts(2 ) (x) L` P (t s)u ds.
103
( )
( )
Estimating the norm of Tt 1 (x) Tt 2 (x) by 2kxk and computing the integral
D
E
XZ t

(1 )
(2 )
2
L` P (t s)u 2 ds
u, Tt (x) Tt (x) u
2kxk2
=
`S 0
2
2kxk2 kuk2
2kxk22kuk2
kP (t)uk2
Iterating the estimate and computing the integral we obtain

D
E
( )
( )
u, Tt 1 (x) Tt 2 (x) u

E
XZ tD
22
L` P (t s)u, Ts(1 ) (x) Ts(2 ) (x) L` P (t s)u ds
`S
2kxk24
224kxk
XZ t

L` P (t s)u 2 ds
`S
kuk .
Therefore, iterating n times, we have

D
E
( )
( )
u, Tt 1 (x) Tt 2 (x) u 222nkxk kuk2.
The statement 2 follows then letting n tend to infinity.
satisfying the hypothesis LL and let (V (t))t0
Theorem 5.22 Let L be an operator in h k
be the unique solution of (5.36). The quantum dynamical semigroup associated with the
cocycle V coincides with the minimal quantum dynamical semigroup T (min) associated with
the operators G, L`+ .
Proof. We use the notation of the proof of Proposition 5.21. Let x be a positive operator
B(h), u a vector in h and t 0. Letting tend to 1 it turns out that the limit quantum
dynamical semigroup satisfies the equations (3.16) and (3.15). Moreover, for [0, 1[, t 0,
by the same argument of the proof of Proposition 5.21, we have
()
Tt
(min)
(x) Tt
(x).
Therefore the limit quantum dynamical semigroup coincides with the minimal quantum
dynamical semigroup T (min) associated with the operators G, L`+ by the minimality of
T (min) (see Theorem 3.22).
Let T be the quantum dynamical semigroup associated with the cocycle V . The second
fundamental formula of quantum stochastic calculus (5.7) shows immediately that T satisfies
the equations (3.16) and (3.15). Therefore, by virtue of Theorem 3.22, we have
(min)
Tt(x) Tt
(x).
On the other hand the lower semicontinuity of the norm with respect to weak convergence
yields

2

hu, Tt(x)ui = x1/2V (t)ue(0)

2

lim inf x1/2V () (t)ue(0)
1
D
E
()
= lim inf u, Tt (x)u
1
D
E
(min)
=
u, Tt
(x)u .
104
5. QUANTUM FLOWS

We can now give necessary and sufficient conditions for the operators V (t) to be isometries.
satisfying the hypothesis LL and let (V (t))t0
be the unique contractive cocycle satisfying (5.36). The following conditions are equivalent:
1. the operators V (t) are isometries for every t 0,
2. for each v, u D and each t 0 we have
hV (t)ve(0), V (t)ue(0)i = hv, ui ,
3. the quantum dynamical semigroup associated with the cocycle V is conservative.
Proof. Clearly the first condition implies the second one.
By virtue of Theorem 5.22, the quantum dynamical semigroup associated with the cocycle V coincides with the minimal quantum dynamical semigroup T (min) associated with
the operators G and L`+ . Therefore we have
D
E
(min)
hV (t)ve(0), xV (t)ue(0)i = v, Tt
(x)u
for every v, u D, x B(h) and t 0. Taking x = 1l it is clear that 2 and 3 are equivalent.
To complete the proof it suffice then to show that 2 implies 1. We will prove by induction
that for each m 0, n, n0 IN with n + n0 = m, v, u D, g, f MS we have
D
E
0
n
V (t)vgn , V (t)uf n = n0 ,n hv, ui hg, fi .
(5.42)
This claim is clearly true for m = 0 because of condition 2. Suppose it has been established
for all n, n0 with n + n0 = m and take n, n0 such that n + n0 = m + 1.
Writing the Ito formula (5.7) of Boson Fock space quantum stochastic calculus using nchaos vectors and taking into the account the induction hypothesis, we obtain the equation
D
E
0
n
V (t)vgn , V (t)uf n = n0 ,n hv, ui hg, fi
Z tnD
E D
E
0
0
V (s)vgn , V (s)Guf n + V (s)Gvgn , V (s)uf n
+
0
D
Eo
0
+ V (s)ELve+ gn , V (s)ELue+ f n
ds
Let > 0 fixed and (as in the proof of Theorem 5.18) define a bounded operator R in h by
Z
E
D
0
hv, Rui =
exp(t) V (t)vgn , V (t)uf n dt.
0
A simple computation (again as in the proof of Theorem 5.18) yields

n
hv, Rui = n0 ,n hv, ui hg, fi + hv,

L(R)ui
i.e., since
L(1l) = 0 (in the form sense)
L(R c1l) = (R c1l)

n
where c = 1 n0 ,n hv, ui hg, fi .
105
Condition 2 (or the equivalent condition 3) and Proposition 3.31 imply then
n
R = 1n0 ,n hg, fi 1l.

Hence (5.42) follows for n + n0 = m + 1. This completes the proof.
Theorem 5.23 together with the results of Chapter 3 allow us to construct a large class of
isometric cocycles. Proposition 5.12 suggests a simple way to determine whether a cocycle
is coisometric. In fact this question could be answered applying Theorem 5.23 to the dual
cocycle.
The following proposition gives a natural condition on the operator L allowing to find
the quantum stochastic differential equation satisfied by the dual cocycle.
Hypothesis LLD We say that the operator L satisfies the hypothesis LLD if it satisfies
the hypothesis LL and:
e for G such that D
e S,G is contained in the domain of F LF ,
1. there exists a core D
e of F L F to D
e S,G satisfies the hypothesis L with D
e and G replacing
2. the restriction L
D and G,
3. there exists a sequence (Ln )n1 of elements of IB such that
s lim Ln = L,
for every DS,G ,
e,
s lim Lbn = L
for every DS,G .
n
n
satisfies the hypothesis LLD. Let

Proposition 5.24 Suppose that the operator L in h k
V and Z be the unique cocycles satisfying the quantum stochastic differential equations
V (0) = 1l,
dZ(t) = Z(t)dL
e(t),
Z(0) = 1l
in DS
eS
in D
Then Z coincides with the dual cocycle Ve of V .

Proof. Let (Ln )n1 be the sequence of bounded approximations of L. For every n 1 let
Vn be the unique cocycle satisfying the quantum stochastic differential equation
dVn(t) = Vn (t)dLn (t),
Vn(0) = 1l
By virtue of Proposition 5.12 the dual cocycle Ven satisfies the quantum stochastic differential
equation
dVen (t) = Ven (t)dLbn (t),
Ven (0) = 1l.
By the hypothesis LLD we can apply Proposition 5.16 and find subsequences (that we
denote as the given sequence for simplicity) such that
w lim Vn (t) = V (t),
n
w lim Ven (t) = Ve (t)

n
Then it can be shown as in the proof of Proposition 5.12 that the cocycle Ve satisfies the
same quantum stochastic differential equation satisfied by the cocycle Z.
Therefore, by Theorem 5.18, the two cocycles coincide.
The following Lemma shows that the sequence (Ln )n1 considered in Proposition 5.15
can often be used to check the hypothesis LLD.
106
5. QUANTUM FLOWS
Lemma 5.25 Let L be an operator in h k satisfying the hypothesis L. Suppose that there
e which is a core for G such that:
exists a domain D
eS ,
1. the domain of the operator L contains D
e (F L F ),
2. we have L
3. for all u D and all n 1, R(n; G)u belongs to the domain of G and the sequence
(G R(n; G)u)n1 converges.
Then the operator L satisfies the hypothesis LLD.
Proof. Let (Ln )n1 be the sequence (5.28).

e on D
eS .
We show that, under 1, 2 and 3 the sequence Lbn n>0 converges strongly to L
e S we have
In fact, for all D
Lbn
= ELbn E + E Lbn E + ELbnE+ + ELbn E+
= (ELE) + nR(n; G)F E+ L E + (E Ln E) F

+E Lbn E+
Clearly the first two terms converge strongly when n tends to infinity. The vector E+
e for all n, m IN , hence we have
can be written in the form ue+ with u D;

E Lb E+ E Lb E+ = k(Gn) u (Gm ) uk .
n
m
e is contained in D(G ), the limit as n tends
Since (Gn) u = n2 R(n; G)G R(n; G)u and D
to infinity of the fourth term exists. The third term can be written in the form
ELbnE+ = (E Ln E) F = (E + ELE) ELn E+ F

where the second identity follows from the equality (5.39). This shows that the sequence

e on D
e S consider
Lbn n>0 is strongly convergent. To show that the limit coincides with L
e S , x D(L).
e Because of 2 we have then,
yD

lim y, Lbn x =
=
=
=
lim hLn F y, F xi
hLF y, F xi
hy, F L F xi
D
E
e
y, Lx
5.5
Inner quantum flows
In this section, following [41], we show how to construct a class of quantum Markov processes with unbounded infinitesimal generator using unitary solutions of quantum stochastic
differential equations. Moreover we give a general condition in order the restrictions to an
abelian sub-*-algebra to be commutative.
Definition 5.26 . Let A be a unital sub-*algebra of B(h). A quantum flow on A is a family
j = (jt )t0 of identity preserving *-homomorphisms from A into B(H) such that:
1. for all x A, j0 (x) = x,
5.5. INNER QUANTUM FLOWS
107
2. for all x A, (jt (x))t0 is an (h, MS )-regular adapted process,

3. there exists a sub-*algebra A of A dense in A for the norm kk and structure maps
`
m
: A 7 A such that the quantum stochastic differential equation
X
`
djt (x) =
jt (m
(x))dm
j0 (x) = x
(5.43)
` (t),
`,mS
in DS is satisfied for all x A .

We refer to [69], [74] for results on quantum flows with bounded structure maps, in
particular, the existence theorem of M. Evans [34].
Definition 5.27 Let A be a commutative sub-*algebra of B(h). A quantum flow (jt)t0 on
A is commutative if, for all x, y A and all s, t 0, we have
[js (y), js+t (x)] = 0.
In the following we shall assume
Hypothesis LLF
An operator L in h k satisfies the hypothesis LLF if:
1. it satisfies LLD,
e is a dense submanifold of D,
2. D
3. V , Ve are the unique unitary solutions of the quantum stochastic differential equations
dVe (t) = Ve (t)dL
e(t),
V (0) = 1l,
Ve (0) = 1l,
in DS ,
eS .
in D
(5.44)
(5.45)
Note that V , Ve are strong limits of cocycles Vn , Ven satisfying the quantum stochastic
differential equation in H
dVen(t) = Ven (t)dL
en (t),
Vn (0) = 1l
Ven (0) = 1l
by Proposition 5.24.
satisfying the hypothesis LLF and let A be
a sub-*-algebra of B(h). Moreover let A be a sub-*-algebra of A which is dense in A for
the uniform norm kk . Suppose that:
e `, m S,
the strong limit
1. for all x A , w D,
s lim
(n) `
m (x)w
`
e
exists and defines a bounded operator m
(x) satisfying, for all u, v D,
v, m (x)u = F LbF ve` , xuem

+ ve` , xLbuem + F Lb F ve` , xLbuem
(5.46)
108
5. QUANTUM FLOWS
we have
2. for all x A , `, m S,

`
sup (n) m
(x)
n
< .
Let V , Ve be the unique unitary solutions of the quantum stochastic differential equations
eS . For all t 0 consider the *-homomorphism jt on A defined by
(5.44), (5.45) in D
jt (x) = V (t)xV (t).
Then (jt )t0 is a quantum flow on A with structure maps given by (5.46) satisfying the
quantum stochastic differential equation (5.43).
Proof. Under the present hypotheses we have
s lim Vn (t) = V (t),
n
s lim Ven (t) = Ve (t),

n
s lim Vn (t) = V (t)

n
for all t 0. Moreover the cocycles V and Ve satisfy quantum stochastic differential equations
(5.44), (5.45). For all n IN , u, v D, f, g MS and all x A we have then
hVn (t)ve(g), xVn (t)ue(f)i = hve(g), xue(f)i
E
X Z tD
`
+
Vn (s)ve(g),(n) m
(x)Vn (s)ue(f) g` (s)f m (s) ds
`,mS
(5.47)
`
the structure maps (n) m
where, for all `, m S,
are given by (5.46) with L replaced by Ln .
(n) `
For all x A the operators
m (x) in h are uniformly bounded by condition 2. Then,
we have
because of condition 1, for each w h and each l, m S,
lim
(n) `
m (x)w
`
= m
(x)w.
(5.48)
Hence, by the dominated convergence theorem, taking the limit of both sides of (5.47) we
obtain (5.43).
The following results give sufficient conditions in order that the restriction to a commutative sub-*-algebra of B(h) of a quantum Markov flow in Fock space with unbounded
structure maps to be commutative.
Proposition 5.29 Let : A 7 A be a linear map such that (1l )(A ) is dense in A
for the norm k k for all greater than a fixed positive number 0 . Let (Tt)t0 be a family
of linear maps Tt : A B(h) with the following properties:
1. for all t 0 and all x A, we have
kTt(x)k kxk ,
e the function t 7 hv, Tt (x)ui is continuous,
2. for all x A and all v, u D
e we have
3. for all x A , v, u D
hv, Tt(x)ui =
hv, Ts((x))ui ds.

0
Then Tt = 0 for all t 0.
(5.49)
5.5. INNER QUANTUM FLOWS
109
Proof. For all > 0 let Tb be the operator on B(h) defined by

D
E Z
b
v, T(x)u =
exp(s) hv, Ts (x)ui ds.

0
e Taking the Laplace transform of both sides of (5.49) and

for all x A and all v, u D.
integrating by parts the right-hand side we obtain
D
E
v, Tb ((1l )(x)) u = 0,
for all x A . This implies that Tb = 0 because of condition 1 and the density of
(1l )(A ) in A. Hence we obtain Tt = 0 for all t 0 by a well-known property of the
Laplace transform.
Remark. If (1l )(A ) is -weakly dense in A and the map x Tt(x) satisfying 1, 2
and 3 is also -weakly continuous and positive the same conclusion holds.
The following result, obtained in [41], generalises the corresponding result of Parthasarthy
and Sinha [75] for quantum flows with bounded structure maps.
`
Theorem 5.30 Let (jt)t0 be a quantum flow on A with structure maps (m
)`,mS with
`
m
: A 7 A
satisfying the quantum stochastic differential equation (5.43). Suppose that (1l+
)(A )
is dense in A for all greater than a positive number 0 . Then the following facts hold:
1. if A is commutative, then the quantum flow (jt )t0 is commutative,

2. the adapted family of operators (jt(x))t0 is the unique solution of the quantum stochastic differential equation (5.43) for all x A .
e
Proof. Fix y A and g, f MS . We will prove that, for all p, q IN , s, t 0, v, u D
and all x A we have
q

vg , [js (y), js+t (x)] uf p = 0
(5.50)
Let us first consider the case when n = 0. For all t 0 from the homomorphism property
of jt we have kjt (a)k kak . Hence we can define bounded operators Tt : A B(h) by
hv, Tt(x)ui = hve(0), [js (y), js+t (x)] ue(0)i
Write js+t (y) as the sum of js (x) and a stochastic integral on (0, t). Since js is a *homomorphism we have
[js (x), js(y)] = js ([x, y]) = 0.
Then, applying the Ito formula (5.7), we obtain the equation
hv, Tt (x)ui =
v, Tr (+
(x))u dr.
and (5.50) follows from Proposition 5.29. The proof can be completed by induction on
n = p + q by the same argument of the case n = 0.
110
5.6
5. QUANTUM FLOWS
Quantum diffusions
In this section we construct an inner quantum flow extending the flow of a classical multidimensional diffusions with smooth covariance and drift. This generalises the result obtained
in [41] for elliptic diffusions.
Let (y(t))t0 be a d-dimensional diffusion, as considered in Chapter 4, Section 2, satisfying the stochastic differential equation
dyj (t) =
d
X
jk (y(t))dwk (t) + bj (y(t))dt,
1jd
(5.51)
k=1
In this section we will always assume that and b satisfy the hypothesis D in Chapter 4, Section 2. The quantum extension of the semigroup obtained there guides the our
construction of the operator cocycle here.
Denote again a = .
e be the linear manifold C (IRd ;C)
Let h = L2 (IRd ;C)
I and let D and D
I of compact
c
support functions with continuous partial derivatives of all orders. Let S = { 1, . . ., d } and
we denote by Lb the restriction to DS
S = S { , + }. For every operator L in h k
of the operator F L F .
Let A = Cl0 (IRd ; IR) and A be the algebra generated by multiplication operators by a
function which is either constant or infinitely differentiable with compact support.
We define now an operator L which is the candidate infinitesimal generator of a cocycle
V satisfying the quantum stochastic differential equation
V (0) = 1l.
e be the restriction to DS of F L F .
Denote by L
e satisfy the hypothesis L then the following identities hold
If both the operators L and L
ELE + EL E + ELEL E = 0
ELE + EL E + EL ELE = 0
ELE+ + ELb E+ + ELELb E+ = 0
E LE+ + E Lb E+ + ELELb E+ = 0
(5.52)
(5.53)
(5.54)
(5.55)
Let us consider:
a dd matrix with entries in Cb4 (IRd ; IR) such that 1l+ is unitary (real orthogonal),
a d-dimensional vector in Cb3 (IRd ; IR),
a d-dimensional vector in Cb2(IRd ; IR) such that
j = bj
X
1X
(k akj ) +
` `j .
2
d
k=1
`=1
(5.56)
Define the operator L, the infinitesimal generator of a cocycle V , through its matrix
elements Lm
` .
D(Lm
` )= h
m
Lm
` = M (` )
) = H 1(IRd ;C)
I
D(L
`
d
X
L
=
`
j=1
if `, m S
j `j + `
if ` S
5.6. QUANTUM DIFFUSIONS
111
where M (`m ) denoted the multiplication operator by the function `m . Condition (5.54)
yields
ELE+ = (1l + ELE) ELb E+
or, in the coordinate notation, for m S,
1
d
D(Lm
I
+ ) = H (IR ; C),
Lm
+
d
X
(`m + `m ) L
`
`=1
d
X
`=1
(`m + `m )
d
X
j=1
`j j + `
where `m = 0 if ` 6= m and `m = 1 if ` = m.
Finally let L
+ be the operator
2
d
D(L
I
+ ) = H (IR ; C),
d
L
+
1 X 1 X
L` (L` ) +
(j j + j j )
2
2 j=1
`=1
Notice that (L
) (resp. (L
+ ) ) coincides with the operator L` (resp. G) given by
`
(4.10) (resp. (4.13)).
The results of Chapter 4, Section 2 allow us to prove immediately the following

+
Proposition 5.31 The operators L
satisfy the hypothesis AA as well as the op+ , L`

`
erators (L+ ) , (L ) . The associated quantum dynamical semigroups are Markov.

Proof. The operators L
+ , (L+ ) are infinitesimal generators of contraction semigroups
in h by Theorem A.3. Indeed these are second order differential operators with sufficiently
regular coefficients.
Moreover the linear manifold Cc (IRd ;C)
I is a core for both. Therefore the identity
(3.28), which is immediately checked for u Cc (IRd ;C)
I by our choice of the coefficients
+
of the differential operators L

and (L
+ , L`
+ ) , (L ) , is fulfilled. Hence condition
AA holds.

Finally, since the coefficients of L
+ , (L+ ) are sufficiently regular (Hypothesis D)
holds, we can apply Theorem 4.6 to concluded that both the quantum dynamical semigroups
+
associated with L
and with (L
+ , L`
+ ) , (L ) are Markov.
Let K and H be the operators

D(K) = H (IR ;C),
I
1 X
K=
L` (L` ) ,
2
D(H) = H 1 (IRd ;C),

I
d
iX
H =
(j j + j j )
2 j=1
`=1
Notice that both K and H are self-adjoint. (K is negative self-adjoint). For every n 1
let
Kn = n2 R(n; K)KR(n; K),
Hn = inHR(in; H)
112
5. QUANTUM FLOWS
Consider bounded approximations of L

(Ln )m
`
(Ln )
`
(Ln )m
+
(Ln )
+
Lm
`
=
=
=
nR(n; K)L
`
nLm
+ R(n; K),
if `, m S,
if ` S,
if m S,
if m = , ` = +.
Kn + iHn
Straightforward algebraic computations show that the operators Ln satisfy identities

(5.24). Therefore there exists a dual pair of unitary cocycles Vn , Ven such that
dVen (t) = Ven (t)dL

e (t).
n
(5.57)
The well-known properties of resolvent operators yield

s lim Ln = L,
n
s lim Lbn = Lb .
n
Theorem 5.32 The sequence (Vn (t))n1 (resp. (Ven (t))n1 ) converges strongly to a unitary
operator V (t) (resp. Ve (t)) for every t 0. The dual cocycles (V (t))t0 (resp. (Ve (t))t0)
satisfy the quantum stochastic differential equations
dVe (t) = Ve (t)dL

e(t)
(5.58)
with initial conditions V (0) = 1l, Ve (0) = 1l.

e is immediately checked by our choice of
Proof. The identity (5.39) for the operators L, L
m
the coefficients of the differential operators L` . Therefore there exists unique contractive
solutions of the above quantum stochastic differential equations.
Since the associated quantum dynamical semigroup are Markov by Proposition 5.31, the
cocycles (V (t))t0 , (Ve (t))t0 are isometric. Thus, being a pair of dual cocycles, they are
unitary.
By virtue of Proposition 5.16 there exists a subsequence (nm )m1 such that
V (t) = w lim Vnm (t),
m
Ve (t) = w lim Venm (t).

m
for every t 0.
The above limits are also strong because both the operators V (t) and Ve (t) are isometries.
Finally we can remove subsequences by the same argument of the proof of Proposition
5.21 because limit cocycles are unique.
We find now, applying Theorem 5.28, the quantum stochastic differential equation satisfied by the quantum flow obtained by conjugation through the unitary cocycle V . In order
`
to show that the operators (n) m
(f) with f A are uniformly bounded in norm we need
the following easy lemma.
Lemma 5.33 For every f A the operators
R(n; K) [K, [K, f]] R(n; K)
are uniformly bounded in norm.
(5.59)
5.6. QUANTUM DIFFUSIONS
113
Proof. A straightforward computation shows that the double commutator [K, [K, f]] is
given by the multiplication operator by
d
d
X
1 X
e
ajk (x)(j k f)(x) +
bk (x)(k f)(x)
2
j,k=1
k=1
where
d
d
X
X
ebk (x) = 1
(m amk )(x) +
m (x)mk (x).
2 m=1
m=1
Indeed the second order differential operator f [K, [K, f]] must coincide with the second
order differential operator given by the choice = 0 in (5.56).
The conclusion follows from the fact that the operators R(n; K) are contractions.
Proposition 5.34 Let A = Cl0(IRd ; IR) and let A be the algebra generated by compact
support functions and constant functions and let (V (t))t0 be the unitary cocycle satisfying
the quantum stochastic differential equation (5.58). The quantum flow
jt (x) = V (t)xV (t) ,
x B(h)
(5.60)
satisfies, for f A , the quantum stochastic differential equation (5.43) with structure
maps
`
m
(f)
m
+
(f)
=
=
0,
`, m S
d
X
(km + km ) kj (j f),
mS

k` + k` kj (j f),
`S
k,j=1
` (f)
d
X
k,j=1
+
(f)
1
ajk(j k f) + bj (j f)
2
where (bj )1jd is given by (5.56).

Proof. Clearly A is dense in A for the uniform norm.
We check the hypotheses 1 and 2 of Theorem 5.28. For every f A , n 1, `, m
{1, . . ., d}, by the basic rules of quantum stochastic calculus, we have
(n) `
m (f)
(n) m
+ (f)
`
m
(f) = 0,
nLm
+ R(n; K)f
+n
d
X
(5.61)
+
nf(L
m ) R(n; K)
km f(L
) R(n; K)
k
(5.62)
k=1
(n)
` (f)
nR(n; K)L
f + nfR(n; K)(L`+ )
`
+nR(n; K)
d
X
`
L
k fk
(5.63)
k=1
(n)
+ (f)
i[Hn, f] + Kn f + fKn
+
d
X
j=1
nR(n; K)L
fn(L
) R(n; K)
`
`
(5.64)
114
5. QUANTUM FLOWS
Therefore hypothesis 1 of Theorem 5.28 follows from the well-known properties of resolvent
operators.
`
In order to check the hypothesis 2 notice first that structure maps (n) m
with `, m
{1, . . ., d} are obviously uniformly bounded in n. Then, due to the structure relation,
(n)
+ (yx)
= y(n) +
(x) +(n) +
(y)x +
d
X
(n)
`
` (y)(n) +
(x)
`=1
it suffices to prove that, for every f A , the operators

in n.
A direct computation from (5.64) yields
(n)
+ (f)
(n)
+ (f)
are uniformly bounded
i[Hn, f] + Kn f + fKn
+
d
X
nR(n; K)L
fn(L
) R(n; K)
`
`
j=1
i[Hn, f] + nR(n; K)+

(f)nR(n; K)
+n2 R(n; K)[R(n; K), f] + n2 [f, R(n; K)]R(n; K)

The second term is obviously uniformly bounded. The first one can be written in the form
i[Hn, f] =
=
=
nHR(in; H)f nfHR(in; H)

nH[f, R(in; H)] + n[f, H]R(in; H)
HR(in; H)[H, f](nR(in; H)) + [f, H](nR(in; H)).
Since [f, H] is a bounded multiplication operator, the operators [Hn, f] are uniformly bounded.
The sum of the third and fourth term can be written in the form
=
=
=
n2(R(n; K))2 [f, n K]R(n; K) + n2 R(n; K)[n K, f](R(n; K))2

(nR(n; K)) (R(n; K)[K, f] [K, f]R(n; K)) (nR(n; K))
(nR(n; K)) [R(n; K), [K, f]] (nR(n; K))
(nR(n; K)) {R(n; K) [K, [K, f]] R(n; K)} (nR(n; K)).
Therefore it is uniformly bounded in norm by Lemma 5.33.

We can now prove the main result of this section.
Theorem 5.35 The restriction of the quantum flow (5.60) to the algebra A coincides with
the flow of a classical diffusion with infinitesimal generator +

.
Proof. The algebra A is a core for +

by Theorem A.1. Moreover the structure maps
`
m (`, m {1, . . ., d, +, }) of the quantum flow (5.60) map A into A. Therefore,
by virtue of Theorem 5.30, the algebra A is invariant under the homomorphisms jt and the
restriction to A of the quantum flow (5.60) is commutative.
Proposition 5.34 implies that it satisfies the quantum stochastic differential equation
djt (f) =
d
X

jt ` (f) ) d`(t) + d
(t) + jt +
(f) )dt
`
`=1
(t))t0 as an operator version of classical

with j0(f) = f. Interpreting (` (t) +
`
Brownian motion via the Segal isometry it turns out that the above equation coincides
with the stochastic differential equation for the flow of a diffusion process with infinitesimal
generator +
.
5.7. CONCLUSION
5.7
115
Conclusion
The boson Fock quantum stochastic calculus developed by Hudson and Parthasarathy [58]
allows to construct several quantum flows by conjugation with unitary cocycles extending
the flow of a classical Markov process to the algebra of all bounded operators on a Hilbert
space which is often an L2 space of the state space. Since q.s.d.e. are driven by brownian
motions and Poisson processes which are mutually non-commuting, in this way we can
realise a classical Markov process as a non-commutative functional of brownian motions and
Poisson processes. Moreover the quantum flow enjoys the quantum Markov property in the
sense of Accardi [1] and satisfies a quantum stochastic differential equation.
A class of quantum flows which are not necessarily of the form
jt(x) = V (t)xV (t)
has been studied by Evans and Hudson in [36]. A detailed account on this subject for
classical Markov processes with bounded generator can be found in the recent books by
Meyer [69] and Parthasarathy [74].
The generalisation to unbounded generators is necessary to construct a non-commutative
extension of flows of most interesting classical Markov processes. Here we described the
general scheme (outlined in [41]) for constructing such extension. The main steps are the
following:
1. from the infinitesimal generator of a classical process find a possible choice of the
structure maps of the quantum stochastic differential equation satisfied by the flow
j and the coefficients of the quantum stochastic differential equation satisfied by the
cocycle V ,
2. show the existence of V by an approximation procedure,
3. show that V is unitary,

4. study a regularity property of a sequence j (n) n1 of approximating quantum flows
with bounded structure maps,
5. show that the homomorphisms {jt }t0 are also homomorphism of a smaller commutative algebra.
This scheme has been successfully followed here to construct quantum flows extending
the classical flow of multidimensional diffusion processes with regular covariance and drift.
The same can be done for discrete state quantum Markov chains (see [71]) and other classical
stochastic processes as for example Bessel processes (see [72] and also [17] with a different
approach for integer dimension index).
The analytical methods described here can be used also to construct quantum diffusions
on manifold studied by Applebaum in [11] and Sauvageot in [80] mainly from the algebraic
point of view.
The success of this scheme depends essentially on the form of infinitesimal generator.
Some techniques remind constructions of classical Markov processes done in [33]. The most
difficult problem arises when it turns out that the cocycle V is not isometric (see, for example,
[19], [38] and [44]). This always occurs when the classical process has some singular state,
for example a boundary point that can be reached in finite time, so that the infinitesimal
generator depends on boundary conditions. In this case the minimal quantum dynamical
semigroup arising from the representation in Lindblad form of the infinitesimal generator
might not be identity preserving. Therefore we expect that the cocycle V satisfies a very
singular quantum stochastic differential equation.
116
5. QUANTUM FLOWS
The attempts to study directly the quantum stochastic differential equation

X
`
djt (x) =
jt (m
(x))dm
` (t)
(5.65)
`,mS
have been successful only when the structure maps are bounded [34] or satisfy and analyticity
condition which is very hard to check in practice [48].
A result of Accardi and Mohari [9] states that, given a family of homomorphisms (jt )t0 ,
if the domain of the infinitesimal generator of the associated quantum Markov semigroup
is an algebra, then the quantum flow must satisfy an equation of the form (5.65). However
there are no known results (besides those in the classical commutative cases and [48] in the
noncommutative one) allowing to prove that given unbounded structure maps satisfying the
structure relations the maps (jt )t0 solving (5.65) are homomorphisms.
The direct study of the equation for (jt )t0 however seems important because, even
in the simplest cases, it might not be possible to find a suitable choice of the coefficients
of equation for V . In fact consider, for example, the classical Markov process on IR with
infinitesimal generator
(Af)(x) = f(0) f(x).
The heuristic discussions in Chapter 4 and 5 lead us to the choice
(01 (f))(x) = (10 (f))(x) = (11 (f))(x) = f(0) f(x)
and it is not possible to find a unitary operator on L2 (IR) such that (SfS )(x) = f(0).
Appendix A
Results on semigroups
We recall here some results on generation and regularity properties of a strongly continuous
semigroup by a second order, possibly degenerated, differential operator.
We study essentially two cases: differential operators acting on a space of continuous
functions and differential operators acting on a L2 space.
These results are perhaps not new but it was impossible to find a full proof of all the
properties we need of such semigroups in a few basic and accessible references. Proofs
essentially follow adapting arguments from [83]. The details have been spelled out in [43].
We denote by Cbn (IRd ; IR) be the Banach space of bounded continuous functions with
bounded continuous partial derivatives up to the order n endowed with the norm

1/2

X

2
kfkC n =
| f|
b

||n
where = (1, . . . , d) IN d is a multindex and || = 1 + . . . + d , f = 11 . . . dd f.

The closed subspace of continuous functions vanishing at infinity with continuous partial
derivatives up to the order n vanishing at infinity will be denoted by C0n(IRd ; IR).
The closed subspace of continuous functions having a limit as |x| tends to infinity with
continuous partial derivatives up to the order n having a limit as |x| tends to infinity will
be denoted by C`n(IRd ; IR).
Let ( > 0) be the functions
: IRd IR,
(x) = 1 + |x|2
/2
n
We denote by C
(IRd ; IR) be the Banach space of bounded continuous functions with
bounded continuous partial derivatives up to the order n such that f vanishes as |x|
goes to infinity endowed with the norm

1/2

X

2
kfkC n =
| f| .

||n
The linear manifold of infinitely differentiable functions with compact support will be
denoted Cc (IRd ; IR).
A similar notation will be used for spaces of continuous complex-valued functions.
The following are the main results proved in the appendix
117
118
APPENDIX A. RESULTS ON SEMIGROUPS
Theorem A.1 Let n 2 be a fixed integer. For x IRd , let (ajk (x))1j,kd be a d d
symmetric non-negative matrix, (bj )1jd a d-dimensional vector and c(x) a real number.
Suppose that:
1. the functions ajk belong to C n (IRd ; IR) and their second order partial derivatives belong
to Cbn2(IRd ; IR) for every j, k {1, . . . , d},
2. the functions bj are differentiable and their first order partial derivatives belong to
Cbn1(IRd ; IR) for every j {1, . . . , d},
3. the function c belongs to Cbn (IRd ; IR).
0
Then, for every 0, the closure in C
(IRd ; IR) of the operator
Af =
d
d
X
1 X
ajkj k f +
bj j f + cf
2
j=1
(A.1)
j,k=1
defined on the domain Cc (IRd ; IR) is the infinitesimal generator of a strongly continuous
0
semigroup (T (t))t0 on C
(IRd ; IR). Moreover
0
1. for every t 0 and every non-negative f C
(IRd ; IR) the function T (t)f is also
non-negative,
k
2. for every t 0 and every k with 2 k n the linear manifold C
(IRd ; IR) is
invariant under T (t) and the resolvent operators R(; A) ( > 0),
3. if = 0 the operators (T (t))t0 are contractive if and only if c 0.

Corollary A.2 Under the assumptions of Theorem A.1 suppose c = 0. Let D be the linear
manifold generated by Cc(IRd ; IR) and constant functions. Then the closure in C`0(IRd ; IR)
of the operator A defined on the domain D by (A.1) is the infinitesimal generator of a
strongly continuous contraction semigroup (T (t))t0 on C`0(IRd ; IR). Moreover the semigroup
(T (t))t0 is Markov (i.e. T (t)1 = 1).
Theorem A.3 Suppose that the hypotheses of Theorem A.1 hold. Then the closure in
L2(IRd ;C)
I of the operator
Gf =
d
d
X
1 X
ajk j k u +
bj j u + cu
2
j=1
(A.2)
j,k=1
defined on the domain Cc (IRd ;C)

I is the infinitesimal generator of a strongly continuous
semigroup (P (t))t0 on L2 (IRd ;C).
I Moreover the semigroup (P (t))t0 is contraction if and
only if G is dissipative i.e.
2<ehu, Gui 0
for every u Cc (IRd ;C).
I
Remark. Notice that, under the assumptions of Theorem A.1, P (t)u = T (t)u for every
u L2 (IRd ;C)
I C00(IRd ; IR) and t 0. In fact P (t) and T (t) are bounded operators
coinciding on the dense (in L2(IRd ;C)
I and C00(IRd ; IR)) subset Cc(IRd ; IR).
We refer to [43] for the proofs of the above results.
Bibliography
[1] L. Accardi: On the quantum Feynman-Kac formula. Rend. Sem. Mat. Fis. Milano,
Vol. XLVIII, 135179 (1978).
[2] L. Accardi: A note on Meyers note. In: Accardi, L., von Waldenfels, W. (eds.) Quantum Probability and Applications III. Proceedings, Oberwolfach 1987. (Lect. Notes
Math., vol. 1303, pp. 15). Berlin, Heidelberg, New York: Springer 1988.
[3] L. Accardi, C. Cecchini: Conditional expectations in von Neumann algebras and a
theorem of Takesaki, J. Funct. Anal. 45, (1982), 245273.
[4] L. Accardi, F. Fagnola, J. Quaegebeur: A representation free quantum stochastic
calculus. J. Funct. Anal., 104 (1992), 149197.
[5] L. Accardi, A. Frigerio, J.T. Lewis: Quantum stochastic processes. Publ. R.I.M.S.
Kyoto Univ., 18 (1982), 97133.
[6] L. Accardi, R. Alicki, A. Frigerio, Y.G. Lu: An invitation to the weak coupling and
low density limits. Quantum Probability and Related Topics VI, (1991), 362.
[7] L. Accardi, J.L. Journe, J.M. Lindsay: On multidimensional Markovian cocycles. In
: Accardi, L., von Waldenfels, W., (eds.) Quantum Probability and Applications IV.
Proceedings Rome 1987 (Lect. Notes Math. Vol. 1396, pp. 59-67) Berlin Heidelberg
New York, Springer, 1989.
[8] L. Accardi, Y.G. Lu, I.V. Volovich, Quantum Theory and its Stochastic Limit. Springer
2000 (to appear).
[9] L. Accardi, A. Mohari: On the Structure of Classical and Quantum Flows. J. Funct.
Anal. 135 (1996), 421455.
[10] R. Alicki, K. Lendi, Quantum Dynamical Semigroups and Applications, Lect. Notes
Phys. 286, Springer-Verlag, 1987.
[11] D. Applebaum: Towards a Quantum Theory of Classical Diffusions on Riemannian
Manifolds. Quantum Probability and Related Topics VI, 93109, (1991).
[12] W.B. Arveson: Subalgebras of C -algebras, Acta Math., 123 (1969), 141224.
[13] S. Attal: Extensions of Quantum Stochastic Calculus. Quantum Probability Summer
School. Grenoble 1998.
[14] S. Attal, P.-A. Meyer: Interpretation probabiliste et extension des integrales stochastiques non commutatives. Seminaire de Probabilites, XXVII, 312327, Lecture Notes
in Math., 1557, Springer, Berlin, 1993.
119
120
BIBLIOGRAPHY
[15] A. Barchielli: Applications of quantum stochastic calculus to quantum optics. Quantum Probability and Related Topics VI, (1991), 111-126.
[16] V.P. Belavkin: A new form and a ?-algebraic structure of quantum stochastic integrals
in Fock space. Rend. Sem. Mat. Fis. Milano, Vol. LVIII (1988), 177193.
[17] Ph. Biane: Quelques proprietes du mouvement brownien non-commutatif. Hommage
a P. A. Meyer et J. Neveu. Asterisque No. 236 (1996), 73101.
`
[18] B.V.R. Bhat, F. Fagnola, K.B. Sinha: On quantum extensions of semigroups of brownian motions on an half-line. Russian J. Math. Phys. 4 (1996), 1328.
[19] B.V.R. Bhat, K.B. Sinha: Examples of unbounded generators leading to nonconservative minimal semigroups. Quantum Probability and Related Topics, IX (1994), 89104.
[20] B.V.R. Bhat, K.R. Parthasarathy: Markov dilations of nonconservative dynamical
semigroups and a quantum boundary theory. Ann. Inst. H. Poincare Probab. Statist.
31 (1995), no. 4, 601651.
[21] O. Bratteli, D.W. Robinson, Operator Algebras and Quantum Statistical Mechanics I.
Springer-Verlag, 1979.
[22] A.M. Chebotarev: The theory of conservative dynamical semigroups and applications.
MIEM Preprint n.1, Moscow, March 1990.
[23] A.M. Chebotarev: Necessary and sufficient conditions of the conservativism of dynamical semigroups, in: Contemporary Problems , of Mathematics. Newest Achievements
36, VINITI, Moscow (1990), 149184.
[24] A.M. Chebotarev: Necessary and sufficient conditions for conservativeness of dynamical semigroups, J. Sov. Math., 56 (1991), 26972719.
[25] A.M. Chebotarev: Sufficient conditions of the conservativism of a minimal dynamical
semigroup. Math. Notes 52 (1993), 10671077.
[26] A.M. Chebotarev, F. Fagnola: Sufficient Conditions for Conservativity of Quantum
Dynamical Semigroups. J. Funct. Anal. 118 (1993), 131153.
[27] A.M. Chebotarev, F. Fagnola: On quantum extensions of the Azema martingale semigroup. Sem. Prob. XXIX (1995), 116, Springer LNM 1613.
[28] A.M. Chebotarev, F. Fagnola: Sufficient Conditions for Conservativity of Minimal
Quantum Dynamical Semigroups. J. Funct. Anal. 153 (1998), 382404.
[29] M. Choi: Positive linear maps on C -algebras, Can. J. Math., XXIV (1972), 520529.
[30] E. Christensen, D.E. Evans: Cohomology of operator algebras and quantum dynamical
semigroups. J. London Math. Soc. 20 (1979), 358368.
[31] K.L. Chung: Markov Chains with Stationary Transition Probability Springer-Verlag,
1960.
[32] E.B. Davies: Quantum dynamical semigroups and the neutron diffusion equation. Rep.
Math. Phys. 11 (1977), 169188.
[33] S.N. Ethier, T.G. Kurtz: Markov Processes. Characterization and convergence. John
Wiley & Sons (1986).
BIBLIOGRAPHY
121
[34] M. Evans: Existence of Quantum Diffusions. Probab. Th. Rel. Fields 81 (1989), 473
483.
[35] D.E. Evans, H. Hanche-Olsen: The generators of positive semi-groups. J. Funct. Anal.
32 (1979), 207212.
[36] M. Evans, R.L. Hudson: Multidimensional quantum diffusions. In: Accardi, L., von
Waldenfels, W. (eds.) Quantum Probability and Applications III. Proceedings, Oberwolfach 1987. (Lect. Notes Math., vol. 1303, pp. 6988). Berlin, Heidelberg, New York:
Springer 1988.
[37] F. Fagnola: On quantum stochastic differential equations with unbounded coefficients.
Probab. Th. Rel. Fields, 86, 501516 (1990).
[38] F. Fagnola: Pure birth and pure death processes as quantum flows in Fock space.
Sankhya 53 (1991), 288297.
[39] F. Fagnola: Unitarity of solutions of quantum stochastic differential equations and
conservativity of the associated semigroups. Quantum Probability and Related Topics,
VII (1992), 139148.
[40] F. Fagnola: Characterisation of isometric and unitary weakly differentiable cocycles
in Fock space. Quantum Probability and Related Topics VIII (1993), 143164.
[41] F. Fagnola: Diffusion processes in Fock space. Quantum Probability and Related
Topics IX (1994), 189214.
[42] F. Fagnola: A simple singular quantum Markov semigroup. To appear in: Anestoc 98
- Proceedings.
[43] F. Fagnola: Quantum Markov Semigroups and Quantum Markov Flows. Tesi di perfezionamento. Scuola Normale Superiore di Pisa, Pisa 1998.
[44] F. Fagnola, R. Monte: A quantum extension of the semigroup Bessel processes. Mat.
Zametki 60 n.5 (1996), p.519537.
[45] F. Fagnola, R. Rebolledo: The approach to equilibrium of a class of quantum dynamical semigroups. Infinite Dimensional Analysis and Quantum Probability 1, n.4 (1998),
561572.
[46] F. Fagnola, R. Rebolledo, C. Saavedra: Quantum flows associated to master equations
in quantum optics. J. Math. Phys. 35 (1994), 112.
[47] F. Fagnola, R. Rebolledo, C. Saavedra: Reduction of Noise by Squeezed Vacuum. In:
R. Rebolledo (ed.), Stochastic Analysis and Mathematical Physics. ANESTOC 96.
World Scientific 1998. 6171.
[48] F. Fagnola, Kalyan B. Sinha: Quantum flows with unbounded structure maps and
finite degrees of freedom. J. London Math. Soc. (2) 48, (1993) p. 537551, .
[49] W. Feller: On the integro-differential equations for purely discontinuous Markov processes. Trans. Am. Math. Soc. 48, 488575 (1940); Errata 58, 474 (1945).
[50] A. Frigerio: Positive contraction semigroups on B (H) and quantum stochastic differential equations. In: Trends in semigroup theory and applications (Ph. Clement,
S. Invernizzi, E. Mitidieri, I.I. Vrabie eds.) Proceedings, Trieste 1987. Marcel Dekker
(1989), 175-188.
122
BIBLIOGRAPHY
[51] W. Feller: An Introduction to Probability Theory and Its Applications. Vol. I. John
Wiley & Sons, Inc., New York, N.Y., 1950.
[52] A. Frigerio: Some applications of quantum probability to stochastic differential equations in Hilbert space. In: Stochastic partial differential equations and applications
(G. Da Prato and L. Tubaro eds.) Proceedings, Trento 1988. Springer LNM 1390
(1989), 7790.
[53] J.C. Garca, R. Quezada: A priori estimates for a class of Quantum Dynamical Semigroups and applications. Cinvestav, Reporte interno n. 235. June 1998.
[54] C.W. Gardiner, P. Zoller: Quantum Noise in Quantum Optics: the Stochastic
Schroedinger Equation. http://xxx.sissa.it/list/quant-ph/9702030.
[55] V. Gorini, A. Kossakowski, E.C.G. Sudarshan: Completely positive dynamical semigroups of N -level systems. J. Math. Phys. 17 (1976), 821825.
[56] A.S. Holevo: On the structure of covariant dynamical semigroups. J. Funct. Anal. 131
(1995), 255278.
[57] R.L. Hudson, J.M. Lindsay: On characterizing quantum stochastic evolutions. Math.
Proc. Cambridge Philos. Soc. 102 (1987), no. 2, 363369
[58] R.L. Hudson, K.R. Parthasarathy: Quantum Itos formula and stochastic evolutions,
Commun. Math. Phys. 93 (1984), 301323.
[59] K. It
o, H.P. McKean, Jr: Diffusion Processes and their Sample Paths, Springer 1965.
[60] P.E.T. Jorgensen: Approximately Reducing Subspaces for Unbounded Linear Operators. J. Funct. Anal. 23 (1976), 392141.
[61] J.-L. Journe: Structure des cocycles markoviens sur lespace de Fock. Probab. Th. Rel.
Fields 75 (1987), 291316.
[62] T. Kato: On the semi-group generated by Kolmogoroffs differential equations, J.
Math. Soc. Japan 6 (1954), 115.
[63] T. Kato: Perturbation theory for linear operators. Springer-Verlag, 1966.
[64] K. Kraus: General States Changes in Quantum Theory, Ann. Phys., 64 (1970), 311
335.
[65] G. Lindablad: On the genarators of Quantum Dynamical Semigroups. Commun. Math.
Phys. 48 (1976), 119130.
[66] J.M. Lindsay: Quantum and noncausal stochastic calculus. Probab. Theory Rel. Fields
97 (1993), no. 1-2, 6580.
[67] H. Maassen: Quantum markov processes on Fock space described by integral kernels.
In: Accardi, L., von Waldenfels, W. (eds.) Quantum Probability and Applications II.
(Lect. Notes Math., vol. 1136, pp. 361374) Berlin, Heidelberg, New York: Springer
1985.
[68] P.A. Meyer: A note on shifts and cocycles. In: Accardi, L., von Waldenfels, W. (eds.)
Quantum Probability and Applications III. Proceedings, Oberwolfach 1987. (Lect.
Notes Math., vol. 1303, pp. 209212). Berlin, Heidelberg, New York: Springer 1988.
BIBLIOGRAPHY
123
[69] P.A. Meyer, Quantum Probability for Probabilists, Lect. Notes Math. 1538, SpringerVerlag, 1994.
[70] A. Mohari, K.R. Parthasarathy: On a class of generalizes Evans-Hudson flows related
to classical markov processes. Quantum Probability and Related Topics, VII (1992),
221249.
[71] A. Mohari, K.B. Sinha: Stochastic dilation of minimal quantum dynamical semigroup.
Proc. Indian Acad. Sci. 102 (1992), 159173.
[72] R. Monte: Sullestensione quantistica dei processi di Markov. Universit`
a di Palermo.
Tesi di dottorato. February 1997.
[73] M.Ohya, D.Petz: Quantum Entropy and its Use, Springer 1995.
[74] K.R. Parthasarathy, An Introduction to Quantum Stochastic Calculus, Monographs in
Mathematics, Vol. 85, 1992.
[75] K.R. Parthasarathy, K.B. Sinha: Markov chains as Evans-Hudson diffusion in Fock
space. Sem. Prob. XXIV (1990), 362369, Springer LNM 1426.
[76] A. Pazy, Semigroups of Linear Operators and Applications to Partial Differential Equations, Springer-Verlag, 1975.
[77] D. Petz: Conditional Expectation in Quantum Probability. In: Accardi, L., von
Waldenfels, W. (eds.) Quantum Probability and Applications III. Proceedings, Oberwolfach 1987. (Lect. Notes Math., vol. 1303, pp. 251260). Berlin, Heidelberg, New
York: Springer 1988.
[78] M. Reed, B. Simon: Methods of Modern Mathematical Physics, Vol. I, Functional
Analysis, Academic Press, 1975.
[79] M. Reed, B. Simon: Methods of Modern Mathematical Physics, Vol. II, Fourier Analysis, Self-Adjointness, Academic Press, 1975.
[80] J.L. Sauvageot: Towards a Quantum Theory of Classical Diffusions on Riemanian
Manifolds. Quantum Probability and Related Topics VII, 299316, (1992).
[81] K.B. Sinha: Quantum Dynamical Semigroups. In: Operator Theory: Advances and
Applications, Vol. 70, 161-169, Birkhauser Verlag Basel, 1994.
[82] W.F. Stinespring: Positive functions on C -algebras, Proc. Am. Math. Soc., 6 (1955),
211216.
[83] D.W. Stroock, S.R.S. Varadhan: Multidimensional Diffusion Processes. Springer,
1979.
[84] S. Wills: Stochastic Calculus for Infinite Dymensional Noises. Ph. D. Thesis. Nottingham 1997.

F. Fagnola, Quantum Markov Semigroups and Quantum Flows

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

F. Fagnola, Quantum Markov Semigroups and Quantum Flows

Uploaded by

Copyright:

Available Formats

QUANTUM MARKOV SEMIGROUPS AND

2 Algebraic Markov Processes

3 Quantum Markov Semigroups

4 Classical and Quantum Semigroups

Algebraic Probability Spaces

Definition 1.1 A -algebra is a complex algebra A with an involution, denoted by , with

1. ALGEBRAIC PROBABILITY SPACES

Classical probability spaces as algebraic

A measurable space (, F) clearly determines uniquely the commutative -algebra A =

Quantum probability spaces

1.3. QUANTUM PROBABILITY SPACES

Therefore X defines a quantum random variable.

where 1B denotes the indicator function of a B F. Thus we have constructed a classical

U f1 (X)f2 (X) . . . fn (X)e = f1 f2 . . . fn .

The operator U is unitary because of the relation

for g integrable with respect to IP which follows immediately from (1.1).

1. ALGEBRAIC PROBABILITY SPACES

be summarized by saying that (A, ) is a quantum probability space containing infinitely

1.3. QUANTUM PROBABILITY SPACES

4. momentum (or electric field) operator

D(p) = D(a) = D(a ),

D(q) = D(a) = D(a ),

([, ] denoting the commutator) yield

allows us to move from one representation to the other.

L2 (IR; 1/4 exp(x2 /2)dx) L2 (IR; 1/4 exp(x2 /2)dx)

1. ALGEBRAIC PROBABILITY SPACES

(exp(itp))v(x) = v(x + t).

Therefore the characteristic functions of p and q in the state are given by

hH0, exp(itq)H0 i = exp(t2 /4),

Examples given in the previous section motivate the following

1.5. TOPOLOGIES ON ALGEBRAS

4. for all a A we have IE [ a | A0 ] = (IE [ a | A0 ]) ,

1. ALGEBRAIC PROBABILITY SPACES

1.5. TOPOLOGIES ON ALGEBRAS

1. T is -weakly continuous (i.e. continuous with respect to the -weak topologies on A

for each u in a linear submanifold of h which is norm-dense in h,

for each v, u in a total subset of h.

|hu u , (T x)ui hu u , (T x)ui|

|hu, (T x)(u u )i hu, (T x)(u u)i|

|hu, (T x)u i hu , (T x)ui| .

it follows then that (1.2) holds for each v, u h.

To this end, for every > 0, take an integer such that

1. ALGEBRAIC PROBABILITY SPACES

The first term, since kT xk kT xk, can be estimated by

e (y) = (T y)/(T 1l),

This completes the proof.

Algebraic Markov Processes

2. ALGEBRAIC MARKOV PROCESSES

and IP is the probability measure on F defined by

where 0 t1 < t2 < . . . < tn . Consider the filtration Ft]

Consider a quantum probability space (A, ) with a filtration At]

The classical process x defines a family of *-homomorphisms

Therefore the classical Markov property

f(y)P (t, t + s, xt, dy)

f(y)P (0, s, x, dy)

2.1. FUNDAMENTAL DEFINITIONS

Proposition 2.3 The following conditions are equivalent:

This completes the proof.

2. ALGEBRAIC MARKOV PROCESSES

which is clearly a positive operator because of property 1 of conditional expectation. This

Completely positive linear maps

2. completely positive if it is n positive for every integer n 1.

2.2. COMPLETELY POSITIVE LINEAR MAPS

In order to give a useful condition equivalent to complete positivity by means of the