The stochastic limit

of quantum spin systems

L. Accardi, S.V. Kozyrev
November 18, 1999

Centro Vito Volterra Universita di Roma Tor Vergata

Abstract
The stochastic limit for the system of spins interacting with a boson field is investigated.
In the finite volume an application of the stochastic golden rule shows that in the limit the
dynamics of a quantum system is described by a quantum white noise equation that after
taking of normal order is equivalent to quantum stochastic differential equation (QSDE). For
the quantum Langevin equation the dynamics is well defined and is a quantum flow on the
infinite lattice system.

1 Introduction
Starting with the work of Glauber [10] the dynamics of infinite classical lattice systems has been
considered by many authors and has led to study the ergodic and equlibrium properties of a new
class of classical Markov semigroups (cf. [11] for a general survey and for further references).
Quantum analogues of these semigroups have also been considered by several authors (e.g. [13],
[14], [12], [15], [16], . . . ). However the problem of deriving these Markovian semigroups, and more
generally the stochastic flows, as limits of Hamiltonian systems, was open both in the classical
and in the quantum case.
On the other hand the stochastic golden rules, which arise in the stochastic limit of quantum
theory as natural generalizations of Fermi golden rule [2], provide a natural tool to associate
a stochastic flow, driven by a white noise equation, to any discrete system interacting with a
quantum field. Moreover another quite general result of the stochastic limit is that the Markov
semigroup, canonically associated to the flow, leaves invariant the abelian subalgebra generated by
the spectral projections of the Hamiltonian of the discrete system so that the associated Markov
process describes the jumps among these energy levels.
Since a quantum spin system in finite volume is obviously a discrete system, the above results
naturally suggest the conjecture that, by coupling such a system to a quantum field via a suitable
interaction, applying the stochastic golden rule and taking the thermodynamic limit, one might
obtain a class of quantum flows which, when restricted to an appropriate abelian subalgebra, gives
rise to the classical interacting particle systems studied in classical statistical mechanics.
In the present paper we will prove this conjecture for general finite range interactions thus
extending a previous result obtained by the authors in the case of the Ising model [6]. The model

1
we consider is essentially the same as the one considered by Martin and Buffet [13] apart from
the minor difference that these authors consider fermion reservoir and we a boson one. Since the
stochastic golden rule holds also in the fermion case [1], [2] there is no difficulty in extending the
present results to the fermion case. We investigate a quantum system with a Hamiltonian of the
form
H = H0 + λHI
where H0 is called the free Hamiltonian and HI the interaction Hamiltonian. The general idea
of the stochastic limit (see [3]) is to make the time rescaling t → t/λ2 in the solution of the
(λ)
Schrödinger (or Heisenberg) equation in interaction picture Ut = eitH0 e−itH , associated to the
Hamiltonian H, i.e.
∂ (λ) (λ)
U = −iλHI (t) Ut
∂t t
with HI (t) = eitH0 HI e−itH0 . This gives the rescaled equation
∂ (λ) i (λ)
Ut/λ2 = − HI (t/λ2 ) Ut/λ2 (1)
∂t λ
and one wants to study the limits, in a topology to be specified,
 
(λ) 1 t
lim U 2 = Ut ; lim HI = Ht (2)
λ→0 t/λ λ→0 λ λ2
After the rescaling t → t/λ2 we consider the simultaneous limit λ → 0, t → ∞ under the
condition that λ2 t tends to a constant (interpreted as a new slow time scale). This limit captures
the dominating contributions to the dynamics, in a regime of long times and small coupling, arising
from the cumulative effects, on a large time scale, of small interactions (λ → 0). The physical idea
is that, looked from the slow time scale of the atom, the field looks like a very chaotic object: a
quantum white noise, i.e. a δ-correlated (in time) quantum field b∗ (t, k), b(t, k) also called a master
field.

2 Physical model
In the present paper we consider the open Ising model, describing the interaction of a system S of
spins (or, more generally, two-level systems) with a reservoir, represented by a bosonic quantum
field. The total Hamiltonian is
H = H0 + λHI = HS + HR + λHI
where HR is the free Hamiltonian of a bosonic reservoir R:
Z
HR = ω(k)a∗ (k)a(k) dk

acting in the representation space F corresponding to the bosonic equilibrium state at temperature
β −1 and chemical potential µ and the assumptions we need on ω(k) are given at the beginning of
Section III below. Thus the reservoir state is Gaussian with mean zero and correlations given by
1
ha∗ (k)a(k 0 )i = δ(k − k 0 )
eβ(ω(k)−µ) −1

2
The spin variables are labeled by the lattice Z d , and, for each finite subset Λ ⊆ Z d , the system
Hilbert space is
HS = HΛ = ⊗r∈Λ C 2
and the system Hamiltonian has the form
1 X
HS = HΛ = − Jrs σrz σsz
2 r,s∈Λ

where σrx , σry , σrz are Pauli matrices (r ∈ Λ) at the r-th site in the tensor product. For any r,
s∈Λ
Jrs = Jsr ∈ R, Jrr = 0
In the present paper we consider a system Hamiltonian that describes the interaction of spin
with a finite number of other spins (finite range potential). The simplest example is the nearest
neighbor interaction (Ising model), discussed in Section V below. The interaction Hamiltonian
HI (acting in HS ⊗ F ) has the form
X Z
HI = σrx ⊗ ψ(gr ), ψ(g) = A(g) + A∗ (g), A(g) = dk g(k)a(k),
r∈Λ

where ψ is a field operator, A(g) is a smeared quantum field with cutoff function (form factor)
g(k). To perform the construction of the stochastic limit one needs to calculate the free evolution
of the interaction Hamiltonian: HI (t) = eitH0 HI e−itH0 . We use the formula
X
σrz = εr |εr ihεr |
εr =±1

where |εr i is the eigenvector of σrz corresponding to the egenvalue εr to rewrite the interaction
Hamiltonian in the form
X X
HI = hεr , σrx µr i|εr ihµr | ⊗ (A(g) + A∗ (g)) = (DΛ + DΛ∗ ) (A(g) + A∗ (g))
r∈Λ εr ,µr =±1

where X
Dr = |1r ih−1r | , Dr∗ = | − 1r ih1r | , DΛ = Dr
r∈Λ

Since X
[HS , |1r ih−1r |] = ∆(r)|1r ih−1r | , ∆(r) = − Jrs σsz
s

the free evolution for Dr has the form

eitHS Dr e−itHS = eit∆(r) Dr

The sum over s here is finite because the spin at r interacts only with a finite number of spins
(only a finite number of Jrs for fixed r is nonzero).
Using the formula for the free evolution of bosonic fields

eitHR a(k)e−itHR = e−itω(k) a(k)

3
we get for the free evolution of the interaction Hamiltonian:
XZ  
HI (t) = dk g(k) Dr e−it(ω(k)−∆(r)) a(k) + Dr∗ e−it(ω(k)+∆(r)) a(k) + h.c. (3)
r∈Λ

and, for the free evolution of Dr :

P Y
−it Jrs σsz z
eit∆(r) Dr = e s∈Σ(r) Dr = Dr e−itJrs σs =
s∈Σ(r)
Y X
= Dr e−itJrs εs |εs ihεs | (4)
s∈Σ(r) εs =±1

where Σ(r) denotes the (finite) set of spins interacting with the spin r (so r 6= Σ(r)). When
no confusion is possible we use the same symbol Σ(r) to denote also the set of configurations of
correponding spins: σ(r) = {εs }(s ∈ Σ(r)).
We denote
X
E(r, σ(r)) = Jrs εs = HS (σrz = −1, σ(r)) − HS (σrz = −1, σ(r)) (5)
s∈Σ(r)

the energy difference between two configurations with all spins fixed, and given by σ(r), with the
exception of εr that changes value from −1 to 1.
With these notations formula (3) takes the form
XZ  X Y
HI (t) = dk g(k) Dr |εs ihεs |e−it(ω(k)+E(r,σ(r))) a(k)+
r∈Λ σ∈Σ(r) s∈Σ(r)

X Y 
+ Dr∗ |εs ihεs |e −it(ω(k)−E(r,σ(r)))
a(k) + h.c. =
σ∈Σ(r) s∈Σ(r)
X X Z 
= dk g(k) Gr,σ(r) e−it(ω(k)+E(r,σ(r))) a(k)+
r∈Λ σ(r)∈Σ(r)

+G∗r,σ(r) e−it(ω(k)−E(r,σ(r))) a(k) + h.c. =
X X Z
∗
= dk g(k)Fr,σ(r) e−it(ω(k)−E(r,σ(r))) a(k) + h.c. (6)
r∈Λ σ(r)∈Σ(r)

where we denote
Y
Gr,σ(r) = Dr |εs ihεs |, Fr,σ(r) = G∗r,−σ(r) + Gr,σ(r) (7)
s∈Σ(r)

To prove (6) we used a suitable rearranging of the indices σ(r) and the property

E(r, −σ(r)) = −E(r, σ(r))

where −σ(r) is the configuration of spins with all spins opposite to the spins in σ(r), i.e. εs (−σ(r)) =
−εs (σ(r)).
The operator Fr,σ(r) flips the spin at site r (or kills a spin configuration).

4
3 The stochastic limit of the model
In this section we will denote R ∈ S the pair of indices (r, σ(r)) where r takes values in Λ. In
these notations the free evolution of the interaction Hamiltonian (6) takes the form
XZ
HI (t) = dk g(k)FR∗ e−it(ω(k)−E(R)) a(k) + h.c. (8)
R∈S

In the stochastic limit the field HI (t) gives rise to a family of quantum white noises, or master
fields. To investigate these noises, let us suppose the following:
1) ω(k) ≥ 0, ∀k;
2) The d − 1–dimensional Lebesgue measure of the surface {k : ω(k) = 0} is equal to zero (so
that δ(ω(k)) = 0) (for example ω(k) = k 2 + m with m ≥ 0).
Now let us investigate the limit of HI (t/λ2 ) using one of the the basic formulae of the stochastic
limit:  
1 it
lim 2 exp 2 f (k) = 2πδ(t)δ(f (k)) (9)
λ→0 λ λ
Since the term δ(f (k)) in (9) is not identically equal to zero only if f (k) = 0 for some k in a set of
nonzero d − 1–dimensional Lebesgue measure (in our case f (k) = ω(k) − E(R)), in the stochastic
limit of the rescaled interaction Hamiltonian (8) only the terms with R ∈ S+ will survive. Here S+
P
(respectively S− ) denotes the spin configurations (εs ) whose mean energy difference s∈Σ(r) Jrs εs is
strictly positive (respectively negative). We will call such spin configurations positive (respectively
negative).
The rescaled interaction in (8) is expressed in terms of the rescaled creation and annihilation
t
operators aλ,R (t, k) = λ1 e−i λ2 (ω(k)−E(R)) a(k). After the stochastic limit every rescaled annihilation
operator correspondent to positive spin configuration generates one non-trivial quantum white
noise
1 t
bR (t, k) = lim aλ,R (t, k) = lim e−i λ2 (ω(k)−E(R)) a(k)
λ→0 λ→0 λ

with the relations

[bR (t, k), b∗R (t0 , k 0 )] = lim [aλ,R (t, k), a∗λ,R (t0 , k 0 )] =
λ→0

1 −i t−t2 0 (ω(k)−E(R))
= lim e λ δ(k − k 0 ) = 2πδ(t − t0 )δ(ω(k) − E(R))δ(k − k 0 ) (10)
λ→0 λ2

[bR (t, k), b∗R0 (t0 , k 0 )] = 0

by (9). Moreover, configurations, corresponding to different values of the energy difference E(R),
are independent. For generic interactions if R 6= R0 then corresponding energy differences E(R) 6=
E(R0 ) and correponding white noises are independent.
The stochastic limit of the interaction Hamiltonian is therefore equal to
  Z X
1 t
h(t) = lim HI = dk g(k) FR∗ bR (t, k) + h.c. (11)
λ→0 λ λ2 R∈S+

Physically such a form of the evolved interaction Hamiltonian can be explained as follows: the
operator FR , where R ∈ S+ , decreases the energy of the spin configuration. Therefore the vertex

5
FR∗ a(k) describes the absorption, by the spin configuration R, of an energy quantum of the field
of momentum k.
The state on the master field (white noise) bR (t, k), corresponding to the equilibrium state of
the field, is the mean zero gaussian state with correlations:
1
hb∗R (t, k)bR (t0 , k 0 )i = 2πδ(t − t0 )δ(ω(k) − E(R))δ(k − k 0 )
eβ(ω(k)−µ) −1
1
hbR (t, k)b∗R (t0 , k 0 )i = 2πδ(t − t0 )δ(ω(k) − E(R))δ(k − k 0 )
1− e−β(ω(k)−µ)
and vanishes for noises corresponding to different positive spin configurations.
Now let us investigate the evolution equation in interaction picture for our model. According to
the general scheme of the stochastic limit, up to possible divergences (due to the thermodynamic
limit) that we will discuss later, we get the (singular) white noise equation

d
Ut = −ih(t)Ut (12)
dt
whose normally ordered form is the quantum stochastic differential equation [4]

dUt = (−idH(t) − Gdt) Ut (13)

where h(t) is the white noise (11) given by the stochastic limit of the interaction Hamiltonian and
X  
dH(t) = FR∗ dBR (t) + FR dBr,σ(r)
∗
(t) (14)
R∈S+

Z Z t+dt
dBR (t) = dk g(k) bR (τ, k)dτ (15)
t

According to the stochastic golden rule, (13) is obtained as follows: the first term in (13) is
just the limit of the iterated series solution for (1)
Z  
1 t+dt τ
lim HI dτ
λ→0 λ t λ2
The second term Gdt, called the drift, is equal to the limit of the expectation value in the reservoir
state of the second term in the iterated series solution for (1)
Z Z    
1 t+dt t1 t1 t2
lim dt1 dt2 hHI HI i
λ→0 λ2 t t λ 2 λ2
Making in this formula the change of variables τ = t2 − t1 we get
Z Z 0    
1 t+dt t1 t1 τ
lim dt1 dτ hHI HI + 2 i (16)
λ→0 λ2 t t−t1 λ 2 λ 2 λ
Computing the (Gaussian) expectation value and using formula (8) and the fact that the limits
ict1
of oscillating factors of the form limλ→0 e λ2 vanish unless the constant c is equal to zero, we see

6
 ict1
that we can have non–zero limit only when all oscillating factors of a kind e λ2 (with t1 ) in (16)
cancel. In conclusion we get
XZ Z 0 
2 1
G= dk |g(k)| dτ FR∗ FR eiτ (ω(k)−E(R)) +
R∈S −∞ 1− e−β(ω(k)−µ)

1
+FR FR∗ e−iτ (ω(k)−E(R))
eβ(ω(k)−µ) − 1
and therefore, from the formula
Z 0 −i 1
eitω = = πδ(ω) − iP.P. (17)
−∞ ω − i0 ω
we get the following expression for the drift G:
XZ 
2 −iFR∗ FR 1
dk |g(k)| +
R∈S ω(k) − E(R) − i0 1 − e−β(ω(k)−µ)


iFR FR∗ 1
+ =
ω(k) − E(R) + i0 eβ(ω(k)−µ) −1
X +
= FR∗ FR (g|g)− ∗
R + FR FR (g|g)R (18)
R∈S

We see from formula (18) that our model also exhibits the Cheshire cat effect discovered in [5] in
a simpler model, that is: the part of the terms in the drift (18) with R ∈ S+ comes from quantum
white noises (or, correspondingly, from stochastic differentials) and describes the self–interaction
of such noises, but products of pairs of operators FRε with the index R corresponding to negative
spin configuration describes self–interaction of virtual noises (corresponding to vertices without
conservation of energy).
The drift term contains sums over R which are divergent for large Λ. Therefore in the
Schrodinger picture we will get a divergence in the thermodynamic limit. In the following section
we will consider the evolution in the Heisenberg picture (the Langevin equation) and we will show
that, in this context, no divergence arises.

4 The Langevin equation

Now we will find the Langevin equation, which is the limit of the Heisenberg evolution, in in-
teraction representation, of a system observable. Let X be a local operator acting on the spin
degrees of freedom, i.e. one acting only on a finite number of spins. The Langevin equation is the
equation satisfied by the stochastic flow jt , defined by:

jt (X) = Ut∗ XUt

where Ut satisfies equation (13) in the previous section, i.e.

dUt = (−idH(t) − Gdt) Ut (19)

7
 X +
G= FR∗ FR (g|g)− ∗
R + FR FR (g|g)R (20)
R∈S

To derive the Langevin equation we consider

djt (X) = jt+dt (X) − jt (X) = dUt∗ XUt + Ut∗ XdUt + dUt∗ XdUt (21)

The only nonvanishing products for quantum stochastic differentials are

dBR (t)dBR∗ (t) = 2Re (g|g)−

R dt, dBR∗ (t)dBR (t) = 2Re (g|g)+
R dt. (22)

Combining terms in (21) and using (19), (14), (20) and (22) we get the Langevin equation
X X
djt (X) = jt ◦ θα (X)dM α (t) = jt ◦ θnR (X)dM nR (t) + jt ◦ θ0 (X)dt (23)
α n=−1,1;R∈S+

where
dM −1,R (t) = dBR (t), θ−1,R (X) = −i[X, FR∗ ], R ∈ S+ (24)
dM 1,R (t) = dBR∗ (t), θ1,R (X) = −i[X, FR ], R ∈ S+ (25)
and X 
(0,−1) (0,1) (−1) (1)
θ0 (X) = θ0,R + θ0,R + θ0,R + θ0,R (X) = (26)
R∈S
X
= −iIm (g|g)− ∗ + ∗
R [X, FR FR ] + iIm (g|g)R [X, FR FR ]+
R∈S

+Re (g|g)− ∗ ∗ + ∗ ∗
R (2FR XFR − {X, FR FR }) + Re (g|g)R (2FR XFR − {X, FR FR })

is a quantum Markovian generator. Notice that the factors Re (g|g)±

R are > 0 only for R ∈ S+
and vanish for R ∈ S− .
We will also use the following notions
(1)
θ0,R = Re (g|g)−
R κ1,R , κ1,R (X) = 2FR∗ XFR − {X, FR∗ FR }; (27)
(−1)
θ0,R = Re (g|g)+
R κ−1,R , κ−1,R = 2FR XFR∗ − {X, FR FR∗ }. (28)
Let us prove that generators above, i.e. the θ–maps satisfy some special properties which will
be crucial to prove the existence of the solution of the flow equation. Omitting, for simplicity, the
(0,±1) (±1)
indices R, we will consider all the above defined operators θ0 , θ1 , θ−1 (= θ1∗ ), θ0 , θ0 , κ1 , κ−1 ,
acting on the Uniformly Hyperfinite (UHF, cf. [9]) C ∗ –algebra BS generated by the identity and
the local operators (those acting only on a finite number of spins). On this algebra the limits, as
Λ → Z d , of all the structure maps are well defined and their domains contain, as a common core,
the unital dense ∗-subalgebra B0 of the local operators which is also invariant under the square
root. All our operators (as well as the operators F and F ∗ ) map B0 into itself. Moreover θ1 , θ−1
(0,±1)
are nonsymmetric derivations and θ0 are symmetric derivations. From the above described
properties and from (3.2.22) of [9] it follows that κ1 , κ−1 are (symmetric) dissipations. But in
fact κ1 , κ−1 have a much stronger properties, than dissipativity. We will use this property in this
section without further comments. The crucial point is the following lemma.

8
Lemma 1. For arbitrary local X, Y , i = ±1 we have

κi (XY ) = κi (X)Y + Xκi (Y ) + 2θ−i (X)θi (Y ) (29)

Proof . Let us prove (29) for i = −1 (for i = 1 the proof is analogous). Formula (28) gives

κ−1 (T ) = 2F T F ∗ − T F F ∗ − F F ∗ T

so that
κ−1 (T ) = [F, T ]F ∗ + F [T, F ∗ ]
For T = XY we get
κ−1 (XY ) = [F, XY ]F ∗ + F [XY, F ∗ ] =
= [F, X]Y F ∗ + X[F, Y ]F ∗ + F [X, F ∗ ]Y + F X[Y, F ∗ ] =
= [F, X]F ∗Y + X[F, Y ]F + F [X, F ∗ ]Y + XF [Y, F ∗] + 2[F, X][Y, F ∗] =
= κ−1 (X)Y + Xκ−1 (Y ) + +2(−i)[X, F ](−i)[Y, F ∗ ]
From (24), (25) it follows that this is exactly (29) and that proves the lemma.
Theorem 1. For any pair of local operators X, Y , the structure maps in the Langevin equation
(23) satisfy the equation
X
θα (XY ) = θα (X)Y + Xθα (Y ) + cβγ
α θβ (X)θγ (Y ) (30)
β,γ

where the structure constants cβγ

α is given by the Ito table
X
dM β (t)dM γ (t) = cβγ α
α dM (t) (31)
α

The conjugation rules of dM α (t) and θα are connected in such a way that formula (23) defines a
∗–flow (∗ ◦ jt = jt ◦ ∗).
Proof . Follows from Lemma 1 and formulae (24), (25), (26) by direct calculation.
The existence of the infinite volume dynamics and and its approximation by finite volume
ones, is discussed in [7], [8].

5 The simplest case: one dimensional nearest neighbor

interaction
In this section in order to compare our results with the known results on Glauber dynamics
we consider the simplest case of one dimensional translationally invariant Hamiltonian with the
nearest neighbor interaction

Jrs = Jr+1,s+1 , Jrs = Jr,r+1 = J > 0

In this case for every r we have 4 configurations σ(r) ∈ Σ(r) of nearest neighbors of the spin at r.
We will denote these configurations ++, +−, −+ and −− (the first symbol is the orientation of

9
the spin on the left of r and the second — on the right). Only the configuration ++ will give a
contribution in the stochastic limit of interaction Hamiltonian (will lie in Σ+ (r)) and the energies
E = E(r, σ) = 2J for different r will be equal. Then (6) takes the form
X Z
(ε,µ)∗
FΛ dk g(k)e−it(ω(k)−2J)) a(k) + h.c. (32)
ε,µ=++,+−,−+,−−

(ε,µ) X
FΛ = Fr(ε,µ)
r∈Λ

Fr(++) = |1r−1ih1r−1 ||1r ih−1r ||1r+1 ih1r+1| + | − 1r−1 ih−1r−1 || − 1r ih1r || − 1r+1 ih−1r+1 |
and the other operators Fr(ε,µ) are defined correspondingly. In this case there is only one quantum
white noise, denoted b(t, k), which is the stochastic limit of
1 − it2 (ω(k)−2J)
e λ a(k)
λ
and satisfies the commutation relations

[b(t, k), b∗ (t0 , k 0 )] = 2πδ(t − t0 )δ(ω(k) − 2J)δ(k − k 0 )

The Langevin equation (23) takes the form

X
djt (X) = jt ◦ θα (X)dM α (t)
α=−1,0,1

where
(++)∗
dM −1 (t) = dB(t), θ−1 (X) = −i[X, FΛ ]
(++)
dM 1 (t) = dB ∗ (t), θ1 (X) = −i[X, FΛ ]
 
(0,−1) (0,1) (−1) (1)
θ0 (X) = θ0 + θ0 + θ0 + θ0 (X) =
X  
= −iIm (g|g)− εµ∗ εµ + εµ εµ∗
(εµ) [X, FΛ FΛ ] + iIm (g|g)(εµ) [X, FΛ FΛ ] +
ε,µ
 
(++)∗ (++) (++)∗ (++)
+Re (g|g)−
(++) 2FΛ XFΛ − {X, FΛ FΛ } +
 
(++) (++)∗ (++) (++)∗
+Re (g|g)+
(++) 2FΛ XFΛ − {X, FΛ FΛ }

ACKNOWLEDGEMENTS
S.Kozyrev is grateful to Luigi Accardi and Centro Vito Volterra where this work was done for
kind hospitality. This work was partially supported by INTAS 96-0698 grant.

10
References
[1] Accardi, L., Frigerio, A., Lu, Y.G., ”The Weak Coupling Limit For Fermions”, Journ. Math.
Phys. 32, pp.1567–1581 (1991).

[2] Accardi, L., Lu, Y.G., Volovich, I.V., ”Interacting Fock spaces and Hilbert module extensions
of the Heisenberg commutation relations”, Publications of IIAS, Kyoto, (1997).

[3] Accardi, L., Lu, Y.G., Volovich, I.V., Quantum Theory and its Stochastic Limit, Springer
Verlag, 2000, to appear.

[4] Accardi, L., Lu, Y.G., Volovich, I.V., ”A white noise approach to classical and quantum
stochastic calculus”, Preprint of Centro Vito Volterra N.375, Rome, July 1999.

[5] Accardi, L., Kozyrev, S.V., Volovich, I.V., ”Dynamics of dissipative two-level system in the
stochastic approximation”, Phys.Rev.A 57, N3 (1997), quant-ph/9706021.

[6] Accardi, L., Kozyrev, S.V., ”Glauber dynamics from stochastic limit”, in: White noise anal-
ysis and related topics, Volume in honor of T.Hida, CESNAM, Kyoto, 1999.

[7] Accardi, L., Kozyrev, S.V., ”The structure of stochastic flows”, submitted to: Infinite dimen-
sional Analysis, Quantum Probability and related topics (1999).

[8] Accardi, L., Kozyrev, S.V., ”Stochastic dynamics of of lattice systems in stochastic limit”, to
appear in: Chaos, Solitons and Fractals.

[9] Bratteli, O., Robinson, D.W., Operator algebras and quantum statistical mechanics 1, New
York Heidelberg Berlin, Springer–Verlag, 1979.

[10] Glauber, R.J., ”Time dependent statistics of the Ising model”, J. Math. Phys. 4, pp.294–307
(1963).

[11] Liggett, T.M., Interacting particle systems, Berlin, Heidelberg, New–York, Springer Verlag,
1985.

[12] Majewski, A.W., Zegarlisnki, B., ”Quantum stochastic dynamics II”, Reviews in Mathemat-
ical Physics 8, N5 pp.689–713 (1996).

[13] Martin, P.A., Buffet, E., ”Dynamics of the Open BCS Model”, Preprint.

[14] Matsui, T., ”Markov semigroups which describe the time evolution of some higher spin quan-
tum models”, J. Funct. Anal. 116, pp.179–198 (1993).

[15] Spitzer, F., ”Random fields and interacting particle systems”, Proc. M.A.A. Summer Seminar,
Math. Ass. Amer., Washington DC 1971.

[16] Sullivan, W.G., ”Mean square relaxation times for evolution of random fields”, Commun.
Math. Phys. 40, pp.249–258 (1975).

11