Anda di halaman 1dari 38

Mathematical theory of non-equilibrium quantum

statistical mechanics
 
V. Jakšić and C.-A. Pillet

Department of Mathematics and Statistics


McGill University
805 Sherbrooke Street West
Montreal, QC, Canada H3A 2K6

PHYMAT
Université de Toulon, B.P. 132
F-83957 La Garde Cedex, France

FRUMAM
CPT-CNRS Luminy, Case 907
F-13288 Marseille Cedex 9, France

April 24, 2002

Dedicated to David Ruelle and Yakov Sinai


on the occasion of their 65th birthday

Abstract

We review and further develop a mathematical framework for non-equilibrium quantum


statistical mechanics recently proposed in [JP4, JP5, JP6, Ru3, Ru4, Ru5, Ru6]. In the alge-
braic formalism of quantum statistical mechanics we introduce notions of non-equilibrium
steady states, entropy production and heat fluxes, and study their properties. Our basic
paradigm is a model of a small (finite) quantum system coupled to several independent
thermal reservoirs. We exhibit examples of such systems which have strictly positive en-
tropy production.

1
2

1 Introduction
The properties of a physical system out of thermal equilibrium are usually described in term
of phenomenological concepts like steady state, entropy production and heat flux. These
notions are related by the fundamental laws of thermodynamics. As an illustration, consider
a model describing a small system  coupled to two infinite heat reservoirs  ,
which
are at temperature  , . Under normal conditions, one expects that the combined system
                    
                
                    
                
                    
 
  
 
  
 
  
 
  
 
  
 
  
 
  
 
  
 
  
 
  
 
     
                
                   
  
  
  
  
  
  
  
  
  
  
   
  
  
  
  
  
  
  
  

                   
   
   
   
   
   
   
   
   
   
   
    
  
  
  
  
  
  
  
  

                    

 

 

 

 

 

  

 

 

 

 

   
                 
                   

 

 

 

 

 

 

 

 

 

 

   
                 
                    
 
  
 
  
 
  
 
  
 
  
 
  
 
  
 
  
 
  
 
  
 
     
                
                   
  
  
  
  
  
  
  
  
  
  
   
  
  
  
  
  
  
  
  

                   
   
   
   
   
   
   
   
   
   
   
    
  
  
  
  
  
  
  
  

                   

 

 

 

 

 

 

 

 

 

 

   
                
                   

 

 

 

 

 

 

 

 

 

 

   
                
                    
  
  
  
  
  
  
  
  
  
  
    
                

Figure 1: A system coupled to two heat reservoirs

will settle into a steady state in which there is a constant flow of heat and entropy from
the hotter to the colder reservoir across the system  . Let  be the heat current flowing
from reservoir 
 into the small system  , and  the entropy production rate in  . In the
steady state, the fundamental laws of thermodynamics read:

  !#"%$


(1)
 
 !)(*+,"%-
& '
The first relation expresses energy conservation (the first law of thermodynamics). The
second asserts that the heat flows from the hotter to the colder reservoir and that the entropy
of  is not decreasing (the second law of thermodynamics).
Our goal is to give a precise mathematical meaning to the notions of non-equilibrium steady
state, entropy production and heat flux, study their properties and prove Relations (1) from
first principles. We will also exhibit examples of combined systems .,/0
which
have strictly positive entropy production and hence non-trivial thermodynamics.
We will work in the mathematical framework of algebraic quantum statistical mechanics
which appears to be particularly well-suited to the study of general structural properties of
non-equilibrium steady states. The basic notions of this algebraic framework are briefly
introduced in Section 2. The reader is referred to the monographs [BR1, BR2, Ha, Si1,
Th] for more detailed expositions. In Section 3, we define non-equilibrium steady states
and discuss their basic structural properties. Section 4 is devoted to the notion of entropy
production. Finally, a simple class of models with strictly positive entropy production is
described in Section 5.
3

This review is based on a series of recent papers [JP4, JP5, JP6, Ru3, Ru4, Ru5, Ru6].
These works rely on a large body of knowledge previously developed by many authors
in various areas of mathematical physics: Equilibrium statistical mechanics, quantum dy-
namical systems, quantum Markovian processes, van Hove limit, linear response theory,
etc... Even though we have tried to provide the reader with the most relevant references to
these earlier works, we do not claim completeness in this respect, and refer the reader to
[BR1, BR2, OP] for an extensive list of references.
Acknowledgment. We are grateful to Jan Derezinski and David Ruelle for useful discus-
sions on the subject of this review. A part of this work has been done during the visit of the
first author to Johns Hopkins University. V.J. is grateful to Steve Zelditch for his friend-
ship and to Mathematics Department of Johns Hopkins University for generous support.
The second author thanks Jean-Pierre Eckmann for hospitality at the University of Geneva
where part of this work has been done, and the Swiss National Science Foundation for
financial support. The research of the first author was partly supported by NSERC.

2 The Framework
In its algebraic formulation, the quantum mechanics of a physical system is described by a
132 2
or a 4 - dynamical system. To avoid technicalities we will consider in this review only
1 2 2
- systems. The non-equilibrium statistical mechanics of 4 - systems will be discussed
elsewhere.
132
- dynamical systems are introduced in Subsections 2.1 and 2.2 below. In Subsection 2.3
we describe some examples of such systems.

2.1 56 - Dynamical Systems


1 2 1 2
A - dynamical system is a pair 798:$<;>= , where 8 is a - algebra and ;%? a strongly
continuous group of @ - automorphisms of 8 (that is, the map ACBDE; ? 7GFH= is norm-continuous
for each FJIK8 ). We assume that 8 has an identity L . The elements of 8 describe
observables of the physical system under consideration and the group ; specifies their time
132
evolution. An example of a - algebra is MN7PO/= , the algebra of all bounded operators on a
1 2
Hilbert space O , equipped with the operator norm topology. Any - algebra is isomorphic
to a subalgebra of MQ7POR= for some Hilbert space O .
2
A state of the system is described by a positive linear functional SIT8 satisfying SU7VL%=!
W
. The number SU7G;%?X7GFQ=Y= is the expected value of the observable F at time A , assuming that
the system was initially in the state S . The set Z:798[= of all states on 8 is a convex, weak-*
2
compact subset of the dual Banach space 8 .
1 2
A state S\I]Z
798[= is called ; - invariant, or steady state, if S_^N;%?H!KS for all A . A -
dynamical system has at least one (and typically many) steady states. We call quantum
dynamical system a triple 798:$<;%$YS`= , where S is a ; - invariant state.
4

A quantum dynamical system 798:$<;%$YS`= is called ergodic if


W k f
acbed 2 2
fhg3i j f SU7Gm ; ? 7GFH=nmo=qprAC!sSU7GFH=tSU7Gm mo=u$
l

for all Fv$<mwI8 . It is said to have the property of return to equilibrium if


aebed 2 2
x x g3i SU7Gm ; ? 7GFQ=nmo=!ySU7GFQ=tSU7Gm mo=u-
?

Thermal equilibrium states are characterized by the KMS condition. Let zK{|" be the
inverse temperature. A state S is 7G;%$Yz&= -KMS if, for all Fv$<mwIT8 , there is a function } ~€ 
d
analytic inside the strip ‚„ƒQ…n".†ˆ‡ ƒ‰†ˆzŠ , bounded and continuous on its closure, and
satisfying the KMS boundary conditions

} ~€  7PAY=C!sSU7GF; ? 7Gmo=Y=u$ } ~€  7PA€‰‹9z&=C!sSU7G; ? 7Gmo=nFQ=u$

for AHIŒ . A KMS state is ; - invariant. The quantum dynamical system 798:$<;%$YS`= , where
S is a 7G;%$Yz&= -KMS state, describes a physical system in thermal equilibrium at temperature
W„Ž
z .
Note that a 7G;%$Yz&= -KMS state is also a z -KMS state for the dynamics defined by ;%?’‘”“–•n‘ .
Even though, in most systems, the physical temperature is a non-negative parameter, it is
mathematically convenient to define KMS state for all z#I—Œ,˜™‚”šœ›]Š . The case z]!"
corresponds to infinite temperature and 7G;%$<"ž= -KMS states (; - invariant traces) are some-
times called chaotic states. Let Ÿ be the generator of ; . The state S is called 7G;%$uš3›,= -KMS
2
state if šQ‹9SU7GF Ÿq7GFH=Y=o ˆ" for all F¡I¢£7¤Ÿ¥= . 7G;%$Yz¦= -KMS states at values z#!K3› and
z§!¨(› are called respectively ground states and ceiling states.
1 2
Let S be a state on the - algebra 8 . We denote by 7POo©€$<ª«©€$X¬­©h= the GNS-representation
of 8 associated to S . An injective representation ª«© is called faithful. A state ®I_Z:798[=
is called S - normal if there is a density matrix ¯ on O°© such that ® 7n±²=*!¨³ ´„7P¯µª«©&7n±²=Y= . The
set ¶
© of all S - normal states is a norm closed convex subset of Z
798[= . ¶
© is sometimes
called the folium of S . Any ®§IR¶ © has a unique normal extension to the enveloping von
Neumann algebra ·y©/!#ª«© 798[=t  . The state S is called factor state (or primary state) if its
enveloping von Neumann algebra is a factor i.e., if its center ·y©¦¸­· © consists of multiples
of the identity.
The effectiveness of the algebraic formalism of quantum statistical mechanics is largely due
to Tomita-Takesaki modular theory of von Neumann algebras. We assume that the reader is
familiar with the basic results of this theory as discussed, for example, in [BR1, BR2, Ha,
OP]. For notational purposes we recall some well-known facts.
The state S is called modular if ¬© is a separating vector for ·y© i.e., if S extends to a
faithful normal state on ·y© . Any KMS state at inverse temperature z¹I#Œ is modular.
Assume that S is a modular state on 8 and denote by º»©0!K¼¾½µ¿ , À and Á the modular
operator, the modular conjugation and the natural cone associated to the pair 7P·y©€$X¬­© = .
5

The operator ÂC© is self-adjoint while À is an anti-unitary involution on Oo© . These operators
are characterized by the relation
2
À ¼ ½ ¿ • FQ¬­©£!yF ¬­©€$ (2)

which holds for any FÃIR· © . The natural cone Á is the norm closure of the set

‚„FœÀFN¬­©… FÄIR·y©€Šr-

An important property of the natural cone is that for every state ®™I¶
© there is a unique
1œ2
vector ¬ÅTI0Á such that ® 7n±²=œ!|7¤¬Ş$<ªÆ©&7n±²=V¬År= . Moreover, if ; is a - dynamics on 8
(not necessarily leaving the state S invariant), then there is a unique self-adjoint operator Ç
on Oo© such that, for all A ,
l
ª«© 7G; ? 7GFQ=Y=!y¼ÉÈ ?eÊ ªÆ©¦7GFQ=t¼ È ?eÊ $
(3)
l
¼ È ?eÊ ÁˆËÁ[-

The operator Ç is called the standard Liouvillean. The first formula in (3) allows us to
extend ; to all of ·y© .
A state ®RIo¶
© is ; - invariant iff Ç­¬ÅN!y" . Thus, the study of S - normal, ; - invariant states
reduces to the study of Ì3¼É´XÇ . This is the first link between quantum statistical mechanics
and modular theory. The second one is Takesaki’s theorem: S is a 7G;%$Yz&= -KMS state iff

ÂC©
!¨( z€ÇU- (4)

The third link is quantum Koopmanism: The spectral properties of the standard Liouvillean
Ç encode the ergodic properties of the quantum dynamical system 798:$<;%$YS`= in complete
analogy with Koopman’s lemma of classical ergodic theory [JP1, JP3]. For example, if the
state S is modular, then 798:$<;%$YS`= is ergodic iff zero is a simple eigenvalue of Ç . More-
over, the system returns to equilibrium if the singular spectrum of Ç reduces to this simple
eigenvalue.

2.2 Local Perturbations


Let 798:$<;%$YS`= be a quantum dynamical system describing a physical system in a steady state.
Our guiding physical principle is that the response of the system to local perturbations will
reveal its basic thermodynamical properties. We now discuss the mathematical formalism
needed to deal with local perturbations.
Let Ÿ be the generator of the dynamical group ; ? . The operator Ÿ is a @ - derivation: Its
domain ¢£7¤Ÿ¥= is a @ - subalgebra of 8 and for Fv$<mÍI/¢£7¤Ÿ¥= ,
2 2
Ÿµ7GFH= !#Ÿq7GF =u$ Ÿq7GFmo=C!Οµ7GFH=nmÏ—FQŸµ7Gmo=u-
6

2
Let Ð be a local perturbation i.e., й!ÃÐ I™8 . The generator of the perturbed dynamics
is ŸÉÑU7n±²= !¨Ÿq7n±²= ‹uÒÓÐ$ɱVÔ . The operator ŸÉÑ is also a @ - derivation and ¢R7¤ŸÉÑC= !Ä¢R7¤Ÿ”= . The
perturbed dynamics is given by the automorphisms ; Ñ ? !y¼ ?eÕ×Ö ,
k k k
Ù ? ?GÜ ?e݄ÞqÜ
; Ñ ? 7GFQ=C!y; ? 7GFQ=€,Ø ‹ Û pžAX Û prAV ¦±É±É± Û pžA Ù Ò ; ? Ý 79Ðß=u$„Ò–±É±É±Ò ; ?GÜ 79л=u$<; ? 7GFH=×ԒԒÔ9-
ÙžÚ 
132
The pair 798:$<;¾ÑC= is a - dynamical system. If S is modular and Ç is the standard Liou-
villean associated to ; , then the standard Liouvillean associated to ;¾Ñ is the self-adjoint
operator given by
Ç Ñ !yǧ,ÐÎ(—À Ð[ÀÆ$
with domain ¢£7GÇ­Ñ=!s¢R7GÇ­= .

2.3 Examples
2.3.1 Finite Quantum Systems
1 2
Let àwáˆâã be a finite dimensional Hilbert space and 8ä!¹MN7Gàœ= . A - dynamics ; is
determined by a self-adjoint operator (Hamiltonian) å on à
l
; ? 7GFQ=!y¼ È ?cæ F¼ È ?cæ -
The dynamics associated with the local perturbation Ð is
Ñ&é l Ñ é
; Ñ ? 7GFH=!s¼ÉÈ ?Gç’æ`è F¼ È ?Gçêæ­è -
A states SÎI™Z
798[= is determined by a density matrix on à which we denote by the same
letter, so SU7GFQ=§!ë³ ´ì7PSFH= . The state S is faithful iff Sí{E" and it is ; - invariant iff
Ò S $<å/Ô !y" .
For any z_I£Œ the density matrix
l Ž l
¼ ‘æ ³ ´ì7P¼ ‘æ =u$
defines the unique 7G;%$Yz&= -KMS state on 8 . On the other hand, if S is a faithful state, then
1 2
for any zK!¡î " there exists a unique - dynamics ï ©  such that S is a 7Gï ©  $Yz&= -KMS
‘ l  a’ðñ ‘
state. This dynamics is generated by the Hamiltonian ( z S .
The GNS-representation 7PO © $<ª © $X¬ © = of 8 associated to the state S can be explicitly con-
structed as follows. Let "ˆ+óò +ô±É±É±:+óò be the eigenvalues of S counted with
ã
multiplicities and denote by õ the corresponding eigenvectors. Fix a complex conjugation
È
õ0BD õ on à . Then
O © !yàsö—à»$

ªÆ© 7GFH=!yF0öyL>$

ã 
¬­©
! Ø ò • õ  ö õ  -
ø÷ 
7

The standard Liouvillean corresponding to the dynamics ; generated by the Hamiltonian


å is
lj!yåwöyLv(LQö å.$
where, by definition, åTõy! å õ . Finally, let us describe the modular structure associated
to a faithful state S . The modular conjugation acts as ÀC7Põ™ö0ù'=! ù»ö õ , and the modular
operator is given by
a’ðñ a’ðñ a’ðñ
©
! ºß©
! S§ösLß(LNö S -

Isolated finite quantum systems have no interesting thermodynamics. However, models


where a finite quantum system is coupled to infinite thermal reservoirs are one of the basic
paradigms of quantum statistical mechanics.

2.3.2 Free Fermi Gases

Let O be the Hilbert space of a single fermion and ú its Hamiltonian. For example, a
free non-relativistic spinless electron of mass û is described by the Hilbert space Oü!
Ž¥j
Ç 7PŒýÆ$<p¾þ'= and its Hamiltonian ú is the operator of multiplication by þ û . Another
example is given by a spinless lattice fermion with Hilbert space Oÿ!

 
7 = and Hamil-
tonian ú
!Ã(º , the usual discrete Laplacian on . 
Let  l 7POR= be the anti-symmetric (fermionic) Fock space over O . For Í  I)O , denote
by h7 = and  7 = the annihilation and creation operators on  l 7PO/= . They are bounded
2
operators satisfying
h7 =
*!
 7 =
!

. As usual,  stands for either  or  . Let
2 2
8 be the - algebra generated by ‚   7   = … IROTŠ . The map
132

; ? 7  7 
 =Y=!  7P¼ È ?  =u$
1 2
extends to a - dynamics on 8 which can be explicitly written as
l
; ? 7GFQ=!y¼ È c? æ F¼ È ?cæ $

where åÿ! `7¤úh= . The 1 2


- dynamical system 798:$<;>= describes a free Fermi gas.
132

One often deals with the even subalgebra 8 , the - algebra generated by

  7   = ±É±É±  7  Ù = …§!s"%$ W $ɱɱɱ   $ɱɱɱ€$  Ù


‚ I/OŠr-

The pair 798  $<;>= is also a 1 2


- dynamical system.
A self-adjoint operator on O , such that "o+— Ä+ÎL , determines a state SI/Z:798[= by
2
  2
   !  !#"
SU7 7 µu= ±É±É± 7 Ù = h7 q = ±É±É± '7 H=Y=!#Ÿ  Ù «¼ ø‚µ7 $V ”=XŠr-
È
$"  &% ! '&(
This, so called quasi-free gauge invariant state, is completely determined by its two-point

   ! ! '
function 2
SU7 7 = '7 %=Y=C!ˆ7 >$V =u- (5)
8

l

A quasi-free gauge invariant states S is ; - invariant iff for all A , ¼ È ? ” U¼ È ? v!] . In particu- 
lar, the quasi-free gauge invariant state determined by y!#}°7¤ú = describes a free Fermi gas
with energy density per unit volume }°7GZ»= . For any zI Œ , the quasi-free gauge invariant
W
state determined by y!)7 _¼É‘ µ=
l 
#
is the unique 7G;%$Yz&= -KMS state on 8 .

)
The GNS-representation and the associated modular structure of the quasi-free gauge in-
variant state S determined by can be explicitly computed (see [AW]). Let

be the
number operator and ¬ the Fock vacuum on l 7PO/= . Fix a complex conjugation on O and
extend it to a complex conjugation on l 7POR= . Set 
O © !  l 7POR= ö* l 7POR=u$

¬­©
!#¬_ö0¬Q$

  '7Y7VLß(
ªÆ© 7 '7 =Y=!


Q= • = öyLQ#7V(3Lq= ã ö * 2 7  •  =u-
The triple 7PO°©€$<ªÆ©$X¬­© = is the GNS-representation of the algebra 8 associated to S . The
modular conjugation acts as ÀC79ö 3=»! Îö ,+ .- + /-
 , where |! - 7V(3L%= ã ç’ã
l té
• . The
modular operator is

ÂC©
!
a’ðñ
ºß©
!­710=ösLß(LNö2`7 0=u$
where !0 a’ðñ
ß7
W
( Q=
l 
. The corresponding Liouvillean is given by

Ǘ! `7¤úh=öyLv(LQö2`7 úh=u-


A quasi-free gauge invariant state is primary. It is modular iff ÌH¼É´Y y!#ÌH¼É´ì7VLN(£ Q=!Ä"qŠ .
From the above discussion it follows easily that, if S is a quasi-free gauge invariant and
; - invariant state, then the quantum dynamical system 798:$<;%$YS`= is ergodic iff ú has no
eigenvalues. Moreover, if the spectrum of ú is purely absolutely continuous, then this
system returns to equilibrium.

2.3.3 Interacting Lattice Fermi Gases


1œ2
Consider the - dynamical system 798:$<;>= describing the free lattice Fermi gas introduced
9 :
in the previous subsection. For 0I 3 4 
, we set 7 =U! 5 63 7 $8 9
N7¤Ÿ µ= and, to any finite subset
Ë , we associate the local algebra 8 <;
generated by ‚ 7 '= … ™I   63 3
Š . The algebra 8
;
is quasi-local with respect to the net ‚”8 ߊ , in particular 8 is the norm closure of ˜ Q8 . =; <;
9 : 
The fermions interact through a real valued pair potential £I

> ? 
7 = . For any finite subset
Ë , the local perturbation

@; 8 Ø A B ; >h763£(DCÆ= 2 763'= 2 76CÆ='76C«=h763 =u$


Ð Î!
9

;
induces a dynamics ;%? on 8 . Moreover, there exists a dynamics ; on 8 FE
=; ; G 
such that, for any
F|Is˜ H8 and for any increasing sequence of finite subsets ÍË

which eventually
contains any finite subset of , one has

E
; ? 7GFH=! H aebcIKd J@L ; H ? 7GFQ=u-
The
1œ2
FE
- dynamical system 798:$<; ”= describes an interacting Fermi gas on the lattice  .
This important system is little understood (see however [Hu]).

2.3.4 Lattice Spin Systems

Let à be a finite dimensional Hilbert space and, for each I 3 4 , let à'8 be a copy of à .
To a finite domain ]Ë G :
we associate the Hilbert space

O H !Îö8 B H OM8q$
132
and the corresponding - algebra

8 H !yMN7PO H =u-
G NGH S PORQ SKTUQ H Ü with a subalge-
If U*Ë ` , then the injection FÎBDEF§ö‰L
132 Ü
allows us to identify 8
H H . We can identify
bra of 8
H
. Let 8 denote the - algebra obtained by completing ˜ 8
the local algebras 8 with subalgebras of 8 .
The algebra 8 describes observables of the infinite spin system, the subalgebra 8 H con-
taining the observables of the spins inside the domain G .
9 ËV  into self-adjoint elements of
9
An interaction is a function  from the finite subsets
8 such that H7 =IT8; . The Hamiltonian

å‰7G =!EØ
9
;XW H H7 =u$

of a finite domain G induces a dynamics


Hé l Hé
; H ? 7GFH=!s¼ È ?cæ*ç F¼ È ?cæ ç $

9 9
on 8 .
Let us denote by … … the number of points in . If the interaction satisfies
Y Z J@ L Ø
H7 9
&^ ; x x
8 B ;\[]8 = <¼ †]›s$ (6)

_
for some [{" , then the limit

; ? 7GFH=! H aebeIKd JL ; H ? 7GFQ=u$


10

H H
exists for any F|Iy˜ 8 . Moreover, ; extends by continuity to a dynamics on 8 . The
132
pair 798:$<;>= is the - dynamical system describing an infinite quantum spin system.
Whenever Condition (6) holds, there exist at least one (and possibly many) 7G;%$Yz&= -KMS

H
states on 8 for any zÎI‰Œ . Such states are constructed as thermodynamic limits of local
KMS states defined on 8 . Under some additional regularity conditions on the interaction
 the KMS state is unique for small enough z .
We refer the reader to [BR2, I, Ru1, Si1] for detailed information on the kinematical struc-
ture and equilibrium thermodynamics of quantum spin systems, and to [Ru3, Ru4, Ru5,
Ru6] for their non-equilibrium statistical mechanics. The dynamical aspects of quantum

case is the one-dimensional


9a`
spin systems are comparatively little studied. To our knowledge the only well-understood
model, see [Ar1, Ar2].

3 Non-equilibrium Steady States


In Subsection 3.1, we define non-equilibrium steady states (NESS) and describe their basic
structural properties. Stability properties of quantum dynamical systems are discussed in
Subsection 3.2. Finally, in Subsection 3.3, we review two approaches to the study of NESS.

3.1 Definition and Basic Properties of NESS


Let 798:$<;%$YS`= be a quantum dynamical system and Ð|I08 a local perturbation. The non-
equilibrium steady states (NESS) of the locally perturbed system 798:$<; Ñ = are the weak- *
limit points, as yD뜛 , of the states
W k f
S Ñ cb
f
Û ST^­; Ñ ? prA - (7)

In other words, S Ñè is a NESS iff there is a sequence Ù Dô3› such that, for all FI.8 ,
one has f
aebed
S Ñ è 7GFQ=! Ù S Ñ Ý 7GFH=u-

d
The set èÑ 7PS`= of NESS is a non-empty weak-* compact subset of Z:798[= whose elements
are ;¾Ñ - invariant.
d l
The set Ñ 7PS`= is defined analogously, by taking #D (› in Equ. (7). Although states in
d l
Ñ 7PS`= are non-physical, they are both technically and conceptually useful.
Obviously, the ; - invariance of S plays no role in the above definitions, and we can define
de
the NESS and the sets Ñ 7P®«= for any initial state ®
IZ:798[= .
Remark. There is a fair amount of arbitrariness in the above definition. The ergodic mean
in Equ. (7) can be replaced by another averaging procedure. Without further assumptions
on the ergodic properties of the system, the resulting set of NESS will generally not coincide
11

d
with èÑ 7P®«= . However, most results in this section are either independent of our specific
choice of averaging, or can be easily adapted to other averagings. For technical reasons,

d e fhg
related to the use of spectral analysis (see [JP5]), a particularly useful alternative is Abelian
averaging. We denote by Ñ  7P®«= the set of weak-* limit points of the state
k i
l i
Û ¼ ? ®3^­; Ñ ? pžAu$
kj e (8)

as iml " .

3.1.1 Dependence on the Initial State

d
Clearly èÑ 7P®«= describes the set of steady states that can be reached starting from ® , and thus
depends on the particular choice of the initial state ® . However, we expect that sufficiently

d
similar initial states should be driven towards the same set of asymptotic states. Indeed,
under some mild regularity assumption, èÑ 7P®«= does not depend on the choice of ® as long
as it remains in the folium ¶ © . The following result will be proved in [JP6].

Theorem 3.1 Let S be a factor state and assume that, for all ®RIo¶ © and Fv$<mwIT8 ,
W k f
aebed
Û ® 7YÒ ; Ñ ? 7GFH=u$<mvÔP=qprA`!y"%$
fhg i
e
holds (weak asymptotic abelianness in the mean). Then Ñ 7P®Æ=U! dne ode
Ñ 7PS`= for all ®§I¶
© .
In particular, if ÌH¼É´<Ç`ÑÄ!¨
î ‚„"qŠ , then it is one-dimensional and there is a unique S - normal
de
;¾Ñ - invariant state S&Ñ such that Ñ 7P®Æ=`!Ă¾S&Ñ­Š for all ®RIo¶
© .

3.1.2 Normal and Singular NESS

From now on, we fix the initial state S , and investigate the structural properties of the states
de
in Ñ 7PS`= . First we remark that if the GNS-representation ª«© is not faithful, we can consider
Ž
the quotient dynamical system on the algebra 8 ÌH¼É´Xª © (which has a faithful representation
on Oo© ). One easily sees that the NESS of the original system are obtained by lifting to
8 the NESS of the quotient system with the help of the canonical projection ¯ 8 D
Ž
p
8 ÌH¼É´Xª«© . Thus, without loss of generality, we may assume that the GNS-representation
ªÆ© is faithful and identify 8 with ªÆ© 798[= .
2
q q sr
A positive linear functional I_8 is called S - normal iff _!ˆò for some ™IR¶
© and r
q
ò—{Ã" . It is called S - singular iff y ¨ù Ä" for some S - normal ù implies ù_!ˆ" . Any
q
state I/Z:798[= has a unique decomposition

qT!q Ù 4 qutø$ (9)

where q Ù and qut are positive linear functionals and q Ù is S - normal while qut is S - singular
(in particular, q Ù and q t are disjoint, see [KR, Ta] for details). Since S - normal states are
12

mapped to S - normal states by ;¾Ñ , the uniqueness of this decomposition implies that if q
is ;¾Ñ - invariant, then both Ù and q
qut q Dd e
are ;¾Ñ - invariant. We say that a NESS _I Ñ 7PS`= is
normal if its S - singular part qt q
is zero and purely singular if its S - normal part Ù is zero.

Theorem 3.2 Assume that qIadÑe 7PS`= , then:

1. If q is a factor state, it is either normal or purely singular.


2. If S is a factor state, either q is purely singular or every S - normal state is also q -
normal.

If the system has settled into a NESS ÍI q od


Ñè 7PS`= , then it is described by the quantum
vq
dynamical system 798:$<;¾Ñ$ = . From the point of view of thermodynamics, a fundamental
question is whether this system has a strictly positive entropy production. At this level of
generality, it is far from obvious even how to define entropy production. We will return to

q
this question in Section 4. We will prove that the entropy production of the normal part
Ù is always zero. Therefore, a necessary condition for non-trivial thermodynamics is that
q
sI
q
î ¶
© . In other words, must be “sufficiently far” from S , so far in fact that it does not
“live” in the same Hilbert space Oo© . Although physically natural, this restriction is one of
the main sources of difficulty in the mathematical study of NESS.
The following result gives useful criteria for a system to have either purely singular or
normal NESS.

Proposition 3.3 Assume that S is modular, and let Ç`Ñ be the Liouvillean of the locally
perturbed system 798:$<; Ñ = .

1. If Ì3¼É´<Ç`Ñ_!ς„"qŠ , then any NESS in deÑ 7PS`= is purely singular.


2. If Ì3¼É´<Ç`Ñ contains a separating vector for ·y© , then
d d
èÑ 7PS`=C!
l
wq
Ñ 7PS`=!Ă ¦Šßˍ¶
©-

We finish with a brief discussion of equilibrium vs. non-equilibrium steady states in quan-
tum statistical mechanics. Consider a quantum dynamical system 798:$<;%$YS`= and assume
that S is a 7G;%$Yz¦= -KMS state for some z { " . The Araki perturbation theory of KMS
states [Ar5, BR2, DJP] yields that for any local perturbation Ð there exist a S - normal
7G;¾Ñ$Yz&= -KMS state S&Ñ . Under normal conditions, it is expected that for any ®ˆIö
© ,
de
Ñ 7P®«=!ó‚¾S Ñ Š . There is no interesting thermodynamics in the sense that the entropy
production of S&Ñ is zero. Hence the main question in thermal equilibrium is whether the
quantum dynamical system 798:$<;¾Ñ$YS&ÑC= is ergodic or returns to equilibrium. Although it is
generally expected that these properties hold for physical systems under normal conditions
(the zeroth law of thermodynamics), there are very few non-trivial models for which this
has been proven (see [BFS, DJ, JP3, M]).
The expected scenario in the non-equilibrium case is quite different. One considers a
quantum dynamical system 798:$<;%$YS`= which is not in thermal equilibrium and a local per-
turbation Ð . Under the influence of the perturbation the system will settle into a NESS
13

S Ñè Iad èÑ 7PS`= . One expects that under normal circumstances


aebed g i
? è
®H^`;%Ñ ? !sS Ñ è for all
® I£¶
© . Moreover, one expects that S Ñ I£
T è î ¶
© and that the entropy production associated
to S Ñè is strictly positive. Establishing this scenario for physically relevant models is one of
the central goals of non-equilibrium quantum statistical mechanics.

3.2 Stability
From a physical point of view, the prominent feature of thermal equilibrium is its stability
under local perturbations. This basic phenomenon appears in two flavors: Structural and
dynamical stability.
Let 798:$<;%$YS`= be a quantum dynamical system where S is a 7G;%$Yz&= -KMS state. It follows
from Araki’s theory [Ar5, BR2, DJP] that for every local perturbation Ð there is a 7G;¾Ñ $Yz&= -
KMS state S&Ñ0Io¶
© such that


<S&э(S
­!yxo7
Ðz
=u- (10)

Moreover the map S_BD S&Ñ is, for fixed Ð , a bijection from the set of 7G;%$Yz&= -KMS states to
the set of 7G;¾Ñ$Yz¦= -KMS states. This shows that the set of thermal equilibria of the system is
structurally stable under local perturbations.
Dynamical stability does not hold without further assumption. However, if we assume that
W k f
aebed

v

Û ìÒÓÐ$<; Ñ ? 7GFQ=×Ô >pžA`!y"%$


fhg i
e
(11)

de
for all F¡I,8 , then any NESS in Ñ 7PS`= is a 7G;¾Ñ $Yz&= -KMS state (this is a simple variant
of Proposition 5.4.6 in [BR2]). Moreover, if 798:$<; Ñ $YS Ñ = is ergodic, then the structural
de
isomorphism S‰BDES&Ñ is dynamically realized as Ñ 7PS`=!Ă¾S¦Ñ­Š .
It is a deep fact that thermal equilibrium is actually characterized by the stability criteria
(10) and (11). There are several results in this direction. We will describe one of them, for
others see [BKR, HKTP, HTP, NT]. The following ergodicity assumption will be needed.

(E) For any self-adjoint element Ð of a norm-dense @ -subalgebra 8 Û Ë]8 there


is a ò Ñ0{" such that
k i

l i

{ v

ìÒÓÐ$<; ? Ñ 7GFH=×Ô >prA*†s›s$ (12)

holds for all … ò&…q†]ò Ñ and FÃIT8 Û ,

Under assumption (E), for any self-adjoint НIT8 Û the strong limits
| {e b {
aebcd l
g i ; Ñ? ^ ; ? $
Ñ
? e
} {e b aebcd l {
Ñ
? e
g i ; ? ^­; ? Ñ $
14

| {e
| {e
7 Ñ =
l 
! } {e l {
exist on 8 for all … ò&…†)ò>Ñ . The Møller morphisms Ñ are @ - automorphisms of 8
Ñ . Since ST^; ? Ñ !sST^; ? ^; ? Ñ , we get { and

~{e Ñ d{e Ñ 7PS`=!Ă¾S Šr$

with S Ñ S°^ {e b } {e Ñ . It follows that S_!sS {e Ñ ^ | {e Ñ !


aebed
?
g
e i S {e Ñ ^&;%? , from which we
get the formula
k i
{
S è Ñ 7GFH=¦(™S {Ñ
l
7GFQ=`!s‹nò
l i
S m{!€‚Ñ vƒ ç²? é 7YÒÓÐ$<; ? 7GFQ=×ÔP=qpžAu- (13)

The stability requirement is:

(S) For any self-adjoint НI8 Û and ò small enough, there exists a ; Ñ - invari- {
{
ant state S Ñ I[¶
© , such that

d{e Ñ 7PS`=!Ă¾S {Ñ Š „†… { aebeg d Û


<S { Ñ (™S
!y"%-

l
Assume that (E) and (S) hold, then S è Ñ ! S Ñ and it follows from Equ. (13) and the { {
dominated convergence theorem that
k i
SU7YÒÓÐ$<; ? 7GQ
F =×ÔP=qpžA`!y"%- (14)
l i

This is the famous stability criterion of [HKTP]. It is a well-known result of Haag and
Trych-Pohlmeyer [HTP], and of Bratteli, Kishimoto and Robinson [BKR] that (14) together
with (E) implies that S is a KMS state. More precisely,

Theorem 3.4 Assume that S is a factor state and that assumption (E) holds. Then (S)
holds if and only if S is a 7G;%$Yz&= -KMS state for some z.IRŒ‰˜T‚”šœ›]Š .
1 2

An example of - dynamical system satisfying condition (E) is the even subalgebra
798 $<;>= of a free Fermi gas introduced in Subsection 2.3.2. Let 8 Û be the @ - subalgebra

of 8 consisting of finite sum of monomials

  7 µX= ±É±É±$  7 ¥ Ù =u$ (15)

such that for all ‹<$ , U‡


 È $Y¼ l È ? &(„=­I/Ç  7PŒ`=ø- 7 (16)
Then Condition (12) holds for any FÃIT8 Û , any self-adjoint ЈIT8 Û and any ò/I/Œ . More-
over, the algebra 8 Û is norm-dense in 8  iff ú has purely absolutely continuous spectrum
(see [BM] for details).
The stability of thermal equilibrium can also be understood in spectral terms. Consider a
quantum dynamical system 798:$<;%$YS`= , and assume that the ; - invariant state S is modular.
Since Ç­¬ © !¨" , zero is an eigenvalue of the standard Liouvillean Ç of the system. There
15

might be other eigenvectors in ÌH¼É´<Ç , associated to other S - normal invariant states. In


physical situations of interest, the zero eigenvalue is embedded in the continuous spectrum
of Ç which covers the entire real line. It is a piece of folklore that, unless prevented by some
symmetry, this eigenvalue will turn into complex resonances under the influence of a local
perturbation Ð . Therefore, one expects that the standard Liouvillean Ç Ñ !#ÇT—Ð#(_À лÀ
of the perturbed system 798:$<;¾ÑC= will have purely continuous spectrum and that there will

{
be no ;¾Ñ - invariant state in the folium ¶
© . Indeed, by computing the Fermi golden rule

{
for Ç Ñ , the stability requirement (14) can be seen to be a condition which ensures that
zero is an eigenvalue of Ç Ñ to second order of perturbation theory. Moreover, the KMS
condition Equ. (4) can be interpreted as a form of symmetry which forces Ç Ñ to have a
zero eigenvalue. To see this, note that since À Ð[À belongs to the commutant 8v , we have
Ñ é l Ñ é l
¼ È ç–Ê%è ? À€Ð[À ¼ È ç–Ê%è ? !s¼ È Êž? À ÐoÀ ¼ È Êµ? -

Therefore, using Equ. (4), we can write

¼ÉÈ çêÊqè
Ñ&é
? Ç­Ñ»¼
l
È ç–Ê%è
рé
? ¬­©
! ‰ˆ¤Ç™0ÐÄ(™¼ÉÈ Êž? À Ð[À'¼ l È Êµ?UŠ ¬­©
!ÎÐv¬­©°(™¼ÉÈ Êž? À Ð߬­©

Œ‹ìÐÄ(º
!
l
Ž
© È ?c•n‘ À Ð 3¬­©-
Ž¥j
The analytic continuation of the right hand side of the last identity to A3!ÿ( ‹×z further
gives
‹ l 
 
Œ‹
ìÐÄ(.º © • À Ð Q¬©
! ”ÐÎ(0Àº © • Ð 3¬­©
!ˆ79ÐÏ(‰Ð =V­
2
¬ ©£!#"%$ 
from which we conclude that

¬*Ñ b ¼
l
‘žçêÊqè
Ñ&é
• ¬©$ (17)

must be in the kernel of Ç­Ñ . In fact, up to normalization, Equ. (17) is nothing but Araki’s
formula for the vector representative of the 7G;¾Ñ$Yz&= -KMS state S&Ñ , see e.g. [Ar5, BR2,
DJP].

3.3 The Study of NESS


From the above discussion, it should be clear that the study of local perturbations of a
quantum dynamical system 798:$<;%$YS`= depends critically on the nature of the state S . We
can distinguish two cases:

1. If S is a 7G;%$Yz&= -KMS state, we are dealing with a system near thermal equilibrium.
We expect that
aebed
? e
g i ®œ^­; Ñ ? 7GFH=!sS¦Ñ*7GFH=u$
16

for any initial state ® in the folium ¶


© and any F¹I 8 , where S&Ñ is a 7G;¾Ñ$Yz&= -KMS
state. This ergodic problem can be reduced to the spectral analysis of the standard
Liouvillean Ç Ñ . Although spectral theory of the standard Liouvillean has been so far
understood only for very few systems, the conceptual framework of ergodic theory
near thermal equilibrium is well-understood.
2. If S does not belong to the folium of a 7G;%$Yz&= -KMS state, the system is far from
equilibrium. In contrast with the former case, the conceptual framework for the study
of NESS is not well-understood. In the remaining part of this subsection, we describe
the two approaches that have been adopted so far in the mathematically rigorous
literature.

3.3.1 The Scattering Approach

Let 798:$<;%$YS`= be a quantum dynamical system and Ð a local perturbation such that for all
F in a norm-dense @ - subalgebra 8 Û Ë]8 ,
k i k i

l i

v

ìÒÓÐ$<; ? 7GFQ=×Ô >pžA*†]›s$


l i
ìÒÓÐ$<; Ñ ? 7GFQ=×Ô >pžA*†]›s-

(18) v

As in the previous subsection, these conditions ensure that the strong limits
} eÑ b g aebcd i l
? e
; ? ^­; Ñ ? $

exist and define @ - automorphisms of 8 . It follows that that dneÑ ~e


7PS`=!Ă¾S Ñ Š and

~e } eÑ acbed
S Ñ !yST^ !
?
g
ei ST^; Ñ ? 7GFH=u-

}e
The Møller morphisms Ñ induce unitary operators Ñ between the GNS spaces Oo© and -<e }

O © . These operators intertwine the representations ªÆ©£^ Ñ and ª © and provide uni- e ‘
Ö
tary equivalences between the standard Liouvilleans associated to S and S Ñ . Under some
Ö
e
additional regularity assumptions one can also show that for suitable F¨I8 ,
k i k k
~e
S Ñ 7GFH=C!ySU7GFQ= Ø
ÙµÚ 
79šN‹n=
Ù
Û
?GÜ
prA  Û prA ±É±É± Û
?e݄޵Ü
e e
pžA Ù SU7YÒ ; ? Ý 79л=u$„Ò–±É±É±€$„Ò ; ?GÜ 79л=u$<FUԒԒÔP=u-

(19)
The expansion (19) can be used for perturbative computations of basic thermodynamic
quantities such as entropy production and heat fluxes.
It is important to remark that the assumptions (18) are very difficult to verify in concrete
physically interesting models, and so far they have been established only for a few exam-
ples. Nevertheless, we believe that the scattering approach is a helpful tool in developing
the conceptual understanding of the subject, much in the spirit of the “chaotic hypothesis”
used in classical non-equilibrium statistical mechanics (see [Ru2]).
17

An example of a system where (18) holds has been provided by Botvich and Malyshev
[BM]. Consider a free Fermi gas with one particle Hilbert space O and one particle energy
ú . Assume that ú has purely absolutely continuous spectrum. Let 8 Û consists of finite sums

     
of monomials
7  = ±É±É± 7 Ù =u$
such that (16) holds. Set Ð b k’ ’
ò , where \I™8 Û is an even polynomial. Then (18) holds
for sufficiently small òI Œ . In this case, each term in the expansion (19) is well-defined,
and the series converges absolutely for all FÃIT8 Û .

3.3.2 The Spectral Approach

As already pointed out, the thermodynamically interesting NESS associated to a local per-
turbation Ð of the quantum dynamical system 798:$<;%$YS`= are not in the folium ¶ © . Since on
the other hand the kernel of Ç`Ñ , and more generally its eigenvectors, provide information
about S - normal ;¾Ñ - invariant states, the thermodynamical content of the spectral theory
of Ç`Ñ is not easily decoded. This explains why, for many years, scattering theory was the
only available approach to the study of singular NESS.
Recently, in [JP5], we have developed a spectral theory of NESS which we describe here
in its simplest setting. According to the remark in Subsection 3, we define NESS using
Abelian averaging (8).
Let 798:$<;%$YS`= be a quantum dynamical system and Ð a local perturbation. Assume that S is
a modular state and that the map

“RABD
Œ Ð
?
!s¼ÉÈ ?e½ ¿ Ð3¼
l
È ?e½ ¿ IR· ©
d W„Ž¥j
extends to an analytic function (in norm) inside the complex strip „‚ "Ć|‡ Ï ƒ † Š ,
1
bounded and continuous on its closure (the set of such Ð is total in · © ). The - Liouvil-
lean of the locally perturbed system is defined by

Ç i b ǧ,ÐÎ(—À Ð l
ȕ
À%-

This operator is closed on ¢£7GÇ i = b d


¢ 7GÇ= . Its spectrum is contained in ‚q… ‡ ƒ £
£ … +

Ð
ȕ
É Š , and its adjoint is given by
2
Ç i !yÇ  ÐÄ(—À Ð ÀÆ-
ȕ
2
The operators ‹×Ç i and ‹×Ç i generate quasi-bounded strongly continuous groups on O°© .
By construction ¬ © I/Ì3¼É´XÇ i , thus

¼ È ?eÊ 5” ¬­©
!#¬­©$

(see Equ. (2)) and it follows from the Trotter product formula that

; Ñ ? 7GFH=!s¼ È ?eÊ” F¼


l
È ?eÊ” !y¼ È ?eÊ @”• F¼
l
È ?eÊ@”• - (20)
18

For ‡
d
—– 2
ƒ°{," , we define linear functionals S QIT8 by the formula

—–
S 7GFQ=b
k i
‹ Û ¼ È ?U
–
S 7G; Ñ ? 7GH
F =Y=qprA -


Clearly, the map ƒBD S is weak- @ analytic in the half-plane ‚¾‡
d
ƒ™{Ã"qŠ . Moreover, for
d
‡ ƒ:{ Ð ˜
e

ȕ
, we have

S 7GFH=! —– 2
ˆ l 
¬­©$<Fß7GÇ i (ƒµ= ¬© - Š
i™l i
This formula and the fact that Abelian NESS )I q ,d e hf g
Ñ  7PS`= are weak-* limit points, as
2
Ž
" , of 7 ‹n=tS
È
j
suggests that NESS are described by zero-resonance eigenvectors of
Ç i . Indeed, it is possible to develop an abstract axiomatic complex deformation tech-
2
nique which allows to directly relate NESS to the zero-resonance eigenvectors of Ç i . This
method has been used in [JP5] to study the NESS of a finite quantum system coupled to
several fermionic reservoirs at different temperatures. This analysis has led to results which
could not be reached by scattering methods. In particular, it allows to obtain precise infor-
1
mation on the relaxation to the NESS from the study of complex resonances of the -
Liouvillean. We briefly describe the model and results of [JP5] in Section 5.3.
The spectral approach to the study of NESS is a recent development which has led to some
insights into the general structure of the non-equilibrium quantum statistical mechanics
and has been a useful tool in the study of some concrete models. The method is still
being developed and its full potential remains to be reached. It should be noticed that this
approach to the dynamical properties of quantum dynamical system is closely related to
the study of the decay of correlations in classical dynamical systems, and in particular to
Ruelle resonances of the transfer operator (see for example [Ru7, Ru8, Ru9] and [E]).

4 Entropy Production

4.1 Phenomenological Considerations


In non-equilibrium thermodynamics, the local entropy production rate ï is defined as the
source term in the local entropy balance equation
š * bœŸž !yï'$
?U›
ž
where
› is the entropy density and the entropy flux. In a stationary state, the total entropy
production rate in a subsystem  is therefore equal to the net entropy flux leaving this
subsystem. If  interacts with independent reservoirs   $Y $ɱɱɱ , which are in thermal
equilibrium at inverse temperature z ¾$Yz> ¥$ɱɱɱ , then entropy leaves  with a rate given by
the formula
(TØ z ¥µ$

19

¢m¢m£m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £¢£¢ ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡ ¡  ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥¤¥¤
¢m¢m£m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £¢£¢ ¡ m¡ m¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m   ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡ ¡  ¥¤m¥¤m¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥¤¥¤
¢m¢m£m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢  £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £¢£¢ ¨ ¡ m¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡ ¡  ¨ ¥¤m¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤  ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥¤¥¤
¢m¢m£m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £¢£¢ ¡m ¡ m¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡ ¡ 

¥¤m¥¤m¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥¤¥¤
¢m¢m£m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £¢£¢ ¡ m¡ m¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡m ¡m  ¡ ¡  ¥¤m¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥¤¥¤
¢m¢m£m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £m¢£m¢ £¢£¢ ¡m  ¡m  ¡m  ¡m  ¡m  ¡m  ¡m  ¡m  ¡m  ¡m  ¡m  ¡  ¥¤m¥¤m¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥m¤¥m¤ ¥¤¥¤
¨ 
§m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §¦§¦
§m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §¦§¦
§m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦  §m¦§m¦ §m¦§m¦  §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §¦§¦
§m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §¦§¦
§m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §m¦§m¦ §¦§¦
Figure 2: The energy currents N . ©
where  denotes the energy current leaving the reservoir : (see Fig. 2). The Hamiltonian
of the combined system  ‰  y±É±É± is

åÿ!#å »Ø «ª 
«¬~­`Ð$
å

where Ð describes the interaction between  and the reservoirs. The Heisenberg equation
for the total energy of the reservoir : leads to the expression #! ( ‹ Ò å$<å Ô
! ¬~­
( ‹uÒÓÐ$<å «¬ ­
Ô . To make the connection with the algebraic formulation we note that

N!Οøq79л=u$

where Ÿø ‹ Òå b «¬ ­
$€±VÔ is the generator of the dynamics of 
 . The entropy production rate
in a stationary state can now be written as q
6q ®q*¯&(§Ø 
C 7 =! z>¥Ÿøq79Ðß= ]- ° (21)

Clearly, the argument leading to this formula physically makes sense only in the idealized
case where the interaction Ð is so small that it does not affect the thermal equilibrium states
of the reservoirs. In concrete models, strictly speaking this will only be the case if either
Ðä!Í" or zœ!z [! ±É±É± . However, if the reservoirs are initially in thermal equilibrium
at different temperatures, we do not expect a local perturbation Ð to be strong enough to
induce a global approach to thermal equilibrium in the combined system T./,±É±É± . In
such circumstances, provided that the restriction to : of the NESS remains sufficiently q
20

close to the initial thermal equilibrium, the quantity defined in Equ. (21) should still carry
useful information on the thermodynamics of the system.
6q
In fact, as we shall see in the next section,  7 = has many of the expected properties of
6q
an entropy production. In particular, C€7 =» ˆ" for an NESS , and the strict positivity q
6q
of C 7 = is a sufficient condition for the existence of energy currents. Thus, we consider
Equ. (21) as a sufficient motivation for the general definition of entropy production given
below. We refer the reader to [Ru4, Ru5, Ru6] for a more detailed discussion of entropy
production in spin systems.

4.2 Definition and Basic Properties


Let 798:$<;%$YS`= be a quantum dynamical system. For any state ®¡Iw¶ © we denote by
~…±%
 „7P®œ…tS`= the relative entropy of ® with respect to S . The basic properties of the rela-
tive entropy are discussed in the monograph [OP]. We recall that  ì7P®3…tS`=H+Î" and that ~…±%
for finite quantum systems

~…±%„7P®3…tS`=!#³€´ 7¤®'7 a’ðñ


 S(
a’ðñ
®«=Y='- (22)

Note also that our notation for the relative entropy differs from the one originally introduced
by Araki in [Ar3, Ar4] by a sign and the order of its two arguments.
We shall need two assumptions. The first one concerns the state S :
132 W
(A1) There exists a - dynamics ïÆ© such that S is a 7Gï«©$¾( = -KMS state.
W
Remark 1. The choice of the reference temperature z§!)( in (A1) is made for mathemat-
1 2
ical convenience. If (A1) holds, then for any zs!Î î " there is a - dynamics ï ©  such that
l ‘
S is 7Gï ©  $Yz&= -KMS state (set ï'© ?  !ï © ?c•n‘ ). Let us also point out that a state satisfying
‘ ‘ l
(A1) is modular and that Takesaki’s theorem shows that ï © ? 7GFQ=C!s¼ È ?e½ ¿ÆF È ?e½ ¿ .
Remark 2. A state which can be factorized into a product of KMS states satisfies Condition

²
(A1). Indeed, if
ã
S_! S µ$ (23)
É÷ Û
l ­W
where S  is 7G;Éž$Yz>= -KMS, then S is 79öHì;  ‘ ? $¾( = -KMS. Since this is a common situation
in non-equilibrium statistical mechanics, our first hypothesis is quite natural.
Let Ÿu© be the generator of ï«© . Our second assumption is a regularity condition for the local
perturbation Ð :

(A2) ЈIR¢R7¤Ÿu©'= .
21

Remark 3. To make the connection with the discussion of the previous section and Formula
(21), note that for the state (23) we get

Ÿu©£!)(TØ z>¥Ÿøµ$


where Ÿø is the generator of ;É . Thus, identifying the factor corresponding to !)" with ³
the system  and the remaining factors with the reservoirs 
 , we can easily reproduce
Formula (21) by setting z Û !#" .
For a ; - invariant state S satisfying Hypothesis (A1) and a local perturbation Ð satisfying
(A2), we introduce the observable

ï Ñ b Ÿu©&79Ðß=IT8:$

and define the entropy production of the locally perturbed system 798:$<; Ñ = in the stationary
state ®£IZ:798[= , with respect to the reference state S , as

€7P®«= b ® 7Gï Ñ­=u-

The following theorem is a simple extension of the main result in [JP4] (see [JP6] for the
proof).

Theorem 4.1 Assume that Hypotheses (A1) and (A2) hold. Then, for any state ®_I™¶
©
one has
~…±%
 ì7P®œ^­; Ñ ? …tS`=!y „7P®œ…tS`=¦(
k
?
…P%
Û ®œ^; Ñ 7GïhÑC=qp $ (24)
t ›
both sides of this equality being either finite or (› .

The proof of Theorem 4.1 is based on the Araki perturbation theory of KMS states [Ar5].
It is however an instructive elementary exercise to verify Relation (24) for finite quantum
systems. We use the notation of Example 2.3.1 and set

®
? b ®œ^­; Ñ ? !s¼
l
È ?Gçêæ­è
Ñ é
®`¼ È ?Gç’æ`è
Ѧé
-

From Equ. (22) we get

…P%„7P®œ^; Ñ ?
 …tS`=C!y³€´„7P®
?
a’ðñ
S`=¦(³ ´„7P®
?
a’ðñ
® =
?
a’ðñ a’ðñ
!y³€´„7P® S`=¦(³ ´„7P® ®«=u$
?
and hence
p
p A
r
~…±%
 „7P®3^ ; Ñ ? t… S`=!s‹ž³ ´ì7YÒ ® $<åwÐHÔ
?
a’ðñ
S`=u-
a’ðñ
Since S is ; - invariant, we have Ò S <$ å/Ô'!#" and the cyclicity of the trace leads to

p
p A
r
 ~…±%ì7P®œ^­; Ñ ? …tS`=!¨(*‹ž³€´”7P®
?
Ò
a’ðñ
S $uÐHÔP=u-
22

a’ðñ
Finally, note that Ÿu©&79л=!s‹uÒ S $uÐHÔ , from which we conclude

p
p A
ž
…P%”7P®3^­; Ñ ? …tS`=!¨(*³ ´„7P®
?
Ÿu© 79л=Y=!¨( ®3^; Ñ ? ¤7 Ÿu©79Ðv=Y=u-

q ad q
In the rest of this section we describe some f basic properties of the entropy production. Let
acbed
I èÑ 7PS`= be such that C7GFH=! Ù S Ý 7GFH= , then one has
W W k f
aebed
~…±% f aebcd
Ù Ù  ì7PST^; Ñ Ý … S`=!¨( Ù Ù Û
Ý
ST^; Ñ ? 7Gï ÑC=qpžA!)(*€7 =u- (25) 6q
The following result follows immediately from this relation.

Theorem 4.2 Assume that Hypotheses (A1) and (A2) hold. Then, any I q ´d èÑ 7PS`= satisfies

6q
€7 =  "%-

Remark. Obviously  7 =:+w" if )I 6q q ,d l


6q
Ñ P7 S`= . In particular, C€7 =[!¡" for any )I q
d zd
èÑ 7PS`=¸
l
Ñ 7PS`= .
6q
As displayed by Formula (25), €7 = describes the rate at which the relative entropy de-
creases along the trajectory S§^­; Ñ ? . The idea of defining entropy production as the asymp-
totic rate of decrease of some relative entropy was already present in the early works [Sp1]
and [LS], although in the slightly different context of quantum semigroups. For quantum
dynamical systems, a similar construction can be found in [OHI, O1, O2], where the posi-
tivity of entropy production is also discussed.
The next result shows that this asymptotic behavior, and thus the value of €7 = , is essen- 6q
tially independent of the choice of the reference state used to compute the relative entropy.

Proposition 4.3 Assume that (A1) holds, then there is a norm-dense set of states ¶ ©  ˍ¶
©
such that, as AD › ,

~…±%„7P®œ^; Ñ ? … S
  =!y~…±%ì7P®œ^­; Ñ ? … S`=2µo7 W =u$
for any S'Io¶]©  and any ®£Io¶
© .

q
Since we believe that a state —I£¶
© describes the same thermodynamics as S , we expect

q
such a state to have vanishing entropy production. This was shown in [JP4] under the
assumption that is faithful. The next statement generalizes this result.

Proposition 4.4 Assume that Hypotheses (A1) and (A2) hold. Then, for any ;¾Ñ - invariant,
q
S - normal state , one has
€7 =C!y"%- 6q
23

Under the assumptions of the above proposition it follows that, for any NESS .I q ´d eÑ 7PS`= ,

6q Vq 
one has
C 7 =! 7Gï Ñ =u$
where qdenotes the S - singular part of q . Thus the singular part of a NESS contains the
full information about its entropy production.
The reverse of Proposition 4.4 holds in a slightly weaker form:

Proposition 4.5 Assume that (A1) and the following condition hold,
aebedb
?
g3i …¶ ~…±%ì7PST^; Ñ ? …tS`=­{Î(›s-

d yd wq
Then, one has
l
èÑ 7PS`=! Ñ P7 S`=!Ă ¦ŠœË¶
©-

Under some weak ergodicity assumption, we can actually characterize S - normality of a


NESS by the vanishing of its entropy production. This result is an effective way to prove
strict positivity of entropy production in some interesting examples, see Section 5.3.

Proposition 4.6 Assume that I q ·d èÑ 7PS`= satisfies


k f
YKZ
f¹¸  Û«º Û Gˆ ST^ ; Ñ ? 7Gï ÑC=(DqC7Gï ÑC= Š ºº †]›s-
ºº prA
ºº
(26)

º
Then q is S - normal if and only if C 76q =!s" .

q Nde 6q
Finally, we wish to briefly come back to the interpretation of entropy production. For a
NESS yI Ñ 7PS`= , we have seen that €7 = describes the divergence rate of the entropy
of S™^*; Ñ ? relative to a “typical” reference state in ¶ © . However, we do not yet have such
an interpretation for an arbitrary ;¾Ñ - invariant state ®ˆIˆZ
798[= (we are grateful to J.L.
Lebowitz for raising this question).
Let ®RI/Z:798[= be an arbitrary state. Since ¶
© is weak-* dense in Z:798[= , there is a sequence
® Ù Io¶ © which converges towards ® in the weak-* topology. Moreover, one easily arranges
…P%
that sequence to satisfy  ”7P® Ù …tS`={Î(N› . By Theorem 4.1, the limit

M0
º 7P®h$YA<= b aebed
Ù ˆ ~…±%„7P® Ù ^; Ñ ? …tS`=&(~…±%„7P® Ù …tS`= Š $ (27)

exists, is independent of the choice of the approximating sequence ® Ù and satisfies

M0
k
? t
º 7P®h$YAY=`!¨( Û œ® ^; Ñ 7Gï Ñ =qp
›-
Hence, if ® is ;¾Ñ - invariant, then  7P®«= is the rate of divergence of the entropy differential
M0
º ,
º 7P® $YA< »0
= !¨(*AqC€7P®Æ=u-
24

4.3 Stability and Entropy Production


Let 798:$<;%$YS`= be a quantum dynamical system such that 798:$<;>= satisfies Assumption (E) of
Section 3.2 for some @ - subalgebra 8 Û . It follows that for any self-adjoint ÐëIÎ8 Û and
… ò …&†¹ò Ñ , one has d {e {e
Ñ 7PS`=H! ‚¾S Ñ Š . Assume that S satisfies Hypothesis (A1) and that
(A2) holds for all self-adjoint НIT8 Û . Finally suppose that the following assumption also
holds:

(A3) For all self-adjoint йIT8 Û

{ ¼ KY { Z ¸ ºº {t
k
x x  Û Û
?
7PS ^­; Ñ 7Gï Ñ`=¦(™S è Ñ 7Gï Ñ`=Y=qp { › ºº †]›s-
Ö ?

ºº ºº
By Theorem 4.1, if C€7PS { è Ñ =C!y" the assumptions of Proposition 4.5 hold and therefore

{
S è Ñ !sS {l Ñ I[¶
©-

Moreover, the well-known entropic inequality


<S {è Ñ (™S
+Î(
j
~…±%ì7PS {è Ñ
 …tS`=u$

together with the upper semi-continuity of the relative entropy, yields the estimate


<S {è Ñ (™S
y! µ[79òh=u-
Thus the stability assumption (S) is satisfied and, provided S is a factor state, Theorem 3.4
yields that S is a 7G;%$Yz¦= -KMS state for some z_IRŒ‰˜T‚”š3›sŠ .
Hence, under sufficient regularity assumptions, S is a KMS state if and only if the entropy
production vanishes for sufficiently many local perturbations Ð . More precisely:

Theorem 4.7 Let 798:$<;%$YS`= be a quantum dynamical system where S is a factor state sat-
isfying (A1). Assume that (E) holds and that (A2) holds for all self-adjoint Ð Iä8 Û .

q d {
Finally assume that (A3) holds. Under these hypotheses, S is a 7G;%$Yz&= -KMS state for some
6q
z¹I#Œ#˜‰‚”š3›]Š if and only if C 7 =o!ÿ" for all ˆI è Ñ 7PS`= , all local perturbations
ЈI§8 Û and all sufficiently small òTI£Œ .

Examples where all conditions of Theorem 4.7 are satisfied can be constructed using the
even subalgebra of a free Fermi gas. We omit the details.

4.4 Time-Dependent Perturbations


In this section we consider time-dependent local perturbations of a quantum dynamical
system 798:$<;%$YS`= . The response of the system to such perturbations will shed an additional
light on the notion of entropy production.
25

1 2
The entropy production for time-dependent perturbations of - dynamical systems have
been previously studied in [OHI, O1, O2]. Some of the basic formulas of this section seem
to be known for a long time (see Remark on page 281 in [PW]).
A time-dependent local perturbation is specified by a norm-continuous, self-adjoint, 8
valued function Ð
7PA<= on Π. The perturbed time evolution is a family of norm-continuous
automorphisms of 8 given by the formula
k k k
; Ñ ? 7GFH= b
; ? 7GFH= Ø
ÙµÚ 
Ù ? ?GÜ
‹ Û prAu Û prAV ¦±É±É± Û
?e݄޵Ü
prA Ù Ò ; ? Ý 79Ð:7PA Ù =Y=u$„Ò–±É±É± $„Ò ; ?GÜ 79Ð
7PAuX=Y=u$<; ? 7GFQ=×ԒԒÔ9-

Note that in general ;%Ñ ? is not a group.


The usual interaction representation of this time evolution is given by

; Ñ ? 7GFQ=`!
2
?Ñ ; ? 7GFH= ?Ñ $   (28)


where ?Ñ is the unitary element of 8 satisfying the differential equation

p
p A
r
 ?Ñ !s‹U ?Ñ ; ? 79Ð
7PA<=Y=u$ (29)
Û
with the initial condition Ñ !¨L . 
Throughout this section, we assume that the state S satisfies (A1) and that Ð:7PA<= satisfies
(A2) for all A . Moreover, the maps A­BD Ð:7PAY= and A`BD Ÿ © 79Ð
7PAYY= = are assumed to be respec-
1  1
tively 7PΠ$u8o= and 7PΠ$u8o= . We set

ï Ñ*7PA<= b Ÿu©¦79а7PA<=Y=u$

and for any state ®


IZ:798[= , we define the rate of entropy production by

C 7P® $YA<= b ® 7Gï Ñ7PAY=Y=u- (30)

The analog of the Theorem 4.1 is the following result.

Theorem 4.8 For any state ®£Io¶


© , such that  ~…±%„7P®œ…tS`={Î(› , one has
…P%ì7P®œ^­; Ñ ?
 …tS`=( ~…±%„7P®œ…tS`=!¨(
k
? t
Û ®œ^­; Ñ 7GïhÑ*7 › =Y=qp › !¨( ‹9® 7 ?Ñ Ÿu©&7 ?Ñ =Y=u-
2
(31)

Its proof follows closely the proof of Theorem 4.1. Time-dependent local perturbations
allow us to discuss some physical aspects of entropy production which could not be dis-
cussed within the framework of Section 4.2. The most interesting of those aspects involves
the relation between entropy production and Carnot’s version of the second law of thermo-
dynamics.
26

Assume that the system 798:$<;%$YS`= is initially in thermal equilibrium – this means that S is
a 7G;%$Yz&= -KMS state, where zy{Ä" is the inverse temperature. The KMS condition implies
that Ÿ © !)( z Ÿ . Assume also that Ð:7PAY= vanishes outside of the interval Ò "%$V *Ô . Relation (31)

~…±%  ˆ  Š
yields that f f f 2
 „7PS§^; Ñ …tS`=C!s‹9zSU7 Ñ Ÿ Ñ = - (32)
On the other hand, the quantity 4 b f
( ‹9SU7 Ñ Ÿ  ˆ  Š
f 2
Ñ = is precisely the work done on the
system by the time-dependent force Ð . Let us briefly elaborate this well-known point (see
Section V.3.3 in [Ha] and Appendix to Section IV.5 in [Si1]). Assume that our system is
finite, that is, that 8 is a finite dimensional algebra. Denote by å the Hamiltonian of the
unperturbed system, so that Ÿµ7n±²=£!|‹uÒ å$ɱ Ô . The Hamiltonian of the perturbed system is
å‰7PA<=!yåwÐ:7PAY= , and its energy at time A is
½ 7PAY=C!sSU7G; Ñ ? 7Gå—7PAY=Y=u-
Since å‰7 H=!yå , the total amount of work done on the system is given by

4 ! ½ ½ f
7 Q=¦( 7G"ž= !ySU7G; Ñ 7Gå =Y=(™SU7Gå =
f

!)( Û ST^­;%Ñ ? 7¤Ÿq79У7PAY=Y=Y=qpžA (33)
f f 2
!)( ‹9SU7 Ñ Ÿµ7 Ñ =Y=u$  
where the last relation is easily derived with the help of (29) and (28). For infinite systems,
the total energy is infinite and å is not well-defined. However, the work 4 remains a
well-defined quantity given by one of the two last formulas in (33). In particular, Equ. (33)
yields the following expression for the instantaneous power dissipated into the system:

*Ñ*7PAY= b ( ST^; Ñ ? ¤7 Ÿq79У7PAY=Y=Y=u-

Equ. (32) is the integrated form of the relation

C€7PS§^­; Ñ ? $YA<=!sz& Ñ 7PAY=u- (34)

Obviously, (32) and (34) are expressions of the thermodynamical relation p _!sz€p 0 ¿ .
We further remark that since the relative entropy is always non-positive and z¨{¹" , (32)
implies that the energy 4 transfered by the external perturbations is always non-negative
– this is Carnot’s version of the second law of thermodynamics which says that one cannot
extract work from a system in thermal equilibrium. With regard to the usual discussion of
passivity, energy transfer and the second law of thermodynamics (see [BR2, Ha, Si1, Th]),
we emphasize the relation of these notions with entropy production.
Consider now two independent systems 798Hµ$<;ɵ$YS = , each of which is in thermal equilib-
rium at inverse temperature z> (S& is a 7G;Éž$Yz = -KMS state). We set

8)!Ï8ß&ö83 ¥$ ;°!y;”&ö—;É ”$ S_!sS ö‰S& ì$


27

and denote by Ÿø the generator of ;É . The generator of ; is given by Ÿ! Ÿìϟø . As
already noticed, Assumption (A1) holds and Ÿu©—!¡(œ7Pz uŸì`]z> ¾Ÿø „= . Let Ð:7PA<= be a time-
dependent local perturbation of 798:$<;%$YS`= , vanishing outside of the time interval Ò "%$V UÔ , and
establishing a temporary link between the two subsystems. In accordance with the above
discussion, the total work done by the external perturbation is

 
f f 2
4 !¨( ‹×SU7 Ñ Ÿq7 Ñ =Y=u-

Obviously, 4 !Î4_ 4™ , where


f f 2

4™N!¨( ‹×SU7 Ñ Ÿø%7 Ñ Y= =u$

³
is interpreted as the work done on the -th subsystem. As in the previous example, it
follows from Theorem 4.8 that

~…±% f
z  4  ‰z 4 !)(* „7PST^; Ñ …tS`=u- (35)

Relation (35) has an interesting consequence [PW]. If  b


l 
z  denotes the temperature of
³
the -th subsystem and &U{— , then S is not a thermal equilibrium state, and in principle
it is possible to have 4_*†" , which means that the first system produces a positive amount
of work during a cyclic process. Since the relative entropy is non-positive, we get from (35)
that 4 †," and
4 &C(
+ -
4  
This is the well-known Carnot’s formula which states that the efficiency of a heat engine
Ž
is limited by 7 H(] ' „= & , where & is the temperature of the heat source and ' the
temperature of the heat sink (compare with remark on page 216 of [BR2] and [PW]).
A deep result of Pusz and Woronowicz [PW] asserts that (under a mild additional regularity
condition and assuming &U{— ) there exist perturbation Ð:7PAY= satisfying all our conditions
and such that 4_*†" (see also Theorem 5.4.28 in [BR2]). It suffices, for example, that the
systems 798Hµ$<;ɵ$YS ”= have the property of return to equilibrium.

5 Thermodynamics of Finite Quantum Systems Cou-


pled to Thermal Reservoirs
In this section we consider a more specific class of models where a finite quantum system
 interacts with several extended reservoirs : which are at thermal equilibrium at differ-
ent inverse temperatures z> . This class of models is a basic paradigm of non-equilibrium
quantum statistical mechanics [JP5, LS, Ru3]. The classical statistical mechanics of such
systems has been developed in [BL, L, LSh].
We define these models and describe their basic properties in Subsection 5.1. The general
features of the scattering approach and spectral analysis of these models are described in
Subsection 5.2. Finally, in Subsection 5.3, we discuss a concrete example.
28

5.1 The Model


The system  is a finite quantum system, as described in 2.3.1. We denote its Hilbert space

by O ˻ 132
and its Hamiltonian by å . The corresponding - dynamical system is 798 $<; '= ª Fª

and we denote its 7G; €$<"ž= -KMS state (i.e., the normalized trace on 8 ) by S Û . ª uª
³
Each reservoir 
 , o!
W
Á
$ɱɱɱ $ , is a quantum dynamical system 7983ž$<;ɵ$YS = , in thermal
equilibrium at inverse temperature z> . Thus S  is a 7G;ɵ$Yz>r= -KMS state. We denote by Ÿø
the generator of ;É .
The combined system T/,±É±É±¾ ÃÂ
is described by the quantum dynamical system
798:$<;%$YS`= where
8 b <ª
8 »ö8»&ös±É±É±”ö8 §$ ŽÂ
and the free (i.e., decoupled) dynamics is given by

;°!#; ª ö—;  ös±É±É±ö‰;  -

We are interested in initial states of the form


S_!yS »ö_SC&ös±É±É±”ö‰S /$  (36)

ª ª ª
where S is a faithful, ; - invariant state on 8 . We denote the set of such product states

by ¶ . If 7PO°©€$<ªÆ©€$X¬©h= is the GNS-representation associated to S]I:¶ , then Oo© and ªÆ© ?ª
do not depend on the choice of S .
The coupling of the finite system  with the reservoir : is specified by a self-adjoint

element ЫßIT8 vö‰83 . Note that Ð> is naturally identified with an element of 8 . We will
use such obvious identification without further comment. The complete interaction is given
by the local perturbation
Â
ò Ш!Îò Ø Ð>ž$
É÷ 
where ò is a coupling constant. We assume that Ð>ÏI\¢R7¤Ÿø= for all , and therefore
{ ³
ЈI§¸¦¾¢R7¤Ÿør= . The generator of the locally perturbed dynamics ; Ñ is given by
Â
Ÿ { Ñ ! Ø Àª
Ÿø*‰‹uÒ å »òhÐ$ ±VÔ9- (37)
É÷ 

{
The goal is to study the NESS of 798:$<; Ñ = associated to the initial states S¡Is¶ . As ª
remarked in Subsection 4.1, the observable describing the heat flux from the 
 into the
system  is
 b
ò Ÿøq79л=!Îò Ÿøq79Ð>=u-
It follows from Equ. (37) that the heat fluxes satisfy the energy balance relation
 { «ª
Ø  N!#Ÿ Ñ 7Gå »0òhл=u$
É÷ 
29

from which we immediately obtain the following expression of the first law of thermody-
namics

Proposition 5.1 For any ;¾Ñ - invariant state ® , one has


Â
Ø ®'79=!y"%-
ø÷ 

³
It is also possible to define the heat flux from the -th reservoir into the system  as the
”change of energy” of  due to the coupling to : . We then obtain the observables
p Ä b { ­ ª = ? ÷ Û !s‹uÒÓòhЫµ$<å ª Ô9$
pžA
; ? Ñ 7Gå

ºº (38)

 Ä
which satisfy the energy balance relation

Ø H!#Ÿ { Ñ 7GåÀª'=u- (39)


É÷ 

Ä
The relation with the previously defined heat flux is given by
{
Ä
 N!ÎN(_ò Ÿ Ñ 79Ð>¥=€‰‹nò ÒÓÐ$uÐ>¾Ô9-
µ Ä µ
(40)
Note that both   and  are [79òh= , while  C(  is o79ò = , up to a total derivative. From
Equ. (39) and (40) we obtain the following result.

Proposition 5.2 For any ;¾Ñ - invariant state ® , one has


 Ä
Ø ®'7 =!y"%-
ø÷ 
Moreover, if ÒÓÐ$uÐ>¾Ô'!#" , then ®'79=!s® 7 = .
Ä
Finally, we relate the heat fluxes to entropy production. For this purpose, it is convenient to
choose the reference state

S Û !yS Û ö—S ös±É±É±¥ö_S T$ Â
—ª
where S Û is the KMS state of  at z‰!)" . Then we have Ÿu© 3!Í( ÆÅ aÇ  z>”Ÿø . One shows
that for any SIo¶ , ª
~…±% ~…±%
W
 „7PST^; Ñ ? …tS`=!# „7PST^; Ñ ? …tS Û = o7 = :x
as AD › , and the following result is immediate.

Proposition 5.3 Assume that SI°¶ ?ª , then for any qIad èÑ 7PS`= one has
Â
Ø z  C
q 79¥=!¨(*C€76 q =u- (41)
É÷ 

6q q
Formula (41) is the basic thermodynamic relation between heat fluxes and entropy produc-
tion. In particular, if  7 =­{" , the NESS carries non-vanishing energy currents.
30

5.2 The Study of NESS


In Section 3.3 we have discussed the scattering and spectral approaches to the study of

ÃÂ
NESS for general quantum systems. In this subsection we describe features of these meth-
ods which are particular to the specific model  —/&s±É±É±”— .

5.2.1 The Scattering Approach

The scattering approach described in Section 3.3.1 has to be slightly modified to accommo-
date for the finite subsystem  . The necessary changes are described in [Ru3]. Setting

8 ¬ b 8»&ös±É±É±”ö8ŽÂ$
;F¬ b ;” ös±É±É±”ö0;wÂ$

S ¬ b S
‘ È
ös±É±É±¥ö‰S $
‘ Ü
the - dynamical system 798¬H$<;FU ¬ = describes
1œ2
˨8 and 8<¬ Û ËÃ8<¬ such that, for FwI‰8 Û and
free non-interacting reservoirs. Assume that
there exists norm-dense subalgebras 8 Û
F\¬ÎI§8¬ ,Û
k i k i
ì
ÒÓÐ$<; ? 7VLÆ[
ª ö—F\¬ =×Ôv>
pžA †]›s$ l i ì
ÒÓÐ$<; Ñ ? 7GFQ=×Ôv>
pžA*†]›s- (42)
l i

The first condition ensures that the limits


| Ñe 7GF\¬= b g aebed i ; Ñ l ? ^­; ? 7VLƪoö—FX¬*=u$
? e
exist in norm and define @ - morphisms | Ñ e p'8<¬ÄD 8 . The second condition ensures that
the strong limits
} eÑ b g aebcd i ; l ? ^­; Ñ ? $
? e
exist and define @ - morphisms Ñe h } p 8)Dë8 . One shows that for FÄIT8 ,
} Ñe 7GFH=!¨L ª ö } Ñeɬ 7GFH=u$
where } Ñe¹¬ !ˆ7 | Ñ e = . Thus } Ñeɬ are in fact @ - isomorphisms. The sets d Ñe 7PS`= consist of
l 
single NESS S~Ñ e and
S~Ñe !sST^ } Ñe !sS ¬ ^ } Ñe¹¬ -
Under some additional assumption one can also construct perturbative expansions of S Ñe
analogous to (19). For details and additional information we refer the reader to [Ru3].
The abstract scattering method is elegant and provides a fairly complete framework for
the study of NESS. However, in concrete models, the verification of Conditions (42) (or
suitable variants of them) is a rather difficult mathematical problem.
31

There exists a number of interesting, essentially exactly solvable models, for which Condi-

studied
9a`
tions (42) (or suitable variants of them) can be verified. One such model is the extensively
spin chain [Ar1, Ar2]. Another one involves a single spin coupled to free
fermionic reservoirs [AM]. It would be interesting to study the NESS of these models in
detail and show that they have strictly positive entropy production.

5.2.2 The Spectral Approach

In this section we take into account the additional structure of the model Ho/”/±É±É±Go ÃÂ
ª ª ª
to elaborate the spectral theory of NESS described in 3.3.2. Let 7PO $<ª $X¬ = be the GNS-
representation of 8 ª Fª uª
associated to the 7G; €$<"ž= -KMS state S Û . The Liouvillean Ç is then ~ª
Rª «ª aª
given by
Ç R!yå ßö#Lv(0LHö å €$
ª ª
and if ‚„Z Š is the spectrum of å , the spectrum of Ç consists of the eigenvalue differences
Ÿ( È
‚„Z (_Z ¥Š . In particular, " is an eigenvalue whose multiplicity is at least equal to the di-
È
Ȼ
mension of O . Let 7PO
ž$<ª ž$X¬ ¥= be the GNS-representation of 83 associated to S  and let
ÇC be the corresponding Liouvillean. We assume that the reservoirs are sufficiently ergodic
so that Ç` has a simple eigenvalue zero, the rest of its spectrum being purely absolutely
continuous and filling the entire real line. The GNS-representation 7PO°©$<ªÆ©€$X¬©h= associ-
ated to a state SI[¶?ª is obtained by taking tensor products of the GNS-representations of
individual subsystems. In particular,

Ȼ
Oo©
!sO »ö‰O€öy±É±É±¥ö‰O /- ?Â
The corresponding Liouvillean is


lj!yÇ »—Ç*€y±É±É±”—Ç T- Â
The eigenvalues of Ç coincide with the eigenvalues of Ç Rª
and the rest of its spectrum is
purely absolutely continuous and covers the real line, see Figure 3.

Figure 3: The spectrum of Ê


The standard Liouvillean for the locally perturbed dynamics ; Ñ is {
Ç { !yǧòhÐÏ(_ò'À ÐoÀÆ-

If the temperatures of the reservoirs are different, then one expects that for small non-
zero ò all the eigenvalues of Ç turn into resonances and that the spectrum of Ç is purely {
absolutely continuous (see Figure 4). This implies that in non-equilibrium situations there
is no S - normal, ; Ñ - invariant state. In fact, by Theorem 3.3, any NESS is purely singular.
32

Figure 4: The spectrum and resonances of Ê { , far from equilibrium


To prove the above spectral results in concrete models one uses either complex deformation
techniques or Mourre theory, see [DJ] for a general description of these methods.
{
If the temperatures of all the reservoirs are equal, " remains an eigenvalue of Ç with
{
{
an eigenvector corresponding to the perturbed KMS state S Ñ . Apart from this simple
eigenvalue, the spectrum of Ç is expected to be purely absolutely continuous, all the other
eigenvalues of Ç turning into complex resonances, see Figure 5. This yields that, in thermal
equilibrium, the combined system enjoys strong ergodic properties and in particular, has the
property of return to equilibrium. See [DJ, JP3] and the next section for application of this
strategy to concrete models.

Figure 5: The spectrum and resonances of Ê { , near equilibrium


In the non-equilibrium case the above argument shows that, for small non-zero coupling,
1
there is no normal NESS. To study singular NESS one uses the - Liouvillean defined by

Ç { i !yÇ ò ÐÄ(_ò À€º Ü Ðœº


S
l Ü
ÀÆ$
S (43)

{
where º and À are the modular operator and conjugation of the unperturbed system. For
sufficiently regular Ð , the operator Ç  i is well-defined and closed. It is not self-adjoint
{
and satisfies Ç  i ¬­©™!" . The expected spectral picture is now more delicate. Roughly,
there is a Banach space M , densely and continuously embedded in O°© , with the following
property. For all yI/M , the matrix elements
l 
79N$„7Gƒv(Ç  i = U=u$ { (44)
d
originally defined for large ‡ ƒ:{s" , have an analytic continuation to the entire half-plane
d
‡ ƒ
{s" , and a meromorphic continuation across the real axis to a second Riemann sheet.

{ {
The only singularity on the real axis is a simple pole at " . The residue at this pole is given by
79N$X¬­©h='7¤¬ $u*= , where ¬
2 2
I§M is a “resonance eigenvector” of Ç  i . For a norm-dense {
Ë 6Ë
subset ÍËs8 such that ªÆ© 7 o=V¬­©TËM , the formula
{
S è 7GFH=!)7¤¬ { $<ªÆ©&7GFH=V¬­©'=u$
defines the unique NESS of the perturbed system. Moreover, ¬ { {
and hence S è will have
convergent expansions in powers of ò .
33

For details and additional information concerning the above heuristic description of spectral
theory of NESS we refer the reader to [JP5]. In the next section we describe a non-trivial
model to which the above spectral approach can be effectively applied.

5.3 An Example
In this section we consider a concrete model ™/'±É±É±ø ÃÂ
, where the reservoirs are
identical free Fermi gases. For concreteness, we assume that the Hilbert space of a single

fermion (in momentum representation) is O !¨Ç 7PŒý«$<pÉþ'= and that its Hamiltonian is the
operator of multiplication by SU7²þ'=!ˆ… þ¦…
Ž ¥j
û . Denote by Æ7 = Ì  b   ? 
2
qq7 =  7 = the field
operator of : . The coupling Ы is given by
y¿3öNÌ %7 } =u$
ЫH!
}
where ßIRO is a form-factor and ¿3Q!,¿  IT8ª . The total perturbation is
2

ò Ш!ÎòNØ Ð>ž$

where ò is a coupling constant.
{
Davies investigated the van Hove weak coupling limit ò
W
" , A y› with A l
To our knowledge, the dynamical system 798:$<; Ñ = was first studied by Davies in [Da1].

ò A! ÃÍ b
8£7 = . This limit yields information in the first non-trivial order of perturbation theory.
The paper [Da1] was followed by a substantial body of literature concerning dynamics of
open quantum systems in the van Hove limit (for references and additional information
see [GFV]). In particular, the non-equilibrium thermodynamics in the van Hove limit have
been studied in detail in [LS], while the linear response theory was developed in [DS].
{
The tools to study the dynamical system 798:$<; Ñ = for finite small ò have been developed
only recently. In the case where all the reservoirs have the same temperature, it has been
shown in [JP3] that the system enjoys strong ergodic properties. In fact, the system studied
in [JP3] differs from the one considered here by the bosonic nature of the reservoir. How-
ever the techniques extend immediately to the fermionic case (see also [BFS, DJ, M] for
additional developments). The non-equilibrium case has been considered in [JP5], using
the spectral approach outlined in the previous section. In the rest of this section, we briefly
summarize the results of [JP5]. For reasons of space, we will not specify here all the tech-
nical conditions we need - the interested reader may consult [JP5] for precise statements
and additional information.
We need two assumptions on the model. The first one is a non-degeneracy condition which
ensures the uniqueness of the NESS.

«ª ¿ ¿Â
(ND) The commutant ‚„å €$ »¾$É-É-É-¾$ TŠ¾ in 8 <ª
is trivial, namely consists
only of multiples of the identity. Moreover
k i

l i
l
¼ È ?GçÏÎkÐ Î
l é
1Ì } 1Ì }
Ý S µ7 Æ7 ¥=n;  ? 7 €Æ7 ¥=YY= =qpžA3!y
î "%$ (45)
34

\" ǻ
for Z œ$<Z Ù IRï7Gå = .

Note that if
9 Iy‚„åÀª$ß¿ ¾$É-É-É-ì$¿ÂŠ¾ , then ;%Ñ ? 7 9
= !
œ
9
. Thus, the non-triviality of the
above commutant would lead to an artificial multiplicity of NESS. Relations (45) ensure
that the reservoirs induces transitions between eigenstates of the small system.

}
The second assumption we need is of a more technical nature. It requires the form-factors
 to be analytic in a suitable sense. This condition allows us to use a complex deformation
technique to investigate the analytic structure of the resolvent (44). It is possible that some
of the technical developments in [BFS, DJ] can be used to relax this condition.
Fix an initial state SˆI™¶ ?ª
and assume that not all z> are the same. The main results of
[JP5] are summarized as follows: there exists a constant y{" such that, for "o†Ä… ò&…%†G, NG
the following holds:

d{
(1) è Ñ 7PS`= consists of a unique purely singular NESS S è Ñ . Moreover, for all {
®£I°¶ © and FÃIT8 ,
aebed {
g3i ®œ^­; ? Ñ 7GFH=!sS è Ñ 7GFH=u- { (46)
?

|
(2) The limit (46) is exponentially fast in the following sense. There exist
79òh=
{w" , a norm-dense set of states ¶ Û Ë)¶
© and a norm-dense @ - subal-
gebra 8 Û Ë]8 such that, for ®
I°¶ Û and FÄI§8 Û ,
{ { 1 l é { @Ñ {
| ºº |
®œ^­; ? Ñ 7GFH=¦(™S è Ñ 7GFQ= +
ºº
~ Å  ¼ ç ? -
|
(47)

Moreover, 79òh=!
| Ò
Û ò /8£79ò ì= as ò£DE" , where Û {" is a computable con-

Ä {
stant. In fact, ( 79òh= is equal to the imaginary part of the non-zero resonance
of the operator Ç  i closest to the real axis.
(3)  ž$ ž$<ï Ñ b ÆÅ
Ÿu© ¥79Ðß=IT8 Û . Hence, Proposition 4.6 applies and gives
{
€7PS è Ñ ={"%-

(4) There exist operators Ó ª —p ª <ª


  8 /D 8 , completely determined by the sec-
{
ond order perturbation theory (Fermi Golden Rule) of the resonances of Ç  i
ÓÀª b Ç Ó ª
   is precisely the generator of the Markovian dy-
ª ª
and such that
namics in the van Hove limit. Namely, for any F I‰8
uª =Â S ?ª S
and any initial state
S‰!sS »ö‰S ös±É±É±”ö—S ¡Io¶
{ {
we have

{ aebeg d Û ST^; l ?c• ^ ; { ? •Ñ 7GF\ª[öyLƬ=!ySuªz‹É¼ l ?!Ô=Õ FXªs»-


(5) For FëI¨8 Û , the function òÎBD S { è Ñ 7GFH= is analytic for … ò …U†ŒG . More
precisely, there exist linear functionals S  è h p 8 Û D â such that, for F¨IT8 Û ,
S { è Ñ 7GFH=C!ÎØ

ò S  è 7GFQ=u- (48)
Ú Û
35

The first term S Û è belongs to ¶ ?ª and hence has the form


S Ûè !ySuª  1Ö ö‰SC€öy±É±É±”ö‰S=ÂT- (49)
—ª UÖ Fª
The state S  is ; - invariant. It is a solution of the equation
2
Ó ª uª 1Ö
7PS  =!#" .

—ª UÖ
The non-degeneracy condition (ND) ensures that scalar multiples of the density
matrix S  are the only solution of this equation. Although the formulas
for the higher order terms in Equ. (48) become quickly very complicated, in
principle it is possible to compute all terms of this expansion.
1
The proofs of (1)-(5) are based on the spectral analysis of the - Liouvillean Ç  i and {
follow the strategy described in Section 3.3.2. The necessary modular structures needed
1
to compute the - Liouvillean are described in 2.3.1 and 2.3.2. The technical analysis is

×ÔØ
based on the complex deformation technique previously developed in [JP2, JP3]. Condition
l
(ND) is related to the ergodic properties of the Markovian semigroup ¼ ? (see [Fr1, Fr2,
GFV, Sp2]).

{
The results (3)-(5) allow to compute heat fluxes and entropy production perturbatively.
Since S Û è is invariant under the unperturbed dynamics, we have S Û è 7Gï Ñ =`!yS Û è 79=!#" .
Hence,
{
Å Å
C 7PS è =!Îò S è 7¤Ÿu© 79л=Y=ò ý S è 7¤Ÿu© 79л=Y=s±É±É±$

{
S è 79=!Îò S è 7¤Ÿø%79л=Y=ò ý S è 7¤Ÿø%79л=Y=s±É±É±-
(50)


Of particular importance are the ò -contributions (we will call them the Fermi Golden Rule
terms). They can be used to give a perturbative proof of the strict positivity of the entropy
production, as we will describe below. With
C
!sS è 7¤Ÿ © ÆÅ 79Ðß=Y=u$
H!sS è 7¤Ÿø%79Ðß=Y=u$
we can write

!¨( Ø z> r-

From Equ. (38),(40) it easily follows that
H!sS —ª  1Ö 7Ó ª  %7G嫪h=Y=u-
The operators Óª   can be explicitly computed and a somewhat long computation shows
that, as long as the z> are not all identical, one has
/{"%-
This gives another proof of the strict positivity of the entropy production for small ò , which
has the advantage of providing a concrete estimate.
Some of the principles of phenomenological thermodynamics hold for the state S  . Of —ª UÖ
particular interest are the Onsager reciprocity relations. For the proof of these relations and
additional discussion we refer the reader to [LS].
36

References
[AM] Aizenstadt, V.V., Malyshev, V.A.: Spin interaction with an ideal Fermi gas. J. Stat.
Phys. 48, 51 (1987).
[Ar1] Araki, H.: On the
9a`
-model on two-sided infinite chain. Publ. Res. Inst. Math.
Sci. Kyoto Univ. 20, 277 (1984).
[Ar2] Araki, H.: Dynamic and ergodic properties of the
9a`-model, in Critical phenom-
ena (Brasov, 1983), Progr. Phys. 11, 287, Birkhäuser, Boston, MA, (1985).
[Ar3] Araki, H.: Relative entropy of states of von Neumann algebras. Publ. Res. Inst.
Math. Sci. Kyoto Univ. 11, 809, (1975/76).
[Ar4] Araki, H.: Relative entropy for states of von Neumann algebras. II. Publ. Res. Inst.
Math. Sci. Kyoto Univ. 13, 173, (1977/78).
[Ar5] Araki, H.: Relative Hamiltonian for faithful normal states of a von Neumann alge-
bra. Publ. Res. Inst. Math. Sci. Kyoto Univ. 9, 165 (1973).
[AW] Araki, H., Wyss, W.: Representations of canonical anti-commutation relations.
Helv. Phys. Acta 37, 136 (1964).
[BKR] Bratteli, O., Kishimoto, A., Robinson, D.W.: Stability properties and the KMS
condition. Commun. Math. Phys. 61, 209 (1978).
[BM] Botvich, D.D., Malyshev, V.A.: Unitary equivalence of temperature dynamics for
ideal and locally perturbed Fermi gas. Commun. Math. Phys. 91, 301 (1983).
[BL] Bargmann, P.G., Lebowitz, J.L.: New approach to nonequilibrium process. Phys.
Rev. 99, 578 (1955).
[BFS] Bach, V., Fröhlich, J., Sigal, I.: Return to equilibrium. J. Math. Phys. 41, 3985
(2000).
[BR1] Bratteli, O., Robinson D.W.: Operator Algebras and Quantum Statistical Mechan-
ics 1. Springer-Verlag, Berlin, (1987).
[BR2] Bratteli, O., Robinson D.W.: Operator Algebras and Quantum Statistical Mechan-
ics 2. Springer-Verlag, Berlin, (1996).
[Da1] Davies, E.B.: Markovian master equations. Commun. Math. Phys. 39, 91 (1974).
[DS] Davies, E.B., Spohn, H.: Open quantum systems with time-dependent Hamiltonians
and their linear response. J. Stat. Phys. 19, 511 (1978).
[DJ] Derezinski, J., Jakšić, V.: Spectral theory of Pauli-Fierz operators. J. Func. Anal.
180, 243 (2001).
2
[DJP] Derezinski, J., Jakšić, V., Pillet, C.-A.: Perturbation theory of 4 - dynamics, Liou-
villeans and KMS states. Submitted.
[E] Eckmann, J.-P.: Resonances in dynamical systems. Proceedings of the IXth Inter-
national Congress of Mathematical Physics (Swansea 1988). Hilger, Bristol (1989).
37

[Fr1] Frigerio, A.: Quantum dynamical semigroups and approach to equilibrium. Lett.
Math. Phys. 2, 79 (1977).
[Fr2] Frigerio, A.: Stationary states of quantum dynamical semigroups. Commun. Math.
Phys. 63, 269 (1978).
[GFV] Gorini, V., Frigerio A., Verri, M., Kossakowski, A., Sudarshan, E.C.G.: Properties
of quantum Markovian master equations. Rep. Math. Phys. 13, 149 (1978).
[Ha] Haag, R.: Local Quantum Physics. Springer-Verlag, Berlin (1993).
[HKTP] Haag, R., Kastler, D., Trych-Pohlmeyer E.: Stability and equilibrium states. Com-
mun. Math. Phys. 38, 213 (1974).
[HTP] Haag, R., Trych-Pohlmeyer E.: Stability properties of equilibrium states. Commun.
Math. Phys. 56, 273 (1977).
[Hu] Hugenholz, N.M.: Derivation of the Boltzmann equation for a Fermi gas. J. Stat.
Phys. 32, 231 (1983).
[I] Israel, R. B.: Convexity in the Theory of Lattice Gases. Princeton University Press,
Princeton, N.J., (1979).
[JP1] Jakšić, V., Pillet, C.-A.: Spectral theory of thermal relaxation. J. Math. Phys. 38,
1757 (1997).
[JP2] Jakšić, V., Pillet, C.-A.: On a model for quantum friction II. Fermi’s golden rule
and dynamics at positive temperature. Commun. Math. Phys. 176, 619 (1996).
[JP3] Jakšić, V., Pillet, C.-A.: On a model for quantum friction III. Ergodic properties of
the spin-boson system. Commun. Math. Phys. 178, 627 (1996).
[JP4] Jakšić, V., Pillet, C.-A.: On entropy production in quantum statistical mechanics.
Commun. Math. Phys. 217, 285 (2001).
[JP5] Jakšić, V., Pillet, C.-A.: Non-equilibrium steady states for finite quantum systems
coupled to thermal reservoirs. Commun. Math. Phys. 226, 131 (2002).
[JP6] Jakšić, V., Pillet, C.-A.: In preparation.
[KR] Kadison, R.V., Ringrose, J.R. Fundamentals of the Theory of Operator Algebras.
Vol. II. Advanced theory. Academic Press, Orlando, FL, (1986).
[L] Lebowitz, J.L.: Stationary nonequilibrium Gibbsian ensembles. Phys. Rev. 114,
1192 (1959).
[LS] Lebowitz, J.L., Spohn, S.: Irreversible thermodynamics for quantum systems
weakly coupled to thermal reservoirs. Adv. Chem. Phys. 38, 109 (1978).
[LSh] Lebowitz, J.L., Shimony, A.: Statistical mechanics of open systems. Phys. Rev. 128,
1945 (1962).
[M] Merkli, M.: Positive commutators in non-equilibrium quantum statistical mechan-
ics. Commun. Math. Phys. 223, 327 (2001).
[NT] Narnhofer, H., Thirring, W.: On the adiabatic theorem in quantum statistical me-
chanics. Phys. Rev. A 26, 3646 (1982).
38

[OHI] Ojima, I., Hasegawa, H., Ichiyanagi, M.: Entropy production and its positivity in
nonlinear response theory of quantum dynamical systems. J. Stat. Phys. 50, 633
(1988).
[O1] Ojima, I.: Entropy production and non-equilibrium stationarity in quantum dynam-
ical systems: physical meaning of van Hove limit. J. Stat. Phys. 56, 203 (1989).
[O2] Ojima, I: Entropy production and non-equilibrium stationarity in quantum dynami-
cal systems, in Proceedings of International Workshop on Quantum Aspects of Op-
tical Communications. Lecture Notes in Physics 378, 164, Springer-Verlag, Berlin
(1991).
[OP] Ohya, M., Petz, D.: Quantum Entropy and its Use. Springer-Verlag, Berlin (1993).
[PW] Pusz, W., Woronowicz, S.L.: Passive states and KMS states for general quantum
systems. Commun. Math. Phys. 58, 273 (1978).
[Ru1] Ruelle, D.: Statistical Mechanics. Rigorous Results. Benjamin, New York, (1969).
[Ru2] Ruelle, D.: Smooth dynamics and new theoretical ideas in nonequilibrium statistical
mechanics. J. Stat. Phys. 95, 393 (1999).
[Ru3] Ruelle, D.: Natural nonequilibrium states in quantum statistical mechanics. J. Stat.
Phys. 98, 57 (2000).
[Ru4] Ruelle, D.: Entropy production in quantum spin systems. Commun. Math. Phys.
224, 3 (2001).
[Ru5] Ruelle, D.: How should one define entropy production for nonequilibrium quantum
spin systems? Preprint (2001), mp-arc 01-258.
[Ru6] Ruelle, D.: Topics in quantum statistical mechanics and operator algebras. Preprint
(2001), mp-arc 01-257.
[Ru7] Ruelle, D.: One-dimensional Gibbs states and Axiom A diffeomorphisms. J. Diff.
Geom. 25, 117 (1987).
[Ru8] Ruelle, D.: Resonances for axiom A flows. Axiom A diffeomorphisms. J. Diff.
Geom. 25, 99 (1987).
[Ru9] Ruelle, D.: Resonances of chaotic dynamical systems. Phys. Rev. Lett. 56, 405
(1986).
[Si1] Simon, B.: Statistical Mechanics of Lattice Gases. Princeton University Press,
Princeton, N.J., (1993).
[Sp1] Spohn, H.: Entropy production for quantum dynamical semigroups. J. Math. Phys.
19, 227 (1978).
[Sp2] Spohn, H.: An algebraic condition for the approach to equilibrium of an open ) -
level system. Lett. Math. Phys 2, 33 (1977/78).
[Ta] Takesaki, M.: Theory of Operator Algebras I. Springer-Verlag, New York (1979).
[Th] Thirring, W.: Quantum Mechanics of Large Systems. Springer-Verlag, Wien (1980).

Anda mungkin juga menyukai