Micro-Macro Duality in Quantum Physics∗

Dedicated to Professor Takeyuki Hida

on the occasion of his 77th birthday

arXiv:math-ph/0502038v1 14 Feb 2005

Izumi Ojima
RIMS, Kyoto University, Kyoto, Japan

Abstract
Micro-Macro Duality means here the universal mutual relations
between the microscopic quantum world and various macroscopic clas-
sical levels, which can be formulated mathematically as categorical ad-
junctions. It underlies a unified scheme for generalized sectors based
upon selection criteria proposed by myself in 2003 to control differ-
ent branches of physics from a unified viewpoint, which has played
essential roles in extending the Doplicher-Haag-Roberts superselection
theory to various situations with spontaneously as well as explicitly
broken symmetries.
Along this line of thought, the state correlations between a system
and a measuring apparatus necessary for measurements can canoni-
cally be formulated within the context of group duality; the obtained
measurement scheme is not restricted to the quantum mechanical sit-
uations with finite number of particles but can safely be applied to
quantum field theory with infinite degrees of freedom whose local sub-
algebras are given by type III von Neumann algebras.

1 Why & what is Micro-Macro Duality?

– Vital roles played by Macro –
In spite of their ubiquitous (but implicit) relevance to quantum theory,
the importance of macroscopic classical levels is forgotten in current
trends of microscopic quantum physics (owing to the overwhelming belief in
the ultimate unification at the Planck scale?). Without those levels, how-
ever, neither measurement processes nor theoretical descriptions of micro-
scopic quantum world would be possible! For instance, a state ω : A → C
as one of the basic ingredients of quantum theory is nothing but a micro-
macro interface assigning macroscopically measurable expectation value
∗
Invited talk at International Conference on Stochastic Analysis, Classical and Quan-
tum held in Nagoya in November 2004

1
ω(A) to each microscopic quantum observable A ∈ A. Also physical inter-
pretations of quantum phenomena are impossible without vocabularies (e.g.,
spacetime x, energy-momentum p, mass m, charge q, particle numbers n;
entropy S, temperature T , etc., etc.), whose communicative powers rely on
their close relationship with macroscopic classical levels of nature.
– Universality of Macro due to Micro-Macro duality –
Then one is interested in the question as to why and how macroscopic
levels play such essential roles: the answer is found in the universality of
“Macro” in the form of universal connections of a special Macro with generic
Micro’s. To equip this notion with a precise mathematical formulation we
F
introduce the notion of a categorical adjunction Q⇄C which controls the
E
mutual relations between [unknown generic objects Q (: microscopic quan-
tum side) to be described, classified and interpreted] and [special familiar
model C (: macroscopic classical side) for describing, classifying and inter-
preting], related by a pair of functors E(: c→q) and F (: q→c), mutually
η ε
inverse up to homotopy I → EF , F E → I, via a natural isomorphism:

εa F (·)
Q(ω, E(a)) ⇄ C(F (ω), a) ,
E(·)ηω

so that

an ‘equation’ E(a) ∼ ω in Q to compare an unknown object ω with

controlled ones E(a) specified by known parameters a in C can be
‘solved’ to give a solution a ∼ F (ω) which allows ω to be interpreted
in the vocabulary a in C in the context and up to the accuracy specified,
respectively, by (E, F ) and (η, ε).

Abstract mathematical essence of “Micro-Macro Duality” can be seen in

this notion of adjunction, whose concrete meanings are seen in the following
discussion. What to be emphasized before going into details is the vast
freedom in the choices of categories Q, C and functors E, F which are not
to be fixed but adjusted and modified flexibly so that our descriptions are
adapted to each focused context of given physical situations and to the
aspects to be examined. This point should be contrasted to the rigidity
inherent to the ultimate “Theory of Everything”. The simplest example of
duality is given by the Gel’fand isomorphism,

CommC ∗ Alg(A, C0 (M )) ≃ HausSp(M, Spec(A)), (1)

between a commutative C*-algebra and a Hausdorff space defined by [ϕ∗ (x)](A) :=

ϕ ϕ∗
[ϕ(A)](x) for [A → C0 (M )] ⇄ [M → Spec(A) = {χ : A → C ; χ: character

2
s.t. χ(AB) = χ(A)χ(B)}] and for A ∈ A, x ∈ M . Through our discus-
sion on the Micro-Macro duality below, we will encounter various kinds of
fundamental adjunctions appearing in quantum physics as follows:
1) Basic duality between algebras/ groups and states / representations
“Micro-Macro Duality” underlies “a unified scheme for generalized sec-
tors based upon selection criteria” [14] proposed by myself in 2003 to control
various branches of physics from a unified viewpoint. Extracted from a new
general formulation of local thermal states in relativistic QFT (Buchholz,
IO and Roos [2]), this scheme has played essential roles in my recent work to
extend the Doplicher-Haag-Roberts superselection theory [5, 6] to recover a
field algebra F and its (global) gauge group G from the G-invariant observ-
able algebra A = FG and its selected family of states, according to which its
range of applicability restricted to unbroken symmetries has been extended
to not only spontaneously but also explicitly broken symmetries [15].
2) Adjunction as a selection criterion to select states of physical relevance
to a specific physical situation, which ensures at the same time the phys-
ical interpretations of selected states. This is just the core of the present
approach to Micro-Macro Duality between microscopic quantum and
macroscopic classical worlds formulated mathematically by categorical ad-
junctions:
q−c
(generic:) Micro ⇄ Macro (: special model space with universality),
c−q

where c → q (q → c) means a c → q (q → c) channel to transform classical

states into quantum ones (vice versa).
3) Symmetry breaking patterns constituting such a hierarchy as unbro-
ken / spontaneously broken / explicitly broken symmetries: the adjunction
relevant here describes and controls the relation between [broken ⇄ un-
broken], playing essential roles in formulating the criterion for symmetry
breakings in terms of order parameters. Through a Galois extension, an
augmented algebra can be defined as a composite system consising of the
object physical system and of its macroscopic environments including ex-
ternalized breaking terms, where broken symmetries are “recovered” and
the couplings with external fields responsible for symmetry breaking are
naturally described.
4) If we succeed in extrapolating this line of thoughts to attain an adjunc-
homotopical dilation
tion between [irreversible historical process] ⇄ [stabilized hi-
erarchical domains with reversible dynamics] through enough controls over
mutual connections among different physical theories describing different
domains of nature, we would be able to envisage a perspective towards a

3
theoretical framework to describe the historical process of the cosmic evolu-
tion.

2 Basic scheme for Micro/Macro correspondence

2.1 Definition of sectors and order parameters
In the absence of an intrinsic length scale to separate quantum and classical
domains, the distinctions between Micro and Macro and between quantum
and classical are to some extent ‘independent’ of each other, admitting such
interesting phenomena as “macroscopic quantum effects”. Since this kind
of “mixtures” can be taken as ‘exceptional’, however, we put in parallel
micro//quantum//non-commutative and macro//classical//commutative, re-
spectively, in generic situations. The essence of Micro/Macro correspon-
dence is then seen in the fundamental duality between non-commutative
algebras of quantum observables and their states, where the latter transmit
the microscopic data encoded in the former at invisible quantum levels into
the visible macroscopic form. While the relevance of duality is evident from
such prevailing opposite directions as between maps ϕ : A1 → A2 of algebras
and their dual maps of states, ϕ∗ : EA2 ∋ ω 7−→ ϕ∗ (ω) = ω ◦ ϕ ∈ EA1 , their
relation cannot, however, be expressed in such a simple clear-cut form as
the Gel’fand isomorphism Eq.(1) valid for commutative algebras, because of
the difficulty in recovering algebras on the micro side from the macro data
of states. The essence of the following discussion consists, in a sense, in the
efforts of circumventing this obstacle for recovering Micro from Macro.
Starting from a given C*-algebra A of observables describing a Micro
quantum system, we find, as a useful mediator between algebras and states,
the category RepA of representations π = (π, Hπ ) of A with intertwiners
T , T π1 (A) = π2 (A)T (∀A ∈ A), as arrows ∈ RepA(π1 , π2 ), which is nicely
connected with the state space EA of A via the GNS construction: ω ∈
1:1 up to
EA ←→ (πω , Hω ) ∈ RepA with Ωω ∈ Hω s.t. ω(A) = hΩω | πω (A)Ωω i
unitary equiv.
(∀A ∈ A) and πω (A)Ωω = Hω . Two representations π1 , π2 without (non-
p
zero) connecting arrows are said to be disjoint and denoted by π1 ◦ π2 , i.e.,
p def
π1 ◦ π2 ⇐⇒ RepA(π1 , π2 ) = {0}. The opposite situation to disjointness
can be found in the definition of quasi-equivalence, π1 ≈ π2 , which can be
simplified into

π1 ≈ π2 (: unitary equivalence up to multiplicity)

⇐⇒ π1 (A)′′ ≃ π2 (A)′′ ⇐⇒ c(π1 ) = c(π2 ) ⇐⇒ W ∗ (π1 )∗ = W ∗ (π2 )∗ .

To explain the central support c(π) of a representation π, we introduce the

universal enveloping W*-algebra A∗∗ ≃ πu (A)′′ := W ∗ (A) of C*-algebra A
which contains all (cyclic) representations of A as W*-subalgebras W ∗ (π) :=

4
π(A)′′ ⊂ W ∗ (A). In the universal Hilbert space Hu := ⊕ω∈EA Hω , W ∗ (A) and
W ∗ (π) are realized, respectively, by the universal representation (πu , Hu ),
πu := ⊕ω∈EA πω , and by its subrepresentations π(A) := P (π)πu (A) ↾P (π)
(∀A ∈ A) in Hπ = P (π)Hu with P (π) ∈ W ∗ (A)′ . W ∗ (A) is characterized by
universality via adjunction,

W ∗ Alg(W ∗ (A), M) ≃ C ∗ Alg(A, E(M)),

between categories C ∗ Alg, W ∗ Alg of C*- and W*-algebras (with forgetful

functor E to treat M as C*-algebra E(M) forgetting its W*-structure due
ηA
to the predual M∗ ) with a canonical embedding map A ֒→ E(W ∗ (A)), so
that any C*-homomorphism ∀ϕ : A → E(M) is factored ϕ = E(ψ) ◦ ηA
through ηA with a uniquely existing W*-homomorphism ψ : W ∗ (A) → M:

A
ηA
↓  ց∀ϕ .
E(W ∗ (A)) 99K E(M)
E(ψ)

In this situation, the central support c(π) of the representation π is defined

by the minimal central projection majorizing P (π) in the centre Z(W ∗ (A)) :=
W ∗ (A) ∩ W ∗ (A)′ of W ∗ (A).

i) Basic scheme for Micro-Macro correspondence in terms of sec-

tors and order parameters: The Gel’fand spectrum Spec(Z(W ∗ (A))) of
the centre Z(W ∗ (A)) := W ∗ (A) ∩ W ∗ (A)′ can be identified with a factor
⌢
spectrum A of A:
⌢
Spec(Z(W ∗ (A))) ≃ A := FA/ ≈: factor spectrum,

defined by all quasi-equivalence classes of factor states ω ∈ FA (with trivial

centres Z(W ∗ (πω )) := W ∗ (πω ) ∩ W ∗ (πω )′ = C1Hω in the GNS representa-
tions (πω , Hω )).

Definition 1 A sector of observable algebra A is defined by a quasi-

equivalence class of factor states of A.

In view of the commutativity of Z(W ∗ (A)) and of the role of its spectrum,
we can regard
⌢
• Spec(Z(W ∗ (A))) ≃ A as the classifying space of sectors to distin-
guish among different sectors, and

• Z(W ∗ (A)) as the algebra of macroscopic order parameters to spec-

ify sectors.

5
Then the map
⌢ ⌢
Micro: A∗ ⊃ EA ։ P rob( A) ⊂ L∞ (A)∗ : Macro,
⌢
defined as the dual of embedding Z(W ∗ (A)) ≃ L∞ ( A) ֒→ W ∗ (A), can be
interpreted as a universal q→c channel, transforming microscopic quantum
⌢
states ∈ EA to macroscopic classical states ∈ P rob( A) identified with prob-
abilities. This basic q → c channel,
⌢
EA ∋ ω 7−→ µω = ω ′′ ↾Z(W ∗ (A)) ∈ EZ(W ∗ (A)) = M 1 (Spec(Z(W ∗ (A)))) = P rob( A) ,

describes the probability distribution µω of sectors contained in the central

decomposition of a state ω of A:
⌢
A ⊃ ∆ 7−→ ω ′′ (χ∆ ) = µω (∆) = P rob(sector ∈ ∆ | ω),

where ω ′′ denotes the normal extension of ω ∈ EA to W ∗ (A). While it tells

us as to which sectors appear in ω, it cannot specify as to precisely which
representative factor state appears within each sector component of ω.

ii) [MASA] To detect this intrasectorial data, we need to choose a max-

imal abelian subalgebra (MASA) N of a factor M, defined by the condition
N′ ∩ M = N ∼ = L∞ (Spec(N)). Using a tensor product M ⊗ N (acting on the
Hilbert-space tensor product c(π)Hu ⊗ L2 (Spec(N))) with a centre given by

Z(M ⊗ N) = Z(M) ⊗ N = 1 ⊗ L∞ (Spec(N)),

we find a conditional sector structure described by spectrum Spec(N) of a

chosen MASA N.

iii) [Measurement scheme as group duality] Since the W*-algebra N

is generated by its unitary elements U (N), the composite algebra M⊗ N can
be seen in the context of a certain group action which can be related with a
coupling of M with the probe system N as seen in my simplified version [14]
of Ozawa’s measurement scheme [17]. To be more explicit, a reformulation in
terms of a multiplicative unitary [1] can exhibit the universal essence of the
problem. In the context of a Hopf-von Neumann algebra M (⊂ B(H)) [7] with
a coproduct Γ : M → M ⊗M , a multiplicative unitary V ∈ U ((M ⊗M∗ )− ) ⊂
U (H ⊗ H) implementing Γ, Γ(x) = V ∗ (1 ⊗ x)V , is characterized by the
pentagonal relation, V12 V13 V23 = V23 V12 , on H ⊗ H ⊗ H, expressing the
coassociativity of Γ, where subscripts i, j of Vij indicate the places in H⊗H⊗
H on which the operator V acts. It plays fundamental roles as an intertwiner,
V (λ ⊗ ι) = (λ ⊗ λ)V , showing the quasi-equivalence between tensor powers
of the regular representation λ : M∗ ∋ ω 7−→ λ(ω) := (i ⊗ ω)(V ) ∈ M̂ , a
generalized Fourier transform, λ(ω1 ∗ ω2 ) = λ(ω1 )λ(ω2 ), of the convolution

6
algebra M∗ , ω1 ∗ ω2 := ω1 ⊗ ω2 ◦ Γ. On these bases the duality for Kac
algebras as a generalization of group duality can be formulated. In the case
of M = L∞ (G, dg) with a locally compact group G with the Haar measure
dg, the multiplicative unitary V is explicitly specified on L2 (G × G) by

(V ξ)(s, t) := ξ(s, s−1 t) for ξ ∈ L2 (G × G), s, t ∈ G, (2)

or symbolically in the Dirac-type notation,

V |s, ti = |s, sti. (3)

Identifying M with the Hopf-von Neumann algebra L∞ (G) = N correspond-

ing to G := \U (N) given by the character group of our abelian group U (N)
(assumed to be locally compact), we adapt this machinery to the present
context of the MASA N, by considering a crossed product M ⋊α G :=
[C ⊗ λ(G)′′ ] ∨ α(M) [9] defined as the von Neumann algebra generated
by C ⊗ λ(G)′′ = C ⊗ N b and by the image α(M) of M under an iso-
morphism α of M into M ⊗ L∞ (G) ≃ L∞ (G, M) ≃ M ⊗ N given by
[α(B)](γ) := Adγ (B) = φγ Bφ∗γ , γ ∈ G, B ∈ M where φγ is an action
of γ ∈ G on L∞ (M). By definition, M ⋊ι G = M ⊗ N b is evident for the
trivial G-action ι with ι(M) = M.
R The crossed product M⋊α G is generated
by the representation φ(V ) = G dE(γ) ⊗ λγ of V on L2 (M) ⊗ L2 (G) with
the spectral measure E(∆) = E(χ∆ ) of N (for Borel sets ∆ in Spec(N))
defined by the embedding homomorphism E : N ∼ = L∞ (G) ֒→ M of N into
∗ b
R as seen from (ω ⊗ i)(φ(V )) = λ(E ω) ∈ C ⊗ N and (i ⊗ Ω)(φ(V )) =
M,
G dE(γ)Ω(λγ ) ∈ α(M). The action of φ(V ) corresponding to Eq.(3) can be
expressed by

φ(V )(ξγ ⊗ |χi) = ξγ ⊗ |γχi for γ, χ ∈ G, (4)

satisfying the modified version of the pentagonal relation, φ(V )12 φ(V )13 V23 =
∗ = φ(V ) φ(V ) . Under the as-
V23 φ(V )12 , or equivalently, V23 φ(V )12 V23 12 13
sumption that U (N) is locally compact, the spectral measure E consti-
tutes an imprimitivity system, φγ (EA (∆))φ∗γ = EA (γ∆), w.r.t. a represen-
tations φ of G on L2 (M), from which the following intertwining relation
follows: φ(V )(φγ ⊗ I) = (φγ ⊗ λγ )φ(V ), for γ ∈ G. While the role of
a multiplicative unitary is to put an arbitrary representation ρ in quasi-
equivalence relation ≈ with the regular representation λ by tensoring with
λ: ρ ⊗ λ ∼= Uρ (ι ⊗ λ)Uρ∗ ≈ λ, the above relation allows us to proceed further
to
φ ≈ φ(V )(φ ⊗ ι)φ(V )∗ = φ ⊗ λ ∼ = Uφ (ι ⊗ λ)Uφ∗ ≈ λ.
The important operational meaning of Eq.(4) can clearly be seen in the
case where G is a discrete group which is equivalent to the compactness of the
group U (N) in its norm topology (or, the almost periodicity of functions on

7
it). In the present context of group duality with G as an abelian group gen-
erated by Spec(N), the unit element ι ∈ G naturally enters to describe the
neutral position of measuring pointer in addition to Spec(N), in contrast
to the usual approach to measurements. Then Eq.(4) is seen just to create
the required correlation (“perfect correlation” due to Ozawa [18]) between
the states ξγ of microscopic system M to be observed and that |γi of the
measuring probe system N coupled to the former: P φ(V )(ξγ ⊗ |ιi) = ξγ ⊗ |γi
for ∀γ ∈ G. Applying it to a generic state1 ξ = γ∈G cγ ξγ of M, an initial
uncorrelated state ξ ⊗ |ιi is transformed by φ(V ) to a correlated one:
X
φ(V )(ξ ⊗ |ιi) = cγ ξγ ⊗ |γi.
γ∈G

The created perfect correlation establishes a one-to-one correspondence be-

tween the state ξγ of the system M and the measured data γ on the pointer,
which would not hold without the maximality of N as an abelian subalge-
bra of M. On these bases, we can define the notion of an instrument I
unifying all the ingredients relevant to a measurement as follows:

I(∆|ωξ )(B):=(ωξ ⊗ | ιih ι|)(φ(V )∗ (B ⊗ χ∆ )φ(V ))

= (h ξ| ⊗ h ι|)φ(V )∗ (B ⊗ χ∆ )φ(V )(|ξi ⊗ | ιi).

In the situation with a state ωξ = h ξ| (−)ξi of M as an initial state of the

system, the instrument describes simultaneously the probability p(∆|ωξ ) =
I(∆|ωξ )(1) for measured values of observables in N to be found in a Borel
set ∆ and the final state I(∆|ωξ )/p(∆|ωξ ) realized through the detection
of measured values [17]. While this measurement scheme of Ozawa’s is for-
mulated originally in quantum-mechanical contexts with finite degrees of
freedom where M is restricted to type I, its applicability to general situ-
ations without such restrictions is now clear from the above formulation
which applies equally to non-type I algebras describing such general quan-
tum systems with infinite degrees of freedom as QFT. Since instruments
do not exclude “generalized observables” described by “positive operator-
valued measures (POM)”, it may be interesting to examine the possibility
to replace the spectral measure dE(γ) with such a POM as corresponding to
a non-homomorphic completely positive map for embedding a commutative
subalgebra N into M.
In what follows, the above new formulation will be seen to provide a pro-
totype of more general situations found in various contexts involving sectors,
such as Galois-Fourier duality in the DHR sector theory and its extension
to broken symmetries with augmented algebras (see below). It is impor-
tant there to control such couplings between Micro (M) and Macro (N as
1
Note that any normal state of M in the standard form can be expressed as a vectorial
state without loss of generality.

8
measuring apparatus) as φ(V ) ∈ M ⋊ G, whose Lie generators in infinitesi-
mal version consist of Ai ∈ M and their “conjugate” variables to transform
G ∋ χ 7−→ γi χ ∈ G. This remarkable feature exhibited already in von Neu-
mann’s measurement model, is related with a Heisenberg group as a central
extension of an abelian group with its dual and is found universally in such
a form as Onsager’s dissipation functions, (currents)×(external forces), as
a linearized version of general entropy production [11], etc. To be precise,
what is described here is the state-changing processes caused by this type
of interaction terms φ(V ) between the observed system M and the probing
external system N, with the intrinsic (=“unperturbed”) dynamics of the for-
mer being neglected. While the validity of this approximation is widely taken
for granted (especially in the context of measurement theory), the problem
as to how to justify it seems to be a conceptually interesting and important
issue which will be discussed elsewhere.

iv) [Central measure as a c→q channel] Here we note that, from

the spectral measure in iii), a central measure µ is defined R ⊕ and achieves a
central decomposition of M ⊗ N = L∞ (Spec(N), M) = Spec(N) Ma dµ(a),
where µ(∆) := ω0 (E(∆)) with ω0 a state of N supported by Spec(N) being
faithful to ensure the equivalence µ(∆) = 0 ⇐⇒ E(∆) = 0.
A central measure µ is characterized as a special case of orthogonal mea-
sures by the following relations according to a general theorem due to Tomita
(see [3] Theorem 4.1.25): for a state ω ∈ EA of a unital C*-algebra A there is
a 1-1 correspondence between the followinghR three items, 1)i (sub)central mea-
R ′ ′ ′ ′ p R ′ ′

sures µ on EA s.t. ω = EA ω dµ(ω ) and EA S ω dµ(ω ) ◦ S ω dµ(ω ) for
∀∆: Borel set in EA, 2) W*-subalgebras B of the centre: B ⊂ Z(W ∗ (πω )) =
πω (A)′ ∩ πω (A)′′ , 3) projections P on Hω s.t. P Ωω = Ωω , P πω (A)P ⊂
{P πω (A)P }′ . If µ, B, P are in correspondence, they are related mutually
as follows:

1. B = {P }′ ∩ Z(W ∗ (πω ));

2. P = [BΩω ];

3. µ(Â1 Â2 · · · Ân ) = hΩω | πω (A1 )P πω (A2 )P · · · P πω (An )Ωω i, where  ∈

C(EA) for A ∈ A denotes a map Â(ϕ) = ϕ(A) for ϕ ∈ EA;

4. B is *-isomorphic to the image of κµ : L∞ (E R A, µ) ∋ f 7−→ κµ (f ) ∈

πω (A)′ defined by hΩω | κµ (f )πω (A)Ωω i = dµ(ω ′ )f (ω ′ )ω ′ (A), and,
for A, B ∈ A, κµ (Â)πω (B)Ωω = πω (B)P πω (A)Ωω .

When B = {P }′ ∩Z(W ∗ (πω )) = Z(W ∗ (πω )), or equivalently, Z(W ∗ (πω )) ⊂

{P }′ , µ is called a central measure, for which we can derive the following
result from the above fact:

9
Proposition 2 ([16]) A map Λµ defined by

Λµ : πω (A)′′ ∋ πω (A) 7−→ κµ (Â) ∈ Z(W ∗ (πω ))

is a conditional expectation characterized by

Λµ (Z1 πω (A)Z2 ) = Z1 Λµ (πω (A))Z2 for Zi ∈ Z(W ∗ (πω )) (i = 1, 2).

To summarize, we have established the following logical connections:

1) As dual of embedding Z(W ∗ (A)) ֒→ W ∗ (A) of the centre, we obtain
⌢
a basic q→c channel EA ։ P rob(Spec(Z(W ∗ (A))) = P rob( A) with a factor
⌢
spectrum A = FA/ ≈ as the classifying space of sectors.R
2) A central measure µω with a barycentre ω = EA ω ′ dµω (ω ′ ) ∈ EA
specifies a conditional expectation Λµω : W ∗ (πω ) ∋ πω (A) 7−→ κµω (Â) =
[Spec(Z(W ∗ (πω ))) ∋ ω ′ 7−→ ω ′ (A)] ∈ Z(W ∗ (πω )), whose dual

Λ∗µω : P rob(Spec(Z(W ∗ (πω ))) → EW ∗ (πω )

⌢
defines a c→q channel given by Spec(Z(W ∗ (πω )))[⊂ A] ∋ γ 7−→ ωγ :=
Λ∗µω (δγ ) = δγ ◦ Λµω ∈ supp(µω ) ⊂ FA[⊂ EA] as a (local) section of the
⌢
bundle FA ։ [FA/ ≈] = A.
3) Operationally, this corresponds just to a choice of a selection cri-
terion to select out states of relevance and we have realized that the more
internal structure to be detected, the larger algebra we need, which requires
the Galois extension scheme just in parallel with DHR sector theory and
with my propsal of general augmented algebra, as seen below.

2.2 Selection criteria to choose an appropriate family of sec-

tors
Now we come to a “unified scheme for generalized sectors based on selection
criteria” [13, 14], extracted from a new general formulation of local thermal
states in relativistic QFT [2, 12]. What I have worked out so far in this
direction can be summarized as follows:
   
A) Non-equilibrium B) DHR sector theory of
 local states:   unbroken internal symmetry: 
continuous sectors discrete sectors
 ↓  ւ  
C) Sector structure of D) Unified scheme for
 broken symmetry:  →  Micro-Macro based 
discrete & continuous on selection criteria

A) General formulation of non-equilibrium local states in QFT [2, 12, 13];

10
 B) Reformulation [14] of DHR-DR sector theory [5, 6] of unbroken internal
symmetry;

C) Extension of B) to spontaneously or explicitly broken symmetry

[14, 15].

The results obtained in A), B) and C) naturally lead us to

D) Unified scheme for describing Micro-Macro relations based on

selection criteria [12, 13, 14]:
   
q : generic states c : reference model system with
i) =⇒ii)
of object system ↑ classifying space of sectors
↑

iii) a map to compare i) with ii)

 ⇑   ⇓ 
state preparation & classification &
  adjunction  
iv) selection criterion:  ⇄  interpretation of ,
ii) =⇒ i) i) w.r.t. ii): i) =⇒ ii)
c-q q-c

which can be seen as a natural generalization of

Example 3 The formulation of a manifold M based on local charts {(Uλ , ϕλ :

Uλ → Rn )} consisting of
i)= local neighbourhoods Uλ of M constituting a covering M = ∪Uλ ,
ii)= model space Rn ,
iii)= local homeomorphisms ϕλ : Uλ → Rn ,
iv)= interpretation of the atlas in terms of geometrical invariants such as
homology, cohomology, homotopy, K-groups, characteristic classes, etc., etc.

Example 4 Non-equilibrium local states in A) [2, 12, 13] are character-

ized by localizing the following generalized equilibrium states with fluctuating
thermal parameters:
i) = the set Ex of states ω at a spacetime point x satisfying certain energy
bound locally [ω((1 + HO )m ) < ∞ with “local Hamiltonian” HO ],
ii) = the space BK of thermodynamic parameters (β, µ) to distinguish among
different thermodynamic pure phases and the space M+ (BK ) =: T h of prob-
ability measures ρ on BK to describe fluctuations of (β, µ),
iii) = comparison
R of an unknown state ω with members of standard states
ωρ = C ∗ (ρ) = BK dρ(β, µ)ωβ,µ with parameters ρ belonging to reference sys-
tem, in terms of the criterion ω ≡ C ∗ (ρ) through “quantum fields at x” ∈ Tx
Tx
(justified by energy bound in i)).
iv) = adjunction
q⇄c
Ex /Tx (ω, C ∗ (ρ)) ≃ T h/C(Tx )((C ∗ )−1 (ω), ρ)

11
with q→c channel (C ∗ )−1 as a “left adjoint” to the c→q channel C ∗ (from the
classical reference system to generic quantum states): as a localized form of
the zeroth law of thermodynamics, this adjunction achieves simultaneously
the two goals of identifying generalized equilibrium local states and of giving
the thermal interpretation (C ∗ )−1 (ω) ≡ ρ of a selected generic state ω in
C ∗ (Tx )
the vocabulary of a standard known object ρ ∈ T h.

What we have discussed so far can be summarized as follows:

1. Classification of quantum states/representations by quasi-equivalence

(= unitary equivalence up to multiplicity): achieved by means of sec-
tors labelled by macroscopic order parameters as points in the
spectrum of centre, where a sector is defined by a quasi-equivalence
class of factor states ω ∈ FA with trivial centres Z(W ∗ (πω )) :=
W ∗ (πω ) ∩ W ∗ (πω )′ = C1Hω . In short, a sector = all density-matrix
states within a factor representation = a folium of a factor state.
⇓

2. A mixed phase = non-factor state = non-trivial centre Z(W ∗ (A)) 6=

C1H: allows “simultaneous diagonalization” as a central decomposition
arising from non-trivial sector structure.
=⇒ Z(W ∗ (A)): the set of all macroscopic order parameters to
distinguish among different sectors;
Spec(Z(W ∗ (A))): a classifying space to parametrize sectors com-
pletely in the sense that quasi-equivalent sectors correspond to one
and the same point and that disjoint sectors to the different points.
⇓

3. Micro-macro relation:
Intersector level controlled by Z(W ∗ (A)): macroscopic situations pre-
vail, which are macroscopically observable and controllable;
Inside a sector : microscopic situations prevail (e.g., for a pure state in
a sector, as found in the vacuum situations, it represents a “coherent
subspace” with superposition principle being valid).

4. Selection criterion = physically and operationally meaningful char-

acterization as to how and which sectors should be picked up for dis-
cussing a specific physical domain. E.g., DHR criterion for states ω
with localizable charges (based upon “Behind-the-Moon” argument)
πω ↾A(O′ ) ∼
= π0 ↾A(O′ ) in reference to the vacuum representation π0 .
A suitably set up criterion determines the associated sector structure
so that natural physical interpretations of a theory are provided in a
physical domain specified by it.

12
3 Sectors and symmetry: Galois-Fourier duality
To control the relations among algebras with group actons, their extensions
and corresponding representations, we need the Galois-Fourier duality as
an important variation of our main theme Micro-Macro Duality. The essence
of DHR-DR theory [5, 6] of sectors associated with an unbroken internal
symmetry can be seen in this duality which enables one to reconstruct a
field algebra F as a dynmaical system F xG with the action of an internal
symmetry group G from its fixed-point subalgebra A = FG consisting of
G-invariant observables in combination with data of a family T of states
∈ EA specified by the above DHR selection criterion:
Invisible micro Visible macro
   
G∼ = RepT Fourier duality T ∼
= RepG
 y  ⇄  y .
⇄
F∼ = A ⋊ Ĝ Galois duality A=F ∼ G
=F⋊G

In my recent reformulation, its applicability range restricted to unbroken

symmetries has been extended to not only spontaneously but also explicitly
broken symmetries.
In B) DHR-DR sector theory, we see

1. Sector structure:

H = ⊕ (Hγ ⊗ Vγ );
γ∈Ĝ

π(A) = ⊕ (πγ (A)′′ ⊗ 1Vγ ) = U (G)′ ,

′′
γ∈Ĝ

U (G) = ⊕ (1Hγ ⊗ γ(G)′′ ) = π(A)′ .

′′
γ∈Ĝ

2. Z(π(A)′′ ) = ⊕ C(1Hγ ⊗ 1Vγ ) = l∞ (Ĝ); Ĝ = Spec(Z(π(A)′′ )) =⇒

γ∈Ĝ
vocabulary for interpretation of sectors in terms of G-charges.
1−1
3. (πγ , Hγ ): sector of A ←→ (γ, Vγ ) ∈ Ĝ : equiv. class of irred. unitary
representations of a compact Lie group G of unbroken internal sym-
metry of field algebra F := A ⊗ Od with a Cuntz algebra generated
OdG
by isometries.

4. (π, U, H): covariant irred. vacuum representation of C*-dynamical sys-

tem F x G, s.t. π(τg (F )) = U (g)π(F )U (g)∗ .
τ

5. A, G, F: triplet of Galois extension F of A = FG by Galois group

G = Gal(F/A), determining one term from two.
How to solve two unknowns G & F from A?: DHR selection criterion

13
 Tannaka-Krein
=⇒ T (⊂ End(A)): DR tensor category ∼
= RepG =⇒ G =⇒
duality
F∼
= A ⋊ Ĝ.

Similar schemes hold also for C) with spontaneously and/or explicitly

broken symmetries. For instance, in the case of SSB, we have [14]
   
broken: G ⊃ H: unbroken T ∼
= RepH ∼ Ĥ ֒→ ∐gH∈G/H g Ĥg−1
 y y 
  ⇄ 
 y ↓ 

 b ∼ 
 F ⊃ F = Ad ⋊ Ĥ   
 Ad = FH G/H: degenerate vacua ,

 
 k  ⇄  ∪ ·· 
[ = Ad ⋊ Ĝ
F ⋊ (H\G) A = FG sector bundle

with Zπ̄ (Ad ) = L∞ (H\G; dġ)⊗Zπ (Ad ) = L∞ (H\G; dġ)⊗l∞ (Ĥ) and the base
space G/H of the sector bundle, Spec(Zπ̄ (Ad )) = ∐gH∈G/H gĤg−1 ։ G/H,
corresponds mathematically to the “roots” in Galois theory of equations and
physically to the degenerate vacua characteristic to SSB.

3.1 Hierarchy of symmetry breaking patterns and augmeneted

algebras
Extension of B) to broken symmetries [14, 15]: In my attempts to extend
DHR-DR sector theory with unbroken symmetries to the broken cases, the
adjunction,
augmented algebra
Broken ⇄ U nbroken,
has been important, as seen in my criterion of symmetry breaking:

Definition 5 ([14]) A symmetry described by a (strongly continous) auto-

morphic G-action τ : G y F(: field algebra), is unbroken in a given repre-
τ
sentation (π, H) of F if the spectrum Spec(Zπ (F)) of centre Zπ (F) := π(F)′′ ∩
π(F)′ is pointwise invariant (µ-a.e. w.r.t. the central measure µ which
decomposes π into factor representations) under the G-action induced on
Spec(Zπ (F)). If the symmetry is not unbroken in (π, H), it is said to be
broken there.

Remark 6 Since macroscopic order parameters Spec(Zπ (F)) emerge in low-

energy infrared regions, a symmetry breaking means the“infrared(=Macro)
instability” along the direction of G-action.

Remark 7 Since a representation π with broken symmetry can still contain

unbroken and broken subrepresentations, further decomposition of Spec(Zπ (F))
is possible into G-invariant domains. A minimal G-invariant domain is
characterized by G-ergodicity which means central ergodicity. =⇒ π is

14
decomposed into a direct sum (or, direct integral) of unbroken factor rep-
resentations and broken non-factor representations, each component
of which is centrally G-ergodic. =⇒ phase diagram on Spec(Zπ (F)).

Thus the essence of broken symmetry is found in the conflict between

factoriality and unitary implementability. In the usual approaches,
the former is respected at the expense of the latter. Taking the opposite
choice to respect implementability, we encounter a non-trivial centre which
provides convenient tools for analyzing sector structure and flexible treat-
ment of macroscopic order parameters to distinguish different sectors.
non-trivial centre
Namely, the adjunction holds between [Broken ⇄ Unbroken2 ],
controlled by a canonical homotopy η from [F x G with non-implementable
broken symmetry G in a pure phase] to [F b x G with unitarily implemented
symmetry G → U (G) in a mixed phase with a non-trivial centre], where F b
is an augmented algebra [14] defined by F \ as a crossed
b := F ⋊ (H\G),
product of F by the coaction of H\G (: degenerate vacua) arising from the
symmetry breaking from G to its unbroken subgroup H.
Note here that the above criterion does not touch upon the relation
between the symmetry group G and the dynamics of the physical system
described by the algebra F in relation with spacetime; if the latter is pre-
served by the former, the breakdown of symmetry G is called spontaneous
(SSB for short). Otherwise, it is explicit, associated with some parameter
changes involving changes of physical constants appearing in the specifi-
cation of a physical system. For instance, we can formulate such an ex-
plicitly broken symmetry as broken scale invariance associated with
temperature β as order parameter [15], where augmented algebra of ob-
\
servables  = A ⋊ (SO(3)\(R+ ⋊ L↑+ ) is the scaling algebra due to Buch-
holz and Verch [4] to accommodate the notion of renormalization group
(in combination with components arising from SSB of Lorentz boost sym-
metry due to thermal equilibrium [10] to accommodate relative velocity
uµ := β µ /β ∈ SO(3)\L↑+ ). What is scaled here is actually Boltzmann con-
stant kB !! In this way, we are led to the hierarchy of symmetry break-
ing patterns ranging from unbroken symmetries, spontaneous and explicit
breakdown of symmetries, the latter of which would be related with more
general treatments of transformations, such as semigroups or groupoids.
An eminent feature emerging through the hierarchy of symmetry break-
ing patterns is the phenomena of externalization of internal degrees of free-
dom in the form of order parameters and breaking parameters, along which
external degrees of freedom coupled to the system are incorporated through
Galois extension into the augmented algebra: it describes a composite
system consisting of the microscopic object system and its macroscopic “en-
2
To be precise, “unbroken” should be understood as “unitarily implemented”.

15
vironments”, which canonically emerge at the macroscopic levels consist-
ing of macroscopic order parameters classifying different sectors and of
symmetry breaking terms such as mass m and kB , etc. This formulation
allows us to describe the coupling between the system and external
fields in a universal way (e.g., measurement couplings).

4 From [thermality ⇄ geometry] towards [history

of Nature]
Although the modular structure of a W*-algebra in standard form has not
been explicitly mentioned so far, it plays fundamental roles almost every-
where in the above discussion, responsible for the homotopical extension
mechanism: this is crucial, for instance, in the formulation of group duality
and of scaling as well as conformal aspects. From the viewpoint that the
notion of quasi-equivalence fundamental to our whole discussion is just a
form of homotopy, we show here the Galois-theoretical aspects of modular
p p
structure M ⇄ M′ arising from canonical homotopy ηπ : π → π ◦◦ to move
to standard form.

Theorem 8 ([16]) i) In the universal representation (πu , Hu = ⊕ Hω )

ω∈EA
p
of a C*-algebra A, we define the maximal representation π ◦ disjoint from a
representation π = (π, Hπ ) ∈ RepA by
p p
π ◦ := sup{ρ ∈ RepA; ρ ≤ πu , ρ ◦ π}.

Then we have the following relations in terms of the projection P (π) ∈

W ∗ (A)′ on the representation space Hπ of π and its central support c(π):
p p p p p p p p
π1 ≤ π2 =⇒ π1◦ ≥ π2◦ , π ◦ = π ◦◦◦ and π ≤ π ◦◦ ,
p
P (π ◦ ) = c(π)⊥ := 1 − c(π),
p p
P (π ◦◦ ) = c(π)⊥⊥ = c(π) = ∨ uPπ u∗ ∈ P(Z(W ∗ (A))).
u∈U (π(A)′ )

ii) Quasi-equivalence π1 ≈ π2 (⇐⇒ π1 (A)′′ ≃ π2 (A)′′ ⇐⇒ c(π1 ) = c(π2 )

⇐⇒ W ∗ (π1 )∗ = W ∗ (π2 )∗ ) is equivalent to
p p p p
π1◦◦ = π2◦◦ .
p p p p
iii) The representation (π ◦◦ , c(π)Hu ) of W*-algebra W ∗ (π) ≃ π ◦◦ (A)′′ in
p p
the Hilbert space c(π)Hu = P (π ◦◦ )Hu gives the standard form of W ∗ (π)
associated with a normal faithful semifinite weight ϕ and the corresponding

16
Tomita-Takesaki modular structure (Jϕ , ∆ϕ ). It is characterized by the
universality:
p p
Std(π ◦◦ , σ) ≃ RepA(π, σ),
where Std denotes the caterogy of representations of A in standard form;
according to this relation, any intertwiner T : π → σ to a representation
p p
(σ, Hσ ) in standard form of W ∗ (σ) is uniquely factored T = T ◦◦ ◦ ηπ through
p p
the canonical homotopy ηπ : π → π ◦◦ with a uniquely determined intertwiner
p p p p
T ◦◦ : π ◦◦ → σ.
iv) The quasi-equivalence relation π1 ≈ π2 defines a classifying groupoid Γ≈
consisting of invertible intertwiners in the category RepA of representa-
tions of A, which reduces on each π ∈ RepA to Γ≈(π, π) ≃ Isom(W ∗ (π)∗ ),
the group of isometric isomorphisms of predual W ∗ (π)∗ as a Banach space.
The modular structure in iii) of W*-algebra W ∗ (π) =: M in the standard
p p
form in (π ◦◦ , c(π)Hu ) can be understood as the minimal implemention
by the unitary group U (M′ ) of a normal subgroup GM := Isom(M∗ )M ⊳
Isom(M∗ ) fixing M pointwise: namely, for γ ∈ GM, there exists Uγ′ ∈
U (M′ ) s.t.

hγω, xi = hω, γ ∗ (x)i = hω, Uγ′∗ xUγ′ i for ω ∈ M∗ ,

and Uγ′∗ xUγ′ = x ⇐⇒ x ∈ M. For M of type III, we can verify Galois-type

relations involving crossed product by a coaction of the group GM ≃ U (M′ )
as follows:
d
Z(M)′ = M ∨ M′ = M ⋊ GM: Galois extension of M,

M = (M ∨ M′ )GM : fixed-point subalgebra under GM,

GM = Gal(Z(M)′ /M): Galois group of M ֒→ Z(M)′ ,

according to which factoriality Z(M) = C1 of M can be seen as the ergod-

icity of M under Aut(M) or GM:

C1 = M ∩ M′ = M′ ∩ U (M′ )′ = (M′ )GM ⊃ (M′ )Aut(M) .

In view of the dominant roles of thermal or modular-theoretical notions

mentioned above, this theorem suggests possible paths from thermality to ge-
ometry to explain different geometries at macroscopic classical levels emerg-
ing from the invisible microscopic quantum world; it would explain the origin
of universality of Macro put in Micro-Macro Duality in our theoretical de-
scriptions of physical worlds. A typical example of this sort can be seen in
the formulation of group duality which exhibits its essence as a homotopi-
cal duality involving interpolation spaces [8]. Moreover, we can develop
a framework to go into a step from the above modular homotopy to the
generalized version of classifying spaces or classifying toposes [16]. Along

17
this line of thoughts, we can envisage such a perspective that theoretical
descriptions of physical nature can be mapped into a “categorical bundle
of physical theories” over a base category consisting of selection criteria to
characterize each theory as a fibre, which are mutually connected by meta-
morphisms of intertheory deformation arrows parametrized by fundamental
physical constants like ~, c, kB ; κ, e, etc., controlled by the “method of vari-
ations of natural constants” (work in progress). One of the most important
virtues of the above augmented algebra is found in the possibility that such
physical constants can be treated on the same footing as various physical
variables responsible for changing the symmetry properties of the systems; in
such contexts, they represent controlling parameters of deformations among
different selection criteria to determine theories corresponding to stabilized
hierarchical domains. Then the most crucial step will be to formulate each
selection criterion as an integrability condition in terms of generalized cat-
egorical connections, through which the framework can accommodate such
an adjunction as
  homotopical dilation  
irreversible stabilized hierarchical domains
⇄
historical process with reversible dynamics
to be found among such adjunctions as to put a generic category with non-
invertible arrows (describing an irreversible open system in a historical pro-
cess) in a relation adjoint to a groupoid with invertible arrows (correspond-
ing to a reversible closed system with repeatable dynamics in a specific
hierarchical domain). This kind of theoretical framework would provide
an appropriate stage on which the natural history of cosmic evolution be
developed.

References
[1] Baaj, S. and G. Skandalis, Ann. Scient. Ècole Norm. Sup. 26 (1993),
425-488.
[2] Buchholz, D., Ojima, I. and Roos, H., Ann. Phys. (N.Y.) 297 (2002),
219 - 242.
[3] Bratteli, O. and Robinson, D.W., Operator Algebras and Statistical Me-
chanics, vol. 1, Springer-Verlag (1979).
[4] Buchholz, D. and Verch, R., Rev. Math. Phys. 7 (1995), 1195-1240.
[5] Doplicher, S., Haag, R. and Roberts, J.E., Comm. Math. Phys. 13
(1969), 1-23; 15 (1969), 173-200; 23 (1971), 199-230; 35 (1974), 49-85.
[6] Doplicher, S. and Roberts, J.E., Comm. Math. Phys. 131 (1990), 51-
107; Ann. Math. 130 (1989), 75-119; Inventiones Math. 98 (1989),
157-218.

18
 [7] Enock, M. and Schwartz, J.-M., Kac Algebras and Duality of Locally
Compact Groups, Springer, 1992.

[8] Maumary, S. and Ojima, I., in preparation.

[9] Nakagami, Y. and Takesaki, M., Lec. Notes in Math. 731, Springer,
1979.

[10] Ojima, I., Lett. Math. Phys. 11 (1986), 73-80.

[11] Ojima, I., J. Stat. Phys. 56(1989), 203-226; Lec. Notes in Phys. 378,
pp.164-178, Springer, 1991.

[12] Ojima, I., pp. 48-67 in Proc. of Japan-Italy Joint Workshop on Funda-
mental Problems in Quantum Physics, Sep. 2001, eds. Accardi, L. and
Tasaki, S., World Scientific (2003).

[13] Ojima, I., pp.365-384 in “A Garden of Quanta”, World Scientific (2003);

e-print: cond-mat/0302283.

[14] Ojima, I., Open Systems and Information Dynamics, 10 (2003), 235-
279; math-ph/0303009.

[15] Ojima, I., Publ. RIMS 40, 731-756 (2004).

[16] Ojima, I., in preparation.

[17] Ozawa, M., J. Math. Phys. 25, 79-87 (1984); Publ. RIMS, Kyoto Univ.
21, 279-295 (1985); Ann. Phys. (N.Y.) 259, 121-137 (1997).

[18] Ozawa, M., quant-phys/0310072, to appear in Phys. Lett. A.

19