INSTRUCTIONS
The text between the lines BODY and ENDBODY is made of
1306 lines and 54406 bytes (not counting or )
In the following table this count is broken down by ASCII code;
immediately following the code is the corresponding character.
34992 lowercase letters
2897 uppercase letters
1206 digits
2 ASCII characters 9
6764 ASCII characters 32
18 ASCII characters 33 !
15 ASCII characters 34 "
95 ASCII characters 35 #
1280 ASCII characters 36 $
516 ASCII characters 37 %
31 ASCII characters 38 &
120 ASCII characters 39 '
523 ASCII characters 40 (
523 ASCII characters 41 )
30 ASCII characters 42 *
20 ASCII characters 43 +
490 ASCII characters 44 ,
253 ASCII characters 45 -
646 ASCII characters 46 .
15 ASCII characters 47 /
31 ASCII characters 58 :
4 ASCII characters 59 ;
7 ASCII characters 60 <
130 ASCII characters 61 =
7 ASCII characters 62 >
2 ASCII characters 64 @
73 ASCII characters 91 [
1911 ASCII characters 92 \
73 ASCII characters 93 ]
130 ASCII characters 94 ^
374 ASCII characters 95 _
96 ASCII characters 96 `
513 ASCII characters 123 {
20 ASCII characters 124 |
514 ASCII characters 125 }
85 ASCII characters 126 ~
BODY
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%% AN ALGEBRAIC SPIN AND STATISTICS THEOREM. I %%%%%%%%%%
%%%%%% D. Guido - R. Longo, 15 pages, plain TeX %%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%% Special Fonts %%%%%%%
%Number Sets % Blackboard bold
\def\Co{{\bf C}} %\def\Co{{\Bbb C}}
\def\Na{{\bf N}} %\def\Na{{\Bbb N}}
\def\Re{{\bf R}} %\def\Re{{\Bbb R}}
\def\To{{\bf T}} %\def\To{{\Bbb T}}
\def\Ze{{\bf Z}} %\def\Ze{{\Bbb Z}}
%%%%%% SYMBOLS %%%%%%%%%%%%%
\def\A{{\cal A}}
\def\a{\alpha}
\def\B{{\cal B}}
\def\b{\beta}
\def\C{{\cal C}}
\def\d{\delta}
\def\D{\Delta}
\def\cD{{\cal D}}
\def\f{\varphi}
\def\F{{\cal F}}
\def\g{\gamma}
\def\G{\Gamma}
\def\bG{{\bf G}}
\def\H{{\cal H}}
\def\k{\kappa}
\def\K{{\cal K}}
\def\bK{{\bf K}}
\def\l{\lambda}
\def\L{\Lambda}
\def\m{\mu}
\def\n{\nu}
\def\o{\omega}
\def\O{{\cal O}}
\def\p{\pi}
\def\Q{\Omega}
\def\r{\rho}
\def\s{\sigma}
\def\S{{\cal S}}
\def\t{\tau}
\def\T{\Theta}
\def\th{\vartheta}
\def\W{{\cal W}}
\def\x{\xi}
\def\Z{{\cal Z}}
\def\Lp{{\cal L}_+} %Lorentz proper
\def\Lpo{{\cal L}_+^\uparrow} %Lorentz proper orthochronous
\def\Pp{{\cal P}_+} %Poincare' proper
\def\Ppo{{\cal P}_+^\uparrow} %Poincare' proper orthochronous
\def\Sp{\widetilde{\cal P}_+} %covering of Poincare' proper
\def\Spo{\widetilde{\cal P}_+^\uparrow} %covering of Poincare' prop. ort.
\def\imply{\Rightarrow}
\def\Ind#1#2{{\rm Ind}_{#1}^{#2}}
%%%%%%%% Macros for theorems %%%%%%%%
\font\sectionfont=cmbx10 scaled\magstep1
\def\begProof{\medskip\noindent{\bf Proof.}\quad}
\def\begProofof#1{\medskip\noindent{\bf Proof of #1.}\quad}
\def\square{\hbox{$\sqcap\!\!\!\!\sqcup$}}
\def\endProof{\hskip-4mm\hfill\vbox{\vskip3.5mm\square\vskip-3.5mm}\bigskip}
\def\titlea#1{\vskip0pt plus.3\vsize\penalty-75
\vskip0pt plus -.3\vsize\bigskip\bigskip
\noindent{\sectionfont #1}\nobreak\smallskip\noindent}
\def\titleb#1{\medskip\noindent{\it#1}\qquad}
\def\claim#1#2{\vskip.1in\medbreak\noindent{\bf #1.} {\sl #2}\par
\ifdim\lastskip<\medskipamount\removelastskip\penalty55\medskip\fi}
\def\begProof{\medskip\noindent{\bf Proof.}\quad}
\def\begProofof#1{\medskip\noindent{\bf Proof of #1.}\quad}
\def\square{\hbox{$\sqcap\!\!\!\!\sqcup$}}
\def\endProof{\hbox{\hskip-4mm\hfill\vbox{\vskip3.5mm\square\vskip-3.5mm}}
\bigskip}
\def\beglemma#1#2\endlemma{\claim{#1 Lemma}{#2}}
\def\begdefinition#1#2\enddefinition{\claim{#1 Definition}{#2}}
\def\begtheorem#1#2\endtheorem{\claim{#1 Theorem}{#2}}
\def\begcorollary#1#2\endcorollary{\claim{#1 Corollary}{#2}}
\def\begremark#1#2\endremark{\claim{#1 Remark}{#2}}
\def\begproposition#1#2\endproposition{\claim{#1 Proposition}{#2}}
\def\refno#1#2{\item{[#1]}{#2}}
\def\begref#1#2{\titlea{#1}}
\def\endref{}
\newcount\FNOTcount \FNOTcount=1
\def\numfnot{\number\FNOTcount}
\def\addfnot{\global\advance\FNOTcount by 1}
\def\fonote#1{\footnote{$^\numfnot$}{#1}\addfnot}
%%%%%%%% AUTO REF. %%%%%%%%%%%%%
\newcount\REFcount \REFcount=1
\def\numref{\number\REFcount}
\def\addref{\global\advance\REFcount by 1}
\def\wdef#1#2{\expandafter\xdef\csname#1\endcsname{#2}}
\def\wdch#1#2#3{\ifundef{#1#2}\wdef{#1#2}{#3}
\else\write16{!!doubly defined#1,#2}\fi}
\def\wval#1{\csname#1\endcsname}
\def\ifundef#1{\expandafter\ifx\csname#1\endcsname\relax}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\def\rfr(#1){\wdef{q#1}{yes}\ifundef{r#1}$\diamondsuit$#1
\write16{!!ref #1 was never defined!!}\else\wval{r#1}\fi}
\def\inputreferences{
\def\REF(##1)##2\endREF{\wdch{r}{##1}{\numref}\addref}
\REFERENCES}
\def\references{
\def\REF(##1)##2\endREF{
\ifundef{q##1}\write16{!!ref. [##1] was never quoted!!}\fi
% \item{[\rfr(##1)]}##2}
\refno{\rfr(##1)}##2}
\begref{References}{99}\REFERENCES\endref}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%% TITLE %%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\font\ftitle=cmbx10 scaled\magstep1
\centerline{\ftitle AN ALGEBRAIC SPIN AND STATISTICS THEOREM. I}
\bigskip
\centerline{Daniele Guido$^1$\footnote{$^*$}
{ Supported in part by MURST and CNR-GNAFA.} and Roberto Longo$^{1,2*}$}
\footnote{}{E-mail:\ guido@mat.utovrm.it, longo@mat.utovrm.it }
\bigskip
\item{$(^1)$} Dipartimento di Matematica, Universit\`a di Roma ``Tor Vergata''
\par via della Ricerca Scientifica, I--00133 Roma, Italia.
\item{$(^2)$} Centro Linceo Interdisciplinare,
\par via della Lungara 10, I--00165 Roma, Italia
\bigskip\bigskip
\noindent{\bf Abstract.} A spin-statistics theorem and a PCT theorem are
obtained in the context of the superselection sectors in Quantum Field Theory
on a 4-dimensional space-time.
Our main assumption is the requirement that the modular groups of the
von Neumann algebras of local observables associated with wedge regions act
geometrically as pure Lorentz transformations.
Such a property, satisfied by the local algebras generated
by Wightman fields because of the Bisognano-Wichmann theorem, is regarded as
a natural primitive assumption.
%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%% REFERENCES %%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%
\def\REFERENCES{
\REF(AHKT1)Araki H., Haag R., Kastler D., Takesaki M., ``{\it
Extensions of KMS states and chemical potential}'', Commun.
Math. Phys. {\bf 53} (1977), 97-134.
\endREF
\REF(BiWi1)Bisognano J., Wichmann E., ``{\it On the duality
condition for a Hermitian scalar field}", J. Math. Phys. {\bf
16} (1975), 985-1007.
\endREF
\REF(BiWi2)Bisognano J., Wichmann E., ``{\it On the duality
condition for quantum fields}", J. Math. Phys. {\bf
17} (1976), 303-321.
\endREF
\REF(BjDr1)Bjorken J.D., Drell S.D., ``{\it Relativistic
Quantum Fields}'', McGraw-Hill, New York 1965.
\endREF
\REF(Borc2)Borchers H.J., ``{\it Local rings and the
connection between spin and statistics}", Commun. Math. Phys. {\bf 1} (1965),
281-307
\endREF
\REF(Borc1)Borchers H.J., ``{\it The CPT theorem in
two-dimensional theories of local observables}", Commun. Math.
Phys. {\bf 143} (1992), 315.
\endREF
\REF(Borc3)Borchers H.J., ``{\it On the converse of the Reeh-Schlieder
theorem}'', Commun. Math. Phys. {\bf 10} (1968), 269-273.
\endREF
\REF(BGL1)Brunetti R., Guido D., Longo R., ``{\it Modular
structure and duality in conformal quantum field theory}",
Commun. Math. Phys. {\bf 156} (1993), 201-219.
\endREF
\REF(BGL2)Brunetti R., Guido D., Longo R., ``{\it Group
cohomology, modular theory and space-time symmetries}",
to appear in Rev. Math. Phys.
\endREF
\REF(BuEp1)Buchholz D., Epstein H., ``{\it Spin and statistics
of quantum topological charges}'', Fizika {\bf 3} (1985),
329-343.
\endREF
\REF(BuFr1)Buchholz D., Fredenhagen K., ``{\it Locality and
structure of particle states}'', Commun. Math. Phys. {\bf 84}
(1982) 1-54.
\endREF
\REF(BuSu1)Buchholz D., Summers S.J., ``{\it An algebraic
characterization of vacuum states in Minkowski space}", Commun. Math. Phys.
{\bf 155} (1993), 442-458.
\endREF
\REF(Burgoyne)Burgoyne N., ``{\it On the connection of spin with
statistics}'', Nuovo Cimento {\bf 8} (1958), 807.
\endREF
\REF(Dell'Antonio)Dell'Antonio G.F., ``{\it On the connection of spin with
statistics}'', Ann. Phys. {\bf 16} (1961), 153.
\endREF
\REF(DHR1)Doplicher S., Haag R., Roberts J.E., ``{\it Local
observables and particle statistics I}'', Commun. Math. Phys.
{\bf 23} (1971), 199-230.
\endREF
\REF(DHR2)Doplicher S., Haag R., Roberts J.E., ``{\it Local
observables and particle statistics II}'', Commun. Math. Phys.
{\bf 35} (1974), 49-85.
\endREF
\REF(DoLo1)Doplicher S., Longo R., ``{\it Standard and split
inclusions of von Neumann algebras}", Invent. Math. {\bf 73}
(1984), 493-536.
\endREF
\REF(DoRo)Doplicher S., Roberts J.E., ``{\it Why there is a
field algebra with a compact gauge group describing the
superselection structure in particle physics}'', Commun. Math.
Phys. {\bf 131} (1990), 51-107.
\endREF
\REF(Dyson)Dyson F.J., ``{\it On the connection of weak local
commutativity and regularity of Wightman functions}'', Phys. Rev.
{\bf 110} (1958), 579.
\endREF
\REF(Fie)Fierz M., ``{\it \"Uber die relativische Theorie
kr\"aftfreier Teilchen mit beliebigem spin}'', Helv. Phys. Acta
{\bf 12} (1939), 3.
\endREF
\REF(Epst1)Epstein H., ``{\it CTP invariance in a theory of
local observables}'', J. Math. Phys. {\bf 8} (1967), 750.
\endREF
\REF(GL1)Guido D., Longo R., ``{\it Relativistic invariance
and charge conjugation in quantum field theory}'', Commun. Math.
Phys. {\bf 148} (1992), 521-551.
\endREF
\REF(GL3)Guido D., Longo R., ``{\it An algebraic Spin and Statistics
theorem. II}'' in preparation.
\endREF
\REF(Haag1)Haag R., {\it Local Quantum Physics}, Springer
Verlag, Berlin Heidelberg 1992.
\endREF
\REF(HHW)Haag R., Hugenoltz N.M., Winnink M., ``{\it On the equilibrium states
in quantum statistical mechanics}'', Commun. Math.
Phys. {\bf 5} (1967), 215.\endREF
\REF(HaKa1)Haag R., Kastler D., ``{\it An
algebraic approach to Quantum Field Theory}'', J. Math. Phys. {\bf 5} (1964),
848-861.
\endREF
\REF(Jost1)Jost R., ``{\it The general theory of Quantized Fields}''
Amer. Math Soc., Providence RI 1965.
\endREF
\REF(Jos57)Jost R., ``{\it Eine Bemerkung zu CTP Theorem}'',
Elv. Phys. Acta {\bf 30} (1957), 409.
\endREF
\REF(Kay)Kay B.S., ``{\it The free field on the wedge and uniqueness
of KMS one-particle structure}'' preprint Feb. 1983.
\endREF
\REF(Kuck1)Kuckert B., ``{\it PCT und Kovarianz als modulare Strukturen}'',
Diploma Thesis, Hamburg, in preparation.
\endREF
\REF(Lipsman)Lipsman R.L., ``{\it Group representations}'' Lect. Notes
in Math. {\bf 388}, Springer Verlag, New
York--Heidelberg--Berlin 1974.
\endREF
\REF(Land1)Landau L., ``{\it On local functions of fields}'' Commun. Math.
Phys. {\it 39} (1974), 49-62.
\endREF
\REF(Long1)Longo R., ``{\it Index of subfactors and statistics
of quantum fields. }'', $I$ Commun. Math. Phys. {\bf 126} (1989), 217-247;
$II$ {\bf 130} (1990), 285-309.
\endREF
\REF(Lud54)L\"uders G., ``{\it Vertaugschungsrelationen zwischen
verschiedenen Feldern}'', Z. Naturforsch. {\bf 13a} (1958), 254.
\endREF
\REF(LudersAndZumino)L\"uders G., Zumino B., ``{\it Connection between spin
and statistics}'', Phys. Rev. {\bf 110} (1958), 1450.
\endREF
\REF(Pau)Pauli W., ``{\it On the connection between spin and statistics}'',
Phys. Rev. {\bf 58} (1940), 716.
\endREF
\REF(Pau55)Pauli W., ``{\it Exclusion principle, Lorentz group and
reflection of space time and charge}'', in Niels Bohr and the Development of
Physics, W. Pauli (ed.) Pergamon Press, New York, 1955.
\endREF
\REF(Schwinger)Schwinger J., ``{\it On the theory of quantized fields I}'',
Physics Reviews {\bf 82} (1951), 914.
\endREF
\REF(Sewell)Sewell G.L., ``{\it Relativity of temperature and Hawking
effect}'', Phys. Lett. {\bf79A} (1980), 23.
\endREF
\REF(StZs1)Str\u atil\u a S., Zsido L., {\it Lectures on von~Neumann
algebras}, Abacus press, England 1979.
\endREF
\REF(Streater)Streater R.F., ``{\it Local fields with the wrong
connection between Spin and Statistics}'', Commun. Math. Phys.
{\bf 5} (1967), 88-96.
\endREF
\REF(StWi1)Streater R.F., Wightman A.S., ``{\it PCT, Spin and Statistics,
and all that}'', Benjamin, Reading (MA) 1964.
\endREF
\REF(Take1)Takesaki M., ``{\it Tomita theory of modular Hilbert
algebras }'', Lect. Notes in Math. {\bf 128}, Springer Verlag, New
York--Heidelberg--Berlin 1970.
\endREF
\REF(Wigh)Wightman A.S., ``{\it Quantum field theory in terms
of vacuum expectation values}'', Phys. Rev. {\bf 101}, (1956) 860.
\endREF
\REF(Wies1)Wiesbrock H.V., ``{\it A comment on a recent work of
Borchers}", Lett. Math. Phys. {\bf 25}, (1992), 157-159.
\endREF
\REF(Yngv1)Yngvason J., ``{\it A note on essential
duality}'', preprint.
\endREF}
\inputreferences
\titlea{Introduction}
In this paper we shall reconsider from an intrinsic point of view two well
known fashinating theorems in Quantum Field Theory: the PCT theorem
and the Spin and Statistics theorem.
Both of them have a long history, see [\rfr(Jost1),\rfr(StWi1)]. The spin and
statistics theorem first appeared in the context of free fields in the work of
Fierz [\rfr(Fie)] and Pauli [\rfr(Pau)]: one cannot second quantize particles
with integer spin by anticommuting fields, i.e. fields obeying Fermi statistics,
nor particles with half-integer spin by local fields, i.e. fields obeying Bose
statistics.
The PCT theorem originated in [\rfr(Lud54)] as a relation between the
existence of the space-inversion symmetry P and the existence of the product of
the charge and the time-inversion symmetry CT. Pauli proved in
[\rfr(Pau55)] that PCT is always a symmetry of Lorentz invariant field
equations.
It was a success of the Wightman axiomatic approach [\rfr(Wigh)] to extabilish
model independent results: the connection between spin and statistics was
obtained by Burgoyne [\rfr(Burgoyne)], see also
[\rfr(LudersAndZumino),\rfr(Dell'Antonio)], and a PCT theorem by Jost
[\rfr(Jos57)], see also [\rfr(Dyson), \rfr(Schwinger)], both relying on the
general holomorphic properties of the $n$-point functions. A spin and
statistics theorem in the algebraic approach [\rfr(HaKa1)], see also
[\rfr(Borc2)], was later given by Epstein [\rfr(Epst1)] and has a version for
(Doplicher-Haag-Roberts) DHR superselection sectors [\rfr(DHR2)] and for
more general topological charges [\rfr(BuEp1)].
All these approaches heavily rely on arguments of analytic continuation, whose
nature give some mysterious effectiveness to the results. Moreover they make
use of certain detailed structures, either because they deal with Wightman
tempered distributions or because they treat the case of finite mass
degeneracy, where a supeselection sector has to contain only finitely many
particles of the same type and all of them are assumed to have strictly
positive mass.
The approach to Quantum Field Theory by local observable algebras
[\rfr(HaKa1)] suggests however that a PCT symmetry and spin-statistics
correspondence should be intrinsically associated with the net of local
algebras and manifest itself as the consequence of the locality
principle\fonote{The spin-statistics relation depends on sharp locality. In
the second quantization of Bose particles by anti-commuting fields, microscopic
causality is still asymptotically present and its violation is sizeable only
at distances comparable to the Compton wave lenght [\rfr(BjDr1)]. This is
perhaps an indication that the spin-statistics relation might be different in
contexts like Quantum Gravity where a sharp causality principle does not
occur.}.
>From the mathematical point of view the spin-statistics correspondence is a
relation between two quantities of different nature, the univalence and
the statistics phase, and one is led to tie up these concepts on a general
ground, somehow in the spirit of an index theorem.
We shall extablish both a PCT and a spin-statistics theorem in the following
general context.
Let $\O\to\A(\O)$ be a net of von~Neumann algebras on a Hilbert space
$\H$, i.e. an inclusion preserving association between regions
$\O$ in the four-dimensional Minkowski space and von Neumann
algebras of local observables, that we assume here to be irreducible.
We make the following assumptions.
{\it Locality.} If $\O_1$ and $\O_2$ are space-like separated regions, then
$\A(\O_1)$ and $\A(\O_2)$ commute elementwise.
{\it Modular covariance.} There is a vector $\Q\in\H$, the vacuum
vector, cyclic for the algebras $\A(W)$ associated with all wedge region $W$ in
the Minkowski space, such that
$$
\D_W^{it}\A(\O)\D_W^{-it}=\A(\L_W(t)\O)\ ,\qquad t\in\Re
$$
where $\O$ is any region, $\D$ is the Tomita-Takesaki modular operator
[\rfr(Take1),\rfr(StZs1)] associated with $(\A(W),\Q)$ and $\L_W$ is the
one-parameter rescaled group of pure Lorentz transformations preserving $W$.
{\it Reeh-Schlieder property.} The vacuum vector $\Q\in\H$ is
also cyclic for the
algebras $\A(\S)$ associated with all space-like cones $\S$.
Locality is the well-known expression of Einstein causality and we do not
dwell on it. The Reeh Schlieder property is
known to hold for the vacuum vector in a Poincar\`e covariant theory as a
consequence of the positivity of the energy and the weak additivity assumption
for the local algebras [\rfr(Borc3)].
Modular covariance needs however further comments. Postponing for a while the
justification for such assumption, we recall that this entails the net to be
covariant with respect to the universal covering $\Spo$ of the Poincar\'e group
$\Ppo$, with positive energy [\rfr(BGL2)]. Indeed we shall prove here that it
is actually covariant with respect to $\Ppo$ as a special case of our general
Spin and Statistics theorem and taking to completion our previous work.
Therefore modular covariance is a way to intrinsically encode the Poincar\'e
covariance property in the net structure, providing a canonical
representation of the Poincar\`e group $\Ppo$ (cf. also [\rfr(BuSu1)]).
Let now $\r$ be a superselection sector of $\A$ in the sense of
Doplicher-Haag-Roberts [\rfr(DHR1)] or more generally of Buchholz-Fredenhagen
[\rfr(BuFr1)].
An index-statistics relation [\rfr(Long1)] shows that
$$
{\rm Ind}(\r)=d(\r)^2
$$
where ${\rm Ind}(\r)$ is the Jones index of $\r$ and $d(\r)$ is the DHR
statistical dimension, namely
$$
d(\r)=|\l_\r|^{-1}
$$
where the statistical parameter $\l_\r\in\Re$ classifies the statistics in
$3+1$ space-time dimensions [\rfr(DHR1)].
Therefore the index is an intrinsic quantity that determines the statistics up
to the Fermi-Bose alternative, i.e. the sign of $\l_\r$.
On the other hand, the Poincar\'e representation in the vacuum sector being
fixed by modular covariance, the representation of $\Spo$ associated with a
covariant irreducible sector $\r$ is uniquely determined, therefore the
univalence (integer or half-integer spin alternative) is intrinsically
associated with $\r$. Since $\r$ is automatically $\Spo$-covariant if
$d(\r)<\infty$ (assuming a regularity property for the net [\rfr(GL1)]), it is
natural to expect a general algebraic Spin and Statistics theorem connecting
these two intrinsic quantities for any sector with finite statistics.
Our result in this respect will in fact show that on these general grounds
$$
{\rm sign}(\l_\r)=U_\r(2\pi)
$$
where $U_\r$ is the representation of $\Spo$ in the sector $\r$ and
$U_\r(2\pi)$ denotes the corresponding rotation by $2\pi$.
Modular covariance also implies that the anti-unitary involution $\T$,
definable by the modular theory according to the Bisognano-Wichmann
prescription [\rfr(BiWi2)], implements a complete space-time reflection. As
shown in [\rfr(GL1)], this entails that $\T$ intertwines a sector with its
conjugate. We therefore obtain a PCT symmetry.
We come now back to the origin of the modular covariance
property. Its main justification certainly comes from the
Bisognano-Wichmann theorem [\rfr(BiWi1),\rfr(BiWi2)] to the effect that this
property holds if the local algebras are constructed from Wightman fields.
An algebraic version of the Bisognano-Wichmann theorem
does not exist yet, except in the case of conformal theories where it holds in
full generality [\rfr(BGL1)].
However a theorem of Borchers [\rfr(Borc1)] shows part of the geometric
properties of the modular group for wedge regions to be always present and in
particular every $1+1$ dimensional Poincar\'e covariant net satisfying essential
duality has the modular covariance property.
At the present time no counter-example to modular covariance is known to
exist within Poincar\`e covariant theories (see however
[\rfr(Land1),\rfr(Yngv1)]). There are nevertheless counter-examples to the
spin-statistics theorem [\rfr(Streater)]: these are constructed by infinite
multiplicity fields where the Poincar\'e group representation is not unique. It
turns out that the wrong connection between spin and statistics depends on the
wrong choice of the Poincar\`e group representation, while our canonical choice
for the latter has the desired properties. We remark that an intrinsic way to
eliminate pathological examples of the above kind comes by requiring the split
property [\rfr(DoLo1)]; this indeed implies the uniqueness of the Poincar\`e
group representation [\rfr(BGL1)] and we propose it as a natural candidate for a
derivation of the modular covariance property by first principles.
On the physical side modular covariance manifests an interesting analogy with
the Unruh effect and with the Hawking black hole thermal radiation, as first
noticed by Sewell [\rfr(Sewell)]. We sketch the essential ideas, see also
[\rfr(Haag1)]. As is known the modular group of a von Neumann algebra with
respect to a given state is characterized by the Kubo-Martin-Schwinger
condition [\rfr(Take1)] and, on the other hand, KMS condition is peculiar of
thermal equilibrium states in Statistical Mechanics [\rfr(HHW)]. By the
Bisognano-Wichmann theorem the boosts satify the KMS condition with respect to
the vacuum, as automorphisms of the von Neumann algebra of the corresponding
wedge $W$ and, on the other hand, the orbits of the boosts are the trajectories
of a uniformly accelerated motion for which the "Rindler universe" $W$ is a
natural horizon; the equivalence principle in Relativity Theory then allows an
interpretation of the thermal outcome as a gravitational effect. On this basis
Haag has long proposed to derive the Bisognano-Wichmann theorem.
The role of the modular covariance assumption may be also understood by its
consequences. Among other things, it implies the positivity of the energy for
the constructed Poincar\'e group representation [\rfr(Wies1),\rfr(BGL2)]. As is
known the positivity of the energy is lost on a curved space-time, and the
modular covariance seems to be the appropriate substitute in this case.
Moreover, as already mentioned, it gives rise to the KMS condition, namely an
analytic continuation property. It turns out that this analytic aspect of the
modular covariance assumption encorporates all the holomorphic properties
present in Quantum Field Theory. But, as a matter of facts, the modular group
is an algebraic object, a manifestation of the $^*$-operation, thus providing
us with an algebraic approach to our problems.
We pass now to a description of the methods of our work. This paper relies on
the modular theory of Tomita and Takesaki and on an analysis with the unitary
representations of ${\rm SL}(2,\Re)$.
We shall find a key relation arising from the comparison of the modular
groups of different algebras, and we shall regard it as an identity concerning
operators in the space of a representation of ${\rm SL}(2,\Re)$, because of the
well known fact that the 2+1 dimensional Lorentz group is isomorphic to
${\rm SL}(2,\Re)/\{1,-1\}$. Section~1 contains the proof of this identity by
the Mackey machine of the induced representations (e.g. [\rfr(Lipsman)]) and a
free field verification.
In section~2 the PCT theorem and the Spin and Statistics relation are proven
in the context of the field algebras,\fonote{We have recently been informed by
Kuckert of an independent complementary analysis based on assumptions of weak
geometric type for the modular conjugation [\rfr(Kuck1)].} where the formalism
is close to the classical formulation.
Then, in section~3, we obtain our result in the context of local
observables. This is done by rephrasing the statements in terms of the
Doplicher-Roberts field algebra [\rfr(DoRo)]. This last step has certain
pedagogical advantages, but has to be avoided in order
to extend our work to more
general settings where the field algebra does not exists.
In a forthcoming paper [\rfr(GL3)] we shall indeed provide a more intrinsic
approach in terms of local observables only, that will cover low dimensional and
conformal theories in particular. The general picture will be clarified by
examples.
\titlea{1. An identity for operators associated with
representations of ${\rm SL}(2,\Re)$}
Let us consider two one-parameter subgroups of ${\rm SL}(2,\Re)$
$$
\mu(t)\equiv\left (\matrix{\cosh\pi t&-\sinh\pi t\cr-\sinh\pi t&\cosh\pi
t\cr}\right)\ ,\quad
\nu(t)\equiv\left (\matrix{e^{-\pi t}&0\cr0&e^{\pi t}\cr}\right)\ .
$$
If $U$ is a unitary representation of ${\rm SL}(2,\Re)$ on a Hilbert space
$\H$ we look at the corresponding selfadjoint infinitesimal generators
$$
H = H_U = i{d\over dt}U(\mu(t))|_{t=0}
$$
$$
K = K_U = i{d\over dt}U(\nu(t))|_{t=0}.
$$
We shall denote by $\bG$ the group ${\rm PSL}(2,\Re)$ given by the quotient of
${\rm SL}(2,\Re)$ with its center $\{-1,1\}$ and by $\widetilde\bG$ the
universal covering of $\bG$, which is of course the universal covering of ${\rm
SL}(2,\Re)$ too. We take the same definition for $H$ and $K$ in the case of a
unitary representation $U$ of the universal covering group $\widetilde\bG$.
We will consider the following property for a representation $U$ of
$\widetilde\bG$:
$$
\eqalign{
T_t&\equiv e^{{ 1\over 2}K}e^{itH}e^{-{1\over 2}K}\subset e^{-itH}\cr
{\rm and}\quad T_t&\hbox{\rm is densely defined}\quad\forall t\in\Re\cr}
\eqno(1.1)
$$
where the symbol $\subset$ denotes the extension of operators.
Property~(1.1) refers to a representation $U$, but we omit the symbol
$U$ when no confusion arises.
\begtheorem{1.1} Property~(1.1) holds for all the unitary
representations of $\widetilde\bG$.
\endtheorem
In order to prove this theorem, we first observe that it is enough to check
Property~(1.1) on dense sets of vectors, not necessarily on the full domain of
$T_t$.
\beglemma{1.2} Let us assume that, for each real $t$, there is a dense subset
${\cal D}_t$ of the domain of $T_t$ such that $T_t|_{{\cal D}_t}\subset
e^{-itH}$. Then Property~(1.1) holds for the given representation.
\endlemma
\begProof Note first that the matrix
$\left(\matrix{0&1\cr-1&0\cr}\right )\in {\rm SL}(2,\Re)$
conjugates $\mu(t)$ with $\mu(-t)$ and $\nu(t)$ with $\nu(-t)$
therefore the assumption of the lemma remains true if we replace $K$ with
$-K$ and $H$ with $-H$, in particular $e^{ -{1\over 2}K}e^{itH}e^{{1\over 2}K}$
is densely defined. Now
$$
T_t^*\supset e^{ -{1\over 2}K}e^{-itH}e^{{1\over 2}K}
$$
therefore $T_t^*$ is densely defined and $T_t$ is closable. Since $T_t\xi
= e^{-itH}\xi$ for all $\xi$ in a dense set and $e^{-itH}$ is bounded,
the equality $T_t\xi=e^{-itH}\xi$ must hold for all $\xi$ in the domain of
$T_t$.
\endProof
\begcorollary{1.3}Let $U_1,U_2$ and $U$ be unitary representations of
$\widetilde\bG$.
\medskip
\item{$(a)$} Property~(1.1) holds for $U_1\otimes U_2$ iff it holds for both
$U_1$ and $U_2$.
\item{$(b)$}If $U = \int^{\oplus}U(\lambda)dm(\lambda)$ is a direct integral
decomposition of $U$, then Property~(1.1) holds for $U$ iff it holds
for $U(\l)$, for $m$-almost all $\lambda$.
\medskip
\endcorollary
\begProof Part $(b)$ and the implication $\Leftarrow$ of part $(a)$ are
immediate by Lemma~1.2 since one can check Property~(1.1) on natural dense sets.
If Property~(1.1) holds for $U_1\otimes U_2$ then it holds for $U_1$ and $U_2$
up to a constant, namely, considering for example the representation $U_1$,
there exists a one dimensional character $z(\cdot)$ of $\Re$ such that
$$
T_t\subset z(t)e^{-itH}
$$
and $T_t$ is densely defined. Of course $z(\cdot)$ remains unchanged if
we replace $\mu$ and $\nu$ by a pair of conjugate one-parameter subgroups. As
in the proof of Lemma~1.2 we may thus replace $\mu(t)$ by $\mu(-t)$ and thus
$z(t)$ by $z(-t)$, hence $z(t)=z(-t)=1$ and the proof is complete.
\endProof
We need now to verify Property~(1.1) in some specific representation.
To this end recall that [\rfr(StWi1)], if $W_i$ is the wedge in the
$3$-dimensional space-time along the axis $x_i$, $i=1,2$, and $\L_i(t)$ is the
associated one-parameter group of pure Lorentz transformations
(see section~2), there is an isomorphism of ${\rm PSL}(2,\Re)$ with the
$2+1$-dimensional Lorentz group $\Lpo(3)$ determined by
$$
\m(t)\to\L_1(t)\quad{\rm and}\quad
\n(t)\to\L_2(t) .
$$
Accordingly, we
shall identify $\bG$ with a subgroup of the 2+1-dimensional Poincar\'e group
$\Ppo(3)$.
\beglemma{1.4} Let $V\equiv V_{m,0}$ be the positive energy
representation of $\Ppo(3)$ of spin 0 and mass $m>0$. Then Property~(1.1) holds
for the restriction $U\equiv V|_{\bG}$ of $V$ to $\bG$.
\endlemma
\begProof As is known, $V$ extends to a (anti-)representation of the proper
Poincar\'e group $\Pp(3)$, namely there exists a anti-unitary involution
$\Theta$, on the same Hilbert space, that commutes with $U$ and
implements the change of sign on the translation operators in any space-time
direction.
By the one-particle version of the Bisognano Wichmann theorem
(which follows of course from the Bisognano-Wichmann theorem in the free field
setting, see [\rfr(Kay)] for a short direct verification of this special case)
we may identify the rescaled boost transformations with the modular group of
the real Hilbert subspace of the one-particle Hilbert space associated to the
corresponding wedge region. Then Property~(1.1) holds because it is equivalent
to the commutativity of $\Theta$ with the boosts (see Proposition~2.6).
\endProof
\begremark{1.5} The proof of Lemma~1.4 makes use of a one-particle version
of the Bisognano-Wichmann theorem; as we mentioned this can be proved directly
by mimicking the proof of the Bisognano-Wichmann
in this special case. Since this quick verification requires an analytic
continuation argument that does not fit with the spirit of this paper, we sketch
here an algebraic derivation of Lemma~1.4. To begin with note that Lemma~1.4
would hold if $V$ were the irreducible positive-energy massless representation
of $\Ppo(3)$ with helicity 0. Indeed in this case $V$ extends to a
representation
of the conformal group and the algebraic argument in [\rfr(BGL1)] applies. Now
$V\otimes V$ has a direct integral decomposition into irreducible
representations where massive representations occur and thus Lemma~1.2 implies
Lemma~1.4 for some $m>0$. However the representation $U\equiv V|_{\bG}$
in Lemma~1.4 does not depend on $m>0$ up to unitary equivalence by the following
proposition, hence Lemma~1.4 holds for all $m>0$.
\endremark
\begproposition{1.6} The representation $U\equiv V|_{\bG}$
in Lemma~1.4
is equivalent to the quasi-regular representation of $\bG$
corresponding to the rotation subgroup
$\bK\equiv\left\{\left(\matrix{\cos\theta
&\sin\theta\cr-\sin\theta&\cos\theta\cr}\right),
0\leq\theta<2\pi\right\}/\{1,-1\}$,
namely $U$ is the representation of $\bG$ induced by the identity
representation of $\bK$.
\endproposition
\begProof The $m>0$ hyperboloid
$H_m=\{\vec{x}\in\Re^3/x_0^2-x_1^2- x_2^2 = m^2,\, x_0>0\}$
is a homogeneous space for $\bG$ whose stability subgroup at the point
$(m,0,0)$ is $\bK$. $U$ is the corresponding representation on
$L^2(H_m,\mu_m)$, with $\mu_m$ the Lorentz invariant measure on $H_m$, and this
is by definition the quasi-regular represention with respect to $\bK$.
\endProof
\begcorollary{1.7} With $U\equiv V|_{\bG}$ the representation in Lemma~1.4,
$U\otimes U$ is equivalent to an infinite multiple of the regular representation
$\lambda$ of $\bG$
$$
U\otimes U = \infty\cdot\lambda.
$$
In particular Property~(1.1) holds for $\lambda$.
\endcorollary
\begProof The first statement is a consequence of the Mackey tensor product
theorem for induced representations, see [\rfr(Lipsman), Theorem~2 and
Example~5]. Property~(1.1) then holds for $\lambda$ because of
Lemma~1.2.\endProof
Here is an alternative verification of Property~(1.1) for $\lambda$.
By Proposition~1.6, taking tensor products and making use of Corollary~1.3,
we check that Property~(1.1) is valid for the irreducible representation
$V_{m,s}$ of $\Ppo(3)$ of any mass $m>0$ and any integral spin
$s$. Now $U_{m,s}\equiv V_{m,s}|_\bG$ is the induced representation
$$
U_{m,s}=\Ind{\bK}{\bG}(\chi_s)
$$
where $\chi_s$ is character of $\bK\simeq\To$ associated
with the integer $s$. By inducing at stages one has
$$
\lambda=\Ind{\{1\}}{\bG}(id)=\Ind{\bK}{\bG}(\lambda_\bK)
$$
where $\lambda_\bK=\Ind{\{1\}}{\bK}(id)$ is the
regular representation of \bK, hence
$$
\lambda =\Ind{\bK}{\bG}(\lambda_\bK)
= \Ind{\bK}{\bG}\bigoplus_{s=-\infty}^{\infty}\chi_s =
\bigoplus_{s=-\infty}^{\infty}\Ind{\bK}{\bG}(\chi_s) =
\bigoplus_{s=-\infty}^{\infty}U_{m,s}
$$
and the statement follows by Corollary~1.4.
\begProofof{of Theorem~1.1} Suppose first that $U$ is a representation
of $\bG$. By the absorbing property of the regular representation
$\lambda$, $U\otimes\lambda$ is equivalent to a multiple of $\lambda$,
therefore Property~(1.1) holds for $U$ because of Lemma~1.2 and Corollary~1.7.
If now $U$ is a more general representation of $\widetilde\bG$, then $U$
detemines a projective representation of $\bG$, namely $U(l(\cdot))$ with $l$
a Borel section for the quotient map of $\widetilde\bG$ modulo $\bG$. The
tensor product $U\otimes\overline U$ of $U$ with the conjugate representation
is a true representation of $\bG$, hence Property~(1.1) holds for
$U\otimes\overline U$. Then, by Lemma~1.2, Property~(1.1) holds for $U$.
\endProof
\titlea{2. PCT, Spin and Statistics on the field algebras}
In this section we consider a pre-cosheaf $\O\to\F(\O)$ of von~Neumann algebras
acting on a Hilbert space $\H$, where $\O$ is any open region of
the $4$-dimensional Minkowski space $M$.
We assume the following properties:
\medskip
\item{$(1)$} {\it Reeh-Schlieder property} for space-like cones:
there is a vector $\Q\in\H$ ({\it vacuum}) which is cyclic for the
algebras associated with all space-like cones.
\item{$(2)$} {\it Normal commutation relations}: there is a
vacuum-preserving self-adjoint unitary $\G$ (statistics operator)
that implements an automorphism on every local von Neumaann algebra and the
normal commutation relations between Bose and Fermi fields hold, i.e. setting
$$
\F_{\pm}(\O):=\{A\in\F(\O):\G A\G=\pm A\}
$$
we have that if $\O_1$ and $\O_2$ are space-like separated then
$\F_+(\O_1)$ commutes with $\F(\O_2)$ and $\F_-(\O_1)$
anticommutes with $\F_-(\O_2)$.
\item{$(3)$} {\it Modular covariance property} with respect to the vacuum
vector $\Q$ (cf. Definition~2.3).
\medskip
\begproposition{2.1}{\rm(Twisted Locality)} Let $Z$ be the unitary
operator defined by
$$
Z={I+i\G\over 1+i}.\eqno(2.1)
$$
Then
$$
Z\F(\O)Z^*\subset\F(\O')'.\eqno(2.2)
$$
\endproposition
\begProof A direct computation shows that
$$
\eqalign{
ZBZ^*=B,\qquad&B\in\F_+\cr
ZFZ^*=i\G F\quad&F\in\F_-\cr}.\eqno(2.3)
$$
Hence, if $\O_1$ and $\O_2$ are space-like separated and
$F_j\in\F_-(\O_j)$, $j=1,2$ we have
$$
[ZF_1Z^*,F_2]=i\G(F_1F_2+F_2F_1)=0
$$
and the thesis holds.
\endProof
We recall that a {\it wedge} region is any Poincar\'e
transformed of the region $W_1:=\{\vec{x}\in\Re^n:|x_0|