As promised here I comment on the difference between Peter and
myself in the weighting of the importance of
differential geometric and topological ideas in the present and future
development of local quantum physics (please
not axiomatic QFT since Haag and I intensely disliked the
terminology already since the beginning of the 60s,
because it gives a completely wrong impression of what we tried to
achieve). Starting from 1972 I began to look at
conformal QFT from a "Lie-field" point of view (for a brief review of
those old concepts see which are coming from
Wally Greenberg and John Lowenstein see "Two-dimensional models
as testing grounds for principles and concepts of
local quantum physics" published in the February 2006 issue of Ann. of
Phys.) and after realizing that the 2-dim.
energy momentum tensor is such a Lie-field, I As promised in my email
from last Friday, I wanted to return to some
differences started to look together with Swieca at more general fields
with anomalous scale dimensions (which
necessarily live on the covering of the Dirac-Weyl compactified
Minkowski spacetime). In contrast to naive
expectations they are reducible under the center of the covering
(all published in 1974/75 in Phys. Rev. D) and the
ensuing decomposition theory global fields into its central irreducible
components (the global fields fulfilling the
Irvine Segal concept of global causality on the covering are being
decomposed into operator field valued sections on
the compactified Minkowski spacetime, or to put it into more physical
terms the fields living in the hells and
heavens (the Luescher-Mack fields) are decomposed into objects which
live in our laboratory world where the non
conformal invariant hardware is localized). Only in d=1+1 where the
covering space decomposes into a product of two
coverings of the circle we found nontrivial examples (the exponential
Bose fields). The component fields were 10
years later called "conformal blocks" by BPZ. These component fields
were strange objects because they had a central
source and range projector and if the source projector did not match
the vacuum quantum numbers the application to
the vacuum was zero, not a property which Lagrangian or Wightman fields
have and which is consistent with the
Osterwalder-Schrader Euclideanization. As a result of our prejudice in
favor of Euclidean functional integral
representations at that time we left the exploration of that
interesting decomposition theory and looked at
geometric aspects of functional integrals.
During a visit of the CERN theory group 1976/77 I looked again at the
beautiful Lowenstein-Swieca operator
presentation of the 2-dim. Schwinger model. When I observed a general
connection between the zero modes of the
Euclidean spinor field in a generic vector potential with the winding
number of that vectorpotential I got extremely
excited (my rather late Sturm and Drang period had arrived) and thought
that this must be the tip of an iceberg. I
remember when I presented my concrete observations and made some of
these sweeping claims which came from my gut
feelings (but for which I had no convincing arguments at that time),
Roman Jackiw rightfully challenged my
speculative statements. I roamed the CERN theory library about elliptic
differential equations in topologically
interesting external fields but only after I went down to the Geneva
University mathematical library I finally found
something under the name Atiyah Singer which remotely resembled what I
was looking for. It took me several month
and the invaluable help of Hartmann Roemer (who came to CERN from the
University of Freiburg) to get a rough
impression of what it was about. When I finally wrote a paper with N.
K. Nielsen on the Euclidean functional
integral approach to the Schwinger model with special emphasis on these
new topological structures I used these new
words but there were still a lot of things outside my intellectual
range. When Swieca visited CERN for a short time
I infected him with my enthusiasm and soon several of his younger
Brazilian collaborators were trying these ideas
on all kind of two-dimensional models. What facilitated the situation
was the fact that Leo Kadanoff already had
prepared the ground by preempting many two-dimensional structures on
the lattice. Actually Swieca, having had the
same thesis adviser (Rudolf Haag) but 3 years after me, knew much more
about algebraic QFT than I at that time and
he once asked me whether we should not have a look at the
representation theoretical operator approach of Doplicher
Haag and Roberts of how to get the charged fields by knowing only the
(by definition) neutral local observables. I
told him that this may be a loss of time in view of the interesting
Euclidean functional&geometry approach, a
suggestion which years later I should regret
To make a long story short, I continued this topological geometrical
functional integral-based approach after
Swieca's death (end of 1980) but my confidence was waning. Around
1982/83 together with Roger Picken (who spend more
than a year at the FU Berlin) we tried to understand the Schulman
paradoxon that if on does the quantum mechanical
rigid top (movement on a compact group) via functional integrals and
takes into account the infinitely many saddle
points but stops at the quadratic fluctuations around them (i.e. simply
ignore the nonvanishing higher fluctuation
terms) one obtains the rigorous operator QM result! There was something
strange going on here, some sort of
"topological protection?". We found that this problem has a finite
dimensional counterpart: the Duistermaat-Heckmann
theory and of course the formal application to the infinite dimensional
Euclidean Feynman would confirm this; but
there does not exist a D-H theory in infinite dimension unless one
finds a regularization (approximation by a finite
integral) which maintains the D-H situation. We could not, and I think
nobody did find such a thing after us. So how
could you think of doing group-valued sigma models by functional
integration if not even this quantum mechanical
trivial example fits into this setting? It that time Luis Alvarez Gaume
wrote a nice little paper of a "derivation"
of the Atiyah-Singer index theorem in terms of a quantum
mechanical supersymmetric path integral but all this was
formal; it does not amount to a proof of the A-S theorem inasmuch as
the work with Picken did not resolve the
Schulman paradox (i.e. no proof of an infinite dim. extension of
Duistermaat-Heckmann). This was the end of my 8
year old love affair with differential geometry&Euclidean
functional integrals. I cleaned up my desk and started to
go back to those interesting algebraic structures I found from the
central decoposition theory with Swieca. This was
around 1983/84. I had the good luck to have a collaborator of
phantastic scientific qualities, Karl Henning Rehren.
He came from Pohlmeyer with an excellent background on integrable
systems, in particular with algebraic structures
arising from infinitely many conserved charges (in particular the ones
associated with the Nambu-Goto model) and was
enthusiastically looking at those new structures in the FQS and the BPZ
papers on 2-dim. conformal QFT whereas I
told him about the old algebraic structures with Swieca. There was no
doubt in our mind that the two things belong
together, but the details turned out to be tricky since the conceptual
and mathematical setting was quite different.
I remember that I got excited when it became clear to us that the
structures which appeared in our old conformal
decomposition theory where new fields with source and range projections
which formed "exchange algebras" with braid
group commutation relation. I am still proud of a little paper where we
constructed explicitly these exchange fields
in the conformal Ising field theory and computed with their help
explicit analytic expressions for n-point
functions. Our collaboration culminated in a paper which gave a
setting for chiral quantum field theory in which
the new exchange algebra structures are simply new realizations of the
old causality principles which also underly
the Lagrangian quantization- or the Wightman- setting and in
order to highlight the fact that all these new
structures still obey the same principles we chose the title "Einstein
causality and Artin braids". This was
published at approximately the same time as the Moore-Seiberg paper
which was based on a categorical analysis of the
braid group statistics setting and gained particular popularity by
mathematicians for who causality and localization
properties are outside their conceptual radar screen. with the
exception of a group in Rome and of the Japanese
mathematician Kawahigashi who got most of their powerful new results
from localization (spacetime-indexed nets of
observable algebras). Geometry has a role in local quantum
physics but it is totally subject to the local quantum
principles. The most important concept which is capable to convert
abstract algebraic properties (domain and range
properties of local operator algebras) into geometry (inner symmetries,
spacetime symmetries) is the Tomita-Takesaki
modular theory of operator algebras and I am convinced that in will
gain importance in the near future. My
constructive results which I obtained over the last couple of years on
factorizing models, lightfront and double
cone holography and results on localization entropy (including a very
interesting result which I will post on hep-th
on wednesday) all depend on the use of this amazingly powerful theory.
It is often said by outsiders that the algebraic approach is weak if it
comes to computation. This is a
misunderstanding. It can do all the computations (by causal
perturbation theory) one does with Feynman rules but it
avoids functional integrals for two simple reasons. First in functional
integral representations one is limited to
those special covariant representations which allow an Euler Lagrange
description and as already Weinberg has shown
in the 60s most covariantizations of the unique (m,s) Wigner
representations do not have this property (causal
perturbation does not depend on it) and second the correlation
functions of strictly renormalizable theories after
renormalization are simply not representable in the Feynman-Kac
functional integral form; so a functional integral
is only a formal starter of a computation, afterwards you follow the
intrinsic logic and do not give a damn for F-K
representability (renormalization cannot be done functionally). This is
similar to the old Bohr-Sommerfeld quantum
theory which was more on the artistic and less on the
mathematical-conceptual side. Scattering theory and cross
sections are not only a part of AQFT but AQFT was born together with
LSZ scattering theory being at its cradle. That
algebraic field theorist do not do much computing is another story;
violin builders like Strativari were mostly not
good fiddlers. AQFT extends standard QFT; example: for doing
perturbation theory in curved spacetime the usual
rules are useless, in particular there is no functional integral
representation. One has to separate the algebraic
structure from the structure of states on algebras and the standard
approach is incapable of doing this. But in the
algebraic approach this is very natural and one can find the
appropriate formalism in recent works of Wald,
Hollands, Brunetti-Fredenhagen...There are two interesting results. On
the one hand one understands why the
expectations in individual states cannot be diffeomorphism-invariant,
rather this invariance is a property of the
particular folium of states to which the given state belongs. The
second result is perhaps the most recent really
great achievement: for the first time we are able to formulate
Einstein's classical local covariance principle in
the setting of QFT in curved spacetime. This leads to a completely new
view of what a QFT is: it is an object which
lives simultaneously on all hyperbolic spacetimes, i.e. an abstract
algebraic substrate which (not unlike
stem-cells) can be spacetime-indexed by any globally hyperbolic manfold
(this corresponding to the spatially
organized organs you can grow from stem cells).
Perhaps the greatest conceptual achievement of AQFT is the derivation
of the origin of inner (charged) symmetries
from the (neutral) observable. These phantastic contributions of
Doplicher, Haag and Roberts are like Mark Kac's
"hearing the shape of a drum" i.e. reconstructing the whole structure
of QFT from its observable shadow. In this
case the problem cannot even be formulated in the lagrangian setting,
less solved.
I could continue the list with other results of the AQFT setting which
you cannot hope to obtain in other way, but I
think the relevant question to be asked here is:
What is the explanation why
such a rich setting is suffering from so much prejudice?
This question I asked already a long time ago and one of my reputable
colleagues gave me an answer which I find very
convincing. Sometime in the early 60s somewhere at an international
conference somebody (whose name I will not
mention sincs he is still alive) made a very stupid remark after a talk
by Weinberg and he claimed that what he was
telling Weinberg is a rigorous result of AQFT (which in that time was
called axiomatic, to the distast of Haag and
myself). This caused an enormous damage and I don't blame Weinberg for
considering a theory which produces such a
nonsense as not worthwhile. Weinberg's opinion carry a considerable
weight in the community and string theory would
never enjoy its present status without his support.
There is also a positive aspect to this prejudice. Only mature and very
qualified physicists who were able to make a
qualified judgement entered this area. In all papers from AQFT people
exactly deliver what they promise and mediocre
papers are virtually nonexistent.
The reason why I wrote my essay is that times really have worsened from
tolerating a small dedicated group to ethnic
cleansing of everything which is not stringy.
I cannot make an attempt to talk Weinberg out of his prejudice, but
maybe it is possible to try this in the case of
a younger not totally polarized person as Aaron Bergman.
To Aaron Bergman
I have the following suggestion: take some time, maybe 2 weeks and look
at math-ph/ 0511042 (not because I am a
co-author but because this is probably the simplest presentation of
modular theory in its spatial form). Afterwards
follow some of the reference e.g. a more mathematical presentation by
Brunetti, Guido and Longo or go directly to
the operator version of modular theory and its applications to
factorizing models. There is no proselyting on my
part (to enter research in this area one needs anyhow several years of
investment), I only want to let you know that
there are small time investments for getting rid of prejudices.
To Thomas Larsson
There is a conceptual error right at the beginning of your paper;
Lorentz covariance, quantum theory and the cluster
property do not lead to QFT. You find the counterexamples in my work
"An Anthology of non-local Quantum Field Theory
and Quantum Field Theory on non-commutative Spacetime" (published in
AOP towards the end of last year). These
so-called "direct particle interaction" models fulfill the
cluster-decomposition property yet they do not have a
representation in terms of a second quantization setting i.e. there is
no universal (n-independent) operator which
applied to the n-particle states maps to n+1 or n-1 particle states.
With other words this claim by Weinberg (based
on your 3 properties) on which there are many papers (including
philosophical treatises is incorrect, but nobody
checks what Weinberg says). I have not red your paper and perhaps you
later arguments and perhaps they do not depend
on what you write at the beginning. In any case considerable progress
to find analoga of those structures which one
has in chiral theories (energy-momentum virasoro structure, current
algebras) are well on their way in recent
articles of Todorov et al.
To Peter Woit
Coming back to Peter's viewpoint about the use of differential geometry
in quantum physics, I think it does not go
significantly beyond free quantum fields coupled to external
potentials. In no sense is the intrinsicness of diff.
geometry (coordinate independence) related to the intrinsicness problem
of QFT (independence of the use of
coordinatization by quantum fields). AQFT addresses the latter problem
and this is considerably more subtle (in
particular this is where modular operator theory comes in) than the
former. In fact I do not believe in the use of
any outside mathematics unless the principles of QFT ask for it (which
definitely is the case with modular theory,
in fact large parts of it were discovered independently by physicists.)
Isn't it ironic that string theorists accuse the AQFT of the use of
gratitious mathematics when they themselves are
using structures like gerbes and algebraic geometry? I cannot think of
more useless mathematical structures for
quantum theory (which deals with operators and states) than those.
Hypocricy? Hubris? Double standards?