# Why not diffeomorphism group representation theory?

by lugita15
Tags: diffeomorphism, representation, theory
 P: 1,583 For some reason, diffeomorphism invariance seems to be treated like a second-class citizen in the land of symmetries. In nonrelativistic quantum mechanics, we consider Galilean invariance so important that we form our Hilbert space operators from irreducible representations of the Galilei group. In relativistic quantum mechanics, I think we can do the same thing with representations of the Poincare group. (Could someone back me up on that? How do operators work in Fock space?) But when it comes time to consider quantum gravity, we do not grant diffeomorphism invariance an analogous role, citing the fact that it is a gauge symmetry and thus indicating a mere superfluousness in our mathematical description of physical states. I have a few issues with that. First of all, gauge symmetries can be quite important; it is the gauge invariance of Maxwell's equation that gives rise via Noether's theorem to (local) conservation of electric charge. What's the Noether charge for diffeomorphism invariance? Second of all, it seems to me that diffeomorphism invariance is more than just a gauge symmetry. Any statement that has rotational invariance, translational invariance, Lorentz invariance etc. (all locally) as it's implications surely has some physical significance. Can't we easily imagine a universe in which the laws of physics looked profoundly different in different coordinate systems? Has there been any work in building quantum gravity from the representation theory of the diffeomorphism group? Any help would be greatly appreciated. Thank You in Advance. P.S. Can someone recommend a good book on Lie group representations, direct sums, tensor products, and all that jazz?
 Sci Advisor P: 869 I have to think more about your questions, but perhaps you like the following two articles: [*] "Pseudoduality", Van Proeyen and Hull (about the difference between "proper symmetries" and "pseudo symmetries", formulated in terms of sigma models) [*] Black hole entropy is Noether charge, Robert Wald (about how one could assign a Noether charge to diffeomorphisms, but how to reconcile that with the first article is not yet clear to me)
 Sci Advisor P: 5,295 When looking at loop quantum gravity it becomes clear that (spatial) diffeomorphisms play a prominent role in its construction. The problem is that this construction is neither complete nor indisputable.
P: 1,583

## Why not diffeomorphism group representation theory?

 Quote by tom.stoer When looking at loop quantum gravity it becomes clear that (spatial) diffeomorphisms play a prominent role in its construction. The problem is that this construction is neither complete nor indisputable.
In LQG does the representation theory of the diffeomorphism group play a role in constructing Hilbert space operators?
P: 5,295
 Quote by lugita15 In LQG does the representation theory of the diffeomorphism group play a role in constructing Hilbert space operators?
not in the usual sense
P: 3,711
 it is the gauge invariance of Maxwell's equation that gives rise via Noether's theorem to (local) conservation of electric charge. What's the Noether charge for diffeomorphism invariance?
General covariance (or diffeomorphism covariance if you must call it that) can be thought of as the 'local' generalization of the translation group. That is, in place of global translations, xμ → xμ + εaμ, where aμ = const, an infinitesimal coordinate transformation may be written as xμ → xμ + εξμ where ξμ = ξ(x)μ. This is analogous to the gauge groups of electromagnetism we have gauge transformations Aμ → Aμ + λμ which are global or local depending on whether or not λ is a constant.

Question: for electromagnetism, which type of gauge transformation (global or local) generates the conserved current? The short answer is: they both do. But in different ways. And so to state the short answer to the OP, both the general covariance group and the translation group generate the same conserved quantity, namely (EDIT: the stress-energy tensor, and its charge,) the energy-momentum vector.

A very clear exposition of this topic can be found here. The point is that there are, in fact, two Noether theorems. The first deals with constant parameters (global gauge groups) while the second deals with function parameters (local gauge groups). Both of them lead to the same conserved current, but the equations of motion are required to be satisfied in the first case, while in the second case they are not.
P: 1,583
 Quote by tom.stoer not in the usual sense
So then what is the role for irreducible representations of diffeomorphism group in LQG? If Loop Quantum Gravity doesn't really use it, is there any other approach that does?
 Sci Advisor P: 5,295 In LQG it is claimed that the final version of the theory is diff. inv. b/c the symmetry has been reduced to the identity; however the way towards this rep. is slightly obscure and the implementation of all constraints including the Hamiltonian is not yet fully understood. Anyway, the diff. inv. symmetry is "unphysical" just like a gauge symmetry, therefore physical obervables and Hilbert space vectors should live in the trivial representation. Compare this to QCD: yes, you are right, in that case you start with SU(3) representations, quarks live in the triplet, gluons live in the octet, in order to construct the theory. But the hysical states are constrained by the "color Gauss law" which means that all physical states are gauge invariant states i.e.live in the singlet. So even in QCD (when it comes to physical states and observables) no representation but the trivial one is used. Regarding your last question: I do not know any other theory that uses representation theory of the spacetime diffeomorphims group.
P: 1,583
 Quote by tom.stoer In LQG it is claimed that the final version of the theory is diff. inv. b/c the symmetry has been reduced to the identity; however the way towards this rep. is slightly obscure and the implementation of all constraints including the Hamiltonian is not yet fully understood. Anyway, the diff. inv. symmetry is "unphysical" just like a gauge symmetry, therefore physical obervables and Hilbert space vectors should live in the trivial representation. Compare this to QCD: yes, you are right, in that case you start with SU(3) representations, quarks live in the triplet, gluons live in the octet, in order to construct the theory. But the hysical states are constrained by the "color Gauss law" which means that all physical states are gauge invariant states i.e.live in the singlet. So even in QCD (when it comes to physical states and observables) no representation but the trivial one is used.
But SU(3) invariance for QCD really is merely a gauge symmetry reflecting a redundancy in our mathematical description. In contrast, in any quantum field theory we get e.g. the momentum and angular momentum operators by considering irreducible representations of the Poincare group, because the Poincare group is a set of physically meaningful symmetries. But if that is the case, then surely the diffeomorphism group, which contains the Poincare group and much more things (like accelerated frames), should also be thought of in the same way. Is there any way to treat the diffeomorphism group in any way other than as a gauge group?

By the way, how are things done in quantum field theory on curved spacetimes? What is the symmetry group used there to construct Hilbert space operators?
P: 5,295
 Quote by lugita15 But SU(3) invariance for QCD really is merely a gauge symmetry reflecting a redundancy in our mathematical description. In contrast, in any quantum field theory we get e.g. the momentum and angular momentum operators by considering irreducible representations of the Poincare group, because the Poincare group is a set of physically meaningful symmetries. But if that is the case, then surely the diffeomorphism group, which contains the Poincare group and much more things (like accelerated frames), should also be thought of in the same way. Is there any way to treat the diffeomorphism group in any way other than as a gauge group?
You have to carefully distinguish two concepts:

Global symmetries like e.g. SU(N) flavor and global Lorentz or Poincare invariance in special relativity for which you can construct (well-known) representations; the physical Hilbert can be decomposed accordingly.

Local symmetries like gauge symmetries and diffeomorphim invariance; it is true that you have Poincare invariance in GR, but this becomes a local (gauge) symmetry (I will try to find papers for you). Global or "rigid" Lorentz or Poincare invariance is not a symmetry of GR, only as a special subset of local diffeomorphism invariance.

It should be clear that global Lorentz invariance requires a globally flat spacetime; it is also well-known that the concept of energy, momentum and angular momentum (as integral constants of motion) do no longer exist in GR in general.
 Sci Advisor P: 869 @tom.stoer: Especially in papers about supergravity this exchange of local translations and gct's is adressed, like in Samtleben's introduction. See also here, http://www.physicsforums.com/showthread.php?t=463190 at post #9.
P: 1,583
 Quote by tom.stoer You have to carefully distinguish two concepts: Global symmetries like e.g. SU(N) flavor and global Lorentz or Poincare invariance in special relativity for which you can construct (well-known) representations; the physical Hilbert can be decomposed accordingly. Local symmetries like gauge symmetries and diffeomorphim invariance; it is true that you have Poincare invariance in GR, but this becomes a local (gauge) symmetry (I will try to find papers for you). Global or "rigid" Lorentz or Poincare invariance is not a symmetry of GR, only as a special subset of local diffeomorphism invariance. It should be clear that global Lorentz invariance requires a globally flat spacetime; it is also well-known that the concept of energy, momentum and angular momentum (as integral constants of motion) do no longer exist in GR in general.
Yes, I am aware that nontrivial global symmetries are incompatible with space-time curvature. But are all local symmetries "gauge" symmetries in the pejorative sense, i.e. just indicating mathematical redundancy not physical content? I would think local Lorentz invariance is pretty physically significant; it indicates that space-time is locally Minkowskian. Can't you still use local symmetries to find local conservation laws, and can't you form operators using the representation theory of a local symmetry group?

I think quantum field theory in curved space-time would be relevant to all these issues.
 P: 1,583 I found out more information on this now. It seems that the generators of the local diffeomorphism group on a manifold are the vector fields defined on the manifold. Is there any way to interpret these vector fields as operators on the Hilbert space? This is especially interesting, because wave functions are scalar fields ψ(x,t), i.e. scalar fields on spacetime. So how could you find represent vector fields on a manifold as operators on the space of scalar fields of the manifold ?
 P: 33 "But are all local symmetries "gauge" symmetries in the pejorative sense, i.e. just indicating mathematical redundancy not physical content?" For a (philosophical) discussion of this issue, please see 'Gauging What's Real' by Richard Healey, OUP.
P: 5,295
 Quote by malreux But are all local symmetries "gauge" symmetries ... indicating mathematical redundancy ... ?
Yes; I do not know any counter example.

 Quote by malreux ... not physical content?
No; diffeomorphism invariance has of course a physical content, namely independence of reference frames.
P: 1,583
 Quote by tom.stoer Yes; I do not know any counter example. No; diffeomorphism invariance has of course a physical content, namely independence of reference frames.
For the record that was my quote; malreux was quoting me. I'm a bit confused. Some symmetries like translational invariance are said to be physically meaningful, and so the momentum operator is constructed using the representation theory of the local translation group (i.e. the associated Lie algebra). On the other hand, people do not do the same thing for the local diffeomorphism group; their justification for this is to dismiss this as a "gauge symmetry", by which they do not just mean that it is local, but that it is just a redundancy in our mathematical description of the physical world, not something physically meaningful.

But I agree with you: diffeomorphism invariance is a physically significant assertions about the laws of physics taking the same form in all reference frames. Thus it stands to reason that the generators of infinitesimal diffeomorphisms should have irreducible representations as Hilbert space operators.

Am I on the right track concerning vector fields on a manifold being the generators of diffeomorphisms on the manifold? If so, how could you construct Hilbert space operators out of vector fields? Also, am I right to assume that quantum mechanics in curved spacetime is relevant?
 P: 33 "On the other hand, people do not do the same thing for the local diffeomorphism group; their justification for this is to dismiss this as a "gauge symmetry", by which they do not just mean that it is local, but that it is just a redundancy in our mathematical description of the physical world, not something physically meaningful." -Lugita 15 The book I referenced can clarify some of the issues, such as what is meant when mathematical structure is considered unphysical or 'pure gauge'. It is written from the perspective of philosophy of physics rather than first order physics, but it is very clear with relatively simple formalism (although not toy examples). Its conclusion is basically "'Local' gauge symmetry is a purely formal feature of [e.g. Yang-Mills theories]" (p.155). This is not tackling what your asking about head on, by any stretch. But I found Healey's examination of classical and quantum field theories, and in particular what the transliteration of Weyl's 'gauge' means in the very different contexts of e.g. QFT and general relativity, very useful, very subtle, and illuminating. One last thing, there are many nice papers about QFT on curved spacetime, but I don't think I quite follow your reasoning here: "how could you construct Hilbert space operators out of vector fields?"
P: 1,583
 Quote by malreux "On the other hand, people do not do the same thing for the local diffeomorphism group; their justification for this is to dismiss this as a "gauge symmetry", by which they do not just mean that it is local, but that it is just a redundancy in our mathematical description of the physical world, not something physically meaningful." -Lugita 15
Here's a tip, malreux: if you want to quote someone's post, all you need to do is click the "Quote" button next to the post.
 One last thing, there are many nice papers about QFT on curved spacetime, but I don't think I quite follow your reasoning here: "how could you construct Hilbert space operators out of vector fields?"
I'm not doing anything unusual here, just following standard quantum procedure. For example, consider nonrelativistic QM. The laws of physics are invariant under spatial translation, so we look at the representation theory of the translation group. Ultimately we find the generators of infinitesimal translations, and represent these generators as operators on the Hilbert space, in this case the momentum operators. And we find that the laws of physics are invariant under time translation, so we represent the generator of infinitesimal time translations as an operator on the Hilbert space, the Hamiltonian operator. Finally we have rotational invariance, so we construct operators on the Hilbert space, the angular momentum operators, which generate infinitesimal rotations.

In each case, the pattern is clear. We find a continuous set of symmetry transformations, which form a Lie group. We then look at the infinitesimal transformations that are in this group, AKA the Lie algebra. We find the generators of these infinitesimal transformations. Finally we try to find self-adjoint operators on our Hilbert space which correspond to these generators. (I'm skipping a step: we often first try finding unitary operators to represent elements of the whole Lie group, and then we use those to find the self-adjoint operators representing the infinitesimal generators. For instance, we first find the unitary time evolution operators U(t) which represent the Lie group of time translations, then we express the Hamiltonian operator in terms of U(dt).)

In the present case the symmetry we are concerned with is local diffeomorphism invariance, so I would like to similarly construct Hilbert space operators using the representation theory of the diffeomorphism group. Based on some searching, it seems that the generators of infinitesimal diffeomorphisms on a manifold are vector fields on the manifold. So somehow these vector fields have to be represented as operators on the Hilbert space. This is somewhat interesting, because the Hilbert space is the set of wave functions, which are themselves scalar fields on the manifold. So we have to represent vector fields acting on the manifold as self-adjoint linear operators acting on scalar fields acting on the manifold.

 Related Discussions Linear & Abstract Algebra 3 Calculus & Beyond Homework 7 Linear & Abstract Algebra 3