LQG discreteness no conflict with Lorentz contraction

[marcus]

My impression was that only misinformed people argue that LQG discreteness is in conflict with Lorentz invariance. This was resolved a long time ago, i thought.

Does anyone disagree? Want to talk about it? Explain something to me that I am missing?

http://arxiv.org/abs/gr-qc/0205108
Reconcile Planck-scale discreteness and the Lorentz-Fitzgerald contraction
Carlo Rovelli, Simone Speziale
12 pages, 3 figures
(Submitted on 25 May 2002)

"A Planck-scale minimal observable length appears in many approaches to quantum gravity. It is sometimes argued that this minimal length might conflict with Lorentz invariance, because a boosted observer could see the minimal length further Lorentz contracted. We show that this is not the case within loop quantum gravity. In loop quantum gravity the minimal length (more precisely, minimal area) does not appear as a fixed property of geometry, but rather as the minimal (nonzero) eigenvalue of a quantum observable. The boosted observer can see the same observable spectrum, with the same minimal area. What changes continuously in the boost transformation is not the value of the minimal length: it is the probability distribution of seeing one or the other of the discrete eigenvalues of the area. We discuss several difficulties associated with boosts and area measurement in quantum gravity. We compute the transformation of the area operator under a local boost, propose an explicit expression for the generator of local boosts and give the conditions under which its action is unitary."

Discreteness in LQG is a serious issue. For various reasons I've always thought of it as a bit of a liability for the theory. FOR ONE THING THE LOOP COMMUNITY IS ALMOST ALL WORKING ON things like spinfoam, groupfieldtheory, simplicial (e.g. Loll CDT), renorm'ble quantum metric (e.g. Reuter) and these are THINGS WHICH DON'T HAVE MINIMAL LENGTH.

It is a strange anomaly to have a community called Loop and have almost nobody working in canonical LQG proper, and to have canonical Loop the only approach with e.g. an area operator with discrete spectrum. It makes it seem as if the community connects to discreteness in geometrical measurement only by its historical accident name.

That is bad enough but it is worse when we get people who don't know what they are talking come in and say that LQG can't possibly be right BECAUSE discrete spectrum of a geometric observable breaks LORENTZ. It simply does not. That is just something badmouth people say who don't know any better.

 

[marcus]

I can sympathise with Dittrich Thiemann wanting to break LQG out of the discrete geometry bind.

Not because of Lorentz invariance which is a hyped-up NON-ISSUE. But because it would bring a lot of change and maybe convergence with other approaches.

But they failed to bring it off. Rovelli demolished Dittrich Thiemann argument in a gentle respectful way that leave lots of room to appreciate their valuable contribution.
And it is definitely valuable, but my reading of what Rovelli posted today is that he still demolished their case. If you want, check it out:
http://arxiv.org/abs/0708.2481
Comment on "Are the spectra of geometrical operators in Loop Quantum Gravity really discrete?" by B. Dittrich and T. Thiemann

Nevertheless the discreteness issue is still a hot one. Set aside Lorentz and set aside Dittrich Thiemann, and it is still demanding attention.

Let's consider two recent papers by Brunnemann and Rideout that show that the LQG volume spectrum approaches zero arbitrarily close. these are numbercrunch papers.

http://arxiv.org/abs/0706.0469
Properties of the Volume Operator in Loop Quantum Gravity I: Results
Johannes Brunnemann, David Rideout
37 pages, 7 figures
(Submitted on 4 Jun 2007)

"We analyze the spectral properties of the volume operator of Ashtekar and Lewandowski in Loop Quantum Gravity, which is the quantum analogue of the classical volume expression for regions in three dimensional Riemannian space. Our analysis considers for the first time generic graph vertices of valence greater than four. Here we find that the geometry of the underlying vertex characterizes the spectral properties of the volume operator, in particular the presence of a 'volume gap' (a smallest non-zero eigenvalue in the spectrum) is found to depend on the vertex embedding. We compute the set of all non-spatially diffeomorphic non-coplanar vertex embeddings for vertices of valence 5--7, and argue that these sets can be used to label spatial diffeomorphism invariant states. We observe how gauge invariance connects vertex geometry and representation properties of the underlying gauge group in a natural way. Analytical results on the spectrum on 4-valent vertices are included, for which the presence of a volume gap is proved. This paper presents our main results; details are provided by a companion paper arXiv:0706.0382v1.

http://arxiv.org/abs/0706.0382
Properties of the Volume Operator in Loop Quantum Gravity II: Detailed Presentation
Johannes Brunnemann, David Rideout
95 pages, 65 figures
(Submitted on 4 Jun 2007)

"The properties of the Volume operator in Loop Quantum Gravity, as constructed by Ashtekar and Lewandowski, are analyzed for the first time at generic vertices of valence greater than four. The present analysis benefits from the general simplified formula for matrix elements of the Volume operator derived in gr-qc/0405060, making it feasible to implement it on a computer as a matrix which is then diagonalized numerically. The resulting eigenvalues serve as a database to investigate the spectral properties of the volume operator. Analytical results on the spectrum at 4-valent vertices are included. This is a companion paper to arXiv:0706.0469, providing details of the analysis presented there."

And here is Rideout's invited plenary talk at Loops 07
http://www.matmor.unam.mx/eventos/loops07/talks/PL4/Rideout.pdf
http://www.matmor.unam.mx/eventos/loops07/talks/PL4/Rideout.mp3

They found there is no smallest nonzero volume in the spectrum of the LQG volume operator---define, according to what i understand, on the kinematical states as usual. this bothers me. I find it at first a surprising idea, and then a nagging one. The positive eigenvalues might well be a DISCRETE set and yet they get arbitrarily close to zero. Like the numbers 1/n are a discrete set but there is no smallest one. there is no volume "gap". does anyone want to explain this or put it in context?

did anybody here attend Rideout's talk at Loops 07 in June?

 

[Demystifier]

Rovelli also had a nice analogy:
Does a discrete spectra of the angular momentum in ordinary nonrelativistic QM violates rotational invariance? The answer is - of course not.

Or even a simpler analogy:
Does a discrete spectra of momentum of a particle in a box violates translational invariance? Again, of course not.

One should also not forget about the Ehrenfest theorem:
Even though the spectrum of the eigenvalues may be discrete, the average value satisfies classical equations of motion and may attain an arbitrary value (provided that the zero value is also one of the eigen-values in the spectrum).

So, LQG is nothing special in that regard. It is just another example of QM.

 

[etera]

Or even a simpler analogy:
Does a discrete spectra of momentum of a particle in a box violates translational invariance? Again, of course not.

hum... doesn't the box itself break translation invariance??

Rovelli also had a nice analogy:
Does a discrete spectra of the angular momentum in ordinary nonrelativistic QM violates rotational invariance? The answer is - of course not.

So, LQG is nothing special in that regard. It is just another example of QM.

Yes, Carlo's analogy is great! the "minimal length" is not a problem as soon as the "length" is not simply considered as a classical object but as a quantum operator. I tried to give some Lorentz invariant toy models to illustrate this point in an old paper with Daniele Oriti, gr-qc/0405085. We use models very similar to the non-commutative geometry of the deformed special relativity type. However, it is all in flat space-time... so still quite far from LQG and the diffeomorphism invariance!

Finally, i would like to say that we do not need to go to quantum gravity to get a notion of Lorentz invariant minimal length. We can simply look at the QFT of a massive field, there's the Compton length which is a minimal scale for all observers... The way that this minimal length is implemented does not involve gravity but simply the quantum fluctuations of the field at high energy (small scales).
The difference with quantum gravity is that the qg minimal length is supposed to depend on the gravitational constant and not of any details of the matter fields... but the mechanism could be similar... or am i missing something?

 

[Demystifier]

hum... doesn't the box itself break translation invariance??

Actually yes, so maybe this analogy is not so perfect. But it is still fine, because in classical mechanics the momentum of the particle is arbitrary (not discrete), which can be obtained by boosts of the momentum. The box does not change it.

 

[Mike2]

Can a metric be quantized but the coordinates be continuous? Or is this a contradiction of terms? We can assign a continuous sequence of numbers to a variable; each number differs from the previous by a digit in the infinith decimal place. But perhaps the distance formula between these barely different numbers may be quite a bit larger than the previous pair of barely different numbers. Does this make sense? Thanks.

 

[Demystifier]

Can a metric be quantized but the coordinates be continuous?

Yes.
First, if something is quantized, it does not mean that it has a discrete spectrum.
Second, if a quantum operator has a discrete spectrum, the average values still can attain continuous values.

 

[arivero]

What I do not understand, was this new scale suppossed to be different of Planck scale?

Because Planck scale is simply Newton Constant in natural units, and as far as I know, Newton constant is perfectly compatible with general relativity. Of course it is not compatible with special relativity; if it were, why should be need general relativity at all?

 

[Mike2]

Yes.
First, if something is quantized, it does not mean that it has a discrete spectrum.
Second, if a quantum operator has a discrete spectrum, the average values still can attain continuous values.

I guess I'm confused about how you can say that numbers can actually be "different" without a difference function, or metric. What property of 5 and 7 makes them different if we don't take into account 7-5=2?

 

[Demystifier]

I guess I'm confused about how you can say that numbers can actually be "different" without a difference function, or metric. What property of 5 and 7 makes them different if we don't take into account 7-5=2?

Think about spacetime points as elements of a set (not numbers!). Two points are either different or not. But if they are different, in general you cannot say how much they are different. To do that, you must introduce an additional structure on this set - the metric. Take some mathematics textbook to see general axioms that a metric on a set must satisfy.

 

[Thomas Larsson]

Spacetime has more structure than a set - it has a smooth structure, which allows us to define partial derivatives. Hence the distance between two spacetime points can have three values: zero, finite, and infinitesimal.

This reminds me of one of my pet peeves. The anomalous dimension, i.e. the eigenvalue of the dilatation operator D, is background independent, because D = x^u d/dx^u does not depend of the metric.

 

[arivero]

There is a letter of 1981 from Brink and Schwarz titled "quantum superspace", http://www.slac.stanford.edu/spires/find/hep/www?j=PHLTA,B100,310 , that makes interesting reading. Pity is not available for free online.

 

[Rebel]

Yes.
First, if something is quantized, it does not mean that it has a discrete spectrum.
Second, if a quantum operator has a discrete spectrum, the average values still can attain continuous values.

May you give examples where something is quantized not having a discrete spectrum? I don't know if you were meanning more that are many things in quantum mechanics and QFT that are conitnuus, but doesn't mean that they are quantized, as the case of momentum and position.

Also, maybe a better and much simpler example for the second is how the operators (here for the metric) having discrete spectrum have always a representation in terms of continuus variables.

 

[marcus]

Rebel, I see you have a question for Demystifier. Sorry to butt in, I had something more to say about the general topic of the thread. But want to preserve your question:[QUOTE Demystifier;1407476]Yes.
First, if something is quantized, it does not mean that it has a discrete spectrum.
Second, if a quantum operator has a discrete spectrum, the average values still can attain continuous values
==endquote==

[QUOTE Rebel;2272676]May you give examples where something is quantized not having a discrete spectrum? I don't know if you were meanning more that are many things in quantum mechanics and QFT that are conitnuus, but doesn't mean that they are quantized, as the case of momentum and position.

Also, maybe a better and much simpler example for the second is how the operators (here for the metric) having discrete spectrum have always a representation in terms of continuus variables.
==endquote==

I wanted to say that the thread was started in 2007 right after Dittrich and Thiemann had challenged the Loop result about discrete spectrum of the area and volume operators. I would say a few things differently now from what I said in 2007.
At that time my intuition about that was wavering off track a little. At that time people were using Spinfoam to calculate with but Spinfoam had not yet shown that it had a place for Immirzi parameter and could confirm the earlier results about spectra.

Since then the spinfoam approach has strongly confirmed the key results of earlier (1990s) LQG. So now there is no dissonance.

Now when one compares Loop on the one hand and Reuter AsymSafe and Loll Triangulations on the other hand, it seems to me that on balance Loop now has more going for it. All these approaches are providing hints and clues about what is going on with geometry at very small scale, so you don't actually want to be "picking winners", but with Modesto's result Loop has the dimensional reduction at small scale which for a while it looked like only Reuter's and Loll's approaches had. Less dissonance there too. And Loop let's you do a lot more calculation. Have to see about running coupling constants. Rovelli's lectures in September may resolve that.

 

[Rebel]

Since then the spinfoam approach has strongly confirmed the key results of earlier (1990s) LQG. So now there is no dissonance.

Now when one compares Loop on the one hand and Reuter AsymSafe and Loll Triangulations on the other hand, it seems to me that on balance Loop now has more going for it. All these approaches are providing hints and clues about what is going on with geometry at very small scale, so you don't actually want to be "picking winners", but with Modesto's result Loop has the dimensional reduction at small scale which for a while it looked like only Reuter's and Loll's approaches had. Less dissonance there too. And Loop let's you do a lot more calculation. Have to see about running coupling constants. Rovelli's lectures in September may resolve that.

Oh I see, I did not saw when those posts were put and I feel embarrassed.

But now that it comes... is the question of volume operator settled by now in LQG? I don't see much about this, maybe some indicators from Ashtekar and Rovelli a decade ago, but guess they obtained different results. Sorry for don't being that updated. Also I heard once Modesto saying that this result about that paper of 2005 with Rovelli about running scale dimensions is what they wanted to do from the start, something like if Rovelli wanted to have a fractal-like structure for dimensions. And guess that was motivated exactly to relate with Loll's work, and don't know if ultimately he wanted to obtain something more for arguing more about time existence as some papers of him (an attitude about time in QG that as far as I know smolin has changed), but things did not resulted in the dimensions they wanted to obtain at first.

 

[marcus]

I for one don't see any cause for embarrassment! I'm glad you renewed activity on this thread. It's an important point, that a lot of people seem not to realize.
No one has ever been able to show that (in 4D) LQG leads to any sort of Lorentz violation, or even any deformation of Lorentz invariance (like DSR which has no distinguished frame of reference, all observers see the same speed c.) They've given loose arguments, and theyve thought they could show it, but it has never "stuck".

Loop researchers have tried hard to find a way to derive some kind of bending of Lorentz and they so far can't do it. It would be great if they could, give something to test observationally. So DSR is a separate thing that has to be tested separately.

I'm glad you brought the thread up.

The only reason I mentioned that my earlier contribution to it was 2007 is that my perspective has changed some. Just wanted to update a bit.

 

[tom.stoer]

The argumentation relies heavily on the states; if we want to discuss Lorentz symmetry, the operator algebra itself has to be discussed

A few comments are in order:
- the angular momentum is a perfect example where a continuous symmetry leads to discrete eigenvalues
- a symmetry is manifest after quantization if the operator algebra remains valid = if there are no anomalies
- the area and volume operator and not diff. invariant, so I can't say if these arguments apply (this is a major difference to angular momentum!)
- in the final theory the generator of infinitesimal diffeomorphisms does no longer exist, so comparison to to angular is not straightforward
- I don't know if the algrabra of constraints itself closes or if it closes only weakly = within the physical Hilbert space; therefore the question of anomalies is not clear to me

 

[Eelco]

A few comments are in order:
- the angular momentum is a perfect example where a continuous symmetry leads to discrete eigenvalues

To which id like to add: a closed universe only permits discrete eigenvalues.

Continuous spectra, move over.

 

[marcus]

...
- I don't know if the algrabra of constraints itself closes or if it closes only weakly = within the physical Hilbert space; therefore the question of anomalies is not clear to me

It sounds like you are talking about the old LQG. The canonical approach.
I think the best survey of the current LQG situation is the talk that Rovelli gave to Strings 2008 at Geneva. He also re-wrote his Living Reviews article last year.

Have you watched the Strings 2008 talk?

I guess the situation could seem a bit confusing---there was a little "revolution" in 2007, when a spinfoam formulation appeared which preserved all the kinematic features of canonical (oldstyle) LQG. So one could throw out the old canonical approach---which few people have worked on in recent years---and move completely to the path integral or covariant approach.
I haven't seen Rovelli or anyone else call it a "revolution" (which would sound a bit pompous and self-important) but it has been referred to as a "new look".
There was a redefinition of the field/program in or around 2007. One could call it a "restyling" or a "make-over"---a bit lighter and less solemn sounding than "revolution".

I'm not sure how much of this you already know. I would have assumed you know the whole story but I'll recap---when Rovelli gives his introduction to LQG to the strings 2008 audience he gives a brief account of oldstyle LQG including the area and volume operator results (which are kinematic) up to but not including dynamics. Then he says "now we come to dynamics, you can forget everything I just told you, what follows is a completely new derivation of LQG".

That is, as he defines the program, LQG is no longer based on the canonical formalism---it is entirely based on path integral. This is what most of the Loop community seems to have wanted for the past 10 years, since they almost all worked on spinfoam and the related group field theory (see Oriti's papers for explanation of GFT). What enabled the desired switch was in 2007 a new version of spinfoam was found to contain the Immirzi parameter and to MATCH the oldstyle canonical results (that were purely kinematic.) They found in 2007 that one could move over to a different definition of LQG without sacrificing anything valuable of the earlier results.

If you look at the LQG papers from 2006 onwards that study the graviton propagator and the large-scale limit you will see that they are all using spinfoam (and related group field theory) methods---the active research on the key questions is essentially all path integral. Perhaps this is partly sociological: the majority of the young researchers in the field have come out of Marseille. But I think that the more important reason is that Thiemann's heroic effort to save the canonical dynamics did not succeed, they might come back to it after the path integral approach is worked out.

Another element to the picture is a further new look hinted at in Rovelli's course description for his lecture series for the Corfu school. If that gambit succeeds it will further re-define LQG, if it fails then they will continue with the redefinition as of 2007.
You probably have the links, but in case anyone wants them I will list a few.
Strings 2008 intro to LQG video:
http://cdsweb.cern.ch/record/1121957?ln=en
Slides:
http://indico.cern.ch/getFile.py/access?contribId=30&resId=0&materialId=slides&confId=21917

Page of helpful excerpts from recent seminar talk quoted here:
https://www.physicsforums.com/showthread.php?p=2236221#post2236221

Abstracts for upcoming Corfu school (September 2009):
http://www.maths.nottingham.ac.uk/qg/CorfuSS.html

2008 Living Reviews article on LQG:
http://relativity.livingreviews.org/Articles/lrr-2008-5/

 

[atyy]

Yes, I am totally confused - I've often read spin foams is supposed to reproduce the results of the canonical approach. If spin foams is new, how is it motivated? Or if spin foams is equivalent to the canonical approach, am I allowed to use the old motivation?

 

[marcus]

Yes, I am totally confused - I've often read spin foams is supposed to reproduce the results of the canonical approach. If spin foams is new, how is it motivated? Or if spin foams is equivalent to the canonical approach, am I allowed to use the old motivation?

I'm not entirely sure about the dates, and as I'm just a bystander watching QG as a whole I can't speak for the experts. As I recall the canonical approach failed around 1996 or 1997. Lewandowski was one of the authors that pointed out something wrong. The kinematics were successful but a roadblock was encountered at the dynamics.

The spinfoam or path integral approach was introduced around 1998. It is rather similar to Loll's CDT, which also appeared (but only at the 2D spacetime level) in 1998.
I think it was Baez who proposed the spinfoam approach. You can probably find motivation in his original article. It is an obvious thing to do because it is covariant. Treats spacetime geometry, as a path or trajectory. You may remember Stephen Hawking had already been pushing "Euclidean QG" which is also a path integral approach. There was no need for LQG path integral to have a special motivation.

The canonical approach did not have a representation of spacetime, it treated a slice.
Instead of a Hamiltonian operator you had a Hamiltonian constraint. It is possible to justify this, and motivate it, but it is really the canonical approach that takes special motivation.

So from 1998 to 2007 the majority of LQG research was spinfoam or path integral (that's my impression based on familiarity with the literature, not on hard statistics).

However the pre-1998 development of canonical (to a kinematic descriptive level) had produced certain valuable results (the Loop cosmology spinoff, replacing bang by bounce, the geometric operators discrete spectra, the interesting Immirzi parameter, the study of black hole entropy...all the pre-1998 stuff).

Note that Loll still does not have anything to speak of on black holes, or pre-bang cosmology, or area and volume operator spectra.

So they couldn't just throw out canonical pre-1998 LQG. So for roughly 10 years there was an uneasy mix. Until 2007 it did not seem as if spinfoam was compatible with pre-1998. The researcher's intuition was that it would turn out to be compatible, that it really was the same theory just two different ways of calculating. But they did not have that compatibility down on paper. Then in 2007 a slight change was made in the spinfoam model and suddenly they had compatibility. So it was time to "go public" or "come out" with the spin foam approach and make that the official basis of LQG.

This is what Rovelli did at Strings 2008---one could say he "rewrote the book" on LQG.

There may be a further change in store---look at the abstract of his lecture series at the Corfu school. It mentions scale dependence. That sounds like running couplings. How can this be compatible? But Modesto showed LQG dimensional reduction based on the discreteness of the geometric operator (area etc) spectra. And Carlip discussed this at his Planck Scale talk. How could that all work out?

Atyy, you say "...if spinfoam is equivalent to the canonical..." It is not equivalent it is more and better. It appears more complete because of having dynamics. The current work on the largescale limit is all based on spinfoam. Like Rovelli said, you can forget the canonical approach (for all practical purposes.)

Spinfoam is not equivalent---that would be very limiting and restrictive. It reproduces some valuable earlier results. It matches canonical in the limited things that canonical achieved. But it does other stuff besides.

Spin networks are still a good intuituive tool. In a spinfoam the initial and final states of geometry are, in fact, spin networks. The spinfoam just describes an evolution from initial to final state, and gives a way to calculate an amplitude of that evolution happening.
A spinfoam is somewhat like a Feynman diagram of geometry.
The initial spin network describes the geometry going in, the final spin network describes the geometry coming out. So spin networks are still good to think with. But you don't have to think of them only in the frozen-time canonical Hamiltonian constraint context.

I think the clearest way to think about all these methods of calculation (triangulation, Regge, spinfoam, spinnetwork...) is what is sketched in the quotes in this post:
https://www.physicsforums.com/showthread.php?p=2236221#post2236221

 

[tom.stoer]

Marcus,

I know some of the discussion and results rearding spin foams (nevertheless: thanks for the hints).

I agree that the spin foam models are much more promising in deriving graviton properties, IR limit etc. Nevertheless my expectation is that a path integral must be (for the very definition) be based on an Hamiltonian. In order to construct the path integral you need operators and states. Otherwise you can motivate the path integral only via reproducing the correct results - but in our case this is the long distance limit, only.

Therefore I don't think that you simply can forget about all problems you had in the canonical approach. They are still there; you are not able to derive the correct Hibert space with Hamiltonian (you write down a space of states plus some interaction and - wow - it works - but why?)

Sooner or later the community should provide the missing link (compare it to QCD: in the early ninetees a lot of people startet to work on the canonical approach; it was not only to derive new physics, but to have a more solid basis for the whole theory)

 

[marcus]

... Nevertheless my expectation is that a path integral must be (for the very definition) be based on an Hamiltonian. In order to construct the path integral you need operators and states...

Yes, and perhaps time as well. To get a really satisfactory path integral of the kind we know and love one should probably have time.
Or there should be a generalized path integral idea.
I agree that they still have a whole lot to do. All the work is unfinished. And the current research papers are hard to read. Here is a sample. I don't want to say "read this paper"--I have only chewed on little bits of it myself, but I think it probably represents where we are currently in the development of a path integral QG approach.

Here is just the abstract of Freidel-Conrady http://arxiv.org/abs/0809.2280
==quote Freidel abstract==
On the semiclassical limit of 4d spin foam models
We study the semiclassical properties of the Riemannian spin foam models with Immirzi parameter that are constructed via coherent states. We show that in the semiclassical limit the quantum spin foam amplitudes of an arbitrary triangulation are exponentially suppressed, if the face spins do not correspond to a discrete geometry. When they do arise from a geometry, the amplitudes reduce to the exponential of i times the Regge action. Remarkably, the dependence on the Immirzi parameter disappears in this limit.
==endquote==
You can see that they are trying to get the spinfoam amplitudes to be e^iAction
or what you might like to see. It is work in progress. I would not urge anyone to read the whole paper but only to be aware of this as a research direction.

There is also Freidel-Conrady http://arxiv.org/abs/0806.4640
==quote Freidel abstract==
Path integral representation of spin foam models of 4d gravity
We give a unified description of all recent spin foam models introduced by Engle, Livine, Pereira and Rovelli (ELPR) and by Freidel and Krasnov (FK). We show that the FK models are, for all values of the Immirzi parameter, equivalent to path integrals of a discrete theory and we provide an explicit formula for the associated actions...
=endquote==
Freidel is in a way representative of the current status---the Marcel Grossmann 12 invited him to give the plenary LQG talk (Paris 2009) and the GRG 18 invited him to give it (Sydney 2007) just as they had Renate Loll give the CDT plenary. These are the big triennial international meetings on general relativity and gravitation. (MG 11 [Berlin 2006] had Ashtekar do the Loop plenary, and GRG 19 [Mexico 2010] will have Rovelli do it. So Freidel was like filling in for them.) One can think of his work as somewhat indicative of where the program is. I admit it is hard to read, but here is another fragment to mull over, of a more speculative character.

This is from page 28 of Freidel-Conrady http://arxiv.org/abs/0902.0351
==quote Freidel conclusions 0902.0351==
... Livine and Speziale’s coherent state method: namely, the attempt to express quantum states in terms of geometrical quantities. ...we go yet another step in this direction: now the states are not only labelled by spins and normal vectors, but, in addition, these normal vectors close, so they truly define a geometrical tetrahedron. In this sense, these states are closer to the notion of geometry than the spin network intertwiners, and one is more justified to call them states of quantum geometry.
We believe that the use of such states could have various advantages in spin foam models and also in canonical loop quantum gravity. It could give access to perturbation theory and Feynman diagrams, since the amplitude is now conveniently expressed as a function of the on–shell background and an “off–shell parameter”. Thus, the path integral resembles a lattice path integral over a tetrad field, a connection and further parameters, and it seems conceivable that one can derive Feynman diagrams in a similar way as in lattice gauge theory.
The methods developed in our paper should also shed new light on the coherent states of canonical loop quantum gravity [37]. ...
==endquote==

It seems that the direction of the program is somehat like what you suggest in your post, namely after the spinfoam approach is worked out--if it can be satisfactorily worked out--then one could try to fit it together with a canonical LQG version. But in this process canonical LQG might be changed. You can see Freidel speculatively hinting at this, in the fragment quoted here.

 

[tom.stoer]

... To get a really satisfactory path integral ... one should probably have time.

I agree; you need time to define the propagation - and need time to do the calculation. Unfortunately for the former you can go to a dt=0 limit, whereas for the latter one there seems to be a natural limit of months or even years which cannot be reduced further.

If you look at the canonical approach it compares quite well to formulations of canonical non-abelian quantum gauge theories. The main difference (and difficulty) is the construction of a well-defined hamiltonian. The definition of the Hamiltonian requires to introduce a granular structure which carries the nodes of the spin networks. In the "continuum limit" the resulting interactions should be unique "moves" within a spin network, but unfortunately this is not the case. In addition the SU(2) representations used in the spin networks cannot be deduced uniquely from this approach.

These two problems are addressed via the "new vertex"; it seems that in the long distance limit this new vertex reproduces "gravitons propagating on flat space". So this is a kind of consistency check, but not a unique construction (or I missed this point in Rovelli's papers).

The interesting point is that the non-uniqueness of the Hamiltonian carries over to the path integrals and appears as a non-unique vertex. So by fixing the vertex you have fixed the Hamiltonian as well. The different vertices seem to define a set of candidates of path integrals for quantum gravity. There seems to be no principle which tells you upfront which vertex you have to chose. In the canocical approach it's quite different. There it seems that if you are able to define the quantization and the limiting procedure correctly, you get the correct Hamiltonian automatically.

So in one approach the non-uniqueness is due to a lack of mathematical rigour, in the other approach it seems to be a basic principle of choice. The canocical approach is more restrictive but still incomplete