Short question about diffeomorphism invariance

[nrqed]

I am posting my question in this forum because it is about a basic conceptual aspect of LQG discussed in Rovelli's book Quantum Gravity.
He makes the following statement on page 67 (here, "e" refers to the vierbein):

Now, if e is a solution of the equations of motion, and if the equations of motion are generally covariant, then \tilde{e} is also a solution of the equations of motion. This is because of the relation between active diffeomorphisms and changes of coordinates: we can always find two different coordinate systems on M, say x and y, such that the function e^I_{\mu} (x) that represents e in the coordinate system x is the same function as the function \tilde{e}^I{\mu}(y) in the coordinate system y. Since the equations of motion are in the same in the two coordinates, the fact that this function satisfies The Einstein equations implies that e as well as \tilde{e} are physical solutions.

I do not understand the part in boldface. First, he means that the *functional form* of e and \tilde{e} is the same, when he says that the two functions are equal, right? (which is different from saying e(x) = \tilde{e}(y(x))).

If that's the case, then I don't follow the logic of the argument. First, I don't see in what way the relation with active diffeomorphisms plays a role...is he assuming that the theory is invariant under active diffeomorphisms? It seems to me that one only needs to use the freedom to make changes of coordinates to obtain the result.

A second question is:if we had a scalar function f instead of a one-form like e, then it seemes to me that we could not make the argument that we can always find a different coordinate system such that f and f' can be made equal. Am I missing something?


Thanks in advance

 

[marcus]

A passive diffeo (or change of coords) is a map R^d --> R^d
(see top of page 64 for additional details, I won't give the full definitions)

An active diffeo is a map M --> M (other details at bottom of page 63)

It may seem strange to you that he should be so careful about explaining the fact that

invariance under change of coords implies invariance under active diffeo. But there is something to prove because they are different animals.

So you might get confused just because he is taking his time and going slow.

It is the invariance under (active) diffeo that eventually (on page 68) implies that points in spacetime have no meaning. Because active diffeos stir and smoosh the whole manifold M around and can take point A to point B. You have looked at the "Hole Argument" on page 68? There's that famous Einstein quote from 1916 that the principle of general covariance deprives time and space of the last remnant of physical reality.

So coordinate change (which doesn't move points of the manifold M) looks harmless.
But what Rovelli is explaining, where you asked about it, is that you cannot buy coordinate change invariance without also getting full diffeomorphism invariance.

Formally the two kinds of maps look different. One is M --> M and the other is R^d --> R^d. So there is something to prove, even though it may be intuitively obvious to you.

Should we go over the argument? Paraphrase it? Maybe some other poster will step in. Otherwise I will tomorrow (bed-time now). Or maybe it is clear already nrqed?

 

[nrqed]

A passive diffeo (or change of coords) is a map R^d --> R^d
(see top of page 64 for additional details, I won't give the full definitions)

An active diffeo is a map M --> M (other details at bottom of page 63)

It may seem strange to you that he should be so careful about explaining the fact that

invariance under change of coords implies invariance under active diffeo. But there is something to prove because they are different animals.

So you might get confused just because he is taking his time and going slow.

It is the invariance under (active) diffeo that eventually (on page 68) implies that points in spacetime have no meaning. Because active diffeos stir and smoosh the whole manifold M around and can take point A to point B. You have looked at the "Hole Argument" on page 68? There's that famous Einstein quote from 1916 that the principle of general covariance deprives time and space of the last remnant of physical reality.

So coordinate change (which doesn't move points of the manifold M) looks harmless.
But what Rovelli is explaining, where you asked about it, is that you cannot buy coordinate change invariance without also getting full diffeomorphism invariance.

Formally the two kinds of maps look different. One is M --> M and the other is R^d --> R^d. So there is something to prove, even though it may be intuitively obvious to you.

Should we go over the argument? Paraphrase it? Maybe some other poster will step in. Otherwise I will tomorrow (bed-time now). Or maybe it is clear already nrqed?

Thank you Marcus for your time. It is appreciated.

I do understand the overall idea, which you summarized nicely.
As you said they are different animals so indeed there is something to prove. It is the proof that gives me trouble. The cornerstone of the proof reside in the step that I boldfaced in my OP. It is that sentence that I am trying to understand. I guess I have two questions:

By saying ''the function e and \tilde{e} are the same", what does he mean? I think he means that the functional forms are the same (and not the value evaluated at the same point is the same)

How does one see that one can always have two different coordinate systems that give
e = \tilde{e} ?

Thanks again

 

[marcus]

Let's see if we can prove it. You can check me to see if my notation is understandable and whether I have it right or not. We may want to make an additional assumption like the active diffeo stays within the coordinate patch, just for convenience. Let the manifold be M and let systems of coordinates be written with letters like x or y

Here are two different systems of coordinates
x: M --> R^d
y: M --> R^d

We have a diffeo f: M --> M
and we want to find a change of coordinates, from x to y, that will undo the effect of the diffeo on what you call the functional form of e(m)

In x coords, the functional form of e(m) is e(x^-1(s))
You take a point s in R^d and map it up to point m in the manifold and then find e(m)

The diffeo has this effect, e(f(m)). You moosh with the diffeo first and then do e.

What we want is a new coord system y such that the functional form of e(f(m)) is the same as the old functional form.

e(f(y^-1(s)) should = e(x^-1(s)) for any point s in some appropriate neighborhood in R^d

I think that means that given x and the diffeo f, we have to find coords y such that
f(y^-1(s)) = x^-1(s)

This is just half done. I have to go temporarily back later.

 

[nrqed]

Let's see if we can prove it. You can check me to see if my notation is understandable and whether I have it right or not. We may want to make an additional assumption like the active diffeo stays within the coordinate patch, just for convenience. Let the manifold be M and let systems of coordinates be written with letters like x or y

Here are two different systems of coordinates
x: M --> R^d
y: M --> R^d

We have a diffeo f: M --> M
and we want to find a change of coordinates, from x to y, that will undo the effect of the diffeo on what you call the functional form of e(m)

In x coords, the functional form of e(m) is e(x^-1(s))
You take a point s in R^d and map it up to point m in the manifold and then find e(m)

Hi Marcus,

Thanks for your help. I understand your approach, it fits well with my understanding of the problem. But there is one point that I need to clarify. It seems to me that when Rovelli is talking about the functional form of the function e(x), he is talking about something that maps a point in R^d to the space of one-forms. So in your notation, we should be talking about e(s) and not e(m). Do you see what I mean?

 

[atyy]

I don't understand Rovelli's general argument, but the particular part you mention is also in Wald, beginning on the bottom of p438.

 

[nrqed]

I don't understand Rovelli's general argument, but the particular part you mention is also in Wald, beginning on the bottom of p438.

Thank you! I will read that.

 

[atyy]

Thank you! I will read that.

Also Carroll's notes http://nedwww.ipac.caltech.edu/level5/March01/Carroll3/Carroll5.html , the part beginning at "We are now in a position to explain the relationship between diffeomorphisms and coordinate transformations." (I haven't read this part carefully, but I quick glance seems to show it's the same argument as Wald's)

 

[nrqed]

Also Carroll's notes http://nedwww.ipac.caltech.edu/level5/March01/Carroll3/Carroll5.html , the part beginning at "We are now in a position to explain the relationship between diffeomorphisms and coordinate transformations." (I haven't read this part carefully, but I quick glance seems to show it's the same argument as Wald's)

Thanks again! Sounds well explained.

I am still interested in understanding Rovelli's argument.

Thanks,

Patrick

 

[marcus]

whew! Atyy, thanks so much for getting me off the hook! I had to be out for much of today and got distracted. Very glad you took over.
AFAICS Rovelli does not prove the relation between diffeo and coord transformation. That is more appropriate to a GR textbook. He motivates it with an example (French and English coordinates for a map of air temperature) and invokes it, leading up to the "Hole argument" on page 68.

The whole thing is motivational and introductory, giving perspective. If you buy that gravity theory is invariant under coord change then you buy that it is diffeo invariant. Then (what he really wants to show) if it is diffeo invariant then the points of spacetime have no physical meaning---no objective reality. Something Einstein pointed out in 1916, quotes from a paper and a letter. So then we need a new formulation of physics in which there is no underlying continuum for fields to be defined on.

The discussion is driving towards the two short paragraphs on page 75 that conclude that section of the book.
If you think of what is talked about on pages 65-69 as leading up to that, it gets clearer I think.

 

[atyy]

I am still interested in understanding Rovelli's argument.

With respect to Rovelli's argument (which I don't understand), there are some other sources which deal with the issue, which I have found helpful.

One is MTW's statement that although general covariance was a founding principle of GR, it is in fact not, and the founding principle is "no prior geometry".

The distinction between general covariance and "no prior geometry" is found in Mattingly's http://relativity.livingreviews.org/Articles/lrr-2005-5/ : "There are three general principles in general relativity relevant to Lorentz violation: general covariance (which implies both passive and active diffeomorphism invariance [247]), the equivalence principle, and lack of prior geometry. As we saw in Section 2, general covariance is automatically a property of an appropriately formulated Lorentz violating theory, even in flat space. The fate of the equivalence principle we deal with below in Section 2.5. The last principle, lack of prior geometry, is simply a statement that the metric is a dynamical object on the same level as any other field."

An extensive discussion is given in Giulini's http://arxiv.org/abs/gr-qc/0603087.

 

[nrqed]

With respect to Rovelli's argument (which I don't understand), there are some other sources which deal with the issue, which I have found helpful.

One is MTW's statement that although general covariance was a founding principle of GR, it is in fact not, and the founding principle is "no prior geometry".

The distinction between general covariance and "no prior geometry" is found in Mattingly's http://relativity.livingreviews.org/Articles/lrr-2005-5/ : "There are three general principles in general relativity relevant to Lorentz violation: general covariance (which implies both passive and active diffeomorphism invariance [247]), the equivalence principle, and lack of prior geometry. As we saw in Section 2, general covariance is automatically a property of an appropriately formulated Lorentz violating theory, even in flat space. The fate of the equivalence principle we deal with below in Section 2.5. The last principle, lack of prior geometry, is simply a statement that the metric is a dynamical object on the same level as any other field."

An extensive discussion is given in Giulini's http://arxiv.org/abs/gr-qc/0603087.

Thanks again Atyy for the very interesting references!


Thanks to you and Marcus for the exchange. I figured out what was bugging me.
It was not the relation between coordinate changes and active diffeomorphisms, it was the meaning of general covariance. Now it's clear (I think!)

Thanks!

Patrick

 

[nrqed]

EDIT: For some reason, when I put a tilde on the new function T, it did not show up. SO I used T' instead.



Unfortunately, I now realize that there is still one point that is unclear to me.
However, I have understood enough to make my question well-defined, so hopefully it will be much easier to answer it.

My confusion comes from the following point. A coordinate system is a map from M to R^d. A scalar function T is a map from M to R. So they are not, essentially, different objects. My problem comes from the fact that they are treated differently under an active diffeomorphism.

What I mean is this: Rovelli states clearly that under an active diffeomorphism, the function T(P) changes (where P is a point in the manifold). Therefore, if we see an active diffeomorphism as dragging the points of the manifold to new positions, the scalar function does not "follow", it stays where it initially was, so that we get a new function T'(P). (of course, in his example of winds and temperature, he does not drag the manifold, he drags the temperature field instead so it is the inverse to what I just described but the key point is the same: the points in the manifodl get moved relative to the scalar function).

Now I can state the source of my confusion. Consider now including a coordinate system on the manifold, that is a map X: M \rightarrow R^d, before applying the active diffeomorphism. We can now introduce a temperature function t(x) that maps R^d to R.

Ok, now we apply the active diffeomorphism. The scalar function T(P) moves relative to the manifold, as we said before. The question is: what happens to the coordinate system X(P)?

a) At first, it seems as if it should behave the same way as the scalar function T(P).
However, if this was the case, the induced function t(x) would not change of functional form!

b) The coordinate system is dragged along the manifold. In other words, each point P remains
asisgned to the same coordinate x^\mu. In that case, the scalar function
t(x) will change of functional form since a given coordinate is no longer assign to the same
point in the function. To be more mathematical, what can do here is to assign a coordinate system
to th emanifold *after* the active diffeomorphism and then we use the pullback of this mapping to
the original manifold (before the active diffeomorphism) to get a coordinate system on the original
manifold. In this way any given point is asisgned the same coordinates before and after the diffeomorphism.



It is clear that to get equivalence between active diffeomorphisms and changes of coordinates, we must adopt
the second point of view. But then my question is: why do we treat differently the coordinate system X, which is just
a mapping from the manifold to R^d and the saclar function T, which is a mapping from the manifold
to R? Why the different rules for these two mappings?

I hope my question is clearer now.

Thanks!

Patrick

 

[Finbar]

I fail to see the physical significance of general covariance or diffeomorphisms. Newtonian physics can be expressed as a diffeomorphic invariant theory. What is physically interesting about GR is that there is no a priori geometry; that the geometry is dynamical.

I'm some what confused by the whole active/passive diffeomorphism thing. What exactly is the difference?

 

[atyy]

I fail to see the physical significance of general covariance or diffeomorphisms. Newtonian physics can be expressed as a diffeomorphic invariant theory. What is physically interesting about GR is that there is no a priori geometry; that the geometry is dynamical.

I'm some what confused by the whole active/passive diffeomorphism thing. What exactly is the difference?

How about as applied to the field in the action (rather than the equations of motion written in generally covariant form) - so Maxwell's equations would not come from a genrally covariant action, whereas the Einstein equations do?

 

[atyy]

I hope my question is clearer now.

I fail to see the physical significance of general covariance or diffeomorphisms. Newtonian physics can be expressed as a diffeomorphic invariant theory. What is physically interesting about GR is that there is no a priori geometry; that the geometry is dynamical.

I'm some what confused by the whole active/passive diffeomorphism thing. What exactly is the difference?

Is Rovelli's point really that active diff = passive diff? I think he actually wants to say active diff is *not* equivalent to passive diff, and GR is invariant under active and passive diff, whereas other theories are invariant only under passive diff.

I believe he is using the term "active diff" *differently* from Wald and Caroll - whose point is simply that you can use a diffeomorphism to change coordinates. Rovelli's point is really "no prior geometry", and his definition of an active diff is a diff on the "dynamical fields" only, whereas Wald and Carroll's active diff is a diff on everything that is geometrical including the metric, even if it is not dynamical. In SR, a Rovelli "active diff" will move everything except the metric, since the metric is not dynamical, so the physics will be changed, since the metric is an essential part of the theory, just that it's not dynamical. In GR, a Rovelli "active diff" will move everything including the metric, since the metric is dynamical, and so the physics will not change. (In short, I agree with Finbar.)

I think Rovelli said it much more clearly here (maybe he read MTW or Finbar between 2003 and 2008 ): "Conventional field theories are not invariant under a diffeomorphism acting on the dynamical fields. (Every field theory, suitably formulated, is trivially invariant under a diffeomorphism acting on everything.)" http://relativity.livingreviews.org/Articles/lrr-2008-5/

So why is Rovelli using such confusing terms (apart from being confused - he really wants to do Asymptotic Safety, not LQG). I think it may be that although general covariance of the dynamical equations is not a principle solely of GR (since all theories can be written in generally covariant form, even SR and Newton) - the general covariance of the action is a distinguishing feature of GR (this is one the key ingredients in the definition of Asymptotic Safety).

 

[nrqed]

Is Rovelli's point really that active diff = passive diff? I think he actually wants to say active diff is *not* equivalent to passive diff, and GR is invariant under active and passive diff, whereas other theories are invariant only under passive diff.

My interpretation of what he is saying in his book is that active diffeomorphism *is* equivalent to passive diff. That, in itself, does not say anything special about GR or physics.
*Then*, the fact that Einstein's equations are generally covariant (that's the physical input)
implies that they are invariant under passive diffeomorphisms. Hence they are invariant under active diffeomorphisms, hence the points in the manifold have no physical meaning. That's my understanding of his explanations.

So why is Rovelli using such confusing terms (apart from being confused - he really wants to do Asymptotic Safety, not LQG). I think it may be that although general covariance of the dynamical equations is not a principle solely of GR (since all theories can be written in generally covariant form, even SR and Newton)

Really? If we write Newton's equations in a given coordinate system and then do a change of coordinates, the functional form will no longer be the same. I mean, that's why we have things like the Coriolis force in a rotating frame. Maybe I am missing something!

 

[Fra]

I find this discussion closely related to https://www.physicsforums.com/showthread.php?t=418467.

Loosely speaking, to me the conceptual meaning of ANY so called "passive transformation" be it poincare, diff or anything else, is in the context of a given observer simply a "relabelling" of events. So if you acknowledge that that the labes(coordinates, event index etc) are just for bookkeeping and that the choice of label has no physical significance, since type of invariance is somehow obvious.

What I associate with "active transformations" rather means, conceptually that you are transforming the observer (or the event manifold (and not just the lables).

The connection is that mathematically the same transformations defines the set of coordiante systems, as defines the set of observer.

IMO, the "physics" lies in the statement that the set of possible observers are generated exactly by the same mathematical transformation that relabels the events (the passive ones).

I think this statement can be discussed and there are issues with it, because from the inference perspective, one certainly wonder where is the information about these transformations stored, and what physical process allows it's inference?

I think this question is ultimately related to what's measurable and not, and who is measuring it. This is less problemativ in classic GR, but I think it's nontrivial when you take measurements and information also about transformations more serious than Einstein did in the "classical world".

I definitely think that one should make a clear distinction between passive and active transformations, beucase it's not clear to me that the coincidental mathematical similarity is quite correct and will survive the transition to QG.

Edit: the point one is tempted to make is that the "choice of observer" is also just a relabelling in the extended sense. But there is a problem with that, in the measurement perspective, since the observer is central. The vision of an observer independent transformation that fomr a birds perspective generates all possible observers, is not unproblematic. So the choice of observer in a seriously constructed intrinsic measurement theory, can IMHO not quite be put on par with the "choice of event LABELS".

This disitincion wasn't there for Einstein as he was looking for a deterministic classical model.

/Fredrik

 

[Fra]

hence the points in the manifold have no physical meaning.

I don't like how this is usually put. It's like saying that the observer has no physical meaning. While that makes sense in GR; please explain how it makes sense for a measurement theory.

I personally think it's more correct to say the point on the manifold have no objective(observer-independent) meaning.

The problem with Rovelli is that he ASSUMES that there is some objective meaning to the set of possible observers, and the generating transformations. Except for the mathematics I don't see how this is a inferrable statement, therefor I think it shold be reformulated.

/Fredrk

 

[Finbar]

So if I begin with one set of coordinates relating to some observer a passive diffeomorphism
would simply re-label the events as seen by this observer. On the other hand an active diffeomorphism is like moving to a coordinate system associated to a different observer?

This seems to make sense since under both diffeomorphisms the objective physics is still invariant. However for the active case we seem to have changed the order of events or possibly the gravitational field(metric).


The other idea is that diffeomorphisms are active because the metric is a dynamical field. This makes sense also but I'm not sure if this need be a prerequisite to define an active diffeomorphism.


With respect to the maxwell equations you can write them in a generally covariant form. For example if you were on a fixed curved space-time.

 

[atyy]

Really? If we write Newton's equations in a given coordinate system and then do a change of coordinates, the functional form will no longer be the same. I mean, that's why we have things like the Coriolis force in a rotating frame. Maybe I am missing something!

This is done in section 12.4 of MTW.

Also helpful, I think, is section 3.2 of Malament's http://arxiv.org/abs/gr-qc/0506065.

 

[nrqed]

This is done in section 12.4 of MTW.

Also helpful, I think, is section 3.2 of Malament's http://arxiv.org/abs/gr-qc/0506065.

Thanks again for the interesting references.


But then I don't understand Rovelli's point. Why is he stressing the general covariance of GR and what point is he making, exactly? Or are you saying that Rovelli's arguments are baseless?


Thanks

 

[atyy]

But then I don't understand Rovelli's point. Why is he stressing the general covariance of GR and what point is he making, exactly? Or are you saying that Rovelli's arguments are baseless?

I've never understood his argument there - perhaps it's right and poorly presented, or perhaps it is wrong, I don't know which - the only thing I understood from his book, is that there is an issue, but as to which the right statements and arguments are I only gathered from other books - I really think he said it much better in his later review, which is exactly what all modern textbooks say:

"Conventional field theories are not invariant under a diffeomorphism acting on the dynamical fields. (Every field theory, suitably formulated, is trivially invariant under a diffeomorphism acting on everything.)" http://relativity.livingreviews.org/...es/lrr-2008-5/

So the statement "Conventional field theories are not invariant under a diffeomorphism acting on the dynamical fields" is equivalent to MTW's "no prior geometry".

And the statement "Every field theory, suitably formulated, is trivially invariant under a diffeomorphism acting on everything" is what Wald and Carroll mean when they say you can use a diffeomorphism to change coordinates.

 

[nrqed]

Is Rovelli's point really that active diff = passive diff? I think he actually wants to say active diff is *not* equivalent to passive diff, and GR is invariant under active and passive diff, whereas other theories are invariant only under passive diff.

I believe he is using the term "active diff" *differently* from Wald and Caroll - whose point is simply that you can use a diffeomorphism to change coordinates. Rovelli's point is really "no prior geometry", and his definition of an active diff is a diff on the "dynamical fields" only, whereas Wald and Carroll's active diff is a diff on everything that is geometrical including the metric, even if it is not dynamical. In SR, a Rovelli "active diff" will move everything except the metric, since the metric is not dynamical, so the physics will be changed, since the metric is an essential part of the theory, just that it's not dynamical. In GR, a Rovelli "active diff" will move everything including the metric, since the metric is dynamical, and so the physics will not change. (In short, I agree with Finbar.)

That might be the key but could you make this more precise? Rovelli defines an active diffeomorphism as a map from M to M. How would you define mathematically a "dynamical" field and a "nondynamical" field?

I think Rovelli said it much more clearly here (maybe he read MTW or Finbar between 2003 and 2008 ): "Conventional field theories are not invariant under a diffeomorphism acting on the dynamical fields. (Every field theory, suitably formulated, is trivially invariant under a diffeomorphism acting on everything.)" http://relativity.livingreviews.org/Articles/lrr-2008-5/

This may be the answer to the question I posted earlier, when I asked why the coordinate system is dragged with the points when Rovelli is performing an active diffeomorphism while the field T is not (again, he actually drags the field, leaving the manifold unmoved but that's completely equivalent to dragging the manifold and leaving the field there). So I guess that, using your terminology, a "dynamical field" would be something which is *not* dragged together with the manifold under an active diffeomorphism. In that case, a dynamical field gets assigned to a different point in the manifold under such (nontrivial) active diffeomorphism. Of course, if the dyamical fields are dragged with the points in the manifold, nothing has changed and we get a "trivial active diffeomorphism".



I am surprised that the whole terminology is still unclear given how old Einstein's hole argument is!

 

[atyy]

That might be the key but could you make this more precise? Rovelli defines an active diffeomorphism as a map from M to M. How would you define mathematically a "dynamical" field and a "nondynamical" field?

I think I better not discuss Rovelli's 2003 terminology since I've never understood it. Yes, it's not possible to make more than a heuristic distinction between dynamical fields and nondynamical fields - that's what the Giulini article I linked to says.

So GR really cannot be "no prior objects" - it it really is "no prior 4D spacetime geometry".

Should we generalise or not? I think the correct thinking is really not with LQG (as far as motivation goes, even though the formalism and especially the link to group field theory, non-commutative geometry, and condensed matter may be very fruitful). The correct conceptual thinking is stated by Polchinski "There are two possible resolutions. The first is ... a nontrivial UV fixed point [asymptotic safety]. The second is that ... the extrapolation to arbitrarily high energies is incorrect [emergent gravity]" http://books.google.com/books?id=k4ZQ04viGWIC&dq=polchinski+string&source=gbs_navlinks_s

 

[nrqed]

This is done in section 12.4 of MTW.

Also helpful, I think, is section 3.2 of Malament's http://arxiv.org/abs/gr-qc/0506065.

Thanks again for the references. I have read MTW (and section 12.5 as well) and have looked at the paper. I see what you mean, that any physical theory can be cast in a geometric language, but this is not what I had in mind by general covariance, and I have the impression that this is not what Rovelli has in mind either.

What I mean by general covariance is general covariance of the equations in the sense that they must take the same form expressed in any coordinate system. For example, consider
F_x = m a_x. If we go to an accelerated frame and define the force and acceleration in that frame with primes, they will *not* obey F_x' = m a_x'. It is in that sense that I would say that the theory is not generally covariant. If we do a general coordinate transformation in all the quantities appearing in Einstein's equation, the equation obeyed by the prime quantities will be exactly the same as the equation obeyed by the initial variables. This is the key point for Rovelli, I think. This is what he uses to say that if e^I_{\mu}(x) is a solution, then the transformed quantity e^I_{\mu}'(y) is also a solution. This is my understanding of his argument.

Thanks again for all the very useful feedback. It helped me greatly, both by providing me insightful comments and valuable references but also by forcing me to sharpen my questions and my arguments.

 

[atyy]

What I mean by general covariance is general covariance of the equations in the sense that they must take the same form expressed in any coordinate system.

But what is the same "form"? Can't we say that the Coriolis term is in fact present in a non-rotating frame, but it just happens to be zero? I can't provide a detailed example for the Newtonian theory of the top of my head, but if you take Maxwell's equations on flat Minkowski spacetime, write them in Lorentz covariant form with the metric explicitly in the equations, then turn all partial derivatives to covariant derivatives (comma to semicolon rule), wouldn't one get a generally covariant form of the equations? In an arbitrary coordinate system, the Christoffel symbols in the covariant derivatives will be non-zero, but in a Lorentz inertial coordinate system, the Christoffel symbols will be zero.

 

[Haelfix]

See for instance Weinberg "Gravitation" p92 about Newton's laws and general covariance for another view.

The terminology about general (in/co)variance, active vs passive vs mathematical diffeomorphisms vs the equivalence principle has a long, tired and grueling history of debate. It is quite rare to see two authors use the same sets of definitions which is why it is worthwhile to stick to one text lest you end up hopelessly confused.

There is nothing particularly profound that has come out of any of those debates, the physics and mathematics has mostly remained the same and not in dispute, and its just a question of interpretation.

I was taught the way Carrol presents it, which is the more modern way of teaching GR I think and more useful b/c it goes straight into differential geometry and is readily translated into the fiber bundle language and gauge theories..

 

[Fra]

There is nothing particularly profound that has come out of any of those debates, the physics and mathematics has mostly remained the same and not in dispute, and its just a question of interpretation.

I can't help but ask if this is good or bad? QG puzzle has also remained unsolved, but maybe it's a coincidence.

This is similar to the questions of the QM foundations. It's true that in the end, not much has changed beyond interpretations. Maybe they theory needs to change and be reconstructed with the old theory beeing just a limiting case.

Somehow this discussion is about the meaning of symmetry, mathematical vs physical vs inferrable symmetry and what is to be considered as subjec to observations - events or symmetries, or both? This is I think possibly a key point towards QG as well.

Maybe the way GR was constructed, should even the reconstructed in terms of a measurement theory. Then reconsidering some of thes admittedly old arguments, like hole argument in a different setting may provide new light? I think I most probably represent a very small minority but at least anyone ask wonder why thq QG puzzle hasn't made more progress, and wether some of the things we "know" about the structure of physical models need to be changed or reconstructed.

/Fredrik

 

[Fra]

It seems to me the interesting core of this thread isn't at all about GR. It's about symmetry about sets of distinguishable events, and symmetry and sets of distinguishable observers and their relation, and what the physical and observable status is of these symmetries; why a particular symmetry and is it static or evolving?

And what does this mean, in the context of a measurement theory? How does the inference go from observed symmetry of events, to observed symmetry of observers without resorting to disturbing realism?

/Fredrik

 

[nrqed]

But what is the same "form"? Can't we say that the Coriolis term is in fact present in a non-rotating frame, but it just happens to be zero? I can't provide a detailed example for the Newtonian theory of the top of my head, but if you take Maxwell's equations on flat Minkowski spacetime, write them in Lorentz covariant form with the metric explicitly in the equations, then turn all partial derivatives to covariant derivatives (comma to semicolon rule), wouldn't one get a generally covariant form of the equations? In an arbitrary coordinate system, the Christoffel symbols in the covariant derivatives will be non-zero, but in a Lorentz inertial coordinate system, the Christoffel symbols will be zero.

Ok, point well taken.


But then this implies that general covariance is a red herring, right?
So you are saying that Rovelli's discussion is completely incorrect? (I am not being provocative here, I am asking an honest question because I truly want to understand the hole argument and the ''no prior geometry'' argument) What about Einstein's hole argument? Is it a red herring too? Was Einstein wrong too about making a big deal out of general covariance, or does Rovelli misrepresent what happened historically? If the hole argument is not a red herring, what is the correct way to phrase it (not invoking general covariance)?

I guess that what I really want to understand is the argument for no prior geometry. Rovelli seems to rely on general covariance to reach that conclusion. I am willing to accept that general covariance should not play a role in the argument. But then, what *is* the argument?? Some reference cited by Atyy says that no prior geometry is simply a consequence of the metric being a dynamical field. But what is the definition of a dynamical field, and how does that imply that there is no prior geometry?

It seems as if people think of dynamical vs non-dynamical fields as being related to how field behave under an active diffeomorphisms. Is that correct? What is the exact definition?
What Carroll says about active diffeomorphisms does not clarify the issue because he does not talk about how fields transform, he just talks about the transformation of the points in the manifold and of the coordinates (at least in the link that Atyy provided).

I tried to make my questions as specific as possible. Thanks for the feedback.

 

[nrqed]

Let me ask a follow-up question:
The description of the hole argument made on Wikipedia (and I have seen this argument made in other places) says:

Einstein noticed that if the equations of gravity are generally covariant, then the metric cannot be determined uniquely by its sources as a function of the coordinates of spacetime. The argument is obvious: consider a gravitational source, such as the sun. Then there is some gravitational field described by a metric g(r). Now perform a coordinate transformation r r' where r' is the same as r for points which are inside the sun but r' is different from r outside the sun. The coordinate description of the interior of the sun is unaffected by the transformation, but the functional form of the metric for coordinate values outside the sun is changed.

I don't understand the point made here. I mean, of course the functional form of the metric will change since we changed the coordinate system! We could say the same thing about an electric field and a charge distribution, it seems to me. I mean:

"The argument is obvious: consider a charge distribution confined to a certain volume. Then there is some electric field described by E(r). Now perform a coordinate transformation r-> r' where r' is the same as r for points which are inside thcharge distribution but r' is different from r outside. The coordinate description of the interior of the volume containing the charge distribution is unaffected by the transformation, but the functional form of the electric field for coordinate values outside is changed"

So what? This does not imply anything special in E&M, why does it have a profound meaning in GR? (I know that I am missing something important!) Can someone elaborate?

 

[Finbar]

It seems as if people think of dynamical vs non-dynamical fields as being related to how field behave under an active diffeomorphisms. Is that correct? What is the exact definition?

The metric is a dynamical field in GR because you have to solve the Einstein solutions to find it (or rather the equivalence class of metrics). If it was non-dynamical then you would have to put in some metric by hand.


Maybe the point is that in other theories you can get away with writing equations in a non-generally covariant form. But in gravity you can't. Further more when you go to quantum theory symmetries become much more important than in classical theory. For example observables are the generators of symmetries i.e. momentum generates space translations etc. Quite possibly diffeomorphisms generate some kind of observables? Entropy? Also if one wants to quantize the metric, in a path integral say, you need to somehow integrate over only the class of gauge equivalent metrics. So maybe the general point is that when the metric is dynamical you need to worry more about general covariance since it plays an
"active" role. Where as in a theory with a fixed metric general covariance is more passive.

I think a related issue, that Fra is getting at (?), is to split diffeomorphisms into two classes
those which correspond to moving to co-ordinates for which correspond to the same observer's point of view and those for which relate to a different observer. For example in flat space we could move from a non-accelerating frame to an accelerating one and we would observe a different vacuum in each. Again in the quantum theory this distinction is much more meaningful than in classical theory where there is no concept of the vacuum( i.e. no creation or annihilation operators).

We could say the same thing about an electric field and a charge distribution, it seems to me.

Yes, I think the hole argument was just Einstein coming to terms with gauge invariance and gauge fixing. So its not special to gravity.

 

[atyy]

But then this implies that general covariance is a red herring, right?
So you are saying that Rovelli's discussion is completely incorrect? (I am not being provocative here, I am asking an honest question because I truly want to understand the hole argument and the ''no prior geometry'' argument) What about Einstein's hole argument? Is it a red herring too? Was Einstein wrong too about making a big deal out of general covariance, or does Rovelli misrepresent what happened historically? If the hole argument is not a red herring, what is the correct way to phrase it (not invoking general covariance)?

I don't know whether Rovelli's arguemnt is wrong, but I am pretty sure that general covariance (defined here as covariance under an arbitary change of coordinates) is a red herring - all modern textbooks agree on this point.
-MTW: The "no prior geometry" demand actually fathered general relativity, but ... disguised as "general covariance", it also fathered half a century of confusion."
-Weinberg: "It should be stressed that general covariance by itself is empty of physical content" [Weinberg however does define a "Principle of General Covariance" which is meaningful, but it is not general covariance and corresponds to what other people call the "Principle of Equivalence" or "minimal coupling".]
-Carroll: "Since diffeomorphisms are just active coordinate transformations, this is a highbrow way of saying that the theory is coordinate invariant. Although such a statement is true, it is a source of great misunderstanding, for the simple fact that it conveys very little information. Any semi-respectable theory of physics is coordinate invariant, including those based on special relativity or Newtonian mechanics; GR is not unique in this regard."

I guess that what I really want to understand is the argument for no prior geometry. Rovelli seems to rely on general covariance to reach that conclusion. I am willing to accept that general covariance should not play a role in the argument. But then, what *is* the argument?? Some reference cited by Atyy says that no prior geometry is simply a consequence of the metric being a dynamical field. But what is the definition of a dynamical field, and how does that imply that there is no prior geometry?

That the metric is a dynamical field is the definition of no prior geometry. I don't know if a general definition of a dynamical field exists, but if we restrict to theories derivable from a Lagrangian, then I would say a dynamical field is defined by a Lagrangian and its symmetries, with GR's principle being that the Lagrangian contains scalars made from the 4D spacetime curvature tensor. In general, there will be an infinite number of such possible terms, but taking an effective theory point of view, the Einstein-Hilbert action which contains only the lowest order term is good enough at low energies. (Asymptotic Safety is the hypothesis that a Lagrangian containing additional terms made from the 4D spacetime curvature tensor will have a non-trivial fixed point as energy increases; emergent gravity such as string theory is the hypothesis that Lagrangian terms made from the 4D curvtaure tensor are insufficient at high energies, and new degrees of freedom must enter.)

It seems as if people think of dynamical vs non-dynamical fields as being related to how field behave under an active diffeomorphisms. Is that correct? What is the exact definition?
What Carroll says about active diffeomorphisms does not clarify the issue because he does not talk about how fields transform, he just talks about the transformation of the points in the manifold and of the coordinates (at least in the link that Atyy provided).

The most common definition of "active diffeomorphism" is Carroll's and Wald, which is just a "diffeomorphism" and has nothing to do with "no prior geometry". Rovelli has used a differrent definition of "active diffeomorphism" by which he means "no prior geometry" (eg. last paragraph of section 4.1 of http://arxiv.org/abs/gr-qc/9910079). Anyone is allowed to make up their own terminology, no matter how confusing. In his 2003 book, he claims to be using the Carroll and Wald definition of "active diffeomorphism", but the passage makes no sense to me unless he is not using it. I personally think it best to avoid Rovelli's definition.

 

[Finbar]

Ok the following is a personal view.


I don't think that the *best* way to think about gravity is that it is the curvature of space-time. Although this interpretation is the most straight forward mathematically i think it some what misses the real meaning of general relativity. Instead if you think back to Einstein's original motivation, before he realized that Riemannian geometry was the language was elegantly expressed in, he was trying to generalise the idea that all motion is relative. So that all that is physical is the interaction between matter and forces. In this way gravity is just a force interacting with other forces and matter fields. What is special about gravity is that it actually "gauges away" space-time; it removes space-time as being physical in any sense. There is no space-time. I find it much more intuitive to think of gravitational interactions this way; thinking of the force of gravity acting directly on matter. What is physical then is just the relation of the matter to the gravitational force both of which are dynamical and not to any idea of space-time.


Viewing it this way then explains why the idea of space-time being emergent seems to be imbedded, already, in classical general relativity e.g. its relation to thermodynamics. The idea that there is no space-time is already there in the original motivations of Einstein it was just lost in the language of Riemannian geometry.

 

[haushofer]

I also opened a topic about this a while ago:

https://www.physicsforums.com/showthread.php?t=280232

Also, the paper of Norton is mentioned; he wrote a lot about the interpretation of passive and active coordinate transfo's and the history of it.

I still don't understand it completely, and I don't think I've ever met someone who does.

 

[nrqed]

I don't know whether Rovelli's arguemnt is wrong, but I am pretty sure that general covariance (defined here as covariance under an arbitary change of coordinates) is a red herring - all modern textbooks agree on this point.
-MTW: The "no prior geometry" demand actually fathered general relativity, but ... disguised as "general covariance", it also fathered half a century of confusion."
-Weinberg: "It should be stressed that general covariance by itself is empty of physical content" [Weinberg however does define a "Principle of General Covariance" which is meaningful, but it is not general covariance and corresponds to what other people call the "Principle of Equivalence" or "minimal coupling".]
-Carroll: "Since diffeomorphisms are just active coordinate transformations, this is a highbrow way of saying that the theory is coordinate invariant. Although such a statement is true, it is a source of great misunderstanding, for the simple fact that it conveys very little information. Any semi-respectable theory of physics is coordinate invariant, including those based on special relativity or Newtonian mechanics; GR is not unique in this regard."

Thank you. You have pretty much convinced me that general covariance is indeed a red herring.

And thank you for the link to the Rovelli paper you provided. In that paper Rovelli is much clearer than in his book! He makes very explicit his definitions (and does not talk about general covariance. He may have received comments and questions after the publication of his book and rephrased his presentation to make it more transparent).

So let me set aside the discussion in his book and use the paper instead. There he says explicitly that all theories can be made invariant under passive diffeomorphisms (i.e. arbitrary changes of coordinates) but only GR (and not QED, QCD, etc) is invariant under active diffeomorphisms. And he defines clearly active and passive diffeomorphisms. In a passive diffeomorphism, any given point in the manifold is assigned to the same value of the field. In an active diffeomorphism, each point in the manifold is assigned to a *different* value of the field, so the mapping T:M to R (let's consider a scalar field) is changed. He says that GR is special in that it is invariant under those changes, and this is what implies that we cannot assign any significance to the points in the manifold.

I have to say that his presentation in that paper makes a whole lot of sense to me (much more than in his book, although I am sure he meant the same thing).



Now, Carroll says that active diffeo are completely equivalent to coordinates changes.
But in that section he talks only about diffeo of the manifold, there is no field. I would love to know what he does with fields under active diffeomorphisms. If he uses the same definition as Rovelli, then one cannot say anymore that active diffeo are equivalent to coordinates transfos. Maybe he uses a different definition. But then I am curious about how he would call what Rovelli defines as active diffeo.

That the metric is a dynamical field is the definition of no prior geometry.

Yes. I think I used "no prior geometry" incorrectly. I am probably confused about the meaning of the term. It seems to me that there are two issues at hand (maybe they are facets of the same thing and I don't realize it)

a) The metric is dynamical so it does not make sense to talk about spacetime intervals between points on the manifold as being defined prior to solving the dynamics.

b) But it seems that Rovelli (and the hole argument) seems to be implying more than that.
That one cannot assign a physical meaning to the points in the manifold. Isn't it what the hole argument is about?

In other words, that a) implies that we cannot talk about the distance between two points in a manifold until we solve the dynamics. But b) states that we should ot even talk about points in the manifold has having a physical reality in the first place. Am I wrong?

The most common definition of "active diffeomorphism" is Carroll's and Wald, which is just a "diffeomorphism" and has nothing to do with "no prior geometry". Rovelli has used a differrent definition of "active diffeomorphism" by which he means "no prior geometry" (eg. last paragraph of section 4.1 of http://arxiv.org/abs/gr-qc/9910079). Anyone is allowed to make up their own terminology, no matter how confusing. In his 2003 book, he claims to be using the Carroll and Wald definition of "active diffeomorphism", but the passage makes no sense to me unless he is not using it. I personally think it best to avoid Rovelli's definition.

Thank you again for this useful reference. As I said earlier in my post, I can't really compare Carroll's definition of active diffeomorphism with Rovelli's because Carroll did not define how fields (or tensors) transform under his active diffeomorphism. As long as we just look at how the manifold changes (and the coordinates, which are defined through a pullback), then it seems to me that they are saying the same thing. What is needed, however, is to compare how they define the transformation of fields.


Thank you again for all your feedback. It is very much appreciated.

 

[nrqed]

Ok the following is a personal view.


I don't think that the *best* way to think about gravity is that it is the curvature of space-time. Although this interpretation is the most straight forward mathematically i think it some what misses the real meaning of general relativity. Instead if you think back to Einstein's original motivation, before he realized that Riemannian geometry was the language was elegantly expressed in, he was trying to generalise the idea that all motion is relative. So that all that is physical is the interaction between matter and forces. In this way gravity is just a force interacting with other forces and matter fields. What is special about gravity is that it actually "gauges away" space-time; it removes space-time as being physical in any sense. There is no space-time. I find it much more intuitive to think of gravitational interactions this way; thinking of the force of gravity acting directly on matter. What is physical then is just the relation of the matter to the gravitational force both of which are dynamical and not to any idea of space-time.


Viewing it this way then explains why the idea of space-time being emergent seems to be imbedded, already, in classical general relativity e.g. its relation to thermodynamics. The idea that there is no space-time is already there in the original motivations of Einstein it was just lost in the language of Riemannian geometry.

Very interesting. But if there is no spacetime, how do gravitational waves propagate?
How is the information about the presence of the Sun over there conveyed to Saturn, say?
I read the hole argument as saying that we cannot identify spacetime points, i.e. we cannot think of a given spacetime point as having an independent existence. However, I think that the actual manifold exists. Is that your interpretation too?

 

[marcus]

I also opened a topic about this a while ago:

https://www.physicsforums.com/showthread.php?t=280232

Also, the paper of Norton is mentioned; he wrote a lot about the interpretation of passive and active coordinate transfo's and the history of it.

I still don't understand it completely, and I don't think I've ever met someone who does.

That's good, haushofer. Your thread in the GR forum is a good place to go into technical detail about GR.

This thread has focused on a few pages of Rovelli's QG book, and involved different people interpreting and understanding it different ways. Probably at times misinterpreting or simply not understanding. But the thread is too bulky to scrutinize systematically, so I will try to make a clean start with a fresh summary.

Rovelli's conceptual framework here is pretty standard. A less condensed treatment is in his 1999 paper with Marcus Gaul. "LQG and the meaning of diffeomorphism invariance." Really good clear introductory paper. The book gives a compressed version which is more difficult to follow.

MTW's book "Gravitation" makes an excellent observation which amplifies the point Rovelli is making. Diffeomorphism invariance means something different from what people normally understand by general covariance.
Einstein and others saying "g.c." when they meant "d.i." led to a half century of confusion, say from 1915 to 1975.

When you say g.c. people think of coordinate change, which does not change the distances between points or anything essential about the geometry of the manifold itself.
Other physics besides GR can be formulated in a way that allows coordinate change.

A diffeomorphism φ: M --> N is just an invertible smooth map, which
can be between two different manifolds with completely different shape/curvature and it can completely change the distances. Points m m' can get mapped to points n n' with different distance from each other.

The case we are considering here is φ: M --> M, and the same thing is true.
Points m m' m" which are certain distances apart can get mapped to points
φ(m) φ(m') φ(m") which are completely different distances apart.

Whatever physical theory you want to talk about, Newton, Maxwell, QED, QCD it is not diffeomorphism invariant.

But GR is. And people seem to have been slow to realize this.

This is the "anonymity" that caused the "half century of confusion" that MTW talk about. Einstein did not come out and say "diffeomorphism invariance", but he in effect concealed the radicalism of the concept by calling it "covariance". So people kept thinking it was mere coordinate change stuff.

==MTW quote==
Einstein described both demands by a single phrase, "general covariance." The "no prior geometry" demand actually fathered general relativity, but by doing so anonymously, disguised as "general covariance", it also fathered half a century of confusion.
==endquote==

αβγδεζηθικλμνξοπρσςτυφχψωΓΔΘΛΞΠΣΦΨΩ∏∑

 

[marcus]

Very interesting. But if there is no spacetime, how do gravitational waves propagate?
How is the information about the presence of the Sun over there conveyed to Saturn, say?
I read the hole argument as saying that we cannot identify spacetime points, i.e. we cannot think of a given spacetime point as having an independent existence. However, I think that the actual manifold exists. Is that your interpretation too?

Nrqed you are getting to the heart of it!

In GR there is no manifold, no physical objective existing continuum. There is geometry. Waves propagate in the geometry.

People get the idea from popular/simplified accounts of GR that there is a fixed spacetime manifold and the geometry is a metric g_mu_vu on that manifold.

That is not true. In GR the gravitatonal field is an equivalence class of metrics under diffeomorphism.

This effectively denies the manifold (and spacetime points) any objective physical existence.
Einstein remarked on this very explicitly in a couple of 1915 quotes.

The gravitational field is not some particular metric, but rather the whole class of metrics which you can get by mooshing and mapping the manifold around with diffeomorphisms.
And of course the matter gets mooshed around when you do that, as well.

For an online source of the 1915 Einstein quotes, see page 43 of this pdf at a University of Minnesota website
www.tc.umn.edu/~janss011/pdf%20files/Besso-memo.pdf[/URL]

==quote from the source material==
...In the introduction of the paper on the perihelion motion presented on 18 November 1915, Einstein wrote about the assumption of general covariance “by which time and space are robbed of the last trace of objective reality” (“durch welche Zeit und Raum der letzten Spur objektiver Realität beraubt werden,” Einstein 1915b, 831). In a letter to Schlick, he again wrote about general covariance that
“thereby time and space lose the last vestige of physical reality” (“Dadurch verlieren Zeit & Raum den letzter Rest von physikalischer Realität.” Einstein to Moritz Schlick, 14 December 1915 [CPAE 8, Doc. 165]).
==endquote==

In fact it was the GR invariance under diffeomorphisms that led to those radical conclusions. You could moosh and morph a solution and it would still be a solution.
Coord change is different. The manifold and its shape stay the same and you just used different coords.

With coord change you change coordinates but you still have the same rubber sheet :biggrin: but [B]with GR there is no rubber sheet[/B].

 

[atyy]

Thank you. You have pretty much convinced me that general covariance is indeed a red herring.

And thank you for the link to the Rovelli paper you provided. In that paper Rovelli is much clearer than in his book! He makes very explicit his definitions (and does not talk about general covariance. He may have received comments and questions after the publication of his book and rephrased his presentation to make it more transparent).

So let me set aside the discussion in his book and use the paper instead. There he says explicitly that all theories can be made invariant under passive diffeomorphisms (i.e. arbitrary changes of coordinates) but only GR (and not QED, QCD, etc) is invariant under active diffeomorphisms. And he defines clearly active and passive diffeomorphisms. In a passive diffeomorphism, any given point in the manifold is assigned to the same value of the field. In an active diffeomorphism, each point in the manifold is assigned to a *different* value of the field, so the mapping T:M to R (let's consider a scalar field) is changed. He says that GR is special in that it is invariant under those changes, and this is what implies that we cannot assign any significance to the points in the manifold.

I have to say that his presentation in that paper makes a whole lot of sense to me (much more than in his book, although I am sure he meant the same thing).



Now, Carroll says that active diffeo are completely equivalent to coordinates changes.
But in that section he talks only about diffeo of the manifold, there is no field. I would love to know what he does with fields under active diffeomorphisms. If he uses the same definition as Rovelli, then one cannot say anymore that active diffeo are equivalent to coordinates transfos. Maybe he uses a different definition. But then I am curious about how he would call what Rovelli defines as active diffeo.

Quick check on terminology:

1) Which paper of Rovelli's do you mean for definitions http://arxiv.org/abs/gr-qc/9910079 or http://relativity.livingreviews.org/Articles/lrr-2008-5/ ?

2) In special relativity, eg. Maxwell's equations on flat Minkowski spactime, do you consider the metric a field or not?

 

[nrqed]

That's good, haushofer. Your thread in the GR forum is a good place to go into technical detail about GR.

This thread has focused on a few pages of Rovelli's QG book, and involved different people interpreting and understanding it different ways. Probably at times misinterpreting or simply not understanding. But the thread is too bulky to scrutinize systematically, so I will try to make a clean start with a fresh summary.

Rovelli's conceptual framework here is pretty standard. A less condensed treatment is in his 1999 paper with Marcus Gaul. "LQG and the meaning of diffeomorphism invariance." Really good clear introductory paper. The book gives a compressed version which is more difficult to follow.

I agree completely. Now I use only the paper to understand Rovelli's arguments.


But even then, things are not as clear cut. Carroll says

Let's put some of these ideas into the context of general relativity. You will often hear it proclaimed that GR is a "diffeomorphism invariant" theory. What this means is that, if the universe is represented by a manifold M with metric g_{\mu \nu} and matter fields \psi, and \phi : M \rightarrow Mis a diffeomorphism, then the sets (M, g_{\mu \nu},\psi)and (M, \phi_{*}g_{\mu \nu}, \phi_{*} \psi)represent the same physical situation. Since diffeomorphisms are just active coordinate transformations, this is a highbrow way of saying that the theory is coordinate invariant. Although such a statement is true, it is a source of great misunderstanding, for the simple fact that it conveys very little information. Any semi-respectable theory of physics is coordinate invariant, including those based on special relativity or Newtonian mechanics; GR is not unique in this regard. When people say that GR is diffeomorphism invariant, more likely than not they have one of two (closely related) concepts in mind: the theory is free of "prior geometry", and there is no preferred coordinate system for spacetime.

So Carroll does not use the same definition for active diffeomorphisms as Rovelli does!
You see how hard it is for a neophyte like me not to get confused!


I *think* that the difference between the two views is in the definition of how the fields transform. Carroll transform them by pulling them back using the mapping. I think that Rovelli does not pull them back.

MTW's book "Gravitation" makes an excellent observation which amplifies the point Rovelli is making. Diffeomorphism invariance means something different from what people normally understand by general covariance.
Einstein and others saying "g.c." when they meant "d.i." led to a half century of confusion, say from 1915 to 1975.

When you say g.c. people think of coordinate change, which does not change the distances between points or anything essential about the geometry of the manifold itself.
Other physics besides GR can be formulated in a way that allows coordinate change.

All this is (finally!) clear to me now. Nice summary.

A diffeomorphism φ: M --> N is just an invertible smooth map, which
can be between two different manifolds with completely different shape/curvature and it can completely change the distances. Points m m' can get mapped to points n n' with different distance from each other.

The case we are considering here is φ: M --> M, and the same thing is true.
Points m m' m" which are certain distances apart can get mapped to points
φ(m) φ(m') φ(m") which are completely different distances apart.

I understand the idea and it makes sense to me.
In order for me to *really* understand, though, I would like to see the precise (i.e. mathematical) transformation of the metric under such a diffeomorphism.

Whatever physical theory you want to talk about, Newton, Maxwell, QED, QCD it is not diffeomorphism invariant.

But GR is. And people seem to have been slow to realize this.

This is the "anonymity" that caused the "half century of confusion" that MTW talk about. Einstein did not come out and say "diffeomorphism invariance", but he in effect concealed the radicalism of the concept by calling it "covariance". So people kept thinking it was mere coordinate change stuff.

Yes, it is finally sinking in now!

To really understand the invariance under diffeomorphisms (as you, and Rovelli, define it) I would need to see the proof. It is probably very short. Do you know a good reference where it is done clearly?

Thanks!

 

[nrqed]

Nrqed you are getting to the heart of it!

In GR there is no manifold, no physical objective existing continuum. There is geometry. Waves propagate in the geometry.

People get the idea from popular/simplified accounts of GR that there is a fixed spacetime manifold and the geometry is a metric g_mu_vu on that manifold.

That is not true. In GR the gravitatonal field is an equivalence class of metrics under diffeomorphism.

This effectively denies the manifold (and spacetime points) any objective physical existence.
Einstein remarked on this very explicitly in a couple of 1915 quotes.

The gravitational field is not some particular metric, but rather the whole class of metrics which you can get by mooshing and mapping the manifold around with diffeomorphisms.

...

Interesting stuff deleted

...

In fact it was the GR invariance under diffeomorphisms that led to those radical conclusions. You could moosh and morph a solution and it would still be a solution.
Coord change is different. The manifold and its shape stay the same and you just used different coords.

With coord change you change coordinates but you still have the same rubber sheet but with GR there is no rubber sheet.

That's extremely interesting. But it seems to me that this may be overstating the case.
I mean, if there is no manifold, then we cannot even define a diffeomorphism to start with!
I thought that the conclusion was that we cannot think of points in the manifold as having any physical meaning. In other words, we could not imagine making dots on the manifold and thinking of these dots as representing points in spacetime. Because diffeomorphism can move them around. But the manifold itself still exits, doesn't it? Maybe we are saying the same thing but using a different language. If there is no manifold, I don't see how to define a diffeomorphism to start with.

 

[nrqed]

Quick check on terminology:

1) Which paper of Rovelli's do you mean for definitions http://arxiv.org/abs/gr-qc/9910079 or http://relativity.livingreviews.org/Articles/lrr-2008-5/ ?

I apologize. I meant the paper on the archives. His section 4.1 seems to give the clearest exposition of his argument that I have seen.

2) In special relativity, eg. Maxwell's equations on flat Minkowski spactime, do you consider the metric a field or not?

I don't consider it a field, no.

Let me make my position more clear.
Let's say that we see spacetime as a rubber sheet. I identify points in spacetime by marking a few dots with a pen. Now we do GR. GR says that the metric is dynamical, its value depends on the energy/mass/momentum distribution. Ok, so I picture this by deformations of the rubber sheet. If a mass is present, the rubber sheet is deformed and the distance between the marked points changes. So the metric is dynamical here, but it still does not seem to imply that we must deny reality to the points in the manifold (the points marked with the pen). At least it does not seem to imply that to me. This is why I would say that having a dynamical metric does not imply that we must deny the physical reality of the points in spacetime.

However, the hole argument and the active diffeomorphism invariance (in the language of Rovelli) seem to imply something stronger: that we cannot even think of marking points in the rubber sheet and assigning them an objective reality (although I do not understand thsi deeply since I haven't found an explicit and clear demonstration of this). To me, this is more than just saying that the rubber sheet may be deformed (i.e. that the metric is dynamical)

I don't know what the statement "no prior geometry" means. I thought it meant simply that the spacetime interval between spacetime points depends on energy/mass, i.e. that the rubber sheet is not stiff (as in other theories) but may be deformed. However, the hole argument and the business about diffeomorphisms seem to say more: that we cannot mark points in the rubber sheet (in other words, that there is no objective meaning to a given point in the manifold).


I hope this makes sense.

 

[MTd2]

Let me see from a spin foam point of view. Minimal size means the term locality is restricted the link be nodes, so a ruler can measure a minimum size between nodes. But this size can be arbitrarily small if inferred in other reference frame, so the set node + link + node is just another example of extended object. Right?

 

[nrqed]

Now, Carroll says that active diffeo are completely equivalent to coordinates changes.
But in that section he talks only about diffeo of the manifold, there is no field. I would love to know what he does with fields under active diffeomorphisms. If he uses the same definition as Rovelli, then one cannot say anymore that active diffeo are equivalent to coordinates transfos. Maybe he uses a different definition. But then I am curious about how he would call what Rovelli defines as active diffeo.

Let me take this back. I had not realized that Carroll does define how his fields transform under active diffeomorphisms (see my post number 42 where I quote the relevant passage).
He uses the pullback of the transformation to define the fields in the original manifold. In that case I can see how such an active diffeomorphism has no more content than a simpel change of coordinate system. However, Rovelli uses a different definition, so this is one question that is answered. One down, 20 to go!

 

[atyy]

I don't consider it a field, no.

OK, so in SR the metric is not a field - but, in GR presumably it is a field?

 

[nrqed]

OK, so in SR the metric is not a field - but, in GR presumably it is a field?

Yes. And it is a dynamical field, I agree with that. However, I don't quite see how the "metric as a dynamical field" issue is related to the hole argument. To me they seem like separate issues. See my post number 44 where I try to clarify my question as much as possible.

Thanks!

NOTE ADDED: I think it would help me understand greatly if I could see a clear definition of the hole argument (not just stated n words but with clear definitions of the transformations of all the quantities involved). Any good reference anyone might suggest?

 

[atyy]

Yes. And it is a dynamical field, I agree with that. However, I don't quite see how the "metric as a dynamical field" issue is related to the hole argument. To me they seem like separate issues. See my post number 44 where I try to clarify my question as much as possible.

OK, that's all that I understand then. The difference between SR and GR is that the metric is a dynamical field which is all that I mean when I say that GR is distinguished by "no prior geometry."

I think the argument about points in the manifold having no physical existence is baseless - quite simply, we need the manifold to define GR. But since GR models physical reality by manifold and metric, only manifolds that are isometric describe the same physical reality (eg. Hawking and Ellis, p56). In the most common sense of diffeomorphism (Hawking and Ellis, Wald, Carroll, but not Rovelli's), two manifolds related by a diffeomorphism do not necessarily represent the same physical spacetime - only manifolds related by isometric diffeomorphisms are - but that's common sense since to start we postulated that a metric is physically important - and this is a property of both SR and GR (ie. I have never understood why need to dig ourselves into the hole problem then dig ourselves out).

With respect to the larger issue of background independence, let me note that in GR topology and signature are fixed backgrounds. And in Group Field Theory, which is a current approach to trying to define dynamics for LQG, the theory is defined on a fixed metric of the group manifold ( http://arxiv.org/abs/gr-qc/0607032 ). So I find Rovelli's motivation for LQG poorly conceived, and the criticism that string theory is not background independent also baseless (unless one is also willing to criticize group field theory in the same way).

 

[atyy]

NOTE ADDED: I think it would help me understand greatly if I could see a clear definition of the hole argument (not just stated n words but with clear definitions of the transformations of all the quantities involved). Any good reference anyone might suggest?

There is a clear write up in Wuthrich's thesis, p49. http://philosophy.ucsd.edu/faculty/wuthrich/pub/WuthrichChristianPhD2006Final.pdf

In the later chapters he does say "The fact that GTR has Diff(M) as gauge symmetry means, due to the hole argument as discussed in Chapter 4, that the view which takes the spacetime manifold to be a substance has come under pressure."

However, what he means by this has nothing to do with "points of the manifold having no physical reality". He makes clear what he means before that "Thus, although there still exists a spacetime, it is no longer absolute, but it is demoted to being just another physical field.", which is completely in line with all standard textbooks saying that the distinguishing feature of GR is that the metric is a dynamical field, which is the definition of "no prior geometry".

To make clear that there is nothing mysterious aobut being "demoted" to a field, we could just as easily have said "promoted" to a field. eg. this paper talks about the Barbero–Immirzi parameter being "promoted" to a field http://arxiv.org/abs/0902.2764. What I dislike about Rovelli's work is that he makes simple things so obscure so that he can say LQG is really addressing deep conceptual issues that other quantum gravity people don't.