Two seemingly unrelated arguments against superposition in GR

  • Thread starter bcrowell
  • Start date
bcrowell
Staff Emeritus
Science Advisor
Insights Author
Gold Member
6,723
422

Main Question or Discussion Point

Here are two seemingly unrelated arguments to explain why GR is a nonlinear theory:

(1) By the equivalence principle, any form of mass-energy should cause gravitational fields. Since gravitational fields carry energy, they should cause gravitational fields.

(2) GR doesn't care what coordinates we use. Therefore there is no natural way to identify the points of one spacetime with the points of another. This means that we can't add two metrics in any natural way.

I think #1 is pretty standard. #2 is raised by more than one author in Callender 2001. It's an anthology about quantum gravity, but this particular argument seems to me to apply just as well to classical superposition as to quantum-mechanical superposition.

I haven't seen #2 applied to classical GR before. Am I correct to do so? Is either 1 or 2 more fundamental? They seem unrelated; is there any connection I'm missing?

Craig Callender and Nick Huggett, eds., Physics Meets Philosophy at the Planck Scale: Contemporary Theories in Quantum Gravity, 2001.
 

Answers and Replies

3,506
26
Wrt #2, if any physical theory can be formulated in a coordinate independent way, that would mean that all physical theories are non-linear.
Classical electrodynamics can be formulated in coordinate independent way, but is linear, so that seems a counterexample to #2.

The problems of #1 have been amply discussed in this forum.
 
Bill_K
Science Advisor
Insights Author
4,155
194
(2) GR doesn't care what coordinates we use. [STRIKE]Therefore there is no natural way to identify the points of one spacetime with the points of another. This means that we can't add two metrics in any natural way.[/STRIKE]
Therefore there are an infinite number of ways to identify the points of one spacetime with another. This means that we can add two metrics in an infinite number of ways.
 
3,506
26
Therefore there are an infinite number of ways to identify the points of one spacetime with another. This means that we can add two metrics in an infinite number of ways.
so I guess you don't consider (2) a good way to tell whether a theory is linear, do you? IOW, you are not implying GR is linear, right?
 
martinbn
Science Advisor
1,601
414
Wrt #2, if any physical theory can be formulated in a coordinate independent way, that would mean that all physical theories are non-linear.
Classical electrodynamics can be formulated in coordinate independent way, but is linear, so that seems a counterexample to #2.
But it is not defined in a background independent way.
 
Bill_K
Science Advisor
Insights Author
4,155
194
You can always define a new spacetime by adding the g's, gμν = g(1)μν + g(2)μν. It doesn't have any physical meaning. More to the point, you can add sources: Tμν = T(1)μν + T(2)μν, and presumably the metric resulting from this could be calculated. Both of these operations are always possible, and neither of them means the theory is linear.

The relationship between field and source is nonlocal, something like ◻gμν = Tμν. In a linear theory, there's a linear response. Meaning that when you add the T's, there's a variable g describing the field that also adds. For a theory to be linear they must both happen together.
 
martinbn
Science Advisor
1,601
414
You can always define a new spacetime by adding the g's, gμν = g(1)μν + g(2)μν. It doesn't have any physical meaning. More to the point, you can add sources: Tμν = T(1)μν + T(2)μν, and presumably the metric resulting from this could be calculated. Both of these operations are always possible, and neither of them means the theory is linear.
You can add the metrics only if the underlining manifold is the same, but if they are different it is meaningless. How do you add the standard euclidean metric of the plane and the standard metric of the sphere?
 
3,506
26
But it is not defined in a background independent way.
Sure, but background independence is not mentioned in the OP.
 
3,506
26
You can add the metrics only if the underlining manifold is the same, but if they are different it is meaningless. How do you add the standard euclidean metric of the plane and the standard metric of the sphere?
By adding different manifold metrics, are you referring to adding different solutions of the EFE? In that case I agree, but I guess Bill is referring to superpositions within the same solution.
 
martinbn
Science Advisor
1,601
414
By adding different manifold metrics, are you referring to adding different solutions of the EFE? In that case I agree, but I guess Bill is referring to superpositions within the same solution.
Yes, that's what I mean, metrics on different manifolds. If we have the same manifold of course we can add tensors (of the same type). Why do you think he means within the same solution? In the context of some quantum theory one would probably need superpositions of different solutions.
 
martinbn
Science Advisor
1,601
414
Sure, but background independence is not mentioned in the OP.
Yes, my bad, but that's how I understood it.
 
atyy
Science Advisor
13,646
1,704
The 2 arguments seem to see GR in different ways. #1 sees it as a field in flat spacetime, since in curved spacetime GR, the gravitational field does not have energy. #2 sees it as a curved spacetime, since in flat spacetime presumably one could identify points.

To the extent that one believes that the EP itself indicates its limitation as a local principle, then the EP would indicate curved spacetime, and that would connect arguments #1 and 2. (Perhaps #2 should really be that there is no global inertial frame in GR, since every theory even SR is generally covariant?)

Another thought is that Deser's derivation of GR uses both the EP (gravity couples to all "energy") and gauge invariance (which is analogous to general covariance): http://arxiv.org/abs/gr-qc/0411023, http://arxiv.org/abs/0910.2975 .
 
Last edited:
3,506
26
#1 sees it as a field in flat spacetime, since in curved spacetime GR, the gravitational field does not have energy.
There is only curved spacetime GR, so it is the only way to see it.
 
bcrowell
Staff Emeritus
Science Advisor
Insights Author
Gold Member
6,723
422
Therefore there are an infinite number of ways to identify the points of one spacetime with another. This means that we can add two metrics in an infinite number of ways.
Right, that's what I mean when I say, "This means that we can't add two metrics in any natural way."

The 2 arguments seem to see GR in different ways. #1 sees it as a field in flat spacetime, since in curved spacetime GR, the gravitational field does not have energy. #2 sees it as a curved spacetime, since in flat spacetime presumably one could identify points.
I would seem both 1 and 2 as describing the standard conceptual picture of GR, with curved spacetime.

Another thought is that Deser's derivation of GR uses both the EP (gravity couples to all "energy") and gauge invariance (which is analogous to general covariance): http://arxiv.org/abs/gr-qc/0411023, http://arxiv.org/abs/0910.2975 .
Interesting...this sparks a vague memory of something from the Feynman Lectures on Gravitation (not the undergrad Feynman lectures, a different book). I think he shows that the gravitational field has to be self-interacting.
 
3,506
26
Right, that's what I mean when I say, "This means that we can't add two metrics in any natural way."
But you were saying that shows nonlinearity. How does (2) serve as an argument to explain that GR is nonlinear? Take SR, it doesn't care either what coordinates we use, and it is linear.
 
atyy
Science Advisor
13,646
1,704
Interesting...this sparks a vague memory of something from the Feynman Lectures on Gravitation (not the undergrad Feynman lectures, a different book). I think he shows that the gravitational field has to be self-interacting.
Yes, the Deser derivation is in the same spirit as Feynman's.

I think from the view of a field in flat spacetime, gauge invariance is not necessary, and the EP is all that is needed to argue for a nonlinear equation. The resaon for saying this is that Nordstrom's theory demonstrates a consistent relativistic theory of gravity that obeys the EP but uses a scalar field. If we require gauge invariance as an additional condition, then we do recover GR via arguments in the Feynman-Deser line.

From the point of view of curved spacetime, the EP would say every where is locally flat, but globally curved. If spacetime is globally curved, then there are no global inertial frames, and general covariance is the only covariance, and there are no canonical ways of adding curved objects. (Both Nordstrom's theory (spin 0) and GR (spin 2) have dual formulations as curved spacetime.)
 

Related Threads for: Two seemingly unrelated arguments against superposition in GR

Replies
17
Views
2K
Replies
8
Views
640
  • Last Post
Replies
4
Views
2K
Replies
8
Views
524
  • Last Post
Replies
9
Views
1K
  • Last Post
2
Replies
37
Views
5K
Top