Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

About Tensor Calculus

  1. Apr 5, 2010 #1
    Dear Friends

    I have two questions to do about Tensor Calculus:

    1) Is there any program to calculate Christoffel Symbols, Riemann and Ricci Tensors and everything about Tensor Calculus (Free or paid)?

    2) When in an exercise anyone asks to use the Euclidean metric or Riemann metric, what is the difference? Is it only the presence of € (indicator)?

    Thank you for your help.

    Good day.
     
  2. jcsd
  3. Apr 5, 2010 #2

    dx

    User Avatar
    Homework Helper
    Gold Member

    1. Yes. http://grtensor.phy.queensu.ca/ [/URL]

    2. I don't know what € is, but a Riemannian metric is any metric with signature (n,0), and a Euclidean metric is a special Riemannian metric of the form ds² = dx[SUB]1[/SUB]² + dx[SUB]2[/SUB]² + ... + dx[SUB]n[/SUB]².
     
    Last edited by a moderator: Apr 25, 2017
  4. Apr 5, 2010 #3
    If I've undrestood this right, a Euclidean metric takes that form in an orthonormal coordinate system, but other forms in non-orthonormal coordinate systems, e.g. [itex]ds^2 = dx^2 + dy^2[/itex] becomes [itex]ds^2 = dr^2 + r^2 \; d \theta^2[/itex] in plane polar coordinates. Is there an intrinsic definition of Euclidean metric that doesn't refer to a special coordinate system, or is the full definition of a Euclidean metric: a metric that has the above form in an orthonormal coordinate system? Similarly for a Euclidean metric tensor. Is it defined as a metric tensor whose matrix representation, in orthonormal coordinates, is a diagonal matrix with diagonal entries all 1, or is there a definition that doesn't refer to coordinates?
     
  5. Apr 5, 2010 #4

    dx

    User Avatar
    Homework Helper
    Gold Member

    I'm not sure if 'Euclidean metric' has a conventionally accepted frame invariant meaning. I think its just a term used to describe things which look like the Pythagorean formula in some coordinate system. Of course any such coordinate system will be orthonormal since

    g(∂i, ∂j) = (dx1² + dx2² + ... + dxn²)(∂i, ∂j) = δij.
     
  6. Apr 5, 2010 #5
    Okay, thanks. So if a metric has Pythagorean form, it's Euclidean, and the form guarrantees that an orthonormal coordinate system is meant.

    Since the word tensor is used in this context to mean a tensor (or tensor field) with respect to the tangent space (or if tensor field, the tangent bundle) of a manifold, presumably a Euclidean metric tensor is a frame invariant object that can have coordinate representations besides the diagonal matrix [itex]I[/itex] with entries [itex]\delta_{ij}[/itex] that is its coordinate representation in an orthonormal system. And since the metric can be worked out from the metric tensor, I'm guessing a Euclidean metric can also have non-Pythagorean representations, and that it's the existence of a Pythagorean representation that makes a metric Euclidean, rather than the metric happening to be in Pythagorean form. Is that right?
     
  7. Apr 5, 2010 #6

    atyy

    User Avatar
    Science Advisor

    For Euclidean, how about a Riemannian manifold with signature X and all components of the curvature tensor zero erverywhere?
     
  8. Apr 5, 2010 #7
    Thanks for your help.

    Till the next. Cheers.
     
  9. Apr 5, 2010 #8
    I think I may like this notation more than the notation I'm familiar with. Can you reference a text that utilizes it?
     
  10. Apr 5, 2010 #9

    dx

    User Avatar
    Homework Helper
    Gold Member

    You mean using ∂i for the tangent vector ∂/∂xi?
     
  11. Apr 5, 2010 #10
    Should I be thinking then of a metric generally as a coordinate-dependent thing: orthonormal metric (i.e. "Euclidean metric"), polar metric etc., rather than as a feature of the underlying set of the manifold? Is the following correct? Iff the underlying set of a Riemannian manifold is the point set of Euclidean space En, it can be given a global coordinate system in which the metric takes this form, the "Euclidean metric", everywhere. But in any Riemannian manifold, even one whose underlying set is not the point set of Euclidean space, it's possible, for any given point, to pick a coordinate system in which the metric is the "Euclidean metric" at that point, but not necessarily at all other points. And a Euclidean space En can be referred to coordinate systems in which the metric is not the "Euclidean metric" everywhere, and even to coordinate systems in which the metric is not the "Euclidean metric" anywhere. (And if so, would "orthonormal metric" be a suitably unambiguous alternative name?)
     
  12. Apr 5, 2010 #11

    dx

    User Avatar
    Homework Helper
    Gold Member

    No, it is better and more correct to think of a metric as a coordinate indepent object.

    Since in special relativity, we say that the metric is locally Minkowski, I think it would be consistent to say that the metric on a Riemannian manifold is locally Euclidean. But still, as I said, I don't know if there is a conventional used definition.
     
    Last edited: Apr 6, 2010
  13. Apr 5, 2010 #12
    Ah, okay. Sounds like I was closer to the mark in #3 and #5 than #10.

    Is this more or less what you mean by "define it in this way"? A metric field is a function-valued field (in the sense scalar, vector, tensor, etc. field), or perhaps a field whose value at each point is defined as some kind of coordinate-independent equivalence class of functions... Its value at some point, in some coordinate system might be given by, for example, [itex]ds^2 = dr^2 + r^2 \; d \theta^2[/itex]. The metric (i.e. distance function which is the value of the metric field) is said to be Euclidean at this point iff there exists a coordinate system (namely an orthonormal one) in which the coordinates of the metric tensor give the metric the form of Pythagoras's formula. Iff for some n-manifold, there exists a single, global coordinate system according to which the metric everywhere has Pythagorean form, then the manifold is the Euclidean space En. Iff for each point in the underlying set of the manifold, there exists a coordinate system (not necessarily global, i.e. not necessarily the same coordinate system as for all other points) according to which the metric is Euclidean at that point, then the manifold is Riemannian.
     
    Last edited: Apr 5, 2010
  14. Apr 5, 2010 #13

    bcrowell

    User Avatar
    Staff Emeritus
    Science Advisor
    Gold Member

    For software, I use maxima and ctensor. It does everything I need, and it's free and open source.
     
  15. Apr 6, 2010 #14
    Sean Carroll uses the ∂i notation for taking derivatives, but not for coordinate basis vectors. Instead, he moves quickly to using

    [tex]\hat{e}_{(\mu)} = \partial_{\mu}[/tex]

    and

    [tex]\hat{\theta}^{(\mu)} = d^{\mu} \ ,[/tex]

    as bases in a manner I found too hard to follow. I would be interested in a presentation where the bases dxu are presented as true infinitesimals, and --- I'm not sure what the ∂u would be.
     
  16. Apr 6, 2010 #15

    dx

    User Avatar
    Homework Helper
    Gold Member

    I wasn't using notation in which the dxu were infinetesimals. I was using operator notation, where dx1² + dx2² + ... + dxn² is short for the operator

    [tex] dx_1 \otimes dx_1 + dx_2 \otimes dx_2 + ... + dx_n \otimes dx_n [/tex]

    and the dxu are coordinate 1-forms. So (dx1² + dx2² + ... + dxn²)(∂i, ∂j) would be

    [tex] (dx_1 \otimes dx_1 + dx_2 \otimes dx_2 + ... + dx_n \otimes dx_n) (\partial_i, \partial_j) = \sum_{\alpha} dx_{\alpha}(\partial_i)dx_{\alpha}(\partial_j) = \sum_{\alpha} \delta_{\alpha i} \delta_{\alpha j} = \delta_{ij} [/tex]

    If you're interested in GR books that don't use the modern notation of 1-forms and multilinear operators, the only book I can recommend is Landau and Lifgarbagez's "Classical Theory of Fields". That's probably the only book written in 'old tensor' that is still worth reading.
     
    Last edited: Apr 6, 2010
  17. Apr 6, 2010 #16
    Sorry. I wasn't clear. I'm looking for something modern using both differential forms and infinitesimal calculus, although the infinitesimal calculus is secondary. It would be good if exterior derivatives and wedge products were given some notice, as well.

    Perhaps this is not such a small wish. You certainly have more modern training than I've had, so I was hoping you might point me in the right direction.
     
    Last edited: Apr 6, 2010
  18. Apr 6, 2010 #17

    dx

    User Avatar
    Homework Helper
    Gold Member


    Some books (that use modern 'calculus on manifolds') that I like are:

    "Spacetime, Geometry, Cosmology" by William Burke
    "Gravitation" by Misner, Thorne and Wheeler
    "Gauge Fields, Knots and Gravity" by John Baez and Javier Muniain

    Baez is good for the algebraically oriented, while the other two are good for geometrically oriented readers.
     
    Last edited: Apr 6, 2010
  19. Apr 6, 2010 #18
  20. Apr 6, 2010 #19
    William Burke I have heard of. Which one have you studied from, formally?
     
  21. Apr 6, 2010 #20

    dx

    User Avatar
    Homework Helper
    Gold Member

    I've studied Burke thoroughly, doing all the exercises etc., and I've studied part II and part III of MTW ("Physics in Flat Spacetime" and "Mathematics of Curved Spacetime").

    To get a gut feeling for 1-forms and tangent vectors, I know of no better book than Burke. He doesn't go too deeply into curvatre and Einstein's equations, instead focusing on studying given spacetimes and cosmological models. There are some unconventional ideas in Burke that I think are worth learning about, like the 'wave diagram' and 'Huygen's construction'.

    Burke can serve as excellent preparation for a more detailed discussion of differential geometry and Einstein field equations given in MTW. The first part of Baez is a good presentation of differential forms from an algebraic point of view, while MTW emphasizes the geometry.
     
    Last edited: Apr 6, 2010
  22. Apr 6, 2010 #21
    "curvatre"? Are you British? It's 6 am for me.

    OK. I'll buy Burke. Thanks so much for the help. I've have MTW sitting on the self since before I understood special relativity and I'm still too frightened to crack it open. Baez never made any sense to me. I guess I'm one of those geometrical thinkers.
     
  23. Apr 6, 2010 #22
    Could we, in principle, choose some other ordered set of basis vectors for the cotangent space, a basis for which this equation isn't true? Would you still call the cotangent vectors of such a basis "coordinate 1-forms"? Would it be necessary to use a different notation then; in other words, does the notation imply this equality?

    If this is old-fashioned notation, how would it be written in modern notation? Is there an ambiguity in the notation, in that [itex]\partial_i[/itex] might stand for some particular coordinate basis tangent vector, conceived of as a derivative operator, e.g. [itex]\partial_1[/itex] or [itex]\partial_2[/itex], but in abstract (slot-naming) index notation, [itex]\partial_i[/itex] could stand for the set [itex]\left \{ \partial_1, \partial_2, ..., \partial_n \right \}[/itex], which some people call the gradient?
     
  24. Apr 6, 2010 #23

    dx

    User Avatar
    Homework Helper
    Gold Member

    For any coordinate system {xμ}, the coordinate 1-forms dxμ and the coordinate basis vectors ∂ν always satisfy dxμ⋅∂ν = δμν

    No, everything I wrote was in modern notation.
     
    Last edited: Apr 6, 2010
  25. Apr 6, 2010 #24

    dx

    User Avatar
    Homework Helper
    Gold Member

    If we want to express a directional derivative along some vector v in terms of ∂i, we use vii, where summation is implied. Is that what you meant?
     
  26. Apr 6, 2010 #25
    This seems to treat the Euclidean metric as coordinate-dependent, contra to dx's post #11. Since [itex]ds^2 = dr^2 + r^2 \; d \theta^2[/itex] isn't the Pythagorean formula (and its inputs aren't even all real numbers, being rather one real variable and two differentials), the author of the article would presumably not call it the Euclidean metric, even though with a change of coordinates, such a metric might take Pythagorean form after all. On the other hand, it says the Euclidean metric is "given by" the identity matrix in Euclidean 3-space, making me wonder if it means that the Euclidean metric would be given by some other matrix in some other space, or whether it's suggesting it's only defined in Euclidean space, and what these definitions would look like in a treatment that distinguished between Euclidean space and its coordinate representations. The article conflates Euclidean n-space with the set of real n-tuples, whereas I gather more precise treatments distinguish between the set of points (the point set) of Euclidean space, its algebraic structure (conceived either as an affine space with an inner product, or as a manifold), and the various ways of representing its points as coordinate n-tuples in Rn.
     
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook