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

Covariant derivative of connection coefficients?

  1. Feb 17, 2012 #1
    The connection [tex]\nabla[/tex] is defined in terms of its action on tensor fields. For example, acting on a vector field Y with respect to another vector field X we get

    [tex]\nabla_X Y = X^\mu ({Y^\alpha}_{,\mu} + Y^\nu {\Gamma^\alpha}_{\mu\nu})e_\alpha
    = X^\mu {Y^\alpha}_{;\mu}e_\alpha[/tex]

    and we call [tex]{Y^\alpha}_{;\mu}={Y^\alpha}_{,\mu} + Y^\nu {\Gamma^\alpha}_{\mu\nu}[/tex] the covariant derivative of the components of Y. We can similarly form the covariant derivative of the components of any rank tensor, by including other appropriate terms with the connection coefficients.

    So what does it mean to take the covariant derivative of the connection coefficients themselves? They are not components of a tensor? I have just come across a reference to [tex]{\Gamma^\alpha}_{\mu\nu;\lambda}[/tex] and don't know what to do with it.
    Last edited: Feb 17, 2012
  2. jcsd
  3. Feb 17, 2012 #2
    I figured this out. Apparently, its covariant derivative does have the same form as the covariant derivative of the components of a (1,2) tensor. But if someone can confirm this result is correct, I would appreciate it.
  4. Feb 17, 2012 #3

    Ben Niehoff

    User Avatar
    Science Advisor
    Gold Member

    That's a very bastard notation, and whoever wrote it down should explain what they mean. As you say, the connection coefficients are not a covariant object, so it is not sensible to talk about their covariant derivatives.

    My guess is someone probably noticed they could write down the formula for the Riemann tensor in a kind of shorthand. It is technically incorrect.
  5. Feb 17, 2012 #4

    Ben Niehoff

    User Avatar
    Science Advisor
    Gold Member

    By the way, I'm not sure of your level of knowledge, but if you're still learning this stuff, I would say to avoid getting in the habit of using "comma, semicolon" notation, for two reasons:

    1. Since covariant derivatives do not commute, it is unclear what is meant by objects such as

    [tex]A^\mu{}_{;\nu\rho} = \nabla_\nu \nabla_\rho A^\mu \quad \text{or} \quad \nabla_\rho \nabla_\nu A^\mu \; \text{?}[/tex]
    2. On the printed page, little marks like commas and semicolons can be hard to see, especially in photocopies.

    Whoever invented the notation thought they were being clever by saving space, but seems to have forgotten that the main purpose of scientific papers is to communicate...
  6. Feb 18, 2012 #5
    Thanks, guys. Yeah, I never liked the semi-colon notation either.
  7. May 23, 2012 #6
    Sorry to drag this up, but in trying to verify the formula for the components of the Riemann tensor in a non--coordinate basis, I need to know how to take the covariant derivative of the connection coefficients. Pellman, can you let me know the resource that confirmed that

    [tex]\nabla_a \Gamma^b{}_{cd} = \partial_a \Gamma^b{}_{cd} + \Gamma^b{}_{ma}\Gamma^m{}_{cd} - \Gamma^m{}_{ca}\Gamma^b{}_{md}- \Gamma^m{}_{da}\Gamma^b{}_{cm}

  8. May 23, 2012 #7
    Ah, working backward from the definition of the Riemann tensor, it would appear that

    \nabla_d \Gamma^a{}_{bc} = \partial_d \Gamma^a{}_{bc}

  9. May 23, 2012 #8
    Since the connection coefficients aren't a tensor, taking a covariant derivative of them doesn't really make sense.
  10. May 23, 2012 #9
    OK, but [itex]\Gamma^a{}_{bc}e^b[/itex] is a vector, so it makes sense to take its covariant derivative.
  11. May 24, 2012 #10
    A covariant derivative is the covariant analogue of a regular derivative. But if you use the affine connection as the thing to operate on, even if it has a form looking like a covariant derivative, it still will not be--it will not be a tensor.

    You can do the operation anyway, and if it has physical usefulness it will still have physical usefulness even though it is not covariant.
    Last edited: May 24, 2012
  12. May 24, 2012 #11
    Sorry. I never found an independent confirmation. I just proved it to my own satisfaction. I don't recall the details now either.
  13. May 24, 2012 #12
    I realise the connection coefficients are not the components of a tensor. However, in the case of the vector [itex]\nabla_c e_b = \Gamma^a{}_{bc}e_a[/itex] I'm pretty sure you can just treat the connection coefficient as the component of the vector: [itex]\Gamma^a{}_{bc} = [\nabla_c e_b]^a[/itex]. Hence

    [tex]\nabla_d (\nabla_c e_b) = \partial_d \Gamma^a{}_{bc} + \Gamma^a{}_{fd}\Gamma^f{}_{bc}[/tex].

    I guess my case is different to yours.
  14. May 24, 2012 #13
    Sorry, I meant [itex]\Gamma^a{}_{bc}e_a[/itex] in this case, as in the post above.
  15. May 24, 2012 #14


    User Avatar
    Science Advisor
    Homework Helper

    The connection coefficients are not the components of any tensor. The covariant derivative, if applied onto this set of components, would lose their meaning and purpose as a derivative. I haven't seen any source in geometry defining a covariant derivative to the connection coefficients.
  16. May 24, 2012 #15
    Yes, as before, I understand this. However, as in the previous post, [itex]\nabla_d (\nabla_c e_b) = \nabla_d(\Gamma^a{}_{bc}e_a)[/itex] is a valid equation. Is my previous result correct?
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook