1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Levi-Civita proofs for divergence of curls, etc

  1. Sep 17, 2008 #1
    I've also posted this in the Math forum as it is math as well.

    ---

    I want to know if I'm on the right track here. I'm asked to prove the following.

    a) [tex]\nabla \cdot (\vec{A} \times \vec{B}) = \vec{B} \cdot (\nabla \times \vec{A}) - \vec{A} \cdot (\nabla \times \vec{B})[/tex]
    b) [tex]\nabla \times (f \vec{A}) = f(\nabla \times \vec{A}) - \vec{A} \times (\nabla f)[/tex] (where f is a scalar function)

    And want (read: need to, due to a professor's insistence) to prove these using Levi-Civita notation. I've used the following for reference:
    http://www.uoguelph.ca/~thopman/246/indicial.pdf [Broken] and http://folk.uio.no/patricg/teaching/a112/levi-civita/

    Here's my attempts - I need to see if I have this notation down correctly...

    a) [tex]\nabla \cdot (\vec{A} \times \vec{B}) [/tex]

    = [tex]\partial_i \hat{u}_i \cdot \epsilon_{jkl} \vec{A}_j \vec{B}_k \hat{u}_l[/tex]

    = [tex]\partial_i \vec{A}_j \vec{B}_k \hat{u}_i \cdot \hat{u}_l \epsilon_{jkl}[/tex]

    Now I thought it'd be wise to use the identity that [tex]\hat{u}_i \cdot \hat{u}_l = \delta_{il}[/tex].

    = [tex]\partial_i \vec{A}_j \vec{B}_k \delta_{il} \epsilon_{jkl}[/tex]

    In which we make i = l (and the [tex]\delta_{il}[/tex] goes to 1).

    = [tex]\partial_i \vec{A}_j \vec{B}_k \epsilon_{jki} [/tex]

    Then using 'scalar derivative product rules' we get two terms. Now, here's where I get a little mixed up. I'm wondering if we rearrange the terms and then modify the epsilon to go in order the the terms.

    = [tex] \vec{B}_k \partial_i \vec{A}_j \epsilon_{kij} + \vec{A}_j \partial_i \vec{B}_k \epsilon_{jik}[/tex]

    Now since the first epsilon is 'even' it remains positive, the other epsilon is 'odd' so that term becomes negative and we end up with the required result.

    = [tex] \vec{B} (\nabla \times \vec{A}) - \vec{A} (\nabla \times \vec{B}) [/tex]

    b) [tex]\nabla \times (f \vec{A}) [/tex] (where f is a scalar function)

    = [tex] \partial_i f \vec{A}_j \hat{u}_k \epsilon_{ijk}[/tex]

    = [tex]f \partial_i \vec{A}_j \hat{u}_k \epsilon_{ijk}+ \vec{A}_j \partial_i f \hat{u}_k \epsilon_{jik}[/tex]

    Once again, the first epsilon is the positive ('even') while the other is negative ('odd').

    = [tex] f (\nabla \times \vec{A}) - \vec{A}(\nabla f)[/tex]

    Man, my hands hurt from all that tex work :P Been awhile for me.
    Since my teacher refuses to tell me if this is the correct method (he's only willing to show the concepts, and while I can appreciate that I don't want my mark to go to hell), can anyone help me out?
     
    Last edited by a moderator: May 3, 2017
  2. jcsd
Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook

Can you offer guidance or do you also need help?
Draft saved Draft deleted



Similar Discussions: Levi-Civita proofs for divergence of curls, etc
  1. Spacefuel etc. (Replies: 3)

Loading...