Good morning/evening to everybody.

I'm interested in proving this proposition from the Green's first identity, which reads that, for any sufficiently differentiable vector field $\mathbf{\Gamma}$ and scalar field $\psi$ it holds:

$$ \int_U \nabla \cdot \mathbf{\Gamma} \, \psi \, dU = \int_{\partial U} (\mathbf{\Gamma} \cdot \mathbf{n}) \, \psi \, dS - \int_U \mathbf{\Gamma} \cdot \nabla \psi \, dU.$$

I've been told that, for $\mathbf{u}, \vec{\omega} \in \mathbb{R}^2$, it is true that (which I can prove quite easily using index notation):

$$ \int_U \Delta u_j \omega_j \, dU = \int_{\partial U} \frac{\partial u_j}{\partial n} \, \omega_j \, dS - \int_U \nabla u_j \cdot \nabla \omega_j \, dU,$$ which is equivalent to:

$$\int_U (\nabla \cdot \nabla \mathbf{u}) \cdot \, \vec{\omega} \, dU = \int_{\partial U} \frac{\partial {\mathbf{u}}}{\partial n} \cdot \vec{\omega}\, dS - \int_U \nabla \mathbf{u} : \nabla \vec{\omega} \, dU, \quad j =1,2 . $$

In every one of the equations above, $U$ is a closed region and $\partial U$ its border, being $\mathbf{n}$ its outer normal unit vector and $:$ stands for tensor contraction. I would like to prove that, for any given tensor $\mathbf{T}$, it holds:

$$\int_U (\nabla \cdot \mathbf{T}) \cdot \, \vec{\omega} \, dU = \int_{\partial U} (\mathbf{T} \cdot \mathbf{n}) \cdot \vec{\omega}\, dS - \int_U \mathbf{T} : \nabla \vec{\omega} \, dU , $$

but I'm not very proficient at dealing with index notation, yet. How could I proceed?

Any help will be much appreciated.


Edit and hint: Maybe divergence theorem is useful here? Since the Green's first identity is derived from it.

  • 1
    $\begingroup$ Nothing yet? :( $\endgroup$ – Dmoreno Apr 9 '14 at 10:07

It appears that I misread the question the first time. In any case, this seems to follow from the generalisation of the divergence theorem for rank $n$ tensors:

$$\int_{U}{\frac{\partial T_{i_1i_2 \ldots i_q \ldots i_n}}{\partial x_{i_q}} dU}=\int_{\partial U}{T_{i_1i_2 \ldots i_q \ldots i_n}n_{i_q} dS}$$

applied to $T \cdot \omega$ instead of $T$.

\begin{align} \int_{\partial U}{T_{i_1i_2 \ldots i_q \ldots i_r \ldots i_n}n_{i_q} \omega_{i_r} dS}& =\int_{U}{\frac{\partial (T_{i_1i_2 \ldots i_q \ldots i_r \ldots i_n}\omega_{i_r})}{\partial x_{i_q}} dU} \\ &=\int_{U}{\frac{\partial T_{i_1i_2 \ldots i_q \ldots i_r \ldots i_n}}{\partial x_{i_q}}\omega_{i_r} dU}+\int_{U}{T_{i_1i_2 \ldots i_q \ldots i_r \ldots i_n}\frac{\partial \omega_{i_r}}{\partial x_{i_q}} dU} \end{align}

Hopefully I haven't messed up again.

  • 1
    $\begingroup$ Hi there, Ivan. Is $\mathbf{F}$ either a vector or a tensor field? Is divergence theorem valid for a tensor field? $\endgroup$ – Dmoreno Apr 10 '14 at 18:55
  • 1
    $\begingroup$ well $\psi$ is a scalar field and $\boldsymbol{\Gamma}$ is a vector field so $\boldsymbol{F}=\psi \boldsymbol{\Gamma}$ is a vector field. $\endgroup$ – Ivan Loh Apr 10 '14 at 19:04
  • 1
    $\begingroup$ Then I cannot see how this can help me in order to prove the 'extension' of the Green's first identity to tensor fields :( $\endgroup$ – Dmoreno Apr 10 '14 at 19:06
  • 1
    $\begingroup$ Oh, it seems I misread your question. In that case, it seems that a more general version of divergence theorem applies, e.g. in the last page of cs.uu.nl/groups/AA/virtual/surgery/thesis/appendix.pdf, though I haven't worked it out explicitly myself. $\endgroup$ – Ivan Loh Apr 10 '14 at 19:12
  • 1
    $\begingroup$ Thank you very much, Ivan. Is this document a phd appendix or is it from a book? I will try to work this out tomorrow. $\endgroup$ – Dmoreno Apr 10 '14 at 19:16

Let me answer myself so I can see if I got this right.

Divergence theorem tells us that:

$$\int_S d \vec{s} \cdot \square = \int_V dV \, \nabla \cdot \square, $$

in the particular case of a given second order tensor, $\mathbf{T}$, and a vector field $\omega$, it follows:

$$\int_S d \vec{s} \cdot (\mathbf{T} \cdot \vec{\omega}) = \int_V dV \, \nabla \cdot (\mathbf{T} \cdot \vec{\omega}),$$

since, and now I will use index notation:

$$ \nabla \cdot (\mathbf{T} \cdot \vec{\omega}) = \partial_q (T_{pq} \omega_p) = T_{pq} \partial_q\omega_p + (\partial_q T_{pq}) \omega_p , $$

so I come up with the vector identity:

$$\nabla \cdot(\mathbf{T} \cdot \vec{\omega} ) = T:\nabla\vec{\omega} + (\nabla \cdot \mathbf{T}) \cdot\vec{\omega}, $$

so my original assumption holds.

Am I right?

Cheers and thanks!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.