Linear static elasticity provides a local equation $-\mathrm{div}\sigma=f$, the constitutive law $\sigma=2\mu\epsilon+\lambda \mathrm{tr}(\epsilon)I$ as well as the strain-displacement relationship $\epsilon=1/2(\mathrm{grad}u+\mathrm{grad}^T u)$ (boundary conditions are ignored) where:

  • $\sigma$ is the second-order stress tensor
  • $\epsilon$ is the second-order strain tensor
  • $f$ is the vector of external distributed force
  • $I$ is the second-order identity tensor
  • $u$ is the (space-dependent) displacement vector
  • $\mu$ and $\lambda$ are Lamé's coefficients

The Lamé-Navier version of the local equation is solely expressed in terms of vector $u$ as follows:$$ \mu \Delta u+ (\lambda+\mu) \mathrm{grad}(\mathrm{div}u)+f=0$$ Derivations found in books all use a Cartesian coordinate system to derive this identity which is then re-expressed in an "intrinsic" coordinate-free vector form. Is there a way (or a reference) detailing how to derive this identity directly in a coordinate-free fashion?

  • $\begingroup$ Is this Lame-Navier or Navier-Cauchy? Wikipedia says Navier-Cauchy. $\endgroup$ – timur Jun 18 at 15:00

You're essentially asking to compute $\nabla\cdot\boldsymbol{\sigma}$ in coordinate-free terms. Here's a possible way, although it requires some tricks, in particular the identities $$\text{tr}(\nabla \mathbf{a})=\text{tr}(\nabla \mathbf{a}^\mathsf{T})=\nabla\cdot \mathbf{a}$$ $$\nabla\cdot\nabla \mathbf{a}=\Delta \mathbf{a}$$ $$\nabla\cdot\nabla\mathbf{a}^\mathsf{T}=\nabla(\nabla\cdot\mathbf{a})$$ and $$\nabla\cdot(b\mathbf{I})=\nabla b$$ for vector $\mathbf{a}$, scalar $b$, and identity $\mathbf{I}$. It's then straightforward to get $$\nabla\cdot\boldsymbol{\sigma}=\nabla\cdot\left(2\mu\left(\frac{\nabla\mathbf{u}+\nabla \mathbf{u}^\mathsf{T}}{2}\right)+\lambda\text{tr}\left(\frac{\nabla\mathbf{u}+\nabla \mathbf{u}^\mathsf{T}}{2}\right)\mathbf{I}\right)$$ $$=\mu\nabla\cdot\nabla\mathbf{u}+\mu\nabla(\nabla\cdot\mathbf{u})+\lambda\nabla\cdot((\nabla\cdot\mathbf{u})\mathbf{I})$$ $$=\mu\Delta\mathbf{u}+\mu\nabla(\nabla\cdot\mathbf{u})+\lambda\nabla(\nabla\cdot\mathbf{u})$$ from which $$\mathbf{f}+\mu\Delta\mathbf{u}+(\lambda+\mu)\nabla(\nabla\cdot\mathbf{u})=\mathbf{0}$$ follows.

If anyone has a more straightforward way to demonstrate this in index-free notation, please feel free to post it as another answer! Index-free notation gets difficult once you deal with nontrivial objects. For example, with two matrix functions on space $\mathbf{A}(\mathbf{r})$, $\mathbf{B}(\mathbf{r})$, you have $$\nabla(\mathbf{A}\cdot\mathbf{B})=\sigma_{132}((\sigma_{132}\nabla\mathbf{A})\cdot\mathbf{B})+\mathbf{A}\cdot\nabla\mathbf{B}$$ where $\sigma_{132}$ is the index transposing operator Transpose[#,{1,3,2}]& from Mathematica, whereas with comma derivatives you just use product rule to get $$(A_{ij}B_{jk})_{,l}=A_{ij,l}B_{jk}+A_{ij}B_{jk,l}$$.

EDIT: In index notation, the proof reads $$\sigma_{ij,j}=2\mu\frac{u_{i,jj}+u_{j,ij}}{2}+\lambda u_{k,kj}\delta_{ij}$$ $$=\mu u_{i,jj}+(\lambda+\mu)u_{j,ij}.$$

This is an example of why index notation is better. You don't need to remember a bunch of vector and matrix calculus identites, you just take derivatives.

| cite | improve this answer | |
  • $\begingroup$ Nice. I was missing some of the identities you show. Still, how do we prove them? Should be have to use a specific coordinate system? $\endgroup$ – pluton Mar 3 '14 at 17:14
  • $\begingroup$ @pluton: The 4 index-free identities I used are readily proved by writing them out in matrix form, and recalling that the $\nabla$ operator acts like a vector multiplying from the right; of course, this means their proof requires using indices, but that doesn't prevent you from applying them in index-free form. $\endgroup$ – DumpsterDoofus Mar 3 '14 at 17:21
  • $\begingroup$ Interesting: I'm kindda wondering whether it is possible to prove all these coordinate-free identities without never coming back to a coordinate system? $\endgroup$ – pluton Mar 3 '14 at 22:54
  • $\begingroup$ @pluton: Not that I know of, but it might be possible. In any case, indicial notation is much easier to use, even though it looks uglier and has lots of subscripts floating around. $\endgroup$ – DumpsterDoofus Mar 4 '14 at 2:51
  • $\begingroup$ but index notation will depend on the coordinate system if I am not wrong. $\endgroup$ – pluton Mar 4 '14 at 15:16

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.