I have two questions on the computation of $\nabla \times (\nabla \times \mathbf{A}) $ with Einstein summation notation based on http://www.physics.ohio-state.edu/~ntg/263/handouts/tensor_intro.pdf. It considers the $i$th component. All colours have been added by me. $$ (\color{green}{\nabla} \color{red}{\times} \color{purple}{(\nabla \times \mathbf{A})})_i = \color{red}{\epsilon_{ijk}}\color{green}{\partial_j} \color{purple}{(\nabla \times \mathbf{A})}_k \tag{*}$$

$$ = \color{red}{\epsilon_{ijk}}\color{green}{\partial_j} \color{purple}{\epsilon_{klm}\partial_lA_m}$$ $$ = \epsilon_{ijk}\epsilon_{klm}\partial_j \partial_lA_m $$ $$ = \epsilon_{ijk}\epsilon_{lmk}\partial_j \partial_lA_m $$ $$ = (\delta_{il}\delta_{jm} - \delta_{im}\delta_{jl}) \partial_j \partial_lA_m $$ Now, choose $j \rightarrow m$ and $ l \rightarrow i $ in the first term and let $j \rightarrow l$ and $ m \rightarrow i $ in the second:

$$ = \partial_m \partial_iA_m - \color{blue}{\partial_l \partial_lA_i}$$ $$ = \partial_i(\partial_m A_m) - \color{blue}{(\partial_l \partial_l\mathbf{A})_i}\tag{**}$$

$\Large{\text{Question 1}}$. In (*), where does the subscript $k$ on the RHS come from? Aren't we looking at the $i$th component? I know that $\mathbf{a} \times \mathbf{b} = a_gb_h\epsilon_{ghi}\mathbf{\hat{e_i}} = (..., \underbrace{a_gb_h\epsilon_{ghi}}_{i\text{th component}}, ...)\Longrightarrow (\mathbf{a} \times \mathbf{b})_i = a_gb_h\epsilon_{ghi} $.

$\Large{\text{Question 2}}$. In (**), how does $ \color{blue}{\partial_l \partial_lA_i = (\partial_l \partial_l\mathbf{A})_i} $ ?


I like your format.

Fisrt question: (*) is just applying the definition of $$ (\mathbf{a} \times \mathbf{b})_i = a_j b_k \epsilon_{jki} = a_j b_k \epsilon_{ijk} $$ Let $\mathbf{a} = \nabla = (\partial_1, \partial_2,\partial_3)$, and $a_j = \partial_j$. $\mathbf{b} = \nabla \times \mathbf{A}$ here, hence $$ (\nabla\times(\nabla\times \mathbf{A}))_i = \partial_j(\nabla\times \mathbf{A})_k \epsilon_{ijk} $$

Notice repeated subscripts get canceled, so here $i$ ceases to appear as a subscript on the RHS of (*), but it's in the Levi-Civita symbol. Also, $k$ is just a dummy summation subscript. What you have is still the $i$-th component of $\nabla\times(\nabla\times \mathbf{A})$, for subscript $i$ is in the permutation symbol.

Second question: notice $$\partial_l \partial_l A_i = \sum_{l=1}^3 \partial_l \partial_l A_i $$ while $$ \partial_l \partial_l \mathbf{A} = \sum_{l=1}^3 \partial_l \partial_l (\sum_{i=1}^3 A_i \mathbf{e}_i) = \sum_{i=1}^3\left(\sum_{l=1}^3 \partial_l \partial_l A_i\right)\mathbf{e}_i $$ therefore $$(\partial_l \partial_l \mathbf{A})_i = \partial_l \partial_l A_i$$

Finally the identity: $$ \nabla\times(\nabla\times \mathbf{A}) = -\Delta \mathbf{A} + \nabla \nabla\cdot\mathbf{A}. $$ I don't know if there is a direct proof without expanding the whole thing as sum.

  • $\begingroup$ Thank you very much for your detailed answer. I've just 1 follow-up question. You wrote that "Notice repeated subscripts get canceled...," but what got cancelled in the equations preceding your sentence? I don't see any terms that became cancelled? $\endgroup$ – Greek - Area 51 Proposal May 6 '13 at 15:47
  • $\begingroup$ @LaPrevoyance Pardon the ambiguity in my answer. "repeated subscripts get canceled" means whenever we have repeated subscript $a_i b_i c_j = \sum^n_{i=1} a_i b_i c_j = d_j$, whose subscript follows $j$, has no subscript $i$ any more. The proper way to phrase this should be "repeat indices are summed out and disappeared in the result". $\endgroup$ – Shuhao Cao May 6 '13 at 15:54
  • $\begingroup$ Thank you for your response. Last question: So had you been referring to the subscript $i$? It doesn't appear on the RHS of (*) as a subscript of the whole expression anymore, but only in the Levi-Civita symbol? Also, thank you for liking my formatting! $\endgroup$ – Greek - Area 51 Proposal May 6 '13 at 16:09

Based on Shuhao Cao's detailed and helpful answer, I realised that a simpler way to understand (**) is by first grasping that $[\mathbf{F}]_i = F_i $.

This is true because $\mathbf{F} = \sum_{i = 1}F_i\mathbf{\hat{e_i}} \Rightarrow [\mathbf{F}]_i = F_i$.


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.