Vector Identities

As an example, we will derive the simple vector identities using \bgroup\color{black}$ \epsilon_{ijk}$\egroup.

First, lets do the scalar triple product of vectors. Note that this product is completely symmetric among the three vectors once its written in our notation. Its simply cyclic combinations have a plus sign and anticyclic have a minus sign.

$\displaystyle \vec{A}\cdot\left(\vec{B}\times\vec{C}\right)$ $\displaystyle = A_kB_iC_j\epsilon_{ijk} = A_iB_jC_k\epsilon_{ijk}$    
  $\displaystyle =\vec{B}\cdot\left(\vec{C}\times\vec{A}\right) = \vec{C}\cdot\left(\vec{A}\times\vec{B}\right)$    
  $\displaystyle = -\vec{A}\cdot\left(\vec{C}\times\vec{B}\right) = -\vec{B}\cdot\left(\vec{A}\times\vec{C}\right) = -\vec{C}\cdot\left(\vec{B}\times\vec{A}\right)$    

The triple vector product will require us to ``derive'' an identity involving the product of two \bgroup\color{black}$ \epsilon_{ijk}$\egroups.

\bgroup\color{black}$\displaystyle \left[\vec{A}\times\left(\vec{B}\times\vec{C}...
...ht)\epsilon_{ijk} = A_iB_mC_n \left(\epsilon_{mnj}\epsilon_{ijk}\right) $\egroup

Since the \bgroup\color{black}$ \epsilon$\egroups are well defined and have simple properties, we must be able to compute a simple expression for \bgroup\color{black}$ \epsilon_{mnj}\epsilon_{ijk}$\egroup. Note that the index \bgroup\color{black}$ j$\egroup is repeated and therefore summed over. We just need to do the sum but some thought will be required. Nevertheless, you could do this thinking if stranded on a dessert island and in need of vector identities.

We now do the sum \bgroup\color{black}$ \epsilon_{mnj}\epsilon_{ijk}=-\epsilon_{mnj}\epsilon_{ikj}$\egroup. First of all \bgroup\color{black}$ \epsilon_{ijk}$\egroup will be zero if any of the indices are repeated. So me must have \bgroup\color{black}$ m\neq n$\egroup and \bgroup\color{black}$ i\neq k$\egroup. We will sum over \bgroup\color{black}$ j$\egroup, but starting from the first \bgroup\color{black}$ \epsilon$\egroup, \bgroup\color{black}$ j$\egroup must be the other index besides the two used up for \bgroup\color{black}$ m$\egroup and \bgroup\color{black}$ n$\egroup so there is only one non-zero term in the sum. This is also true for the second \bgroup\color{black}$ \epsilon$\egroup. Since the index \bgroup\color{black}$ j$\egroup in the two tensors is the same we must have \bgroup\color{black}$ m$\egroup and \bgroup\color{black}$ n$\egroup using up the same two indices that \bgroup\color{black}$ i$\egroup and \bgroup\color{black}$ k$\egroup do. So the two possibilities are

$\displaystyle m=i$ $\displaystyle n=k$    
  $\displaystyle {\mathrm or}$      
$\displaystyle m=k$ $\displaystyle n=i$    

There is only one term in the sum. We can write this result in terms of Kronecker \bgroup\color{black}$ \delta$\egroups.
\bgroup\color{black}$ \displaystyle \epsilon_{mnj}\epsilon_{ikj} = \delta_{mi}\delta_{nk} - \delta_{mk}\delta_{ni} $\egroup

\bgroup\color{black}$\displaystyle -\epsilon_{mnj}\epsilon_{ikj} = \delta_{mk}\delta_{ni} - \delta_{mi}\delta_{nk} $\egroup

We can now easily compute the triple vector product identities.

$\displaystyle \left[\vec{A}\times\left(\vec{B}\times\vec{C}\right)\right]_k$ $\displaystyle = A_iB_mC_n \left(\epsilon_{mnj}\epsilon_{ijk}\right) = A_iB_mC_n...
...ht) = B_k\left(\vec{A}\cdot\vec{C}\right) - C_k\left(\vec{A}\cdot\vec{B}\right)$    
$\displaystyle \vec{A}\times\left(\vec{B}\times\vec{C}\right)$ $\displaystyle =\vec{B}\left(\vec{A}\cdot\vec{C}\right) - \vec{C}\left(\vec{A}\cdot\vec{B}\right)$    

For expressions involving derivatives in 3D, these tools become even more useful.

Jim Branson 2012-10-21