1 Einstein Summation Convention
1.1 Expression, Terms, Indices
\[ \begin{equation} A_{im}x_{m}+b_i+C_{ikq}x_{k}y_{q} \end{equation} \]
The expression above contains three terms. The symbols \(i,m,k,q\) denote indices. They typically run from one to number of space dimensions, i.e., two or three.
The complexity of the terms can be indefinitely increased. \[ \begin{equation} a_{ij} x_ix_jx_k \end{equation} \]
Free and dummy indices
Definition 1 A free index is an index that appears only once.
Definition 2 A dummy index is an index that appears exactly twice.
\[ \begin{equation} \underbrace { A_{im}x_{m} } _{ \text{Free}=\{i\}\\ \text{Dummy}=\{m\} } +\underbrace{ b_i} _{ \text{Free}=\{i\}\\ \text{Dummy}=\emptyset }+\underbrace{C_{ikq} x_{k} y_{q}}_{ \text{Free}=\{i\}\\ \text{Dummy}=\{k,q\} } \end{equation} \]
Rules
Rule 1.1: A dummy index implies a sum w.r.t that index.
Rule 1.2: In Einstein summation convention, an index cannot appear more than twice
Examples
- \[\begin{align} x_{i}y_{i} & = x_1y_1+ x_2 y_2+ x_3 y_3 \end{align}\]
- \[\begin{align} m^{j}v_{j} & = m^{1}v_{1}+m^{2}v_{2}+m^{3}v_{3} \end{align}\]
- \[ \begin{align} a_{ij}x_i y_j&=x_ia_{i1} y_1+x_ia_{i2} y_2+x_ia_{i3} y_3\\ &=(x_1a_{11}+x_2a_{21}+x_3a_{31}) y_1 +(x_1a_{12}+x_2a_{22}+x_3a_{32}) y_2+ (x_1a_{13}+x_2a_{23}+x_3a_{33}) y_3\\ &=(a_{11}x_1y_1+a_{21}x_2y_1+a_{31}x_3y_1 +a_{12}x_1y_2+a_{22}x_2y_2+a_{32}x_3y_2 +a_{13}x_1y_3+a_{23}x_2y_3+a_{33}x_3y_3) \end{align} \]
- \[\begin{align}a_{ijk}x_ix_j x_k&=\sum_{i=1}^{3} \sum_{j=1}^{3} \sum_{k=1}^{3} a_{ijk}x_ix_j x_k \end{align} \]
- \[ \begin{align} x_i a_j x_j a_i&=\sum_{i=1}^{3} \sum_{j=1}^{3} x_i a_j x_j a_i\\ &= \sum_{i=1}^{3}x_i a_i \sum_{j=1}^{3} x_j a_j \end{align} \]
Non-examples
\[ \begin{equation} x_{jj}y_j \quad \text{(Not okay; j repeated thrice)} \end{equation} \]
Rule 2.a (Renaming of dummy indices): A dummy index can be renamed, as long as the new name does not conflict with the names of the other dummy or free indices in the same term
Examples
\[ \begin{align} x_iy_i&\Rightarrow x_j y_j\\ \end{align} \]
\[ \begin{align} a_ix_jy_j&\Rightarrow a_ix_k y_k, \quad (\text{OK, no conflict with free index}) \\ a_ix_jy_j&\not\Rightarrow a_ix_i y_i \quad (\text{conflict with free index})\\ \end{align} \]
\[ \begin{align} A_{im}x_{m}+b_i+C_{ikq}x_{k}y_{q} & \Rightarrow A_{im}x_{m}+b_i+C_{ipq}x_{p}y_{q}\quad (\text{OK, no conflict}) \\ A_{im}x_{m}+b_i+C_{ikq}x_{k}y_{q} & \Rightarrow A_{im}x_{m}+b_i+C_{imq}x_{m}y_{q}\quad (\text{OK, no conflict}) \\ A_{im}x_{m}+b_i+C_{ikq}x_{k}y_{q} & \not\Rightarrow A_{im}x_{m}+b_i+C_{iiq}x_{i}y_{q}\quad (\text{conflict with free index}) \\ A_{im}x_{m}+b_i+C_{ikq}x_{k}y_{q} & \not\Rightarrow A_{im}x_{m}+b_i+C_{iqq}x_{q}y_{q}\quad (\text{conflict with dummy index}) \\ A_{im}x_{m}+b_i+C_{ikq}x_{k}y_{q} &\not \Rightarrow A_{ii}x_{i}+b_i+C_{ikq}x_{k}y_{q}\quad (\text{conflict with free index}) \\ \end{align} \]
Rule 2.b: Different terms can have same dummy indices.
Examples
\[ \begin{equation} \underbrace { A_{im}x_m } _{ \{i\}\\ \{m\} } +\underbrace{ b_i} _{ \{i\}\\ \emptyset }+\underbrace{C_{ikq} x_{k} y_{q}}_{ \{i\}\\ \{k,q\} } \end{equation} \]
\(\Rightarrow\)
\[ \begin{equation} \underbrace { A_{im}x_m } _{ \{i\}\\ \{m\} } +\underbrace{ b_i} _{ \{i\}\\ \emptyset }+\underbrace{C_{ikq} x_{k} y_{q}}_{ \{i\}\\ \{k \rightarrow m,q\} } \end{equation} \]
\[ \begin{equation} \underbrace { A_{im}x_m } _{ \{i\}\\ \{m\} } +\underbrace{ b_i} _{ \{i\}\\ \emptyset }+\underbrace{C_{imq} x_{m} y_{q}}_{ \{i\}\\ \{m,q\} } \end{equation} \]
\(\Rightarrow\)
\[ \begin{equation} \underbrace { A_{im}x_m } _{ \{i\}\\ \{m\} } +\underbrace{ b_i} _{ \{i\}\\ \emptyset }+\underbrace{C_{imq} x_{m} y_{q}}_{ \{i\}\\ \{m \rightarrow i,q\} } \end{equation} \]
\(\Rightarrow\)
\[ \begin{equation} \underbrace { A_{im}x_m }_{ \{i\}\\ \{m\} } +\underbrace{ b_i} _{ \{i\}\\ \emptyset }+\underbrace{C_{iiq} x_{i} y_{q}}_{ \{i\}\\ \{i,q\} } \end{equation} \] (Not okay - conflict with free index)
Rule 3 (Free Indices): A free index denotes a set of \(n_{\rm{sd}}\) terms
Examples
\[ \begin{align} a_{i}&\Rightarrow\{a_1,a_2,a_3\}\\ a_{i}=b_{i}&\Rightarrow\{a_1=b_1,a_2=b_2,a_3=b_3\}\\ a_{i}+b_{i}&\Rightarrow\{a_1+b_1,a_2+b_2,a_3+b_3\}\\ \end{align} \]
Denoting ordered sets
\[(\left(A_{ij})_{j\in \mathcal{I}}\right)_{i\in \mathcal{I}}=(A_{ij})_{i,~j\in \mathcal{I}}\]
\[\left(\left(A_{1j}\right)_{j \in \mathcal{I}}, \left(A_{2j}\right)_{j\in \mathcal{I}}, \left(A_{3j}\right)_{j\in \mathcal{I}}\right) \]
Notice that in each expression, all terms have the same set of free indices. That is, take a look at the expression,
\[ \begin{equation} \underbrace { A_{im}x_{m} } _{ \text{Free}=\{i\}\\ \text{Dummy}=\{m\} } +\underbrace{ b_i} _{ \text{Free}=\{i\}\\ \text{Dummy}=\emptyset }+\underbrace{C_{ikq} x_{k} y_{q}}_{ \text{Free}=\{i\}\\ \text{Dummy}=\{k,q\} } \end{equation} \]
This is a consequence of the following rule:
Rule 4: All terms in an expression should have the same set of free indices.
Examples
\[ \begin{align} a_i +b_i\quad (\text{OK})\\ a_j +b_j\quad (\text{OK})\\ a_i +b_j\quad (\text{Not OK})\\ M_i =C_{mi}b_m\quad (\text{OK})\\ M_i =C_{mj}b_m\quad (\text{Not OK})\\ \end{align} \]
Rule 5: All terms in an expression should have the same set of free indices.
1.2 Kronecker-Delta symbol
Kronecker delta, denoted by \(\delta_{ij}\), is defined as:
$$ \[\begin{equation} \delta_{ij} = \begin{cases} 1 & i=j \\ 0 & i \neq j \end{cases} \end{equation}\]
$$
That is,
\(\delta_{11}=1\) \(\delta_{12}=0\) \(\delta_{13}=0\) \(\delta_{21}=0\) \(\delta_{22}=1\) \(\delta_{23}=0\) \(\delta_{31}=0\) \(\delta_{32}=0\) \(\delta_{33}=1\)
For example, \(\delta_{12}=0\) and \(\delta_{33}=1\).
Identity Matrix
\[ \mathbb{\mathsf{I}} = \left(\delta_{ij}\right)= \begin{pmatrix} 1 & 0 & 0 \\ 0 & 1 & 0\\ 0 & 0 & 1 \end{pmatrix}\\ \text{Read paranthesis as ``matrix representation of the Kronecker-delta symbol``} \] \[\left[\delta_{ij}\right]= \begin{bmatrix} 1 & 0 & 0 \\ 0 & 1 & 0\\ 0 & 0 & 1 \end{bmatrix}\]
Properties of the Kronecker Delta
- Replacement Property:
\[ \begin{align} \delta_{1m}a_m =\delta_{11}a_1+\delta_{12}a_2+\delta_{13}a_3\ \\ =1a_1+0a_2+0a_3\ \\ =a_1 \end{align} \]
\[ \begin{align} \delta_{2m}a_m = \delta_{21}a_1+\delta_{22}a_2+0a_3\ \\ =a_2 \end{align} \]
\[ \begin{align} \delta_{3m}a_m & = \delta_{31}a_1+\delta_{32}a_2+\delta_{33}a_3\ \\ =0a_1+0a_2+a_3\ \\ =a_3 \end{align} \]
So, \(\delta_{im}a_m = a_i\), \((\delta_{1m}a_m, \delta_{2m}a_m , \delta_{3m}a_m) = (a_1, a_2, a_3)\), \(\delta_{1m}a_m = a_i\).
\[ \begin{align} \delta_{1m}T_{mj} & = \delta_{11}T_{1j}+ \delta_{12}T_{2j}+\delta_{13}T_{3j} =T_{1j} \\ \delta_{2m}T_{mj} & = \delta_{21}T_{1j}+ \delta_{22}T_{2j}+\delta_{23}T_{3j} =T_{2j} \\ \delta_{3m}T_{mj} & = \delta_{31}T_{1j}+ \delta_{32}T_{2j}+\delta_{33}T_{3j} =T_{3j} \end{align} \]
\[ \begin{bmatrix} \delta_{1m}T_{mj} = T_{1j} \\ \delta_{2m}T_{mj} = T_{2j} \\ \delta_{3m}T_{mj} = T_{3j} \end{bmatrix} \] \(\Rightarrow\) \(\delta_{1m}T_{mj} = T_{1j}, \delta_{2m}T_{mj} = T_{2j}, \delta_{3m}T_{mj}=T_{3j}\). \[\delta_{im}T_{mj} = T_{ij}\]
We can generalize the last two results. If, in a term, the Kronecker delta and another symbol share an index, the index in that symbol can be replaced by the Kronecker delta’s other index in its symbol. For example:
- \(\delta_{im}a_{mb}\) \(\Rightarrow\) \(\delta_{ib}\)
- \(\delta_{mi}a_{mb}\) \(\Rightarrow\) \(a_{ib}\)
- \(\delta_{im}a_{pqms}\) \(\Rightarrow\) \(a_{pqis}\)
- \(\delta_{im}\delta_{mj}\) \(\Rightarrow\) \(\delta_{ij}\)
- \(\delta_{im}\delta_{mn}\delta_{nj}\) \(\Rightarrow\) \(\delta_{in}\delta_nj\) \(\Rightarrow\) \(\delta_{ij}\)
\[ \begin{align} \delta_{m1}T_{mj} &= \delta_{11}T_{1j} + \delta_{21}T_{2j}+ \delta_{31}T_{3j} = T_{1j} \\ \delta_{m2}T_{mj} &= \delta_{12}T_{1j} + \delta_{22}T_{2j}+ \delta_{32}T_{3j} = T_{2j} \\ \delta_{m3}T_{mj} &= \delta_{13}T_{1j} + \delta_{23}T_{2j}+ \delta_{33}T_{3j} = T_{3j} \end{align} \]
\[ \begin{align} \delta_{m1}T_{mj} &= T_{1j} \\ \delta_{m2}T_{mj} &= T_{2j} \\ \delta_{m3}T_{mj} &= T_{3j} \\ \delta_{mi}T_{mj} &= T_{ij} \end{align} \]