The Summation Convention: Free Indices, Dummy Indices, and What They Build

Writing ∑\sum in front of every matrix or vector product becomes tiresome very quickly. The summation convention, introduced by Einstein, removes the symbol by making one rule do the work:


Within a single term, an index that appears twice is summed over its whole range. The summation sign is understood and never written.


Throughout this lesson every index runs over 1,2,31,2,3 unless stated otherwise.

Free and dummy indices

  • A dummy (or summation) index appears exactly twice in a term. It is summed away, so it never appears in the result and can be renamed freely: aibia_i b_i and ambma_m b_m are the same number.
  • A free index appears exactly once in every term. It is not summed. Each free index adds one dimension to the result, and the free indices on the two sides of an equation must match exactly.
  • An index appearing three or more times in one term has no meaning under the convention. Such an expression is simply not well formed.


Counting free indices tells you the shape of the answer before you compute anything:


Expression

Free indices

Result

xiyix_i y_i

none

a scalar (the dot product)

AijxjA_{ij} x_j

ii

a vector with 3 components

AijBjkA_{ij} B_{jk}

i,ki,k

a 3×33\times 3 matrix (the product ABAB)

AiiA_{ii}

none

a scalar (the trace)

Worked example

Let


A=(2−1043510−2),x=(1,0,2).A=\begin{pmatrix}2&-1&0\\4&3&5\\1&0&-2\end{pmatrix}, \qquad x=(1,0,2).


The trace. In AiiA_{ii} the index ii appears twice, so it is summed:


Aii=A11+A22+A33=2+3+(−2)=3.A_{ii}=A_{11}+A_{22}+A_{33}=2+3+(-2)=3.


A matrix acting on a vector. In yi=Aijxjy_i = A_{ij}x_j the index jj appears twice and is summed, while ii appears once on each side and is free. So this single line is three equations:


y1=2(1)+(−1)(0)+0(2)=2,y2=4(1)+3(0)+5(2)=14,y3=1(1)+0(0)+(−2)(2)=−3.y_1 = 2(1)+(-1)(0)+0(2)=2,\quad y_2 = 4(1)+3(0)+5(2)=14,\quad y_3 = 1(1)+0(0)+(-2)(2)=-3.


Hence y=(2,14,−3)y=(2,14,-3). Notice that the free index ii is exactly what makes this a vector equation rather than a single number.

Reading index order carefully

Index position carries all the structural information, so a swap changes the meaning:


  • (AT)ij=Aji(A^{\mathsf T})_{ij}=A_{ji}, which is the whole content of "transpose".
  • (AB)ik=AijBjk(AB)_{ik}=A_{ij}B_{jk}: the dummy index is adjacent, inner to inner. By contrast AijBkj=(ABT)ikA_{ij}B_{kj}=(AB^{\mathsf T})_{ik}, because the second factor is being read along its rows.
  • AijBji=tr⁡(AB)A_{ij}B_{ji}=\operatorname{tr}(AB): both indices are dummies, so the result is a scalar.

The Kronecker delta

Define δij=1\delta_{ij}=1 if i=ji=j and 00 otherwise, the entries of the identity matrix. Its usefulness is that summing against it performs a substitution: in δijvj\delta_{ij}v_j every term with j≠ij\neq i vanishes, leaving


δijvj=vi.\delta_{ij}v_j = v_i.


Read it as "the delta renames the index". The same move gives δijAjk=Aik\delta_{ij}A_{jk}=A_{ik} and δii=3\delta_{ii}=3, not 11, because δii\delta_{ii} is a sum of three diagonal entries.

A checking habit

Before computing, count. List the indices in your expression and mark each as free or dummy.


If an index appears once, it must appear once on the other side of the equation too; if it appears twice, it must not appear in the answer; if it appears three times, stop and rewrite.


Most errors with this notation are caught by that check alone, before a single number is multiplied.

All indices in this question run over 1,2,31,2,3, and the summation convention is in force.

Consider the expression aijbja_{ij}b_j. Which statement about it is correct?

Choose an answer

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors

Indices run over 1,2,31,2,3 and the summation convention applies. Let

B=(1203−11214),u=(2,1,−1).B=\begin{pmatrix}1&2&0\\3&-1&1\\2&1&4\end{pmatrix}, \qquad u=(2,1,-1).

Evaluate the scalar uiBijuju_i B_{ij} u_j.

Give your answer as a number.

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors

Let AA and BB be 3×33\times 3 matrices, with all indices running over 1,2,31,2,3 and the summation convention in force. Which expression equals the (2,3)(2,3) entry of the product ABAB?

Choose an answer

Create a free account to get started

Unlock full course notes by signing up

  • Video lessons taught by top educators
  • Practice questions with guided solutions
  • Written review notes for reference
  • Expert answers anytime from course tutors