16.
21 Techniques of Structural Analysis and
Design
Spring 2005
Unit #2 Mathematical aside: Vectors,
indicial notation and summation convention
Raúl Radovitzky
February 6, 2005
Indicial notation
In 16.21 we’ll work in a an euclidean threedimensional space R3 .
Free index: A subscript index ()i will be denoted a free index if it is
not repeated in the same additive term where the index appears. Free means
that the index represents all the values in its range.
• Latin indices will range from 1 to, (i, j, k, ... = 1, 2, 3),
• greek indices will range from 1 to 2, (α, β, γ, ... = 1, 2).
Examples:
1. ai1 implies a11 , a21 , a31 . (one free index)
2. xα yβ implies x1 y1 , x1 y2 , x2 y1 , x2 y2 (two free indices).
3. aij implies a11 , a12 , a13 , a21 , a22 , a23 , a31 , a32 , a33 (two free indices implies
9 values).
4.
∂σij
+ bi = 0
∂xj
1
has a free index (i), therefore it represents three equations:
∂σ1j
+ b1 = 0
∂xj
∂σ2j
+ b2 = 0
∂xj
∂σ3j
+ b3 = 0
∂xj
Summation convention: When a repeated index is found in an expression
(inside an additive term) the summation of the terms ranging all the possible
values of the indices is implied, i.e.:
3
�
ai b i = a i b i = a 1 b 1 + a 2 b 2 + a3 b 3
i=1
Note that the choice of index is immaterial:
ai bi = ak bk
Examples:
1. aii = a11 + a22 + a33
2. ti = σij nj implies the three equations (why?):
t1 = σ11 n1 + σ12 n2 + σ13 n3
t2 = σ21 n1 + σ22 n2 + σ23 n3
t3 = σ31 n1 + σ32 n2 + σ33 n3
Other important rules about indicial notation:
1. An index cannot appear more than twice in a single additive term, it’s
either free or repeated only once.
ai = bij cj dj is INCORRECT
2. In an equation the lhs and rhs, as well as all the terms on both sides
must have the same free indices
• ai bk = cij dkj free indices i, k, CORRECT
• ai bk = cij dkj +ei fjj +gk pi qr INCORRECT, second term is missing
free index k and third term has extra free index r
2
Vectors
A basis in R3 is given by any set of linearly independent vectors ei , (e1 , e2 , e3 ).
From now on, we will assume that these basis vectors are orthonormal, i.e.,
they have a unit length and they are orthogonal with respect to each other.
This can be expressed using dot products:
e1 .e1 = 1, e2 .e2 = 1, e3 .e3 = 1,
e1 .e2 = 0, e1 .e3 = 0, e2 .e3 = 0, ...
Using indicial notation we can write these expression in very succinct form
as follows:
ei .ej = δij
In the last expression the symbol δij is defined as the Kronecker delta :
�
1 if i = j,
δij =
� j
0 if i =
Example:
ai δij =a1 δ11 + a2 δ21 + a3 δ31 ,
a1 δ12 + a2 δ22 + a3 δ32 ,
a1 δ13 + a2 δ23 + a3 δ33
=a1 1 + a2 0 + a3 0,
a1 0 + a2 1 + a3 0,
a1 0 + a 2 0 + a 3
=a1 ,
a2 ,
a3
or more succinctly: ai δij = aj , i.e., the Kronecker delta can be thought of
an “index replacer”.
A vector v will be represented as:
v = v i e i = v 1 e1 + v 2 e 2 + v 3 e 3
3
The vi are the components of v in the basis ei . These components are the
projections of the vector on the basis vectors:
v = v j ej
Taking the dot product with basis vector ei :
[Link] = vj (ej .ei ) = vj δji = vi
Transformation of basis
Given two bases ei , ẽk and a vector v whose components in each of these
bases are vi and ṽk , respectively, we seek to express the components in basis
in terms of the components in the other basis. Since the vector is unique:
v = ṽm˜
em = vn en
Taking the dot product with ẽi :
v.˜ em .˜
ei = ṽm (˜ ei ) = vn (en .˜
ei )
But v˜m (˜
em .˜
ei ) = ṽm δmi = ṽi from which we obtain:
ṽi = v.˜
ei = vj (ej .˜
ei )
Note that (ej .ẽi ) are the direction cosines of the basis vectors of one basis
on the other basis:
ei = �ej ��˜
ej .˜ ei � cos e�
j ẽi
