components of a vector shown in Figure 1-1. If we choose a particular coordinate system that has
been rotated so that the x
3
-axis lies along the direction of the vector, then the vector will have
components v ¼ {0, 0, jvj}. For this case, two of the components have been reduced to zero,
while the remaining component becomes the largest possible (the total magnitude).
This situation is most useful for symmetric second-order tensors that eventually represent
the stress and/or strain at a point in an elastic solid. The direction determined by the unit vector
n is said to be a principal direction or eigenvector of the symmetric second-order tensor a
ij
if
there exists a parameter l such that
a
ij
n
j
¼ ln
i
(1:6:1)
where l is called the principal value or eigenvalue of the tensor. Relation (1.6.1) can be
rewritten as
(a
ij
ld
ij
)n
j
¼ 0
and this expression is simply a homogeneous system of three linear algebraic equations in the
unknowns n
1
, n
2
, n
3
. The system possesses a nontrivial solution if and only if the determinant
of its coefficient matrix vanishes; that is:
det[a
ij
ld
ij
] ¼ 0
Expanding the determinant produces a cubic equation in terms of l:
det[a
ij
ld
ij
] ¼l
3
þ I
a
l
2
II
a
l þ III
a
¼ 0(1:6:2)
where
I
a
¼ a
ii
¼ a
11
þ a
22
þ a
33
II
a
¼
1
2
(a
ii
a
jj
a
ij
a
ij
) ¼
a
11
a
12
a
21
a
22
þ
a
22
a
23
a
32
a
33
þ
a
11
a
13
a
31
a
33
III
a
¼ det[a
ij
]
(1:6:3)
The scalars I
a
, II
a
, and III
a
are called the fundamental invariants of the tensor a
ij
, and relation
(1.6.2) is known as the characteristic equation. As indicated by their name, the three invariants
do not change value under coordinate transformation. The roots of the characteristic equation
determine the allowable values for l , and each of these may be back-substituted into relation
(1.6.1) to solve for the associated principal direction n.
Under the condition that the components a
ij
are real, it can be shown that all three roots
l
1
, l
2
, l
3
of the cubic equation (1.6.2) must be real. Furthermore, if these roots are distinct, the
principal directions associated with each principal value are orthogonal. Thus, we can con-
clude that every symmetric second-order tensor has at least three mutually perpendicular
principal directions and at most three distinct principal values that are the roots of the
characteristic equation. By denoting the principal directions n
(1)
, n
(2)
, n
(3)
corresponding to
the principal values l
1
, l
2
, l
3
, three possibilities arise:
1. All three principal values are distinct; thus, the three corresponding principal directions
are unique (except for sense).
2. Two principal values are equal (l
1
6¼ l
2
¼ l
3
); the principal direction n
(1)
is unique
(except for sense), and every direction perpendicular to n
(1)
is a principal direction
associated with l
2
, l
3
.
Mathematical Preliminaries 13