Appendix B

Tensor-indicial notation in the

approximation of elasticity
problems
Introduction
The matrix type of notation used in this volume for description of tensor quantities
such as stresses and strains is compact and we believe easy to understand. However, in
a computer program each quantity often will still have to be identied by appropriate
indices and the conciseness of matrix notation does not always carry over to the pro-
gramming steps. Further, many readers are accustomed to the use of indicial-tensor
notation which is a standard tool in the study of solid mechanics. For this reason we
summarize here the formulation of the nite element arrays in an indicial form.
Some advantages of such reformulation from the matrix setting become apparent
when evaluation of stiffness arrays for isotropic materials is considered. Here some
multiplicationoperations previouslynecessarybecome redundant andthe element mod-
ule programs can be written more economically.
When nite deformation problems in solid mechanics have to be considered the use
of indicial notation is almost essential to formmany of the arrays needed for the residual
and tangent terms.
This appendix adds little new to the discretization ideas it merely repeats in a
different language the results already presented.
Indicial notation: summation convention
A point P in three-dimensional space may be represented in terms of its cartesian
coordinates x
i
, i = 1, 2, 3. The limits that i can take dene its range. To dene these
components we must rst establish an oriented orthogonal set of coordinate directions
as shown in Fig. B.1. The distance from the origin of the coordinate axes to the point
dene a position vector x. If along each of the coordinate axes we dene the set of unit
orthonormal base vectors, i
i
, i = 1, 2, 3 which have the property
i
i
i
j
=
ij
=
_
1 for i = j
0 for i = j
(B.1)
Indicial notation: summation convention 675
x
1
x
2
i
1
i
2
x
x
1
(P)
x
2
(P)
P
Fig. B.1 Orthogonal axes and a point: Cartesian coordinates.
where ( ) ( ) denotes the vector dot product. The components of the position vector
are constructed from the vector dot product
x
i
= i
i
x; i = 1, 2, 3 (B.2)
From this construction it is easy to observe that the vector x may be represented as
x =
3

i=1
x
i
i
i
(B.3)
In dealing with vectors, and later tensors, the form x is called the intrinsic notation of
the coordinates and x
i
i
i
the indicial form. An intrinsic form is a physical entity which
is independent of the coordinate system selected, whereas an indicial form depends on
a particular coordinate system.
To simplify notation we adopt the common convention that any index which is
repeated in any given term implies a summation over the range of the index. Thus, our
short-hand notation for Eq. (B.3) is
x = x
i
i
i
= x
1
i
1
+x
2
i
2
+x
3
i
3
(B.4)
For two-dimensional problems, unless otherwise stated, it will be understood that the
range of the index is two.
Similarly, we can dene the components of the displacement vector u as
u = u
i
i
i
(B.5)
Note that the components (u
1
, u
2
, u
3
) replace the components (u, v, w) used through-
out most of this volume.
To avoid confusion with nodal quantities to which we previously also attached sub-
scripts we shall simply change their position to a superscript. Thus, u
a
2
has the same
meaning as v
a
used previously, etc.
Often in an indicial form of equations the base vectors are omitted from the nal equation.
676 Tensor-indicial notation in the approximation of elasticity problems
Derivatives and tensorial relations
Inindicial notationthe derivative of anyquantitywithrespect toa coordinate component
x
i
is written compactly as

x
i
( )
,i
(B.6)
Thus we can write the gradient of the displacement vector as
u
i
x
j
u
i,j
; i, j = 1, 2, 3 (B.7)
In a cartesian coordinate system the base vectors do not change their magnitude or
direction along any coordinate direction. Accordingly their derivatives with respect to
any coordinate is zero as indicated in Eq. B.8
i
i
x
j
= i
i,j
= 0 (B.8)
Thus, in cartesian coordinates the derivative of the intrinsic displacement u is given by
u
,j
= u
i,j
i
i
+u
i
i
i,j
= u
i,j
i
i
(B.9)
The collection of all the derivatives denes the displacement gradient which we write
in intrinsic notation as
u = u
i,j
i
i
i
j
(B.10)
The symbol denotes the tensor product between two base vectors and since only
two vectors are involved the gradient of the displacement is called second rank. The
notation used to dene a tensor product follows that used in reference 1.
Any second rank intrinsic quantity can be split into symmetric and a skewsymmetric
(anti-symmetric) parts as
A =
1
2
_
A +A
T

+
1
2
_
A A
T

= A
(s)
+A
(a)
(B.11)
where A and its transpose have cartesian components
A = A
ij
i
i
i
j
; A
T
= A
ji
i
i
i
j
(B.12)
The symmetric part of the displacement gradient denes the (small) strain
= u
(s)
=
1
2
_
u +(u)
T

=
1
2
_
u
i,j
+u
j,i

i
i
i
j
=
ij
i
i
i
j
=
ji
i
i
i
j
(B.13)
Note that this denition is slightly different from that occurring in Chapters 2 to 6. Now the shearing strain is
given by
ij
= 1/2
ij
when i = j.
Coordinate transformation 677
and the skew symmetric part gives the (small) rotation
= u
(a)
=
1
2
_
u (u)
T

=
1
2
_
u
i,j
u
j,i

i
i
i
j
=
ij
i
i
i
j
=
ji
i
i
i
j
(B.14)
The strain expression is analogous to Eq. (2.13). The components
ij
and
ij
may be
represented by a matrix as

ij
=
_

11

12

13

21

22

23

31

32

33
_
=
_

11

12

13

12

22

23

13

23

33
_
(B.15)

ij
=
_
0
12

13

21
0
23

31

32
0
_
=
_
0
12

13

12
0
23

13

23
0
_
(B.16)
Coordinate transformation
Consider now the representation of the intrinsic coordinates in a system which has
different orientation than that given in Fig. B.1. We represent the components in the
new system by
x = x

(B.17)
Using Eq. (B.2) we can relate the components in the prime system to those in the origi-
nal system as
x

= i

i
x = i

i
i
j
x
j
=
i

j
x
j
(B.18)
where

j
= i

i
j
= cos(x

, x
j
) (B.19)
dene the direction cosines of the coordinate in a manner similar to that of Eq. (6.18).
Equation (B.18) denes how the cartesian coordinate components transform from
one coordinate frame to another. Recall that summation convention implies
x

=
i

1
x
1
+
i

2
x
2
+
i

3
x
3
i

= 1, 2, 3 (B.20)
In Eq. (B.18) i

is called a free index whereas j is called a dummy index since it may

be replaced by any other unique index without changing the meaning of the term (note
that the notation used does not permit an index to appear more than twice in any term).
Summation convention will be employed throughout the remainder of this discussion
and the reader should ensure that the concept is fully understood before proceeding.
Some examples will be given occasionally to illustrate its use.
Using the notion of the direction cosines, Eq. (B.18) may be used to transform
any vector with three components. Thus, transformation of the components of the
displacement vector is given by
u

=
i

j
u
j
i

, j = 1, 2, 3 (B.21)
678 Tensor-indicial notation in the approximation of elasticity problems
Indeed we can also use the above to express the transformation for the base vectors
since
i

=
_
i

i
j
_
i
j
=
i

j
i
j
(B.22)
Similarly, by interchanging the role of the base vectors we obtain
i
j
=
_
i
j
i

_
i

=
i

j
i

(B.23)
which indicates that the inverse of the direction cosine coefcient array is the same as
its transpose.
The strain transformation follows from the intrinsic form written as
=

i
i
i
j
=
kl
i
k
i
l
(B.24)
Substitution of the base vectors from Eq. (B.23) into Eq. (B.24) gives
=
i

kl

l
i
i
i
j
(B.25)
Comparing Eq. (B.25) with Eq. (B.24) the components of the strain transformaccording
to the relation

=
i

kl

l
(B.26)
Variables that transform according to Eq. (B.21) are called rst rank cartesian ten-
sors whereas quantities that transform according to Eq. (B.26) are called second rank
cartesian tensors. The use of indicial notation in the context of cartesian coordinates
will lead naturally to each mechanics variable being dened in terms of a cartesian
tensor of an appropriate rank.
Stress may be written in terms of its components
ij
which may be written in a matrix
form similar to Eq. (B.15)

ij
=
_

11

12

13

21

22

23

31

32

33
_
; i, j = 1, 2, 3 (B.27)
In intrinsic form stress is given by
=
ij
i
i
i
j
(B.28)
and, using similar logic as used for strain, can be shown to transform as a second rank
cartesian tensor. The symmetry of the components of stress may be established by
summing moments (angular momentum balance) about each of the coordinate axes to
obtain

ij
=
ji
(B.29)
Equilibrium and energy
Introducing a body force vector
b = b
i
i
i
(B.30)
Elastic constitutive equations 679
we can write the static equilibrium equations (linear momentum balance) for a differ-
ential element as
div +b
_

ji,j
+b
i
_
i
i
= 0 (B.31)
where the repeated index again implies summation over the range of the index, i.e.,

ji,j

3

j=1

ji,j
=
1i,1
+
2i,2
+
3i,3
Note that the free index i must appear in each term for the equation to be meaningful.
As a further example of the summation convention consider an internal energy term
W =
ij

ij
(B.32)
This expression implies a double summation; hence summing rst on i gives
W =
1j

1j
+
2j

2j
+
3j

3j
and then summing on j gives nally
W =
11

11
+
12

12
+
13

13
+
21

21
+
22

22
+
23

23
+
31

31
+
32

32
+
33

33
We may use symmetry conditions on
ij
and
ij
to reduce the nine terms to six terms.
Accordingly,
W =
11

11
+
22

22
+
33

33
+2(
12

12
+
23

23
+
31

31
)
=
11

11
+
22

22
+
33

33
+
12

12
+
23

23
+
31

31
(B.33)
Following a similar expansion we can also show the result

ij

ij
0 (B.34)
Elastic constitutive equations
For an elastic material the most general linear relationship we can write for components
of the stressstrain characterization is

ij
= D
ijkl
_

kl

0
kl
_
+
0
ij
(B.35)
Equation (B.35) is the equivalent of Eq. (2.16) but now written in indicial notation. We
note that the elastic moduli which appear in Eq. (B.35) are components of the fourth
rank tensor
D = D
ijkl
i
i
i
j
i
k
i
l
(B.36)
The elastic moduli possess the following symmetry conditions
D
ijkl
= D
jikl
= D
ijlk
= D
klij
(B.37)
680 Tensor-indicial notation in the approximation of elasticity problems
the latter arising from the existence of an internal energy density in the form
2
W() =
1
2

ij
D
ijkl

kl
+
ij
_

0
ij
D
ijkl

0
kl

(B.38)
which yields the stress from

ij
=
W

ij
(B.39)
By writing the constitutive equation with respect to x

and using properties of the

base vectors we can deduce the transformation equation for moduli as
D

=
i

q
D
mnpq
(B.40)
A common notation for the intrinsic form of Eq. (B.35) is
= D :
_

0
_
+
0
(B.41)
in which : denotes the double summation (contraction) between the elastic moduli and
the strains.
The elastic moduli for an isotropic elastic material may be written in indicial form
as
D
ijkl
=
ij

kl
+(
ik

jl
+
il

jk
) (B.42)
where , are the Lam e constants. An isotropic linear elastic material is always
characterized by two independent elastic constants. Instead of the Lam e constants we
can use Youngs modulus, E, and Poissons ratio, , to characterize the material. The
Lam e constants may be deduced from
=
E
2(1 +)
and =
E
(1 +)(1 2)
(B.43)
Finite element approximation
If we now introduce the nite element displacement approximation given by Eq. (2.1),
using indicial notation we may write for a single element
u
i
u
i
= N
a
u
a
i
i = 1, 2, 3; a = 1, 2, . . . , n (B.44)
where n is the total number of nodes on an element. The strain approximation in each
element is given by the denition of Eq. (B.13) as

ij
=
1
2
_
N
a,j
u
a
i
+N
a,i
u
a
j

(B.45)
The internal virtual work for an element is given as
U
I
=
_

e
: d =
_

ij

ij
d (B.46)
Finite element approximation 681
Using Eqs (B.45) and (B.46) and noting symmetries in D
ijkl
we may write the internal
virtual work for a linear elastic material as
U
I
= u
a
i
_

e
N
a,j
D
ijkl
N
b,l
d u
b
k
+ u
a
i
_

e
N
a,j
_

0
ij
D
ijkl

0
kl
_
d
(B.47)
which replaces in indicial notation the matrix form presented in Chapters 2 to 6.
In describing a stiffness coefcient two subscripts have been used previously and
the submatrix K
ab
implied 2 2 or 3 3 entries for the ab nodal pair, depending on
whether two- or three-dimension displacement components were involved. Now the
scalar components
K
ab
ij
i, j = 1, 2, 3; a, b = 1, 2, . . . , n (B.48)
dene completely the appropriate stiffness coefcient with ij indicating the relative
submatrix position (in this case for a three-dimensional displacement).
Note that for a symmetric matrix we have previously required that
K
ab
= K
T
ba
(B.49)
In indicial notation the same symmetry is implied if
K
ab
ij
= K
ba
ji
(B.50)
The stiffness tensor is now dened from Eq. (B.47) as
K
ab
ik
=
_

e
N
a,j
D
ijkl
N
b,l
d (B.51)
When the elastic properties are constant over the element we may separate the integra-
tion from the material constants by dening
W
ab
ij
=
_

e
N
a,i
N
b,j
d (B.52)
and then perform the summations with the material moduli as
K
ab
ik
= W
ab
jl
D
ijkl
(B.53)
In the case of isotropy a particularly simple result is obtained
K
ab
ik
= W
ab
ik
+[W
ab
ki
+
ik
W
ab
jj
] (B.54)
which allows the construction of the stiffness to be carried out using fewer arithmetic
operations as compared with the use of matrix form.
3
Using indicial notation the nal equilibrium equations of the system are written as
K
ab
ik
u
b
k
+f
a
i
= 0 i = 1, 2, 3 (B.55)
and in this scalar formevery coefcient is simply identied. The reader can, as a simple
exercise, complete the derivation of the force terms due to the initial strain
0
ij
, stress

0
ij
, body force b
i
and external traction

t
i
.
Indicial notation is at times useful in clarifying individual terms, and this introduction
should be helpful as a key to reading some of the current literature.
682 Tensor-indicial notation in the approximation of elasticity problems
Table B.1 Mapping between matrix and tensor indices for second
rank symmetric tensors
Form Index number
Matrix 1 2 3 4 5 6
Tensor 11 22 33 12 & 21 23 & 32 31 & 13
xx yy zz xy & yx yz & zy zx & xz
Relation between indicial and matrix notation
The matrix form used throughout most of this volume can be deduced from the indicial
form by a simple transformation between the indices. The relationship between the
indices of the secondranktensors andtheir correspondingmatrixformcanbe performed
by an inspection of the ordering in the matrix for stress and its representation shown
in Eq. (B.27). In the matrix form the stress was given in Chapter 6 as
=
_

11

22

33

12

23

31

T
(B.56)
This form includes use of symmetry of stress components. The mapping of the indices
follows that shown in Table B.1.
Table B.1 may also be used to perform the map of the material moduli by noting that
the components in the energy are associated with the index pairs from the stress and
the strain. Accordingly, the moduli transform as
D
1111
D
11
; D
2233
D
23
; D
1231
D
46
; etc. (B.57)
The symmetry of the stress and strain is imbedded in Table B.1 and existence of an
energy function yields symmetry of the modulus matrix, i.e., D
ab
= D
ba
.
References
1. P. Chadwick. Continuum Mechanics. John Wiley & Sons, NewYork, 1976.
2. I.S. Sokolnikoff. The Mathematical Theory of Elasticity. McGraw-Hill, NewYork, 2nd edition,
1956.
3. A.K. Gupta and B. Mohraz. A method of computing numerically integrated stiffness matrices.
Int. J. Numer. Meth. Eng., 5:8389, 1972.