Chapter 5
Fourth-Order Tensors
5.1 Fourth-Order Tensors as a Linear Mapping
Fourth-order tensors play an important role in continuum mechanics where they
appear as elasticity and compliance tensors. In this section we define fourth-order
tensors and learn some basic operations with them. To this end, we consider a set
Linn of all linear mappings of one second-order tensor into another one within Linn .
Such mappings are denoted by a colon as
Y = A : X, A ∈ Linn , Y ∈ Linn , ∀X ∈ Linn . (5.1)
The elements of Linn are called fourth-order tensors.
Example 5.1. Elasticity and compliance tensors. A constitutive law of a linearly
elastic material establishes a linear relationship between the Cauchy stress tensor
σ and Cauchy strain tensor . Since these tensors are of the second-order a linear
relation between them can be expressed by fourth-order tensors like
σ = C : or = H : σ. (5.2)
The fourth-order tensors C and H describe properties of the elastic material and are
called the elasticity and compliance tensor, respectively.
Linearity of the mapping (5.1) implies that
A : (X + Y) = A : X + A : Y, (5.3)
A : (αX) = α (A : X) , ∀X, Y ∈ Linn , ∀α ∈ R, A ∈ Linn . (5.4)
Similarly to second-order tensors one defines the product of a fourth-order tensor
with a scalar
(αA) : X = α (A : X) = A : (αX) (5.5)
© Springer Nature Switzerland AG 2019 123
M. Itskov, Tensor Algebra and Tensor Analysis for Engineers,
Mathematical Engineering, https://doi.org/10.1007/978-3-319-98806-1_5
124 5 Fourth-Order Tensors
and the sum of two fourth-order tensors by
(A + B) : X = A : X + B : X, ∀X ∈ Linn . (5.6)
Further, we define the zero-tensor O of the fourth-order by
O : X = 0, ∀X ∈ Linn . (5.7)
Thus, summarizing the properties of fourth-order tensors one can write similarly to
second-order tensors
A + B = B + A, (addition is commutative), (5.8)
A + (B + C) = (A + B) + C, (addition is associative), (5.9)
O + A = A, (5.10)
A + (−A) = O, (5.11)
α (βA) = (αβ) A, (multiplication by scalars is associative), (5.12)
1A = A, (5.13)
α (A + B) = αA + αB, (multiplication by scalars is distributive
with respect to tensor addition), (5.14)
(α + β) A = αA + βA, (multiplication by scalars is distributive
with respect to scalar addition), ∀A, B, C ∈ Linn , ∀α, β ∈ R. (5.15)
Thus, the set of fourth-order tensors Linn forms a vector space.
On the basis of the “right” mapping (5.1) and the scalar product of two second-
order tensors (1.147) we can also define the “left” mapping by
(Y : A) : X = Y : (A : X) , Y ∈ Linn , ∀X ∈ Linn . (5.16)
5.2 Tensor Products, Representation of Fourth-Order
Tensors with Respect to a Basis
For the construction of fourth-order tensors from second-order ones we introduce
two tensor products as follows
A ⊗ B : X = AXB, A B : X = A (B : X) , ∀X ∈ Linn , (5.17)
where A, B ∈ Linn . Note, that the tensor product “⊗” (5.17)1 applied to second-
order tensors differs from the tensor product of vectors (1.83). One can easily show
that the mappings described by (5.17) are linear and therefore represent fourth-order
tensors. Indeed, we have, for example, for the tensor product “⊗” (5.17)1
5.2 Tensor Products, Representation of Fourth-Order … 125
A ⊗ B : (X + Y) = A (X + Y) B
= AXB + AYB = A ⊗ B : X + A ⊗ B : Y, (5.18)
A ⊗ B : (αX) = A (αX) B = α (AXB)
= α (A ⊗ B : X) , ∀X, Y ∈ Linn , ∀α ∈ R. (5.19)
With definitions (5.17) in hand one can easily prove the following identities
A ⊗ (B + C) = A ⊗ B + A ⊗ C, (B + C) ⊗ A = B ⊗ A + C ⊗ A, (5.20)
A (B + C) = A B + A C, (B + C) A = B A + C A. (5.21)
For the left mapping (5.16) the tensor products (5.17) yield
Y : A ⊗ B = AT YBT , Y : A B = (Y : A) B. (5.22)
As fourth-order tensors represent vectors they can be given with respect to a basis in
Linn .
Theorem 5.1. Let F = {F1 , F2 , . . . , Fn 2 } and G = {G1 , G2 , . . . , Gn 2 } be two arbi-
trary (not necessarily distinct) bases of Linn . Then, fourth-order tensors Fi
G j i, j = 1, 2, . . . , n 2 form a basis of Linn . The dimension of Linn is thus n 4 .
Proof. See the proof of Theorem 1.7.
A basis in Linn can be represented in another way as by the tensors Fi
G j i, j = 1, 2, . . . , n 2 . To this end, we prove the following identity
(a ⊗ d) (b ⊗ c) = a ⊗ b ⊗ c ⊗ d, (5.23)
where we set
(a ⊗ b) ⊗ (c ⊗ d) = a ⊗ b ⊗ c ⊗ d. (5.24)
Indeed, let X ∈ Linn be an arbitrary second-order tensor. Then, in view of (1.146)
and (5.17)2
(a ⊗ d) (b ⊗ c) : X = (bXc) (a ⊗ d) . (5.25)
For the right hand side of (5.23) we obtain the same result using (5.17)1 and (5.24)
a ⊗ b ⊗ c ⊗ d : X = (a ⊗ b) ⊗ (c ⊗ d) : X = (bXc) (a ⊗ d) . (5.26)
For the left mapping (5.16) it thus holds
Y : a ⊗ b ⊗ c ⊗ d = (aYd) (b ⊗ c) . (5.27)
Now, we are in a position to prove the following theorem.
126 5 Fourth-Order Tensors
Theorem
5.2. Let E = {e1 , e2 , . . . , en }, F = f 1 , f 2 , . . . , f n , G = g 1 , g 2 , . . . ,
g n and finally H = {h1 , h2 , . . . , hn } be four arbitrary (not necessarily distinct)
bases of En . Then, fourth-order tensors ei ⊗ f j ⊗ g k ⊗ hl (i, j, k, l = 1, 2, . . . , n)
represent a basis of Linn .
Proof. In view of (5.23)
ei ⊗ f j ⊗ g k ⊗ hl = (ei ⊗ hl ) f j ⊗ g k .
According to Theorem 1.7 the second-order tensors ei ⊗ hl (i, l = 1, 2, . . . , n) on
the one hand and f j ⊗ g k ( j, k = 1, 2, . . . , n) on the other hand form bases of
Linn . According to Theorem 5.1 the fourth-order tensors (ei ⊗ hl ) f j ⊗ g k and
consequently ei ⊗ f j ⊗ g k ⊗ hl (i, j, k, l = 1, 2, . . . , n) represent thus a basis of
Linn .
As a result of this Theorem any fourth-order tensor can be represented by
A = Ai jkl g i ⊗ g j ⊗ g k ⊗ gl = Ai jkl g i ⊗ g j ⊗ g k ⊗ gl
ij
= A· ·kl g i ⊗ g j ⊗ g k ⊗ gl = . . . (5.28)
The components of A appearing in (5.28) can be expressed by
Ai jkl = g i ⊗ gl : A : g j ⊗ g k , Ai jkl = g i ⊗ gl : A : g j ⊗ g k ,
ij
A· ·kl = g i ⊗ gl : A : g j ⊗ g k , i, j, k, l = 1, 2, . . . , n. (5.29)
By virtue of (1.113), (5.17)1 and (5.22)1 the right and left mappings with a second-
order tensor (5.1) and (5.16) can thus be represented by
A : X = Ai jkl g i ⊗ g j ⊗ g k ⊗ gl : Xq p g q ⊗ g p = Ai jkl X jk g i ⊗ gl ,
X : A = Xq p g q ⊗ g p : Ai jkl g i ⊗ g j ⊗ g k ⊗ gl = Ai jkl Xil g j ⊗ g k . (5.30)
We observe that the basis vectors of the second-order tensor are scalarly multiplied
either by the “inner” (right mapping) or “outer” (left mapping) basis vectors of the
fourth-order tensor.
5.3 Special Operations with Fourth-Order Tensors
Similarly to second-order tensors one defines also for fourth-order tensors some
specific operations which are not generally applicable to conventional vectors in the
Euclidean space.
5.3 Special Operations with Fourth-Order Tensors 127
Composition. In analogy with second-order tensors we define the composition of
two fourth-order tensors A and B denoted by A : B as
(A : B) : X = A : (B : X) , ∀X ∈ Linn . (5.31)
For the left mapping (5.16) one can thus write
Y : (A : B) = (Y : A) : B, ∀Y ∈ Linn . (5.32)
For the tensor products (5.17) the composition (5.31) further yields
(A ⊗ B) : (C ⊗ D) = (AC) ⊗ (DB) , (5.33)
(A ⊗ B) : (C D) = (ACB) D, (5.34)
(A B) : (C ⊗ D) = A CT BDT , (5.35)
(A B) : (C D) = (B : C) A D, A, B, C, D ∈ Linn . (5.36)
For example, the identity (5.33) can be proved within the following steps
(A ⊗ B) : (C ⊗ D) : X = (A ⊗ B) : (CXD)
= ACXDB = (AC) ⊗ (DB) : X, ∀X ∈ Linn ,
where we again take into account the definition of the tensor product (5.17).
For the component representation (5.28) we further obtain
A : B = Ai jkl g i ⊗ g j ⊗ g k ⊗ gl : B pqr t g p ⊗ g q ⊗ gr ⊗ g t
= Ai jkl B jqr k g i ⊗ g q ⊗ gr ⊗ gl . (5.37)
Note that the “inner” basis vectors of the left tensor A are scalarly multiplied with
the “outer” basis vectors of the right tensor B.
The composition of fourth-order tensors also gives rise to the definition of
powers as
Ak = A : A : . . . : A, k = 1, 2, . . . , A0 = I, (5.38)
k times
where I stands for the fourth-order identity tensor to be defined in the next section.
By means of (5.33) and (5.36) powers of tensor products (5.17) take the following
form
(A ⊗ B)k = Ak ⊗ Bk , (A B)k = (A : B)k−1 A B, k = 1, 2, . . . (5.39)
