96 3 Curves and Surfaces in Three-Dimensional Euclidean Space

where C denotes an integration constant. In order to avoid singularity of this solution

at t 2 = π/2 we set C = 0. Thus, by (3.155)

1 pR
f 11 = p R, f 11 =  2 . (3.158)
2 2 R0 + R cos t 2

By virtue of (3.153) and (3.98) it yields

p R 2R0 + R cos t 2 p 2R0 + R cos t 2

f 22 = f 21 g12 + f 22 g22 = , f 22 = .
2 R0 + R cos t 2 2R R0 + R cos t 2
(3.159)
Thus, in view of (3.149) the physical stress components in the wall of the toroidal
vessel take the from

R p R 2R0 + R cos t 2
σ (1) = p , σ (2) = , (3.160)
2h 2h R0 + R cos t 2

where h denotes the thickness of the vessel wall assumed to be small in comparison
to R.
Alternatively, the hoop stress σ (2) can be expressed on the basis of the equilibrium
condition (see also [15]) for a part of the vessel cut out by a vertical cylinder of radius
R0 and a conical surface specified in Fig. 3.7 by a dashed line. The force equilibrium
condition of this part can be written for the vertical direction by
 2  
πp R0 + R cos t 2 − R02 − σ (2) h2π R0 + R cos t 2 cos t 2 = 0, (3.161)

which immediately leads to (3.160)2 . Applying further the Laplace law (3.150) where
the principal curvatures are given by (3.103) we finally obtain (3.160)1 .

Exercises

3.1. Show that a curve r (s) is a straight line if κ (s) ≡ 0 for any s.

3.2. Show that the curves r (s) and r  (s) = r (−s) have the same curvature and
torsion.

3.3. Show that a curve r (s) characterized by zero torsion τ (s) ≡ 0 for any s lies in
a plane.

3.4. Evaluate the Christoffel symbols of the second kind, the coefficients of the first
and second fundamental forms, the Gaussian and mean curvatures for the cylinder
(3.56).
Exercises 97

3.5. Evaluate the Christoffel symbols of the second kind, the coefficients of the first
and second fundamental forms, the Gaussian and mean curvatures for the sphere
(3.58).

3.6. For the so-called hyperbolic paraboloidal surface defined by

  t 1t 2
r t 1 , t 2 = t 1 e1 + t 2 e2 + e3 , c > 0, (3.162)
c
evaluate the tangent vectors to the coordinate lines, the Christoffel symbols of the
second kind, the coefficients of the first and second fundamental forms, the Gaussian
and mean curvatures.

3.7. For a cone of revolution defined by

 
r t 1 , t 2 = ct 2 cos t 1 e1 + ct 2 sin t 1 e2 + t 2 e3 , c = 0, (3.163)

evaluate the vectors tangent to the coordinate lines, the Christoffel symbols of the
second kind, the coefficients of the first and second fundamental forms, the Gaussian
and mean curvatures.

3.8. An elliptic torus is obtained by revolution of an ellipse about a coplanar axis.

The rotation axis is also parallel to one of the ellipse axes the lengths of which are
denoted by 2a and 2b. The elliptic torus can thus be defined by
     
r t 1 , t 2 = R0 + a cos t 2 cos t 1 e1 + R0 + a cos t 2 sin t 1 e2 + b sin t 2 e3 .
(3.164)
Evaluate the vectors tangent to the coordinate lines, the Christoffel symbols of the
second kind, the coefficients of the first and second fundamental forms, the Gaussian
and mean curvatures.

3.9. Using the results of Exercise 3.8 calculate stresses in a thin wall vessel of the
elliptic torus form (3.164) subject to the internal pressure p.

3.10. Verify relation (3.114).

3.11. Prove the product rule of differentiation for the covariant derivative of the
vector f α (3.119)1 by using (3.76) and (3.77).

3.12. Derive relations (3.128) and (3.129) from (3.125) and (3.127) utilizing (3.78),
(3.80), (3.119), (3.120) and (2.105–2.107).

3.13. Write out equilibrium equations (3.143–3.144) of the membrane theory for a
cylindrical shell and a spherical shell.
Chapter 4
Eigenvalue Problem and Spectral
Decomposition of Second-Order Tensors

4.1 Complexification

So far we have considered solely real vectors and real vector spaces. For the purposes
of this chapter an introduction of complex vectors is, however, necessary. Indeed, in
the following we will see that the existence of a solution of an eigenvalue problem
even for real second-order tensors can be guaranteed only within a complex vector
space. In order to define the complex vector space let us consider ordered pairs x, y
of real vectors x and y ∈ En . The sum of two such pairs is defined by [18]
     
x 1 , y1 + x 2 , y2 = x 1 + x 2 , y1 + y2 . (4.1)

Further, we define the product of a pair x, y by a complex number α + iβ by

(α + iβ) x, y = αx − β y, βx + α y , (4.2)

√
where α, β ∈ R and i = −1. These formulas can easily be recovered assuming
that
x, y = x + i y. (4.3)

The definitions (4.1) and (4.2) enriched by the zero pair 0, 0 are sufficient to ensure
that the axioms (A.1–A.4) and (B.1–B.4) of Chap. 1 are valid. Thus, the set of all pairs
z = x, y characterized by the above properties
 formsa vector space referred to as
complex vector space. Every basis G = g 1 , g 2 , . . . , g n of the underlying Euclidean
space En represents simultaneously a basis of the corresponding complexified space.
Indeed, for every complex vector within this space

z = x + i y, (4.4)

where x, y ∈ En and consequently

© Springer Nature Switzerland AG 2019 99

M. Itskov, Tensor Algebra and Tensor Analysis for Engineers,
Mathematical Engineering, https://doi.org/10.1007/978-3-319-98806-1_4
100 4 Eigenvalue Problem and Spectral Decomposition of Second-Order Tensors

x = x i gi , y = y i gi , (4.5)

we can write  
z = x i + iy i g i . (4.6)

Thus, the dimension of the complexified space coincides with the dimension of the
original real vector space. Using this fact we will denote the complex vector space
based on En by Cn . Clearly, En represents a subspace of Cn .
For every vector z ∈ Cn given by (4.4) one defines a complex conjugate counter-
part by
z = x − i y. (4.7)

Of special interest is the scalar product of two complex vectors, say z 1 = x 1 + i y1

and z 2 = x 2 + i y2 , which we define by (see also [4])
     
x 1 + i y1 · x 2 + i y2 = x 1 · x 2 − y1 · y2 + i x 1 · y2 + y1 · x 2 . (4.8)

This scalar product is commutative (C.1), distributive (C.2) and linear in each factor
(C.3). Thus, it differs from the classical scalar product of complex vectors given in
terms of the complex conjugate (see, e.g., [18]). As a result, the axiom (C.4) does
not generally hold. For instance, one can easily imagine a non-zero complex vector
(for example e1 + ie2 ) whose scalar product with itself is zero. For complex vectors
with the scalar product (4.8) the notions of length, orthogonality or parallelity can
hardly be interpreted geometrically.
However, for complex vectors the axiom (C.4) can be reformulated by

z · z ≥ 0, z · z = 0 if and only if z = 0. (4.9)

Indeed, using (4.4), (4.7) and (4.8) we obtain z · z = x · x + y · y. Bearing in mind

that the vectors x and y belong to the Euclidean space this immediately implies (4.9).
As we learned in Chap. 1, the Euclidean space En is characterized by the existence
of an orthonormal basis (1.8). This can now be postulated for the complex vector
space Cn as well, because Cn includes En by the very definition. Also Theorem 1.6
remains valid since it has been proved without making use of the property (C.4).
Thus, we may state that for every basis in Cn there exists a unique dual basis.
The last step in the complexification of the vector space is a generalization of the
linear mapping to complex vectors. This can be achieved by setting for every tensor
A ∈ Linn
A (x + i y) = Ax + i (A y) . (4.10)

In a similar fashion, linear mappings can also be complexified. Complex second-

order tensors are important for example in the context of complex eigenprojections
to be discussed in this chapter. Let us consider pairs Z = X, Y of real second order
tensors X, Y ∈ Linn which map complex vectors as follows
4.1 Complexification 101

X, Y x, y = Xx − Y y, Yx + X y . (4.11)

Linearity of this mapping requires that

X1 , Y1  + X2 , Y2  = X1 + X2 , Y1 + Y2  , (4.12)

(α + iβ) X, Y = αX − βY, αY + βX . (4.13)

A complex conjugate counterpart of the tensor Z = X, Y is defined by

Z = X, −Y . (4.14)

Accordingly, a real second-order tensor is characterized by the property A = A =

A, 0.

4.2 Eigenvalue Problem, Eigenvalues and Eigenvectors

Let A ∈ Linn be a second-order tensor. The equation

Aa = λa, a = 0 (4.15)

is referred to as the eigenvalue problem of the tensor A. The non-zero vector a ∈ Cn

satisfying this equation is called an eigenvector of A; λ ∈ C is called an eigenvalue
of A. It is clear that any product of an eigenvector with any (real or complex) scalar
is again an eigenvector.
The eigenvalue problem (4.15) and the corresponding eigenvector a can be
regarded as the right eigenvalue problem and the right eigenvector, respectively.
In contrast, one can define the left eigenvalue problem by

bA = λb, b = 0, (4.16)

where b ∈ Cn is the left eigenvector. In view of (1.119), every right eigenvector of A

represents the left eigenvector of AT and vice versa. In the following, unless indicated
otherwise, we will mean the right eigenvalue problem and the right eigenvector.
Mapping (4.15) by A several times we obtain

Ak a = λk a, k = 1, 2, . . . (4.17)

This leads to the following (spectral mapping) theorem.


Theorem 4.1. Let λ be an eigenvalue
mof the tensor A and let g (A) = m k=0 ak A
k

be a polynomial of A. Then g (λ) = k=0 ak λ is the eigenvalue of g (A).

k