ResearchGate Logo

Discover the world's research

  • 20+ million members
  • 135+ million publications
  • 700k+ research projects

Join for free

Lecture Notes

Vector Analysis

MATH 332

Ivan Avramidi

New Mexico Institute of Mining and Technology

Socorro, NM 87801

May 19, 2004

Author: Ivan Avramidi; File: vecanal4.tex; Date: July 1, 2005; Time: 13:34

Contents

1 Linear Algebra 1

1.1 Vectors in Rn and Matrix Algebra . . . . . . . . . . . . . . . . . . . 1

1.1.1 Vectors ............................. 1

1.1.2 Matrices............................. 3

1.1.3 Determinant........................... 8

1.1.4 Exercises ............................ 9

1.2 VectorSpaces.............................. 11

1.2.1 Exercises ............................ 13

1.3 Inner Product and Norm . . . . . . . . . . . . . . . . . . . . . . . . 14

1.3.1 Exercises ............................ 15

1.4 LinearOperators ............................ 17

1.4.1 Exercises ............................ 24

2 Vector and Tensor Algebra 27

2.1 MetricTensor.............................. 27

2.2 Dual Space and Covectors . . . . . . . . . . . . . . . . . . . . . . . 29

2.2.1 Einstein Summation Convention . . . . . . . . . . . . . . . . 31

2.3 General Definition of a Tensor . . . . . . . . . . . . . . . . . . . . . 34

2.3.1 Orientation, Pseudotensors and Volume . . . . . . . . . . . . 37

2.4 Operators and Tensors . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.5 Vector Algebra in R 3 .......................... 44

3 Geometry 49

3.1 Geometry of Euclidean Space . . . . . . . . . . . . . . . . . . . . . 49

3.2 Basic Topology of Rn .......................... 53

3.3 Curvilinear Coordinate Systems . . . . . . . . . . . . . . . . . . . . 54

3.3.1 Change of Coordinates . . . . . . . . . . . . . . . . . . . . . 56

3.3.2 Examples............................ 57

3.4 Vector Functions of a Single Variable . . . . . . . . . . . . . . . . . 59

3.5 GeometryofCurves........................... 61

3.6 Geometry of Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . 65

I

II CONTENTS

4 Vector Analysis 69

4.1 Vector Functions of Several Variables . . . . . . . . . . . . . . . . . 69

4.2 Directional Derivative and the Gradient . . . . . . . . . . . . . . . . 71

4.3 ExteriorDerivative ........................... 73

4.4 Divergence ............................... 76

4.5 Curl................................... 77

4.6 Laplacian ................................ 78

4.7 Di erential Vector Identities . . . . . . . . . . . . . . . . . . . . . . 79

4.8 Orthogonal Curvilinear Coordinate Systems in R 3 ........... 80

5 Integration 83

5.1 LineIntegrals .............................. 83

5.2 SurfaceIntegrals ............................ 84

5.3 VolumeIntegrals ............................ 86

5.4 Fundamental Integral Theorems . . . . . . . . . . . . . . . . . . . . 87

5.4.1 Fundamental Theorem of Line Integrals . . . . . . . . . . . . 87

5.4.2 Green's Theorem . . . . . . . . . . . . . . . . . . . . . . . . 87

5.4.3 Stokes's Theorem . . . . . . . . . . . . . . . . . . . . . . . . 87

5.4.4 Gauss's Theorem . . . . . . . . . . . . . . . . . . . . . . . . 87

5.4.5 General Stokes's Theorem . . . . . . . . . . . . . . . . . . . 88

6 Potential Theory 89

6.1 Simply Connected Domains . . . . . . . . . . . . . . . . . . . . . . 90

6.2 Conservative Vector Fields . . . . . . . . . . . . . . . . . . . . . . . 91

6.2.1 Scalar Potential . . . . . . . . . . . . . . . . . . . . . . . . . 91

6.3 Irrotational Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . 92

6.4 Solenoidal Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . 93

6.4.1 Vector Potential . . . . . . . . . . . . . . . . . . . . . . . . . 93

6.5 LaplaceEquation ............................ 94

6.5.1 Harmonic Functions . . . . . . . . . . . . . . . . . . . . . . 94

6.6 PoissonEquation ............................ 95

6.6.1 Dirac Delta Function . . . . . . . . . . . . . . . . . . . . . . 95

6.6.2 PointSources.......................... 95

6.6.3 Dirichlet Problem . . . . . . . . . . . . . . . . . . . . . . . . 95

6.6.4 Neumann Problem . . . . . . . . . . . . . . . . . . . . . . . 95

6.6.5 Green's Functions . . . . . . . . . . . . . . . . . . . . . . . 95

6.7 Fundamental Theorem of Vector Analysis . . . . . . . . . . . . . . . 96

7 Basic Concepts of Dierential Geometry 97

7.1 Manifolds................................ 98

7.2 Di erentialForms............................ 99

7.2.1 Exterior Product . . . . . . . . . . . . . . . . . . . . . . . . 99

7.2.2 Exterior Derivative . . . . . . . . . . . . . . . . . . . . . . . 99

7.3 Integration of Di erentialForms.................... 100

7.4 General Stokes's Theorem . . . . . . . . . . . . . . . . . . . . . . . 101

7.5 Tensors in General Curvilinear Coordinate Systems . . . . . . . . . . 102

vecanal4.tex; July 1, 2005; 13:34; p. 1

CONTENTS III

7.5.1 Covariant Derivative . . . . . . . . . . . . . . . . . . . . . . 102

8 Applications 103

8.1 Mechanics................................ 104

8.1.1 InertiaTensor.......................... 104

8.1.2 Angular Momentum Tensor . . . . . . . . . . . . . . . . . . 104

8.2 Elasticity ................................ 105

8.2.1 StrainTensor.......................... 105

8.2.2 StressTensor.......................... 105

8.3 FluidDynamics............................. 106

8.3.1 Continuity Equation . . . . . . . . . . . . . . . . . . . . . . 106

8.3.2 Tensorof Momentum Flux Density . . . . . . . . . . . . . . 106

8.3.3 Euler's Equations . . . . . . . . . . . . . . . . . . . . . . . . 106

8.3.4 Rate of Deformation Tensor . . . . . . . . . . . . . . . . . . 106

8.3.5 Navier-Stokes Equations . . . . . . . . . . . . . . . . . . . . 106

8.4 Heat and Di usionEquations...................... 107

8.5 Electrodynamics ............................ 108

8.5.1 Tensorof Electromagnetic Field . . . . . . . . . . . . . . . . 108

8.5.2 Maxwell Equations . . . . . . . . . . . . . . . . . . . . . . . 108

8.5.3 Scalar and Vector Potentials . . . . . . . . . . . . . . . . . . 108

8.5.4 WaveEquations......................... 108

8.5.5 D'Alambert Operator . . . . . . . . . . . . . . . . . . . . . . 108

8.5.6 Energy-Momentum Tensor . . . . . . . . . . . . . . . . . . . 108

8.6 Basic Concepts of Special and General Relativity . . . . . . . . . . . 109

Bibliography 111

Notation 113

Index 113

vecanal4.tex; July 1, 2005; 13:34; p. 2

IV CONTENTS

vecanal4.tex; July 1, 2005; 13:34; p. 3

Chapter 1

Linear Algebra

1.1 Vectors in Rn and Matrix Algebra

1.1.1 Vectors

Rn is the set of all ordered n-tuples of real numbers, which can be assembled as

columns or as rows.

Let x 1 ,..., xn be n real numbers. Then the column-vector (or just vector) is an

ordered n -tuple of the form

v=

v1

v2

.

.

.

vn

,

and the row-vector (also called a covector) is an ordered n-tuple of the form

vT =( v1 ,v2 ,..., vn ) .

The real numbers x 1 , . . . xn are called the components of the vectors.

The operation that converts column-vectors into row-vectors and vice versa pre-

serving the order of the components is called the transposition and denoted by

T. That is

v1

v2

.

.

.

vn

T

=(v 1 , v 2 ,..., vn ) and (v1 , v2 ,..., vn ) T =

v1

v2

.

.

.

vn

.

Of course, for any vector v (vT )T = v.

1

2CHAPTER 1. LINEAR ALGEBRA

The addition of vectors is defined by

u+ v=

u1 + v 1

u2 + v 2

.

.

.

un + v n

,

and u+v =(u 1 +v 1 ,..., un +vn ).

Notice that one cannot add a column-vector and a row-vector!

The multiplication of vectors by a real constant, called a scalar, is defined by

av=

av1

av2

.

.

.

avn

,av=(av 1 ,..., avn ) .

The vectors that have only zero elements are called zero vectors, that is

0=

0

0

.

.

.

0

,0T =(0 ,..., 0) .

The set of column-vectors

e1 =

1

0

0

.

.

.

0

,e2 =

0

1

0

.

.

.

0

,··· ,en =

0

.

.

.

0

0

1

and the set of row-vectors

eT

1=(1, 0,..., 0), eT

2=(0,1,...,0), eT

n=(0, 0,..., 1)

are called the standard (or canonical) bases in Rn .

There is a natural product of column-vectors and row-vectors that assigns to a

row-vector and a column-vector a real number

huT ,v i= (u 1 , u 2 ,...,un )

v1

v2

.

.

.

vn

=

n

X

i= 1

uivi =u1 v1 + u2 v2 + · ·· + un vn .

This is the simplest instance of a more general multiplication rule for matrices

which can be summarized by saying that one multiplies row by column.

vecanal4.tex; July 1, 2005; 13:34; p. 4

1.1. VECTORS IN RN AND MATRIX ALGEBRA 3

The product of two column-vectors and the product of two row-vectors, called

the inner product (or the scalar product), is defined then by

(u, v )= ( uT , vT )= h uT ,vi =

n

X

i=1

uivi =u1 v1 + · ·· + un vn .

Finally, we define the norm (or the length) of both column-vectors and row-

vectors is defined by

||v || = ||vT || = p h vT ,v i=

n

X

i=1

v2

i

1/2

=q v2

1+· ·· +v 2

n.

1.1.2 Matrices

A set of n 2real numbers Ai j ,i, j= 1 ,...,n , arranged in an array that has n

columns and n rows

A=

A11 A12 ··· A1 n

A21 A22 ··· A2 n

.

.

..

.

.. . . .

.

.

An1 An2 ··· Ann

is called a square n× n real matrix .

The set of all real square n ×nmatrices is denoted by Mat( n,R).

The number Ai j (also called an entry of the matrix) appears in the i-th row and

the j -th column of the matrix A

A=

A11 A12 ··· A1j ·· · A1 n

A21 A22 ··· A2j ·· · A2 n

.

.

..

.

.. . . .

.

..

.

..

.

.

Ai1 Ai2 ··· Ai j ·· · Ain

.

.

..

.

..

.

..

.

.. . . .

.

.

An1 An2 ··· An j ··· Ann

Remark. Notice that the first index indicates the row and the second index

indicates the column of the matrix.

The matrix whose all entries are equal to zero is called the zero matrix.

The addition of matrices is defined by

A+ B=

A11 + B11 A12 +B12 ··· A1n +B1n

A21 + B21 A22 +B22 ··· A2n +B2n

.

.

..

.

... . .

.

.

An1 +Bn1 An2 +Bn2 ·· · Ann +Bnn

vecanal4.tex; July 1, 2005; 13:34; p. 5

4CHAPTER 1. LINEAR ALGEBRA

and the multiplication by scalars by

cA =

cA11 cA12 · · · cA1 n

cA21 cA22 · · · cA2 n

.

.

..

.

.. . . .

.

.

cAn1 cAn2 · ·· cAnn

The numbers Aii are called the diagonal entries. Of course, there are ndiagonal

entries. The set of diagonal entries is called the diagonal of the matrix A.

The numbers Ai j with i, j are called o -diagonal entries ; there are n (n 1)

o -diagonal entries.

The numbers Ai j with i< j are called the upper triangular entries . The set of

upper triangular entries is called the upper triangular part of the matrix A.

The numbers Ai j with i> j are called the lower triangular entries. The set of

lower triangular entries is called the lower triangular part of the matrix A.

The number of upper-triangular entries and the lower-triangular entries is the

same and is equal to n (n 1)/2.

A matrix whose only non-zero entries are on the diagonal is called a diagonal

matrix. For a diagonal matrix

Ai j =0 if i, j.

The diagonal matrix

A=

λ1 0··· 0

0λ 2 ··· 0

.

.

..

.

.. . . .

.

.

0 0 · ·· λn

is also denoted by A= diag(λ 1 , λ 2 , . . . , λ n )

A diagonal matrix whose all diagonal entries are equal to 1

I=

1 0 ··· 0

0 1 ··· 0

.

.

..

.

.. . . .

.

.

0 0 ··· 1

is called the identity matrix. The elements of the identity matrix are

Ii j =

1, if i= j

0, if i, j .

vecanal4.tex; July 1, 2005; 13:34; p. 6

1.1. VECTORS IN RN AND MATRIX ALGEBRA 5

A matrix Aof the form

A=

∗ ∗ ···

0∗ ···

.

.

..

.

.. . . .

.

.

0 0 ··· ∗

where represents nonzero entries is called an upper triangular matrix . Its

lower triangular part is zero, that is,

Ai j =0 if i< j.

A matrix Aof the form

A=

0 ··· 0

∗ ∗ ··· 0

.

.

..

.

.. . . .

.

.

∗ ∗ ···

whose upper triangular part is zero, that is,

Ai j = 0 if i> j ,

is called a lower triangular matrix.

The transpose of a matrix A whose ij -th entry is Aij is the matrix AT whose

ij-th entry is A ji . That is, AT obtained from A by switching the roles of rows and

columns of A:

AT =

A11 A21 ··· Aj1 ·· · An1

A12 A22 ··· Aj2 ·· · An2

.

.

..

.

.. . . .

.

..

.

..

.

.

A1i A2i ··· Aji · ·· Ani

.

.

..

.

..

.

..

.

.. . . .

.

.

A11 A2n · ·· A jn · ·· Ann

or ( AT )i j =Aji .

A matrix A is called symmetric if

AT =A

and anti-symmetric if AT =A.

The number of independent entries of an anti-symmetric matrix is n (n 1)/2.

The number of independent entries of a symmetric matrix is n (n+ 1)/2.

vecanal4.tex; July 1, 2005; 13:34; p. 7

6CHAPTER 1. LINEAR ALGEBRA

Every matrix Acan be uniquely decomposed as the sum of its diagonal part AD ,

the lower triangular part AL and the upper triangular part AU

A= AD + AL + AU .

For an anti-symmetric matrix

AT

U=A L and A D =0.

For a symmetric matrix AT

U=A L .

Every matrix Acan be uniquely decomposed as the sum of its symmetric part AS

and its anti-symmetric part AA

A= AS + AA ,

where

AS = 1

2(A+ AT ), AA = 1

2(A AT ) .

The product of matrices is defined as follows. The i j -th entry of the product

C= AB of two matrices A and B is

Ci j =

n

X

k=1

Aik Bk j =Ai1 B1j +Ai2 B2j + ·· · + Ain Bn j .

This is again a multiplication of the "i-th row of the matrix Aby the j-th column

of the matrix B".

Theorem 1.1.1 The product of matrices is associative, that is, for any matrices

A, B, C ( AB)C= A(BC) .

Theorem 1.1.2 For any two matrices A and B

(AB )T =BTAT .

A matrix A is called invertible if there is another matrix A 1 such that

AA1 = A 1 A= I .

The matrix A1 is called the inverse of A.

Theorem 1.1.3 For any two invertible matrices A and B

(AB)1 = B1 A1 ,

and ( A 1) T =( A T ) 1 .

vecanal4.tex; July 1, 2005; 13:34; p. 8

1.1. VECTORS IN RN AND MATRIX ALGEBRA 7

A matrix A is called orthogonal if

AT A= AAT = I ,

which means AT = A1 .

The trace is a map tr : Mat(n, R ) that assigns to each matrix A= ( Ai j ) a real

number tr A equal to the sum of the diagonal elements of a matrix

tr A =

n

X

k=1

Akk .

Theorem 1.1.4 The trace has the properties

tr(AB )= tr (BA ) ,

and tr A T = tr A .

Obviously, the trace of an anti-symmetric matrix is equal to zero.

Finally, we define the multiplication of column-vectors by matrices from the left

and the multiplication of row-vectors by matrices from the right as follows.

Each matrix defines a natural left action on a column-vector and a right action

on a row-vector.

For each column-vector vand a matrix A= (Ai j ) the column-vector u= Av is

given by

u1

u2

.

.

.

ui

.

.

.

un

=

A11 A12 ··· A1 n

A21 A22 ··· A2 n

.

.

..

.

.. . . .

.

.

Ai1 Ai2 ··· Ain

.

.

..

.

..

.

..

.

.

An1 An2 ··· Ann

v1

v2

.

.

.

vi

.

.

.

vn

=

A11 v1 +A12 v2 + · · · + A1n vn

A21 v1 +A22 v2 + · · · + A2n vn

.

.

.

Ai1 v1 +Ai2 v2 + · ·· + Ainvn

.

.

.

An1 v1 +An2 v2 + · ·· + Annvn

The components of the vector uare

ui =

n

X

j= 1

Ai j vj = Ai1 v1 + Ai2 v2 + · · · + Ainvn .

Similarly, for a row vector vT the components of the row-vector uT =vT A are

defined by

ui =

n

X

j= 1

vj Aji = v1A1i +v2A2i + · · · + vn A ni .

vecanal4.tex; July 1, 2005; 13:34; p. 9

8CHAPTER 1. LINEAR ALGEBRA

1.1.3 Determinant

Consider the set Zn = { 1, 2 ,...,n }of the first n integers. A permutation ϕ of the

set { 1, 2,..., n} is an ordered n-tuple (ϕ (1), . . . , ϕ ( n )) of these numbers.

That is, a permutation is a bijective (one-to-one and onto) function

ϕ:Zn Zn

that assigns to each number ifrom the set Zn ={ 1,...,n} another number ϕ (i)

from this set.

An elementary permutation is a permutation that exchanges the order of only

two numbers.

Every permutation can be realized as a product (or a composition) of elemen-

tary permutations. A permutation that can be realized by an even number of

elementary permutations is called an even permutation. A permutation that

can be realized by an odd number of elementary permutations is called an odd

permutation.

Proposition 1.1.1 The parity of a permutation does not depend on the repre-

sentation of a permutation by a product of the elementary ones.

That is, each representation of an even permutation has even number of elemen-

tary permutations, and similarly for odd permutations.

The sign of a permutation ϕ , denoted by sign(ϕ) (or simply ( 1)ϕ ), is defined

by

sign(ϕ )= ( 1)ϕ =( + 1, if ϕ is even,

1, if ϕ is odd

The set of all permutations of nnumbers is denoted by Sn .

Theorem 1.1.5 The cardinality of this set, that is, the number of dierent per-

mutations, is

|Sn |= n! .

The determinant is a map det : Mat(n, R ) Rthat assigns to each matrix

A=( Aij ) a real number det A defined by

det A= X

ϕSn

sign(ϕ ) A 1ϕ(1) ·· · Anϕ( n) ,

where the summation goes over all n! permutations.

The most important properties of the determinant are listed below:

Theorem 1.1.6 1. The determinant of the product of matrices is equal to the

product of the determinants:

det(AB )= det A det B .

vecanal4.tex; July 1, 2005; 13:34; p. 10

1.1. VECTORS IN RN AND MATRIX ALGEBRA 9

2. The determinants of a matrix A and of its transpose ATare equal:

det A= det AT .

3. The determinant of the inverse A1 of an invertible matrix A is equal to the

inverse of the determinant of A:

det A 1 = (det A ) 1

4. A matrix is invertible if and only if its determinant is non-zero.

The set of real invertible matrices (with non-zero determinant) is denoted by

GL(n, R). The set of matrices with positive determinant is denoted by GL+ (n, R).

A matrix with unit determinant is called unimodular.

The set of real matrices with unit determinant is denoted by S L (n, R).

The set of real orthogonal matrices is denoted by O ( n).

Theorem 1.1.7 The determinant of an orthogonal matrix is equal to either 1or

1.

An orthogonal matrix with unit determinant (a unimodular orthogonal matrix) is

called a proper orthogonal matrix or just a rotation .

The set of real orthogonal matrices with unit determinant is denoted by SO ( n).

A set G of invertible matrices forms a group if it is closed under taking inverse

and matrix multiplication, that is, if the inverse A1 of any matrix A in Gbelongs

to the set Gand the product AB of any two matrices A and B in G belongs to G.

1.1.4 Exercises

1. Show that the product of invertible matrices is an invertible matrix.

2. Show that the product of matrices with positive determinant is a matrix with positive

determinant.

3. Show that the inverse of a matrix with positive determinant is a matrix with positive

determinant.

4. Show that GL ( n, R ) forms a group (called the general linear group).

5. Show that GL+ ( n, R ) is a group (called the proper general linear group).

6. Show that the inverse of a matrix with negative determinant is a matrix with negative

determinant.

7. Show that: a) the product of an even number of matrices with negative determinant is a

matrix with positive determinant, b) the product of odd matrices with negative determinant

is a matrix with negative determinant.

8. Show that the product of matrices with unit determinant is a matrix with unit determinant.

9. Show that the inverse of a matrix with unit determinant is a matrix with unit determinant.

vecanal4.tex; July 1, 2005; 13:34; p. 11

10 CHAPTER 1. LINEAR ALGEBRA

10. Show that S L ( n, R ) forms a group (called the special linear group or the unimodular

group).

11. Show that the product of orthogonal matrices is an orthogonal matrix.

12. Show that the inverse of an orthogonal matrix is an orthogonal matrix.

13. Show that O ( n ) forms a group (called the orthogonal group).

14. Show that orthogonal matrices have determinant equal to either +1 or 1.

15. Show that the product of orthogonal matrices with unit determinant isan orthogonal ma-

trix with unit determinant.

16. Show that the inverse of an orthogonal matrix with unit determinant is an orthogonal

matrix with unit determinant.

17. Show that SO ( n) forms a group (called the proper orthogonal group or the rotation

group).

vecanal4.tex; July 1, 2005; 13:34; p. 12

1.2. VECTOR SPACES 11

1.2 Vector Spaces

Areal vector space consists of a set E, whose elements are called vectors , and

the set of real numbers R, whose elements are called scalars. There are two

operations on a vector space:

1. Vector addition,+ :E× E E , that assigns to two vectors u, v E

another vector u+ v , and

2. Multiplication by scalars ,· :R×E E , that assigns to a vector v E

and a scalar aR a new vector av E .

The vector addition is an associative commutative operation with an additive

identity. It satisfies the following conditions:

1. u+ v= v+ u , u,v, E

2. (u+ v )+w =u + (v+ w ), u,v, w E

3. There is a vector 0E , called the zero vector, such that for any v E

there holds v+ 0= v .

4. For any vector vE , there is a vector (v )E , called the opposite of v,

such that v+ (v )=0 .

The multiplication by scalars satisfies the following conditions:

1. a ( bv )= (ab)v , v E , a,bR ,

2. (a+ b )v= av+ bv , v E , a,bR ,

3. a (u+ v )= au+ av , u, v E , a R ,

4. 1 v= v v E .

The zero vector is unique.

For any u, v Ethere is a unique vector denoted by w= v u, called the

di erence of v and u , such that u+ w= v .

For any v E,0v= 0, and ( 1)v= v .

Let E be a real vector space and A= {e1 ,..., ek }be a finite collection of vectors

from E . A linear combination of these vectors is a vector

a1 e1 + · ·· + ak e k ,

where { a 1 ,..., an } are scalars.

A finite collection of vectors A= {e1 ,...,ek }is linearly independent if

a1 e1 + · ·· + ak e k = 0

implies a 1 = ··· =ak = 0.

vecanal4.tex; July 1, 2005; 13:34; p. 13

12 CHAPTER 1. LINEAR ALGEBRA

A collection Aof vectors is linearly dependent if it is not linearly independent.

Two non-zero vectors u and v which are linearly dependent are also called par-

allel, denoted by u||v.

A collection Aof vectors is linearly independent if no vector of Ais a linear

combination of a finite number of vectors from A.

Let Abe a subset of a vector space E . The span of A, denoted by span A , is the

subset of Econsisting of all finite linear combinations of vectors from A, i.e.

span A= {v E |v= a 1 e 1 + ·· · + ak e k , e i ∈ A, ai R} .

We say that the subset span A is spanned by A .

Theorem 1.2.1 The span of any subset of a vector space is a vector space.

Avector subspace of a vector space E is a subset S Eof Ewhich is itself a

vector space.

Theorem 1.2.2 A subset S of E is a vector subspace of E if and only if span S =

S.

Span of Ais the smallest subspace of E containing A.

A collection Bof vectors of a vector space Eis a basis of E if Bis linearly

independent and span B=E .

A vector space E is finite-dimensional if it has a finite basis.

Theorem 1.2.3 If the vector space E is finite-dimensional, then the number of

vectors in any basis is the same.

The dimension of a finite-dimensional real vector space E, denoted by dim E , is

the number of vectors in a basis.

Theorem 1.2.4 If {e1 ,...,en }is a basis in E, then for every vector v E there

is a unique set of real numbers ( vi )= ( v1 ,...,vn ) such that

v=

n

X

i=1

vi e i = v1 e1 +· ·· + vn e n .

The real numbers vi ,i= 1 ,...,n , are called the components of the vector v

with respect to the basis { ei } .

It is customary to denote the components of vectors by superscripts , which

should not be confused with powers of real numbers

v2 ,( v)2 = vv, . . . , vn , ( v ) n .

vecanal4.tex; July 1, 2005; 13:34; p. 14

1.2. VECTOR SPACES 13

Examples of Vector Subspaces

Zero subspace {0}.

Line with a tangent vector u:

S1 =span {u } ={v E|v= tu, tR }.

Plane spanned by two nonparallel vectors u1 and u 2

S2 =span {u 1 , u 2 } ={v E|v= tu1 + su2 , t, sR }.

More generally, a k - plane spanned by a linearly independent collection of k vec-

tors { u 1 ,..., uk }

Sk =span { u 1 ,...,u k }= {v E|v= t1 u1 + ·· · + tk u k ,t1 ,...,tk R }.

An (n 1)-plane in an n-dimensional vector space is called a hyperplane.

1.2.1 Exercises

1. Show that if λv= 0, then either v= 0 or λ= 0.

2. Prove that the span of a collection of vectors is a vector subspace.

vecanal4.tex; July 1, 2005; 13:34; p. 15

14 CHAPTER 1. LINEAR ALGEBRA

1.3 Inner Product and Norm

A real vector space Eis called an inner product space if there is a function

(·, · ) : E× E R , called the inner product, that assigns to every two vectors u

and v a real number (u,v) and satisfies the conditions: u,v, w E, a R :

1. (v, v ) 0

2. (v, v )= 0 if and only if v= 0

3. (u, v )= ( v, u )

4. (u+ v,w )= ( u,w )+ (v, w )

5. (a u,v )= (u, av )= a ( u,v)

A finite-dimensional inner product space is called a Euclidean space.

The inner product is often called the dot product, or the scalar product, and is

denoted by ( u,v)= u· v .

All spaces considered below are Euclidean spaces. Henceforth, Ewill denote an

n-dimensional Euclidean space if not specified otherwise.

The Euclidean norm is a function || · || :E R that assigns to every vector

vEa real number || v|| defined by

||v || =p ( v, v ).

The norm of a vector is also called the length.

A vector with unit norm is called a unit vector.

Theorem 1.3.1 For any u, v E there holds

||u+ v ||2 =||u||2 + 2(u, v )+ ||v||2 .

Theorem 1.3.2 Cauchy-Schwarz's Inequality. For any u, v E there holds

|(u, v ) | ≤ ||u || ||v || .

The equality

|( u, v ) |= ||u || ||v||

holds if and only if u and v are parallel.

Corollary 1.3.1 Triangle Inequality. For any u, v E there holds

||u+ v || ≤ ||u || + ||v|| .

vecanal4.tex; July 1, 2005; 13:34; p. 16

1.3. INNER PRODUCT AND NORM 15

The angle between two non-zero vectors u and v is defined by

cos θ= (u, v)

||u || ||v || ,0θπ .

Then the inner product can be written in the form

(u,v )= ||u || || v|| cos θ .

Two non-zero vectors u, vE are orthogonal , denoted by u v , if

(u,v )= 0.

A basis {e1 ,..., en }is called orthonormal if each vector of the basis is a unit

vector and any two distinct vectors are orthogonal to each other, that is,

(ei , ej )=( 1,if i= j

0, if i, j .

Theorem 1.3.3 Every Euclidean space has an orthonormal basis.

Let S Ebe a nonempty subset of E. We say that x Eis orthogonal to S,

denoted by xS , if xis orthogonal to every vector of S.

The set S = {xE | xS }

of all vectors orthogonal to Sis called the orthogonal complement of S.

Theorem 1.3.4 The orthogonal complement of any subset of a Euclidean space

is a vector subspace.

Two subsets A and B of E are orthogonal , denoted by A B , if every vector of

Ais orthogonal to every vector of B.

Let S be a subspace of E and S be its orthogonal complement. If every element

of E can be uniquely represented as the sum of an element of S and an element

of S , then E is the direct sum of S and S , which is denoted by

E= S S .

The union of a basis of Sand a basis of S gives a basis of E.

1.3.1 Exercises

1. Show that the Euclidean norm has the following properties

(a) ||v || ≥ 0, v E ;

(b) ||v || = 0 if and only if v= 0;

vecanal4.tex; July 1, 2005; 13:34; p. 17

16 CHAPTER 1. LINEAR ALGEBRA

(c) ||a v || = | a |||v|| , v E, a R .

2. Parallelogram Law. Show that for any u, v E

||u+ v ||2 +||u v||2 =2 ||u||2 +||v||2

3. Show that any orthogonal system in Eis linearly independent.

4. Gram-Schmidt orthonormalization process. Let G= { u 1 , ··· ,uk } be a linearly inde-

pendent collection of vectors. Let O= { v 1 , · ·· , vk } be a new collection of vectors defined

recursively by

v1 =u1 ,

vj = uj

j1

X

i=1

vi ( vi , uj )

||vi ||2 ,2j k,

and the collection B= {e 1 ,...,ek } be defined by

ei = vi

||vi || .

Show that: a) Ois an orthogonal system and b) Bis an orthonormal system.

5. Pythagorean Theorem. Show that if u v , then

||u+ v ||2 =||u||2 +||v||2 .

6. Let B= {e 1 , ··· en } be an orthonormal basis in E. Show that for any vector v E

v=

n

X

i=1e i (e i ,v)

and

||v||2 =

n

X

i=1

(ei , v )2 .

7. Prove that the orthogonal complement of a subset S of E is a vector subspace of E.

8. Let S be a subspace in E. Prove that

a) E ={0} , b) { 0} =E , c) (S ) =S .

9. Show that the intersection of orthogonal subsets of a Euclidean space is either empty or

consists of only the zero vector. That is, for two subsets A and B , if A B , then A B={ 0}

or .

vecanal4.tex; July 1, 2005; 13:34; p. 18

1.4. LINEAR OPERATORS 17

1.4 Linear Operators

Alinear operator on a vector space E is a mapping L :E Esatisfying the

condition u, v E , a R ,

L(u+ v )= L(u )+ L(v ) and L(av )=a L(v).

Identity operator I on E is defined by

Iv= v, vE

Null operator 0 :E Eis defined by

0v= 0, vE

The vector u=L ( v ) is the image of the vector v.

If S is a subset of E, then the set

L(S )={u E |u = L(v ) for some vS }

is the image of the set S and the set

L1 (S )={v E | L(v )S }

is the inverse image of the set A .

The image of the whole space Eof a linear operator Lis the range (or the image)

of L , denoted by

Im(L )=L (E )= {u E |u=L (v ) for some vE } .

The kernel Ker(L ) (or the null space) of an operator Lis the set of all vectors in

Ewhich are mapped to zero, that is

Ker(L )=L 1 ({ 0} )={v E | L ( v )= 0} .

Theorem 1.4.1 For any operator Lthe sets Im(L ) and Ker (L ) are vector sub-

spaces.

The dimension of the kernel Ker(L) of an operator L

null(L )= dimKer (L)

is called the nullity of the operator L.

The dimension of the range Im(L) of an operator L

rank(L )= dimKer (L)

is called the rank of the operator L.

vecanal4.tex; July 1, 2005; 13:34; p. 19

18 CHAPTER 1. LINEAR ALGEBRA

Theorem 1.4.2 For any operator Lon an n-dimensional Euclidean space E

rank(L )+ null(L )= n

The set L(E ) of all linear operators on a vector space Eis a vector space with

the addition of operators and multiplication by scalars defined by

(L 1 +L 2 )(x )=L 1 (x )+L 2 (x), and (aL )(x )= aL (x ).

The product of the operators A and B is the composition of A and B.

Since the product of operators is defined as a composition of linear mappings,

it is automatically associative, which means that for any operators A ,B and C,

there holds ( AB )C= A (BC ) .

The integer powers of an operator are defined as the multiple composition of the

operator with itself, i.e.

A0 = I A1 = A,A2 =AA, . . .

The operator A on E is invertible if there exists an operator A 1 on E , called the

inverse of A , such that A 1A = AA 1 =I .

Theorem 1.4.3 Let A and B be invertible operators. Then:

(A1 )1 =A , (AB)1 = B1 A 1 .

The operators A and B are commuting if

AB =BA

and anti-commuting if AB = BA .

The operators A and B are said to be orthogonal to each other if

AB =BA = 0 .

An operator A is involutive if A 2 = I

idempotent if A 2 =A ,

and nilpotent if for some integer k

Ak = 0 .

vecanal4.tex; July 1, 2005; 13:34; p. 20

1.4. LINEAR OPERATORS 19

Selfadjoint Operators

The adjoint A of an operator Ais defined by

(A u, v )= (u, A v ), u, vE .

Theorem 1.4.4 For any two operators A and B

(A ) =A , (AB ) = BA .

An operator A is self-adjoint if

A =A

and anti-selfadjoint if

A = A

Every operator A can be decomposed as the sum

A= AS + AA

of its selfadjoint part AS and its anti-selfadjoint part A A

AS =1

2(A+ A ),AA = 1

2(A A ).

An operator A is called unitary if

AA =A A= I .

An operator A on E is called positive , denoted by A 0, if it is selfdadjoint and

v E (A v,v) 0.

Projection Operators

Let S be a subspace of E and E= S S . Then for any u Ethere exist unique

vSand wS such that u=v +w.

The vector v is called the projection of u onto S.

The operator P on E defined by

Pu= v

is called the projection operator onto S.

vecanal4.tex; July 1, 2005; 13:34; p. 21

20 CHAPTER 1. LINEAR ALGEBRA

The operator P defined by

P u= w

is the projection operator onto S .

The operators P and P are called complementary projections. They have the

properties:

P = P,( P ) = P ,

P+ P = I,

P2 = P,( P )2 = P ,

PP =P P= 0 .

Theorem 1.4.5 An operator P is a projection if and only if P is idempotent and

self-adjoint.

More generally, a collection of projections {P1 ,...,Pk }is a complete orthogo-

nal system of complimentary projections if

PiPk = 0if i, k

and k

X

i=1P i =P 1 +· ·· +P k =I.

A complete orthogonal system of projections defines the orthogonal decomposi-

tion of the vector space

E= E1 ⊕ · ·· ⊕ Ek ,

where Ei is the subspace the projection Pi projects onto.

Theorem 1.4.6 1. The dimension of the subspaces Eiare equal to the ranks

of the projections P i

dim Ei = rank Pi .

2. The sum of dimensions of the vector subspaces Eiequals the dimension of

the vector space E

n

X

i=1

dim Ei = dim E 1 + · ·· + dim Ek = dim E .

vecanal4.tex; July 1, 2005; 13:34; p. 22

1.4. LINEAR OPERATORS 21

Spectral Decomposition Theorem

A real number λis called an eigenvalue of an operator Aif there is a unit vector

uEsuch that A u=λu.

The vector u is called the eigenvector corresponding to the eigenvalue λ.

The span of all eigenvectors corresponding to the eigenvalue λ of an operator A

is called the eigenspace of λ.

The dimension of the eigenspace of the eigenvalue λis called the multiplicity

(also called the geometric multiplicity ) of λ.

An eigenvalue of multiplicity 1 is called simple (or non-degenerate).

An eigenvalue of multiplicity greater than 1 is called multiple (or degenerate).

The set of all eigenvalues of an operator is called the spectrum of the operator.

Theorem 1.4.7 Let A be a selfadjoint operator. Then:

1. The number of eigenvalues counted with multiplicity is equal to the dimen-

sion n = dim E of the vector space E.

2. The eigenvectors corresponding to distinct eigenvalues are orthogonal to

each other.

Theorem 1.4.8 Spectral Decomposition of Self-Adjoint Operators. Let A

be a selfadjoint operator on E. Then there exists an orthonormal basis B =

{e1 ,...,en } in E consisting of eigenvectors of Acorresponding to the eigenvalues

{λ1 ,...λ n}, and the corresponding system of orthogonal complimentary projec-

tions { P1 ,..., P n } onto the one-dimensional eigenspaces Ei,

A=

n

X

i= 1

λi Pi .

The projections {P i } are defined by

Pi v= ei (ei , v) .

and satisfy the equations

n

X

i=1P i =I,and P i P j =0 if i ,j.

In other words, for any

v=

n

X

i=1e i (e i ,v),

we have

Av=

n

X

i=1

λi ei (ei ,v) .

vecanal4.tex; July 1, 2005; 13:34; p. 23

22 CHAPTER 1. LINEAR ALGEBRA

Let f :R Rbe a real-valued function on R. Let A be a selfadjoint operator on

a Euclidean space Egiven by its spectral decomposition

A=

n

X

i=1

λi Pi ,

where Pi are the one-dimensional projections. Then one can define a function

of the self-adjoint operator f (A ) on Eby

f(A ) =

n

X

i=1

f(λ i )P i .

The exponential of an operator A is defined by

exp A =

X

k=1

1

k!A k =

n

X

i=1

eλ i P i

Theorem 1.4.9 Let U be a unitary operator on a real vector space E. Then there

exists an ani-selfadjoint operator Asuch that

U=exp A.

Recall that the operators U and A satisfy the equations

U =U1 and A = A.

Let A be a self-adjoint operator with the eigenvalues {λ1 , . . . , λn }. Then the trace

of the operator and the determinant of the operator A are defined by

tr A =

n

X

i= 1

λi , det A = λ1 ·· · λn .

Note that tr I=n, det I = 1 .

The trace of a projection Ponto a vector subspace Sis equal to its rank, or the

dimension of the vector subspace S,

tr P =rank P =dim S .

The trace of a function of a self-adjoint operator Ais then

tr f ( A ) =

n

X

i=1

f(λ i ).

If there are multiple eigenvalues, then each eigenvalue should be counted with

its multiplicity.

vecanal4.tex; July 1, 2005; 13:34; p. 24

1.4. LINEAR OPERATORS 23

Theorem 1.4.10 Let A be a self-adjoint operator. Then

detexp A =e tr A .

Let A be a positive definite operator, A > 0. The zeta-function of the operator

Ais defined by

ζ(s )= tr As =

n

X

i= 1

1

λs

i

.

Theorem 1.4.11 The zeta-functions has the properties

ζ(0) =n ,

and

ζ0 (0) = logdet A .

Examples

Let u be a unit vector and Pu be the projection onto the one-dimensional subspace

(line) Su spanned by u defined by

Pu v= u(u,v) .

The orthogonal complement S

uis the hyperplane with the normal u. The oper-

ator Ju defined by J u =I 2P u

is called the reflection operator with respect to the hyperplane S

u. The reflec-

tion operator is a self-adjoint involution, that is, it has the following properties

J

u=J u ,J 2

u=I.

The reflection operator has the eigenvalue 1 with multiplicity 1 and the eigenspace

Su , and the eigenvalue +1 with multiplicity ( n1) and with eigenspace S

u.

Let u 1 and u 2 be an orthonormal system of two vectors and Pu 1 ,u 2be the projec-

tion operator onto the two-dimensional space (plane) Su 1 ,u 2 spanned by u 1and

u2 P u 1 ,u2 v= u 1 ( u 1 , v)+ u2 (u2 ,v) .

Let Nu 1 ,u 2 be an operator defined by

Nu 1 ,u2 v=u 1( u 2 ,v) u2 (u1 ,v) .

Then N u 1 ,u 2 P u 1 ,u 2 =P u 1 ,u 2 N u 1 ,u 2 =N u 1 ,u 2

and N 2

u1 , u2 =P u 1 , u 2 .

vecanal4.tex; July 1, 2005; 13:34; p. 25

24 CHAPTER 1. LINEAR ALGEBRA

Arotation operator R u 1 ,u 2 (θ ) with the angle θin the plane Su 1 ,u 2is defined by

Ru 1 ,u2 (θ )= I Pu 1 ,u2 +cos θ Pu 1 ,u2 +sin θ Nu 1 ,u2 .

The rotation operator is unitary, that is, it satisfies the equation

R

u1 ,u2 R u 1 , u 2=I.

Theorem 1.4.12 Spectral Decomposition of Unitary Operators on Real Vec-

tor Spaces. Let U be a unitary operator on a real vector space E. Then the only

eigenvalues of U are +1 and 1 (possibly multiple) and there exists an orthogo-

nal decomposition E=E + E V 1 ⊕ ·· · ⊕ V k ,

where E+and E are the eigenspaces corresponding to the eigenvalues 1and

1 , and {V1 ,..., vk } are two-dimensional subspaces such that

dim E= dim E+ + dim E + 2k .

Let P+ ,P ,P1 ,...,P k be the corresponding orthogonal complimentary system

of projections, that is,

P+ +P +

k

X

i=1P i =I.

Then there exists a corresponding system of operators N1 ,...,N ksatisfying the

equations N 2

i=P i ,N i P i =P i N i =N i ,

Ni Pj =Pj Ni = 0,if i ,j

and the angles θ1 ,...θ ksuch that

U= P+ P +

k

X

i= 1

(cos θi Pi + sin θi Ni ) .

1.4.1 Exercises

1. Prove that the range and the kernel of any operator are vector spaces.

2. Show that

(aA +bB ) =aA +bB a, bR ,

(A ) =A

(AB) =B A

3. Show that for any operator Athe operators AA and A +A are selfadjoint.

4. Show that the product of two selfadjoint operators is selfadjoint if and only if they com-

mute.

5. Show that a polynomial p (A ) of a selfadjoint operator Ais a selfadjoint operator.

vecanal4.tex; July 1, 2005; 13:34; p. 26

1.4. LINEAR OPERATORS 25

6. Prove that the inverse of an invertible operator is unique.

7. Prove that an operator Ais invertible if and only if Ker A ={ 0} , that is, A v= 0 implies

v= 0.

8. Prove that for an invertible operator A , Im(A )=E , that is, for any vector vE there is a

vector uE such that v=A u .

9. Show that if an operator Ais invertible, then

(A1 )1 =A .

10. Show that the product AB of two invertible operators A and B is invertible and

(AB)1 =B1 A1

11. Prove that the adjoint A of any invertible operator Ais invertible and

(A )1 =(A1 ) .

12. Prove that the inverse A 1 of a selfadjoint invertible operator is selfadjoint.

13. An operator A on E is called isometric if v E ,

||A v || = ||v|| .

Prove that an operator is unitary if and only if it is isometric.

14. Prove that unitary operators preserves inner product. That is, show that if Ais a unitary

operator, then u, v E (A u,A v)= ( u,v) .

15. Show that for every unitary operator A both A 1 and A are unitary.

16. Show that for any operator Athe operators AA and A A are positive.

17. What subspaces do the null operator 0and the identity operator Iproject onto?

18. Show that for any two projection operators P and Q , PQ =0 if and only if QP = 0.

19. Prove the following properties of orthogonal projections

P =P,( P ) = P , P + P= I,PP = P P= 0.

20. Prove that an operator is projection if and only if it is idempotent and selfadjoint.

21. Give an example of an idempotent operator in R 2which is not a projection.

22. Show that any projection operator Pis positive. Moreover, show that v E

(P v, v )=||P v|| 2 .

23. Prove that the sum P = P 1 + P 2of two projections P 1 and P 2 is a projection operator if

and only if P 1 and P 2 are orthogonal.

24. Prove that the product P = P 1 P 2 of two projections P 1 and P 2 is a projection operator if

and only if P 1 and P 2commute.

25. Find the eigenvalues of a projection operator.

26. Prove that the span of all eigenvectors corresponding to the eigenvalue λof an operator A

is a vector space.

vecanal4.tex; July 1, 2005; 13:34; p. 27

26 CHAPTER 1. LINEAR ALGEBRA

27. Let E (λ)= Ker(A λI ) .

Show that: a) if λ is not an eigenvalue of A , then E (λ )= , and b) if λ is an eigenvalue of

A, then E (λ ) is the eigenspace corresponding to the eigenvalue λ.

28. Show that the operator A λI is invertible if and only if λis not an eigenvalue of the

operator A.

29. Let T be a unitary operator. Then the operators Aand

˜

A= TAT1

are called similar. Show that the eigenvalues of similar operators are the same.

30. Show that an operator similar to a selfadjoint operator is selfadjoint and an operator sim-

ilar to an anti-selfadjoint operator is anti-selfadjoint.

31. Show that all eigenvalues of a positive operator Aare non-negative.

32. Show that the eigenvectors corresponding to distinct eigenvalues of a unitary operator are

orthogonal to each other.

33. Show that the eigenvectors corresponding to distinct eigenvalues of a selfadjoint operator

are orthogonal to each other.

34. Show that all eigenvalues of a unitary operator Ahave absolute value equal to 1.

35. Show that if Ais a projection, then it can only have two eigenvalues: 1 and 0.

vecanal4.tex; July 1, 2005; 13:34; p. 28

Chapter 2

Vector and Tensor Algebra

2.1 Metric Tensor

Let E be a Euclidean space and {ei }= {e1 , . . . en }be a basis (not necessarily

orthonormal). Then each vector vE can be represented as a linear combination

v=

n

X

i= 1

vi e i ,

where vi ,i= 1,...,n , are the components of the vector vwith respect to the

basis { ei } (or contravariant components of the vector v). We stress once again

that contravariant components of vectors are denoted by upper indices (super-

scripts).

Let G= (gij ) be a matrix whose entries are defined by

gij =(e i ,e j ).

These numbers are called the components of the metric tensor with respect to

the basis { ei } (also called covariant components of the metric).

Notice that the matrix Gis symmetric, that is,

gij = gji , GT =G.

Theorem 2.1.1 The matrix G is invertible and

det G> 0 .

The elements of the inverse matrix G 1 =(gij ) are called the contravariant

components of the metric. They satisfy the equations

n

X

j= 1

gij g jk =δ i j,

27

28 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

where δ i jis the Kronecker symbol defined by

δi j=

1, if i= j

0, if i, j

Since the inverse of a symmetric matrix is symmetric, we have

gij = gji .

In orthonormal basis

gij =gi j =δ i j , G= G1 = I .

Let v Ebe a vector. The real numbers

vi =( e i , v)

are called the covariant components of the vector v. Notice that covariant com-

ponents of vectors are denoted by lower indices (subscripts).

Theorem 2.1.2 Let v E be a vector. The covariant and the contravariant

components of v are related by

vi =

n

X

j=1

gij vj , vi =

n

X

j=1

gij vj .

Theorem 2.1.3 The metric determines the inner product and the norm by

(u,v )=

n

X

j=1

n

X

i=1

gij ui vj =

n

X

j=1

n

X

i=1

gij ui vj .

||v||2 =

n

X

i= 1

n

X

j= 1

gij vi vj =

n

X

j=1

n

X

i=1

gij vi vj .

vecanal4.tex; July 1, 2005; 13:34; p. 29

2.2. DUAL SPACE AND COVECTORS 29

2.2 Dual Space and Covectors

A linear mapping ω :E Rthat assigns to a vector v Ea real number hω, vi

and satisfies the condition: u, v E , a R

hω,u+ v i= hω, u i+ hω, v i, and hω, av i= a h ω,v i ,

is called a linear functional.

The space of linear functionals is a vector space, called the dual space of E

and denoted by E , with the addition and multiplication by scalars defined by:

ω, σ E , v E, a R,

hω+ σ, v i= hω, v i+ hσ, v i, and haω, v i=a hω,v i.

The elements of the dual space E are also called covectors or 1-forms . In

keeping with tradition we will denote covectors by Greek letters.

Theorem 2.2.1 The dual space E of a real vector space E is a real vector space

of the same dimension.

Let {ei }= {e1 ,...,en }be a basis in E . A basis {ωi }= {ω1 , . . . ωn }in E such that

hωi ,ej i= δi j

is called the dual basis.

The dual {ωi }of an orthonormal basis is also orthonormal.

Given a dual basis {ωi }every covector σ in E can be represented in a unique

way as

σ=

n

X

i= 1

σi ωi ,

where the real numbers (σi )= ( σ 1 , . . . , σn ) are called the components of the

covector σ with respect to the basis {ωi }.

The advantage of using the dual basis is that it allows one to compute the com-

ponents of a vector vand a covector σby

vi =h ω i , vi

and

σi =hσ ,ei i .

That is

v=

n

X

i=1 e i hω i ,vi

and

σ=

n

X

i=1hσ, e i iω i .

vecanal4.tex; July 1, 2005; 13:34; p. 30

30 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

More generally, the action of a covector σon a vector vhas the form

hσ, v i=

n

X

i=1 hσ,e i ihω i ,vi =

n

X

i=1

σi vi .

The existense of the metric allows us to define the following map

g: E E

that assigns to each vector v a covector g (v ) such that

hg( v), u i=(v,u ).

Then

g( v)=

n

X

i=1

(v, ei ) ωi .

In particular,

g( e k ) =

n

X

i= 1

gki ω i .

Let v be a vector and σbe the corresponding covector, so σ= g ( v ) and v =

g1 (σ ). Then their components are related by

σi =

n

X

j=1

gij vj , vi =

n

X

j=1

gij σ j .

The inverse map g 1 :E Ethat assigns to each covector σa vector g1 (σ)

such that

hσ, u i= (g 1 ( σ), u) ,

can be defined as follows. First, we define

g1 (ω k ) =

n

X

i=1

gki e i .

Then

g1 (σ )=

n

X

k=1

n

X

i=1 hσ, e k ig ki e i .

The inner product on the dual space E is defined so that for any two covectors

αand σ( α,σ)=h α,g1 (σ )i= ( g1 ( α) ,g1 ( α)) .

This definition leads to g ij = ( ω i ,ωj ).

vecanal4.tex; July 1, 2005; 13:34; p. 31

2.2. DUAL SPACE AND COVECTORS 31

Theorem 2.2.2 The inner product on the dual space E is determined by

(α,σ )=

n

X

i= 1

n

X

j=1

gij α i σ j .

In particular,

(ωi ,σ )=

n

X

j=1

gij σ j

The inverse map g 1 can be defined in terms of the inner product of covectors as

g1 (σ )=

n

X

i=1 e i (ω i ,σ) .

Since there is a one-to-one correspondence between vectors and covectors, we

can treat a vector v and the corresponding covector g ( v ) as a single object and

denote the components vi of the vector v and the components of the covector

g( v) by the same letter, that is,

vi =

n

X

j=1

gij vj , vi =

n

X

j=1

gij vj .

We call vi the contravaraint components and vi the covariant components.

This operation is called raising and lowering an index; we use gij to raise an

index and gij to lower an index.

2.2.1 Einstein Summation Convention

In many equations of vector and tensor calculus summation over components of

vectors, covectors and, more generally, tensors, with respect to a given basis fre-

quently appear. Such a summation usually occurs on a pair of equal indices, one

lower index and one upper index, and one sums over all values of indices from 1

to n . The number of summation symbols P n

i=1is equal to the number of pairs of

repeated indices. That is why even simple equations become cumbersome and

uncomfortable to work with. This lead Einstein to drop all summation signs and

to adopt the following summation convention:

1. In any expression there are two types of indices: free indices and repeated

indices.

2. Free indices appear only once in an expression; they are assumed to take

all possible values from 1 to n. For example, in the expression

gij v j

the index i is a free index.

vecanal4.tex; July 1, 2005; 13:34; p. 32

32 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

3. The position of all free indices in all terms in an equation must be the same.

For example, gij vj +α i =σ i

is a correct equation, while the equation

gij vj + α i = σ i

is a wrong equation.

4. Repeated indices appear twice in an expression. It is assumed that there is a

summation over each repeated pair of indices from 1 to n. The summation

over a pair of repeated indices in an expression is called the contraction.

For example, in the expression

gij v j

the index jis a repeated index. It actually means

n

X

j=1

gij vj .

This is the result of the contraction of the indices k and l in the expression

gikvl .

5. Repeated indices are dummy indices: they can be replaced by any other

letter (not already used in the expression) without changing the meaning of

the expression. For example

gij vj = gikvk

just means n

X

j=1

gij vj = gi1 v1 + · ·· + gin vn ,

no matter how the repeated index is called.

6. Indices cannot be repeated on the same level. That is, in a pair of repeated

indices one index is in upper position and another is in the lower position.

For example, vi vi

is a wrong expression.

7. There cannot be indices occuring three or more times in any expression.

For example, the expression giivi

does not make sense.

From now on we will use the Einstein summation convention. We will say that

an equation is written in tensor notation.

vecanal4.tex; July 1, 2005; 13:34; p. 33

2.2. DUAL SPACE AND COVECTORS 33

Examples

First, we list the equations we already obtained above

vi =gij vj ,vj = g ji vi ,

(u, v )=gij ui vj =ui vi =uivi =gij ui vj ,

(α, β )= gij α i β j = α i β i = α i β i = g ij α i β j .

gij g jk =δ k

i.

A contraction of indices one of which belongs to the Kronecker symbol just

renames the index. For example:

δi jv j=v i, δ i jδ j

k=δ i

k,

etc.

The contraction of the Kronecker symbol gives

δi

i=

n

X

i=1

1=n.

vecanal4.tex; July 1, 2005; 13:34; p. 34

34 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

2.3 General Definition of a Tensor

It should be realized that a vector is an invariant geometric object that does not

depend on the basis; it exists by itself independently of the basis. The basis is just

a convenient tool to represent vectors by its components. The componenets of a

vector do depend on the basis. It is this transformation law of the components of

a vector (and, more generally, a tensor as we will see later) that makes an n-tuple

of real numbers (v 1 ,..., vn ) a vector. Not every collection of nreal numbers is

a vector. To represent a vector, a geometric object that does not depend on the

basis, these numbers should transform according to a very special rule under a

change of basis.

Let {ei }= {e1 ,..., en }and {e 0 j}={e 0

1,...,e 0

n}be two dierent bases in E . Obvi-

ously, the vectors from one basis can be decomposed as a linear combination of

vectors from another basis, that is

ei = Λ ji e0 j,

where Λji ,i, j= 1,...,n , is a set of n 2 real numbers, forming the transforma-

tion matrix Λ = ( Λij ). Of course, we also have the inverse transformation

e0 j=˜

Λkj e k

where ˜

Λkj ,k, j = 1,..., n , is another set of n 2 real numbers.

The dual bases {ωi } and {ω 0j } are related by

ω0i = Λij ωj , ωi = ˜

Λij ω0j .

By using the second equation in the first and vice versa we obtain

ei =˜

Λkj Λji ek ,e 0 j= Λ ik˜

Λkj e0

i,

which means that ˜

Λkj Λji = δk

i,Λ i k ˜

Λkj =δ i j.

Thus, the matrix ˜

Λ = (˜

Λij ) is the inverse transformation matrix .

In matrix notation this becomes

˜

ΛΛ = I, Λ ˜

Λ = I ,

which means that the matrix Λis invertible and

˜

Λ = Λ1 .

The components of a vector vwith respect to the basis {e0

i}are

v0i = hω 0i , vi = Λ ij hω j ,vi .

vecanal4.tex; July 1, 2005; 13:34; p. 35

2.3. GENERAL DEFINITION OF A TENSOR 35

This immediately gives v 0i = Λ ij vj .

This is the transformation law of contravariant components. It is easy to

recognize this as the action of the transformation matrix on the column-vector of

the vector components from the left.

We can compute the transformation law of the components of a covector σas

follows

σ0

i=hσ,e 0

ii=˜

Λji hσ,ej i ,

which gives

σ0

i=˜

Λji σj .

This is the transformation law of covariant components. It is the action of

the inverse transformation matrix on the row-vector from the right. That is, the

components of covectors are transformed with the transpose of the inverse trans-

formation matrix!

Now let us compute the transformation law of the covariant components of

the metric tensor gij . By the definition we have

g0

ij =(e 0

i,e 0 j)=˜

Λki ˜

Λlj ( ek ,el ) .

This leads to g 0

ij = ˜

Λki ˜

Λlj gkl .

Similarly, the contravariant components of the metric tensor gij transform

according to g 0ij = Λ ik Λ jl gkl .

The transformation law of the metric components in matrix notation reads

G0 =( Λ1 ) TGΛ1

and G 0−1 = ΛG 1 Λ T .

We denote the determinant of the covariant metric components G= (gij ) by

|g |= det G= det(gi j ) .

Taking the determinant of this equation we obtain the transformation law of

the determinant of the metric

|g 0 |=(det Λ) 2 |g | .

vecanal4.tex; July 1, 2005; 13:34; p. 36

36 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

More generally, a set of real numbers Ti 1 ...ip j 1 ...jq is said to represent components

of a tensor of type (p, q ) ( ptimes contravariant and qtimes covariant) if they

transform under a change of the basis according to

T0i1 ... ip

j1 ... jq = Λ i 1 l 1··· Λ i p l p˜

Λm 1 j 1 ··· ˜

Λmq jq Tl 1 ...lp

m1 ... mq .

This is the general transformation law of the components of a tensor of type

(p, q ).

The rank of the tensor of type (p, q ) is the number ( p+ q ).

Atensor product of a tensor Aof type ( p, q ) and a tensor Bof type (r,s) is a

tensor A B of type (p+ r, q+ s ) with components

(A B) i 1 ...iq l 1 ...lr

j1 ... jq k1 ... ks = A i 1 ... i p

j1 ... jq B l 1 ... l r

k1 ... ks .

The symmetrization of a tensor of type (0,k ) with components Ai 1 ...ik is another

tensor of the same type with components

A(i1 ...ik ) = 1

k!X

ϕSk

Ai ϕ(1)...i ϕ(k ) ,

where summation goes over all permutations of k indices. The symmetrization

is denoted by parenthesis.

The antisymmetrization of a tensor of type (0,k ) with components Ai 1 ...ik is

another tensor of the same type with components

A[i1 ...ik ] = 1

k!X

ϕSk

sign(ϕ ) Ai ϕ (1) ...iϕ (k) ,

where summation goes over all permutations of k indices. The antisymmetriza-

tion is denoted by square brackets.

A tensor Ai 1 ...ik is symmetric if

A( i 1 ...ik ) =Ai 1 ...i k

and anti-symmetric if A [i 1 ...ik ] = Ai 1 ...ik .

Anti-symmetric tensors of type (0, p) are called p -forms.

Anti-symmetric tensors of type (p, 0) are called p -vectors.

A tensor is isotropic if it is a tensor product of gij ,gi j and δi j.

Every isotropic tensor has an even rank.

vecanal4.tex; July 1, 2005; 13:34; p. 37

2.3. GENERAL DEFINITION OF A TENSOR 37

For example, the most general isotropic tensor of rank two is

Aij =aδ i j,

where a is a scalar, and the most general isotropic tensor of rank four is

Aij kl =agij gkl +bδ i

kδ j

l+cδ i

lδ j

k,

where a, b, c are scalars.

2.3.1 Orientation, Pseudotensors and Volume

Since the transformation matrix Λis invertible, then the determinant det Λis

either positive or negative. If det Λ> 0 then we say that the bases { ei } and { e0

i}

have the same orientation, and if det Λ< 0 then we say that the bases { ei } and

{e0

i}have the opposite orientation.

This defines an equivalence relation on the set of all bases on E called the ori-

entation of the vector space E . This equivalence relation divides the set of all

bases in two equivalence classes, called the positively oriented and negatively

oriented bases.

A vector space together with a choice of what equivalence class is positively

oriented is called an oriented vector space.

A set of real numbers Ai 1 ...ip

j1 ... jq is said to represent components of a pseudo-tensor

of type (p,q) if they transform under a change of the basis according to

A0i1 ... i p

j1 ... jq =sign(det Λ ) Λ i 1 l 1·· · Λ i p l p˜

Λm 1 j 1 ··· ˜

Λmq jq Al 1 ...l p

m1 ...mq ,

where sign(x ) = +1 if x> 0 and sign( x )= 1 if x< 0.

The Levi-Civita symbol (also called alternating symbol) is defined by

εi 1 ...in =εi 1 ...in =

+1, if (i 1 ,...,in ) is an even permutation of (1,..., n ) ,

1, if (i 1 ,...,in ) is an odd permutation of (1,..., n ) ,

0, if two or more indices are the same .

The Levi-Civita symbols εi 1 ...in and εi 1 ...in do not represent tensors! They have the

same values in all bases.

Theorem 2.3.1 The determinant of a matrix A =( Aij ) can be written as

det A= εi 1 ...in A 1i 1 . . . Ani n

=εj 1 ... jn A j 1 1 . . . Aj n n

=1

n!ε i 1 ...in ε j 1 ... jn A j 1 i 1 . . . A j n i n .

Here, as usual, a summation over all repeated indices is assumed from 1 to n.

vecanal4.tex; July 1, 2005; 13:34; p. 38

38 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

Theorem 2.3.2 There holds the identity

εi 1 ...in εj 1 ... jn = X

ϕSn

sign(ϕ )δi 1

jϕ(1) ··· δ i n

jϕ(n)

=n!δ i 1

[j 1 ··· δ in

jn ].

The contraction of this identity over k indices gives

εi 1 ...ink m1 ...mk εj 1 ... jnk m1 ...mk =k !(n k )! δi 1

[j 1 ··· δ ink

jnk ].

In particular,

εm 1 ...mn εm 1 ...mn =n! .

Theorem 2.3.3 The sets of real numbers Ei 1 ...in and E i 1 ...in defined by

Ei 1 ...in =p | g| ε i 1 ...in

Ei 1 ...in =1

p| g| εi 1 ...in ,

where | g|= det(gij ) , define (pseudo)-tensors of type (0, n ) and ( n, 0) respectively.

Let {v1 ,..., vn }be an ordered n-tuple of vectors. The volume of the paral-

lelepiped spanned by the vectors { v1 ,...,vn } is a real number defined by

|vol(v 1 ,..., vn ) |= q det((vi , vj )) .

Theorem 2.3.4 Let {ei }be a basis in E, {ωi }be the dual basis, and {v1 ,..., vn }

be a set of n vectors. Let V = ( vij ) be the matrix of contravariant components of

the vectors {v j } v ij =hω i ,v j i ,

and W = ( v i j )be the matrix of covariant components of the vectors {v j }

vij = (e i ,v j )= g ik vkj .

Then the volume of the parallelepiped spanned by the vectors { v1 ,...,v n }is

|vol(v 1 ,..., vn ) |= p | g | | det V |= | det W |

p| g| .

If the vectors {v1 ,..., vn }are linearly dependent, then

vol(v 1 ,..., vn )= 0 .

If the vectors {v1 ,...,vn }are linearly independent, then the volume is a positive

real number that does not depend on the orientation of the vectors.

vecanal4.tex; July 1, 2005; 13:34; p. 39

2.3. GENERAL DEFINITION OF A TENSOR 39

The signed volume of the parallelepiped spanned by an ordered n-tuple of

vectors { v1 ,..., vn } is

vol(v 1 ,..., vn )= p | g| det V

=sign(v 1 ,..., vn )| vol( v1 ,...,vn )| , .

The sign of the signed volume depends on the orientation of the vectors { v 1 ,..., vn } :

sign(v 1 ,...,vn )= sign(det V )=

+1, if { v 1 ,..., vn } is positively oriented

1, if {v1 ,..., vn }is negatively oriented

Theorem 2.3.5 The signed volume is equal to

vol(v 1 ,..., vn )=Ei 1 ...in vi 1 1 · · · vi n n = Ei 1 ...in vi 1 1 ··· vi n n ,

where vij = hω i ,v j iand vi j = (e i ,v j ).

That is why the pseudo-tensor Ei 1 ...in is also called the volume form.

Exterior Product and Duality

The volume form allows one to define the duality of k-forms and (n k)-vectors

as follows. For each k -form Ai 1 ...ik one assigns the dual (n k )-vector by

Aj 1 ... j nk =1

k! E j 1 ... j nki 1 ...ik A i 1 ...ik .

Similarly, for each k -vector Ai 1 ...ik one assigns the dual (n k )-form

Aj 1 ... jnk =1

k! E j 1 ... jnk i1 ...ik A i 1 ...ik .

Theorem 2.3.6 For each k-form αthere holds

∗ ∗ α= ( 1)k (n k) α .

That is,

∗∗ = ( 1)k(n k) .

The exterior product of a k -form A and a m -form B is a (k+ m )-form A B

defined by

(A B )i 1 ...ikj 1 ... jm = (k+ m)!

k!m ! A [ i 1 ...ik B j 1 ... jm ]

Similarly, one can define the exterior product of p-vectors.

Theorem 2.3.7 The exterior product is associative, that is,

(A B )C= A (B C ).

vecanal4.tex; July 1, 2005; 13:34; p. 40

40 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

A collection {v1 ,...,v n1 }of (n 1) vectors defines a covector αby

α=( v1 ∧ · ·· ∧ vn1 )

or, in components,

αj =Eji 1 ...in1 vi 11 ··· vi n1 n1 .

Theorem 2.3.8 Let {v1 ,...,v n1 }be a collection of ( n 1) vectors and S =

span { v 1 , . . . v n1 } be the hyperplane spanned by these vectors. Let ebe a unit

vector orthogonal to S oriented in such a way that {v1 ,..., v n1 , e} is oriented

positively. Then the vector u= g1 (α ) corresponding to the 1 -form α= ( v1

· ·· ∧ v n1 ) is parallel to e(with the same orientation)

u= e|| u||

and has the norm

||u || = vol(v 1 ,..., vn1 , e ) .

In three dimensions, i.e. when n= 3, this defines a binary operation ×, called

the vector product, that is

u= v× w=( v w) ,

or uj =Ejik viwk =p |g |ε jik vi wk .

vecanal4.tex; July 1, 2005; 13:34; p. 41

2.4. OPERATORS AND TENSORS 41

2.4 Operators and Tensors

Let A be an operator on E . Let {ei }= {e1 ,..., en }be a basis in a Euclidean

space and {ωi }= {ω 1 ,...,ωn } be the dual basis in E . The real square matrix

A=( Aij ), i, j= 1 ,..., n, defined by

Aej =Aij e i ,

is called the matrix of the operator A .

Therefore, there is a one-to-one correspondence between the operators on Eand

the real square n× n matrices A= ( Aij ).

It can be computed by

Aij =hω i ,A e j i= gik (e k ,A e j ).

Remark. Notice that the upper index, which is the first one, indicates the row

and the lower index is the second one indicating the column of the matrix. The

convenience of this notation comes from the fact that all upper indices (also

called contravariant indices) indicate the components of vectors and "belong"

to the vector space E while all lower indices (called covariant indices) indicate

components of covectors and "belong" to the dual space E .

The matrix of the identity operator Iis

Iij =δ i j.

For any v Ev=vj e j ,vj =hω j ,vi

we have A v=Aij vj e i .

That is, the components ui of the vector u= A vare given by

ui =Aij vj .

Transformation Law of Matrix of an Operator

Under a change of the basis ei = Λ ji e 0 j, the matrix Aij of an operator Atransforms

according to A 0ij = Λ ik Akm ˜

Λmj ,

which in matrix notation reads

A0 = ΛAΛ1 .

Therefore, the matrix A= ( Aij ) of an operator Arepresents the components of

a tensor of type (1,1). Conversely, such tensors naturally define linear operators

on E . Thus, linear operators on Eand tensors of type (1,1) can be identified.

vecanal4.tex; July 1, 2005; 13:34; p. 42

42 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

The determinant and the trace of the matrix of an operator are invariant under the

change of the basis, that is

det A0 = det A, tr A0 = tr A .

Therefore, one can define the determinant of the operator A and the trace of

the operator A by the determinant and the trace of its matrix, that is,

det A = det A, tr A = tr A .

For self-adjoint operators these definitions are consistent with the definition in

terms of the eigenvalues given before.

The matrix of the sum A +B of two operators A and B is the sum of matrices A

and B of the operatos A and B.

The matrix of a scalar multiple cA is equal to cA , where A is the matrix of the

operator A and c R.

Matrix of the Product of Operators

The matrix of the product C = AB of two operators reads

Cij =hω i ,AB e j i= hω i ,A e k ihω k ,B e j i= Aik Bkj ,

which is exactly the product of matrices A and B.

Thus, the matrix of the product AB of the operators A and B is equal to the

product AB of matrices of these operators in the same order.

The matrix of the inverse A 1 of an invertible operator Ais equal to the inverse

A1 of the matrix A of the operator A.

Theorem 2.4.1 The algebra L(E ) of linear operators on E is isomorphic to the

algebra Mat(n, R ) of real square n ×n matrices.

Matrix of the Adjoint Operator

For the adjoint operator A we have

(ei , A ej )= (A ei , ej )= (ej ,A ei ).

Therefore, the matrix of the adjoint operator is

Akj =gki ( e i , A e j )= gki Ali gl j .

In matrix notation this reads

A = G 1 AT G .

vecanal4.tex; July 1, 2005; 13:34; p. 43

2.4. OPERATORS AND TENSORS 43

Thus, the matrix of a self-adjoint operator Asatisfies the equation

Akj =gki Ali glj or gik Akj =Ali gl j ,

which in matrix notation reads

A= G1 AT G, or GA = AT G .

The matrix of a unitary operator Asatisfies the equation

gki Ali glj Ajm =δk

mor A l i g lj A j m =g im ,

which in matrix notation has the form

G1 AT GA = I, or AT GA = G .

vecanal4.tex; July 1, 2005; 13:34; p. 44

44 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

2.5 Vector Algebra in R 3

We denote the standard orthonormal basis in R 3by

e1 = i, e2 = j, e3 = k ,

so that e i ·ej =δij .

Each vector v is decomposed as

v=v1 i+v2 j+v3 k.

The components are computed by

v1 =v· i, v2 =v· j, v3 =v· k .

The norm of the vector

||v || = q v2

1+v 2

2+v 2

3.

Scalar product is defined by

v· u=v1 u1 +v2 u2 +v3u3 .

The angle between vectors

cos θ=u·v

||u || ||v || .

The orthogonal decomposition of a vector v with respect to a given unit vector u

is v=vk +v ,

where v k = u(u· v), v = v u( u· v) .

We denote the Cartesian coordinates in R 3by

x1 = x, x2 = y, x3 = z.

The radius vector (the position vector) is

r=x i+y j+z k.

The parametric equation of a line parallel to a vector u=a i+b j+c k is

r= r0 +t u,

where r 0 = x 0 i+ y 0 j+ z 0 k is a fixed vector and tis a real parameter. In

components, x=x 0 + at ,y= y 0 + bt ,z= z 0 + ct .

The non-parametric equation of a line (if a, b, c are non-zero) is

x x0

a= yy0

b= zz0

c.

vecanal4.tex; July 1, 2005; 13:34; p. 45

2.5. VECTOR ALGEBRA IN R3 45

The parametric equation of a plane spanned by two non-parallel vectors u and

vis r=r0 +tu +sv,

where t and s are real parameters.

A vector n that is perpendicular to both vectors u and v is normal to the plane.

The non-parametric equation of a plane with the normal n=a i+b j+c k is

(r r 0) ·n= 0

or a (x x 0 )+b( y y 0)+(z z 0 )=0,

which can also be written as

ax + by + cz = d ,

where d=ax 0 + by 0 + cz 0 .

The positive (right-handed) orientation of a plane is defined by the right hand

(or counterclockwise) rule. That is, if u1 and u2 span a plane then we orient the

plane by saying which vector is the first and which is the second. The orientation

is positive if the rotation from u 1 to u 2 is counterclockwise and negative if it is

clockwise. A plane has two sides. The positive side of the plane is the side with

the positive orientation, the other side has the negative (left-handed) orientation.

The vector product of two vectors is defined by

w= u× v=det

i j k

u1u2u3

v1v2v3

,

or, in components,

wi = ε ijk uj vk = 1

2ε ijk (uj vk uk vj ) .

The vector products of the basis vectors are

ei × ej =εijk ek .

If u and v are two nonzero nonparallel vectors, then the vector w= u ×vis

orthogonal to both vectors, u and v , and, hence, to the plane spanned by these

vectors. It defines a normal to this plane.

The area of the parallelogram spanned by two vectors u and vis

area(u,v )= |u × v|= || u || ||v|| sin θ .

vecanal4.tex; July 1, 2005; 13:34; p. 46

46 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

The signed volume of the parallelepiped spanned by three vectors u ,v and w

is

vol(u,v,w )=u· ( v,w )= det

u1u2 u 3

v1v2 v 3

w1w2w3

=εijk ui vj wk .

The signed volume is also called the scalar triple product and denoted by

[u, v, w ]=u· (v× w ) .

The signed volume is zero if and only if the vectors are linearly dependent, that

is, coplanar.

For linearly independent vectors its sign depends on the orientation of the triple

of vectors { u,vw}

vol(u, v, w )= sign(u,v,w)|vol(u, v, w)| ,

where

sign(u, v, w )=( 1 if {u, v,w } is positively oriented

1 if {u, v,w } is negatively oriented

The scalar triple product is linear in each argument, anti-symmetric

[u, v,w ]= [v, u,w ]= [u,w,v ]= [w, v, u]

cyclic [ u,v,w]= [ v,w,u]= [ w, u, v] .

It is normalized so that [i, j, k]= 1.

The orthogonal decomposition of a vector vwith respect to a unit vector ucan

be written in the form

v= u( u· v) u×( u× v) .

The Levi-Civita symbol in three dimensions

εijk =εi jk =

+1 if (i, j, k ) = (1,2,3),(2,3,1),(3, 1, 2)

1 if (i, j, k )= (2,1,3),(3,2,1),(1, 3, 2)

0 otherwise

has the following properties:

εijk = ε jik = εik j = εk ji

εijk = ε jki =εki j

vecanal4.tex; July 1, 2005; 13:34; p. 47

2.5. VECTOR ALGEBRA IN R3 47

εijk εmnl =6 δm

[iδ n

jδ l

k]

=δm

iδ n

jδ l

k+δ m

jδ n

kδ l

i+δ m

kδ n

iδ l jδm

iδ n

kδ l jδm

jδ n

iδ l

kδ m

kδ n

jδ l

i

εijk εmnk =2 δm

[iδ n

j]=δ m

iδ n

jδ m

jδ n

i

εijk εmjk =2 δm

i

εijk εijk = 6

This leads to many vector identities that express double vector product in terms

of scalar product. For example,

u×( v× w)= ( u· w) v( u· v)w

u×( v× w)+ v× ( w× u)+ w×( u× v)= 0

(u× v )× (w× n )= v [ u,w,n ] u[v, w,n ]

(u× v )· (w× n )= (u· w )(v· n ) (u· n )(v· w )

vecanal4.tex; July 1, 2005; 13:34; p. 48

48 CHAPTER 2. VECTOR AND TENSOR ALGEBRA

vecanal4.tex; July 1, 2005; 13:34; p. 49

Chapter 3

Geometry

3.1 Geometry of Euclidean Space

The set Rn can be viewed geometrically as a set of points, that we will denote

by P ,Q , etc. With each point P we associate an ordered n-tuple of real numbers

(xi

P)=(x 1

P,...,x n

P), called the coordinates of the point P. The assignment of

n-tuples of real numbers to the points in space should be bijective. That is,

di erent points are assigned di erent n -tuples, and for every n-tuple there is a

point in space with such coordinates. Such a map is called a coordinate system.

A space Rn with a coordinate system is a Euclidean space if the distance be-

tween any two points P and Q is determined by

d( P ,Q)=v

tn

X

i=1

(xi

Px i

Q) 2 .

Such coordinate system is called Cartesian.

The point Owith the zero coordinates (0 ,..., 0) is called the origin of the Carte-

sian coordinate system.

In Rn it is convenient to associate vectors with points in space. With each point

Pwith Cartesian coordinates ( x1 ,..., xn ) in R n we associate the column-vector

r=(xi ) with the components equal to the Cartesian coordinates of the point P.

We say that this vector points from the origin Oto the point P ; it has its tail

at the point Oand its tip at the point P. This vector is often called the radius

vector, or the position vector, of the point Pand denoted by r .

Similarly, with every two points P and Q with the coordinates ( xi

P) and (x i

Q) we

associate the vector u PQ =rQ rP =(xi

Qx i

P)

that points from the point Pto the point Q .

49

50 CHAPTER 3. GEOMETRY

Obviously, the Euclidean distance is given by

d( P ,Q)=|| r P r Q || .

The standard (orthonormal) basis {e1 ,..., en }of Rn are the unit vectors that con-

nect the origin Owith the points { (1,0,..., 0), ... , (0,..., 0,1)} that have only

one nonzero coordinate which is equal to 1.

The one-dimensional subspaces Li spanned by a single basis vector ei ,

Li = span {e i }= { P|r P = t e i , t R},

are the lines called the coordinate axes . There are n coordinate axes; they are

mutually orthogonal and intersect at only one point, the origin O.

The two-dimensional subspaces Pij spanned by a couple of basis vectors ei and

ej ,Pij =span {ei , ej } ={P |rP =te i + se j , t, s R} ,

are the planes called the coordinate planes. There are n (n 1)/ 2 coordinate

planes; the coordinate planes are mutually orthogonal and intersect along the

coordinate axes.

Let a and b be real numbers such that a< b . The set [a, b ] is a closed interval

in R . A parametrized curve C in Rn is a map C : [a, b ] Rn which assigns a

point in R n

C:r ( t)= ( xi ( t))

to each real number t [a,b].

The positive orientation of the curve C is determined by the standard orientation

of R , that is, by the direction of increasing values of the parameter t.

The point r ( a) is the initial point and the point r (b ) is the endpoint of the

curve.

The curve ( C ) is the parametrized curve with the opposite orientation. If the

curve C is parametrized by r (t ), a t b , then the curve (C ) is parametrized

by (C ) : r (t+ a+ b ) .

The boundary C of the curve C consists of two points C 0 and C 1correspond-

ing to r (a ) and r (b ), that is,

C= C1 C0 .

A curve C is continuous if all the functions xi (t ) are continuous for any ton

[a,b].

vecanal4.tex; July 1, 2005; 13:34; p. 50

3.1. GEOMETRY OF EUCLIDEAN SPACE 51

Let a 1 , a 2 and b 1 ,b 2 be real numbers such that

a1 < b1 , and a2 < b2 .

The set D= [a 1 , b 1] × [ a 2 ,b 2 ] is a closed rectangle in the plane R 2.

Aparametrized surface S in Rn is a map S :D Rn which assigns a point

S:r ( u)= ( xi ( u))

in Rn to each point u= ( u 1 ,u 2) in the rectangle D.

The positive orientation of the surface S is determined by the positive orienta-

tion of the standard basis in R 2. The surface (S ) is the surface with the opposite

orientation.

The boundary S of the surface S consists of four curves S (1),0 ,S (1),1 ,S (2),0 ,

and S (2),1 parametrized by r (a 1 , v ), r ( b 1 , v ), r ( u,a 2 ) and r ( u, b 2 ) respectively.

Taking into account the orientation, the boundary of the surface Sis

S= S (2),0 +S (1),1 S (2),1 S (1),0 .

Let a 1 ,..., ak and b1 ,..., bk be real numbers such that

ai <bi , i=1,..., k .

The set D= [ a 1 , b 1 ] × ··· × [ ak ,bk ] is called a closed k -rectangle in Rk . In

particular, the set [0, 1]k =[0, 1] × ·· · × [0, 1] is the standard k -cube .

Let D= [ a 1 ,b 1 ] × ··· × [ ak , bk ] be a closed rectangle in R k . A parametrized

k-dimensional surface Sin R n is a continuous map S: DR n which assigns

a point S :r (u)= ( xi ( u))

in Rn to each point u= ( u 1 ,..., uk ) in the rectangle D.

A (n 1)-dimensional surface is called the hypersurface . A non-parametrized

hypersurface can be described by a single equation

F( x)= F( x1 ,..., xn )= 0,

where F : Rn R is a real-valued function of ncoordinates.

The boundary S of S consists of (k 1)-surfaces, S (i), 0 and S (i), 1 ,i= 1 ,...,k,

called the faces of the k -surface S . Of course, a k -surface S has 2k faces. The

face S (i), 0 is parametrized by

S(i), 0 :r ( u1 ,..., ui1 , ai ,ui+1 ,..., uk ) ,

where the i-th parameter ui is fixed at the initial point, i.e. ui =ai , and the face

S(i), 0 is parametrized by

S(i), 1 :r ( u1 ,..., ui1 , bi ,ui+1 ,..., uk ) ,

where the i -th parameter ui is fixed at the endpoint, i.e. ui =bi .

vecanal4.tex; July 1, 2005; 13:34; p. 51

52 CHAPTER 3. GEOMETRY

The boundary of the surface Sis defined by

S =

k

X

i=1

(1)i (S (i), 0 S (i), 1 ) .

Let S 1 ,...,Sm be parametrized k -surfaces. A formal sum

S=

m

X

i=1

ai S i

with integer coe cients a 1 ,...,am , is called a k -chain. Usually (but not always)

the integers ai are equal to 1, (1) or 0.

The product of any k -chain S with zero is called the zero chain

0S= 0.

The addition of k-chains and multiplication by integers is defined by

m

X

i=1

ai Si +

m

X

i= 1

bi Si =

m

X

i=1

(ai +bi )Si ,

b

m

X

i=1

ai Si

=

m

X

i=1

(bai )Si .

The boundary of a k -chain S is an (k 1)-chain S defined by

m

X

i= 1

ai Si

=

m

X

i=1

ai Si .

Theorem 3.1.1 For any k-chain S there holds

(S)= 0 .

vecanal4.tex; July 1, 2005; 13:34; p. 52

3.2. BASIC TOPOLOGY OF RN 53

3.2 Basic Topology of R n

Let P 0 be a point in a Euclidean space Rn and ε > 0 be a positive real number.

The open ball Bε ( P 0 ) or radius with the center at P 0is the set of all points

whose distance from the point P 0is less than ε, that is,

Bε (P0 )={ P| d( P ,P0 ) < ε}.

Aneighborhood of a point P 0is any set that contains an open ball centered at

P0 .

Let S be a subset of a Euclidean space Rn . A point P is an interior point of Sif

there is a neighborhood of Pthat lies completely in S.

A point P is an exterior point of S if there is a neighborhood of Pthat lies

completely outside of S.

A point P is a boundary point of S if it is neither an interior nor an exterior

point. If P is a boundary point of S, then every neighborhood of Pcontains

points in S and points not in S.

The set of boundary points of Sis called the boundary of S , denoted by S .

The set of all interior points of Sis called the interior of S , denoted by So .

A set S is called open if every point of Sis an interior point of S, that is, S= So .

A set S is closed if it contains all its boundary points, that is, S= So S .

Henceforth, we will consider only open sets and call them regions of space.

A region S is called connected (or arc-wise connected) if for any two points P

and Q in S there is an arc joining P and Q that lies within S.

A connected region, that is a connected open set, is called a domain.

A domain Sis said to be simply-connected if every closed curve lying within S

can be continuously deformed to a point in the domain without any part of the

curve passing through regions outside the domain.

A domain is simply connected if for any closed curve lying in the domain there

can be found a surface within the domain that has that curve as its boundary.

A domain is said to be star-shaped if there is a point Pin the domain such that

for any other point in the domain the entire line segment joining these two points

lies in the domain.

vecanal4.tex; July 1, 2005; 13:34; p. 53

54 CHAPTER 3. GEOMETRY

3.3 Curvilinear Coordinate Systems

We say that a function f (x )=f (x 1 ,...,xn ) is smooth if it has continuous partial

derivatives of all orders.

Let P be a point with Cartesian coordinates (xi ). Suppose that we assign another

n-tuple of real numbers (qi )= ( q1 ,...,qn ) to the point P, so that

xi = fi ( q),

where fi ( q )=fi (q 1 ,..., qn ) are smooth functions of the variables xi . We will

call this a change of coordinates.

The matrix

J= x i

qj !

is called the Jacobian matrix . The determinant of this matrix is called the Ja-

cobian.

A point P 0at which the Jacobian matrix is invertible, that is, the Jacobian is not

zero, det J, 0, is called a nonsingular point of the new coordinate system (qi ).

Theorem 3.3.1 (Inverse Function Theorem) In a neighborhood of any nonsin-

gular point P0the change of coordinates is invertible.

That is, if xi = fi ( q)and

det xi

qj ! P0

,0,

then for all points suciently close to P0there exist n smooth functions

qi =hi ( x)= hi ( x1 ,..., xn ) ,

of the variables ( xi ) such that

fi ( h1 ( x) ,..., hn ( x)) = xi , hi ( f1 ( q) ,..., fn ( q)) = qi .

The Jacobian matrix of the inverse transformation is the inverse matrix of the

Jacobian matrix of direct transformation, i.e.

xi

qj

qj

xk = δ i

k,and q i

xj

xj

qk = δ i

k.

The curves Ci along which only one coordinate is varied, while all other are

fixed, are called the coordinate curves, that is,

xi = xi ( q1

0,...,qi 1

0,qi ,qi +1

0,...,qn

0)}.

vecanal4.tex; July 1, 2005; 13:34; p. 54

3.3. CURVILINEAR COORDINATE SYSTEMS 55

The vectors e i = r

qi

are tangent to the coordinate curves.

The surfaces Sij along which only two coordinates are varied, while all other are

fixed, are called the coordinate surfaces, that is,

xi =xi ( q1

0,...,q i1

0,qi ,qi+1

0,...,qj 1

0,qj ,qj+1

0,...,qn

0)}.

Theorem 3.3.2 For each point P there are n coordinate curves that pass through

P. The set of tangent vectors { e i }to these coordinate curves is linearly indepen-

dent and forms a basis.

The basis {ei }is not necessarily orthonormal.

The metric tensor is defined as usual by

gij = e i ·e j =

n

X

k=1

xk

qi

xk

qj .

The dual basis of 1-forms is defined by

ωi =dqi =qi

xj σ j ,

where σj is the standard dual basis.

The vector

dr= e i dqi = r

qi dq i

is called the infinitesimal displacement.

The arc length, called the interval, is determined by

ds2 = || dr ||2 = dr· dr= g ij dqidq j .

The volume of the parallelepiped spanned by the vectors {e1 dq1 ,..., en dqn },

called the volume element, is

dV =p | g|dq1 · ·· dqn ,

where, as usual, |g |= det(gij ).

A coordinate system is called orthogonal if the vectors r /∂qi are mutually

orthogonal. The norms of these vectors

hi =

r

qi

are called the scale factors.

vecanal4.tex; July 1, 2005; 13:34; p. 55

56 CHAPTER 3. GEOMETRY

Then one can introduce the orthonormal basis {ei }by

ei =

r

qi

1r

qi = 1

hi

r

qi .

For an orthonormal system the vector components are (there is no dierence

between contravariant and covariant components)

vi =vi =v· e i .

Then the interval has the form

ds2 =

n

X

i=1

h2

i(dq i ) 2 .

The volume element in orthogonal coordinate system is

dV = h1 ··· hndq1 ··· dqn .

3.3.1 Change of Coordinates

Let (qi ) and (q 0 j) be two curvilinear coordinate systems. Then they should be

related by a smooth invertible transformation

q0i = fi ( q)= fi ( q1 ,..., qn ), qi =hi ( q0 )= fi ( q01 ,..., q0n ) ,

such that fi ( h ( q 0 )) =q 0i , hi (f( q)) = qi .

The Jacobian matrices are related by

q0 i

qj

qj

q0k = δ i

k,qi

q0 j

q0 j

qk = δ i

k,

so that the matrix q0i

qj is inverse to qi

q0j .

The basis vectors in these coordinate systems are

e0

i=r

q0i , e i = r

qi .

Therefore, they are related by a linear transformation

e0

i=q j

q0i e j ,ej = q0i

qj e 0 j.

They have the same orientation if the Jacobian of the change of coordinates is

positive and oppositive orientation if the Jacobian is negative.

vecanal4.tex; July 1, 2005; 13:34; p. 56

3.3. CURVILINEAR COORDINATE SYSTEMS 57

Thus, a set of real numbers Ti 1 ...ip j 1 ... jq is said to represent components of a ten-

sor of type ( p,q) ( p times contravariant and q times covariant) if they transform

under a change of coordinates according to

T0i1 ... ip

j1 ... jq =q 0 i1

ql 1 ··· q0i p

ql p

qm 1

q0j1 ··· qmq

q0jq T l 1 ...lp

m1 ...mq .

This is the general transformation law of the components of a tensor of type

(p,q ) with respect to a change of curvilinear coordinates.

A pseudo-tensor has an additional factor equal to the sign of the Jacobian, that

is, the components of a pseudo-tensor of type (p,q) transform as

T0i1 ... i p

j1 ... jq =sign " det q 0 i 1

ql 1 !# q0i 1

ql 1 ··· q0i p

ql p

qm 1

q0j1 ··· qm q

q0jq T l 1 ...lp

m1 ...mq .

This is the general transformation law of the components of a pseudo-tensor

of type (p, q) with respect to a change of curvilinear coordinates.

3.3.2 Examples

The polar coordinates in R 2 are introduced by

x1 =ρcos ϕ , x2 =ρsin ϕ ,

where ρ 0 and 0 ϕ < 2π . The Jacobian matrix is

J= cos ϕ ρ sin ϕ

sin ϕ ρ cos ϕ! .

The Jacobian is det J=ρ .

Thus, the only singular point of the polar coordinate system is the origin ρ= 0.

At all nonsingular points the change of variables is invertible and we have

ρ=p (x 1 )2 +(x 2 )2 , ϕ = cos1 x 1

ρ! =sin1 x 2

ρ! .

The coordinate curves of ρare half-lines (rays) going through origin with the

slope tan ϕ. The coordinate curves of ϕare circles with the radius ρcentered at

the origin.

The cylindrical coordinates in R 3 are introduced by

x1 =ρcos ϕ , x2 =ρsin ϕ , x3 = z

where ρ 0, 0 ϕ < 2π and z R . The Jacobian matrix is

J=

cos ϕ ρ sin ϕ0

sin ϕ ρ cos ϕ0

0 0 1

.

vecanal4.tex; July 1, 2005; 13:34; p. 57

58 CHAPTER 3. GEOMETRY

The Jacobian is det J=ρ .

Thus, the only singular point of the cylindrical coordinate system is the origin

ρ=0. At all nonsingular points the change of variables is invertible and we have

ρ=p (x 1 )2 +(x 2 )2 , ϕ = cos1 x 1

ρ! =sin1 x 2

ρ! ,z= x3 .

The coordinate curves of ρare horizontal half-lines in the plane z= const going

through the z -axis. The coordinate curves of ϕare circles in the plane z= const

of radius ρcentered at the z axis. The coordinate curves of z are vertical lines.

The coordinate surfaces of ρ, ϕ are horizontal planes. The coordinate surfaces of

ρ, z are vertical half-planes going through the z-axis. The coordinate surfaces of

ϕ, z are vertical cylinders centered at the origin.

The spherical coordinates in R 3 are introduced by

x1 = rsin θcos ϕ , x2 = rsin θsin ϕ , x3 = rcos θ

where r 0, 0 ϕ < 2π and 0 θ π. The Jacobian matrix is

J=

sin θcos ϕr cos θcos ϕ r sin θsin ϕ

sin θsin ϕr cos θsin ϕr sin θcos ϕ

cos θ r sin θ 0

.

The Jacobian is det J= r 2 sin θ .

Thus, the singular points of the spherical coordinate system are the points where

either r= 0, which is the origin, or θ= 0 or θ= π , which is the whole z -axis. At

all nonsingular points the change of variables is invertible and we have

r=p ( x1 )2 + ( x2 )2 + ( x3 )2 ,

ϕ=cos1 x 1

ρ! =sin1 x 2

ρ! ,

θ=cos1 x 3

r! ,

where ρ= p (x 1 )2 + (x 2 )2 .

The coordinate curves of rare half-lines going through the origin. The coordi-

nate curves of ϕare circles of radius r sin θ centered at the zaxis. The coordinate

curves of θare vertical half-circles of radius rcentered at the origin. The coor-

dinate surfaces of r, ϕ are half-cones around the z-axis going through the origin.

The coordinate surfaces of r, θ are vertical half-planes going through the z-axis.

The coordinates surfaces of ϕ, θ are spheres of radius rcentered at the origin.

vecanal4.tex; July 1, 2005; 13:34; p. 58

3.4. VECTOR FUNCTIONS OF A SINGLE VARIABLE 59

3.4 Vector Functions of a Single Variable

A vector-valued function is a map v : [a,b ] Efrom an interval [a, b ] of

real numbers to a vector space Ethat assigns a vector v (t ) to each real number

t[ a, b ].

We say that a vector valued function v (t ) has a limit v 0 as t t 0 , denoted by

lim

t t0 v(t )= v 0

if lim

t t0 ||v(t ) v 0 || =0.

A vector valued function v (t ) is continuous at t 0if

lim

t t0 v(t )= v(t 0 ) .

A vector valued function v (t ) is continuous on the interval [a,b] if it is continuous

at every point tof this interval.

A vector valued function v (t ) is di erentiable at t 0 if there exists the limit

lim

h0

v(t 0 + h ) v(t 0)

h.

If this limit exists it is called the derivative of the function v (t ) at t 0 and denoted

by

v0 (t 0 )=d v

dt = lim

h0

v(t 0 + h) v(t 0)

h.

If the function v (t ) is di erentiable at every t in an interval [a, b ], then it is called

di erentiable on that interval.

Let {ei }be a constant basis in Ethat does not depend on t. Then a vector valued

function v (t ) is represented by its components

v(t )= vi ( t ) e i

and the derivative of vcan be computed componentwise

dv

dt = dv i

dt e i .

The derivative is a linear operation, that is,

d

dt (u+ v)= d u

dt + d v

dt , d

dt ( c v)=cd v

dt ,

where c is a scalar constant.

vecanal4.tex; July 1, 2005; 13:34; p. 59

60 CHAPTER 3. GEOMETRY

More generally, the derivative satisfies the product rules

d

dt ( f v)=fd v

dt + d f

dt v ,

d

dt ( u,v)= d u

dt ,v! + u, d v

dt !

Similarly for the exterior product

d

dt ωσ= d ω

dt σ+ω d σ

dt

By taking the dual of this equation we obtain in R 3the product rule for the vector

product d

dt u×v= d u

dt ×v+ u× d v

dt

Theorem 3.4.1 The derivative v 0 (t )of a vector valued function v (t )with the

constant norm is orthogonal to v ( t) . That is, if ||v( t) || = const , then for any t

(v0 (t ),v (t )) = 0 .

vecanal4.tex; July 1, 2005; 13:34; p. 60

3.5. GEOMETRY OF CURVES 61

3.5 Geometry of Curves

Let r= r (t ) be a parametrized curve.

A curve r=r 0 +tu

is a straight line parallel to the vector upassing through the point r 0.

Let u and v be two orthonormal vectors. Then the curve

r= r0 +a(cos t u+sin t v )

is a circle of radius a with the center at r 0 in the plane spanned by the vectors u

and v.

Let {u,v,w }be an orthonormal triple of vectors. Then the curve

r= r0 +b(cos t u+sin t v ) +at w

is the helix of radius bwith the axis passing through the point r 0and parallel to

w.

The vertical distance between the coils of the helix, equal to 2π |a|, is called the

pitch.

Let (qi ) be a curvilinear coordinate system. Then a curve can be described by

qi =qi ( t ), which in Cartesian coordinates becomes r= r ( q ( t )).

The derivative d r

dt = r

qi dqi

dt

of the vector valued function r (t ) is called the tangent vector . If r (t ) represents

the position of a particle at the time t, then r0 is the velocity of the particle.

The norm

dr

dt

=r gij (q( t )) dq i

dt dq j

dt

of the velocity is called the speed. Here, as usual

gij =r

qi · r

qj =

n

X

k=1

xk

qi

xk

qj

is the metric tensor in the coordinate system (qi ).

We will say that a curve r= r (t ) is smooth if:

a) it has continuous derivatives of all orders,

b) there are no self-intersections, and

c) the speed is non-zero, i.e. || r 0 (t) || , 0, at every point on the curve.

vecanal4.tex; July 1, 2005; 13:34; p. 61

62 CHAPTER 3. GEOMETRY

For a curve r : [a, b ] Rn , the possibility that r ( a )=r ( b) is allowed. Then it

is called a closed curve. A closed curve does not have a boundary.

A curve consisting of a finite number of smooth arcs joined together without

self-intersections is called piece-wise smooth, or just regular.

For each regular curve there is a natural parametrization, or the unit-speed

parametrization with a natural parameter ssuch that

dr

ds

=1.

The orientation of a parametrized curve is determined by the direction of in-

creasing parameter. The point r (a ) is called the initial point and the point r ( b)

is called the endpoint.

Nonparametric curves are not oriented.

The unit tangent is determined by

T=

dr

dt

1dr

dt .

For the natural parametrization the tangent is the unit tangent, i.e.

T=d r

ds = r

qi dq i

ds .

The norm of the displacement vector dr= r 0 dt

ds =

dr

dt dt =||dr || = rgi j (q(t )) dqi

dt dq j

dt dt .

is called the length element.

The length of a smooth curve r : [a, b ] Rn is defined by

L=Z C ds =Z b

a

dr

dt dt =Z b

ar g ij (q(t)) dq i

dt dq j

dt dt

For the natural parametrization the length of the curve is simply

L= b a.

That is why, the parameter sis nothing but the length of the arc of the curve

from the initial point r (a ) to the current point r (t )

s( t)=Z t

adτ

dr

dτ

.

vecanal4.tex; July 1, 2005; 13:34; p. 62

3.5. GEOMETRY OF CURVES 63

This means that ds

dt =

dr

dt

,

and d r

dt = ds

dt d r

ds .

The second derivative

r00 =d 2 r

dt2 = d

dt r

qi ! dqi

dt + r

qi d 2 q i

dt2 .

is called the acceleration.

In the natural parametrization this gives the natural rate of change of the unit

tangent d T

ds = d 2 r

ds2 = d

ds r

qi ! dq i

ds + r

qi d 2 q i

ds2 .

The norm of this vector is called the curvature of the curve

κ=

dT

ds

=

dr

dt

1

dT

dt

.

The radius of curvature is defined by

ρ=1

κ.

The normalized rate of change of the unit tangent defines the principal normal

N=ρd T

ds =

dT

dt

1dT

dt .

The unit tangent and the principal normal are orthogonal to each other. They

form an orthonormal system.

Theorem 3.5.1 For any smooth curve r= r(t ), the acceleration r 00 lies in the

plane spanned by the vectors T and N . The orthogonal decomposition of r 00

with respect to T and N has the form

r00 =d|| r 0||

dt T+κ|| r 0||2 N .

The vector d N

ds +κT

is orthogonal to both vectors T and N , and hence, to the plane spanned by these

vectors. In a general space Rn this vector could be decomposed with respect to

a basis in the (n 2)-dimensional subspace orthogonal to this plane. We will

restrict below to the case n= 3.

vecanal4.tex; July 1, 2005; 13:34; p. 63

64 CHAPTER 3. GEOMETRY

In R 3 one defines the binormal

B= T× N.

Then the triple {T, N, B } is a right-handed orthonormal system called a moving

frame.

By using the orthogonal decomposition of the acceleration one can obtain an

alternative formula for the curvature of a curve in R 3as follows. We compute

r0 × r00 =κ|| r0 ||3 B.

Therefore,

κ=|| r0 ×r 00||

|| r 0||3 .

The scalar quantity

τ=B·d N

ds =N · d B

ds

is called the torsion of the curve.

Theorem 3.5.2 (Frenet-Serret Equations) For any smooth curve in R3 there

hold

dT

ds =κN

dN

ds = κT+ τB

dB

ds = τN .

Theorem 3.5.3 Any two curves in R3 with identical curvature and torsion are

congruent.

vecanal4.tex; July 1, 2005; 13:34; p. 64

3.6. GEOMETRY OF SURFACES 65

3.6 Geometry of Surfaces

Let S be a parametrized surface. It can be described in general curvilinear coor-

dinates by qi =qi ( u, v ), where u [a, b ] and v [ c, d ]. Then r= r ( q ( u,v)).

The parameters u and v are called the local coordinates on the surface.

The curves r ( u, v 0) and r ( u 0 , v), with one coordinate being fixed, are called the

coordinate curves.

The tangent vectors to the coordinate curves

ru = r

u= r

qi

qi

u and rv = r

v= r

qi

qi

v

are tangent to the surface.

A surface is smooth if:

a) r (u, v ) has continuous partial derivatives of all orders,

b) the tangent vectors ru and rv are non-zero and linearly independent,

c) there are no self-intersections.

It is allowed that r (a,v )=r (b,v ) and r ( u,c )=r ( u, d ).

A plane TP spanned by the tangent vectors ru and rv at a point Pon a smooth

surface S is called the tangent plane.

A surface is smooth if the tangent plane is well defined, that is, the tangent vec-

tors are linearly independent (nonparallel), which means that it does not degen-

erate to a line or a point at every point of the surface.

A surface is piece-wise smooth if it consists of a finite number of smooth pieces

joined together.

The orientation of the surface is achived by cutting it in small pieces and orient-

ing the small pieces separately. If this can be made consistenly for the whole

surface, then it is called orientable.

The boundary S of the surface r= r ( u, v ), where u [a,b],v [c, d ] consists

of the curves r ( a,v ), r (b, v ), r (u, c ) and r ( u, d ). A surface without boundary

is called closed.

Remark. There are non-orientable smooth surfaces.

In R 3 one can define the unit normal vector to the surface by

n=|| ru ×rv ||1 ru ×rv .

Notice that

|| ru ×rv || =p || ru ||2|| rv ||2 (ru ·rv )2 .

vecanal4.tex; July 1, 2005; 13:34; p. 65

66 CHAPTER 3. GEOMETRY

In components,

(ru × rv )i = g εilm ql

u

qm

v ,

ru ·ru =gi j q i

u

qj

u ,rv · rv =gi j qi

v

qj

v ,ru ·rv =gi j qi

u

qj

v .

The sign of the normal is determined by the orientation of the surface.

For a smooth surface the unit normal vector field nvaries smoothly over the

surface.

The normal to a closed surface in R 3is usually oriented in the outward direction.

In R 3 a surface can also be described by a single equation

F( x, y, z)= 0 .

This equation does not prescribe the orientation though. Then

F

xi

xi

u=0, F

xi

xi

v=0 .

The unit normal vector is then

n=± grad F

|| grad F || .

The sign here is fixed by the choice of the orientation. In components,

ni = F

xi .

Let r (u, v ) be a surface, u= u (t ),v= v (t ) be a curve in the rectangle D =

[a,b ]× [c, d ], and r ((u (t ),v (t )) be the image of that curve on the surface S . Then

the arc length of this curve is

dl =

dr

dt dt ,

or dl 2 = ||dr || 2 = hab duadub ,

where u 1 =u and u 2 = v , and

hab =gij q i

ua

qj

ub ,

is the induced metric on the surface and the indices a, b take only the values

1, 2. In more detail,

dl2 =h11 du2 + 2 h 12 du dv + h22 dv2 ,

and h 11 =ru ·ru ,h 12 =h 21 =ru ·rv ,h 22 =rv ·rv .

vecanal4.tex; July 1, 2005; 13:34; p. 66

3.6. GEOMETRY OF SURFACES 67

The area of a plane spanned by the vectors ru du and rv dv is called the area

element dS = h du dv ,

where h= det hab = || r u || 2 || rv || 2 (ru · rv ) 2 .

In R 3 the area element can also be written as

dS = || r u ×r v || du dv

The area element of a surface in R 3parametrized by

x= u, y= v, z= f( u, v) ,

is

dS = s 1+ f

x! 2

+ f

y! 2 dxdy .

The area element of a surface in R 3described by one equation F (x, y,z )= 0 is

dS =

F

z

1

|| grad F || dx dy

=

F

z

1 s F

x! 2

+ F

y! 2

+ F

z! 2 dxdy

if F /∂z, 0.

The area of a surface Sdescribed by r :D Rn , where D= [a, b ] ×[c, d ], is

S=Z S dS =Z b

aZ d

c

h du dv .

Let S be a parametrized hypersurface defined by r= r ( u )=r ( u 1 ,..., r n1 ).

The tangent vectors to the hypersurface are

r

ua ,

where a= 1, 2,..., (n 1).

The tangent space at a point Pon the hypersurface is the hyperplane equal to the

span of these vectors

T=span ( r

ua ,..., r

ua ) .

The unit normal to the hypersurface Sat the point P is the unit vector northog-

onal to T.

vecanal4.tex; July 1, 2005; 13:34; p. 67

68 CHAPTER 3. GEOMETRY

If the hypersurface is described by a single equation

F( x)= F( x1 ,..., xn )= 0,

then the normal is n=± grad F

|| grad F || .

The sign here is fixed by the choice of the orientation. In components,

(dF )i =F

xi .

The arc length of a curve on the hypersurface is

dl2 = h ab duadub ,

where

hab =gij q i

ua

qj

ub .

The area element of the hypersurface is

dS = h du1 . . . dun1 .

vecanal4.tex; July 1, 2005; 13:34; p. 68

Chapter 4

Vector Analysis

4.1 Vector Functions of Several Variables

The set Rn is the set of ordered n-tuples of real numbers x= (x 1 ,..., xn ). We

call such n -tuples points in the space Rn . Note that the points in space (although

related to vectors) are not vectors themselves!

Let D= [a 1 , b 1 ] × · ·· × [ an , bn ] be a closed rectangle in R n .

Ascalar field is a scalar-valued function of n variables. In other words, it is a

map f :D R , which assigns a real number f (x )=f (x 1 ,...,xn ) to each point

x=( x1 ,...,xn ) of D.

The hypersurfaces defined by f (x)= c,

where c is a constant, are called level surfaces of the scalar field f.

The level surfaces do not intersect.

Avector field is a vector-valued function of nvariables; it is a map v :D Rn

that assigns a vector v (x ) to each point x= (x 1 ,..., xn ) in D.

Atensor field is a tensor-valued function on D.

Let v be a vector field. A point x 0 in Rn such that v (x 0 )=0 is called a singular

point (or a critical point) of the vector field v. A point x is called regular if it

is not singular, that is, if v (x ), 0.

In a neighborhood of a regular point of a vector field vthere is a family of

parametrized curves r (t ) such that at each point the vector vis tangent to the

curves, that is, d r

dt = fv,

where f is a scalar field. Such curves are called the flow lines , or stream lines,

or characteristic curves, of the vector field v.

69

70 CHAPTER 4. VECTOR ANALYSIS

Flow lines do not intersect.

No flow lines pass through a singular point.

The flow lines of a vector field v= vi (x ) ei can be found from the dierential

equations dx 1

v1 =··· = dxn

vn .

vecanal4.tex; July 1, 2005; 13:34; p. 69

4.2. DIRECTIONAL DERIVATIVE AND THE GRADIENT 71

4.2 Directional Derivative and the Gradient

Let P 0 be a point and ube a unit vector. Then r (s )=r 0 + su is the equation of

the oriented line passing through P 0with the unit tangent u.

Let f (x ) be a scalar field. Then the derivative

d

ds f(x(s )) s=0

at s= 0 is called the directional derivative of f at the point P 0in the direction

of u and denoted by

u f= d

ds f(x( s)) s=0.

The directional derivatives in the direction of the basis vectors ei are the partial

derivatives

e i f= f

xi ,

which are also denoted by

i f= f

xi .

More generally, let r (s ) be a parametrized curve in the natural parametrization

and u=d r/ ds be the unit tangent. Then

u f= f

xi dxi

ds

In curvilinear coordinates

u f= f

qi dqi

ds .

The covector (1-form) field with the components f /∂qi is denoted by

df = f

xi dx i = f

qi dq i .

Therefore, the 1-forms dqi form a basis in the dual space of covectors.

The vector field corresponding to the 1-form df is called the gradient of the

scalar field fand denoted by

grad f= f= gij f

qi e j .

The directional derivative is simply the action of the covector d f on the vector u

(or the inner product of the vectors grad f and u)

u f= h df , u i =( grad f, u) .

vecanal4.tex; July 1, 2005; 13:34; p. 70

72 CHAPTER 4. VECTOR ANALYSIS

Therefore,

u f= || grad f || cos θ ,

where θ is the angle between the gradient and the unit tangent u.

Gradient of a scalar field points in the direction of the maximum rate of increase

of the scalar field.

The maximum value of the directional derivative at a fixed point is equal to the

norm of the gradient max

u u f=|| grad f || .

The minimum value of the directional derivative at a fixed point is equal to the

negative norm of the gradient

min

u u f=−|| grad f || .

Let f be a scalar field and P 0be a point where grad f, 0. Let r= r ( s ) be

a curve passing through Pwith the unit tangent u=d r/ ds. Suppose that the

directional derivative vanishes, u f= 0. Then the unit tangent uis orthogonal

to the gradient grad f at P . The set of all such curves forms a level surface

f( x)= c, where c= f( P0 ). The gradient grad f is orthogonal to the tangent

plane to the this surface at P 0.

Theorem 4.2.1 For any smooth scalar field f there is a level surface f ( x)= c

passing through every point where the gradient of f is non-zero, grad f,0 .

The gradient grad f is orthogonal to this surface at this point.

A vector field v is called conservative if there is a scalar field fsuch that

v= grad f .

The scalar field fis called the scalar potential of v.

vecanal4.tex; July 1, 2005; 13:34; p. 71

4.3. EXTERIOR DERIVATIVE 73

4.3 Exterior Derivative

Recall that antisymmetric tensors of type (0,k ) are called the k-forms, and the

antisymmetric tensors of type (k, 0) are called k -vectors. We denote the space of

all k -forms by Λk and the space of all k -vectors by Λk .

The exterior derivative dis an operator

d:Λ k Λ k+1 ,

that assigns a (k+ 1)-form to each k-form. It is defined as follows.

A scalar field can be also called a 0-form. The exterior derivative of a zero form

fis a 1-form

df = f

qi dxi

with components

(df )i = f

qi .

The exterior derivative of a 1-form is a 2-form dσ defined by

(dσ )ij =∂σ j

qi ∂σ i

qj .

The exterior derivative of a k -form σ is a (k+ 1)-form dσ with components

(d σ )i 1 i 2 ...ik+1 = (k+ 1)

q[i1 σ i 2 ...ik+1 ] = X

ϕSk+1

sign(ϕ )

qi ϕ(1) σ i ϕ(2) ...i ϕ(k+ 1) .

Theorem 4.3.1 The exterior derivative of a k-form is a ( k+ 1)-form.

The exterior derivative plays the role of the gradient for k-forms.

Theorem 4.3.2 The exterior derivative has the property

d2 =0

Recall that the duality operator assigns a (n k)-vector to each k-form and an

(n k )-form to each k -vector:

:Λk Λnk , :Λk Λnk .

Therefore, one can define the operator

d:Λ k Λnk 1 ,

which assigns a (n k 1)-vector to each k-form by

(d σ )i 1 ...ink 1 = 1

k! g 1/ 2 ε i 1 ...ink1 j1 j2 ... jk+1

qj 1 σ j 2 ... jk+1

vecanal4.tex; July 1, 2005; 13:34; p. 72

74 CHAPTER 4. VECTOR ANALYSIS

We can also define the operator

d : Λk Λk 1

acting on k -vectors, which assigns a (k 1) vector to a k -vector.

Theorem 4.3.3 For any k-vector A with components Ai 1 ...ik there holds

(d A ) i 1 ...ik1 =(1)nk+1 g1/ 2

qj g 1/2 A ji 1 ...ik1 .

The operator d plays the role of the divergence of k -vectors.

Theorem 4.3.4 The operator d has the property

(d )2 = 0

Let G denote the operator that converts k -vectors to k-forms,

G:Λ k Λ k .

That is, if Aj 1 ... jk are the components of a k-vector, then the corresponding k-form

σ=GA has components

σi 1 ...ik =gi 1 j 1 . . . gi k j k Aj 1 ... jk .

Then the operator G d:Λ k Λ nk 1

assigns a (n k 1)-form to each k-form by

(G d σ ) i 1 ...ink 1 = 1

k! g 1/ 2 g i 1 m 1 ··· g i nk1 m nk1 ε m 1 ...mnk1 j1j2 ... jk+1

qj 1 σ j 2 ... jk+1

The operator d plays the role of the curl of k-forms.

Further, we can define the operator

δ=G dG: Λk Λk1 ,

which assigns a (k 1)-form to each k-form.

Theorem 4.3.5 For any k-form A with components σ i 1 ...ik there holds

(δσ )i 1 ...ik1 = (1)nk+1 g 1/ 2 gi 1 j 1 ··· gi k1 j k1

qj g 1/2 g jp gj 1 m 1 ··· gj k1 m k1 σ pm 1 ...mk1 .

The operator δ plays the role of the divergence of k-forms.

Theorem 4.3.6 The operator δ has the property

δ2 =0 .

vecanal4.tex; July 1, 2005; 13:34; p. 73

4.3. EXTERIOR DERIVATIVE 75

Therefore the operator

L= dδ+ δ d

assigns a k -form to each k-form, that is,

L:Λ k Λ k .

This operator plays the role of the Laplacian of k-forms.

Ak -form σ is called closed if dσ= 0.

Ak -form σ is called exact if there is a (k 1)-form α such that σ=d α .

A 1-form σ corresponding to conservative vector field vis exact, that is, σ= d f .

Every exact k -form is closed.

vecanal4.tex; July 1, 2005; 13:34; p. 74

76 CHAPTER 4. VECTOR ANALYSIS

4.4 Divergence

The divergence of a vector field vis a scalar field defined by

div v= (1)n+1 d v ,

which in local coordinates becomes

div v= g1/ 2

qi ( g 1/2 vi ) .

where g= det gij .

Theorem 4.4.1 For any vector field v the divergence div v is a scalar field.

The divergence of a covector field σis

div σ= g 1/2

qi ( g 1/2 g ij σ j ) .

In Cartesian coordinates this gives simply

div v= i vi .

A vector field v is called solenoidal if

div v= 0 .

The 2-form v dual to a solenoidal vector field vis closed, that is, d v= 0.

Physical Interpretation of Divergence

The divergence of a vector field is the net outflux of the vector field per unit

volume.

vecanal4.tex; July 1, 2005; 13:34; p. 75

4.5. CURL 77

4.5 Curl

Recall that the operator dassigns a ( n k1)-vector to a k -form. In case n= 3

and k= 1 this operator assigns a vector to a 1-form. This enables one to define

the curl operator in R 3 , which assigns a vector to a covector by

curl σ= dσ ,

or, in components,

(curl σ )i = g 1/2 εijk

qj σ k =g 1/ 2 det

e1 e2e3

q1

q2

q 3

σ1 σ2σ 3

.

We can also define the curl of a vector field vby

(curlv)i = g1/ 2 ε ijk

qj (gkm vm ) .

In Cartesian coordinates we have simply

(curl σ )i = εijk j σk .

This can also be written in the form

curl σ= det

i j k

x yz

σ1 σ2 σ3

A vector field v in R 3 is called irrotational if

curlv = 0 .

The one-form σ corresponding to an irrotational vector field v is closed, that is

dσ=0.

Each conservative vector field is irrotational.

Let v be a vector field. If there is a vector field Asuch that

v= curl A ,

when A is called the vector potential of v.

If v has a vector potential, then it is solenoidal.

If A is a vector potential for v, then the 2-form v dual to v is exact, that is,

v=d α, where α is the 1-form corresponding to A.

Physical Interpretation of the Curl

The curl of a vector field measures its tendency to swirl; it is the swirl per unit

area.

vecanal4.tex; July 1, 2005; 13:34; p. 76

78 CHAPTER 4. VECTOR ANALYSIS

4.6 Laplacian

The scalar Laplace operator (or the Laplacian) is the map :C (Rn )

C (R n ) that assigns a scalar field to a scalar field. It is defined by

f =div grad f =g1/ 2 i g1/2 gij j f .

In Cartesian coordinates it is simply

f =ii f.

The Laplacian of a 1-form (covector field) σis defined as follows. First, one

obtains a 2-form dσ by the exterior derivative. Then one take the dual of this

2-form to get a (n 2)-form d σ . Then one acts by exterior derivative to get a

(n 1)-form d d σ , and, finally, by taking the dual again one gets the 1-form

d d σ. Similarly, reversing the order of operations one gets the 1-form dd σ.

The Laplacian is the sum of these 1-forms, i.e.

σ = (G dG d + dG dG )σ .

The expression of this Laplacian in components is too complicated, in general.

The components expression for this is

( v )i = gij

qj g 1/ 2

qk g 1/2 g 1/ 2

qj g 1/2 g pi g q j

qp gqk

+g1/ 2

qj g 1/2 g pj gqi

qp g qk vk .

Of course, in Cartesian coordinates this simpifies significantly

(v )i =jj vi .

In R 3 it can be written as

v = grad div v curl curl v .

Interpretation of the Laplacian

The Laplacian measures the dierence between the value of a scalar field f ( P)

at a point Pand the average of faround this point.

vecanal4.tex; July 1, 2005; 13:34; p. 77

4.7. DIFFERENTIAL VECTOR IDENTITIES 79

4.7 Di erential Vector Identities

The identities below that involve the vector product and the curl apply only for

R3 . Other formulas are valid for arbitrary Rn in arbitrary coordinate systems:

grad ( fh)= ( grad f) h+ f grad h

div( f v )= ( grad f )·v+f div v

grad f (h (x )) = df

dh grad h

curl (f v)= ( grad f)× v+f curlv

div(u× v )= ( curl u )·v u · ( curl v)

curl grad f= 0

div curlv = 0

div( grad f× grad h )= 0

Let ei be the standard basis in Rn ,xi be the Cartesian coordinates, r=xi ei be

the position (radius) vector field and r= || r || =p xi xi . Scalar fields that depend

only on r and vector fields that depend on x and r are called radial fields . Below

ais a constant vector field.

div r= n

curl r = 0

grad ( a· r)= a

curl ( a× r)= 2 a

grad r= r

r

grad f (r )= df

dr r

r

grad 1

r=r

r3

grad rk = krk2 r

f( r) =f00 +(n1)

rf 0

rk =k (k +n 2)rk2

1

rn2 =0

Some useful formulas when working with radial fields are

i xk = δ k

i, δ i

i=n.

vecanal4.tex; July 1, 2005; 13:34; p. 78

80 CHAPTER 4. VECTOR ANALYSIS

4.8 Orthogonal Curvilinear Coordinate Systems in R3

Let (q 1 ,q 2 , q 3) be an orthogonal coordinate system in R 3 and {ˆ

e1 , ˆ

e2 , ˆ

e3 }be the

corresponding orthonormal basis

ˆ

ei =1

hi

r

qi .

where

hi =

r

qi

are the scale factors.

Then for any vector v=vi ˆ

ei the contravariant and the covariant components

coincide vi =vi = ˆ

ei · v .

The displacement vector, the interval and the volume element in the orthogonal

coordinate system are dr=h 1 ˆ

e1 + h2 ˆ

e2 +h3 ˆ

e3 ,

ds2 =h2

1(dq 1 ) 2 +h 2

2(dq 2 ) 2 +h 2

3(dq 3 ) 2 ,

dV = h1 h2h3dq1dq2dq3 .

The dierential operators introduced above take the following form

grad f= ˆ

e1 1

h1

q1 f+ ˆ

e2 1

h2

q2 f+ ˆ

e3 1

h3

q3 f

div v= 1

h1 h2 h3 (

q1 ( h 2 h 3 v 1)+

q2 ( h 3 h 1 v 2 )+

q3 ( h 1 h 2 v 3 ) )

curlv = ˆ

e1 1

h2 h3 "

q2 ( h 3 v 3)

q3 ( h 2 v 2 ) #

+ˆ

e2 1

h3 h1 "

q3 ( h 1 v 1)

q1 ( h 3 v 3 ) #

+ˆ

e3 1

h1 h2 "

q1 ( h 2 v 2)

q2 ( h 1 v 1 ) #

f =1

h1 h2 h3 (

q1 h 2 h 3

h1

q1 ! +

q1 h 2 h 3

h1

q1 ! +

q1 h 2 h 3

h1

q1 !) f

Cylindrical coordinates:

dr= dρˆ

eρ +ρd ϕ ˆ

eϕ +dz ˆ

ez

ds2 = d ρ2 +ρ2 dϕ2 + dz2

vecanal4.tex; July 1, 2005; 13:34; p. 79

4.8. ORTHOGONAL CURVILINEAR COORDINATE SYSTEMS IN R3 81

dV =ρ dρ dϕ dz

grad f= ˆ

eρ ρ f+ ˆ

eϕ 1

ρ ϕ f+ˆ

ez z f

div v= 1

ρ ρ ( ρv ρ )+1

ρ ϕ v ϕ + z vz

curlv = 1

ρ

ˆ

eρ ρ ˆ

eϕ ˆ

ez

ρϕz

vρ ρ vϕvz

f =1

ρ ρ ( ρ ∂ ρ f)+ 1

ρ2 2

ϕf+2

zf

Spherical coordinates:

dr= dr ˆ

er + rdθ ˆ

eθ +r sin θd ϕ ˆ

eϕ

ds2 =dr2 + r2 dθ2 + r2 sin2 θ dϕ 2

dV = r2 sin θ dr dθ d ϕ

grad f= ˆ

er r f+ ˆ

eθ 1

rθ f+ˆ

eϕ 1

rsin θ ϕ f

div v= 1

r2 r ( r 2 v r )+1

rsin θ θ (sin θv θ )+ 1

rsin θ ϕ v ϕ

curlv = 1

r2 sin θ

ˆ

er rˆ

eθ rsin θ ˆ

eϕ

rθϕ

vr rvθ r sin θ vϕ

f =1

r2 r ( r 2 r f)+ 1

r2 sin θ θ (sin θθ f)+ 1

r2 sin2 θ 2

ϕf

vecanal4.tex; July 1, 2005; 13:34; p. 80

82 CHAPTER 4. VECTOR ANALYSIS

vecanal4.tex; July 1, 2005; 13:34; p. 81

Chapter 5

Integration

5.1 Line Integrals

Let C be a smooth curve described by r (t ), where t [a, b ]. The length of the

curve is defined by

L=Z C ds =Z b

a

dr

dt dt .

Let f be a scalar field. Then the line integral of the scalar field f is

ZC f ds = Z b

af(x (t ))

dr

dt dt .

If v is a vector field, then the line integral of the vector field v along the curve

Cis defined by

ZC v·dr= Z b

av(x (t )) ·d r

dt dt .

In components, the line integral of a vector field takes the form

ZC v·dr= ZC vidqi = ZC v 1 dq 1 + · ·· +vn dqn ,

where vi = gij vj are the covariant components of the vector field.

The expression

σ=vi dqi =v1dq1 +· · · +vndqn

is called a di erential 1-form . Each covector naturally defines a dierential

form. That is why it is also called a 1-form.

If C is a closed curve, then the line integral of a vector field is denoted by

IC v·dr

and is called the circulation of the vector field v about the closed curve C.

83

84 CHAPTER 5. INTEGRATION

5.2 Surface Integrals

Let S be a smooth parametrized surface described by r :D Rn , where

D=[ a,b]× [ c, d ]. The surface integral of a scalar field f is

ZS f dS = Z b

aZ d

cf(x (u, v )) h du dv ,

where u 1 = u,u 2 =v ,h= det hab and hab is the induced metric on the surface

hab =gij q i

ua

qi

ub .

Let A be an antisymmetric tensor field of type (0, 2) with components Ai j . It

naturally defines the di erential 2-form

α=X

i< j

Aij dqi dq j

=A12dq1 dq2 +· ·· +A1n dq1 dqn

+A23dq2 dq3 +· ·· +A2n dq2 dqn

+· ·· +An1,n dqn1 dqn .

Then the surface integral of a 2 -form α is defined by

ZS α= ZS X

i< j

Aij dqi dq j =Z b

aZ d

cX

i<j

Aij J i j dudv ,

where

Jij = q i

u

qj

v qj

u

qi

v .

In R 3 every 2-form defines a dual vector. Therefore, one can integrate vectors

over a surface. Let vbe a vector field in R 3. Then the dual two form is

Aij = g ε i jk vk ,

or A 12 = gv 3 , A 13 = gv 2 ,A 23 = g v 1 .

Ttherefore,

α= g v3 dq1 dq2 v2dq1 dq3 +v1 dq2 dq3 .

Then the surface integral of the vector field v, called the total flux of the

vector field through the surface, is

ZS α= ZS v·ndS = Z b

aZ d

c[v, r u ,r v ]du dv ,

vecanal4.tex; July 1, 2005; 13:34; p. 82

5.2. SURFACE INTEGRALS 85

where n=|| ru ×rv ||1 ru ×r v

is the unit normal to the surface and

[v, ru , rv ]= vol(v, ru ,rv )= gε i jk vi q j

u

qk

v .

Similarly, the integrals of a di erential k -form

α=X

i1 <···<ik

Ai 1 ...ik dqi 1 ∧ · ·· ∧ dqi k

with components Ai 1 ...ik over a k -dimensional surface qi =qi ( u 1 ,..., uk ), ui

[ai ,bi ] are defined by

ZS α= Z bk

ak ··· Z b 1

a1 X

i1 <···< i k

Ai 1 ...ik Ji 1 ...ik du1 ··· duk ,

where

Ji 1 ...ik = k! q [i 1

u1 ··· qi k ]

uk .

The surface integral over a closed surface Swithout boundary is denoted by

IS α .

In the case k= n 1 we obtain the integral of a (n 1)-form α over a hypersurface

ZS α= Z bn1

an1 ··· Z b 1

a1 X

i1 <···<in1

Ai 1 ...i n1 Ji 1 ...i n1 du1 ··· dun1 .

Let n be the unit vector orthogonal to the hypersurface and v= α be the vector

field dual to the (n 1)-form α . Then

ZS α= Z bn1

an1 ··· Z b 1

a1

v· n h du1 ··· dun1 .

This defines the total flux of the vector field v through the hypersurface S .

The normal can be determined by

h n j =1

(n 1)! gε ji 1 ...in1

qi 1

u1 ··· qi n1

un1 .

vecanal4.tex; July 1, 2005; 13:34; p. 83

86 CHAPTER 5. INTEGRATION

5.3 Volume Integrals

Let D= [ a 1 ,b 1 ] ×···× [ an ,bn ] be a domain in Rn described in local coordinates

(qi ) by qi [ai ,bi ] .

The volume element in general curvilinear coordinates is

dV = g dq1 · ·· dqn ,

where g= det(gij ).

The volume of the region D is

V=Z D dV =Z b 1

a1 ··· Z b n

an

g dq 1 ··· dqn .

The volume integral of a scalar field f (x ) is

ZD f dV = Zb1

a1 ··· Z b n

an

f( x( q)) g dq1 ··· dq n .

vecanal4.tex; July 1, 2005; 13:34; p. 84

5.4. FUNDAMENTAL INTEGRAL THEOREMS 87

5.4 Fundamental Integral Theorems

5.4.1 Fundamental Theorem of Line Integrals

Theorem 5.4.1 Let C be a smooth curve parametrized by r (t ) , t [a, b ]. Then

for any scalar field f (a 0-form)

ZC df = ZC

f

qi dq i = Z C grad f·dr= f( x( b)) f( x( a)) .

The line integral of a conservative vector field does not depend on the interior of

the curve but only on the endpoints of the curve.

Corollary 5.4.1 The circulation of a smooth conservative vector field over a

closed smooth curve is zero,

IC grad f· dr=0.

5.4.2 Green's Theorem

Theorem 5.4.2 Let x and y be the Cartesian coordinates in R2 . Let U be a

bounded region in R2 with the boundary U, which is a closed curve oriented

couterclockwise. Then for any 1 -form α= Aidxi =A1 dx + A2dy

ZU A 2

x A1

y! dxdy = I U ( A 1 dx + A2dy) .

5.4.3 Stokes's Theorem

Theorem 5.4.3 Let S be a bounded surface in R3 with the boundary S oriented

consistently with the the surface S. Then for any vector field v

ZS curl v ·ndS = IS v·dr.

5.4.4 Gauss's Theorem

Theorem 5.4.4 Let D be a bounded domain in R3 with the boundary D ori-

ented by an outward normal. Then for any vector field v

ZD div vdV = ID v·ndS

vecanal4.tex; July 1, 2005; 13:34; p. 85

88 CHAPTER 5. INTEGRATION

5.4.5 General Stokes's Theorem

Theorem 5.4.5 Let S be a bounded smooth k-dimensional surface in R nwith the

boundary S , which is a closed ( k 1)-dimensional surface oriented consistently

with S. Let

α=X

i1 <··· < ik1

Ai 1 ...ik1 dqi 1 ∧ · ·· ∧ dqi k1

be a smooth ( k 1) -form. Then

ZS dα= IS α .

In components this formula takes the form

ZS

Ai 1 ...ik1

qj

q[ j

u1

qi 1

u2 ··· qi k1]

uk du 1 ··· duk =Z SA i 1 ...ik1

q[ i 1

u1 ··· qi k1]

uk1 du 1 ··· duk1 .

vecanal4.tex; July 1, 2005; 13:34; p. 86

Chapter 6

Potential Theory

89

90 CHAPTER 6. POTENTIAL THEORY

6.1 Simply Connected Domains

vecanal4.tex; July 1, 2005; 13:34; p. 87

6.2. CONSERVATIVE VECTOR FIELDS 91

6.2 Conservative Vector Fields

6.2.1 Scalar Potential

vecanal4.tex; July 1, 2005; 13:34; p. 88

92 CHAPTER 6. POTENTIAL THEORY

6.3 Irrotational Vector Fields

vecanal4.tex; July 1, 2005; 13:34; p. 89

6.4. SOLENOIDAL VECTOR FIELDS 93

6.4 Solenoidal Vector Fields

6.4.1 Vector Potential

vecanal4.tex; July 1, 2005; 13:34; p. 90

94 CHAPTER 6. POTENTIAL THEORY

6.5 Laplace Equation

6.5.1 Harmonic Functions

vecanal4.tex; July 1, 2005; 13:34; p. 91

6.6. POISSON EQUATION 95

6.6 Poisson Equation

6.6.1 Dirac Delta Function

6.6.2 Point Sources

6.6.3 Dirichlet Problem

6.6.4 Neumann Problem

6.6.5 Green's Functions

vecanal4.tex; July 1, 2005; 13:34; p. 92

96 CHAPTER 6. POTENTIAL THEORY

6.7 Fundamental Theorem of Vector Analysis

vecanal4.tex; July 1, 2005; 13:34; p. 93

Chapter 7

Basic Concepts of Di erential

Geometry

97

98 CHAPTER 7. BASIC CONCEPTS OF DIFFERENTIAL GEOMETRY

7.1 Manifolds

vecanal4.tex; July 1, 2005; 13:34; p. 94

7.2. DIFFERENTIAL FORMS 99

7.2 Di erential Forms

7.2.1 Exterior Product

7.2.2 Exterior Derivative

vecanal4.tex; July 1, 2005; 13:34; p. 95

100 CHAPTER 7. BASIC CONCEPTS OF DIFFERENTIAL GEOMETRY

7.3 Integration of Dierential Forms

vecanal4.tex; July 1, 2005; 13:34; p. 96

7.4. GENERAL STOKES'S THEOREM 101

7.4 General Stokes's Theorem

vecanal4.tex; July 1, 2005; 13:34; p. 97

102 CHAPTER 7. BASIC CONCEPTS OF DIFFERENTIAL GEOMETRY

7.5 Tensors in General CurvilinearCoordinate Systems

7.5.1 Covariant Derivative

vecanal4.tex; July 1, 2005; 13:34; p. 98

104 CHAPTER 8. APPLICATIONS

8.1 Mechanics

8.1.1 Inertia Tensor

8.1.2 Angular Momentum Tensor

vecanal4.tex; July 1, 2005; 13:34; p. 99

8.2. ELASTICITY 105

8.2 Elasticity

8.2.1 Strain Tensor

8.2.2 Stress Tensor

vecanal4.tex; July 1, 2005; 13:34; p. 100

106 CHAPTER 8. APPLICATIONS

8.3 Fluid Dynamics

8.3.1 Continuity Equation

8.3.2 Tensor of Momentum Flux Density

8.3.3 Euler's Equations

8.3.4 Rate of Deformation Tensor

8.3.5 Navier-Stokes Equations

vecanal4.tex; July 1, 2005; 13:34; p. 101

8.4. HEAT AND DIFFUSION EQUATIONS 107

8.4 Heat and Diusion Equations

vecanal4.tex; July 1, 2005; 13:34; p. 102

108 CHAPTER 8. APPLICATIONS

8.5 Electrodynamics

8.5.1 Tensor of Electromagnetic Field

8.5.2 Maxwell Equations

8.5.3 Scalar and Vector Potentials

8.5.4 Wave Equations

8.5.5 D'Alambert Operator

8.5.6 Energy-Momentum Tensor

vecanal4.tex; July 1, 2005; 13:34; p. 103

8.6. BASIC CONCEPTS OF SPECIAL AND GENERAL RELATIVITY 109

8.6 Basic Concepts of Special and General Relativity

vecanal4.tex; July 1, 2005; 13:34; p. 104

110 CHAPTER 8. APPLICATIONS

vecanal4.tex; July 1, 2005; 13:34; p. 105

Bibliography

[1] A. I. Borisenko and I. E. Tarapov, Vector and Tensor Analysis, Dover, 1979

[2] D. E. Bourne and P. C. Kendall, Vector Analysis and Cartesian Tensors, Nelson,

1977

[3] H. F. Davis and A. D. Snider, Introduction to Vector Analysis, 7th Edition,

Brown Publishers, 1995

[4] J. H. Hubbard and B. B. Hubbard, Vector Calculus, Linear Algebra, and Di er-

ential Forms, (2nd Edition), Prentice Hall, 2001

[5] H. M. Schey, Div, grad, curl, and all that: an informal text on vector calculus,

W.W. Norton, 1997

[6] Th. Shifrin, Multivariable Mathematics, Wiley, 2005

[7] M. Spivak, Calculus on Manifolds: A Modern Approach to Classical Theorems

of Advanced Calculus, HarperCollins Publishers, 1965

[8] J. L. Synge and A. Schild, Tensor Calculus, Dover, 1978

[9] R. C. Wrede, Introduction to Vector and Tensor Analysis, Dover, 1972

111

112 BIBLIOGRAPHY

vecanal4.tex; July 1, 2005; 13:34; p. 106

ResearchGate has not been able to resolve any citations for this publication.

  • Th
  • Shifrin

Th. Shifrin, Multivariable Mathematics, Wiley, 2005