Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
By ir. J.C.A. Wevers
c (1999, 2008 J.C.A. Wevers Version: May 4, 2008
Dear reader,
This document contains 66 pages with mathematical equations intended for physicists and engineers. It is intended
to be a short reference for anyone who often needs to look up mathematical equations.
This document can also be obtained from the author, Johan Wevers (johanw@vulcan.xs4all.nl).
It can also be found on the WWW on http://www.xs4all.nl/˜johanw/index.html.
This document is Copyright by J.C.A. Wevers. All rights reserved. Permission to use, copy and distribute this
unmodiﬁed document by any means and for any purpose except proﬁt purposes is hereby granted. Reproducing this
document by any means, included, but not limited to, printing, copying existing prints, publishing by electronic or
other means, implies full agreement to the above nonproﬁtuse clause, unless upon explicit prior written permission
of the author.
The C code for the rootﬁnding via Newtons method and the FFT in chapter 8 are from “Numerical Recipes in C ”,
2nd Edition, ISBN 0521431085.
The Mathematics Formulary is made with teT
E
X and L
A
T
E
X version 2.09.
If you prefer the notation in which vectors are typefaced in boldface, uncomment the redeﬁnition of the `vec
command and recompile the ﬁle.
If you ﬁnd any errors or have any comments, please let me know. I am always open for suggestions and possible
corrections to the mathematics formulary.
Johan Wevers
Contents
Contents I
1 Basics 1
1.1 Goniometric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Hyperbolic functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Complex numbers and quaternions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5.1 Complex numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5.2 Quaternions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.6 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6.1 Triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6.2 Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.7 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.8 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.8.1 Expansion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.8.2 Convergence and divergence of series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.8.3 Convergence and divergence of functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.9 Products and quotients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.10 Logarithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.11 Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.12 Primes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2 Probability and statistics 9
2.1 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Probability theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.3.2 Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.4 Regression analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Calculus 12
3.1 Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.1 Arithmetic rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.2 Arc lengts, surfaces and volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.3 Separation of quotients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.4 Special functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.1.5 Goniometric integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2 Functions with more variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2.1 Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.2.2 Taylor series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.3 Extrema . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2.4 The ∇operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2.5 Integral theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.6 Multiple integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.7 Coordinate transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Orthogonality of functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4 Fourier series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
I
II Mathematics Formulary door J.C.A. Wevers
4 Differential equations 20
4.1 Linear differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1.1 First order linear DE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1.2 Second order linear DE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.1.3 The Wronskian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.4 Power series substitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Some special cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.1 Frobenius’ method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.2 Euler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2.3 Legendre’s DE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2.4 The associated Legendre equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2.5 Solutions for Bessel’s equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.2.6 Properties of Bessel functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2.7 Laguerre’s equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2.8 The associated Laguerre equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.9 Hermite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.10 Chebyshev . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.2.11 Weber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3 Nonlinear differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4 SturmLiouville equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.5 Linear partial differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.5.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.5.2 Special cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.5.3 Potential theory and Green’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5 Linear algebra 29
5.1 Vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3 Matrix calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3.1 Basic operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.3.2 Matrix equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5.4 Linear transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.5 Plane and line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5.6 Coordinate transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.7 Eigen values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.8 Transformation types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.9 Homogeneous coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.10 Inner product spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.11 The Laplace transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5.12 The convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.13 Systems of linear differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.14 Quadratic forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.14.1 Quadratic forms in IR
2
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.14.2 Quadratic surfaces in IR
3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
6 Complex function theory 39
6.1 Functions of complex variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6.2 Complex integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6.2.1 Cauchy’s integral formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
6.2.2 Residue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
6.3 Analytical functions deﬁnied by series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.4 Laurent series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.5 Jordan’s theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Mathematics Formulary by J.C.A. Wevers III
7 Tensor calculus 43
7.1 Vectors and covectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
7.2 Tensor algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.3 Inner product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
7.4 Tensor product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
7.5 Symmetric and antisymmetric tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
7.6 Outer product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
7.7 The Hodge star operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.8 Differential operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.8.1 The directional derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.8.2 The Liederivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.8.3 Christoffel symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.8.4 The covariant derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.9 Differential operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
7.10 Differential geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.10.1 Space curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.10.2 Surfaces in IR
3
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
7.10.3 The ﬁrst fundamental tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
7.10.4 The second fundamental tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
7.10.5 Geodetic curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
7.11 Riemannian geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
8 Numerical mathematics 51
8.1 Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.2 Floating point representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.3 Systems of equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
8.3.1 Triangular matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
8.3.2 Gauss elimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
8.3.3 Pivot strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
8.4 Roots of functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
8.4.1 Successive substitution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
8.4.2 Local convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
8.4.3 Aitken extrapolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8.4.4 Newton iteration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8.4.5 The secant method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
8.5 Polynomial interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
8.6 Deﬁnite integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.7 Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.8 Differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
8.9 The fast Fourier transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
IV Mathematics Formulary door J.C.A. Wevers
Chapter 1
Basics
1.1 Goniometric functions
For the goniometric ratios for a point p on the unit circle holds:
cos(φ) = x
p
, sin(φ) = y
p
, tan(φ) =
y
p
x
p
sin
2
(x) + cos
2
(x) = 1 and cos
−2
(x) = 1 + tan
2
(x).
cos(a ±b) = cos(a) cos(b) ∓sin(a) sin(b) , sin(a ±b) = sin(a) cos(b) ±cos(a) sin(b)
tan(a ±b) =
tan(a) ±tan(b)
1 ∓tan(a) tan(b)
The sum formulas are:
sin(p) + sin(q) = 2 sin(
1
2
(p +q)) cos(
1
2
(p −q))
sin(p) −sin(q) = 2 cos(
1
2
(p +q)) sin(
1
2
(p −q))
cos(p) + cos(q) = 2 cos(
1
2
(p +q)) cos(
1
2
(p −q))
cos(p) −cos(q) = −2 sin(
1
2
(p +q)) sin(
1
2
(p −q))
From these equations can be derived that
2 cos
2
(x) = 1 + cos(2x) , 2 sin
2
(x) = 1 −cos(2x)
sin(π −x) = sin(x) , cos(π −x) = −cos(x)
sin(
1
2
π −x) = cos(x) , cos(
1
2
π −x) = sin(x)
Conclusions from equalities:
sin(x) = sin(a) ⇒ x = a ±2kπ or x = (π −a) ±2kπ, k ∈ IN
cos(x) = cos(a) ⇒ x = a ±2kπ or x = −a ±2kπ
tan(x) = tan(a) ⇒ x = a ±kπ and x =
π
2
±kπ
The following relations exist between the inverse goniometric functions:
arctan(x) = arcsin
x
√
x
2
+ 1
= arccos
1
√
x
2
+ 1
, sin(arccos(x)) =
1 −x
2
1.2 Hyperbolic functions
The hyperbolic functions are deﬁned by:
sinh(x) =
e
x
−e
−x
2
, cosh(x) =
e
x
+ e
−x
2
, tanh(x) =
sinh(x)
cosh(x)
From this follows that cosh
2
(x) −sinh
2
(x) = 1. Further holds:
arsinh(x) = ln [x +
x
2
+ 1[ , arcosh(x) = arsinh(
x
2
−1)
1
2 Mathematics Formulary by ir. J.C.A. Wevers
1.3 Calculus
The derivative of a function is deﬁned as:
df
dx
= lim
h→0
f(x +h) −f(x)
h
Derivatives obey the following algebraic rules:
d(x ±y) = dx ±dy , d(xy) = xdy +ydx , d
x
y
=
ydx −xdy
y
2
For the derivative of the inverse function f
inv
(y), deﬁned by f
inv
(f(x)) = x, holds at point P = (x, f(x)):
df
inv
(y)
dy
P
df(x)
dx
P
= 1
Chain rule: if f = f(g(x)), then holds
df
dx
=
df
dg
dg
dx
Further, for the derivatives of products of functions holds:
(f g)
(n)
=
n
¸
k=0
n
k
f
(n−k)
g
(k)
For the primitive function F(x) holds: F
(x) = f(x). An overview of derivatives and primitives is:
y = f(x) dy/dx = f
(x)
f(x)dx
ax
n
anx
n−1
a(n + 1)
−1
x
n+1
1/x −x
−2
ln [x[
a 0 ax
a
x
a
x
ln(a) a
x
/ ln(a)
e
x
e
x
e
x
a
log(x) (xln(a))
−1
(xln(x) −x)/ ln(a)
ln(x) 1/x xln(x) −x
sin(x) cos(x) −cos(x)
cos(x) −sin(x) sin(x)
tan(x) cos
−2
(x) −ln [ cos(x)[
sin
−1
(x) −sin
−2
(x) cos(x) ln [ tan(
1
2
x)[
sinh(x) cosh(x) cosh(x)
cosh(x) sinh(x) sinh(x)
arcsin(x) 1/
√
1 −x
2
xarcsin(x) +
√
1 −x
2
arccos(x) −1/
√
1 −x
2
xarccos(x) −
√
1 −x
2
arctan(x) (1 +x
2
)
−1
xarctan(x) −
1
2
ln(1 +x
2
)
(a +x
2
)
−1/2
−x(a +x
2
)
−3/2
ln [x +
√
a +x
2
[
(a
2
−x
2
)
−1
2x(a
2
+x
2
)
−2
1
2a
ln [(a +x)/(a −x)[
The curvature ρ of a curve is given by: ρ =
(1 + (y
)
2
)
3/2
[y
[
The theorem of De ’l Hˆ opital: if f(a) = 0 and g(a) = 0, then is lim
x→a
f(x)
g(x)
= lim
x→a
f
(x)
g
(x)
Chapter 1: Basics 3
1.4 Limits
lim
x→0
sin(x)
x
= 1 , lim
x→0
e
x
−1
x
= 1 , lim
x→0
tan(x)
x
= 1 , lim
k→0
(1 +k)
1/k
= e , lim
x→∞
1 +
n
x
x
= e
n
lim
x↓0
x
a
ln(x) = 0 , lim
x→∞
ln
p
(x)
x
a
= 0 , lim
x→0
ln(x +a)
x
= a , lim
x→∞
x
p
a
x
= 0 als [a[ > 1.
lim
x→0
a
1/x
−1
= ln(a) , lim
x→0
arcsin(x)
x
= 1 , lim
x→∞
x
√
x = 1
1.5 Complex numbers and quaternions
1.5.1 Complex numbers
The complex number z = a + bi with a and b ∈ IR. a is the real part, b the imaginary part of z. [z[ =
√
a
2
+b
2
.
By deﬁnition holds: i
2
= −1. Every complex number can be written as z = [z[ exp(iϕ), with tan(ϕ) = b/a. The
complex conjugate of z is deﬁned as z = z
∗
:= a −bi. Further holds:
(a +bi)(c +di) = (ac −bd) +i(ad +bc)
(a +bi) + (c +di) = a +c +i(b +d)
a +bi
c +di
=
(ac +bd) +i(bc −ad)
c
2
+d
2
Goniometric functions can be written as complex exponents:
sin(x) =
1
2i
(e
ix
−e
−ix
)
cos(x) =
1
2
(e
ix
+ e
−ix
)
From this follows that cos(ix) = cosh(x) and sin(ix) = i sinh(x). Further follows from this that
e
±ix
= cos(x) ±i sin(x), so e
iz
= 0∀z. Also the theorem of De Moivre follows from this:
(cos(ϕ) +i sin(ϕ))
n
= cos(nϕ) +i sin(nϕ).
Products and quotients of complex numbers can be written as:
z
1
z
2
= [z
1
[ [z
2
[(cos(ϕ
1
+ϕ
2
) +i sin(ϕ
1
+ϕ
2
))
z
1
z
2
=
[z
1
[
[z
2
[
(cos(ϕ
1
−ϕ
2
) +i sin(ϕ
1
−ϕ
2
))
The following can be derived:
[z
1
+z
2
[ ≤ [z
1
[ +[z
2
[ , [z
1
−z
2
[ ≥ [ [z
1
[ −[z
2
[ [
And from z = r exp(iθ) follows: ln(z) = ln(r) +iθ, ln(z) = ln(z) ±2nπi.
1.5.2 Quaternions
Quaternions are deﬁned as: z = a + bi + cj + dk, with a, b, c, d ∈ IR and i
2
= j
2
= k
2
= −1. The products of
i, j, k with each other are given by ij = −ji = k, jk = −kj = i and ki = −ik = j.
4 Mathematics Formulary by ir. J.C.A. Wevers
1.6 Geometry
1.6.1 Triangles
The sine rule is:
a
sin(α)
=
b
sin(β)
=
c
sin(γ)
Here, α is the angle opposite to a, β is opposite to b and γ opposite to c. The cosine rule is: a
2
= b
2
+c
2
−2bc cos(α).
For each triangle holds: α +β +γ = 180
◦
.
Further holds:
tan(
1
2
(α +β))
tan(
1
2
(α −β))
=
a +b
a −b
The surface of a triangle is given by
1
2
ab sin(γ) =
1
2
ah
a
=
s(s −a)(s −b)(s −c) with h
a
the perpendicular on
a and s =
1
2
(a +b +c).
1.6.2 Curves
Cycloid: if a circle with radius a rolls along a straight line, the trajectory of a point on this circle has the following
parameter equation:
x = a(t + sin(t)) , y = a(1 + cos(t))
Epicycloid: if a small circle with radius a rolls along a big circle with radius R, the trajectory of a point on the small
circle has the following parameter equation:
x = a sin
R +a
a
t
+ (R +a) sin(t) , y = a cos
R +a
a
t
+ (R +a) cos(t)
Hypocycloid: if a small circle with radius a rolls inside a big circle with radius R, the trajectory of a point on the
small circle has the following parameter equation:
x = a sin
R −a
a
t
+ (R −a) sin(t) , y = −a cos
R −a
a
t
+ (R −a) cos(t)
A hypocycloid with a = R is called a cardioid. It has the following parameterequation in polar coordinates:
r = 2a[1 −cos(ϕ)].
1.7 Vectors
The inner product is deﬁned by: a
b =
¸
i
a
i
b
i
= [a [ [
b [ cos(ϕ)
where ϕ is the angle between a and
b. The external product is in IR
3
deﬁned by:
a
b =
¸
a
y
b
z
−a
z
b
y
a
z
b
x
−a
x
b
z
a
x
b
y
−a
y
b
x
¸
=
e
x
e
y
e
z
a
x
a
y
a
z
b
x
b
y
b
z
Further holds: [a
b [ = [a [ [
b [ sin(ϕ), and a (
b c ) = (a c )
b −(a
b )c.
Chapter 1: Basics 5
1.8 Series
1.8.1 Expansion
The Binomium of Newton is:
(a +b)
n
=
n
¸
k=0
n
k
a
n−k
b
k
where
n
k
:=
n!
k!(n −k)!
.
By subtracting the series
n
¸
k=0
r
k
and r
n
¸
k=0
r
k
one ﬁnds:
n
¸
k=0
r
k
=
1 −r
n+1
1 −r
and for [r[ < 1 this gives the geometric series:
∞
¸
k=0
r
k
=
1
1 −r
.
The arithmetic series is given by:
N
¸
n=0
(a +nV ) = a(N + 1) +
1
2
N(N + 1)V .
The expansion of a function around the point a is given by the Taylor series:
f(x) = f(a) + (x −a)f
(a) +
(x −a)
2
2
f
(a) + +
(x −a)
n
n!
f
(n)
(a) +R
where the remainder is given by:
R
n
(h) = (1 −θ)
n
h
n
n!
f
(n+1)
(θh)
and is subject to:
mh
n+1
(n + 1)!
≤ R
n
(h) ≤
Mh
n+1
(n + 1)!
From this one can deduce that
(1 −x)
α
=
∞
¸
n=0
α
n
x
n
One can derive that:
∞
¸
n=1
1
n
2
=
π
2
6
,
∞
¸
n=1
1
n
4
=
π
4
90
,
∞
¸
n=1
1
n
6
=
π
6
945
n
¸
k=1
k
2
=
1
6
n(n + 1)(2n + 1) ,
∞
¸
n=1
(−1)
n+1
n
2
=
π
2
12
,
∞
¸
n=1
(−1)
n+1
n
= ln(2)
∞
¸
n=1
1
4n
2
−1
=
1
2
,
∞
¸
n=1
1
(2n −1)
2
=
π
2
8
,
∞
¸
n=1
1
(2n −1)
4
=
π
4
96
,
∞
¸
n=1
(−1)
n+1
(2n −1)
3
=
π
3
32
1.8.2 Convergence and divergence of series
If
¸
n
[u
n
[ converges,
¸
n
u
n
also converges.
If lim
n→∞
u
n
= 0 then
¸
n
u
n
is divergent.
An alternating series of which the absolute values of the terms drop monotonously to 0 is convergent (Leibniz).
6 Mathematics Formulary by ir. J.C.A. Wevers
If
∞
p
f(x)dx < ∞, then
¸
n
f
n
is convergent.
If u
n
> 0 ∀n then is
¸
n
u
n
convergent if
¸
n
ln(u
n
+ 1) is convergent.
If u
n
= c
n
x
n
the radius of convergence ρ of
¸
n
u
n
is given by:
1
ρ
= lim
n→∞
n
[c
n
[ = lim
n→∞
c
n+1
c
n
.
The series
∞
¸
n=1
1
n
p
is convergent if p > 1 and divergent if p ≤ 1.
If: lim
n→∞
u
n
v
n
= p, than the following is true: if p > 0 than
¸
n
u
n
and
¸
n
v
n
are both divergent or both convergent, if
p = 0 holds: if
¸
n
v
n
is convergent, than
¸
n
u
n
is also convergent.
If L is deﬁned by: L = lim
n→∞
n
[n
n
[, or by: L = lim
n→∞
u
n+1
u
n
, then is
¸
n
u
n
divergent if L > 1 and convergent if
L < 1.
1.8.3 Convergence and divergence of functions
f(x) is continuous in x = a only if the upper  and lower limit are equal: lim
x↑a
f(x) = lim
x↓a
f(x). This is written as:
f(a
−
) = f(a
+
).
If f(x) is continuous in a and: lim
x↑a
f
(x) = lim
x↓a
f
(x), than f(x) is differentiable in x = a.
We deﬁne: f
W
:= sup([f(x)[ [x ∈ W), and lim
x→∞
f
n
(x) = f(x). Than holds: ¦f
n
¦ is uniform convergent if
lim
n→∞
f
n
−f = 0, or: ∀(ε > 0)∃(N)∀(n ≥ N)f
n
−f < ε.
Weierstrass’ test: if
¸
u
n

W
is convergent, than
¸
u
n
is uniform convergent.
We deﬁne S(x) =
∞
¸
n=N
u
n
(x) and F(y) =
b
a
f(x, y)dx := F. Than it can be proved that:
Theorem For Demands on W Than holds on W
rows f
n
continuous, f is continuous
¦f
n
¦ uniform convergent
C series S(x) uniform convergent, S is continuous
u
n
continuous
integral f is continuous F is continuous
rows f
n
can be integrated, f
n
can be integrated,
¦f
n
¦ uniform convergent
f(x)dx = lim
n→∞
f
n
dx
I series S(x) is uniform convergent, S can be integrated,
Sdx =
¸
u
n
dx
u
n
can be integrated
integral f is continuous
Fdy =
f(x, y)dxdy
rows ¦f
n
¦ ∈C
−1
; ¦f
n
¦ unif.conv → φ f
= φ(x)
D series u
n
∈C
−1
;
¸
u
n
conv;
¸
u
n
u.c. S
(x) =
¸
u
n
(x)
integral ∂f/∂y continuous F
y
=
f
y
(x, y)dx
Chapter 1: Basics 7
1.9 Products and quotients
For a, b, c, d ∈ IR holds:
The distributive property: (a +b)(c +d) = ac +ad +bc +bd
The associative property: a(bc) = b(ac) = c(ab) and a(b +c) = ab +ac
The commutative property: a +b = b +a, ab = ba.
Further holds:
a
2n
−b
2n
a ±b
= a
2n−1
±a
2n−2
b +a
2n−3
b
2
± ±b
2n−1
,
a
2n+1
−b
2n+1
a +b
=
n
¸
k=0
a
2n−k
b
2k
(a ±b)(a
2
±ab +b
2
) = a
3
±b
3
, (a +b)(a −b) = a
2
+b
2
,
a
3
±b
3
a +b
= a
2
∓ba +b
2
1.10 Logarithms
Deﬁnition:
a
log(x) = b ⇔ a
b
= x. For logarithms with base e one writes ln(x).
Rules: log(x
n
) = nlog(x), log(a) + log(b) = log(ab), log(a) −log(b) = log(a/b).
1.11 Polynomials
Equations of the type
n
¸
k=0
a
k
x
k
= 0
have n roots which may be equal to each other. Each polynomial p(z) of order n ≥ 1 has at least one root in C. If
all a
k
∈ IR holds: when x = p with p ∈ C a root, than p
∗
is also a root. Polynomials up to and including order 4
have a general analytical solution, for polynomials with order ≥ 5 there does not exist a general analytical solution.
For a, b, c ∈ IR and a = 0 holds: the 2nd order equation ax
2
+bx +c = 0 has the general solution:
x =
−b ±
√
b
2
−4ac
2a
For a, b, c, d ∈ IR and a = 0 holds: the 3rd order equation ax
3
+ bx
2
+ cx + d = 0 has the general analytical
solution:
x
1
= K −
3ac −b
2
9a
2
K
−
b
3a
x
2
= x
∗
3
= −
K
2
+
3ac −b
2
18a
2
K
−
b
3a
+i
√
3
2
K +
3ac −b
2
9a
2
K
with K =
9abc −27da
2
−2b
3
54a
3
+
√
3
√
4ac
3
−c
2
b
2
−18abcd + 27a
2
d
2
+ 4db
3
18a
2
1/3
1.12 Primes
A prime is a number ∈ IN that can only be divided by itself and 1. There are an inﬁnite number of primes. Proof:
suppose that the collection of primes P would be ﬁnite, than construct the number q = 1 +
¸
p∈P
p, than holds
q = 1(p) and so Q cannot be written as a product of primes from P. This is a contradiction.
8 Mathematics Formulary by ir. J.C.A. Wevers
If π(x) is the number of primes ≤ x, than holds:
lim
x→∞
π(x)
x/ ln(x)
= 1 and lim
x→∞
π(x)
x
2
dt
ln(t)
= 1
For each N ≥ 2 there is a prime between N and 2N.
The numbers F
k
:= 2
k
+ 1 with k ∈ IN are called Fermat numbers. Many Fermat numbers are prime.
The numbers M
k
:= 2
k
− 1 are called Mersenne numbers. They occur when one searches for perfect numbers,
which are numbers n ∈ IN which are the sum of their different dividers, for example 6 = 1 + 2 + 3. There
are 23 Mersenne numbers for k < 12000 which are prime: for k ∈ ¦2, 3, 5, 7, 13, 17, 19, 31, 61, 89, 107, 127, 521,
607, 1279, 2203, 2281, 3217, 4253, 4423, 9689, 9941, 11213¦.
To check if a given number n is prime one can use a sieve method. The ﬁrst known sieve method was developed by
Eratosthenes. A faster method for large numbers are the 4 Fermat tests, who don’t prove that a number is prime but
give a large probability.
1. Take the ﬁrst 4 primes: b = ¦2, 3, 5, 7¦,
2. Take w(b) = b
n−1
mod n, for each b,
3. If w = 1 for each b, then n is probably prime. For each other value of w, n is certainly not prime.
Chapter 2
Probability and statistics
2.1 Combinations
The number of possible combinations of k elements from n elements is given by
n
k
=
n!
k!(n −k)!
The number of permutations of p from n is given by
n!
(n −p)!
= p!
n
p
The number of different ways to classify n
i
elements in i groups, when the total number of elements is N, is
N!
¸
i
n
i
!
2.2 Probability theory
The probability P(A) that an event A occurs is deﬁned by:
P(A) =
n(A)
n(U)
where n(A) is the number of events when A occurs and n(U) the total number of events.
The probability P(A) that A does not occur is: P(A) = 1 − P(A). The probability P(A ∪ B) that A and
B both occur is given by: P(A ∪ B) = P(A) + P(B) − P(A ∩ B). If A and B are independent, than holds:
P(A∩ B) = P(A) P(B).
The probability P(A[B) that A occurs, given the fact that B occurs, is:
P(A[B) =
P(A∩ B)
P(B)
2.3 Statistics
2.3.1 General
The average or mean value 'x` of a collection of values is: 'x` =
¸
i
x
i
/n. The standard deviation σ
x
in the
distribution of x is given by:
σ
x
=
n
¸
i=1
(x
i
−'x`)
2
n
When samples are being used the sample variance s is given by s
2
=
n
n −1
σ
2
.
9
10 Mathematics Formulary by ir. J.C.A. Wevers
The covariance σ
xy
of x and y is given by::
σ
xy
=
n
¸
i=1
(x
i
−'x`)(y
i
−'y`)
n −1
The correlation coefﬁcient r
xy
of x and y than becomes: r
xy
= σ
xy
/σ
x
σ
y
.
The standard deviation in a variable f(x, y) resulting from errors in x and y is:
σ
2
f(x,y)
=
∂f
∂x
σ
x
2
+
∂f
∂y
σ
y
2
+
∂f
∂x
∂f
∂y
σ
xy
2.3.2 Distributions
1. The Binomial distribution is the distribution describing a sampling with replacement. The probability for
success is p. The probability P for k successes in n trials is then given by:
P(x = k) =
n
k
p
k
(1 −p)
n−k
The standard deviation is given by σ
x
=
np(1 −p) and the expectation value is ε = np.
2. The Hypergeometric distribution is the distribution describing a sampling without replacement in which the
order is irrelevant. The probability for k successes in a trial with A possible successes and B possible failures
is then given by:
P(x = k) =
A
k
B
n −k
A+B
n
The expectation value is given by ε = nA/(A+B).
3. The Poisson distribution is a limiting case of the binomial distribution when p → 0, n → ∞ and also
np = λ is constant.
P(x) =
λ
x
e
−λ
x!
This distribution is normalized to
∞
¸
x=0
P(x) = 1.
4. The Normal distribution is a limiting case of the binomial distribution for continuous variables:
P(x) =
1
σ
√
2π
exp
−
1
2
x −'x`
σ
2
5. The Uniform distribution occurs when a random number x is taken from the set a ≤ x ≤ b and is given by:
P(x) =
1
b −a
if a ≤ x ≤ b
P(x) = 0 in all other cases
'x` =
1
2
(b −a) and σ
2
=
(b −a)
2
12
.
Chapter 2: Probability and statistics 11
6. The Gamma distribution is given by:
P(x) =
x
α−1
e
−x/β
β
α
Γ(α)
if 0 ≤ y ≤ ∞
with α > 0 and β > 0. The distribution has the following properties: 'x` = αβ, σ
2
= αβ
2
.
7. The Beta distribution is given by:
P(x) =
x
α−1
(1 −x)
β−1
β(α, β)
if 0 ≤ x ≤ 1
P(x) = 0 everywhere else
and has the following properties: 'x` =
α
α +β
, σ
2
=
αβ
(α +β)
2
(α +β + 1)
.
For P(χ
2
) holds: α = V/2 and β = 2.
8. The Weibull distribution is given by:
P(x) =
α
β
x
α−1
e
−x
α
if 0 ≤ x ≤ ∞∧ α ∧ β > 0
P(x) = 0 in all other cases
The average is 'x` = β
1/α
Γ((α + 1)α)
9. For a twodimensional distribution holds:
P
1
(x
1
) =
P(x
1
, x
2
)dx
2
, P
2
(x
2
) =
P(x
1
, x
2
)dx
1
with
ε(g(x
1
, x
2
)) =
g(x
1
, x
2
)P(x
1
, x
2
)dx
1
dx
2
=
¸
x
1
¸
x
2
g P
2.4 Regression analyses
When there exists a relation between the quantities x and y of the form y = ax + b and there is a measured set x
i
with related y
i
, the following relation holds for a and b with x = (x
1
, x
2
, ..., x
n
) and e = (1, 1, ..., 1):
y −ax −be ∈< x, e >
⊥
From this follows that the inner products are 0:
(y, x) −a(x, x) −b(e, x) = 0
(y, e ) −a(x, e ) −b(e, e ) = 0
with (x, x) =
¸
i
x
2
i
, (x, y ) =
¸
i
x
i
y
i
, (x, e ) =
¸
i
x
i
and (e, e ) = n. a and b follow from this.
A similar method works for higher order polynomial ﬁts: for a second order ﬁt holds:
y −a
x
2
−bx −ce ∈<
x
2
, x, e >
⊥
with
x
2
= (x
2
1
, ..., x
2
n
).
The correlation coefﬁcient r is a measure for the quality of a ﬁt. In case of linear regression it is given by:
r =
n
¸
xy −
¸
x
¸
y
(n
¸
x
2
−(
¸
x)
2
)(n
¸
y
2
−(
¸
y)
2
)
Chapter 3
Calculus
3.1 Integrals
3.1.1 Arithmetic rules
The primitive function F(x) of f(x) obeys the rule F
(x) = f(x). With F(x) the primitive of f(x) holds for the
deﬁnite integral
b
a
f(x)dx = F(b) −F(a)
If u = f(x) holds:
b
a
g(f(x))df(x) =
f(b)
f(a)
g(u)du
Partial integration: with F and G the primitives of f and g holds:
f(x) g(x)dx = f(x)G(x) −
G(x)
df(x)
dx
dx
A derivative can be brought under the intergral sign (see section 1.8.3 for the required conditions):
d
dy
x=h(y)
x=g(y)
f(x, y)dx
¸
¸
¸ =
x=h(y)
x=g(y)
∂f(x, y)
∂y
dx −f(g(y), y)
dg(y)
dy
+f(h(y), y)
dh(y)
dy
3.1.2 Arc lengts, surfaces and volumes
The arc length of a curve y(x) is given by:
=
1 +
dy(x)
dx
2
dx
The arc length of a parameter curve F(x(t)) is:
=
Fds =
F(x(t))[
˙
x(t)[dt
with
t =
dx
ds
=
˙
x(t)
[
˙
x(t)[
, [
t [ = 1
(v,
t)ds =
(v,
˙
t(t))dt =
(v
1
dx +v
2
dy +v
3
dz)
The surface A of a solid of revolution is:
A = 2π
y
1 +
dy(x)
dx
2
dx
12
Chapter 3: Calculus 13
The volume V of a solid of revolution is:
V = π
f
2
(x)dx
3.1.3 Separation of quotients
Every rational function P(x)/Q(x) where P and Q are polynomials can be written as a linear combination of
functions of the type (x −a)
k
with k ∈ ZZ, and of functions of the type
px +q
((x −a)
2
+b
2
)
n
with b > 0 and n ∈ IN. So:
p(x)
(x −a)
n
=
n
¸
k=1
A
k
(x −a)
k
,
p(x)
((x −b)
2
+c
2
)
n
=
n
¸
k=1
A
k
x +B
((x −b)
2
+c
2
)
k
Recurrent relation: for n = 0 holds:
dx
(x
2
+ 1)
n+1
=
1
2n
x
(x
2
+ 1)
n
+
2n −1
2n
dx
(x
2
+ 1)
n
3.1.4 Special functions
Elliptic functions
Elliptic functions can be written as a power series as follows:
1 −k
2
sin
2
(x) = 1 −
∞
¸
n=1
(2n −1)!!
(2n)!!(2n −1)
k
2n
sin
2n
(x)
1
1 −k
2
sin
2
(x)
= 1 +
∞
¸
n=1
(2n −1)!!
(2n)!!
k
2n
sin
2n
(x)
with n!! = n(n −2)!!.
The Gamma function
The gamma function Γ(y) is deﬁned by:
Γ(y) =
∞
0
e
−x
x
y−1
dx
One can derive that Γ(y + 1) = yΓ(y) = y!. This is a way to deﬁne faculties for nonintegers. Further one can
derive that
Γ(n +
1
2
) =
√
π
2
n
(2n −1)!! and Γ
(n)
(y) =
∞
0
e
−x
x
y−1
ln
n
(x)dx
The Beta function
The betafunction β(p, q) is deﬁned by:
β(p, q) =
1
0
x
p−1
(1 −x)
q−1
dx
with p and q > 0. The beta and gamma functions are related by the following equation:
β(p, q) =
Γ(p)Γ(q)
Γ(p +q)
14 Mathematics Formulary by ir. J.C.A. Wevers
The Delta function
The delta function δ(x) is an inﬁnitely thin peak function with surface 1. It can be deﬁned by:
δ(x) = lim
ε→0
P(ε, x) with P(ε, x) =
0 for [x[ > ε
1
2ε
when [x[ < ε
Some properties are:
∞
−∞
δ(x)dx = 1 ,
∞
−∞
F(x)δ(x)dx = F(0)
3.1.5 Goniometric integrals
When solving goniometric integrals it can be useful to change variables. The following holds if one deﬁnes
tan(
1
2
x) := t:
dx =
2dt
1 +t
2
, cos(x) =
1 −t
2
1 +t
2
, sin(x) =
2t
1 +t
2
Each integral of the type
R(x,
√
ax
2
+bx +c)dx can be converted into one of the types that were treated in
section 3.1.3. After this conversion one can substitute in the integrals of the type:
R(x,
x
2
+ 1)dx : x = tan(ϕ) , dx =
dϕ
cos(ϕ)
of
x
2
+ 1 = t +x
R(x,
1 −x
2
)dx : x = sin(ϕ) , dx = cos(ϕ)dϕ of
1 −x
2
= 1 −tx
R(x,
x
2
−1)dx : x =
1
cos(ϕ)
, dx =
sin(ϕ)
cos
2
(ϕ)
dϕ of
x
2
−1 = x −t
These deﬁnite integrals are easily solved:
π/2
0
cos
n
(x) sin
m
(x)dx =
(n −1)!!(m−1)!!
(m+n)!!
π/2 when m and n are both even
1 in all other cases
Some important integrals are:
∞
0
xdx
e
ax
+ 1
=
π
2
12a
2
,
∞
−∞
x
2
dx
(e
x
+ 1)
2
=
π
2
3
,
∞
0
x
3
dx
e
x
+ 1
=
π
4
15
3.2 Functions with more variables
3.2.1 Derivatives
The partial derivative with respect to x of a function f(x, y) is deﬁned by:
∂f
∂x
x
0
= lim
h→0
f(x
0
+h, y
0
) −f(x
0
, y
0
)
h
The directional derivative in the direction of α is deﬁned by:
∂f
∂α
= lim
r↓0
f(x
0
+r cos(α), y
0
+r sin(α)) −f(x
0
, y
0
)
r
= (
∇f, (sin α, cos α)) =
∇f v
[v[
Chapter 3: Calculus 15
When one changes to coordinates f(x(u, v), y(u, v)) holds:
∂f
∂u
=
∂f
∂x
∂x
∂u
+
∂f
∂y
∂y
∂u
If x(t) and y(t) depend only on one parameter t holds:
∂f
∂t
=
∂f
∂x
dx
dt
+
∂f
∂y
dy
dt
The total differential df of a function of 3 variables is given by:
df =
∂f
∂x
dx +
∂f
∂y
dy +
∂f
∂z
dz
So
df
dx
=
∂f
∂x
+
∂f
∂y
dy
dx
+
∂f
∂z
dz
dx
The tangent in point x
0
at the surface f(x, y) = 0 is given by the equation f
x
(x
0
)(x −x
0
) +f
y
(x
0
)(y −y
0
) = 0.
The tangent plane in x
0
is given by: f
x
(x
0
)(x −x
0
) +f
y
(x
0
)(y −y
0
) = z −f(x
0
).
3.2.2 Taylor series
A function of two variables can be expanded as follows in a Taylor series:
f(x
0
+h, y
0
+k) =
n
¸
p=0
1
p!
h
∂
p
∂x
p
+k
∂
p
∂y
p
f(x
0
, y
0
) +R(n)
with R(n) the residual error and
h
∂
p
∂x
p
+k
∂
p
∂y
p
f(a, b) =
p
¸
m=0
p
m
h
m
k
p−m
∂
p
f(a, b)
∂x
m
∂y
p−m
3.2.3 Extrema
When f is continuous on a compact boundary V there exists a global maximum and a global minumum for f on
this boundary. A boundary is called compact if it is limited and closed.
Possible extrema of f(x, y) on a boundary V ∈ IR
2
are:
1. Points on V where f(x, y) is not differentiable,
2. Points where
∇f =
0,
3. If the boundary V is given by ϕ(x, y) = 0, than all points where
∇f(x, y) +λ
∇ϕ(x, y) = 0 are possible for
extrema. This is the multiplicator method of Lagrange, λ is called a multiplicator.
The same as in IR
2
holds in IR
3
when the area to be searched is constrained by a compact V , and V is deﬁned by
ϕ
1
(x, y, z) = 0 and ϕ
2
(x, y, z) = 0 for extrema of f(x, y, z) for points (1) and (2). Point (3) is rewritten as follows:
possible extrema are points where
∇f(x, y, z) +λ
1
∇ϕ
1
(x, y, z) +λ
2
∇ϕ
2
(x, y, z) = 0.
16 Mathematics Formulary by ir. J.C.A. Wevers
3.2.4 The ∇operator
In cartesian coordinates (x, y, z) holds:
∇ =
∂
∂x
e
x
+
∂
∂y
e
y
+
∂
∂z
e
z
gradf =
∂f
∂x
e
x
+
∂f
∂y
e
y
+
∂f
∂z
e
z
div a =
∂a
x
∂x
+
∂a
y
∂y
+
∂a
z
∂z
curl a =
∂a
z
∂y
−
∂a
y
∂z
e
x
+
∂a
x
∂z
−
∂a
z
∂x
e
y
+
∂a
y
∂x
−
∂a
x
∂y
e
z
∇
2
f =
∂
2
f
∂x
2
+
∂
2
f
∂y
2
+
∂
2
f
∂z
2
In cylindrical coordinates (r, ϕ, z) holds:
∇ =
∂
∂r
e
r
+
1
r
∂
∂ϕ
e
ϕ
+
∂
∂z
e
z
gradf =
∂f
∂r
e
r
+
1
r
∂f
∂ϕ
e
ϕ
+
∂f
∂z
e
z
div a =
∂a
r
∂r
+
a
r
r
+
1
r
∂a
ϕ
∂ϕ
+
∂a
z
∂z
curl a =
1
r
∂a
z
∂ϕ
−
∂a
ϕ
∂z
e
r
+
∂a
r
∂z
−
∂a
z
∂r
e
ϕ
+
∂a
ϕ
∂r
+
a
ϕ
r
−
1
r
∂a
r
∂ϕ
e
z
∇
2
f =
∂
2
f
∂r
2
+
1
r
∂f
∂r
+
1
r
2
∂
2
f
∂ϕ
2
+
∂
2
f
∂z
2
In spherical coordinates (r, θ, ϕ) holds:
∇ =
∂
∂r
e
r
+
1
r
∂
∂θ
e
θ
+
1
r sin θ
∂
∂ϕ
e
ϕ
gradf =
∂f
∂r
e
r
+
1
r
∂f
∂θ
e
θ
+
1
r sin θ
∂f
∂ϕ
e
ϕ
div a =
∂a
r
∂r
+
2a
r
r
+
1
r
∂a
θ
∂θ
+
a
θ
r tan θ
+
1
r sin θ
∂a
ϕ
∂ϕ
curl a =
1
r
∂a
ϕ
∂θ
+
a
θ
r tan θ
−
1
r sin θ
∂a
θ
∂ϕ
e
r
+
1
r sin θ
∂a
r
∂ϕ
−
∂a
ϕ
∂r
−
a
ϕ
r
e
θ
+
∂a
θ
∂r
+
a
θ
r
−
1
r
∂a
r
∂θ
e
ϕ
∇
2
f =
∂
2
f
∂r
2
+
2
r
∂f
∂r
+
1
r
2
∂
2
f
∂θ
2
+
1
r
2
tan θ
∂f
∂θ
+
1
r
2
sin
2
θ
∂
2
f
∂ϕ
2
General orthonormal curvilinear coordinates (u, v, w) can be derived from cartesian coordinates by the transforma
tion x = x(u, v, w). The unit vectors are given by:
e
u
=
1
h
1
∂x
∂u
, e
v
=
1
h
2
∂x
∂v
, e
w
=
1
h
3
∂x
∂w
where the terms h
i
give normalization to length 1. The differential operators are than given by:
gradf =
1
h
1
∂f
∂u
e
u
+
1
h
2
∂f
∂v
e
v
+
1
h
3
∂f
∂w
e
w
Chapter 3: Calculus 17
div a =
1
h
1
h
2
h
3
∂
∂u
(h
2
h
3
a
u
) +
∂
∂v
(h
3
h
1
a
v
) +
∂
∂w
(h
1
h
2
a
w
)
curl a =
1
h
2
h
3
∂(h
3
a
w
)
∂v
−
∂(h
2
a
v
)
∂w
e
u
+
1
h
3
h
1
∂(h
1
a
u
)
∂w
−
∂(h
3
a
w
)
∂u
e
v
+
1
h
1
h
2
∂(h
2
a
v
)
∂u
−
∂(h
1
a
u
)
∂v
e
w
∇
2
f =
1
h
1
h
2
h
3
¸
∂
∂u
h
2
h
3
h
1
∂f
∂u
+
∂
∂v
h
3
h
1
h
2
∂f
∂v
+
∂
∂w
h
1
h
2
h
3
∂f
∂w
Some properties of the ∇operator are:
div(φv) = φdivv + gradφ v curl(φv) = φcurlv + (gradφ) v curl gradφ =
0
div(u v) = v (curlu) −u (curlv) curl curlv = grad divv −∇
2
v div curlv = 0
div gradφ = ∇
2
φ ∇
2
v ≡ (∇
2
v
1
, ∇
2
v
2
, ∇
2
v
3
)
Here, v is an arbitrary vectorﬁeld and φ an arbitrary scalar ﬁeld.
3.2.5 Integral theorems
Some important integral theorems are:
Gauss:
((v n)d
2
A =
(divv )d
3
V
Stokes for a scalar ﬁeld:
(φ e
t
)ds =
(n gradφ)d
2
A
Stokes for a vector ﬁeld:
(v e
t
)ds =
(curlv n)d
2
A
this gives:
((curlv n)d
2
A = 0
Ostrogradsky:
((n v )d
2
A =
(curlv )d
3
A
((φn)d
2
A =
(gradφ)d
3
V
Here the orientable surface
d
2
A is bounded by the Jordan curve s(t).
3.2.6 Multiple integrals
Let A be a closed curve given by f(x, y) = 0, than the surface A inside the curve in IR
2
is given by
A =
d
2
A =
dxdy
Let the surface A be deﬁned by the function z = f(x, y). The volume V bounded by A and the xy plane is than
given by:
V =
f(x, y)dxdy
The volume inside a closed surface deﬁned by z = f(x, y) is given by:
V =
d
3
V =
f(x, y)dxdy =
dxdydz
18 Mathematics Formulary by ir. J.C.A. Wevers
3.2.7 Coordinate transformations
The expressions d
2
A and d
3
V transform as follows when one changes coordinates to u = (u, v, w) through the
transformation x(u, v, w):
V =
f(x, y, z)dxdydz =
f(x(u))
∂x
∂u
dudvdw
In IR
2
holds:
∂x
∂u
=
x
u
x
v
y
u
y
v
Let the surface A be deﬁned by z = F(x, y) = X(u, v). Than the volume bounded by the xy plane and F is given
by:
S
f(x)d
2
A =
G
f(x(u))
∂X
∂u
∂X
∂v
dudv =
G
f(x, y, F(x, y))
1 +∂
x
F
2
+∂
y
F
2
dxdy
3.3 Orthogonality of functions
The inner product of two functions f(x) and g(x) on the interval [a, b] is given by:
(f, g) =
b
a
f(x)g(x)dx
or, when using a weight function p(x), by:
(f, g) =
b
a
p(x)f(x)g(x)dx
The norm f follows from: f
2
= (f, f). A set functions f
i
is orthonormal if (f
i
, f
j
) = δ
ij
.
Each function f(x) can be written as a sum of orthogonal functions:
f(x) =
∞
¸
i=0
c
i
g
i
(x)
and
¸
c
2
i
≤ f
2
. Let the set g
i
be orthogonal, than it follows:
c
i
=
f, g
i
(g
i
, g
i
)
3.4 Fourier series
Each function can be written as a sum of independent base functions. When one chooses the orthogonal basis
(cos(nx), sin(nx)) we have a Fourier series.
A periodical function f(x) with period 2L can be written as:
f(x) = a
0
+
∞
¸
n=1
a
n
cos
nπx
L
+b
n
sin
nπx
L
Due to the orthogonality follows for the coefﬁcients:
a
0
=
1
2L
L
−L
f(t)dt , a
n
=
1
L
L
−L
f(t) cos
nπt
L
dt , b
n
=
1
L
L
−L
f(t) sin
nπt
L
dt
Chapter 3: Calculus 19
A Fourier series can also be written as a sum of complex exponents:
f(x) =
∞
¸
n=−∞
c
n
e
inx
with
c
n
=
1
2π
π
−π
f(x)e
−inx
dx
The Fourier transform of a function f(x) gives the transformed function
ˆ
f(ω):
ˆ
f(ω) =
1
√
2π
∞
−∞
f(x)e
−iωx
dx
The inverse transformation is given by:
1
2
f(x
+
) +f(x
−
)
=
1
√
2π
∞
−∞
ˆ
f(ω)e
iωx
dω
where f(x
+
) and f(x
−
) are deﬁned by the lower  and upper limit:
f(a
−
) = lim
x↑a
f(x) , f(a
+
) = lim
x↓a
f(x)
For continuous functions is
1
2
[f(x
+
) +f(x
−
)] = f(x).
Chapter 4
Differential equations
4.1 Linear differential equations
4.1.1 First order linear DE
The general solution of a linear differential equation is given by y
A
= y
H
+ y
P
, where y
H
is the solution of the
homogeneous equation and y
P
is a particular solution.
A ﬁrst order differential equation is given by: y
(x) + a(x)y(x) = b(x). Its homogeneous equation is y
(x) +
a(x)y(x) = 0.
The solution of the homogeneous equation is given by
y
H
= k exp
a(x)dx
Suppose that a(x) = a =constant.
Substitution of exp(λx) in the homogeneous equation leads to the characteristic equation λ +a = 0
⇒ λ = −a.
Suppose b(x) = αexp(µx). Than one can distinguish two cases:
1. λ = µ: a particular solution is: y
P
= exp(µx)
2. λ = µ: a particular solution is: y
P
= xexp(µx)
When a DE is solved by variation of parameters one writes: y
P
(x) = y
H
(x)f(x), and than one solves f(x) from
this.
4.1.2 Second order linear DE
A differential equation of the second order with constant coefﬁcients is given by: y
(x) + ay
(x) + by(x) = c(x).
If c(x) = c =constant there exists a particular solution y
P
= c/b.
Substitution of y = exp(λx) leads to the characteristic equation λ
2
+aλ +b = 0.
There are now 2 possibilities:
1. λ
1
= λ
2
: than y
H
= αexp(λ
1
x) +β exp(λ
2
x).
2. λ
1
= λ
2
= λ: than y
H
= (α +βx) exp(λx).
If c(x) = p(x) exp(µx) where p(x) is a polynomial there are 3 possibilities:
1. λ
1
, λ
2
= µ: y
P
= q(x) exp(µx).
2. λ
1
= µ, λ
2
= µ: y
P
= xq(x) exp(µx).
3. λ
1
= λ
2
= µ: y
P
= x
2
q(x) exp(µx).
where q(x) is a polynomial of the same order as p(x).
When: y
(x) +ω
2
y(x) = ωf(x) and y(0) = y
(0) = 0 follows: y(x) =
x
0
f(x) sin(ω(x −t))dt.
20
Chapter 4: Diﬀerential equations 21
4.1.3 The Wronskian
We start with the LDE y
(x) +p(x)y
(x) +q(x)y(x) = 0 and the two initial conditions y(x
0
) = K
0
and y
(x
0
) =
K
1
. When p(x) and q(x) are continuous on the open interval I there exists a unique solution y(x) on this interval.
The general solution can than be written as y(x) = c
1
y
1
(x) +c
2
y
2
(x) and y
1
and y
2
are linear independent. These
are also all solutions of the LDE.
The Wronskian is deﬁned by:
W(y
1
, y
2
) =
y
1
y
2
y
1
y
2
= y
1
y
2
−y
2
y
1
y
1
and y
2
are linear independent if and only if on the interval I when ∃x
0
∈ I so that holds:
W(y
1
(x
0
), y
2
(x
0
)) = 0.
4.1.4 Power series substitution
When a series y =
¸
a
n
x
n
is substituted in the LDE with constant coefﬁcients y
(x) + py
(x) + qy(x) = 0 this
leads to:
¸
n
n(n −1)a
n
x
n−2
+pna
n
x
n−1
+qa
n
x
n
= 0
Setting coefﬁcients for equal powers of x equal gives:
(n + 2)(n + 1)a
n+2
+p(n + 1)a
n+1
+qa
n
= 0
This gives a general relation between the coefﬁcients. Special cases are n = 0, 1, 2.
4.2 Some special cases
4.2.1 Frobenius’ method
Given the LDE
d
2
y(x)
dx
2
+
b(x)
x
dy(x)
dx
+
c(x)
x
2
y(x) = 0
with b(x) and c(x) analytical at x = 0. This LDE has at least one solution of the form
y
i
(x) = x
r
i
∞
¸
n=0
a
n
x
n
with i = 1, 2
with r real or complex and chosen so that a
0
= 0. When one expands b(x) and c(x) as b(x) = b
0
+b
1
x+b
2
x
2
+...
and c(x) = c
0
+c
1
x +c
2
x
2
+..., it follows for r:
r
2
+ (b
0
−1)r +c
0
= 0
There are now 3 possibilities:
1. r
1
= r
2
: than y(x) = y
1
(x) ln [x[ +y
2
(x).
2. r
1
−r
2
∈ IN: than y(x) = ky
1
(x) ln [x[ +y
2
(x).
3. r
1
−r
2
= ZZ: than y(x) = y
1
(x) +y
2
(x).
22 Mathematics Formulary by ir. J.C.A. Wevers
4.2.2 Euler
Given the LDE
x
2
d
2
y(x)
dx
2
+ax
dy(x)
dx
+by(x) = 0
Substitution of y(x) = x
r
gives an equation for r: r
2
+ (a −1)r +b = 0. From this one gets two solutions r
1
and
r
2
. There are now 2 possibilities:
1. r
1
= r
2
: than y(x) = C
1
x
r1
+C
2
x
r
2
.
2. r
1
= r
2
= r: than y(x) = (C
1
ln(x) +C
2
)x
r
.
4.2.3 Legendre’s DE
Given the LDE
(1 −x
2
)
d
2
y(x)
dx
2
−2x
dy(x)
dx
+n(n −1)y(x) = 0
The solutions of this equation are given by y(x) = aP
n
(x) + by
2
(x) where the Legendre polynomials P(x) are
deﬁned by:
P
n
(x) =
d
n
dx
n
(1 −x
2
)
n
2
n
n!
For these holds: P
n

2
= 2/(2n + 1).
4.2.4 The associated Legendre equation
This equation follows from the θdependent part of the wave equation ∇
2
Ψ = 0 by substitution of
ξ = cos(θ). Than follows:
(1 −ξ
2
)
d
dξ
(1 −ξ
2
)
dP(ξ)
dξ
+ [C(1 −ξ
2
) −m
2
]P(ξ) = 0
Regular solutions exists only if C = l(l + 1). They are of the form:
P
m
l
(ξ) = (1 −ξ
2
)
m/2
d
m
P
0
(ξ)
dξ
m
=
(1 −ξ
2
)
m/2
2
l
l!
d
m+l
dξ
m+l
(ξ
2
−1)
l
For [m[ > l is P
m
l
(ξ) = 0. Some properties of P
0
l
(ξ) zijn:
1
−1
P
0
l
(ξ)P
0
l
(ξ)dξ =
2
2l + 1
δ
ll
,
∞
¸
l=0
P
0
l
(ξ)t
l
=
1
1 −2ξt +t
2
This polynomial can be written as:
P
0
l
(ξ) =
1
π
π
0
(ξ +
ξ
2
−1 cos(θ))
l
dθ
4.2.5 Solutions for Bessel’s equation
Given the LDE
x
2
d
2
y(x)
dx
2
+x
dy(x)
dx
+ (x
2
−ν
2
)y(x) = 0
also called Bessel’s equation, and the Bessel functions of the ﬁrst kind
J
ν
(x) = x
ν
∞
¸
m=0
(−1)
m
x
2m
2
2m+ν
m!Γ(ν +m+ 1)
Chapter 4: Diﬀerential equations 23
for ν := n ∈ IN this becomes:
J
n
(x) = x
n
∞
¸
m=0
(−1)
m
x
2m
2
2m+n
m!(n +m)!
When ν = ZZ the solution is given by y(x) = aJ
ν
(x) +bJ
−ν
(x). But because for n ∈ ZZ holds:
J
−n
(x) = (−1)
n
J
n
(x), this does not apply to integers. The general solution of Bessel’s equation is given by
y(x) = aJ
ν
(x) +bY
ν
(x), where Y
ν
are the Bessel functions of the second kind:
Y
ν
(x) =
J
ν
(x) cos(νπ) −J
−ν
(x)
sin(νπ)
and Y
n
(x) = lim
ν→n
Y
ν
(x)
The equation x
2
y
(x) + xy
(x) − (x
2
+ ν
2
)y(x) = 0 has the modiﬁed Bessel functions of the ﬁrst kind I
ν
(x) =
i
−ν
J
ν
(ix) as solution, and also solutions K
ν
= π[I
−ν
(x) −I
ν
(x)]/[2 sin(νπ)].
Sometimes it can be convenient to write the solutions of Bessel’s equation in terms of the Hankel functions
H
(1)
n
(x) = J
n
(x) +iY
n
(x) , H
(2)
n
(x) = J
n
(x) −iY
n
(x)
4.2.6 Properties of Bessel functions
Bessel functions are orthogonal with respect to the weight function p(x) = x.
J
−n
(x) = (−1)
n
J
n
(x). The Neumann functions N
m
(x) are deﬁnied as:
N
m
(x) =
1
2π
J
m
(x) ln(x) +
1
x
m
∞
¸
n=0
α
n
x
2n
The following holds: lim
x→0
J
m
(x) = x
m
, lim
x→0
N
m
(x) = x
−m
for m = 0, lim
x→0
N
0
(x) = ln(x).
lim
r→∞
H(r) =
e
±ikr
e
iωt
√
r
, lim
x→∞
J
n
(x) =
2
πx
cos(x −x
n
) , lim
x→∞
J
−n
(x) =
2
πx
sin(x −x
n
)
with x
n
=
1
2
π(n +
1
2
).
J
n+1
(x) +J
n−1
(x) =
2n
x
J
n
(x) , J
n+1
(x) −J
n−1
(x) = −2
dJ
n
(x)
dx
The following integral relations hold:
J
m
(x) =
1
2π
2π
0
exp[i(xsin(θ) −mθ)]dθ =
1
π
π
0
cos(xsin(θ) −mθ)dθ
4.2.7 Laguerre’s equation
Given the LDE
x
d
2
y(x)
dx
2
+ (1 −x)
dy(x)
dx
+ny(x) = 0
Solutions of this equation are the Laguerre polynomials L
n
(x):
L
n
(x) =
e
x
n!
d
n
dx
n
x
n
e
−x
=
∞
¸
m=0
(−1)
m
m!
n
m
x
m
24 Mathematics Formulary by ir. J.C.A. Wevers
4.2.8 The associated Laguerre equation
Given the LDE
d
2
y(x)
dx
2
+
m+ 1
x
−1
dy(x)
dx
+
n +
1
2
(m+ 1)
x
y(x) = 0
Solutions of this equation are the associated Laguerre polynomials L
m
n
(x):
L
m
n
(x) =
(−1)
m
n!
(n −m)!
e
−x
x
−m
d
n−m
dx
n−m
e
−x
x
n
4.2.9 Hermite
The differential equations of Hermite are:
d
2
H
n
(x)
dx
2
−2x
dH
n
(x)
dx
+ 2nH
n
(x) = 0 and
d
2
He
n
(x)
dx
2
−x
dHe
n
(x)
dx
+nHe
n
(x) = 0
Solutions of these equations are the Hermite polynomials, given by:
H
n
(x) = (−1)
n
exp
1
2
x
2
d
n
(exp(−
1
2
x
2
))
dx
n
= 2
n/2
He
n
(x
√
2)
He
n
(x) = (−1)
n
(exp
x
2
d
n
(exp(−x
2
))
dx
n
= 2
−n/2
H
n
(x/
√
2)
4.2.10 Chebyshev
The LDE
(1 −x
2
)
d
2
U
n
(x)
dx
2
−3x
dU
n
(x)
dx
+n(n + 2)U
n
(x) = 0
has solutions of the form
U
n
(x) =
sin[(n + 1) arccos(x)]
√
1 −x
2
The LDE
(1 −x
2
)
d
2
T
n
(x)
dx
2
−x
dT
n
(x)
dx
+n
2
T
n
(x) = 0
has solutions T
n
(x) = cos(narccos(x)).
4.2.11 Weber
The LDE W
n
(x) + (n +
1
2
−
1
4
x
2
)W
n
(x) = 0 has solutions: W
n
(x) = He
n
(x) exp(−
1
4
x
2
).
4.3 Nonlinear differential equations
Some nonlinear differential equations and a solution are:
y
= a
y
2
+b
2
y = b sinh(a(x −x
0
))
y
= a
y
2
−b
2
y = b cosh(a(x −x
0
))
y
= a
b
2
−y
2
y = b cos(a(x −x
0
))
y
= a(y
2
+b
2
) y = b tan(a(x −x
0
))
y
= a(y
2
−b
2
) y = b coth(a(x −x
0
))
y
= a(b
2
−y
2
) y = b tanh(a(x −x
0
))
y
= ay
b −y
b
y =
b
1 +Cb exp(−ax)
Chapter 4: Diﬀerential equations 25
4.4 SturmLiouville equations
SturmLiouville equations are second order LDE’s of the form:
−
d
dx
p(x)
dy(x)
dx
+q(x)y(x) = λm(x)y(x)
The boundary conditions are chosen so that the operator
L = −
d
dx
p(x)
d
dx
+q(x)
is Hermitean. The normalization function m(x) must satisfy
b
a
m(x)y
i
(x)y
j
(x)dx = δ
ij
When y
1
(x) and y
2
(x) are two linear independent solutions one can write the Wronskian in this form:
W(y
1
, y
2
) =
y
1
y
2
y
1
y
2
=
C
p(x)
where C is constant. By changing to another dependent variable u(x), given by: u(x) = y(x)
p(x), the LDE
transforms into the normal form:
d
2
u(x)
dx
2
+I(x)u(x) = 0 with I(x) =
1
4
p
(x)
p(x)
2
−
1
2
p
(x)
p(x)
−
q(x) −λm(x)
p(x)
If I(x) > 0, than y
/y < 0 and the solution has an oscillatory behaviour, if I(x) < 0, than y
/y > 0 and the
solution has an exponential behaviour.
4.5 Linear partial differential equations
4.5.1 General
The normal derivative is deﬁned by:
∂u
∂n
= (
∇u, n)
A frequently used solution method for PDE’s is separation of variables: one assumes that the solution can be written
as u(x, t) = X(x)T(t). When this is substituted two ordinary DE’s for X(x) and T(t) are obtained.
4.5.2 Special cases
The wave equation
The wave equation in 1 dimension is given by
∂
2
u
∂t
2
= c
2
∂
2
u
∂x
2
When the initial conditions u(x, 0) = ϕ(x) and ∂u(x, 0)/∂t = Ψ(x) apply, the general solution is given by:
u(x, t) =
1
2
[ϕ(x +ct) +ϕ(x −ct)] +
1
2c
x+ct
x−ct
Ψ(ξ)dξ
26 Mathematics Formulary by ir. J.C.A. Wevers
The diffusion equation
The diffusion equation is:
∂u
∂t
= D∇
2
u
Its solutions can be written in terms of the propagators P(x, x
, t). These have the property that
P(x, x
, 0) = δ(x −x
). In 1 dimension it reads:
P(x, x
, t) =
1
2
√
πDt
exp
−(x −x
)
2
4Dt
In 3 dimensions it reads:
P(x, x
, t) =
1
8(πDt)
3/2
exp
−(x −x
)
2
4Dt
With initial condition u(x, 0) = f(x) the solution is:
u(x, t) =
G
f(x
)P(x, x
, t)dx
The solution of the equation
∂u
∂t
−D
∂
2
u
∂x
2
= g(x, t)
is given by
u(x, t) =
dt
dx
g(x
, t
)P(x, x
, t −t
)
The equation of Helmholtz
The equation of Helmholtz is obtained by substitution of u(x, t) = v(x) exp(iωt) in the wave equation. This gives
for v:
∇
2
v(x, ω) +k
2
v(x, ω) = 0
This gives as solutions for v:
1. In cartesian coordinates: substitution of v = Aexp(i
k x) gives:
v(x) =
A(k)e
i
k·x
dk
with the integrals over
k
2
= k
2
.
2. In polar coordinates:
v(r, ϕ) =
∞
¸
m=0
(A
m
J
m
(kr) +B
m
N
m
(kr))e
imϕ
3. In spherical coordinates:
v(r, θ, ϕ) =
∞
¸
l=0
l
¸
m=−l
[A
lm
J
l+
1
2
(kr) +B
lm
J
−l−
1
2
(kr)]
Y (θ, ϕ)
√
r
Chapter 4: Diﬀerential equations 27
4.5.3 Potential theory and Green’s theorem
Subject of the potential theory are the Poisson equation ∇
2
u = −f(x) where f is a given function, and the Laplace
equation ∇
2
u = 0. The solutions of these can often be interpreted as a potential. The solutions of Laplace’s
equation are called harmonic functions.
When a vector ﬁeld v is given by v = gradϕ holds:
b
a
(v,
t )ds = ϕ(
b ) −ϕ(a )
In this case there exist functions ϕ and w so that v = gradϕ + curl w.
The ﬁeld lines of the ﬁeld v(x) follow from:
˙
x(t) = λv(x)
The ﬁrst theorem of Green is:
G
[u∇
2
v + (∇u, ∇v)]d
3
V =
(
S
u
∂v
∂n
d
2
A
The second theorem of Green is:
G
[u∇
2
v −v∇
2
u]d
3
V =
(
S
u
∂v
∂n
−v
∂u
∂n
d
2
A
A harmonic function which is 0 on the boundary of an area is also 0 within that area. A harmonic function with a
normal derivative of 0 on the boundary of an area is constant within that area.
The Dirichlet problem is:
∇
2
u(x) = −f(x) , x ∈ R , u(x) = g(x) for all x ∈ S.
It has a unique solution.
The Neumann problem is:
∇
2
u(x) = −f(x) , x ∈ R ,
∂u(x)
∂n
= h(x) for all x ∈ S.
The solution is unique except for a constant. The solution exists if:
−
R
f(x)d
3
V =
(
S
h(x)d
2
A
A fundamental solution of the Laplace equation satisﬁes:
∇
2
u(x) = −δ(x)
This has in 2 dimensions in polar coordinates the following solution:
u(r) =
ln(r)
2π
This has in 3 dimensions in spherical coordinates the following solution:
u(r) =
1
4πr
28 Mathematics Formulary by ir. J.C.A. Wevers
The equation ∇
2
v = −δ(x −
ξ ) has the solution
v(x) =
1
4π[x −
ξ [
After substituting this in Green’s 2nd theorem and applying the sieve property of the δ function one can derive
Green’s 3rd theorem:
u(
ξ ) = −
1
4π
R
∇
2
u
r
d
3
V +
1
4π
(
S
¸
1
r
∂u
∂n
−u
∂
∂n
1
r
d
2
A
The Green function G(x,
ξ ) is deﬁned by: ∇
2
G = −δ(x −
ξ ), and on boundary S holds G(x,
ξ ) = 0. Than G can
be written as:
G(x,
ξ ) =
1
4π[x −
ξ [
+g(x,
ξ )
Than g(x,
ξ ) is a solution of Dirichlet’s problem. The solution of Poisson’s equation ∇
2
u = −f(x) when on the
boundary S holds: u(x) = g(x), is:
u(
ξ ) =
R
G(x,
ξ )f(x)d
3
V −
(
S
g(x)
∂G(x,
ξ )
∂n
d
2
A
Chapter 5
Linear algebra
5.1 Vector spaces
( is a group for the operation ⊗ if:
1. ∀a, b ∈ ( ⇒ a ⊗b ∈ (: a group is closed.
2. (a ⊗b) ⊗c = a ⊗(b ⊗c): a group is associative.
3. ∃e ∈ ( so that a ⊗e = e ⊗a = a: there exists a unit element.
4. ∀a ∈ (∃a ∈ ( so that a ⊗a = e: each element has an inverse.
If
5. a ⊗b = b ⊗a
the group is called Abelian or commutative. Vector spaces form an Abelian group for addition and multiplication:
1 a =a, λ(µa) = (λµ)a, (λ +µ)(a +
b) = λa +λ
b +µa +µ
b.
W is a linear subspace if ∀ w
1
, w
2
∈ W holds: λ w
1
+µ w
2
∈ W.
W is an invariant subspace of V for the operator A if ∀ w ∈ W holds: A w ∈ W.
5.2 Basis
For an orthogonal basis holds: (e
i
, e
j
) = cδ
ij
. For an orthonormal basis holds: (e
i
, e
j
) = δ
ij
.
The set vectors ¦a
n
¦ is linear independent if:
¸
i
λ
i
a
i
= 0 ⇔ ∀
i
λ
i
= 0
The set ¦a
n
¦ is a basis if it is 1. independent and 2. V =<a
1
, a
2
, ... >=
¸
λ
i
a
i
.
5.3 Matrix calculus
5.3.1 Basic operations
For the matrix multiplication of matrices A = a
ij
and B = b
kl
holds with
r
the row index and
k
the column index:
A
r
1
k
1
B
r
2
k
2
= C
r
1
k
2
, (AB)
ij
=
¸
k
a
ik
b
kj
where
r
is the number of rows and
k
the number of columns.
The transpose of A is deﬁned by: a
T
ij
= a
ji
. For this holds (AB)
T
= B
T
A
T
and (A
T
)
−1
= (A
−1
)
T
. For the
inverse matrix holds: (A B)
−1
= B
−1
A
−1
. The inverse matrix A
−1
has the property that A A
−1
= II and can
be found by diagonalization: (A
ij
[II) ∼ (II[A
−1
ij
).
29
30 Mathematics Formulary by ir. J.C.A. Wevers
The inverse of a 2 2 matrix is:
a b
c d
−1
=
1
ad −bc
d −b
−c a
The determinant function D = det(A) is deﬁned by:
det(A) = D(a
∗1
, a
∗2
, ..., a
∗n
)
For the determinant det(A) of a matrix A holds: det(AB) = det(A) det(B). Een 2 2 matrix has determinant:
det
a b
c d
= ad −cb
The derivative of a matrix is a matrix with the derivatives of the coefﬁcients:
dA
dt
=
da
ij
dt
and
dAB
dt
= B
dA
dt
+A
dB
dt
The derivative of the determinant is given by:
d det(A)
dt
= D(
da
1
dt
, ..., a
n
) +D(a
1
,
da
2
dt
, ..., a
n
) +... +D(a
1
, ...,
da
n
dt
)
When the rows of a matrix are considered as vectors the row rank of a matrix is the number of independent vectors
in this set. Similar for the column rank. The row rank equals the column rank for each matrix.
Let
˜
A :
˜
V →
˜
V be the complex extension of the real linear operator A : V → V in a ﬁnite dimensional V . Then A
and
˜
A have the same caracteristic equation.
When A
ij
∈ IR and v
1
+i v
2
is an eigenvector of A at eigenvalue λ = λ
1
+iλ
2
, than holds:
1. Av
1
= λ
1
v
1
−λ
2
v
2
and Av
2
= λ
2
v
1
+λ
1
v
2
.
2. v
∗
= v
1
−iv
2
is an eigenvalue at λ
∗
= λ
1
−iλ
2
.
3. The linear span < v
1
, v
2
> is an invariant subspace of A.
If
k
n
are the columns of A, than the transformed space of A is given by:
R(A) =< Ae
1
, ..., Ae
n
>=<
k
1
, ...,
k
n
>
If the columns
k
n
of a n m matrix A are independent, than the nullspace ^(A) = ¦
0 ¦.
5.3.2 Matrix equations
We start with the equation
A x =
b
and
b =
0. If det(A) = 0 the only solution is
0. If det(A) = 0 there exists exactly one solution =
0.
The equation
A x =
0
has exactly one solution =
0 if det(A) = 0, and if det(A) = 0 the solution is
0.
Cramer’s rule for the solution of systems of linear equations is: let the system be written as
A x =
b ≡a
1
x
1
+... +a
n
x
n
=
b
then x
j
is given by:
x
j
=
D(a
1
, ..., a
j−1
,
b, a
j+1
, ..., a
n
)
det(A)
Chapter 5: Linear algebra 31
5.4 Linear transformations
A transformation A is linear if: A(λx +βy ) = λAx +βAy.
Some common linear transformations are:
Transformation type Equation
Projection on the line <a > P(x) = (a, x)a/(a, a )
Projection on the plane (a, x) = 0 Q(x) = x −P(x)
Mirror image in the line <a > S(x) = 2P(x) −x
Mirror image in the plane (a, x) = 0 T(x) = 2Q(x) −x = x −2P(x)
For a projection holds: x −P
W
(x) ⊥ P
W
(x) and P
W
(x) ∈ W.
If for a transformation A holds: (Ax, y ) = (x, Ay ) = (Ax, Ay ), than A is a projection.
Let A : W → W deﬁne a linear transformation; we deﬁne:
• If S is a subset of V : A(S) := ¦Ax ∈ W[x ∈ S¦
• If T is a subset of W: A
←
(T) := ¦x ∈ V [A(x) ∈ T¦
Than A(S) is a linear subspace of W and the inverse transformation A
←
(T) is a linear subspace of V . From this
follows that A(V ) is the image space of A, notation: 1(A). A
←
(
0 ) = E
0
is a linear subspace of V , the null space
of A, notation: ^(A). Then the following holds:
dim(^(A)) + dim(1(A)) = dim(V )
5.5 Plane and line
The equation of a line that contains the points a and
b is:
x =a +λ(
b −a ) =a +λr
The equation of a plane is:
x =a +λ(
b −a ) +µ(c −a ) =a +λr
1
+µr
2
When this is a plane in IR
3
, the normal vector to this plane is given by:
n
V
=
r
1
r
2
[r
1
r
2
[
A line can also be described by the points for which the line equation : (a, x) + b = 0 holds, and for a plane V:
(a, x) +k = 0. The normal vector to V is than: a/[a[.
The distance d between 2 points p and q is given by d( p, q ) =  p −q .
In IR
2
holds: The distance of a point p to the line (a, x) +b = 0 is
d( p, ) =
[(a, p ) +b[
[a[
Similarly in IR
3
: The distance of a point p to the plane (a, x) +k = 0 is
d( p, V ) =
[(a, p ) +k[
[a[
This can be generalized for IR
n
and C
n
(theorem from Hesse).
32 Mathematics Formulary by ir. J.C.A. Wevers
5.6 Coordinate transformations
The linear transformation A from IK
n
→ IK
m
is given by (IK = IR of C):
y = A
m×n
x
where a column of A is the image of a base vector in the original.
The matrix A
β
α
transforms a vector given w.r.t. a basis α into a vector w.r.t. a basis β. It is given by:
A
β
α
= (β(Aa
1
), ..., β(Aa
n
))
where β(x) is the representation of the vector x w.r.t. basis β.
The transformation matrix S
β
α
transforms vectors from coordinate system α into coordinate system β:
S
β
α
:= II
β
α
= (β(a
1
), ..., β(a
n
))
and S
β
α
S
α
β
= II
The matrix of a transformation A is than given by:
A
β
α
=
A
β
α
e
1
, ..., A
β
α
e
n
For the transformation of matrix operators to another coordinate system holds: A
δ
α
= S
δ
λ
A
λ
β
S
β
α
, A
α
α
= S
α
β
A
β
β
S
β
α
and (AB)
λ
α
= A
λ
β
B
β
α
.
Further is A
β
α
= S
β
α
A
α
α
, A
α
β
= A
α
α
S
α
β
. A vector is transformed via X
α
= S
β
α
X
β
.
5.7 Eigen values
The eigenvalue equation
Ax = λx
with eigenvalues λ can be solved with (A − λII) =
0 ⇒ det(A − λII) = 0. The eigenvalues follow from this
characteristic equation. The following is true: det(A) =
¸
i
λ
i
and Tr(A) =
¸
i
a
ii
=
¸
i
λ
i
.
The eigen values λ
i
are independent of the chosen basis. The matrix of A in a basis of eigenvectors, with S the
transformation matrix to this basis, S = (E
λ
1
, ..., E
λ
n
), is given by:
Λ = S
−1
AS = diag(λ
1
, ..., λ
n
)
When 0 is an eigen value of A than E
0
(A) = ^(A).
When λ is an eigen value of A holds: A
n
x = λ
n
x.
5.8 Transformation types
Isometric transformations
A transformation is isometric when: Ax = x. This implies that the eigen values of an isometric transformation
are given by λ = exp(iϕ) ⇒ [λ[ = 1. Than also holds: (Ax, Ay ) = (x, y ).
When W is an invariant subspace if the isometric transformation Awith dim(A) < ∞, than also W
⊥
is an invariante
subspace.
Chapter 5: Linear algebra 33
Orthogonal transformations
A transformation A is orthogonal if A is isometric and the inverse A
←
exists. For an orthogonal transformation O
holds O
T
O = II, so: O
T
= O
−1
. If A and B are orthogonal, than AB and A
−1
are also orthogonal.
Let A : V → V be orthogonal with dim(V ) < ∞. Than A is:
Direct orthogonal if det(A) = +1. A describes a rotation. A rotation in IR
2
through angle ϕ is given by:
R =
cos(ϕ) −sin(ϕ)
sin(ϕ) cos(ϕ)
So the rotation angle ϕ is determined by Tr(A) = 2 cos(ϕ) with 0 ≤ ϕ ≤ π. Let λ
1
and λ
2
be the roots of the
characteristic equation, than also holds: '(λ
1
) = '(λ
2
) = cos(ϕ), and λ
1
= exp(iϕ), λ
2
= exp(−iϕ).
In IR
3
holds: λ
1
= 1, λ
2
= λ
∗
3
= exp(iϕ). A rotation over E
λ
1
is given by the matrix
¸
1 0 0
0 cos(ϕ) −sin(ϕ)
0 sin(ϕ) cos(ϕ)
¸
Mirrored orthogonal if det(A) = −1. Vectors from E
−1
are mirrored by A w.r.t. the invariant subspace E
⊥
−1
. A
mirroring in IR
2
in < (cos(
1
2
ϕ), sin(
1
2
ϕ)) > is given by:
S =
cos(ϕ) sin(ϕ)
sin(ϕ) −cos(ϕ)
Mirrored orthogonal transformations in IR
3
are rotational mirrorings: rotations of axis <a
1
> through angle ϕ and
mirror plane <a
1
>
⊥
. The matrix of such a transformation is given by:
¸
−1 0 0
0 cos(ϕ) −sin(ϕ)
0 sin(ϕ) cos(ϕ)
¸
For all orthogonal transformations O in IR
3
holds that O(x) O(y ) = O(x y ).
IR
n
(n < ∞) can be decomposed in invariant subspaces with dimension 1 or 2 for each orthogonal transformation.
Unitary transformations
Let V be a complex space on which an inner product is deﬁned. Than a linear transformation U is unitary if U is
isometric and its inverse transformation A
←
exists. A n n matrix is unitary if U
H
U = II. It has determinant
[ det(U)[ = 1. Each isometric transformation in a ﬁnitedimensional complex vector space is unitary.
Theorem: for a n n matrix A the following statements are equivalent:
1. A is unitary,
2. The columns of A are an orthonormal set,
3. The rows of A are an orthonormal set.
Symmetric transformations
A transformation A on IR
n
is symmetric if (Ax, y ) = (x, Ay ). A matrix A ∈ IM
n×n
is symmetric if A = A
T
. A
linear operator is only symmetric if its matrix w.r.t. an arbitrary basis is symmetric. All eigenvalues of a symmetric
transformation belong to IR. The different eigenvectors are mutually perpendicular. If A is symmetric, than A
T
=
A = A
H
on an orthogonal basis.
For each matrix B ∈ IM
m×n
holds: B
T
B is symmetric.
34 Mathematics Formulary by ir. J.C.A. Wevers
Hermitian transformations
A transformation H : V → V with V = C
n
is Hermitian if (Hx, y ) = (x, Hy ). The Hermitian conjugated
transformation A
H
of A is: [a
ij
]
H
= [a
∗
ji
]. An alternative notation is: A
H
= A
†
. The inner product of two vectors
x and y can now be written in the form: (x, y ) = x
H
y.
If the transformations A and B are Hermitian, than their product AB is Hermitian if:
[A, B] = AB −BA = 0. [A, B] is called the commutator of A and B.
The eigenvalues of a Hermitian transformation belong to IR.
A matrix representation can be coupled with a Hermitian operator L. W.r.t. a basis e
i
it is given by L
mn
=
(e
m
, Le
n
).
Normal transformations
For each linear transformation A in a complex vector space V there exists exactly one linear transformation B so
that (Ax, y ) = (x, By ). This B is called the adjungated transformation of A. Notation: B = A
∗
. The following
holds: (CD)
∗
= D
∗
C
∗
. A
∗
= A
−1
if A is unitary and A
∗
= A if A is Hermitian.
Deﬁnition: the linear transformation A is normal in a complex vector space V if A
∗
A = AA
∗
. This is only the case
if for its matrix S w.r.t. an orthonormal basis holds: A
†
A = AA
†
.
If A is normal holds:
1. For all vectors x ∈ V and a normal transformation A holds:
(Ax, Ay ) = (A
∗
Ax, y ) = (AA
∗
x, y ) = (A
∗
x, A
∗
y )
2. x is an eigenvector of A if and only if x is an eigenvector of A
∗
.
3. Eigenvectors of A for different eigenvalues are mutually perpendicular.
4. If E
λ
if an eigenspace from A than the orthogonal complement E
⊥
λ
is an invariant subspace of A.
Let the different roots of the characteristic equation of A be β
i
with multiplicities n
i
. Than the dimension of each
eigenspace V
i
equals n
i
. These eigenspaces are mutually perpendicular and each vector x ∈ V can be written in
exactly one way as
x =
¸
i
x
i
with x
i
∈ V
i
This can also be written as: x
i
= P
i
x where P
i
is a projection on V
i
. This leads to the spectral mapping theorem:
let A be a normal transformation in a complex vector space V with dim(V ) = n. Than:
1. There exist projection transformations P
i
, 1 ≤ i ≤ p, with the properties
• P
i
P
j
= 0 for i = j,
• P
1
+... +P
p
= II,
• dimP
1
(V ) +... + dimP
p
(V ) = n
and complex numbers α
1
, ..., α
p
so that A = α
1
P
1
+... +α
p
P
p
.
2. If A is unitary than holds [α
i
[ = 1 ∀i.
3. If A is Hermitian than α
i
∈ IR ∀i.
Chapter 5: Linear algebra 35
Complete systems of commuting Hermitian transformations
Consider m Hermitian linear transformations A
i
in a n dimensional complex inner product space V . Assume they
mutually commute.
Lemma: if E
λ
is the eigenspace for eigenvalue λ from A
1
, than E
λ
is an invariant subspace of all transformations
A
i
. This means that if x ∈ E
λ
, than A
i
x ∈ E
λ
.
Theorem. Consider m commuting Hermitian matrices A
i
. Than there exists a unitary matrix U so that all matrices
U
†
A
i
U are diagonal. The columns of U are the common eigenvectors of all matrices A
j
.
If all eigenvalues of a Hermitian linear transformation in a ndimensional complex vector space differ, than the
normalized eigenvector is known except for a phase factor exp(iα).
Deﬁnition: a commuting set Hermitian transformations is called complete if for each set of two common eigenvec
tors v
i
, v
j
there exists a transformation A
k
so that v
i
and v
j
are eigenvectors with different eigenvalues of A
k
.
Usually a commuting set is taken as small as possible. In quantum physics one speaks of commuting observables.
The required number of commuting observables equals the number of quantum numbers required to characterize a
state.
5.9 Homogeneous coordinates
Homogeneous coordinates are used if one wants to combine both rotations and translations in one matrix transfor
mation. An extra coordinate is introduced to describe the nonlinearities. Homogeneous coordinates are derived
from cartesian coordinates as follows:
¸
x
y
z
¸
cart
=
¸
¸
¸
wx
wy
wz
w
¸
hom
=
¸
¸
¸
X
Y
Z
w
¸
hom
so x = X/w, y = Y/w and z = Z/w. Transformations in homogeneous coordinates are described by the following
matrices:
1. Translation along vector (X
0
, Y
0
, Z
0
, w
0
):
T =
¸
¸
¸
w
0
0 0 X
0
0 w
0
0 Y
0
0 0 w
0
Z
0
0 0 0 w
0
¸
2. Rotations of the x, y, z axis, resp. through angles α, β, γ:
R
x
(α) =
¸
¸
¸
1 0 0 0
0 cos α −sin α 0
0 sin α cos α 0
0 0 0 1
¸
R
y
(β) =
¸
¸
¸
cos β 0 sin β 0
0 1 0 0
−sin β 0 cos β 0
0 0 0 1
¸
R
z
(γ) =
¸
¸
¸
cos γ −sin γ 0 0
sin γ cos γ 0 0
0 0 1 0
0 0 0 1
¸
3. A perspective projection on image plane z = c with the center of projection in the origin. This transformation
has no inverse.
P(z = c) =
¸
¸
¸
1 0 0 0
0 1 0 0
0 0 1 0
0 0 1/c 0
¸
36 Mathematics Formulary by ir. J.C.A. Wevers
5.10 Inner product spaces
A complex inner product on a complex vector space is deﬁned as follows:
1. (a,
b ) = (
b, a ),
2. (a, β
1
b
1
+β
2
b
2
) = β
1
(a,
b
1
) +β
2
(a,
b
2
) for all a,
b
1
,
b
2
∈ V and β
1
, β
2
∈ C.
3. (a, a ) ≥ 0 for all a ∈ V , (a, a ) = 0 if and only if a =
0.
Due to (1) holds: (a, a ) ∈ IR. The inner product space C
n
is the complex vector space on which a complex inner
product is deﬁned by:
(a,
b ) =
n
¸
i=1
a
∗
i
b
i
For function spaces holds:
(f, g) =
b
a
f
∗
(t)g(t)dt
For eacha the length a  is deﬁned by: a  =
(a, a ). The following holds: a −
b  ≤ a+
b  ≤ a +
b ,
and with ϕ the angle between a and
b holds: (a,
b ) = a  
b  cos(ϕ).
Let ¦a
1
, ..., a
n
¦ be a set of vectors in an inner product space V . Than the Gramian G of this set is given by:
G
ij
= (a
i
, a
j
). The set of vectors is independent if and only if det(G) = 0.
A set is orthonormal if (a
i
, a
j
) = δ
ij
. If e
1
, e
2
, ... form an orthonormal row in an inﬁnite dimensional vector space
Bessel’s inequality holds:
x
2
≥
∞
¸
i=1
[(e
i
, x)[
2
The equal sign holds if and only if lim
n→∞
x
n
−x = 0.
The inner product space
2
is deﬁned in C
∞
by:
2
=
a = (a
1
, a
2
, ...) [
∞
¸
n=1
[a
n
[
2
< ∞
¸
A space is called a Hilbert space if it is
2
and if also holds: lim
n→∞
[a
n+1
−a
n
[ = 0.
5.11 The Laplace transformation
The class LT exists of functions for which holds:
1. On each interval [0, A], A > 0 there are no more than a ﬁnite number of discontinuities and each discontinuity
has an upper  and lower limit,
2. ∃t
0
∈ [0, ∞ > and a, M ∈ IR so that for t ≥ t
0
holds: [f(t)[ exp(−at) < M.
Than there exists a Laplace transform for f.
The Laplace transformation is a generalisation of the Fourier transformation. The Laplace transform of a function
f(t) is, with s ∈ C and t ≥ 0:
F(s) =
∞
0
f(t)e
−st
dt
Chapter 5: Linear algebra 37
The Laplace transform of the derivative of a function is given by:
L
f
(n)
(t)
= −f
(n−1)
(0) −sf
(n−2)
(0) −... −s
n−1
f(0) +s
n
F(s)
The operator L has the following properties:
1. Equal shapes: if a > 0 than
L(f(at)) =
1
a
F
s
a
2. Damping: L(e
−at
f(t)) = F(s +a)
3. Translation: If a > 0 and g is deﬁned by g(t) = f(t − a) if t > a and g(t) = 0 for t ≤ a, than holds:
L(g(t)) = e
−sa
L(f(t)).
If s ∈ IR than holds '(λf) = L('(f)) and ·(λf) = L(·(f)).
For some often occurring functions holds:
f(t) = F(s) = L(f(t)) =
t
n
n!
e
at
(s −a)
−n−1
e
at
cos(ωt)
s −a
(s −a)
2
+ω
2
e
at
sin(ωt)
ω
(s −a)
2
+ω
2
δ(t −a) exp(−as)
5.12 The convolution
The convolution integral is deﬁned by:
(f ∗ g)(t) =
t
0
f(u)g(t −u)du
The convolution has the following properties:
1. f ∗ g ∈LT
2. L(f ∗ g) = L(f) L(g)
3. Distribution: f ∗ (g +h) = f ∗ g +f ∗ h
4. Commutative: f ∗ g = g ∗ f
5. Homogenity: f ∗ (λg) = λf ∗ g
If L(f) = F
1
F
2
, than is f(t) = f
1
∗ f
2
.
5.13 Systems of linear differential equations
We start with the equation
˙
x = Ax. Assume that x = v exp(λt), than follows: Av = λv. In the 2 2 case holds:
1. λ
1
= λ
2
: than x(t) =
¸
v
i
exp(λ
i
t).
2. λ
1
= λ
2
: than x(t) = (ut +v) exp(λt).
38 Mathematics Formulary by ir. J.C.A. Wevers
Assume that λ = α + iβ is an eigenvalue with eigenvector v, than λ
∗
is also an eigenvalue for eigenvector v
∗
.
Decompose v = u +i w, than the real solutions are
c
1
[ucos(βt) − wsin(βt)]e
αt
+c
2
[v cos(βt) +usin(βt)]e
αt
There are two solution strategies for the equation
¨
x = Ax:
1. Let x = v exp(λt) ⇒ det(A−λ
2
II) = 0.
2. Introduce: ˙ x = u and ˙ y = v, this leads to ¨ x = ˙ u and ¨ y = ˙ v. This transforms a ndimensional set of second
order equations into a 2ndimensional set of ﬁrst order equations.
5.14 Quadratic forms
5.14.1 Quadratic forms in IR
2
The general equation of a quadratic form is: x
T
Ax + 2x
T
P + S = 0. Here, A is a symmetric matrix. If Λ =
S
−1
AS = diag(λ
1
, ..., λ
n
) holds: u
T
Λu+2u
T
P +S = 0, so all cross terms are 0. u = (u, v, w) should be chosen
so that det(S) = +1, to maintain the same orientation as the system (x, y, z).
Starting with the equation
ax
2
+ 2bxy +cy
2
+dx +ey +f = 0
we have [A[ = ac − b
2
. An ellipse has [A[ > 0, a parabola [A[ = 0 and a hyperbole [A[ < 0. In polar coordinates
this can be written as:
r =
ep
1 −e cos(θ)
An ellipse has e < 1, a parabola e = 1 and a hyperbola e > 1.
5.14.2 Quadratic surfaces in IR
3
Rank 3:
p
x
2
a
2
+q
y
2
b
2
+r
z
2
c
2
= d
• Ellipsoid: p = q = r = d = 1, a, b, c are the lengths of the semi axes.
• Singlebladed hyperboloid: p = q = d = 1, r = −1.
• Doublebladed hyperboloid: r = d = 1, p = q = −1.
• Cone: p = q = 1, r = −1, d = 0.
Rank 2:
p
x
2
a
2
+q
y
2
b
2
+r
z
c
2
= d
• Elliptic paraboloid: p = q = 1, r = −1, d = 0.
• Hyperbolic paraboloid: p = r = −1, q = 1, d = 0.
• Elliptic cylinder: p = q = −1, r = d = 0.
• Hyperbolic cylinder: p = d = 1, q = −1, r = 0.
• Pair of planes: p = 1, q = −1, d = 0.
Rank 1:
py
2
+qx = d
• Parabolic cylinder: p, q > 0.
• Parallel pair of planes: d > 0, q = 0, p = 0.
• Double plane: p = 0, q = d = 0.
Chapter 6
Complex function theory
6.1 Functions of complex variables
Complex function theory deals with complex functions of a complex variable. Some deﬁnitions:
f is analytical on ( if f is continuous and differentiable on (.
A Jordan curve is a curve that is closed and singular.
If K is a curve in C with parameter equation z = φ(t) = x(t) + iy(t), a ≤ t ≤ b, than the length L of K is given
by:
L =
b
a
dx
dt
2
+
dy
dt
2
dt =
b
a
dz
dt
dt =
b
a
[φ
(t)[dt
The derivative of f in point z = a is:
f
(a) = lim
z→a
f(z) −f(a)
z −a
If f(z) = u(x, y) +iv(x, y) the derivative is:
f
(z) =
∂u
∂x
+i
∂v
∂x
= −i
∂u
∂y
+
∂v
∂y
Setting both results equal yields the equations of CauchyRiemann:
∂u
∂x
=
∂v
∂y
,
∂u
∂y
= −
∂v
∂x
These equations imply that ∇
2
u = ∇
2
v = 0. f is analytical if u and v satisfy these equations.
6.2 Complex integration
6.2.1 Cauchy’s integral formula
Let K be a curve described by z = φ(t) on a ≤ t ≤ b and f(z) is continuous on K. Than the integral of f over K
is:
K
f(z)dz =
b
a
f(φ(t))
˙
φ(t)dt
fcontinuous
= F(b) −F(a)
Lemma: let K be the circle with center a and radius r taken in a positive direction. Than holds for integer m:
1
2πi
K
dz
(z −a)
m
=
0 if m = 1
1 if m = 1
Theorem: if L is the length of curve K and if [f(z)[ ≤ M for z ∈ K, than, if the integral exists, holds:
K
f(z)dz
≤ ML
39
40 Mathematics Formulary by ir. J.C.A. Wevers
Theorem: let f be continuous on an area G and let p be a ﬁxed point of G. Let F(z) =
z
p
f(ξ)dξ for all z ∈ G
only depend on z and not on the integration path. Than F(z) is analytical on G with F
(z) = f(z).
This leads to two equivalent formulations of the main theorem of complex integration: let the function f be analytical
on an area G. Let K and K
be two curves with the same starting  and end points, which can be transformed into
each other by continous deformation within G. Let B be a Jordan curve. Than holds
K
f(z)dz =
K
f(z)dz ⇔
B
f(z)dz = 0
By applying the main theorem on e
iz
/z one can derive that
∞
0
sin(x)
x
dx =
π
2
6.2.2 Residue
A point a ∈ C is a regular point of a function f(z) if f is analytical in a. Otherwise a is a singular point or pole of
f(z). The residue of f in a is deﬁned by
Res
z=a
f(z) =
1
2πi
K
f(z)dz
where K is a Jordan curve which encloses a in positive direction. The residue is 0 in regular points, in singular
points it can be both 0 and = 0. Cauchy’s residue proposition is: let f be analytical within and on a Jordan curve K
except in a ﬁnite number of singular points a
i
within K. Than, if K is taken in a positive direction, holds:
1
2πi
K
f(z)dz =
n
¸
k=1
Res
z=a
k
f(z)
Lemma: let the function f be analytical in a, than holds:
Res
z=a
f(z)
z −a
= f(a)
This leads to Cauchy’s integral theorem: if F is analytical on the Jordan curve K, which is taken in a positive
direction, holds:
1
2πi
K
f(z)
z −a
dz =
f(a) if a inside K
0 if a outside K
Theorem: let K be a curve (K need not be closed) and let φ(ξ) be continuous on K. Than the function
f(z) =
K
φ(ξ)dξ
ξ −z
is analytical with nth derivative
f
(n)
(z) = n!
K
φ(ξ)dξ
(ξ −z)
n+1
Theorem: let K be a curve and G an area. Let φ(ξ, z) be deﬁned for ξ ∈ K, z ∈ G, with the following properties:
1. φ(ξ, z) is limited, this means [φ(ξ, z)[ ≤ M for ξ ∈ K, z ∈ G,
2. For ﬁxed ξ ∈ K, φ(ξ, z) is an analytical function of z on G,
Chapter 6: Complex function theory 41
3. For ﬁxed z ∈ G the functions φ(ξ, z) and ∂φ(ξ, z)/∂z are continuous functions of ξ on K.
Than the function
f(z) =
K
φ(ξ, z)dξ
is analytical with derivative
f
(z) =
K
∂φ(ξ, z)
∂z
dξ
Cauchy’s inequality: let f(z) be an analytical function within and on the circle C : [z −a[ = Rand let [f(z)[ ≤ M
for z ∈ C. Than holds
f
(n)
(a)
≤
Mn!
R
n
6.3 Analytical functions deﬁnied by series
The series
¸
f
n
(z) is called pointwise convergent on an area G with sum F(z) if
∀
ε>0
∀
z∈G
∃
N
0
∈IR
∀
n>n
0
¸
f(z) −
N
¸
n=1
f
n
(z)
< ε
¸
The series is called uniform convergent if
∀
ε>0
∃
N
0
∈IR
∀
n>n
0
∃
z∈G
¸
f(z) −
N
¸
n=1
f
n
(z)
< ε
¸
Uniform convergence implies pointwise convergence, the opposite is not necessary.
Theorem: let the power series
∞
¸
n=0
a
n
z
n
have a radius of convergence R. R is the distance to the ﬁrst nonessential
singularity.
• If lim
n→∞
n
[a
n
[ = L exists, than R = 1/L.
• If lim
n→∞
[a
n+1
[/[a
n
[ = L exists, than R = 1/L.
If these limits both don’t exist one can ﬁnd R with the formula of CauchyHadamard:
1
R
= lim
n→∞
sup
n
[a
n
[
6.4 Laurent series
Taylor’s theorem: let f be analytical in an area G and let point a ∈ G has distance r to the boundary of G. Than
f(z) can be expanded into the Taylor series near a:
f(z) =
∞
¸
n=0
c
n
(z −a)
n
with c
n
=
f
(n)
(a)
n!
valid for [z − a[ < r. The radius of convergence of the Taylor series is ≥ r. If f has a pole of order k in a than
c
1
, ..., c
k−1
= 0, c
k
= 0.
Theorem of Laurent: let f be analytical in the circular area G : r < [z −a[ < R. Than f(z) can be expanded into
a Laurent series with center a:
f(z) =
∞
¸
n=−∞
c
n
(z −a)
n
with c
n
=
1
2πi
K
f(w)dw
(w −a)
n+1
, n ∈ ZZ
42 Mathematics Formulary by ir. J.C.A. Wevers
valid for r < [z −a[ < R and K an arbitrary Jordan curve in G which encloses point a in positive direction.
The principal part of a Laurent series is:
∞
¸
n=1
c
−n
(z −a)
−n
. One can classify singular points with this. There are 3
cases:
1. There is no principal part. Than a is a nonessential singularity. Deﬁne f(a) = c
0
and the series is also valid
for [z −a[ < R and f is analytical in a.
2. The principal part contains a ﬁnite number of terms. Than there exists a k ∈ IN so that
lim
z→a
(z − a)
k
f(z) = c
−k
= 0. Than the function g(z) = (z − a)
k
f(z) has a nonessential singularity in a.
One speaks of a pole of order k in z = a.
3. The principal part contains an inﬁnite number of terms. Then, a is an essential singular point of f, such as
exp(1/z) for z = 0.
If f and g are analytical, f(a) = 0, g(a) = 0, g
(a) = 0 than f(z)/g(z) has a simple pole (i.e. a pole of order 1) in
z = a with
Res
z=a
f(z)
g(z)
=
f(a)
g
(a)
6.5 Jordan’s theorem
Residues are often used when solving deﬁnite integrals. We deﬁne the notations C
+
ρ
= ¦z[[z[ = ρ, ·(z) ≥ 0¦ and
C
−
ρ
= ¦z[[z[ = ρ, ·(z) ≤ 0¦ and M
+
(ρ, f) = max
z∈C
+
ρ
[f(z)[, M
−
(ρ, f) = max
z∈C
−
ρ
[f(z)[. We assume that f(z) is
analytical for ·(z) > 0 with a possible exception of a ﬁnite number of singular points which do not lie on the real
axis, lim
ρ→∞
ρM
+
(ρ, f) = 0 and that the integral exists, than
∞
−∞
f(x)dx = 2πi
¸
Resf(z) in ·(z) > 0
Replace M
+
by M
−
in the conditions above and it follows that:
∞
−∞
f(x)dx = −2πi
¸
Resf(z) in ·(z) < 0
Jordan’s lemma: let f be continuous for [z[ ≥ R, ·(z) ≥ 0 and lim
ρ→∞
M
+
(ρ, f) = 0. Than holds for α > 0
lim
ρ→∞
C
+
ρ
f(z)e
iαz
dz = 0
Let f be continuous for [z[ ≥ R, ·(z) ≤ 0 and lim
ρ→∞
M
−
(ρ, f) = 0. Than holds for α < 0
lim
ρ→∞
C
−
ρ
f(z)e
iαz
dz = 0
Let z = a be a simple pole of f(z) and let C
δ
be the half circle [z −a[ = δ, 0 ≤ arg(z −a) ≤ π, taken from a +δ
to a −δ. Than is
lim
δ↓0
1
2πi
C
δ
f(z)dz =
1
2
Res
z=a
f(z)
Chapter 7
Tensor calculus
7.1 Vectors and covectors
A ﬁnite dimensional vector space is denoted by 1, J. The vector space of linear transformations from 1 to J is
denoted by L(1, J). Consider L(1,IR) := 1
∗
. We name 1
∗
the dual space of 1. Now we can deﬁne vectors in 1
with basis c and covectors in 1
∗
with basis
ˆ
c. Properties of both are:
1. Vectors: x = x
i
c
i
with basis vectors c
i
:
c
i
=
∂
∂x
i
Transformation from system i to i
is given by:
c
i
= A
i
i
c
i
= ∂
i
∈ 1 , x
i
= A
i
i
x
i
2. Covectors:
ˆ
x = x
i
ˆ
c
i
with basis vectors
ˆ
c
i
ˆ
c
i
= dx
i
Transformation from system i to i
is given by:
ˆ
c
i
= A
i
i
ˆ
c
i
∈ 1
∗
, x
i
= A
i
i
x
i
Here the Einstein convention is used:
a
i
b
i
:=
¸
i
a
i
b
i
The coordinate transformation is given by:
A
i
i
=
∂x
i
∂x
i
, A
i
i
=
∂x
i
∂x
i
From this follows that A
i
k
A
k
l
= δ
k
l
and A
i
i
= (A
i
i
)
−1
.
In differential notation the coordinate transformations are given by:
dx
i
=
∂x
i
∂x
i
dx
i
and
∂
∂x
i
=
∂x
i
∂x
i
∂
∂x
i
The general transformation rule for a tensor T is:
T
q
1
...q
n
s
1
...s
m
=
∂x
∂u
∂u
q
1
∂x
p
1
∂u
q
n
∂x
p
n
∂x
r
1
∂u
s
1
∂x
r
m
∂u
s
m
T
p
1
...p
n
r
1
...r
m
For an absolute tensor = 0.
43
44 Mathematics Formulary by ir. J.C.A. Wevers
7.2 Tensor algebra
The following holds:
a
ij
(x
i
+y
i
) ≡ a
ij
x
i
+a
ij
y
i
, but: a
ij
(x
i
+y
j
) ≡ a
ij
x
i
+a
ij
y
j
and
(a
ij
+a
ji
)x
i
x
j
≡ 2a
ij
x
i
x
j
, but: (a
ij
+a
ji
)x
i
y
j
≡ 2a
ij
x
i
y
j
en (a
ij
−a
ji
)x
i
x
j
≡ 0.
The sum and difference of two tensors is a tensor of the same rank: A
p
q
± B
p
q
. The outer tensor product results in
a tensor with a rank equal to the sum of the ranks of both tensors: A
pr
q
B
m
s
= C
prm
qs
. The contraction equals two
indices and sums over them. Suppose we take r = s for a tensor A
mpr
qs
, this results in:
¸
r
A
mpr
qr
= B
mp
q
. The inner
product of two tensors is deﬁned by taking the outer product followed by a contraction.
7.3 Inner product
Deﬁnition: the bilinear transformation B : 1 1
∗
→ IR, B(x,
ˆ
y ) =
ˆ
y(x) is denoted by < x,
ˆ
y >. For this pairing
operator < , >= δ holds:
ˆ
y(x) =< x,
ˆ
y >= y
i
x
i
, <
ˆ
c
i
, c
j
>= δ
i
j
Let G : 1 → 1
∗
be a linear bijection. Deﬁne the bilinear forms
g : 1 1 → IR g(x, y ) =< x, Gy >
h : 1
∗
1
∗
→ IR h(
ˆ
x,
ˆ
y ) =< G
−1
ˆ
x,
ˆ
y >
Both are not degenerated. The following holds: h(Gx, Gy ) =< x, Gy >= g(x, y ). If we identify 1 and 1
∗
with
G, than g (or h) gives an inner product on 1.
The inner product (, )
Λ
on Λ
k
(1) is deﬁned by:
(Φ, Ψ)
Λ
=
1
k!
(Φ, Ψ)
T
0
k
(V)
The inner product of two vectors is than given by:
(x, y ) = x
i
y
i
<c
i
, Gc
j
>= g
ij
x
i
x
j
The matrix g
ij
of G is given by
g
ij
ˆ
c
j
= Gc
i
The matrix g
ij
of G
−1
is given by:
g
kl
c
l
= G
−1
ˆ
c
k
For this metric tensor g
ij
holds: g
ij
g
jk
= δ
k
i
. This tensor can raise or lower indices:
x
j
= g
ij
x
i
, x
i
= g
ij
x
j
and du
i
=
ˆ
c
i
= g
ij
c
j
.
Chapter 7: Tensor calculus 45
7.4 Tensor product
Deﬁnition: let  and 1 be two ﬁnite dimensional vector spaces with dimensions m and n. Let 
∗
1
∗
be the
cartesian product of  and 1. A function t : 
∗
1
∗
→ IR; (
ˆ
u;
ˆ
v ) → t(
ˆ
u;
ˆ
v ) = t
αβ
u
α
u
β
∈ IR is called a tensor
if t is linear in
ˆ
u and
ˆ
v. The tensors t form a vector space denoted by  ⊗ 1. The elements T ∈ 1 ⊗ 1 are called
contravariant 2tensors: T = T
ij
c
i
⊗c
j
= T
ij
∂
i
⊗∂
j
. The elements T ∈ 1
∗
⊗1
∗
are called covariant 2tensors:
T = T
ij
ˆ
c
i
⊗
ˆ
c
j
= T
ij
dx
i
⊗ dx
j
. The elements T ∈ 1
∗
⊗ 1 are called mixed 2 tensors: T = T
.j
i
ˆ
c
i
⊗c
j
=
T
.j
i
dx
i
⊗∂
j
, and analogous for T ∈ 1 ⊗1
∗
.
The numbers given by
t
αβ
= t(
ˆ
c
α
,
ˆ
c
β
)
with 1 ≤ α ≤ m and 1 ≤ β ≤ n are the components of t.
Take x ∈  and y ∈ 1. Than the function x ⊗y, deﬁnied by
(x ⊗y)(
ˆ
u,
ˆ
v) =< x,
ˆ
u >
U
< y,
ˆ
v >
V
is a tensor. The components are derived from: (u ⊗v )
ij
= u
i
v
j
. The tensor product of 2 tensors is given by:
2
0
form: (v ⊗ w)(
ˆ
p,
ˆ
q) = v
i
p
i
w
k
q
k
= T
ik
p
i
q
k
0
2
form: (
ˆ
p ⊗
ˆ
q)(v, w) = p
i
v
i
q
k
w
k
= T
ik
v
i
w
k
1
1
form: (v ⊗
ˆ
p)(
ˆ
q, w) = v
i
q
i
p
k
w
k
= T
i
k
q
i
w
k
7.5 Symmetric and antisymmetric tensors
A tensor t ∈ 1 ⊗1 is called symmetric resp. antisymmetric if ∀
ˆ
x,
ˆ
y ∈ 1
∗
holds: t(
ˆ
x,
ˆ
y ) = t(
ˆ
y,
ˆ
x) resp. t(
ˆ
x,
ˆ
y ) =
−t(
ˆ
y,
ˆ
x).
A tensor t ∈ 1
∗
⊗1
∗
is called symmetric resp. antisymmetric if ∀x, y ∈ 1 holds: t(x, y ) = t(y, x) resp.
t(x, y ) = −t(y, x). The linear transformations o and / in 1 ⊗J are deﬁned by:
ot(
ˆ
x,
ˆ
y ) =
1
2
(t(
ˆ
x,
ˆ
y) +t(
ˆ
y,
ˆ
x))
/t(
ˆ
x,
ˆ
y ) =
1
2
(t(
ˆ
x,
ˆ
y) −t(
ˆ
y,
ˆ
x))
Analogous in 1
∗
⊗1
∗
. If t is symmetric resp. antisymmetric, than ot = t resp. /t = t.
The tensors e
i
∨e
j
= e
i
e
j
= 2o(e
i
⊗e
j
), with 1 ≤ i ≤ j ≤ n are a basis in o(1 ⊗1) with dimension
1
2
n(n +1).
The tensors e
i
∧e
j
= 2/(e
i
⊗e
j
), with 1 ≤ i ≤ j ≤ n are a basis in /(1 ⊗1) with dimension
1
2
n(n −1).
The complete antisymmetric tensor ε is given by: ε
ijk
ε
klm
= δ
il
δ
jm
−δ
im
δ
jl
.
The permutationoperators e
pqr
are deﬁned by: e
123
= e
231
= e
312
= 1, e
213
= e
132
= e
321
= −1, for all other
combinations e
pqr
= 0. There is a connection with the ε tensor: ε
pqr
= g
−1/2
e
pqr
and ε
pqr
= g
1/2
e
pqr
.
7.6 Outer product
Let α ∈ Λ
k
(1) and β ∈ Λ
l
(1). Than α ∧ β ∈ Λ
k+l
(1) is deﬁned by:
α ∧ β =
(k +l)!
k!l!
/(α ⊗β)
If α and β ∈ Λ
1
(1) = 1
∗
holds: α ∧ β = α ⊗β −β ⊗α
46 Mathematics Formulary by ir. J.C.A. Wevers
The outer product can be written as: (a
b)
i
= ε
ijk
a
j
b
k
, a
b = G
−1
∗(Ga ∧ G
b ).
Take a,
b, c,
d ∈ IR
4
. Than (dt ∧ dz)(a,
b ) = a
0
b
4
− b
0
a
4
is the oriented surface of the projection on the tzplane
of the parallelogram spanned by a and
b.
Further
(dt ∧ dy ∧ dz)(a,
b, c) = det
a
0
b
0
c
0
a
2
b
2
c
2
a
4
b
4
c
4
is the oriented 3dimensional volume of the projection on the tyzplane of the parallelepiped spanned by a,
b and c.
(dt ∧ dx ∧ dy ∧ dz)(a,
b, c,
d) = det(a,
b, c,
d) is the 4dimensional volume of the hyperparellelepiped spanned by
a,
b, c and
d.
7.7 The Hodge star operator
Λ
k
(1) and Λ
n−k
(1) have the same dimension because
n
k
=
n
n−k
for 1 ≤ k ≤ n. Dim(Λ
n
(1)) = 1. The choice
of a basis means the choice of an oriented measure of volume, a volume µ, in 1. We can gauge µ so that for an
orthonormal basis e
i
holds: µ(e
i
) = 1. This basis is than by deﬁnition positive oriented if µ =
ˆ
e
1
∧
ˆ
e
2
∧...∧
ˆ
e
n
= 1.
Because both spaces have the same dimension one can ask if there exists a bijection between them. If 1 has no extra
structure this is not the case. However, such an operation does exist if there is an inner product deﬁned on 1 and the
corresponding volume µ. This is called the Hodge star operator and denoted by ∗. The following holds:
∀
w∈Λ
k
(V)
∃
∗w∈Λ
k−n
(V)
∀
θ∈Λ
k
(V)
θ ∧ ∗w = (θ, w)
λ
µ
For an orthonormal basis in IR
3
holds: the volume: µ = dx ∧ dy ∧ dz, ∗dx ∧ dy ∧ dz = 1, ∗dx = dy ∧ dz,
∗dz = dx ∧ dy, ∗dy = −dx ∧ dz, ∗(dx ∧ dy) = dz, ∗(dy ∧ dz) = dx, ∗(dx ∧ dz) = −dy.
For a Minkowski basis in IR
4
holds: µ = dt ∧ dx ∧ dy ∧ dz, G = dt ⊗ dt − dx ⊗ dx − dy ⊗ dy − dz ⊗ dz, and
∗dt ∧ dx ∧ dy ∧ dz = 1 and ∗1 = dt ∧ dx ∧ dy ∧ dz. Further ∗dt = dx ∧ dy ∧ dz and ∗dx = dt ∧ dy ∧ dz.
7.8 Differential operations
7.8.1 The directional derivative
The directional derivative in point a is given by:
L
a
f =<a, df >= a
i
∂f
∂x
i
7.8.2 The Liederivative
The Liederivative is given by:
(L
v
w)
j
= w
i
∂
i
v
j
−v
i
∂
i
w
j
7.8.3 Christoffel symbols
To each curvelinear coordinate system u
i
we add a system of n
3
functions Γ
i
jk
of u, deﬁned by
∂
2
x
∂u
i
∂u
k
= Γ
i
jk
∂x
∂u
i
These are Christoffel symbols of the second kind. Christoffel symbols are no tensors. The Christoffel symbols of the
second kind are given by:
i
jk
:= Γ
i
jk
=
∂
2
x
∂u
k
∂u
j
, dx
i
Chapter 7: Tensor calculus 47
with Γ
i
jk
= Γ
i
kj
. Their transformation to a different coordinate system is given by:
Γ
i
j
k
= A
i
i
A
j
j
A
k
k
Γ
i
jk
+A
i
i
(∂
j
A
i
k
)
The ﬁrst term in this expression is 0 if the primed coordinates are cartesian.
There is a relation between Christoffel symbols and the metric:
Γ
i
jk
=
1
2
g
ir
(∂
j
g
kr
+∂
k
g
rj
−∂
r
g
jk
)
and Γ
α
βα
= ∂
β
(ln(
[g[)).
Lowering an index gives the Christoffel symbols of the ﬁrst kind: Γ
i
jk
= g
il
Γ
jkl
.
7.8.4 The covariant derivative
The covariant derivative ∇
j
of a vector, covector and of rank2 tensors is given by:
∇
j
a
i
= ∂
j
a
i
+ Γ
i
jk
a
k
∇
j
a
i
= ∂
j
a
i
−Γ
k
ij
a
k
∇
γ
a
α
β
= ∂
γ
a
α
β
−Γ
ε
γβ
a
α
ε
+ Γ
α
γε
a
ε
β
∇
γ
a
αβ
= ∂
γ
a
αβ
−Γ
ε
γα
a
εβ
−Γ
ε
γβ
a
αε
∇
γ
a
αβ
= ∂
γ
a
αβ
+ Γ
α
γε
a
εβ
+ Γ
β
γε
a
αε
Ricci’s theorem:
∇
γ
g
αβ
= ∇
γ
g
αβ
= 0
7.9 Differential operators
The Gradient
is given by:
grad(f) = G
−1
df = g
ki
∂f
∂x
i
∂
∂x
k
The divergence
is given by:
div(a
i
) = ∇
i
a
i
=
1
√
g
∂
k
(
√
g a
k
)
The curl
is given by:
rot(a) = G
−1
∗ d Ga = −ε
pqr
∇
q
a
p
= ∇
q
a
p
−∇
p
a
q
The Laplacian
is given by:
∆(f) = div grad(f) = ∗d ∗ df = ∇
i
g
ij
∂
j
f = g
ij
∇
i
∇
j
f =
1
√
g
∂
∂x
i
√
g g
ij
∂f
∂x
j
48 Mathematics Formulary by ir. J.C.A. Wevers
7.10 Differential geometry
7.10.1 Space curves
We limit ourselves to IR
3
with a ﬁxed orthonormal basis. A point is represented by the vector x = (x
1
, x
2
, x
3
). A
space curve is a collection of points represented by x = x(t). The arc length of a space curve is given by:
s(t) =
t
t
0
dx
dτ
2
+
dy
dτ
2
+
dz
dτ
2
dτ
The derivative of s with respect to t is the length of the vector dx/dt:
ds
dt
2
=
dx
dt
,
dx
dt
The osculation plane in a point P of a space curve is the limiting position of the plane through the tangent of the
plane in point P and a point Q when Q approaches P along the space curve. The osculation plane is parallel with
˙
x(s). If
¨
x = 0 the osculation plane is given by:
y = x +λ
˙
x +µ
¨
x so det(y −x,
˙
x,
¨
x) = 0
In a bending point holds, if
...
x= 0:
y = x +λ
˙
x +µ
...
x
The tangent has unit vector
=
˙
x, the main normal unit vector n =
¨
x and the binormal
b =
˙
x
¨
x. So the main
normal lies in the osculation plane, the binormal is perpendicular to it.
Let P be a point and Q be a nearby point of a space curve x(s). Let ∆ϕ be the angle between the tangents in P
and Q and let ∆ψ be the angle between the osculation planes (binormals) in P and Q. Then the curvature ρ and the
torsion τ in P are deﬁned by:
ρ
2
=
dϕ
ds
2
= lim
∆s→0
∆ϕ
∆s
2
, τ
2
=
dψ
ds
2
and ρ > 0. For plane curves ρ is the ordinary curvature and τ = 0. The following holds:
ρ
2
= (
,
) = (
¨
x,
¨
x) and τ
2
= (
˙
b,
˙
b)
Frenet’s equations express the derivatives as linear combinations of these vectors:
˙
= ρn ,
˙
n = −ρ
+τ
b ,
˙
b = −τn
From this follows that det(
˙
x,
¨
x,
...
x ) = ρ
2
τ.
Some curves and their properties are:
Screw line τ/ρ =constant
Circle screw line τ =constant, ρ =constant
Plane curves τ = 0
Circles ρ =constant, τ = 0
Lines ρ = τ = 0
7.10.2 Surfaces in IR
3
A surface in IR
3
is the collection of end points of the vectors x = x(u, v), so x
h
= x
h
(u
α
). On the surface are 2
families of curves, one with u =constant and one with v =constant.
The tangent plane in a point P at the surface has basis:
c
1
= ∂
1
x and c
2
= ∂
2
x
Chapter 7: Tensor calculus 49
7.10.3 The ﬁrst fundamental tensor
Let P be a point of the surface x = x(u
α
). The following two curves through P, denoted by u
α
= u
α
(t),
u
α
= v
α
(τ), have as tangent vectors in P
dx
dt
=
du
α
dt
∂
α
x ,
dx
dτ
=
dv
β
dτ
∂
β
x
The ﬁrst fundamental tensor of the surface in P is the inner product of these tangent vectors:
dx
dt
,
dx
dτ
= (c
α
, c
β
)
du
α
dt
dv
β
dτ
The covariant components w.r.t. the basis c
α
= ∂
α
x are:
g
αβ
= (c
α
, c
β
)
For the angle φ between the parameter curves in P: u = t, v =constant and u =constant, v = τ holds:
cos(φ) =
g
12
√
g
11
g
22
For the arc length s of P along the curve u
α
(t) holds:
ds
2
= g
αβ
du
α
du
β
This expression is called the line element.
7.10.4 The second fundamental tensor
The 4 derivatives of the tangent vectors ∂
α
∂
β
x = ∂
α
c
β
are each linear independent of the vectors c
1
, c
2
and
N,
with
N perpendicular to c
1
and c
2
. This is written as:
∂
α
c
β
= Γ
γ
αβ
c
γ
+h
αβ
N
This leads to:
Γ
γ
αβ
= (c
γ
, ∂
α
c
β
) , h
αβ
= (
N, ∂
α
c
β
) =
1
det [g[
det(c
1
, c
2
, ∂
α
c
β
)
7.10.5 Geodetic curvature
A curve on the surface x(u
α
) is given by: u
α
= u
α
(s), than x = x(u
α
(s)) with s the arc length of the curve. The
length of
¨
x is the curvature ρ of the curve in P. The projection of
¨
x on the surface is a vector with components
p
γ
= ¨ u
γ
+ Γ
γ
αβ
˙ u
α
˙ u
β
of which the length is called the geodetic curvature of the curve in p. This remains the same if the surface is curved
and the line element remains the same. The projection of
¨
x on
N has length
p = h
αβ
˙ u
α
˙ u
β
and is called the normal curvature of the curve in P. The theorem of Meusnier states that different curves on the
surface with the same tangent vector in P have the same normal curvature.
A geodetic line of a surface is a curve on the surface for which in each point the main normal of the curve is the
same as the normal on the surface. So for a geodetic line is in each point p
γ
= 0, so
d
2
u
γ
ds
2
+ Γ
γ
αβ
du
α
ds
du
β
ds
= 0
50 Mathematics Formulary by ir. J.C.A. Wevers
The covariant derivative ∇/dt in P of a vector ﬁeld of a surface along a curve is the projection on the tangent plane
in P of the normal derivative in P.
For two vector ﬁelds v(t) and w(t) along the same curve of the surface follows Leibniz’ rule:
d(v, w)
dt
=
v,
∇ w
dt
+
w,
∇v
dt
Along a curve holds:
∇
dt
(v
α
c
α
) =
dv
γ
dt
+ Γ
γ
αβ
du
α
dt
v
β
c
γ
7.11 Riemannian geometry
The Riemann tensor R is deﬁned by:
R
µ
ναβ
T
ν
= ∇
α
∇
β
T
µ
−∇
β
∇
α
T
µ
This is a
1
3
tensor with n
2
(n
2
−1)/12 independent components not identically equal to 0. This tensor is a measure
for the curvature of the considered space. If it is 0, the space is a ﬂat manifold. It has the following symmetry
properties:
R
αβµν
= R
µναβ
= −R
βαµν
= −R
αβνµ
The following relation holds:
[∇
α
, ∇
β
]T
µ
ν
= R
µ
σαβ
T
σ
ν
+R
σ
ναβ
T
µ
σ
The Riemann tensor depends on the Christoffel symbols through
R
α
βµν
= ∂
µ
Γ
α
βν
−∂
ν
Γ
α
βµ
+ Γ
α
σµ
Γ
σ
βν
−Γ
α
σν
Γ
σ
βµ
In a space and coordinate system where the Christoffel symbols are 0 this becomes:
R
α
βµν
=
1
2
g
ασ
(∂
β
∂
µ
g
σν
−∂
β
∂
ν
g
σµ
+∂
σ
∂
ν
g
βµ
−∂
σ
∂
µ
g
βν
)
The Bianchi identities are: ∇
λ
R
αβµν
+∇
ν
R
αβλµ
+∇
µ
R
αβνλ
= 0.
The Ricci tensor is obtained by contracting the Riemann tensor: R
αβ
≡ R
µ
αµβ
, and is symmetric in its indices:
R
αβ
= R
βα
. The Einstein tensor G is deﬁned by: G
αβ
≡ R
αβ
−
1
2
g
αβ
. It has the property that ∇
β
G
αβ
= 0. The
Ricciscalar is R = g
αβ
R
αβ
.
Chapter 8
Numerical mathematics
8.1 Errors
There will be an error in the solution if a problem has a number of parameters which are not exactly known. The
dependency between errors in input data and errors in the solution can be expressed in the condition number c. If
the problem is given by x = φ(a) the ﬁrstorder approximation for an error δa in a is:
δx
x
=
aφ
(a)
φ(a)
δa
a
The number c(a) = [aφ
(a)[/[φ(a)[. c < 1 if the problem is wellconditioned.
8.2 Floating point representations
The ﬂoating point representation depends on 4 natural numbers:
1. The basis of the number system β,
2. The length of the mantissa t,
3. The length of the exponent q,
4. The sign s.
Than the representation of machine numbers becomes: rd(x) = s m β
e
where mantissa m is a number with t
βbased numbers and for which holds 1/β ≤ [m[ < 1, and e is a number with q βbased numbers for which holds
[e[ ≤ β
q
−1. The number 0 is added to this set, for example with m = e = 0. The largest machine number is
a
max
= (1 −β
−t
)β
β
q
−1
and the smallest positive machine number is
a
min
= β
−β
q
The distance between two successive machine numbers in the interval [β
p−1
, β
p
] is β
p−t
. If x is a real number and
the closest machine number is rd(x), than holds:
rd(x) = x(1 +ε) with [ε[ ≤
1
2
β
1−t
x = rd(x)(1 +ε
) with [ε
[ ≤
1
2
β
1−t
The number η :=
1
2
β
1−t
is called the machineaccuracy, and
ε, ε
≤ η
x −rd(x)
x
≤ η
An often used 32 bits ﬂoat format is: 1 bit for s, 8 for the exponent and 23 for de mantissa. The base here is 2.
51
52 Mathematics Formulary by ir. J.C.A. Wevers
8.3 Systems of equations
We want to solve the matrix equation Ax =
b for a nonsingular A, which is equivalent to ﬁnding the inverse matrix
A
−1
. Inverting a nn matrix via Cramer’s rule requires too much multiplications f(n) with n! ≤ f(n) ≤ (e−1)n!,
so other methods are preferable.
8.3.1 Triangular matrices
Consider the equation Ux = c where U is a rightupper triangular, this is a matrix in which U
ij
= 0 for all j < i,
and all U
ii
= 0. Than:
x
n
= c
n
/U
nn
x
n−1
= (c
n−1
−U
n−1,n
x
n
)/U
n−1,n−1
.
.
.
.
.
.
x
1
= (c
1
−
n
¸
j=2
U
1j
x
j
)/U
11
In code:
for (k = n; k > 0; k)
{
S = c[k];
for (j = k + 1; j < n; j++)
{
S = U[k][j] * x[j];
}
x[k] = S / U[k][k];
}
This algorithm requires
1
2
n(n + 1) ﬂoating point calculations.
8.3.2 Gauss elimination
Consider a general set Ax =
b. This can be reduced by Gauss elimination to a triangular form by multiplying the
ﬁrst equation with A
i1
/A
11
and than subtract it from all others; now the ﬁrst column contains all 0’s except A
11
.
Than the 2nd equation is subtracted in such a way from the others that all elements on the second row are 0 except
A
22
, etc. In code:
for (k = 1; k <= n; k++)
{
for (j = k; j <= n; j++) U[k][j] = A[k][j];
c[k] = b[k];
for (i = k + 1; i <= n; i++)
{
L = A[i][k] / U[k][k];
for (j = k + 1; j <= n; j++)
{
A[i][j] = L * U[k][j];
}
b[i] = L * c[k];
}
}
Chapter 8: Numerical mathematics 53
This algorithm requires
1
3
n(n
2
− 1) ﬂoating point multiplications and divisions for operations on the coefﬁcient
matrix and
1
2
n(n −1) multiplications for operations on the righthand terms, whereafter the triangular set has to be
solved with
1
2
n(n + 1) operations.
8.3.3 Pivot strategy
Some equations have to be interchanged if the corner elements A
11
, A
(1)
22
, ... are not all = 0 to allow Gauss elimina
tion to work. In the following, A
(n)
is the element after the nth iteration. One method is: if A
(k−1)
kk
= 0, than search
for an element A
(k−1)
pk
with p > k that is = 0 and interchange the pth and the nth equation. This strategy fails only
if the set is singular and has no solution at all.
8.4 Roots of functions
8.4.1 Successive substitution
We want to solve the equation F(x) = 0, so we want to ﬁnd the root α with F(α) = 0.
Many solutions are essentially the following:
1. Rewrite the equation in the form x = f(x) so that a solution of this equation is also a solution of F(x) = 0.
Further, f(x) may not vary too much with respect to x near α.
2. Assume an initial estimation x
0
for α and obtain the series x
n
with x
n
= f(x
n−1
), in the hope that lim
n→∞
x
n
=
α.
Example: choose
f(x) = β −ε
h(x)
g(x)
= x −
F(x)
G(x)
than we can expect that the row x
n
with
x
0
= β
x
n
= x
n−1
−ε
h(x
n−1
)
g(x
n−1
)
converges to α.
8.4.2 Local convergence
Let α be a solution of x = f(x) and let x
n
= f(x
n−1
) for a given x
0
. Let f
(x) be continuous in a neighbourhood
of α. Let f
(α) = A with [A[ < 1. Than there exists a δ > 0 so that for each x
0
with [x
0
−α[ ≤ δ holds:
1. lim
n→∞
n
n
= α,
2. If for a particular k holds: x
k
= α, than for each n ≥ k holds that x
n
= α. If x
n
= α for all n than holds
lim
n→∞
α −x
n
α −x
n−1
= A , lim
n→∞
x
n
−x
n−1
x
n−1
−x
n−2
= A , lim
n→∞
α −x
n
x
n
−x
n−1
=
A
1 −A
The quantity A is called the asymptotic convergence factor, the quantity B = −
10
log [A[ is called the asymptotic
convergence speed.
54 Mathematics Formulary by ir. J.C.A. Wevers
8.4.3 Aitken extrapolation
We deﬁne
A = lim
n→∞
x
n
−x
n−1
x
n−1
−x
n−2
A converges to f
(a). Than the row
α
n
= x
n
+
A
n
1 −A
n
(x
n
−x
n−1
)
will converge to α.
8.4.4 Newton iteration
There are more ways to transform F(x) = 0 into x = f(x). One essential condition for them all is that in a
neighbourhood of a root α holds that [f
(x)[ < 1, and the smaller f
(x), the faster the series converges. A general
method to construct f(x) is:
f(x) = x −Φ(x)F(x)
with Φ(x) = 0 in a neighbourhood of α. If one chooses:
Φ(x) =
1
F
(x)
Than this becomes Newtons method. The iteration formula than becomes:
x
n
= x
n−1
−
F(x
n−1
)
F
(x
n−1
)
Some remarks:
• This same result can also be derived with Taylor series.
• Local convergence is often difﬁcult to determine.
• If x
n
is far apart from α the convergence can sometimes be very slow.
• The assumption F
(α) = 0 means that α is a simple root.
For F(x) = x
k
−a the series becomes:
x
n
=
1
k
(k −1)x
n−1
+
a
x
k−1
n−1
This is a wellknown way to compute roots.
The following code ﬁnds the root of a function by means of Newton’s method. The root lies within the interval
[x1, x2]. The value is adapted until the accuracy is better than ±eps. The function funcd is a routine that
returns both the function and its ﬁrst derivative in point x in the passed pointers.
float SolveNewton(void (*funcd)(float, float*, float*), float x1, float x2, float eps)
{
int j, max_iter = 25;
float df, dx, f, root;
root = 0.5 * (x1 + x2);
for (j = 1; j <= max_iter; j++)
{
(*funcd)(root, &f, &df);
dx = f/df;
Chapter 8: Numerical mathematics 55
root = dx;
if ( (x1  root)*(root  x2) < 0.0 )
{
perror("Jumped out of brackets in SolveNewton.");
exit(1);
}
if ( fabs(dx) < eps ) return root; /* Convergence */
}
perror("Maximum number of iterations exceeded in SolveNewton.");
exit(1);
return 0.0;
}
8.4.5 The secant method
This is, in contrast to the two methods discussed previously, a twostep method. If two approximations x
n
and x
n−1
exist for a root, than one can ﬁnd the next approximation with
x
n+1
= x
n
−F(x
n
)
x
n
−x
n−1
F(x
n
) −F(x
n−1
)
If F(x
n
) and F(x
n−1
) have a different sign one is interpolating, otherwise extrapolating.
8.5 Polynomial interpolation
A base for polynomials of order n is given by Lagrange’s interpolation polynomials:
L
j
(x) =
n
¸
l=0
l=j
x −x
l
x
j
−x
l
The following holds:
1. Each L
j
(x) has order n,
2. L
j
(x
i
) = δ
ij
for i, j = 0, 1, ..., n,
3. Each polynomial p(x) can be written uniquely as
p(x) =
n
¸
j=0
c
j
L
j
(x) with c
j
= p(x
j
)
This is not a suitable method to calculate the value of a ploynomial in a given point x = a. To do this, the Horner
algorithm is more usable: the value s =
¸
k
c
k
x
k
in x = a can be calculated as follows:
float GetPolyValue(float c[], int n)
{
int i; float s = c[n];
for (i = n  1; i >= 0; i)
{
s = s * a + c[i];
}
return s;
}
After it is ﬁnished s has value p(a).
56 Mathematics Formulary by ir. J.C.A. Wevers
8.6 Deﬁnite integrals
Almost all numerical methods are based on a formula of the type:
b
a
f(x)dx =
n
¸
i=0
c
i
f(x
i
) +R(f)
with n, c
i
and x
i
independent of f(x) and R(f) the error which has the form R(f) = Cf
(q)
(ξ) for all common
methods. Here, ξ ∈ (a, b) and q ≥ n + 1. Often the points x
i
are chosen equidistant. Some common formulas are:
• The trapezoid rule: n = 1, x
0
= a, x
1
= b, h = b −a:
b
a
f(x)dx =
h
2
[f(x
0
) +f(x
1
)] −
h
3
12
f
(ξ)
• Simpson’s rule: n = 2, x
0
= a, x
1
=
1
2
(a +b), x
2
= b, h =
1
2
(b −a):
b
a
f(x)dx =
h
3
[f(x
0
) + 4f(x
1
) +f(x
2
)] −
h
5
90
f
(4)
(ξ)
• The midpoint rule: n = 0, x
0
=
1
2
(a +b), h = b −a:
b
a
f(x)dx = hf(x
0
) +
h
3
24
f
(ξ)
The interval will usually be split up and the integration formulas be applied to the partial intervals if f varies much
within the interval.
A Gaussian integration formula is obtained when one wants to get both the coefﬁcients c
j
and the points x
j
in an
integral formula so that the integral formula gives exact results for polynomials of an order as high as possible. Two
examples are:
1. Gaussian formula with 2 points:
h
−h
f(x)dx = h
¸
f
−h
√
3
+f
h
√
3
+
h
5
135
f
(4)
(ξ)
2. Gaussian formula with 3 points:
h
−h
f(x)dx =
h
9
¸
5f
−h
3
5
+ 8f(0) + 5f
h
3
5
+
h
7
15750
f
(6)
(ξ)
8.7 Derivatives
There are several formulas for the numerical calculation of f
(x):
• Forward differentiation:
f
(x) =
f(x +h) −f(x)
h
−
1
2
hf
(ξ)
Chapter 8: Numerical mathematics 57
• Backward differentiation:
f
(x) =
f(x) −f(x −h)
h
+
1
2
hf
(ξ)
• Central differentiation:
f
(x) =
f(x +h) −f(x −h)
2h
−
h
2
6
f
(ξ)
• The approximation is better if more function values are used:
f
(x) =
−f(x + 2h) + 8f(x +h) −8f(x −h) +f(x −2h)
12h
+
h
4
30
f
(5)
(ξ)
There are also formulas for higher derivatives:
f
(x) =
−f(x + 2h) + 16f(x +h) −30f(x) + 16f(x −h) −f(x −2h)
12h
2
+
h
4
90
f
(6)
(ξ)
8.8 Differential equations
We start with the ﬁrst order DE y
(x) = f(x, y) for x > x
0
and initial condition y(x
0
) = x
0
. Suppose we ﬁnd
approximations z
1
, z
2
, ..., z
n
for y(x
1
), y(x
2
),..., y(x
n
). Than we can derive some formulas to obtain z
n+1
as
approximation for y(x
n+1
):
• Euler (single step, explicit):
z
n+1
= z
n
+hf(x
n
, z
n
) +
h
2
2
y
(ξ)
• Midpoint rule (two steps, explicit):
z
n+1
= z
n−1
+ 2hf(x
n
, z
n
) +
h
3
3
y
(ξ)
• Trapezoid rule (single step, implicit):
z
n+1
= z
n
+
1
2
h(f(x
n
, z
n
) +f(x
n+1
, z
n+1
)) −
h
3
12
y
(ξ)
RungeKutta methods are an important class of singlestep methods. They work so well because the solution y(x)
can be written as:
y
n+1
= y
n
+hf(ξ
n
, y(ξ
n
)) with ξ
n
∈ (x
n
, x
n+1
)
Because ξ
n
is unknown some “measurements” are done on the increment function k = hf(x, y) in well chosen
points near the solution. Than one takes for z
n+1
− z
n
a weighted average of the measured values. One of the
possible 3rd order RungeKutta methods is given by:
k
1
= hf(x
n
, z
n
)
k
2
= hf(x
n
+
1
2
h, z
n
+
1
2
k
1
)
k
3
= hf(x
n
+
3
4
h, z
n
+
3
4
k
2
)
z
n+1
= z
n
+
1
9
(2k
1
+ 3k
2
+ 4k
3
)
and the classical 4th order method is:
k
1
= hf(x
n
, z
n
)
k
2
= hf(x
n
+
1
2
h, z
n
+
1
2
k
1
)
k
3
= hf(x
n
+
1
2
h, z
n
+
1
2
k
2
)
k
4
= hf(x
n
+h, z
n
+k
3
)
z
n+1
= z
n
+
1
6
(k
1
+ 2k
2
+ 2k
3
+k
4
)
Often the accuracy is increased by adjusting the stepsize for each step with the estimated error. Step doubling is
most often used for 4th order RungeKutta.
58 Mathematics Formulary by ir. J.C.A. Wevers
8.9 The fast Fourier transform
The Fourier transform of a function can be approximated when some discrete points are known. Suppose we have
N successive samples h
k
= h(t
k
) with t
k
= k∆, k = 0, 1, 2, ..., N − 1. Than the discrete Fourier transform is
given by:
H
n
=
N−1
¸
k=0
h
k
e
2πikn/N
and the inverse Fourier transform by
h
k
=
1
N
N−1
¸
n=0
H
n
e
−2πikn/N
This operation is order N
2
. It can be faster, order N
2
log(N), with the fast Fourier transform. The basic idea is
that a Fourier transform of length N can be rewritten as the sum of two discrete Fourier transforms, each of length
N/2. One is formed from the evennumbered points of the original N, the other from the oddnumbered points.
This can be implemented as follows. The array data[1..2*nn] contains on the odd positions the real and on the
even positions the imaginary parts of the input data: data[1] is the real part and data[2] the imaginary part of
f
0
, etc. The next routine replaces the values in data by their discrete Fourier transformed values if isign = 1,
and by their inverse transformed values if isign = −1. nn must be a power of 2.
#include <math.h>
#define SWAP(a,b) tempr=(a);(a)=(b);(b)=tempr
void FourierTransform(float data[], unsigned long nn, int isign)
{
unsigned long n, mmax, m, j, istep, i;
double wtemp, wr, wpr, wpi, wi, theta;
float tempr, tempi;
n = nn << 1;
j = 1;
for (i = 1; i < n; i += 2)
{
if ( j > i )
{
SWAP(data[j], data[i]);
SWAP(data[j+1], data[i+1]);
}
m = n >> 1;
while ( m >= 2 && j > m )
{
j = m;
m >>= 1;
}
j += m;
}
mmax = 2;
while ( n > mmax ) /* Outermost loop, is executed log2(nn) times */
{
istep = mmax << 1;
theta = isign * (6.28318530717959/mmax);
wtemp = sin(0.5 * theta);
wpr = 2.0 * wtemp * wtemp;
wpi = sin(theta);
Chapter 8: Numerical mathematics 59
wr = 1.0;
wi = 0.0;
for (m = 1; m < mmax; m += 2)
{
for (i = m; i <= n; i += istep) /* DanielsonLanczos equation */
{
j = i + mmax;
tempr = wr * data[j]  wi * data[j+1];
tempi = wr * data[j+1] + wi * data[j];
data[j] = data[i]  tempr;
data[j+1] = data[i+1]  tempi;
data[i] += tempr;
data[i+1] += tempi;
}
wr = (wtemp = wr) * wpr  wi * wpi + wr;
wi = wi * wpr + wtemp * wpi + wi;
}
mmax=istep;
}
}
c 1999, 2008 J.C.A. Wevers Dear reader,
Version: May 4, 2008
This document contains 66 pages with mathematical equations intended for physicists and engineers. It is intended to be a short reference for anyone who often needs to look up mathematical equations. This document can also be obtained from the author, Johan Wevers (johanw@vulcan.xs4all.nl). It can also be found on the WWW on http://www.xs4all.nl/˜johanw/index.html. This document is Copyright by J.C.A. Wevers. All rights reserved. Permission to use, copy and distribute this unmodiﬁed document by any means and for any purpose except proﬁt purposes is hereby granted. Reproducing this document by any means, included, but not limited to, printing, copying existing prints, publishing by electronic or other means, implies full agreement to the above nonproﬁtuse clause, unless upon explicit prior written permission of the author. The C code for the rootﬁnding via Newtons method and the FFT in chapter 8 are from “Numerical Recipes in C ”, 2nd Edition, ISBN 0521431085.
A The Mathematics Formulary is made with teTEX and LTEX version 2.09.
If you prefer the notation in which vectors are typefaced in boldface, uncomment the redeﬁnition of the \vec command and recompile the ﬁle. If you ﬁnd any errors or have any comments, please let me know. I am always open for suggestions and possible corrections to the mathematics formulary. Johan Wevers
Contents
Contents 1 Basics 1.1 Goniometric functions . . . . . . . . . . . . . . 1.2 Hyperbolic functions . . . . . . . . . . . . . . . 1.3 Calculus . . . . . . . . . . . . . . . . . . . . . . 1.4 Limits . . . . . . . . . . . . . . . . . . . . . . . 1.5 Complex numbers and quaternions . . . . . . . . 1.5.1 Complex numbers . . . . . . . . . . . . 1.5.2 Quaternions . . . . . . . . . . . . . . . . 1.6 Geometry . . . . . . . . . . . . . . . . . . . . . 1.6.1 Triangles . . . . . . . . . . . . . . . . . 1.6.2 Curves . . . . . . . . . . . . . . . . . . 1.7 Vectors . . . . . . . . . . . . . . . . . . . . . . 1.8 Series . . . . . . . . . . . . . . . . . . . . . . . 1.8.1 Expansion . . . . . . . . . . . . . . . . . 1.8.2 Convergence and divergence of series . . 1.8.3 Convergence and divergence of functions 1.9 Products and quotients . . . . . . . . . . . . . . 1.10 Logarithms . . . . . . . . . . . . . . . . . . . . 1.11 Polynomials . . . . . . . . . . . . . . . . . . . . 1.12 Primes . . . . . . . . . . . . . . . . . . . . . . . I 1 1 1 2 3 3 3 3 4 4 4 4 5 5 5 6 7 7 7 7 9 9 9 9 9 10 11 12 12 12 12 13 13 14 14 14 15 15 16 17 17 18 18 18
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . .
2 Probability and statistics 2.1 Combinations . . . . 2.2 Probability theory . . 2.3 Statistics . . . . . . . 2.3.1 General . . . 2.3.2 Distributions 2.4 Regression analyses .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
3 Calculus 3.1 Integrals . . . . . . . . . . . . . . . . . . 3.1.1 Arithmetic rules . . . . . . . . . 3.1.2 Arc lengts, surfaces and volumes . 3.1.3 Separation of quotients . . . . . . 3.1.4 Special functions . . . . . . . . . 3.1.5 Goniometric integrals . . . . . . . 3.2 Functions with more variables . . . . . . 3.2.1 Derivatives . . . . . . . . . . . . 3.2.2 Taylor series . . . . . . . . . . . 3.2.3 Extrema . . . . . . . . . . . . . . 3.2.4 The operator . . . . . . . . . . 3.2.5 Integral theorems . . . . . . . . . 3.2.6 Multiple integrals . . . . . . . . . 3.2.7 Coordinate transformations . . . . 3.3 Orthogonality of functions . . . . . . . . 3.4 Fourier series . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . I
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . .
. .12 The convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . 4. 6. . . . . . . . . . . . . . . . .2 Complex integration . . . . 5. . . . . . . . . . . . . . . . . . . . . . . . .2 Residue . . . . .1 Basic operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 The Wronskian . . . . . . . . . . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . Wevers 4 Differential equations 4. . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . . . .11 The Laplace transformation . . . . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . 4. . . . . . . . .5. .8 The associated Laguerre equation . . . . . .2. . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . .2. . . 4. . . . . . .2 Some special cases . . . . . .1 Functions of complex variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Properties of Bessel functions . . . . . . . . . . . . . . . . . . . . .1 Vector spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Frobenius’ method . 6. . . . . . . . . . . . . . . . .2 Basis . . . 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10 Inner product spaces . . . . . 4.1 Cauchy’s integral formula . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. . . . . .4 SturmLiouville equations . . . .14. . . . . . . . . . . . . . .1 First order linear DE . . . . . . . . . . . . . . . . . .C. . . . . . . . . . . . . . . . . . . . . . .5 Plane and line . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . .1. . . . . . . . . . . . . . . . . . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 Linear partial differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14. . 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . 4. . . . . . . . . . . . . . . . . . . . . .9 Hermite . . . . . . . .10 Chebyshev . . . . . . . . . . . . . . . . . . .1. . . . . . . . . . . 6. . . . . . . . . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 The associated Legendre equation . . . . . . . . . .5. . . . . . . . . . . . . 6 . . . . . .3 Nonlinear differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13 Systems of linear differential equations . . .3 Potential theory and Green’s theorem Linear algebra 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6. . . . . .3 Legendre’s DE .14 Quadratic forms . . . . .4 Laurent series . . . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. . . . .1 Linear differential equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . . . . . . 5.6 Coordinate transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11 Weber . . . . . . . . . . . . . .2 Special cases . . . . . . . . . . . . . . . . . 5. . . . .A. . . . . . .II Mathematics Formulary door J.3 Matrix calculus . . . . .2 Quadratic surfaces in I 3 . 5. . . . . . . . . 4. . . . . . . . . . .2. . . . . . . . . . . . . 4. . . . . . . . . .9 Homogeneous coordinates . . . . . . . . . . 4. . . . . . . . . . . . . . . 4. . . . . . . . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 Quadratic forms in I 2 . . . . . . . . . . . . . 5. . 5. . . . . . . . . . . . 5. . . . . . . . . . . . . . . . . 5. . . . . . . . 5. . . . . . . . . . . . . . . . . . . . . . .2 Euler . . . . . . . . . . . . . . . . . 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Second order linear DE . . .3. . . . . . . . . 4. . . .2. . . . . .2 Matrix equations . . . . . .1 General . . . . . . . . . . . . . . . . . . . . . . . . . .7 Laguerre’s equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2. . . . 20 20 20 20 21 21 21 21 22 22 22 22 23 23 24 24 24 24 24 25 25 25 25 27 29 29 29 29 29 30 31 31 32 32 32 35 36 36 37 37 38 38 38 39 39 39 39 40 41 41 42 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . R Complex function theory 6. . . . . . .7 Eigen values . . . . . . . . . . . . .5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 Transformation types . .3 Analytical functions deﬁnied by series 6. . . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Linear transformations . . . . . . . . . . . . . . . . . . . . . . . . .4 Power series substitution . . . . . . 4. . . . . . . . . . . . .5 Solutions for Bessel’s equation . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . . . .2. . . . . . . . . . . . . . . . . . .1. . . . . . . . . . . . . . . . . . . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. . . . . . . . . . . . . . . . . .5 Jordan’s theorem . 4. 5. . R 5.
. . . . . . . . . . . . . . . . . . . . .5 Polynomial interpolation .7 The Hodge star operator . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Floating point representations . . . . . . . . . .4. . . . . . . . . . . . . .3 Inner product . . . . . . .11 Riemannian geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9 Differential operators . . . . . . . . . . .10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8. . . . . . . . . . . . .1 Successive substitution 8. .2 The Liederivative . . . . . . . . . . . . . .2 Surfaces in I 3 . . . . . . . 7. . . . . . . . . . .10. . . .8. . . . . . . 7. . . . . . . . . . . . . . . . Wevers III 7 Tensor calculus 7. . . . . . . . . . . . . .8. . . . . . . . . . . . .10 Differential geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Aitken extrapolation . . . . . . . . . . . . . . .1 Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . .6 Outer product . . . . . . . . . . . .3. . . . . . . . . . . 7. . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . .5 Symmetric and antisymmetric tensors . . . . . . . . . . . . . . . . . . . . . . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .Mathematics Formulary by J. . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .C. . . . .1 Space curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Christoffel symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 The second fundamental tensor 7. R 7. . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . 7. . . . . .3 Systems of equations . . .5 The secant method . . . . . . . . . . . . . . . . .4 Tensor product . . .5 Geodetic curvature . . 7. . . . . . . . . 7. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . . . . . .1 The directional derivative . . . . . . . . . . . . . . . . . . . . . . .2 Local convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 Derivatives . . . . . . 8. . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Gauss elimination . . . . . . . . . . . . . . . . . . 8. .3 Pivot strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10. .9 The fast Fourier transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . .A. . . . .8. . . . . . . . . . . . . . . . . . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . . .4 Roots of functions . 7. . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . . . .8 Differential operations . . . . . . . . . . .8. . . . . . . . . 8. . . . . . . . . . . . . . . . . 7. . . . .1 Triangular matrices . . . . . . .4. . . . . . . . . . . . . . . . . . . . . . . . . 8. . . . . . . . . 8. . . . . . . . . . . . . . . . .10. . . . . . . . . . . . . . . . . . . . . . . . . . . . .10. . . . . . . . . . . . . . . . .3 The ﬁrst fundamental tensor . . . . .6 Deﬁnite integrals . . . . . . . . . . . . . . . . . . . . . 8. .8 Differential equations . . .4. . 7. . . . . . . . . . . . . . . . . . . 8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8. . . . . . . . . 43 43 44 44 45 45 45 46 46 46 46 46 47 47 48 48 48 49 49 49 50 51 51 51 52 52 52 53 53 53 53 54 54 55 55 56 56 57 58 8 Numerical mathematics 8. . . . . . . . . . . . . . 8. . . 8. 7. . . . . . . . . . . . . . . . . . . . . . . . .1 Vectors and covectors . . . . . . .4 The covariant derivative . . . . . . . .4 Newton iteration . 8. . . . . . . . . . . . . . . . . . . . . . . . . .2 Tensor algebra . . . . . . . .3. . .
IV Mathematics Formulary door J. Wevers .A.C.
. 2 sinh(x) cosh(x) The hyperbolic functions are deﬁned by: sinh(x) = cosh(x) = tanh(x) = From this follows that cosh2 (x) − sinh2 (x) = 1. sin(arccos(x)) = 1 − x2 ⇒ ⇒ . sin(a ± b) = sin(a) cos(b) ± cos(a) sin(b) tan(a ± b) = The sum formulas are: sin(p) + sin(q) sin(p) − sin(q) cos(p) + cos(q) cos(p) − cos(q) From these equations can be derived that 2 cos2 (x) = 1 + cos(2x) sin(π − x) = sin(x) 1 sin( 2 π − x) = cos(x) Conclusions from equalities: x = a ± 2kπ or x = (π − a) ± 2kπ. cos(a ± b) = cos(a) cos(b) sin(a) sin(b) .1 Goniometric functions cos(φ) = xp . k ∈ I N x = a ± 2kπ or x = −a ± 2kπ π tan(x) = tan(a) ⇒ x = a ± kπ and x = ± kπ 2 The following relations exist between the inverse goniometric functions: sin(x) = sin(a) cos(x) = cos(a) arctan(x) = arcsin √ x x2 +1 = arccos √ 1 x2 +1 . tan(φ) = sin2 (x) + cos2 (x) = 1 and cos−2 (x) = 1 + tan2 (x).Chapter 1 Basics 1. 2 ex + e−x . sin(φ) = yp . Further holds: arsinh(x) = ln x + x2 + 1 . arcosh(x) = arsinh( x2 − 1) 1 . .2 Hyperbolic functions ex − e−x . 2 sin2 (x) = 1 − cos(2x) cos(π − x) = − cos(x) cos( 1 π − x) = sin(x) 2 = 2 sin( 1 (p + q)) cos( 1 (p − q)) 2 2 = 2 cos( 1 (p + q)) sin( 1 (p − q)) 2 2 1 = 2 cos( 2 (p + q)) cos( 1 (p − q)) 2 = −2 sin( 1 (p + q)) sin( 1 (p − q)) 2 2 tan(a) ± tan(b) 1 tan(a) tan(b) yp xp For the goniometric ratios for a point p on the unit circle holds: 1.
deﬁned by f inv (f (x)) = x. holds at point P = (x. J. f (x)): df inv (y) dy Chain rule: if f = f (g(x)). d(xy) = xdy + ydx .A. Wevers 1. then is lim o f (x) f (x) = lim g(x) x→a g (x) .3 Calculus df f (x + h) − f (x) = lim dx h→0 h The derivative of a function is deﬁned as: Derivatives obey the following algebraic rules: d(x ± y) = dx ± dy .2 Mathematics Formulary by ir. then holds df dg df = dx dg dx Further. d x y = ydx − xdy y2 For the derivative of the inverse function f inv (y). for the derivatives of products of functions holds: n · P df (x) dx =1 P (f · g) (n) = k=0 n (n−k) (k) f ·g k For the primitive function F (x) holds: F (x) = f (x).C. An overview of derivatives and primitives is: y = f (x) axn 1/x a ax ex a log(x) ln(x) sin(x) cos(x) tan(x) sin−1 (x) sinh(x) cosh(x) arcsin(x) arccos(x) arctan(x) (a + x2 )−1/2 (a2 − x2 )−1 dy/dx = f (x) anxn−1 −x−2 0 ax ln(a) ex (x ln(a))−1 1/x cos(x) − sin(x) cos−2 (x) − sin−2 (x) cos(x) cosh(x) sinh(x) √ 1/ √ − x2 1 −1/ 1 − x2 (1 + x2 )−1 −x(a + x2 )−3/2 2x(a2 + x2 )−2 (1 + (y )2 )3/2 y  x→a f (x)dx a(n + 1)−1 xn+1 ln x ax ax / ln(a) ex (x ln(x) − x)/ ln(a) x ln(x) − x − cos(x) sin(x) − ln  cos(x) ln  tan( 1 x) 2 cosh(x) sinh(x) √ x arcsin(x) + √1 − x2 x arccos(x) − 1 − x2 x arctan(x) − 1 ln(1 + x2 ) 2 √ ln x + a + x2  1 ln (a + x)/(a − x) 2a The curvature ρ of a curve is given by: ρ = The theorem of De ’l Hˆ pital: if f (a) = 0 and g(a) = 0.
By deﬁnition holds: i2 = −1.Chapter 1: Basics 3 1. b. a x→∞ x→0 x x lim arcsin(x) =1 .5.5. x↓0 x→∞ = en lnp (x) ln(x + a) = 0 . x→0 x→0 x→0 k→0 x x x lim xa ln(x) = 0 . Every complex number can be written as z = z exp(iϕ). = z1  · z2 (cos(ϕ1 + ϕ2 ) + i sin(ϕ1 + ϕ2 )) z1  = (cos(ϕ1 − ϕ2 ) + i sin(ϕ1 − ϕ2 )) z2  1. lim =a . d ∈ I and i2 = j 2 = k 2 = −1. x→0 x xp = 0 als a > 1. with a. The complex conjugate of z is deﬁned as z = z ∗ := a − bi. lim x→∞ 1.4 lim Limits lim 1+ n x x ex − 1 tan(x) sin(x) = 1 . b the imaginary part of z. jk = −kj = i and ki = −ik = j. c. Further follows from this that e±ix = cos(x) ± i sin(x). x→∞ ax lim lim √ x x=1 x→0 lim a1/x − 1 = ln(a) . lim = 1 . The products of R i. ln(z) = ln(z) ± 2nπi. Products and quotients of complex numbers can be written as: z1 · z2 z1 z2 The following can be derived: z1 + z2  ≤ z1  + z2  . j. z1 − z2  ≥  z1  − z2   And from z = r exp(iθ) follows: ln(z) = ln(r) + iθ. z = a2 + b2 .2 Quaternions Quaternions are deﬁned as: z = a + bi + cj + dk. with tan(ϕ) = b/a. so eiz = 0∀z. lim = 1 .5 Complex numbers and quaternions 1.1 Complex numbers √ The complex number z = a + bi with a and b ∈ I a is the real part. . R. Also the theorem of De Moivre follows from this: (cos(ϕ) + i sin(ϕ))n = cos(nϕ) + i sin(nϕ). Further holds: (a + bi)(c + di) = (ac − bd) + i(ad + bc) (a + bi) + (c + di) = a + c + i(b + d) a + bi (ac + bd) + i(bc − ad) = c + di c2 + d2 Goniometric functions can be written as complex exponents: sin(x) cos(x) = = 1 ix (e − e−ix ) 2i 1 ix (e + e−ix ) 2 From this follows that cos(ix) = cosh(x) and sin(ix) = i sinh(x). lim (1 + k)1/k = e . k with each other are given by ij = −ji = k.
The external product is in I 3 deﬁned by: R ay bz − az by a × b = az bx − ax bz = ax by − ay bx ex ax bx ey ay by ez az bz Further holds: a × b  = a  · b  sin(ϕ). y = a(1 + cos(t)) Epicycloid: if a small circle with radius a rolls along a big circle with radius R.6. and a × (b × c ) = (a · c )b − (a · b )c.2 Curves Cycloid: if a circle with radius a rolls along a straight line. For each triangle holds: α + β + γ = 180◦ . It has the following parameterequation in polar coordinates: r = 2a[1 − cos(ϕ)].4 Mathematics Formulary by ir. β is opposite to b and γ opposite to c. y = −a cos a R−a t + (R − a) cos(t) a A hypocycloid with a = R is called a cardioid. 1.C.6. The cosine rule is: a2 = b2 +c2 −2bc cos(α). α is the angle opposite to a.6 Geometry 1.7 Vectors ai bi = a  · b  cos(ϕ) i The inner product is deﬁned by: a · b = where ϕ is the angle between a and b.1 Triangles The sine rule is: a b c = = sin(α) sin(β) sin(γ) Here. s(s − a)(s − b)(s − c) with ha the perpendicular on 1. . Wevers 1.A. J. the trajectory of a point on this circle has the following parameter equation: x = a(t + sin(t)) . the trajectory of a point on the small circle has the following parameter equation: x = a sin R+a t + (R + a) sin(t) . the trajectory of a point on the small circle has the following parameter equation: x = a sin R−a t + (R − a) sin(t) . y = a cos a R+a t + (R + a) cos(t) a Hypocycloid: if a small circle with radius a rolls inside a big circle with radius R. Further holds: 1 tan( 2 (α + β)) a+b = 1 a−b tan( 2 (α − β)) 1 The surface of a triangle is given by 2 ab sin(γ) = 1 aha = 2 1 a and s = 2 (a + b + c).
. 1 π2 = . 6 k=1 (−1)n+1 π2 = . k!(n − k)! n By subtracting the series k=0 rk and r n k=0 rk one ﬁnds: n rk = k=0 ∞ 1 − rn+1 1−r 1 . n n→∞ n un also converges.8.8 Series 1. n4 90 n=1 ∞ ∞ k 2 = 1 n(n + 1)(2n + 1) .Chapter 1: Basics 5 1. The expansion of a function around the point a is given by the Taylor series: f (x) = f (a) + (x − a)f (a) + where the remainder is given by: Rn (h) = (1 − θ)n and is subject to: M hn+1 mhn+1 ≤ Rn (h) ≤ (n + 1)! (n + 1)! From this one can deduce that (1 − x)α = n=0 ∞ (x − a)n (n) (x − a)2 f (a) + · · · + f (a) + R 2 n! hn (n+1) f (θh) n! α n x n 1 π6 = n6 945 n=1 (−1)n+1 = ln(2) n n=1 (−1)n+1 π3 = (2n − 1)3 32 n=1 ∞ ∞ ∞ One can derive that: 1 π2 = . n2 6 n=1 n ∞ 1 π4 = . un is divergent. n2 12 n=1 ∞ 1 = 2−1 4n n=1 ∞ 1 2 . (2n − 1)4 96 n=1 1.8.1 Expansion The Binomium of Newton is: (a + b)n = k=0 n n n−k k a b k where n k := n! .2 Convergence and divergence of series If n un  converges. If lim un = 0 then An alternating series of which the absolute values of the terms drop monotonously to 0 is convergent (Leibniz). 1−r and for r < 1 this gives the geometric series: k=0 N rk = The arithmetic series is given by: n=0 1 (a + nV ) = a(N + 1) + 2 N (N + 1)V . (2n − 1)2 8 n=1 ∞ 1 π4 = .
Than holds: {fn } is uniform convergent if x→∞ lim fn − f = 0. J.C. than f (x) is differentiable in x = a. y)dxdy D series integral ∂f /∂y continuous fy (x.A. y)dx := F . then is un un divergent if L > 1 and convergent if n 1. then n fn is convergent. un continuous f is continuous fn can be integrated. un (x) and F (y) = a f (x. un conv. than un is also convergent. np n=1 vn are both divergent or both convergent. than b un is uniform convergent. ln(un + 1) is convergent. n If un > 0 ∀n then is n un convergent if If un = cn xn the radius of convergence ρ of n ∞ un is given by: 1 = lim ρ n→∞ n cn  = lim n→∞ cn+1 . Wevers If ∞ p f (x)dx < ∞. un ∞ W Weierstrass’ test: if We deﬁne S(x) = n=N is convergent. y)dx .and lower limit are equal: lim f (x) = lim f (x). Than it can be proved that: Theorem For rows series integral rows C Demands on W fn continuous.conv → φ un ∈C−1 . un can be integrated f is continuous {fn } ∈C−1 .8.3 Convergence and divergence of functions f (x) is continuous in x = a only if the upper . x↑a x↓a We deﬁne: f n→∞ W := sup(f (x) x ∈ W ). or: ∀(ε > 0)∃(N )∀(n ≥ N ) fn − f < ε. {fn } uniform convergent S(x) uniform convergent. F dy = f = φ(x) S (x) = Fy = un (x) Sdx = un dx f (x. if n un = p. n→∞ nn . un u.c. Than holds on W f is continuous S is continuous F is continuous fn can be integrated. n→∞ n n n If L is deﬁned by: L = lim L < 1. {fn } uniform convergent S(x) is uniform convergent. than the following is true: if p > 0 than un and vn n p = 0 holds: if vn is convergent. or by: L = lim n→∞ un+1 . and lim fn (x) = f (x). If f (x) is continuous in a and: lim f (x) = lim f (x). f (x)dx = lim fn dx n→∞ I series integral rows S can be integrated.6 Mathematics Formulary by ir. {fn } unif. This is written as: x↑a x↓a f (a− ) = f (a+ ). cn The series If: lim 1 is convergent if p > 1 and divergent if p ≤ 1.
Further holds: a2n − b2n = a2n−1 ± a2n−2 b + a2n−3 b2 ± · · · ± b2n−1 .9 Products and quotients For a. Proof: N suppose that the collection of primes P would be ﬁnite. a±b a2n+1 − b2n+1 = a+b n a2n−k b2k k=0 (a ± b)(a2 ± ab + b2 ) = a3 ± b3 .12 Primes p∈P A prime is a number ∈ I that can only be divided by itself and 1. Each polynomial p(z) of order n ≥ 1 has at least one root in C . than p∗ is also a root. c. for polynomials with order ≥ 5 there does not exist a general analytical solution. There are an inﬁnite number of primes. b.Chapter 1: Basics 7 1. For logarithms with base e one writes ln(x). Polynomials up to and including order 4 R have a general analytical solution. This is a contradiction. b. For a. ab = ba. (a + b)(a − b) = a2 + b2 .10 Logarithms Deﬁnition: a log(x) = b ⇔ ab = x. log(a) − log(b) = log(a/b). b.11 Polynomials n Equations of the type ak xk = 0 k=0 have n roots which may be equal to each other. 1. log(a) + log(b) = log(ab). d ∈ I and a = 0 holds: the 3rd order equation ax3 + bx2 + cx + d = 0 has the general analytical R solution: x1 x2 = x∗ 3 3ac − b2 b − 9a2 K 3a √ K 3ac − b2 b 3 = − + − +i 2 18a2 K 3a 2 = K− K+ 3ac − b2 9a2 K 1/3 with K = 9abc − 27da2 − 2b3 + 54a3 √ √ 3 4ac3 − c2 b2 − 18abcd + 27a2 d2 + 4db3 18a2 1. c. d ∈ I holds: R The distributive property: (a + b)(c + d) = ac + ad + bc + bd The associative property: a(bc) = b(ac) = c(ab) and a(b + c) = ab + ac The commutative property: a + b = b + a. c ∈ I and a = 0 holds: the 2nd order equation ax2 + bx + c = 0 has the general solution: R √ −b ± b2 − 4ac x= 2a For a. than holds q = 1(p) and so Q cannot be written as a product of primes from P . . Rules: log(xn ) = n log(x). If all ak ∈ I holds: when x = p with p ∈ C a root. a3 ± b3 = a2 a+b ba + b2 1. than construct the number q = 1 + p.
9689. for example 6 = 1 + 2 + 3. If w = 1 for each b. 31.C.8 Mathematics Formulary by ir. N The numbers Mk := 2k − 1 are called Mersenne numbers. 4423. for each b. then n is probably prime. A faster method for large numbers are the 4 Fermat tests. 2281. 1. 17. 127. 521. 3. The ﬁrst known sieve method was developed by Eratosthenes. 7. n is certainly not prime. than holds: π(x) = 1 and x→∞ x/ ln(x) lim For each N ≥ 2 there is a prime between N and 2N . 5. 19. 4253. To check if a given number n is prime one can use a sieve method. 2203. 3.A. 2. Wevers If π(x) is the number of primes ≤ x. 89. The numbers Fk := 2k + 1 with k ∈ I are called Fermat numbers. 9941. 7}. They occur when one searches for perfect numbers. 61. 5. who don’t prove that a number is prime but give a large probability. 3217. 1279. 13. Many Fermat numbers are prime. J. 3. 11213}. Take w(b) = bn−1 mod n. which are numbers n ∈ I which are the sum of their different dividers. There N are 23 Mersenne numbers for k < 12000 which are prime: for k ∈ {2. 607. 107. For each other value of w. Take the ﬁrst 4 primes: b = {2. x→∞ x 2 lim π(x) dt ln(t) =1 .
3.3 Statistics xi /n. The probability P (A ∪ B) that A and B both occur is given by: P (A ∪ B) = P (A) + P (B) − P (A ∩ B). when the total number of elements is N . The probability P (AB) that A occurs. is N! ni ! i The number of possible combinations of k elements from n elements is given by = n! k!(n − k)! 2. The probability P (¬A) that A does not occur is: P (¬A) = 1 − P (A). than holds: P (A ∩ B) = P (A) · P (B). If A and B are independent. is: P (AB) = P (A ∩ B) P (B) 2.1 General The average or mean value x of a collection of values is: x = distribution of x is given by: n i (xi − x )2 σx = i=1 n n σ2 . n−1 When samples are being used the sample variance s is given by s2 = 9 .2 Probability theory n(A) n(U ) The probability P (A) that an event A occurs is deﬁned by: P (A) = where n(A) is the number of events when A occurs and n(U ) the total number of events. given the fact that B occurs.Chapter 2 Probability and statistics 2.1 Combinations n k The number of permutations of p from n is given by n! n = p! (n − p)! p The number of different ways to classify ni elements in i groups. The standard deviation σx in the 2.
3. 4.y) = ∂f σx ∂x 2 + ∂f σy ∂y 2 + ∂f ∂f σxy ∂x ∂y 2. The standard deviation in a variable f (x. The probability for success is p. The Uniform distribution occurs when a random number x is taken from the set a ≤ x ≤ b and is given by: P (x) = 1 x = 2 (b − a) and σ 2 = 1 if a ≤ x ≤ b b−a P (x) = 0 in all other cases (b − a)2 . J. The Poisson distribution is a limiting case of the binomial distribution when p → 0. The Binomial distribution is the distribution describing a sampling with replacement. The Hypergeometric distribution is the distribution describing a sampling without replacement in which the order is irrelevant.2 Distributions 1. 2. The probability P for k successes in n trials is then given by: P (x = k) = The standard deviation is given by σx = n k p (1 − p)n−k k np(1 − p) and the expectation value is ε = np. The probability for k successes in a trial with A possible successes and B possible failures is then given by: A B k n−k P (x = k) = A+B n The expectation value is given by ε = nA/(A + B). λx e−λ P (x) = x! ∞ This distribution is normalized to x=0 P (x) = 1. y) resulting from errors in x and y is: 2 σf (x.3.A. The Normal distribution is a limiting case of the binomial distribution for continuous variables: P (x) = 1 1 √ exp − 2 σ 2π x− x σ 2 5. n → ∞ and also np = λ is constant. 12 .C. Wevers The covariance σxy of x and y is given by:: n (xi − x )(yi − y ) σxy = i=1 n−1 The correlation coefﬁcient rxy of x and y than becomes: rxy = σxy /σx σy .10 Mathematics Formulary by ir.
The Beta distribution is given by: α−1 (1 − x)β−1 P (x) = x if 0 ≤ x ≤ 1 β(α. a and b follow from this. (x. y ) = i xi yi . e ) = i i xi and (e. the following relation holds for a and b with x = (x1 . . . x ) = 0 (y... x ) − a(x. β) P (x) = 0 everywhere else and has the following properties: x = For P (χ2 ) holds: α = V /2 and β = 2. 2 (α + β + 1) α+β (α + β) P (x1 . 1): y − ax − be ∈< x. n 1 The correlation coefﬁcient r is a measure for the quality of a ﬁt. 1. In case of linear regression it is given by: r= n (n x2 −( xy − x y y2 −( y)2 ) x)2 )(n . e ) = n. e ) = 0 with (x. A similar method works for higher order polynomial ﬁts: for a second order ﬁt holds: y − ax2 − bx − ce ∈< x2 . The Gamma distribution is given by: P (x) = xα−1 e−x/β if 0 ≤ y ≤ ∞ β α Γ(α) with α > 0 and β > 0. x2 )dx1 dx2 = x1 x2 α αβ . x2 )dx1 g·P 2. 7. (x. The Weibull distribution is given by: α α P (x) = xα−1 e−x if 0 ≤ x ≤ ∞ ∧ α ∧ β > 0 β P (x) = 0 in all other cases The average is x = β 1/α Γ((α + 1)α) 9.. For a twodimensional distribution holds: P1 (x1 ) = with ε(g(x1 . The distribution has the following properties: x = αβ.. 8. σ 2 = αβ 2 . P2 (x2 ) = P (x1 . x ) − b(e... xn ) and e = (1.. x.. x ) = i x2 . σ2 = .Chapter 2: Probability and statistics 11 6. x2 . x2 ).. . e >⊥ From this follows that the inner products are 0: (y. e >⊥ with x2 = (x2 . x2 )P (x1 . e ) − a(x. e ) − b(e. x2 )) = g(x1 . x2 )dx2 .4 Regression analyses When there exists a relation between the quantities x and y of the form y = ax + b and there is a measured set xi with related yi .
t)ds = The surface A of a solid of revolution is: A = 2π y 1+ dy(x) dx 2 1+ dy(x) dx 2 dx F ds = ˙ F (x(t))x(t)dt ˙ dx x(t) = .3 for the required conditions): x=h(y) x=h(y) d dy x=g(y) f (x.1.1 Integrals 3. surfaces and volumes The arc length of a curve y(x) is given by: = The arc length of a parameter curve F (x(t)) is: = with t= (v. y) dg(y) dh(y) dx − f (g(y).Chapter 3 Calculus 3. t(t))dt = dx 12 . y)dx = x=g(y) ∂f (x.1 Arithmetic rules The primitive function F (x) of f (x) obeys the rule F (x) = f (x).1. t  = 1 ˙ ds x(t) (v1 dx + v2 dy + v3 dz) ˙ (v. With F (x) the primitive of f (x) holds for the deﬁnite integral b f (x)dx = F (b) − F (a) a If u = f (x) holds: b f (b) g(f (x))df (x) = a f (a) g(u)du Partial integration: with F and G the primitives of f and g holds: f (x) · g(x)dx = f (x)G(x) − G(x) df (x) dx dx A derivative can be brought under the intergral sign (see section 1.8. y) + f (h(y).2 Arc lengts. y) ∂y dy dy 3.
1. px + q ((x − a)2 + b2 )n with b > 0 and n ∈ I . (x − a)k p(x) = ((x − b)2 + c2 )n n k=1 Ak x + B ((x − b)2 + c2 )k Recurrent relation: for n = 0 holds: (x2 dx 1 2n − 1 x = + n+1 2 + 1)n + 1) 2n (x 2n (x2 dx + 1)n 3.1.3 Separation of quotients Every rational function P (x)/Q(x) where P and Q are polynomials can be written as a linear combination of functions of the type (x − a)k with k ∈ Z and of functions of the type Z. q) = 0 xp−1 (1 − x)q−1 dx with p and q > 0.4 Special functions Elliptic functions Elliptic functions can be written as a power series as follows: 1 − k 2 sin2 (x) = 1 − 1 1 − k 2 sin2 (x) with n!! = n(n − 2)!!. Further one can derive that ∞ √ π (n) 1 Γ(n + 2 ) = n (2n − 1)!! and Γ (y) = e−x xy−1 lnn (x)dx 2 0 The Beta function The betafunction β(p. The Gamma function The gamma function Γ(y) is deﬁned by: ∞ ∞ (2n − 1)!! k 2n sin2n (x) (2n)!!(2n − 1) n=1 (2n − 1)!! 2n 2n k sin (x) (2n)!! n=1 ∞ =1+ Γ(y) = 0 e−x xy−1 dx One can derive that Γ(y + 1) = yΓ(y) = y!. q) = Γ(p)Γ(q) Γ(p + q) . The beta and gamma functions are related by the following equation: β(p. This is a way to deﬁne faculties for nonintegers.Chapter 3: Calculus 13 The volume V of a solid of revolution is: V =π f 2 (x)dx 3. q) is deﬁned by: 1 β(p. So: N p(x) = (x − a)n n k=1 Ak .
y) is deﬁned by: ∂f ∂x = lim x0 h→0 f (x0 + h. y0 + r sin(α)) − f (x0 .5 Goniometric integrals When solving goniometric integrals it can be useful to change variables. dx = dϕ of cos(ϕ) cos2 (ϕ) These deﬁnite integrals are easily solved: π/2 cosn (x) sinm (x)dx = 0 (n − 1)!!(m − 1)!! · (m + n)!! π/2 when m and n are both even 1 in all other cases Some important integrals are: ∞ xdx π2 = . y0 ) ∂f = lim = ( f. cos(x) = . dx = cos(ϕ)dϕ of x= 1 sin(ϕ) . cos α)) = r↓0 ∂α r f ·v v . y0 ) − f (x0 . J. −∞ −∞ F (x)δ(x)dx = F (0) 3. sin(x) = dx = 1 + t2 1 + t2 1 + t2 √ Each integral of the type R(x.A. ax + 1 e 12a2 ∞ π2 x2 dx = .1.14 Mathematics Formulary by ir. R(x.1 Derivatives The partial derivative with respect to x of a function f (x. (sin α. R(x. x) = ε→0 0 for x > ε 1 when x < ε 2ε Some properties are: ∞ ∞ δ(x)dx = 1 . x + 1)2 (e 3 ∞ x3 dx π4 = x+1 e 15 0 −∞ 0 3.1.2. x2 + 1)dx 1 − x2 )dx x2 − 1)dx : : : x = tan(ϕ) . ax2 + bx + c)dx can be converted into one of the types that were treated in section 3. x) with P (ε. Wevers The Delta function The delta function δ(x) is an inﬁnitely thin peak function with surface 1.2 Functions with more variables 3. After this conversion one can substitute in the integrals of the type: R(x. It can be deﬁned by: δ(x) = lim P (ε. dx = dϕ of cos(ϕ) x2 + 1 = t + x 1 − x2 = 1 − tx x2 − 1 = x − t x = sin(ϕ) .3.C. The following holds if one deﬁnes 1 tan( 2 x) := t: 1 − t2 2t 2dt . y0 ) h The directional derivative in the direction of α is deﬁned by: f (x0 + r cos(α).
y) on a boundary V ∈ I 2 are: R 1. y. If the boundary V is given by ϕ(x. The tangent plane in x0 is given by: fx (x0 )(x − x0 ) + fy (x0 )(y − y0 ) = z − f (x0 ). A boundary is called compact if it is limited and closed. λ is called a multiplicator. b) = m=0 p m p−m ∂ p f (a.Chapter 3: Calculus 15 When one changes to coordinates f (x(u. 2. b) h k m ∂xm ∂y p−m 3. Point (3) is rewritten as follows: possible extrema are points where f (x.2. y) = 0. Possible extrema of f (x. y) = 0 are possible for extrema. y. y. . The same as in I 2 holds in I 3 when the area to be searched is constrained by a compact V .2 Taylor series A function of two variables can be expanded as follows in a Taylor series: f (x0 + h. This is the multiplicator method of Lagrange. v)) holds: ∂f ∂f ∂x ∂f ∂y = + ∂u ∂x ∂u ∂y ∂u If x(t) and y(t) depend only on one parameter t holds: ∂f dx ∂f dy ∂f = + ∂t ∂x dt ∂y dt The total differential df of a function of 3 variables is given by: df = So ∂f ∂f dy ∂f dz df = + + dx ∂x ∂y dx ∂z dx The tangent in point x0 at the surface f (x. z) for points (1) and (2). y) + λ ϕ(x. and V is deﬁned by R R ϕ1 (x. y0 + k) = 1 p! p=0 n h ∂p ∂p +k p p ∂x ∂y f (x0 . z) = 0 and ϕ2 (x. y) is not differentiable. y) = 0 is given by the equation fx (x0 )(x − x0 ) + fy (x0 )(y − y0 ) = 0. v).2. y. than all points where f (x. z) + λ1 ϕ1 (x. y. Points on V where f (x. z) + λ2 ϕ2 (x. y(u. 3. y. z) = 0. z) = 0 for extrema of f (x. y0 ) + R(n) with R(n) the residual error and h ∂p ∂p +k p p ∂x ∂y p f (a. Points where f = 0.3 Extrema When f is continuous on a compact boundary V there exists a global maximum and a global minumum for f on this boundary. ∂f ∂f ∂f dx + dy + dz ∂x ∂y ∂z 3.
2. y. v.4 The operator In cartesian coordinates (x. z) holds: = gradf = ∂ 1 ∂ ∂ er + eϕ + ez ∂r r ∂ϕ ∂z ∂f 1 ∂f ∂f er + eϕ + ez ∂r r ∂ϕ ∂z ar 1 ∂aϕ ∂az ∂ar + + + ∂r r r ∂ϕ ∂z ∂aϕ ∂ar ∂az 1 ∂az − er + − r ∂ϕ ∂z ∂z ∂r 2 2 2 ∂ f 1 ∂f 1 ∂ f ∂ f + + 2 + 2 ∂r2 r ∂r r ∂ϕ2 ∂z div a = curl a = 2 eϕ + ∂aϕ aϕ 1 ∂ar + − ∂r r r ∂ϕ ez f = In spherical coordinates (r. J. The unit vectors are given by: eu = 1 ∂x 1 ∂x 1 ∂x . ev = .C.16 Mathematics Formulary by ir. Wevers 3. w). z) holds: = gradf = ∂ ∂ ∂ ex + ey + ez ∂x ∂y ∂z ∂f ∂f ∂f ex + ey + ez ∂x ∂y ∂z ∂ax ∂ay ∂az + + ∂x ∂y ∂z ∂az ∂ay ∂ax ∂az − ex + − ∂y ∂z ∂z ∂x 2 2 2 ∂ f ∂ f ∂ f + 2 + 2 2 ∂x ∂y ∂z div a = curl a = 2 ey + ∂ay ∂ax − ∂x ∂y ez f = In cylindrical coordinates (r. θ. v. w) can be derived from cartesian coordinates by the transformation x = x(u. ϕ) holds: = gradf = 1 ∂ 1 ∂ ∂ er + eθ + eϕ ∂r r ∂θ r sin θ ∂ϕ 1 ∂f 1 ∂f ∂f er + eθ + eϕ ∂r r ∂θ r sin θ ∂ϕ ∂ar 2ar 1 ∂aθ aθ 1 ∂aϕ + + + + ∂r r r ∂θ r tan θ r sin θ ∂ϕ 1 ∂aϕ aθ 1 ∂aθ 1 ∂ar ∂aϕ aϕ + − er + − − r ∂θ r tan θ r sin θ ∂ϕ r sin θ ∂ϕ ∂r r ∂aθ aθ 1 ∂ar + − eϕ ∂r r r ∂θ 2 ∂f 1 ∂2f 1 ∂f 1 ∂2f ∂2f + + 2 2 + 2 + 2 2 2 ∂r r ∂r r ∂θ r tan θ ∂θ r sin θ ∂ϕ2 div a = curl a = eθ + 2 f = General orthonormal curvilinear coordinates (u. ϕ. The differential operators are than given by: gradf = 1 ∂f 1 ∂f 1 ∂f eu + ev + ew h1 ∂u h2 ∂v h3 ∂w . ew = h1 ∂u h2 ∂v h3 ∂w where the terms hi give normalization to length 1.A.
2. 3. than the surface A inside the curve in I 2 is given by R A= d2 A = dxdy Let the surface A be deﬁned by the function z = f (x.6 Multiple integrals Let A be a closed curve given by f (x. v is an arbitrary vectorﬁeld and φ an arbitrary scalar ﬁeld. y) = 0.5 Integral theorems Some important integral theorems are: Gauss: Stokes for a scalar ﬁeld: Stokes for a vector ﬁeld: this gives: Ostrogradsky: (v · n)d2 A = (φ · et )ds = (v · et )ds = (divv )d3 V (n × gradφ)d2 A (curlv · n)d2 A (curlv · n)d2 A = 0 (n × v )d2 A = (φn )d2 A = (curlv )d3 A (gradφ)d3 V Here the orientable surface d2 A is bounded by the Jordan curve s(t).Chapter 3: Calculus 17 div a = curl a = 2 f = 1 ∂ ∂ ∂ (h2 h3 au ) + (h3 h1 av ) + (h1 h2 aw ) h1 h2 h3 ∂u ∂v ∂w 1 ∂(h3 aw ) ∂(h2 av ) 1 ∂(h1 au ) ∂(h3 aw ) − eu + − h2 h3 ∂v ∂w h3 h1 ∂w ∂u 1 ∂(h2 av ) ∂(h1 au ) − ew h1 h2 ∂u ∂v ∂ h2 h3 ∂f ∂ h3 h1 ∂f ∂ h1 h2 ∂f 1 + + h1 h2 h3 ∂u h1 ∂u ∂v h2 ∂v ∂w h3 ∂w operator are: curl(φv) = φcurlv + (gradφ) × v curl curlv = grad divv − 2 v 2 v ≡ ( 2 v1 . 2 v3 ) ev + Some properties of the div(φv) = φdivv + gradφ · v div(u × v) = v · (curlu) − u · (curlv) div gradφ = 2 φ curl gradφ = 0 div curlv = 0 Here. 2 v2 . y)dxdy = dxdydz . y)dxdy The volume inside a closed surface deﬁned by z = f (x.2. The volume V bounded by A and the xy plane is than given by: V = f (x. 3. y). y) is given by: V = d3 V = f (x.
Let the set gi be orthogonal. bn = 1 L −L f (t) sin nπt L dt .18 Mathematics Formulary by ir. w): V = In I 2 holds: R f (x. v).A.C. A set functions fi is orthonormal if (fi .2. A periodical function f (x) with period 2L can be written as: ∞ f (x) = a0 + n=1 an cos nπx nπx + bn sin L L Due to the orthogonality follows for the coefﬁcients: L L L a0 = 1 2L −L f (t)dt . than it follows: ci = f. Wevers 3. z)dxdydz = f (x(u)) ∂x dudvdw ∂u ∂x = ∂u xu yu xv yv Let the surface A be deﬁned by z = F (x. v. y. gi (gi . When one chooses the orthogonal basis (cos(nx). v. g) = a p(x)f (x)g(x)dx The norm f follows from: f 2 = (f. y) = X(u. F (x. an = 1 L −L f (t) cos nπt L dt . w) through the transformation x(u. f ). by: b (f. sin(nx)) we have a Fourier series. gi ) 3. when using a weight function p(x). y. g) = a f (x)g(x)dx or. y)) 1 + ∂x F 2 + ∂y F 2 dxdy × ∂u ∂v S G G 3.4 Fourier series Each function can be written as a sum of independent base functions. J. ∞ Each function f (x) can be written as a sum of orthogonal functions: f (x) = i=0 ci gi (x) and c2 ≤ f i 2 . fj ) = δij . Than the volume bounded by the xy plane and F is given by: ∂X ∂X f (x)d2 A = f (x(u)) dudv = f (x. b] is given by: (f.7 Coordinate transformations The expressions d2 A and d3 V transform as follows when one changes coordinates to u = (u.3 Orthogonality of functions b The inner product of two functions f (x) and g(x) on the interval [a.
f (a+ ) = lim f (x) x↑a x↓a For continuous functions is 1 2 [f (x+ ) + f (x− )] = f (x).Chapter 3: Calculus 19 A Fourier series can also be written as a sum of complex exponents: ∞ f (x) = n=−∞ cn einx with 1 cn = 2π π f (x)e−inx dx −π ˆ The Fourier transform of a function f (x) gives the transformed function f (ω): ∞ 1 ˆ f (ω) = √ 2π The inverse transformation is given by: f (x)e−iωx dx −∞ ∞ 1 1 f (x+ ) + f (x− ) = √ 2 2π ˆ f (ω)eiωx dω −∞ where f (x+ ) and f (x− ) are deﬁned by the lower .and upper limit: f (a− ) = lim f (x) . .
Chapter 4 Differential equations 4. The solution of the homogeneous equation is given by yH = k exp Suppose that a(x) = a =constant. and than one solves f (x) from this. x When: y (x) + ω 2 y(x) = ωf (x) and y(0) = y (0) = 0 follows: y(x) = 0 f (x) sin(ω(x − t))dt.2 Second order linear DE A differential equation of the second order with constant coefﬁcients is given by: y (x) + ay (x) + by(x) = c(x). a(x)dx 4. 2. λ = µ: a particular solution is: yP = exp(µx) 2. where q(x) is a polynomial of the same order as p(x). If c(x) = p(x) exp(µx) where p(x) is a polynomial there are 3 possibilities: 1. λ2 = µ: yP = q(x) exp(µx). Its homogeneous equation is y (x) + a(x)y(x) = 0. λ1 = λ2 = µ: yP = x2 q(x) exp(µx). λ1 = µ.1 First order linear DE The general solution of a linear differential equation is given by yA = yH + yP . A ﬁrst order differential equation is given by: y (x) + a(x)y(x) = b(x). λ1 = λ2 = λ: than yH = (α + βx) exp(λx). Than one can distinguish two cases: 1. λ2 = µ: yP = xq(x) exp(µx). 2.1 Linear differential equations 4. Substitution of y = exp(λx) leads to the characteristic equation λ2 + aλ + b = 0. There are now 2 possibilities: 1. 3. where yH is the solution of the homogeneous equation and yP is a particular solution. If c(x) = c =constant there exists a particular solution yP = c/b. Suppose b(x) = α exp(µx). λ = µ: a particular solution is: yP = x exp(µx) When a DE is solved by variation of parameters one writes: yP (x) = yH (x)f (x). λ1 .1. 20 .1. Substitution of exp(λx) in the homogeneous equation leads to the characteristic equation λ + a = 0 ⇒ λ = −a. λ1 = λ2 : than yH = α exp(λ1 x) + β exp(λ2 x).
1.. This LDE has at least one solution of the form ∞ yi (x) = xri n=0 an xn with i = 1. 2. The general solution can than be written as y(x) = c1 y1 (x) + c2 y2 (x) and y1 and y2 are linear independent. These are also all solutions of the LDE. Special cases are n = 0.2 Some special cases 4.3 The Wronskian We start with the LDE y (x) + p(x)y (x) + q(x)y(x) = 0 and the two initial conditions y(x0 ) = K0 and y (x0 ) = K1 . r1 − r2 ∈ I : than y(x) = ky1 (x) ln x + y2 (x).1 Frobenius’ method Given the LDE d2 y(x) b(x) dy(x) c(x) + + 2 y(x) = 0 dx2 x dx x with b(x) and c(x) analytical at x = 0. 1. y2 ) = y1 y1 y2 y2 = y1 y2 − y2 y1 y1 and y2 are linear independent if and only if on the interval I when ∃x0 ∈ I so that holds: W (y1 (x0 ).. 2.1. r1 = r2 : than y(x) = y1 (x) ln x + y2 (x). 4. and c(x) = c0 + c1 x + c2 x2 + . y2 (x0 )) = 0. N 3. 2 with r real or complex and chosen so that a0 = 0..Chapter 4: Diﬀerential equations 21 4.2. When one expands b(x) and c(x) as b(x) = b0 + b1 x + b2 x2 + . Z: .. r1 − r2 = Z than y(x) = y1 (x) + y2 (x). it follows for r: r2 + (b0 − 1)r + c0 = 0 There are now 3 possibilities: 1. When p(x) and q(x) are continuous on the open interval I there exists a unique solution y(x) on this interval. 4.4 Power series substitution When a series y = leads to: an xn is substituted in the LDE with constant coefﬁcients y (x) + py (x) + qy(x) = 0 this n(n − 1)an xn−2 + pnan xn−1 + qan xn = 0 n Setting coefﬁcients for equal powers of x equal gives: (n + 2)(n + 1)an+2 + p(n + 1)an+1 + qan = 0 This gives a general relation between the coefﬁcients.. The Wronskian is deﬁned by: W (y1 .
3 Legendre’s DE Given the LDE dy(x) d2 y(x) − 2x + n(n − 1)y(x) = 0 2 dx dx The solutions of this equation are given by y(x) = aPn (x) + by2 (x) where the Legendre polynomials P (x) are deﬁned by: dn (1 − x2 )n Pn (x) = n dx 2n n! (1 − x2 ) 2 For these holds: Pn = 2/(2n + 1). There are now 2 possibilities: x2 1. J.C.2. r1 = r2 : than y(x) = C1 xr1 + C2 xr2 .A. They are of the form: Pl For m > l is Pl m m (ξ) = (1 − ξ 2 )m/2 dm P 0 (ξ) (1 − ξ 2 )m/2 dm+l 2 = (ξ − 1)l 2l l! dξ m dξ m+l (ξ) = 0. Some properties of Pl0 (ξ) zijn: 1 Pl0 (ξ)Pl0 (ξ)dξ = −1 2 δll 2l + 1 ∞ . 2. From this one gets two solutions r1 and r2 . Than follows: (1 − ξ 2 ) d dξ (1 − ξ 2 ) dP (ξ) dξ 2 Ψ = 0 by substitution of + [C(1 − ξ 2 ) − m2 ]P (ξ) = 0 Regular solutions exists only if C = l(l + 1).2 Euler Given the LDE d2 y(x) dy(x) + ax + by(x) = 0 2 dx dx Substitution of y(x) = xr gives an equation for r: r2 + (a − 1)r + b = 0.2. r1 = r2 = r: than y(x) = (C1 ln(x) + C2 )xr .2.22 Mathematics Formulary by ir. Wevers 4. and the Bessel functions of the ﬁrst kind x2 ∞ Jν (x) = xν m=0 22m+ν m!Γ(ν (−1)m x2m + m + 1) .5 Solutions for Bessel’s equation Given the LDE dy(x) d2 y(x) +x + (x2 − ν 2 )y(x) = 0 dx2 dx also called Bessel’s equation.2.4 The associated Legendre equation This equation follows from the θdependent part of the wave equation ξ = cos(θ). 4. l=0 Pl0 (ξ)tl = 1 1 − 2ξt + t2 This polynomial can be written as: π Pl0 (ξ) 1 = π 0 (ξ + ξ 2 − 1 cos(θ))l dθ 4. 4.
But because for n ∈ Z holds: Z Z J−n (x) = (−1)n Jn (x). The general solution of Bessel’s equation is given by y(x) = aJν (x) + bYν (x). Jn+1 (x) − Jn−1 (x) = −2 x dx π 1 Jm (x) = 2π 0 1 exp[i(x sin(θ) − mθ)]dθ = π 0 cos(x sin(θ) − mθ)dθ 4. r x→∞ lim Jn (x) = 2 cos(x − xn ) . The Neumann functions Nm (x) are deﬁnied as: Nm (x) = 1 1 Jm (x) ln(x) + m 2π x ∞ αn x2n n=0 The following holds: lim Jm (x) = xm . where Yν are the Bessel functions of the second kind: Yν (x) = Jν (x) cos(νπ) − J−ν (x) and Yn (x) = lim Yν (x) ν→n sin(νπ) The equation x2 y (x) + xy (x) − (x2 + ν 2 )y(x) = 0 has the modiﬁed Bessel functions of the ﬁrst kind Iν (x) = i−ν Jν (ix) as solution. and also solutions Kν = π[I−ν (x) − Iν (x)]/[2 sin(νπ)].6 Properties of Bessel functions Bessel functions are orthogonal with respect to the weight function p(x) = x. x→0 x→0 x→0 r→∞ lim H(r) = e±ikr eiωt √ . 2 Jn+1 (x) + Jn−1 (x) = The following integral relations hold: 2π 2n dJn (x) Jn (x) . lim N0 (x) = ln(x). lim Nm (x) = x−m for m = 0. Hn (x) = Jn (x) − iYn (x) 4. Sometimes it can be convenient to write the solutions of Bessel’s equation in terms of the Hankel functions (1) (2) Hn (x) = Jn (x) + iYn (x) . πx x→∞ lim J−n (x) = 2 sin(x − xn ) πx 1 with xn = 2 π(n + 1 ).2. J−n (x) = (−1)n Jn (x).2.Chapter 4: Diﬀerential equations 23 for ν := n ∈ I this becomes: N ∞ Jn (x) = xn m=0 22m+n m!(n (−1)m x2m + m)! When ν = Z the solution is given by y(x) = aJν (x) + bJ−ν (x).7 Laguerre’s equation Given the LDE x d2 y(x) dy(x) + (1 − x) + ny(x) = 0 dx2 dx Solutions of this equation are the Laguerre polynomials Ln (x): Ln (x) = ex dn (−1)m n m xn e−x = x n n! dx m! m m=0 ∞ . this does not apply to integers.
given by: Hn (x) = (−1)n exp 1 2 x 2 1 √ dn (exp(− 2 x2 )) = 2n/2 Hen (x 2) n dx Hen (x) = (−1)n (exp x2 √ dn (exp(−x2 )) = 2−n/2 Hn (x/ 2) n dx 4.2.A.10 Chebyshev The LDE (1 − x2 ) has solutions of the form Un (x) = The LDE (1 − x2 ) has solutions Tn (x) = cos(n arccos(x)).11 Weber The LDE Wn (x) + (n + 1 2 1 − 1 x2 )Wn (x) = 0 has solutions: Wn (x) = Hen (x) exp(− 4 x2 ). J.2.8 The associated Laguerre equation Given the LDE d2 y(x) + dx2 m+1 −1 x dy(x) + dx n + 1 (m + 1) 2 x y(x) = 0 Solutions of this equation are the associated Laguerre polynomials Lm (x): n Lm (x) = n (−1)m n! −x −m dn−m e x e−x xn (n − m)! dxn−m 4. Wevers 4.9 Hermite The differential equations of Hermite are: dHn (x) dHen (x) d2 Hen (x) d2 Hn (x) − 2x −x + 2nHn (x) = 0 and + nHen (x) = 0 2 2 dx dx dx dx Solutions of these equations are the Hermite polynomials. 4 4.3 Nonlinear differential equations y y y y y y = a y 2 + b2 = a y 2 − b2 = a b2 − y 2 = a(y 2 + b2 ) = a(y 2 − b2 ) = a(b2 − y 2 ) b−y y = ay b y y y y y y = b sinh(a(x − x0 )) = b cosh(a(x − x0 )) = b cos(a(x − x0 )) = b tan(a(x − x0 )) = b coth(a(x − x0 )) = b tanh(a(x − x0 )) b y= 1 + Cb exp(−ax) Some nonlinear differential equations and a solution are: .2.C. d2 Un (x) dUn (x) − 3x + n(n + 2)Un (x) = 0 dx2 dx sin[(n + 1) arccos(x)] √ 1 − x2 d2 Tn (x) dTn (x) −x + n2 Tn (x) = 0 dx2 dx 4.24 Mathematics Formulary by ir.2.
Chapter 4: Diﬀerential equations 25 4.5. t) = [ϕ(x + ct) + ϕ(x − ct)] + 2 2c x−ct Ψ(ξ)dξ .5 Linear partial differential equations 4. 0)/∂t = Ψ(x) apply. The normalization function m(x) must satisfy b m(x)yi (x)yj (x)dx = δij a When y1 (x) and y2 (x) are two linear independent solutions one can write the Wronskian in this form: W (y1 . than y /y > 0 and the solution has an exponential behaviour. given by: u(x) = y(x) p(x).4 SturmLiouville equations d dx dy(x) dx SturmLiouville equations are second order LDE’s of the form: − p(x) + q(x)y(x) = λm(x)y(x) The boundary conditions are chosen so that the operator L=− d dx p(x) d dx + q(x) is Hermitean.5. the LDE transforms into the normal form: d2 u(x) 1 + I(x)u(x) = 0 with I(x) = 2 dx 4 p (x) p(x) 2 − 1 p (x) q(x) − λm(x) − 2 p(x) p(x) If I(x) > 0. n) ∂n A frequently used solution method for PDE’s is separation of variables: one assumes that the solution can be written as u(x.2 Special cases The wave equation The wave equation in 1 dimension is given by ∂2u ∂2u = c2 2 ∂t2 ∂x When the initial conditions u(x. t) = X(x)T (t). 0) = ϕ(x) and ∂u(x. y2 ) = y1 y1 y2 y2 = C p(x) where C is constant. By changing to another dependent variable u(x). if I(x) < 0. When this is substituted two ordinary DE’s for X(x) and T (t) are obtained. 4. than y /y < 0 and the solution has an oscillatory behaviour.1 General The normal derivative is deﬁned by: ∂u = ( u. 4. the general solution is given by: x+ct 1 1 u(x.
x . t). This gives for v: 2 v(x. x . ω) = 0 This gives as solutions for v: 1. t )P (x. t) = 1 exp 8(πDt)3/2 −(x − x )2 4Dt −(x − x )2 4Dt With initial condition u(x. x . t) = √ 2 πDt In 3 dimensions it reads: P (x. ϕ) = l=0 m=−l Y (θ. x . ϕ) √ r . 0) = δ(x − x ). t) = G f (x )P (x. t)dx The solution of the equation ∂2u ∂u − D 2 = g(x. 2. t − t ) The equation of Helmholtz The equation of Helmholtz is obtained by substitution of u(x. In cartesian coordinates: substitution of v = A exp(ik · x ) gives: v(x ) = with the integrals over k 2 = k 2 . t) = v(x) exp(iωt) in the wave equation. In polar coordinates: ∞ ··· A(k)eik·x dk v(r.A. J. In spherical coordinates: ∞ l 1 [Alm Jl+ 2 (kr) + Blm J−l− 1 (kr)] 2 v(r. These have the property that P (x. t) = dt dx g(x .C. θ. In 1 dimension it reads: 1 exp P (x. x . Wevers The diffusion equation The diffusion equation is: ∂u =D ∂t 2 u Its solutions can be written in terms of the propagators P (x. ϕ) = m=0 (Am Jm (kr) + Bm Nm (kr))eimϕ 3. t) ∂t ∂x is given by u(x. 0) = f (x) the solution is: u(x.26 Mathematics Formulary by ir. ω) + k 2 v(x. x .
The ﬁeld lines of the ﬁeld v(x ) follow from: ˙ x (t) = λv(x ) The ﬁrst theorem of Green is: [u G 2 v + ( u.3 Potential theory and Green’s theorem Subject of the potential theory are the Poisson equation 2 u = −f (x ) where f is a given function. t )ds = ϕ(b ) − ϕ(a ) a In this case there exist functions ϕ and w so that v = gradϕ + curlw.Chapter 4: Diﬀerential equations 27 4. The solution exists if: − R f (x )d3 V = S h(x )d2 A A fundamental solution of the Laplace equation satisﬁes: 2 u(x ) = −δ(x ) This has in 2 dimensions in polar coordinates the following solution: u(r) = ln(r) 2π This has in 3 dimensions in spherical coordinates the following solution: u(r) = 1 4πr . It has a unique solution. The Dirichlet problem is: 2 u(x ) = −f (x ) . The solutions of Laplace’s equation are called harmonic functions. and the Laplace equation 2 u = 0. ∂u(x ) = h(x ) for all x ∈ S. The Neumann problem is: 2 u(x ) = −f (x ) . The solutions of these can often be interpreted as a potential. x ∈ R . When a vector ﬁeld v is given by v = gradϕ holds: b (v. A harmonic function with a normal derivative of 0 on the boundary of an area is constant within that area. x ∈ R . ∂n The solution is unique except for a constant. v)]d3 V = S u ∂v 2 d A ∂n The second theorem of Green is: [u G 2 v−v 2 u]d3 V = S u ∂v ∂u −v ∂n ∂n d2 A A harmonic function which is 0 on the boundary of an area is also 0 within that area. u(x ) = g(x ) for all x ∈ S.5.
ξ ) G(x.28 Mathematics Formulary by ir. ξ )f (x )d3 V − S g(x ) ∂G(x. ξ ) 2 d A ∂n . Wevers The equation 2 v = −δ(x − ξ ) has the solution v(x ) = 1 4πx − ξ  After substituting this in Green’s 2nd theorem and applying the sieve property of the δ function one can derive Green’s 3rd theorem: u(ξ ) = − 1 4π R 2 u r d3 V + 1 4π S 1 ∂u ∂ −u r ∂n ∂n 1 r d2 A The Green function G(x.C. is: u(ξ ) = R 2 u = −f (x ) when on the G(x. The solution of Poisson’s equation boundary S holds: u(x ) = g(x ). ξ ) = 0. ξ ) is deﬁned by: 2 G = −δ(x − ξ ). and on boundary S holds G(x. Than G can be written as: 1 + g(x. J. ξ ) is a solution of Dirichlet’s problem.A. ξ ) = 4πx − ξ  Than g(x.
a2 . 4. independent and 2. The set vectors {an } is linear independent if: λi ai = 0 ⇔ ∀i λi = 0 i The set {an } is a basis if it is 1. V =< a1 .. >= λi ai . ∃e ∈ G so that a ⊗ e = e ⊗ a = a: there exists a unit element. 5. w2 ∈ W holds: λw1 + µw2 ∈ W . For an orthonormal basis holds: (ei .1 Vector spaces G is a group for the operation ⊗ if: 1. . For the ij inverse matrix holds: (A · B)−1 = B −1 · A−1 . 5.2 Basis For an orthogonal basis holds: (ei . a ⊗ b = b ⊗ a the group is called Abelian or commutative.. λ(µa) = (λµ)a. ∀a. (λ + µ)(a + b) = λa + λb + µa + µb. The inverse matrix A−1 has the property that A · A−1 = I and can I be found by diagonalization: (Aij I ∼ (I −1 ). W is an invariant subspace of V for the operator A if ∀w ∈ W holds: Aw ∈ W . ∀a ∈ G∃a ∈ G so that a ⊗ a = e: each element has an inverse. 2. (a ⊗ b) ⊗ c = a ⊗ (b ⊗ c): a group is associative. The transpose of A is deﬁned by: aT = aji . (AB)ij = k aik bkj where r is the number of rows and k the number of columns. Vector spaces form an Abelian group for addition and multiplication: 1 · a = a.Chapter 5 Linear algebra 5.1 Basic operations For the matrix multiplication of matrices A = aij and B = bkl holds with r the row index and k the column index: Ar1 k1 · B r2 k2 = C r1 k2 . b ∈ G ⇒ a ⊗ b ∈ G: a group is closed. I) IAij 29 . W is a linear subspace if ∀w1 . ej ) = δij .3 Matrix calculus 5. If 5.3. For this holds (AB)T = B T AT and (AT )−1 = (A−1 )T . 3. ej ) = cδij .
) dt dt dt dt When the rows of a matrix are considered as vectors the row rank of a matrix is the number of independent vectors in this set. Cramer’s rule for the solution of systems of linear equations is: let the system be written as A · x = b ≡ a1 x1 + . When Aij ∈ I and v1 + iv2 is an eigenvector of A at eigenvalue λ = λ1 + iλ2 . an ) det(A) . . The equation A·x=0 has exactly one solution = 0 if det(A) = 0.. and if det(A) = 0 the solution is 0.. If det(A) = 0 there exists exactly one solution = 0.. than the nullspace N (A) = {0 }... 5. If det(A) = 0 the only solution is 0. kn > If the columns kn of a n × m matrix A are independent. v2 > is an invariant subspace of A... .. a∗2 . . ˜ ˜ ˜ Let A : V → V be the complex extension of the real linear operator A : V → V in a ﬁnite dimensional V . Similar for the column rank. If kn are the columns of A.. .. a∗n ) For the determinant det(A) of a matrix A holds: det(AB) = det(A) · det(B).. . . Een 2 × 2 matrix has determinant: det a b c d = ad − cb The derivative of a matrix is a matrix with the derivatives of the coefﬁcients: daij dAB dA dB dA = and =B +A dt dt dt dt dt The derivative of the determinant is given by: da1 da2 dan d det(A) = D( ...2 Matrix equations We start with the equation A·x=b and b = 0. + an xn = b then xj is given by: xj = D(a1 . Aen >=< k1 . 3... + D(a1 . .C. than the transformed space of A is given by: R(A) =< Ae1 .. an ) + D(a1 . .30 Mathematics Formulary by ir...3....A. than holds: R 1. The row rank equals the column rank for each matrix... v ∗ = v1 − iv2 is an eigenvalue at λ∗ = λ1 − iλ2 ... aj+1 . J. Then A ˜ and A have the same caracteristic equation. 2. b. aj−1 . .. Wevers The inverse of a 2 × 2 matrix is: a b c d −1 = 1 ad − bc d −b −c a The determinant function D = det(A) is deﬁned by: det(A) = D(a∗1 .. Av1 = λ1 v1 − λ2 v2 and Av2 = λ2 v1 + λ1 v2 .. The linear span < v1 . an ) + .
) = (a. x ) = 0 Mirror image in the line < a > Mirror image in the plane (a. V ) = (a. x ) = 0 Equation P (x ) = (a. Let A : W → W deﬁne a linear transformation. x )a/(a. Then the following holds: dim(N (A)) + dim(R(A)) = dim(V ) 5. The distance d between 2 points p and q is given by d(p. Some common linear transformations are: Transformation type Projection on the line < a > Projection on the plane (a. the null space of A. than A is a projection. x ) + k = 0 is R d(p. In I 2 holds: The distance of a point p to the line (a.Chapter 5: Linear algebra 31 5. p ) + b a Similarly in I 3 : The distance of a point p to the plane (a. A← (0 ) = E0 is a linear subspace of V . R . notation: N (A). x) + b = 0 holds. y ) = (x. a ) Q(x ) = x − P (x ) S(x ) = 2P (x ) − x T (x ) = 2Q(x ) − x = x − 2P (x ) For a projection holds: x − PW (x ) ⊥ PW (x ) and PW (x ) ∈ W . q ) = p − q . notation: R(A). x) + k = 0. and for a plane V: (a.4 Linear transformations A transformation A is linear if: A(λx + βy ) = λAx + βAy. If for a transformation A holds: (Ax. Ay ). From this follows that A(V ) is the image space of A. we deﬁne: • If S is a subset of V : A(S) := {Ax ∈ W x ∈ S} • If T is a subset of W : A← (T ) := {x ∈ V A(x ) ∈ T } Than A(S) is a linear subspace of W and the inverse transformation A← (T ) is a linear subspace of V . Ay ) = (Ax.5 Plane and line x = a + λ(b − a ) = a + λr The equation of a line that contains the points a and b is: The equation of a plane is: x = a + λ(b − a ) + µ(c − a ) = a + λr1 + µr2 When this is a plane in I 3 . p ) + k a This can be generalized for I n and C n (theorem from Hesse). x ) + b = 0 is R d(p. The normal vector to V is than: a/a. the normal vector to this plane is given by: R nV = r1 × r2 r1 × r2  A line can also be described by the points for which the line equation : (a.
. ..A. Aβ en α α α δ β α β For the transformation of matrix operators to another coordinate system holds: Aδ = Sλ Aλ Sα ..r. . a basis α into a vector w. The following is true: det(A) = λi and Tr(A) = aii = λi .32 Mathematics Formulary by ir. Eλn ). When λ is an eigen value of A holds: An x = λn x.. Than also holds: (Ax. .. It is given by: α Aβ = (β(Aa1 ).. . J. A vector is transformed via Xα = Sα Xβ .8 Transformation types Isometric transformations A transformation is isometric when: Ax = x . β(Aan )) α where β(x ) is the representation of the vector x w..t.C..6 Coordinate transformations y = Am×n x The linear transformation A from I n → I m is given by (I = I of C ): K K K R where a column of A is the image of a base vector in the original. a basis β. α α α β 5. When W is an invariant subspace if the isometric transformation A with dim(A) < ∞.7 Eigen values Ax = λx The eigenvalue equation with eigenvalues λ can be solved with (A − λI = 0 ⇒ det(A − λI = 0. Wevers 5. Ay ) = (x. This implies that the eigen values of an isometric transformation are given by λ = exp(iϕ) ⇒ λ = 1. .. Aα = Sβ Aβ Sα α α β β β and (AB)λ = Aλ Bα ...t. 5.. β The transformation matrix Sα transforms vectors from coordinate system α into coordinate system β: β Sα := I α = (β(a1 ). α β β α β Further is Aβ = Sα Aα ..r. with S the transformation matrix to this basis.. β(an )) Iβ β α and Sα · Sβ = I I The matrix of a transformation A is than given by: Aβ = Aβ e1 . is given by: Λ = S −1 AS = diag(λ1 . The eigenvalues follow from this I) I) characteristic equation.t. y ). ..r. The matrix Aβ transforms a vector given w. λn ) When 0 is an eigen value of A than E0 (A) = N (A). The matrix of A in a basis of eigenvectors. Aα = Aα Sβ . than also W ⊥ is an invariante subspace. basis β. i i i The eigen values λi are independent of the chosen basis. S = (Eλ1 .
A n × n matrix is unitary if U H U = I It has determinant I. Ay ). The matrix of such a transformation is given by: −1 0 0 0 cos(ϕ) − sin(ϕ) 0 sin(ϕ) cos(ϕ) For all orthogonal transformations O in I 3 holds that O(x ) × O(y ) = O(x × y ). than also holds: (λ1 ) = (λ2 ) = cos(ϕ). A describes a rotation. λ2 = exp(−iϕ). Each isometric transformation in a ﬁnitedimensional complex vector space is unitary. Theorem: for a n × n matrix A the following statements are equivalent: 1.r. A rotation in I 2 through angle ϕ is given by: R R= cos(ϕ) − sin(ϕ) sin(ϕ) cos(ϕ) So the rotation angle ϕ is determined by Tr(A) = 2 cos(ϕ) with 0 ≤ ϕ ≤ π. R I n (n < ∞) can be decomposed in invariant subspaces with dimension 1 or 2 for each orthogonal transformation. Vectors from E−1 are mirrored by A w.r. A 1 1 2 mirroring in I in < (cos( 2 ϕ). A R M linear operator is only symmetric if its matrix w. λ2 = λ∗ = exp(iϕ). All eigenvalues of a symmetric transformation belong to I The different eigenvectors are mutually perpendicular. I. If A and B are orthogonal. A is unitary. The columns of A are an orthonormal set. 2. an arbitrary basis is symmetric. the invariant subspace E−1 . Symmetric transformations A transformation A on I n is symmetric if (Ax. For an orthogonal transformation O holds OT O = I so: OT = O−1 . 3. A rotation over Eλ1 is given by the matrix R 3 1 0 0 0 cos(ϕ) − sin(ϕ) 0 sin(ϕ) cos(ϕ) ⊥ Mirrored orthogonal if det(A) = −1. A matrix A ∈ I n×n is symmetric if A = AT . M .  det(U ) = 1. Than A is: Direct orthogonal if det(A) = +1. The rows of A are an orthonormal set. than AB and A−1 are also orthogonal. and λ1 = exp(iϕ). y ) = (x. than AT = R.t. Than a linear transformation U is unitary if U is isometric and its inverse transformation A← exists. Let λ1 and λ2 be the roots of the characteristic equation. Let A : V → V be orthogonal with dim(V ) < ∞. sin( 2 ϕ)) > is given by: R S= cos(ϕ) sin(ϕ) sin(ϕ) − cos(ϕ) Mirrored orthogonal transformations in I 3 are rotational mirrorings: rotations of axis < a1 > through angle ϕ and R mirror plane < a1 >⊥ . In I 3 holds: λ1 = 1. R Unitary transformations Let V be a complex space on which an inner product is deﬁned. If A is symmetric. A = AH on an orthogonal basis.t. For each matrix B ∈ I m×n holds: B T B is symmetric.Chapter 5: Linear algebra 33 Orthogonal transformations A transformation A is orthogonal if A is isometric and the inverse A← exists.
. y ) = xH y. Hy ). Normal transformations For each linear transformation A in a complex vector space V there exists exactly one linear transformation B so that (Ax.. The eigenvalues of a Hermitian transformation belong to I R. y ) = (x. Deﬁnition: the linear transformation A is normal in a complex vector space V if A∗ A = AA∗ .. Wevers Hermitian transformations A transformation H : V → V with V = C n is Hermitian if (Hx. + Pp = I I. • P1 + . y ) = (A∗ x. than their product AB is Hermitian if: [A.t.. αp so that A = α1 P1 + . A∗ y ) 2. Ay ) = (A∗ Ax.34 Mathematics Formulary by ir. Let the different roots of the characteristic equation of A be βi with multiplicities ni . y ) = (x.r. + αp Pp . Than the dimension of each eigenspace Vi equals ni . B] = AB − BA = 0. Than: 1. If the transformations A and B are Hermitian. . If Eλ if an eigenspace from A than the orthogonal complement Eλ is an invariant subspace of A. with the properties • Pi · Pj = 0 for i = j.r. Eigenvectors of A for different eigenvalues are mutually perpendicular. If A is unitary than holds αi  = 1 ∀i. If A is normal holds: 1.C.. There exist projection transformations Pi .. The following holds: (CD)∗ = D∗ C ∗ . This leads to the spectral mapping theorem: let A be a normal transformation in a complex vector space V with dim(V ) = n. 1 ≤ i ≤ p. This is only the case if for its matrix S w.. By ). These eigenspaces are mutually perpendicular and each vector x ∈ V can be written in exactly one way as x= i xi with xi ∈ Vi This can also be written as: xi = Pi x where Pi is a projection on Vi . + dimPp (V ) = n and complex numbers α1 . The Hermitian conjugated transformation AH of A is: [aij ]H = [a∗ ]. an orthonormal basis holds: A† A = AA† . W. • dimP1 (V ) + . [A. 3. 3. A∗ = A−1 if A is unitary and A∗ = A if A is Hermitian. Len ). For all vectors x ∈ V and a normal transformation A holds: (Ax.A.. This B is called the adjungated transformation of A. An alternative notation is: AH = A† . a basis ei it is given by Lmn = (em . A matrix representation can be coupled with a Hermitian operator L. y ) = (AA∗ x. 2. J. B] is called the commutator of A and B. The inner product of two vectors ji x and y can now be written in the form: (x.t. If A is Hermitian than αi ∈ I ∀i. R . x is an eigenvector of A if and only if x is an eigenvector of A∗ . Notation: B = A∗ .. ⊥ 4.
z axis. Deﬁnition: a commuting set Hermitian transformations is called complete if for each set of two common eigenvectors vi . Z0 . Homogeneous coordinates are derived from cartesian coordinates as follows: X wx x Y wy y = = Z wz z cart w hom w hom so x = X/w. Lemma: if Eλ is the eigenspace for eigenvalue λ from A1 . Rotations of the x.Chapter 5: Linear algebra 35 Complete systems of commuting Hermitian transformations Consider m Hermitian linear transformations Ai in a n dimensional complex inner product space V . β. than Eλ is an invariant subspace of all transformations Ai . Consider m commuting Hermitian matrices Ai . This transformation has no inverse. y. resp. The columns of U are the common eigenvectors of all matrices Aj . Assume they mutually commute. Y0 . vj there exists a transformation Ak so that vi and vj are eigenvectors with different eigenvalues of Ak . In quantum physics one speaks of commuting observables. A perspective projection on image plane z = c with the center of projection in the origin. An extra coordinate is introduced to describe the nonlinearities. w0 ): w0 0 T = 0 0 0 w0 0 0 0 0 w0 0 X0 Y0 Z0 w0 2. Transformations in homogeneous coordinates are described by the following matrices: 1. 5. γ: 1 0 0 0 0 cos α − sin α 0 Rx (α) = 0 sin α cos α 0 Ry (β) = 0 0 0 1 cos γ − sin γ 0 sin γ cos γ 0 Rz (γ) = 0 0 1 0 0 0 cos β 0 − sin β 0 0 0 0 1 0 sin β 1 0 0 cos β 0 0 0 0 0 1 3. Translation along vector (X0 . than Ai x ∈ Eλ . Usually a commuting set is taken as small as possible. If all eigenvalues of a Hermitian linear transformation in a ndimensional complex vector space differ. through angles α.9 Homogeneous coordinates Homogeneous coordinates are used if one wants to combine both rotations and translations in one matrix transformation. than the normalized eigenvector is known except for a phase factor exp(iα). The required number of commuting observables equals the number of quantum numbers required to characterize a state. Theorem. 1 0 0 0 0 1 0 0 P (z = c) = 0 0 1 0 0 0 1/c 0 . Than there exists a unitary matrix U so that all matrices U † Ai U are diagonal. y = Y /w and z = Z/w. This means that if x ∈ Eλ .
b1 . b ) = i=1 a∗ bi i For function spaces holds: b (f. A]. The Laplace transformation is a generalisation of the Fourier transformation. ∞ > and a.36 Mathematics Formulary by ir. ... product is deﬁned by: n (a. . R Than there exists a Laplace transform for f . b2 ∈ V and β1 . The set of vectors is independent if and only if det(G) = 0. (a. Than the Gramian G of this set is given by: Gij = (ai .C. M ∈ I so that for t ≥ t0 holds: f (t) exp(−at) < M .11 The Laplace transformation The class LT exists of functions for which holds: 1.. A > 0 there are no more than a ﬁnite number of discontinuities and each discontinuity has an upper . The following holds: a − b ≤ a+ b ≤ a + b . is deﬁned in C ∞ by: ∞ 2 = 2 a = (a1 . a ).. Let {a1 . A set is orthonormal if (ai . J. b 1 ) + β2 (a. (a.and lower limit. x )2 n→∞ xn − x = 0. aj ) = δij . . a ) ≥ 0 for all a ∈ V . b 2 ) for all a. form an orthonormal row in an inﬁnite dimensional vector space Bessel’s inequality holds: ∞ x The equal sign holds if and only if lim The inner product space 2 2 ≥ i=1 (ei . a ) ∈ I The inner product space C n is the complex vector space on which a complex inner R.. n→∞ 5. β2 ∈ C . e2 . with s ∈ C and t ≥ 0: ∞ F (s) = 0 f (t)e−st dt . β1 b1 + β2 b 2 ) = β1 (a. a ) = 0 if and only if a = 0. 2. Wevers 5. 3. (a.10 Inner product spaces A complex inner product on a complex vector space is deﬁned as follows: 1. a2 .. (a. If e1 . 2. b ) = (b. aj ). a ). and with ϕ the angle between a and b holds: (a. ∃t0 ∈ [0. On each interval [0.)  n=1 an 2 < ∞ A space is called a Hilbert space if it is and if also holds: lim an+1 − an  = 0. an } be a set of vectors in an inner product space V . g) = a f ∗ (t)g(t)dt For each a the length a is deﬁned by: a = (a.A.. Due to (1) holds: (a. The Laplace transform of a function f (t) is. b ) = a · b cos(ϕ).
˙ We start with the equation x = Ax. Equal shapes: if a > 0 than L (f (at)) = 2. λ1 = λ2 : than x(t) = 2. − sn−1 f (0) + sn F (s) The operator L has the following properties: 1. Damping: L (e−at f (t)) = F (s + a) 3. R For some often occurring functions holds: f (t) = tn at e n! eat cos(ωt) eat sin(ωt) δ(t − a) F (s) = L(f (t)) = (s − a)−n−1 s−a (s − a)2 + ω 2 ω (s − a)2 + ω 2 exp(−as) 1 F a s a 5.. . Distribution: f ∗ (g + h) = f ∗ g + f ∗ h 4. L(f ∗ g) = L(f ) · L(g) 3. Commutative: f ∗ g = g ∗ f 5. Translation: If a > 0 and g is deﬁned by g(t) = f (t − a) if t > a and g(t) = 0 for t ≤ a. In the 2 × 2 case holds: 1. f ∗ g ∈LT 2. If s ∈ I than holds (λf ) = L( (f )) and (λf ) = L( (f )). λ1 = λ2 : than x(t) = (ut + v) exp(λt). 5. than holds: L (g(t)) = e−sa L(f (t)).13 Systems of linear differential equations vi exp(λi t). Assume that x = v exp(λt).12 The convolution t The convolution integral is deﬁned by: (f ∗ g)(t) = 0 f (u)g(t − u)du The convolution has the following properties: 1. than follows: Av = λv.Chapter 5: Linear algebra 37 The Laplace transform of the derivative of a function is given by: L f (n) (t) = −f (n−1) (0) − sf (n−2) (0) − .. than is f (t) = f1 ∗ f2 . Homogenity: f ∗ (λg) = λf ∗ g If L(f ) = F1 · F2 .
p = q = −1. • Hyperbolic cylinder: p = d = 1. c are the lengths of the semi axes. 5. Here. r = −1. u = (u. z).1 Quadratic forms in IR2 The general equation of a quadratic form is: xT Ax + 2xT P + S = 0. to maintain the same orientation as the system (x. d = 0. • Doublebladed hyperboloid: r = d = 1. Rank 1: py 2 + qx = d • Parabolic cylinder: p. q = −1. than λ∗ is also an eigenvalue for eigenvector v ∗ . In polar coordinates this can be written as: ep r= 1 − e cos(θ) An ellipse has e < 1. r = −1. .. r = 0. d = 0. Introduce: x = u and y = v. J. Starting with the equation ax2 + 2bxy + cy 2 + dx + ey + f = 0 we have A = ac − b2 . • Double plane: p = 0. b.38 Mathematics Formulary by ir.. . d = 0. Decompose v = u + iw. I) 2. r = d = 0. d = 0. w) should be chosen so that det(S) = +1.14. y. p • Hyperbolic paraboloid: p = r = −1.C. a parabola A = 0 and a hyperbole A < 0.2 Quadratic surfaces in IR3 Rank 3: x2 y2 z2 +q 2 +r 2 =d 2 a b c • Ellipsoid: p = q = r = d = 1. If Λ = S −1 AS = diag(λ1 . • Cone: p = q = 1. q = −1. • Parallel pair of planes: d > 0. Let x = v exp(λt) ⇒ det(A − λ2 I = 0. than the real solutions are c1 [u cos(βt) − w sin(βt)]eαt + c2 [v cos(βt) + u sin(βt)]eαt ¨ There are two solution strategies for the equation x = Ax: 1. a.14 Quadratic forms 5.14. q = 1. so all cross terms are 0. This transforms a ndimensional set of second ˙ ˙ ¨ ˙ ¨ ˙ order equations into a 2ndimensional set of ﬁrst order equations. • Elliptic cylinder: p = q = −1. r = −1. λn ) holds: uT Λu + 2uT P + S = 0. a parabola e = 1 and a hyperbola e > 1. Wevers Assume that λ = α + iβ is an eigenvalue with eigenvector v. p = 0. 5. q = d = 0. • Pair of planes: p = 1. p • Singlebladed hyperboloid: p = q = d = 1. Rank 2: x2 y2 z +q 2 +r 2 =d a2 b c • Elliptic paraboloid: p = q = 1. this leads to x = u and y = v.. An ellipse has A > 0. v. A is a symmetric matrix. q = 0.A. q > 0.
if the integral exists. ∂x ∂y These equations imply that 2 ∂u ∂v =− ∂y ∂x u= 2 v = 0.2. y) the derivative is: f (z) = ∂u ∂v ∂u ∂v +i = −i + ∂x ∂x ∂y ∂y z→a f (z) − f (a) z−a Setting both results equal yields the equations of CauchyRiemann: ∂v ∂u = . than the length L of K is given by: b L= a dx dt 2 + dy dt 2 b b dt = a dz dt = dt a φ (t)dt The derivative of f in point z = a is: f (a) = lim If f (z) = u(x. y) + iv(x. A Jordan curve is a curve that is closed and singular. Than holds for integer m: 1 2πi K dz = (z − a)m 0 if m = 1 1 if m = 1 Theorem: if L is the length of curve K and if f (z) ≤ M for z ∈ K. holds: f (z)dz ≤ M L K 39 . Some deﬁnitions: f is analytical on G if f is continuous and differentiable on G. than. If K is a curve in C with parameter equation z = φ(t) = x(t) + iy(t). Than the integral of f over K is: b f (z)dz = K a ˙ f (φ(t))φ(t)dt f continuous = F (b) − F (a) Lemma: let K be the circle with center a and radius r taken in a positive direction. f is analytical if u and v satisfy these equations.Chapter 6 Complex function theory 6. 6.1 Cauchy’s integral formula Let K be a curve described by z = φ(t) on a ≤ t ≤ b and f (z) is continuous on K. a ≤ t ≤ b.1 Functions of complex variables Complex function theory deals with complex functions of a complex variable.2 Complex integration 6.
which is taken in a positive direction. this means φ(ξ. holds: f (z) 1 f (a) if a inside K dz = 0 if a outside K 2πi z−a K Theorem: let K be a curve (K need not be closed) and let φ(ξ) be continuous on K.and end points.C. Cauchy’s residue proposition is: let f be analytical within and on a Jordan curve K except in a ﬁnite number of singular points ai within K. Than the function f (z) = K φ(ξ)dξ ξ−z is analytical with nth derivative f (n) (z) = n! K φ(ξ)dξ (ξ − z)n+1 Theorem: let K be a curve and G an area. Let B be a Jordan curve. with the following properties: 1. z ∈ G. z) is limited. z) is an analytical function of z on G. Than holds f (z)dz = K K z f (z)dz ⇔ B f (z)dz = 0 By applying the main theorem on eiz /z one can derive that ∞ sin(x) π dx = x 2 0 6. z ∈ G. Than. Otherwise a is a singular point or pole of f (z). Let F (z) = p f (ξ)dξ for all z ∈ G only depend on z and not on the integration path. Let φ(ξ. Than F (z) is analytical on G with F (z) = f (z).A. in singular points it can be both 0 and = 0. z) be deﬁned for ξ ∈ K. For ﬁxed ξ ∈ K. z) ≤ M for ξ ∈ K. J. The residue of f in a is deﬁned by Res f (z) = z=a 1 2πi K f (z)dz where K is a Jordan curve which encloses a in positive direction. if K is taken in a positive direction. 2. Wevers Theorem: let f be continuous on an area G and let p be a ﬁxed point of G. which can be transformed into each other by continous deformation within G. . The residue is 0 in regular points.2 Residue A point a ∈ C is a regular point of a function f (z) if f is analytical in a. Let K and K be two curves with the same starting . holds: 1 2πi K n f (z)dz = k=1 z=ak Res f (z) Lemma: let the function f be analytical in a. This leads to two equivalent formulations of the main theorem of complex integration: let the function f be analytical on an area G.2. than holds: Res z=a f (z) = f (a) z−a This leads to Cauchy’s integral theorem: if F is analytical on the Jordan curve K. φ(ξ.40 Mathematics Formulary by ir. φ(ξ.
• If lim an+1 /an  = L exists. n→∞ If these limits both don’t exist one can ﬁnd R with the formula of CauchyHadamard: 1 = lim sup n an  R n→∞ 6. than R = 1/L.. Than f (z) can be expanded into the Taylor series near a: ∞ f (z) = n=0 cn (z − a)n with cn = f (n) (a) n! valid for z − a < r. • If lim n n→∞ an  = L exists. Than f (z) can be expanded into a Laurent series with center a: ∞ f (z) = n=−∞ cn (z − a)n with cn = 1 2πi K f (w)dw . z)/∂z are continuous functions of ξ on K. z) dξ ∂z Cauchy’s inequality: let f (z) be an analytical function within and on the circle C : z −a = R and let f (z) ≤ M for z ∈ C. Theorem of Laurent: let f be analytical in the circular area G : r < z − a < R. n∈Z Z (w − a)n+1 .4 Laurent series Taylor’s theorem: let f be analytical in an area G and let point a ∈ G has distance r to the boundary of G. the opposite is not necessary. For ﬁxed z ∈ G the functions φ(ξ. ∞ Theorem: let the power series n=0 an z n have a radius of convergence R. If f has a pole of order k in a than c1 . The radius of convergence of the Taylor series is ≥ r. Than the function f (z) = K φ(ξ. z) and ∂φ(ξ.. ck = 0.. Than holds M n! f (n) (a) ≤ n R 6. ck−1 = 0. z)dξ is analytical with derivative f (z) = K ∂φ(ξ. R is the distance to the ﬁrst nonessential singularity.Chapter 6: Complex function theory 41 3. . than R = 1/L.3 Analytical functions deﬁnied by series fn (z) is called pointwise convergent on an area G with sum F (z) if N The series ∀ε>0 ∀z∈G ∃N0 ∈I ∀n>n0 R The series is called uniform convergent if f (z) − n=1 fn (z) < ε N ∀ε>0 ∃N0 ∈I ∀n>n0 ∃z∈G R f (z) − n=1 fn (z) < ε Uniform convergence implies pointwise convergence.
f (a) = 0. f ) = 0. a is an essential singular point of f . We deﬁne the notations Cρ = {zz = ρ. f ) = 0. J. a pole of order 1) in z = a with f (a) f (z) = Res z=a g(z) g (a) 6.e. M (ρ. lim ρM + (ρ. 3. We assume that f (z) is analytical for (z) > 0 with a possible exception of a ﬁnite number of singular points which do not lie on the real axis. f ) = max f (z). Than holds for α > 0 ρ→∞ ρ→∞ + Cρ lim f (z)eiαz dz = 0 Let f be continuous for z ≥ R. f ) = max f (z). ∞ The principal part of a Laurent series is: cases: n=1 c−n (z − a)−n . (z) ≥ 0} and − + − Cρ = {zz = ρ. than ρ→∞ ∞ f (x)dx = 2πi −∞ Resf (z) in (z) > 0 Replace M + by M − in the conditions above and it follows that: ∞ f (x)dx = −2πi −∞ Resf (z) in (z) < 0 Jordan’s lemma: let f be continuous for z ≥ R. f ) = 0 and that the integral exists. If f and g are analytical. The principal part contains a ﬁnite number of terms. The principal part contains an inﬁnite number of terms. Than holds for α < 0 ρ→∞ ρ→∞ − Cρ lim f (z)eiαz dz = 0 Let z = a be a simple pole of f (z) and let Cδ be the half circle z − a = δ.5 Jordan’s theorem + z∈Cρ − z∈Cρ + Residues are often used when solving deﬁnite integrals. One can classify singular points with this. Than is 1 1 lim f (z)dz = 2 Res f (z) z=a δ↓0 2πi Cδ .42 Mathematics Formulary by ir. Deﬁne f (a) = c0 and the series is also valid for z − a < R and f is analytical in a. g(a) = 0. Wevers valid for r < z − a < R and K an arbitrary Jordan curve in G which encloses point a in positive direction.A. Than a is a nonessential singularity. 2. g (a) = 0 than f (z)/g(z) has a simple pole (i. Then. 0 ≤ arg(z − a) ≤ π.C. There are 3 1. such as exp(1/z) for z = 0. taken from a + δ to a − δ. Than the function g(z) = (z − a)k f (z) has a nonessential singularity in a. Than there exists a k ∈ I so that N lim (z − a)k f (z) = c−k = 0. (z) ≥ 0 and lim M + (ρ. (z) ≤ 0 and lim M − (ρ. There is no principal part. z→a One speaks of a pole of order k in z = a. (z) ≤ 0} and M (ρ.
xi = Ai xi i i i i ˆ 2..qn Ts1 .sm = ∂x ∂u ∂uq1 ∂uqn ∂xr1 ∂xrm p .. 43 .1 Vectors and covectors A ﬁnite dimensional vector space is denoted by V.. Consider L(V. Ai = i i ∂x ∂xi k From this follows that Ai · Ak = δl and Ai = (Ai )−1 .I := V ∗ ..Chapter 7 Tensor calculus 7. W. Covectors: x = xiˆ with basis vectors ˆ c c ∂ ∂xi ˆ i = dxi c Transformation from system i to i is given by: ˆ i = Ai ˆ i ∈ V ∗ .pn · · · pn · · · · sm Tr11.rm p1 s1 ∂x ∂x ∂u ∂u For an absolute tensor = 0.. 1. The vector space of linear transformations from V to W is denoted by L(V. i i k l In differential notation the coordinate transformations are given by: dxi = ∂xi i dx ∂xi and ∂ ∂xi ∂ = i ∂x ∂xi ∂xi The general transformation rule for a tensor T is: q1 .. We name V ∗ the dual space of V. xi = Ai xi c i c i Here the Einstein convention is used: ai bi := i ai bi The coordinate transformation is given by: Ai = i ∂xi ∂xi . Vectors: x = xi ci with basis vectors ci : ci = Transformation from system i to i is given by: ci = Ai ci = ∂i ∈ V ... Now we can deﬁne vectors in V R) ∗ with basis c and covectors in V with basis ˆ Properties of both are: c. W).
44
Mathematics Formulary by ir. J.C.A. Wevers
7.2
Tensor algebra
The following holds: aij (xi + yi ) ≡ aij xi + aij yi , but: aij (xi + yj ) ≡ aij xi + aij yj and (aij + aji )xi xj ≡ 2aij xi xj , but: (aij + aji )xi yj ≡ 2aij xi yj en (aij − aji )xi xj ≡ 0.
p The sum and difference of two tensors is a tensor of the same rank: Ap ± Bq . The outer tensor product results in q pr m prm a tensor with a rank equal to the sum of the ranks of both tensors: Aq · Bs = Cqs . The contraction equals two mpr mp indices and sums over them. Suppose we take r = s for a tensor Aqs , this results in: Ampr = Bq . The inner qr r
product of two tensors is deﬁned by taking the outer product followed by a contraction.
7.3
Inner product
ˆ ˆ ˆ Deﬁnition: the bilinear transformation B : V × V ∗ → I B(x, y ) = y(x ) is denoted by < x, y >. For this pairing R, operator < ·, · >= δ holds: ˆ ˆ y(x) =< x, y >= yi xi , < cˆi , cj >= δ i
j
Let G : V → V ∗ be a linear bijection. Deﬁne the bilinear forms g :V ×V →I R h:V ×V →I R
∗ ∗
g(x, y ) =< x, Gy > ˆ ˆ ˆ ˆ h(x, y ) =< G−1 x, y >
Both are not degenerated. The following holds: h(Gx, Gy ) =< x, Gy >= g(x, y ). If we identify V and V ∗ with G, than g (or h) gives an inner product on V. The inner product (, )Λ on Λk (V) is deﬁned by: (Φ, Ψ)Λ = The inner product of two vectors is than given by: (x, y ) = xi y i < ci , Gcj >= gij xi xj The matrix gij of G is given by gij ˆ = Gci c The matrix g ij of G−1 is given by: g kl cl = G−1ˆ c
k j
1 0 (Φ, Ψ)Tk (V) k!
k For this metric tensor gij holds: gij g jk = δi . This tensor can raise or lower indices:
xj = gij xi , xi = g ij xj and dui = ˆ = g ij cj . c
i
Chapter 7: Tensor calculus
45
7.4
Tensor product
Deﬁnition: let U and V be two ﬁnite dimensional vector spaces with dimensions m and n. Let U ∗ × V ∗ be the ˆ ˆ cartesian product of U and V. A function t : U ∗ × V ∗ → I (u; v ) → t(u; v ) = tαβ uα uβ ∈ I is called a tensor R; ˆ ˆ R ˆ and v. The tensors t form a vector space denoted by U ⊗ V. The elements T ∈ V ⊗ V are called ˆ if t is linear in u contravariant 2tensors: T = T ij ci ⊗ cj = T ij ∂i ⊗ ∂j . The elements T ∈ V ∗ ⊗ V ∗ are called covariant 2tensors: i j i T = Tij ˆ ⊗ ˆ = Tij dxi ⊗ dxj . The elements T ∈ V ∗ ⊗ V are called mixed 2 tensors: T = T .j ˆ ⊗ cj = c c c Ti.j dxi ⊗ ∂j , and analogous for T ∈ V ⊗ V ∗ . The numbers given by
α β tαβ = t(ˆ , ˆ ) c c i
with 1 ≤ α ≤ m and 1 ≤ β ≤ n are the components of t. Take x ∈ U and y ∈ V. Than the function x ⊗ y, deﬁnied by ˆ ˆ ˆ ˆ (x ⊗ y)(u, v) =< x, u >U < y, v >V is a tensor. The components are derived from: (u ⊗ v )ij = ui v j . The tensor product of 2 tensors is given by: 2 form: 0 0 form: 2 1 form: 1 ˆ q) (v ⊗ w)(p, ˆ = v i pi wk qk = T ik pi qk ˆ q)(v, w) = pi v i qk wk = Tik v i wk (p ⊗ ˆ
i ˆ q, (v ⊗ p)(ˆ w) = v i qi pk wk = Tk qi wk
7.5
Symmetric and antisymmetric tensors
ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ A tensor t ∈ V ⊗ V is called symmetric resp. antisymmetric if ∀x, y ∈ V ∗ holds: t(x, y ) = t(y, x ) resp. t(x, y ) = ˆ x ). ˆ −t(y, A tensor t ∈ V ∗ ⊗ V ∗ is called symmetric resp. antisymmetric if ∀x, y ∈ V holds: t(x, y ) = t(y, x ) resp. t(x, y ) = −t(y, x ). The linear transformations S and A in V ⊗ W are deﬁned by: ˆ ˆ St(x, y ) = ˆ ˆ At(x, y ) =
1 ˆ ˆ 2 (t(x, y) 1 ˆ ˆ 2 (t(x, y)
ˆ ˆ + t(y, x )) ˆ ˆ − t(y, x ))
Analogous in V ∗ ⊗ V ∗ . If t is symmetric resp. antisymmetric, than St = t resp. At = t.
1 The tensors ei ∨ ej = ei ej = 2S(ei ⊗ ej ), with 1 ≤ i ≤ j ≤ n are a basis in S(V ⊗ V) with dimension 2 n(n + 1). 1 The tensors ei ∧ ej = 2A(ei ⊗ ej ), with 1 ≤ i ≤ j ≤ n are a basis in A(V ⊗ V) with dimension 2 n(n − 1).
The complete antisymmetric tensor ε is given by: εijk εklm = δil δjm − δim δjl . The permutationoperators epqr are deﬁned by: e123 = e231 = e312 = 1, e213 = e132 = e321 = −1, for all other combinations epqr = 0. There is a connection with the ε tensor: εpqr = g −1/2 epqr and εpqr = g 1/2 epqr .
7.6
Outer product
(k + l)! A(α ⊗ β) k!l!
Let α ∈ Λk (V) and β ∈ Λl (V). Than α ∧ β ∈ Λk+l (V) is deﬁned by: α∧β =
If α and β ∈ Λ1 (V) = V ∗ holds: α ∧ β = α ⊗ β − β ⊗ α
46
Mathematics Formulary by ir. J.C.A. Wevers
The outer product can be written as: (a × b)i = εijk aj bk , a × b = G−1 · ∗(Ga ∧ Gb ). Take a, b, c, d ∈ I 4 . Than (dt ∧ dz)(a, b ) = a0 b4 − b0 a4 is the oriented surface of the projection on the tzplane R of the parallelogram spanned by a and b. Further a0 (dt ∧ dy ∧ dz)(a, b, c) = det a2 a4 b0 b2 b4 c0 c2 c4
is the oriented 3dimensional volume of the projection on the tyzplane of the parallelepiped spanned by a, b and c. (dt ∧ dx ∧ dy ∧ dz)(a, b, c, d) = det(a, b, c, d) is the 4dimensional volume of the hyperparellelepiped spanned by a, b, c and d.
7.7
The Hodge star operator
n Λk (V) and Λn−k (V) have the same dimension because n = n−k for 1 ≤ k ≤ n. Dim(Λn (V)) = 1. The choice k of a basis means the choice of an oriented measure of volume, a volume µ, in V. We can gauge µ so that for an 1 2 n e orthonormal basis ei holds: µ(ei ) = 1. This basis is than by deﬁnition positive oriented if µ = ˆ ∧ˆ ∧...∧ˆ = 1. e e
Because both spaces have the same dimension one can ask if there exists a bijection between them. If V has no extra structure this is not the case. However, such an operation does exist if there is an inner product deﬁned on V and the corresponding volume µ. This is called the Hodge star operator and denoted by ∗. The following holds: ∀w∈Λk (V) ∃∗w∈Λk−n (V) ∀θ∈Λk (V) θ ∧ ∗w = (θ, w)λ µ For an orthonormal basis in I 3 holds: the volume: µ = dx ∧ dy ∧ dz, ∗dx ∧ dy ∧ dz = 1, ∗dx = dy ∧ dz, R ∗dz = dx ∧ dy, ∗dy = −dx ∧ dz, ∗(dx ∧ dy) = dz, ∗(dy ∧ dz) = dx, ∗(dx ∧ dz) = −dy. For a Minkowski basis in I 4 holds: µ = dt ∧ dx ∧ dy ∧ dz, G = dt ⊗ dt − dx ⊗ dx − dy ⊗ dy − dz ⊗ dz, and R ∗dt ∧ dx ∧ dy ∧ dz = 1 and ∗1 = dt ∧ dx ∧ dy ∧ dz. Further ∗dt = dx ∧ dy ∧ dz and ∗dx = dt ∧ dy ∧ dz.
7.8
Differential operations
7.8.1 The directional derivative
The directional derivative in point a is given by: La f =< a, df >= ai ∂f ∂xi
7.8.2 The Liederivative
The Liederivative is given by: (Lv w)j = wi ∂i v j − v i ∂i wj
7.8.3 Christoffel symbols
To each curvelinear coordinate system ui we add a system of n3 functions Γi of u, deﬁned by jk ∂2x ∂x = Γi jk i ∂uk ∂u ∂ui These are Christoffel symbols of the second kind. Christoffel symbols are no tensors. The Christoffel symbols of the second kind are given by: ∂2x i := Γi = , dxi jk jk ∂uk ∂uj
Their transformation to a different coordinate system is given by: jk kj Γi j k = Ai Aj Ak Γi + Ai (∂j Ai ) i i k jk k j The ﬁrst term in this expression is 0 if the primed coordinates are cartesian.4 The covariant derivative The covariant derivative j of a vector.Chapter 7: Tensor calculus 47 with Γi = Γi . There is a relation between Christoffel symbols and the metric: Γi = 1 g ir (∂j gkr + ∂k grj − ∂r gjk ) jk 2 and Γα = ∂β (ln( βα g)).8. jk 7.9 Differential operators The Gradient is given by: grad(f ) = G−1 df = g ki The divergence is given by: div(ai ) = i ia ∂f ∂ ∂xi ∂xk 1 √ = √ ∂k ( g ak ) g The curl is given by: rot(a) = G−1 · ∗ · d · Ga = −εpqr The Laplacian is given by: ∆(f ) = div grad(f ) = ∗d ∗ df = ig ij q ap = q ap − p aq ∂j f = g ij i jf 1 ∂ =√ g ∂xi √ g g ij ∂f ∂xj . Lowering an index gives the Christoffel symbols of the ﬁrst kind: Γi = g il Γjkl . covector and of rank2 tensors is given by: i ja j ai α γ aβ γ aαβ αβ γa = ∂j ai + Γi ak jk = ∂j ai − Γk ak ij = ∂γ aα − Γε aα + Γα aε β γβ ε γε β = ∂γ aαβ − Γε aεβ − Γε aαε γα γβ = ∂γ aαβ + Γα aεβ + Γβ aαε γε γε αβ Ricci’s theorem: γ gαβ = γg =0 7.
) = (x. n = −ρ + τ b . Let P be a point and Q be a nearby point of a space curve x(s). Then the curvature ρ and the torsion τ in P are deﬁned by: ρ2 = dϕ ds 2 = lim ∆s→0 ∆ϕ ∆s 2 . x3 ). x. v). ˙ ˙ = ρn . A point is represented by the vector x = (x1 . If x = 0 the osculation plane is given by: ˙ ¨ ˙ ¨ y = x + λx + µx so det(y − x. The following holds: ˙ ˙ ¨ ¨ ρ2 = ( . one with u =constant and one with v =constant.A. Wevers 7.2 Surfaces in IR3 A surface in I 3 is the collection of end points of the vectors x = x(u. x ) = ρ2 τ . the main normal unit vector n = x and the binormal b = x × x. b = −τ n ˙ ¨ From this follows that det(x.. if x= 0: ˙ y = x + λx + µ x . Let ∆ϕ be the angle between the tangents in P and Q and let ∆ψ be the angle between the osculation planes (binormals) in P and Q. The osculation plane is parallel with ˙ ¨ x(s). Some curves and their properties are: Screw line Circle screw line Plane curves Circles Lines τ /ρ =constant τ =constant.10 Differential geometry 7. τ = 0 ρ=τ =0 7. In a bending point holds.1 Space curves We limit ourselves to I 3 with a ﬁxed orthonormal basis. So the main normal lies in the osculation plane.C. b) Frenet’s equations express the derivatives as linear combinations of these vectors: ˙ . For plane curves ρ is the ordinary curvature and τ = 0.. the binormal is perpendicular to it.. ρ =constant τ =0 ρ =constant.48 Mathematics Formulary by ir. τ2 = dψ ds 2 and ρ > 0. so xh = xh (uα ). x ) = 0 . x ) and τ 2 = (b..10. The arc length of a space curve is given by: t s(t) = t0 dx dτ 2 + dy dτ 2 + dz dτ 2 dτ The derivative of s with respect to t is the length of the vector dx/dt: ds dt 2 = dx dx . ˙ ¨ ˙ ¨ The tangent has unit vector = x.10. The tangent plane in a point P at the surface has basis: c1 = ∂1 x and c2 = ∂2 x . On the surface are 2 R families of curves. A R space curve is a collection of points represented by x = x(t). dt dt The osculation plane in a point P of a space curve is the limiting position of the plane through the tangent of the plane in point P and a point Q when Q approaches P along the space curve.. x.. J. x2 .
dt dt dx dv β = ∂β x dτ dτ The ﬁrst fundamental tensor of the surface in P is the inner product of these tangent vectors: dx dx . with N perpendicular to c1 and c2 . v =constant and u =constant. ∂α cβ ) . The ¨ ¨ length of x is the curvature ρ of the curve in P . hαβ = (N .Chapter 7: Tensor calculus 49 7. ∂α cβ ) = αβ 1 det g det(c1 . cβ ) duα dv β dt dτ The covariant components w. uα = v α (τ ). have as tangent vectors in P duα dx = ∂α x .5 Geodetic curvature A curve on the surface x(uα ) is given by: uα = uα (s). v = τ holds: cos(φ) = √ For the arc length s of P along the curve uα (t) holds: ds2 = gαβ duα duβ This expression is called the line element. This is written as: ∂α cβ = Γγ cγ + hαβ N αβ This leads to: Γγ = (c γ .r. So for a geodetic line is in each point pγ = 0.10. g12 g11 g22 7. ∂α cβ ) 7. the basis cα = ∂α x are: gαβ = (cα . denoted by uα = uα (t). The projection of x on the surface is a vector with components p γ = u γ + Γγ u α u β ¨ αβ ˙ ˙ of which the length is called the geodetic curvature of the curve in p. dt dτ = (cα . c2 and N . cβ ) For the angle φ between the parameter curves in P : u = t. This remains the same if the surface is curved ¨ and the line element remains the same. The theorem of Meusnier states that different curves on the surface with the same tangent vector in P have the same normal curvature.10. The projection of x on N has length p = hαβ uα uβ ˙ ˙ and is called the normal curvature of the curve in P . than x = x(uα (s)) with s the arc length of the curve. so d2 uγ duα duβ + Γγ =0 αβ ds2 ds ds . A geodetic line of a surface is a curve on the surface for which in each point the main normal of the curve is the same as the normal on the surface.t.4 The second fundamental tensor The 4 derivatives of the tangent vectors ∂α ∂β x = ∂α cβ are each linear independent of the vectors c1 .3 The ﬁrst fundamental tensor Let P be a point of the surface x = x(uα ).10. c2 . The following two curves through P .
The Einstein tensor G is deﬁned by: Gαβ ≡ Rαβ − 2 g αβ . w ) = dt Along a curve holds: dt (v α cα ) = dv γ duα β + Γγ v cγ αβ dt dt v. Wevers The covariant derivative /dt in P of a vector ﬁeld of a surface along a curve is the projection on the tangent plane in P of the normal derivative in P . If it is 0. .C. J. It has the property that β Gαβ = 0. the space is a ﬂat manifold.50 Mathematics Formulary by ir. For two vector ﬁelds v(t) and w(t) along the same curve of the surface follows Leibniz’ rule: d(v. w dt + w. µ β ]Tν µ σ σ µ = Rσαβ Tν + Rναβ Tσ The Riemann tensor depends on the Christoffel symbols through α Rβµν = ∂µ Γα − ∂ν Γα + Γα Γσ − Γα Γσ βν βµ σµ βν σν βµ In a space and coordinate system where the Christoffel symbols are 0 this becomes: α 1 Rβµν = 2 g ασ (∂β ∂µ gσν − ∂β ∂ν gσµ + ∂σ ∂ν gβµ − ∂σ ∂µ gβν ) The Bianchi identities are: λ Rαβµν + ν Rαβλµ + µ Rαβνλ = 0. This tensor is a measure 3 for the curvature of the considered space.A. It has the following symmetry properties: Rαβµν = Rµναβ = −Rβαµν = −Rαβνµ The following relation holds: [ α. The αβ Ricciscalar is R = g Rαβ . v dt 7.11 Riemannian geometry µ Rναβ T ν = βT µ The Riemann tensor R is deﬁned by: α − β αT µ This is a 1 tensor with n2 (n2 − 1)/12 independent components not identically equal to 0. µ The Ricci tensor is obtained by contracting the Riemann tensor: Rαβ ≡ Rαµβ . and is symmetric in its indices: 1 Rαβ = Rβα .
The largest machine number is amax = (1 − β −t )β β and the smallest positive machine number is amin = β −β q q −1 The distance between two successive machine numbers in the interval [β p−1 . The sign s. 8. and e is a number with q βbased numbers for which holds e ≤ β q − 1. 51 . If x is a real number and the closest machine number is rd(x). for example with m = e = 0. c 1 if the problem is wellconditioned. The length of the mantissa t. and 2 ε. The length of the exponent q. 8 for the exponent and 23 for de mantissa.2 Floating point representations The ﬂoating point representation depends on 4 natural numbers: 1. ε ≤ η x − rd(x) ≤η x An often used 32 bits ﬂoat format is: 1 bit for s. The dependency between errors in input data and errors in the solution can be expressed in the condition number c. 2. 4. The basis of the number system β. The base here is 2. than holds: rd(x) = x(1 + ε) x = rd(x)(1 + ε ) with with ε ≤ 1 β 1−t 2 ε  ≤ 1 β 1−t 2 The number η := 1 β 1−t is called the machineaccuracy. Than the representation of machine numbers becomes: rd(x) = s · m · β e where mantissa m is a number with t βbased numbers and for which holds 1/β ≤ m < 1. The number 0 is added to this set. 3. β p ] is β p−t . If the problem is given by x = φ(a) the ﬁrstorder approximation for an error δa in a is: aφ (a) δa δx = · x φ(a) a The number c(a) = aφ (a)/φ(a).1 Errors There will be an error in the solution if a problem has a number of parameters which are not exactly known.Chapter 8 Numerical mathematics 8.
j < n. } = (c1 − j=2 U1j xj )/U11 1 This algorithm requires 2 n(n + 1) ﬂoating point calculations. j++) { A[i][j] = L * U[k][j]. J. k > 0. . Inverting a n×n matrix via Cramer’s rule requires too much multiplications f (n) with n! ≤ f (n) ≤ (e−1)n!. k <= n. . j++) { S = U[k][j] * x[j]. i++) { L = A[i][k] / U[k][k]. This can be reduced by Gauss elimination to a triangular form by multiplying the ﬁrst equation with Ai1 /A11 and than subtract it from all others. Than the 2nd equation is subtracted in such a way from the others that all elements on the second row are 0 except A22 . } b[i] = L * c[k].A. i <= n. for (j = k + 1. n xn x1 In code: for (k = n.2 Gauss elimination Consider a general set Ax = b. j <= n. 8. } x[k] = S / U[k][k]. 8. . .3 Systems of equations We want to solve the matrix equation Ax = b for a nonsingular A. so other methods are preferable.52 Mathematics Formulary by ir. Wevers 8. now the ﬁrst column contains all 0’s except A11 . etc. . k++) { for (j = k. j++) U[k][j] = A[k][j]. for (i = k + 1. k) { S = c[k]. j <= n. this is a matrix in which Uij = 0 for all j < i. Than: = cn /Unn xn−1 = (cn−1 − Un−1.n xn )/Un−1. and all Uii = 0.3.1 Triangular matrices Consider the equation U x = c where U is a rightupper triangular. for (j = k + 1.C. In code: for (k = 1.3. } } .n−1 . c[k] = b[k]. which is equivalent to ﬁnding the inverse matrix A−1 .
in the hope that lim xn = n→∞ α. Let f (x) be continuous in a neighbourhood of α. (1) 8.3. One method is: if Akk = 0. α − xn−1 lim xn − xn−1 =A .2 Local convergence Let α be a solution of x = f (x) and let xn = f (xn−1 ) for a given x0 . so we want to ﬁnd the root α with F (α) = 0.. = β = xn−1 − ε h(xn−1 ) g(xn−1 ) F (x) h(x) =x− g(x) G(x) 8. than for each n ≥ k holds that xn = α. lim nn = α.1 Successive substitution We want to solve the equation F (x) = 0. . 2. n→∞ 2. xn−1 − xn−2 lim α − xn A = xn − xn−1 1−A n→∞ n→∞ n→∞ The quantity A is called the asymptotic convergence factor. are not all = 0 to allow Gauss elimina(k−1) tion to work. Than there exists a δ > 0 so that for each x0 with x0 − α ≤ δ holds: 1. the quantity B = −10 log A is called the asymptotic convergence speed.4. f (x) may not vary too much with respect to x near α.3 Pivot strategy Some equations have to be interchanged if the corner elements A11 . If xn = α for all n than holds lim α − xn =A . In the following.Chapter 8: Numerical mathematics 53 This algorithm requires 1 n(n2 − 1) ﬂoating point multiplications and divisions for operations on the coefﬁcient 3 matrix and 1 n(n − 1) multiplications for operations on the righthand terms. Example: choose f (x) = β − ε than we can expect that the row xn with x0 xn converges to α. 8. Rewrite the equation in the form x = f (x) so that a solution of this equation is also a solution of F (x) = 0.4. A(n) is the element after the nth iteration. If for a particular k holds: xk = α. Let f (α) = A with A < 1. A22 . whereafter the triangular set has to be 2 1 solved with 2 n(n + 1) operations. . Further.4 Roots of functions 8. Many solutions are essentially the following: 1.. This strategy fails only if the set is singular and has no solution at all. than search (k−1) for an element Apk with p > k that is = 0 and interchange the pth and the nth equation. Assume an initial estimation x0 for α and obtain the series xn with xn = f (xn−1 ).
float SolveNewton(void (*funcd)(float. &f.3 Aitken extrapolation We deﬁne A = lim A converges to f (a). j++) { (*funcd)(root. f. For F (x) = xk − a the series becomes: xn = This is a wellknown way to compute roots. float*). root = 0. The following code ﬁnds the root of a function by means of Newton’s method. The function funcd is a routine that returns both the function and its ﬁrst derivative in point x in the passed pointers. the faster the series converges. If one chooses: Φ(x) = 1 F (x) Than this becomes Newtons method. One essential condition for them all is that in a neighbourhood of a root α holds that f (x) < 1. • The assumption F (α) = 0 means that α is a simple root.4 Newton iteration There are more ways to transform F (x) = 0 into x = f (x). max_iter = 25.A. dx = f/df. float x1. A general method to construct f (x) is: f (x) = x − Φ(x)F (x) with Φ(x) = 0 in a neighbourhood of α. &df). float x2. Than the row αn = xn + will converge to α.4. The value is adapted until the accuracy is better than ±eps. 1 k (k − 1)xn−1 + a xk−1 n−1 F (xn−1 ) F (xn−1 ) . • Local convergence is often difﬁcult to determine. for (j = 1. float eps) { int j. root.C.5 * (x1 + x2). float*. The root lies within the interval [x1. and the smaller f (x). The iteration formula than becomes: xn = xn−1 − Some remarks: • This same result can also be derived with Taylor series. Wevers 8. x2]. • If xn is far apart from α the convergence can sometimes be very slow.54 Mathematics Formulary by ir.4. dx. An (xn − xn−1 ) 1 − An n→∞ xn − xn−1 xn−1 − xn−2 8. j <= max_iter. J. float df.
Lj (xi ) = δij for i.4.1. a twostep method. Each Lj (x) has order n.5 The secant method This is. than one can ﬁnd the next approximation with xn+1 = xn − F (xn ) xn − xn−1 F (xn ) − F (xn−1 ) If F (xn ) and F (xn−1 ) have a different sign one is interpolating. 1. return 0.. } if ( fabs(dx) < eps ) return root. 3. int n) { int i.root)*(root .Chapter 8: Numerical mathematics 55 root = dx.").5 Polynomial interpolation n A base for polynomials of order n is given by Lagrange’s interpolation polynomials: Lj (x) = l=0 l=j x − xl xj − xl The following holds: 1. /* Convergence */ } perror("Maximum number of iterations exceeded in SolveNewton. } After it is ﬁnished s has value p(a).0. } 8. To do this. 2. 8."). i >= 0. . the Horner algorithm is more usable: the value s = k ck xk in x = a can be calculated as follows: float GetPolyValue(float c[]. } return s. If two approximations xn and xn−1 exist for a root. i) { s = s * a + c[i]... otherwise extrapolating. n.0 ) { perror("Jumped out of brackets in SolveNewton.x2) < 0. Each polynomial p(x) can be written uniquely as n p(x) = j=0 cj Lj (x) with cj = p(xj ) This is not a suitable method to calculate the value of a ploynomial in a given point x = a. float s = c[n]. if ( (x1 . exit(1). j = 0. in contrast to the two methods discussed previously. . for (i = n . exit(1).
A Gaussian integration formula is obtained when one wants to get both the coefﬁcients cj and the points xj in an integral formula so that the integral formula gives exact results for polynomials of an order as high as possible. x1 = b. J. ci and xi independent of f (x) and R(f ) the error which has the form R(f ) = Cf (q) (ξ) for all common methods. x2 = b.A. h = 2 (b − a): 2 b f (x)dx = a h5 h [f (x0 ) + 4f (x1 ) + f (x2 )] − f (4) (ξ) 3 90 • The midpoint rule: n = 0. Wevers 8. x0 = a.7 Derivatives There are several formulas for the numerical calculation of f (x): • Forward differentiation: f (x) = f (x + h) − f (x) 1 − 2 hf (ξ) h .6 Deﬁnite integrals b Almost all numerical methods are based on a formula of the type: n f (x)dx = a i=0 ci f (xi ) + R(f ) with n. Some common formulas are: • The trapezoid rule: n = 1. b) and q ≥ n + 1. h = b − a: b f (x)dx = a h h3 [f (x0 ) + f (x1 )] − f (ξ) 2 12 1 • Simpson’s rule: n = 2. Here.C. x0 = a. Gaussian formula with 3 points: h f (x)dx = −h h 5f 9 −h 3 5 + 8f (0) + 5f h 3 5 + h7 f (6) (ξ) 15750 8. ξ ∈ (a. Often the points xi are chosen equidistant. x0 = 1 (a + b). h = b − a: 2 b f (x)dx = hf (x0 ) + a h3 f (ξ) 24 The interval will usually be split up and the integration formulas be applied to the partial intervals if f varies much within the interval. Two examples are: 1.56 Mathematics Formulary by ir. x1 = 1 (a + b). Gaussian formula with 2 points: h f (x)dx = h f −h −h √ 3 +f h √ 3 + h5 (4) f (ξ) 135 2.
y(xn ). zn + 2 k1 ) 2 3 3 hf (xn + 4 h.. One of the possible 3rd order RungeKutta methods is given by: k1 k2 k3 zn+1 and the classical 4th order method is: k1 k2 k3 k4 zn+1 = = = = = hf (xn . y(x2 ). Than we can derive some formulas to obtain zn+1 as approximation for y(xn+1 ): • Euler (single step. zn ) 1 hf (xn + 2 h.. xn+1 ) Because ξn is unknown some “measurements” are done on the increment function k = hf (x.Chapter 8: Numerical mathematics 57 • Backward differentiation: f (x) = • Central differentiation: f (x) − f (x − h) 1 + 2 hf (ξ) h f (x + h) − f (x − h) h2 − f (ξ) 2h 6 • The approximation is better if more function values are used: f (x) = f (x) = −f (x + 2h) + 8f (x + h) − 8f (x − h) + f (x − 2h) h4 (5) + f (ξ) 12h 30 There are also formulas for higher derivatives: f (x) = −f (x + 2h) + 16f (x + h) − 30f (x) + 16f (x − h) − f (x − 2h) h4 (6) + f (ξ) 12h2 90 8.. implicit): 1 zn+1 = zn + 2 h(f (xn . They work so well because the solution y(x) can be written as: yn+1 = yn + hf (ξn .. zn + 1 k2 ) 2 2 hf (xn + h. zn + k3 ) zn + 1 (k1 + 2k2 + 2k3 + k4 ) 6 = = = = hf (xn . zn ) + f (xn+1 .. explicit): zn+1 = zn−1 + 2hf (xn .8 Differential equations We start with the ﬁrst order DE y (x) = f (x. zn for y(x1 ). zn+1 )) − h2 y (ξ) 2 h3 y (ξ) 3 h3 y (ξ) 12 RungeKutta methods are an important class of singlestep methods. Than one takes for zn+1 − zn a weighted average of the measured values. y) for x > x0 and initial condition y(x0 ) = x0 . zn ) 1 hf (xn + 1 h. zn + 4 k2 ) zn + 1 (2k1 + 3k2 + 4k3 ) 9 Often the accuracy is increased by adjusting the stepsize for each step with the estimated error. explicit): zn+1 = zn + hf (xn . y) in well chosen points near the solution. .. z2 . zn + 1 k1 ) 2 hf (xn + 1 h. y(ξn )) with ξn ∈ (xn .. Suppose we ﬁnd approximations z1 . Step doubling is most often used for 4th order RungeKutta. . zn ) + • Midpoint rule (two steps. zn ) + • Trapezoid rule (single step.
wtemp = sin(0. theta = isign * (6. 1. One is formed from the evennumbered points of the original N .C. J.9 The fast Fourier transform The Fourier transform of a function can be approximated when some discrete points are known.A. The next routine replaces the values in data by their discrete Fourier transformed values if isign = 1. theta. i < n.0 * wtemp * wtemp. wpi. It can be faster.h> #define SWAP(a.(b)=tempr void FourierTransform(float data[].58 Mathematics Formulary by ir.. wpi = sin(theta).5 * theta). #include <math. The basic idea is that a Fourier transform of length N can be rewritten as the sum of two discrete Fourier transforms. for (i = 1. tempi. order N ·2 log(N ).2*nn] contains on the odd positions the real and on the even positions the imaginary parts of the input data: data[1] is the real part and data[2] the imaginary part of f0 . each of length N/2.(a)=(b). wr. unsigned long nn. i.b) tempr=(a). . with the fast Fourier transform..28318530717959/mmax). Wevers 8. Than the discrete Fourier transform is given by: N −1 Hn = k=0 hk e2πikn/N and the inverse Fourier transform by hk = 2 1 N N −1 Hn e−2πikn/N n=0 This operation is order N . m >>= 1. The array data[1. data[i+1]). Suppose we have N successive samples hk = h(tk ) with tk = k∆. wpr = 2. . j. nn must be a power of 2. the other from the oddnumbered points. float tempr. and by their inverse transformed values if isign = −1. } mmax = 2. n = nn << 1. is executed log2(nn) times */ { istep = mmax << 1. wi. wpr. i += 2) { if ( j > i ) { SWAP(data[j].. istep. while ( m >= 2 && j > m ) { j = m. k = 0. This can be implemented as follows. N − 1. 2. data[i]). double wtemp.. int isign) { unsigned long n. mmax. j = 1. etc. SWAP(data[j+1]. } j += m. m. } m = n >> 1. while ( n > mmax ) /* Outermost loop.
wi * wpi wi = wi * wpr + wtemp * wpi + wi. data[i] += tempr. m < mmax.0. i <= n. m += 2) { for (i = m. wi = 0.wi tempi = wr * data[j+1] + wi data[j] = data[i] . data[j+1] = data[i+1] .tempi. } wr = (wtemp = wr) * wpr . + wr.0. .tempr. tempr = wr * data[j] . * data[j]. } } DanielsonLanczos equation */ * data[j+1].Chapter 8: Numerical mathematics 59 wr = 1. for (m = 1. i += istep) /* { j = i + mmax. } mmax=istep. data[i+1] += tempi.
This action might not be possible to undo. Are you sure you want to continue?