main January 23, 2006 16:6

Chapter 2 Mathematics
2

2.1 Notation
Mathematics is, of course, a vast subject, and so here we concentrate on those mathematical
methods and relationships that are most often applied in the physical sciences and engineering.
Although there is a high degree of consistency in accepted mathematical notation, there
is some variation. For example the spherical harmonics, Yl m , can be written Ylm , and there
is some freedom with their signs. In general, the conventions chosen here follow common
practice as closely as possible, whilst maintaining consistency with the rest of the handbook.
In particular:

scalars a general vectors a

unit vectors â scalar product a·b
vector cross-product a× b gradient operator ∇
df
Laplacian operator ∇2 derivative etc.
dx
∂f derivative of r with
partial derivatives etc. ṙ
∂x respect to t
n

d f
nth derivative closed loop integral dl
dxn L


closed surface integral ds matrix A or aij
S  
n
mean value (of x) x binomial coefficient
r
factorial ! unit imaginary (i2 = −1) i
exponential constant e modulus (of x) |x|
natural logarithm ln log to base 10 log10
 main January 23, 2006 16:6

20 Mathematics

2.2 Vectors and matrices

Vector algebra
Scalar producta a · b = |a||b|cosθ (2.1)
  a
 x̂ ŷ ẑ 

a× b = |a||b|sinθ n̂ = ax az 
b
Vector product ay (2.2) θ
bx by bz  b
n̂ (in)
a·b=b·a (2.3)
a× b = −b× a (2.4)
Product rules
a · (b + c) = (a · b) + (a · c) (2.5)
a× (b + c) = (a× b) + (a× c) (2.6)

Lagrange’s
(a× b) · (c× d) = (a · c)(b · d) − (a · d)(b · c) (2.7)
identity
 
ax ay az 

(a× b) · c = bx by bz  (2.8)
Scalar triple  cx cy cz  a
product = (b× c) · a = (c× a) · b (2.9) c
= volume of parallelepiped (2.10)
b

Vector triple (a× b)× c = (a · c)b − (b · c)a (2.11)

product a× (b× c) = (a · c)b − (a · b)c (2.12)

a = (b× c)/[(a× b) · c] (2.13)


b = (c× a)/[(a× b) · c] (2.14)
Reciprocal vectors
c = (a× b)/[(a× b) · c] (2.15)
(a · a) = (b · b) = (c · c) = 1 (2.16)

Vector a with
respect to a a = (e1 · a)e1 + (e2 · a)e2 + (e3 · a)e3 (2.17)
nonorthogonal
basis {e1 ,e2 ,e3 }c
a Also known as the “dot product” or the “inner product.”
b Also known as the “cross-product.” n̂ is a unit vector making a right-handed set with a and b.
c The prime ( ) denotes a reciprocal vector.
 main January 23, 2006 16:6

2.2 Vectors and matrices 21

Common three-dimensional coordinate systems

z
ρ 2

point P
θ
r

y
x φ

ρ = (x2 + y 2 )1/2
x = ρcosφ = r sinθ cosφ (2.18)
r = (x2 + y 2 + z 2 )1/2 (2.22)
y = ρsinφ = r sinθ sinφ (2.19)
θ = arccos(z/r)
z = r cosθ
φ = arctan(y/x)

coordinate system: rectangular spherical polar cylindrical polar

coordinates of P : (x,y,z) (r,θ,φ) (ρ,φ,z)
volume element: dx dy dz r2 sinθ dr dθ dφ ρ dρ dz dφ
metric elementsa (h1 ,h2 ,h3 ): (1,1,1) (1,r,r sinθ) (1,ρ,1)
a Inan orthogonal coordinate system (parameterised by coordinates q1 ,q2 ,q3 ), the differential line
element dl is obtained from (dl)2 = (h1 dq1 )2 + (h2 dq2 )2 + (h3 dq3 )2 .

Gradient
Rectangular ∂f ∂f ∂f f scalar field
∇f = x̂ + ŷ + ẑ
coordinates ∂x ∂y ∂z ˆ unit vector

Cylindrical ∂f 1 ∂f ∂f ρ distance from the

∇f = ρ̂ + φ̂ + ẑ
coordinates ∂ρ r ∂φ ∂z z axis

Spherical polar ∂f 1 ∂f 1 ∂f
∇f = r̂ + θ̂ + φ̂
coordinates ∂r r ∂θ r sinθ ∂φ
General
q̂ 1 ∂f q̂ 2 ∂f q̂ 3 ∂f qi basis
orthogonal ∇f = + +
coordinates h1 ∂q1 h2 ∂q2 h3 ∂q3 hi metric elements
 main January 23, 2006 16:6

22 Mathematics

Divergence
A vector field
Rectangular ∂Ax ∂Ay ∂Az
∇·A= + + Ai ith component
coordinates ∂x ∂y ∂z of A
Cylindrical 1 ∂(ρAρ ) 1 ∂Aφ ∂Az ρ distance from
∇·A= + +
coordinates ρ ∂ρ ρ ∂φ ∂z the z axis

1 ∂(r2 Ar ) 1 ∂(Aθ sinθ) 1 ∂Aφ

Spherical polar ∇·A= + +
coordinates r2 ∂r r sinθ ∂θ r sinθ ∂φ
(2.31)

1 ∂ ∂
General ∇·A= (A1 h2 h3 ) + (A2 h3 h1 ) qi basis
h1 h2 h3 ∂q1 ∂q2
orthogonal  hi metric
coordinates ∂ elements
+ (A3 h1 h2 )
∂q3

Curl
 
 x̂ ẑ 
ˆ unit vector
 ŷ
Rectangular   vector field
∇× A = ∂/∂x ∂/∂z 
A
coordinates ∂/∂y
  Ai ith component
 Ax Ay Az  of A
 
 ρ̂/ρ ẑ/ρ 
 φ̂
Cylindrical   ρ distance from
coordinates ∇× A = ∂/∂ρ ∂/∂φ ∂/∂z  the z axis
 
 Aρ ρAφ Az 
 
r̂/(r2 sinθ) φ̂/r 
 θ̂/(r sinθ)
Spherical polar  
coordinates ∇× A =  ∂/∂r ∂/∂θ ∂/∂φ 
 
 Ar rAθ rAφ sinθ
 
 q̂ h q̂ 3 h3 
General  1 1 q̂ 2 h2 qi basis
1  
orthogonal ∇× A = ∂/∂q1 ∂/∂q2 ∂/∂q3 
coordinates h1 h2 h3  
hi metric
elements
 h1 A1 h2 A2 h3 A3 

Radial formsa
−r
r ∇(1/r) =
∇r = r3
r 1
∇·r =3 ∇ · (r/r2 ) = 2
r
∇r2 = 2r −2r
∇(1/r ) = 4
2

∇ · (rr) = 4r r
∇ · (r/r3 ) = 4πδ(r) (2.44)
a Note that the curl of any purely radial function is zero. δ(r) is the Dirac delta function.
 main January 23, 2006 16:6

2.2 Vectors and matrices 23

Laplacian (scalar)
Rectangular ∂2 f ∂2 f ∂2 f
coordinates ∇2
f = + + f scalar field
∂x2 ∂y 2 ∂z 2
Cylindrical 1 ∂

∂f

1 ∂2 f ∂2 f ρ distance 2
coordinates ∇2
f = ρ + + from the
ρ ∂ρ ∂ρ ρ2 ∂φ2 ∂z 2 z axis
   
Spherical 1 ∂ 2 ∂f 1 ∂ ∂f 1 ∂2 f
polar ∇2
f = 2
r + 2
sinθ + 2
r ∂r ∂r r sinθ ∂θ ∂θ r2 sin θ ∂φ2
coordinates
    
1 ∂ h2 h3 ∂f ∂ h3 h1 ∂f
General ∇ f=
2
+ qi basis
h1 h2 h3 ∂q1 h1 ∂q1 ∂q2 h2 ∂q2
orthogonal   hi metric
coordinates ∂ h1 h2 ∂f elements
+ (2.48)
∂q3 h3 ∂q3

Differential operator identities

∇(fg) ≡ f∇g + g∇f
∇ · (fA) ≡ f∇ · A + A · ∇f
∇× (fA) ≡ f∇× A + (∇f)× A
∇(A · B) ≡ A× (∇× B) + (A · ∇)B + B× (∇× A) + (B · ∇)A (2.52)
∇ · (A× B) ≡ B · (∇× A) − A · (∇× B) f,g scalar fields
∇× (A× B) ≡ A(∇ · B) − B(∇ · A) + (B · ∇)A − (A · ∇)B A,B vector fields

∇ · (∇f) ≡ ∇ f ≡
2
f
∇× (∇f) ≡ 0
∇ · (∇× A) ≡ 0
∇× (∇× A) ≡ ∇(∇ · A) − ∇2 A

Vector integral transformations

A vector field
Gauss’s 

dV volume element
(Divergence) (∇ · A) dV = A · ds
V Sc Sc closed surface
theorem
V volume enclosed
S surface


Stokes’s ds surface element
(∇× A) · ds = A · dl
theorem S L L loop bounding S
dl line element


(f∇g) · ds = ∇ · (f∇g) dV
Green’s first S
V f,g scalar fields
theorem
= [f∇2 g + (∇f) · (∇g)] dV (2.62)
V


Green’s second [f(∇g) − g(∇f)] · ds = (f∇2 g − g∇2 f) dV
theorem S V
(2.63)
 main January 23, 2006 16:6

24 Mathematics

Matrix algebraa
 
a11 a12 ··· a1n
 a21 a22 ··· a2n 
  A m by n matrix
Matrix definition A= . .. ..  (2.64)
 .. . ··· .  aij matrix elements

am1 am2 ··· amn

Matrix addition C = A + B if cij = aij + bij (2.65)

C = AB if cij = aik bkj (2.66)

Matrix
(AB)C = A(BC) (2.67)
multiplication
A(B + C) = AB + AC (2.68)

ãij = aji (2.69) ãij transpose matrix

Transpose matrixb  = Ñ... B̃Ã (sometimes aT 
(AB...N) (2.70) ij , or aij )

∗ complex conjugate (of

Adjoint matrix A† = Ã∗ (2.71) each component)
(definition 1)c (AB...N)† = N† ...B† A† (2.72) † adjoint (or Hermitian
conjugate)

Hermitian matrixd H† = H (2.73) H Hermitian (or

self-adjoint) matrix
examples:
   
a11 a12 a13 b11 b12 b13
   
A = a21 a22 a23  B = b21 b22 b23 
a31 a32 a33 b31 b32 b33

   
a11 a21 a31 a11 + b11 a12 + b12 a13 + b13
   
à = a12 a22 a32  A + B = a21 + b21 a22 + b22 a23 + b23 
a13 a23 a33 a31 + b31 a32 + b32 a33 + b33

 
a11 b11 + a12 b21 + a13 b31 a11 b12 + a12 b22 + a13 b32 a11 b13 + a12 b23 + a13 b33
 
AB = a21 b11 + a22 b21 + a23 b31 a21 b12 + a22 b22 + a23 b32 a21 b13 + a22 b23 + a23 b33 
a31 b11 + a32 b21 + a33 b31 a31 b12 + a32 b22 + a33 b32 a31 b13 + a32 b23 + a33 b33
a Terms

are implicitly summed over repeated suffices; hence aik bkj equals k aik bkj .
b See also Equation (2.85).
c Or “Hermitian conjugate matrix.” The term “adjoint” is used in quantum physics for the transpose conjugate of

a matrix and in linear algebra for the transpose matrix of its cofactors. These definitions are not compatible, but
both are widely used [cf. Equation (2.80)].
d Hermitian matrices must also be square (see next table).
 main January 23, 2006 16:6

2.2 Vectors and matrices 25

Square matricesa
trA = aii (2.74) A square matrix
Trace aij matrix elements
tr(AB) = tr(BA) (2.75) 
aii implicitly = i aii
2
detA = ijk... a1i a2j a3k ... (2.76) tr trace
= (−1)i+1 ai1 Mi1 (2.77) det determinant (or |A|)
Determinantb Mij minor of element aij
= ai1 Ci1 (2.78)
Cij cofactor of the
det(AB...N) = detAdetB...detN (2.79) element aij

adj adjoint (sometimes

Adjoint matrix
adjA = C̃ij = Cji (2.80) written Â)
(definition 2)c ∼ transpose
Cji adjA
a−1
ij = = (2.81)
Inverse matrix detA detA
(detA = 0) AA−1 = 1 (2.82) 1 unit matrix
−1 −1 −1 −1
(AB...N) =N ...B A (2.83)

Orthogonality aij aik = δjk (2.84) δjk Kronecker delta (= 1

−1
condition i.e., Ã = A (2.85) if i = j, = 0 otherwise)

If A = Ã, A is symmetric (2.86)

Symmetry
If A = −Ã, A is antisymmetric (2.87)

unitary matrix
Unitary matrix U† = U−1 (2.88) U
† Hermitian conjugate
examples:
 
a11 a12 a13  
  b11 b12
A = a21 a22 a23  B=
b21 b22
a31 a32 a33
trA = a11 + a22 + a33 trB = b11 + b22

detA = a11 a22 a33 − a11 a23 a32 − a21 a12 a33 + a21 a13 a32 + a31 a12 a23 − a31 a13 a22

detB = b11 b22 − b12 b21

 
a22 a33 − a23 a32 −a12 a33 + a13 a32 a12 a23 − a13 a22
−1 1  
A = −a21 a33 + a23 a31 a11 a33 − a13 a31 −a11 a23 + a13 a21 
detA
a21 a32 − a22 a31 −a11 a32 + a12 a31 a11 a22 − a12 a21
 
−1 1 b22 −b12
B =
detB −b21 b11
a Terms

are implicitly summed over repeated suffices; hence aik bkj equals k aik bkj .
b
ijk... is defined as the natural extension of Equation (2.443) to n-dimensions (see page 50). Mij is the determinant
of the matrix A with the ith row and the jth column deleted. The cofactor Cij = (−1)i+j Mij .
c Or “adjugate matrix.” See the footnote to Equation (2.71) for a discussion of the term “adjoint.”
 main January 23, 2006 16:6

26 Mathematics

Commutators
Commutator [A,B] = AB − BA = −[B,A] (2.89) [·,·] commutator
definition

Adjoint [A,B]† = [B† ,A† ] (2.90) † adjoint

Distribution [A + B,C] = [A,C] + [B,C] (2.91)

Association [AB,C] = A[B,C] + [A,C]B (2.92)

Jacobi identity [A,[B,C]] = [B,[A,C]] − [C,[A,B]] (2.93)

Pauli matrices
   
0 1 0 −i
σ1 = σ2 = σi Pauli spin matrices
1 0 i 0
Pauli matrices     1 2 × 2 unit matrix
1 0 1 0 i i2 = −1
σ3 = 1= (2.94)
0 −1 0 1
Anticommuta- σ i σ j + σ j σ i = 2δij 1 δij Kronecker delta
tion

Cyclic σ i σ j = iσ k
permutation (σ i )2 = 1

Rotation matricesa
 
Rotation 1 0 0 Ri (θ) matrix for rotation

about x1 R1 (θ) = 0 cosθ sinθ  about the ith axis

0 −sinθ cosθ θ rotation angle
 
Rotation cosθ 0 −sinθ
about x2 R2 (θ) =  0 1 0 
sinθ 0 cosθ
 
Rotation cosθ sinθ 0 α rotation about x3

about x3 R3 (θ) =  −sinθ cosθ 0 β rotation about x2

0 0 1 γ rotation about x3
Euler angles R rotation matrix
 
cosγ cosβ cosα − sinγ sinα cosγ cosβ sinα + sinγ cosα −cosγ sinβ
R(α,β,γ) = −sinγ cosβ cosα − cosγ sinα −sinγ cosβ sinα + cosγ cosα sinγ sinβ 
sinβ cosα sinβ sinα cosβ

a Angles are in the right-handed sense for rotation of axes, or the left-handed sense for rotation of vectors. i.e., a
vector v is given a right-handed rotation of θ about the x3 -axis using R3 (−θ)v → v  . Conventionally, x1 ≡ x, x2 ≡ y,
and x3 ≡ z.
 main January 23, 2006 16:6

2.3 Series, summations, and progressions 27

2.3 Series, summations, and progressions

Progressions and summations

Sn = a + (a + d) + (a + 2d) + ··· n number of terms 2
+ [a + (n − 1)d] (2.102) Sn sum of n successive
Arithmetic n terms
progression = [2a + (n − 1)d] (2.103) a first term
2
n d common difference
= (a + l) (2.104) l last term
2

Sn = a + ar + ar2 + ··· + arn−1 (2.105)

Geometric 1 − rn
=a (2.106) r common ratio
progression 1−r
a
S∞ = (|r| < 1) (2.107)
1−r
Arithmetic 1
x a = (x1 + x2 + ··· + xn ) (2.108) . a arithmetic mean
mean n
Geometric
x g = (x1 x2 x3 ...xn )1/n (2.109) . g geometric mean
mean
 −1
1 1 1
Harmonic mean x h = n + + ··· + (2.110) . h harmonic mean
x1 x2 xn
Relative mean
x a ≥ x g ≥ x h if xi > 0 for all i (2.111)
magnitudes

n
n
i = (n + 1) (2.112)
2
i=1
n
n
i2 = (n + 1)(2n + 1) (2.113)
6
i=1
n
n2
i3 = (n + 1)2 (2.114)
4
i=1

Summation n
n
i4 = (n + 1)(2n + 1)(3n2 + 3n − 1) (2.115) i dummy integer
formulas 30
i=1
∞
(−1)i+1 1 1 1
= 1 − + − + ... = ln2 (2.116)
i 2 3 4
i=1
∞
(−1)i+1 1 1 1 π
= 1 − + − + ... = (2.117)
2i − 1 3 5 7 4
i=1
∞
1 1 1 1 π2
2
= 1 + + + + ... = (2.118)
i 4 9 16 6
i=1
 
Euler’s 1 1 1
γ = lim 1 + + + ··· + − lnn (2.119) γ Euler’s constant
constanta n→∞ 2 3 n
a γ  0.577215664...
 main January 23, 2006 16:6

28 Mathematics

Power series
Binomial n(n − 1) 2 n(n − 1)(n − 2) 3
(1 + x)n = 1 + nx + x + x + ··· (2.120)
seriesa 2! 3!
 
Binomial n n!
n
Cr ≡ ≡ (2.121)
coefficientb r r!(n − r)!
n  

Binomial n n−k k
(a + b)n = a b (2.122)
theorem k
k=0

Taylor series x2 (2) xn−1 (n−1)

f(a + x) = f(a) + xf (1) (a) + f (a) + ··· + f (a) + ··· (2.123)
(about a)c 2! (n − 1)!

Taylor series (x · ∇)2 (x · ∇)3

f(a + x) = f(a) + (x · ∇)f|a + f|a + f|a + ··· (2.124)
(3-D) 2! 3!

Maclaurin x2 (2) xn−1 (n−1)

f(x) = f(0) + xf (1) (0) + f (0) + ··· + f (0) + ··· (2.125)
series 2! (n − 1)!
a If n is a positive integer the series terminates and is valid for all x. Otherwise the (infinite) series is convergent for

|x| < 1.
b The coefficient of xr in the binomial series.
c xf (n) (a) is x times the nth derivative of the function f(x) with respect to x evaluated at a, taken as well behaved

around a. (x · ∇)n f|a is its extension to three dimensions.

Limits
nc xn → 0 as n → ∞ if |x| < 1 (for any fixed c) (2.126)

xn /n! → 0 as n→∞ (for any fixed x) (2.127)

(1 + x/n)n → ex as n→∞ (2.128)

xlnx → 0 as x→0 (2.129)

sinx
→ 1 as x→0 (2.130)
x
f(x) f (1) (a)
If f(a) = g(a) = 0 or ∞ then lim = (1) (l’Hôpital’s rule) (2.131)
x→a g(x) g (a)
 main January 23, 2006 16:6

2.3 Series, summations, and progressions 29

Series expansions
x2 x3
exp(x) 1+x+ + + ··· (2.132) (for all x)
2! 3!
x2 x3 x4 2
ln(1 + x) x− + − + ··· (2.133) (−1 < x ≤ 1)
2 3 4
   
1+x x3 x5 x7
ln 2 x + + + + ··· (2.134) (|x| < 1)
1−x 3 5 7
x2 x4 x6
cos(x) 1− + − + ··· (2.135) (for all x)
2! 4! 6!
x3 x5 x7
sin(x) x− + − + ··· (2.136) (for all x)
3! 5! 7!
x3 2x5 17x7
tan(x) x+ + + ··· (2.137) (|x| < π/2)
3 15 315
x2 5x4 61x6
sec(x) 1+ + + + ··· (2.138) (|x| < π/2)
2 24 720
1 x 7x3 31x5
csc(x) + + + + ··· (2.139) (|x| < π)
x 6 360 15120
1 x x3 2x5
cot(x) − − − − ··· (2.140) (|x| < π)
x 3 45 945
1 x3 1 · 3 x5 1 · 3 · 5 x7
arcsin(x)a x+ + + ··· (2.141) (|x| < 1)
2 3 2·4 5 2·4·6 7


 x3 x5 x7 (|x| ≤ 1)

 x − + − + ···

 3 5 7

arctan(x)b π 1 1 1
 − + 3 − 5 + ··· (2.142) (x > 1)

 2 x 3x 5x


 π
− − + 1 1 1
− + ··· (x < −1)
2 x 3x3 5x5
x2 x4 x6
cosh(x) 1 + + + + ··· (2.143) (for all x)
2! 4! 6!
x3 x5 x7
sinh(x) x+ + + + ··· (2.144) (for all x)
3! 5! 7!
x3 2x5 17x7
tanh(x) x− + − + ··· (2.145) (|x| < π/2)
3 15 315
a arccos(x) = π/2 − arcsin(x). Note that arcsin(x) ≡ sin−1 (x) etc.
b arccot(x) = π/2 − arctan(x).
 main January 23, 2006 16:6

30 Mathematics

Inequalities
|a1 | − |a2 | ≤ |a1 + a2 | ≤ |a1 | + |a2 |; (2.146)
Triangle  n 
   n
inequality  
 ai ≤ |ai | (2.147)
 
i=1 i=1

if a1 ≥ a2 ≥ a3 ≥ ... ≥ an (2.148)
Chebyshev and b1 ≥ b2 ≥ b3 ≥ ... ≥ bn (2.149)
 n  n 
inequality n  
then n ai bi ≥ ai bi (2.150)
i=1 i=1 i=1
 n 2
Cauchy  
n 
n

inequality ai bi ≤ a2i b2i (2.151)

i=1 i=1 i=1
 b 2  b  b
Schwarz
inequality f(x)g(x) dx ≤ [f(x)]2 dx [g(x)]2 dx (2.152)
a a a

2.4 Complex variables

Complex numbers
z complex variable
Cartesian form z = x + iy (2.153) i i2 = −1
x,y real variables
r amplitude (real)
Polar form z = reiθ = r(cosθ + isinθ) (2.154)
θ phase (real)

|z| = r = (x2 + y 2 )1/2 (2.155)

Modulusa |z| modulus of z
|z1 · z2 | = |z1 | · |z2 | (2.156)

y
θ = argz = arctan (2.157)
Argument x argz argument of z
arg(z1 z2 ) = argz1 + argz2 (2.158)

z ∗ = x − iy = re−iθ (2.159)
Complex arg(z ∗ ) = −argz (2.160) z∗ complex conjugate of
conjugate z = reiθ
∗
z · z = |z| 2
(2.161)

Logarithmb lnz = lnr + i(θ + 2πn) (2.162) n integer

a Or “magnitude.”
b The principal value of lnz is given by n = 0 and −π < θ ≤ π.
 main January 23, 2006 16:6

2.4 Complex variables 31

Complex analysisa
if f(z) = u(x,y) + iv(x,y) z complex variable
Cauchy– ∂u ∂v i i2 = −1
Riemann then =
∂x ∂y
(2.163) x,y real variables 2
equationsb ∂u ∂v f(z) function of z
=− (2.164) u,v real functions
∂y ∂x
Cauchy–

Goursat f(z) dz = 0 (2.165)
theoremc c


1 f(z) (n) nth derivative
Cauchy f(z0 ) = dz (2.166)
2πi c z − z0 an Laurent coefficients
integral

n! f(z) a−1 residue of f(z) at z0
formulad f (n) (z0 ) = dz (2.167)
2πi c (z − z0 )n+1 z dummy variable

∞
 y
f(z) = an (z − z0 )n (2.168) c2
Laurent n=−∞
expansione
c1
1 f(z  )
where an = 
dz  (2.169) c z0
2πi c (z − z0 )n+1


Residue x
f(z) dz = 2πi enclosed residues (2.170)
theorem c
a Closed contour integrals are taken in the counterclockwise sense, once.
b Necessary condition for f(z) to be analytic at a given point.
c If f(z) is analytic within and on a simple closed curve c. Sometimes called “Cauchy’s theorem.”
d If f(z) is analytic within and on a simple closed curve c, encircling z .
0
e Of f(z), (analytic) in the annular region between concentric circles, c and c , centred on z . c is any closed curve
1 2 0
in this region encircling z0 .
 main January 23, 2006 16:6

32 Mathematics

2.5 Trigonometric and hyperbolic formulas

Trigonometric relationships
sin(A ± B) = sinAcosB ± cosAsinB (2.171)
cos(A ± B) = cosAcosB ∓ sinAsinB (2.172)
tanA ± tanB
tan(A ± B) = (2.173)
1 ∓ tanAtanB
1
cosAcosB = [cos(A + B) + cos(A − B)] (2.174)
2
1
sinAcosB = [sin(A + B) + sin(A − B)] (2.175)
2
1
sinAsinB = [cos(A − B) − cos(A + B)] (2.176) 2
2

cos2 A + sin2 A = 1 (2.177) 1

sin
cos

x
sec2 A − tan2 A = 1 (2.178) 0

csc A − cot A = 1
2 2
(2.179)
−1

tan x
sin2A = 2sinAcosA (2.180) −2
−6 −4 −2 0 2 4 6
cos2A = cos A − sin A
2 2
(2.181) x

2tanA
tan2A = (2.182)
1 − tan2 A
sin3A = 3sinA − 4sin3 A (2.183)
cos3A = 4cos3 A − 3cosA (2.184)

A+B A−B
sinA + sinB = 2sin cos (2.185)
2 2
A+B A−B
sinA − sinB = 2cos sin (2.186) 4
2 2
csc x
sec x

2
A+B A−B
co

cosA + cosB = 2cos cos (2.187)

tx

2 2 0

A+B A−B
cosA − cosB = −2sin sin (2.188) −2
2 2
−4
2 1
cos A = (1 + cos2A) (2.189) −6 −4 −2 0 2 4 6
2 x
1
sin A = (1 − cos2A)
2
(2.190)
2
1
cos3 A = (3cosA + cos3A) (2.191)
4
1
sin3 A = (3sinA − sin3A) (2.192)
4
 main January 23, 2006 16:6

2.5 Trigonometric and hyperbolic formulas 33

Hyperbolic relationshipsa
sinh(x ± y) = sinhxcoshy ± coshxsinhy (2.193)
cosh(x ± y) = coshxcoshy ± sinhxsinhy (2.194)
tanhx ± tanhy
2
tanh(x ± y) = (2.195)
1 ± tanhxtanhy
1
coshxcoshy = [cosh(x + y) + cosh(x − y)] (2.196)
2
1
sinhxcoshy = [sinh(x + y) + sinh(x − y)] (2.197)
2
1
sinhxsinhy = [cosh(x + y) − cosh(x − y)] (2.198)
2
4
co
cosh2 x − sinh2 x = 1 (2.199) 2
sh
x
tanhx
sech2 x + tanh2 x = 1 (2.200) 0
tanhx
coth2 x − csch2 x = 1 (2.201) −2

hx
sin
−4
sinh2x = 2sinhxcoshx (2.202)
−3 −2 −1 0 1 2 3
cosh2x = cosh2 x + sinh2 x (2.203) x

2tanhx
tanh2x = (2.204)
1 + tanh2 x
sinh3x = 3sinhx + 4sinh3 x (2.205)
cosh3x = 4cosh x − 3coshx
3
(2.206)

x+y x−y
sinhx + sinhy = 2sinh cosh (2.207)
2 2 4
x+y x−y
sinhx − sinhy = 2cosh sinh (2.208) 2
2 2 sech x
cothx

x+y x−y 0
coshx + coshy = 2cosh cosh (2.209) cs
ch
2 2 −2
x

x+y x−y
coshx − coshy = 2sinh sinh (2.210) −4
2 2
−3 −2 −1 0 1 2 3
1
cosh2 x = (cosh2x + 1) (2.211) x
2
1
sinh2 x = (cosh2x − 1) (2.212)
2
1
cosh3 x = (3coshx + cosh3x) (2.213)
4
1
sinh3 x = (sinh3x − 3sinhx) (2.214)
4
a These can be derived from trigonometric relationships by using the
substitutions cosx → coshx and sinx → isinhx.
 main January 23, 2006 16:6

34 Mathematics

Trigonometric and hyperbolic definitions

de Moivre’s theorem (cosx + isinx)n = einx = cosnx + isinnx (2.215)

1  ix −ix  1  x −x 
cosx = e +e (2.216) coshx = e +e (2.217)
2 2
1  ix −ix  1  x −x 
sinx = e −e (2.218) sinhx = e −e (2.219)
2i 2
sinx sinhx
tanx = (2.220) tanhx = (2.221)
cosx coshx

cosix = coshx (2.222) coshix = cosx (2.223)

sinix = isinhx (2.224) sinhix = isinx (2.225)

cotx = (tanx)−1 (2.226) cothx = (tanhx)−1 (2.227)

secx = (cosx)−1 (2.228) sechx = (coshx)−1 (2.229)

cscx = (sinx)−1 (2.230) cschx = (sinhx)−1 (2.231)

1.6
Inverse trigonometric functionsa
  ar
cc
os
x x
arcsinx = arctan (2.232) 1 nx
(1 − x2 )1/2 arcta
  in
x
(1 − x2 )1/2 ar
cs
arccosx = arctan (2.233)
x
 
1 0 1
arccscx = arctan (2.234) x
(x2 − 1)1/2
  1.6
arcsecx = arctan (x2 − 1)1/2 (2.235)
arccsc x

cx
arcse
arc

 
cot

1
x

1
arccotx = arctan (2.236)
x
π
arccosx = − arcsinx (2.237)
2
a Valid in the angle range 0 ≤ θ ≤ π/2. Note that arcsinx ≡ sin−1 x etc. 0 1 2 3 4 5
x
 main January 23, 2006 16:6

2.6 Mensuration 35

Inverse hyperbolic functions

  2

arsinhx ≡ sinh−1 x = ln x + (x2 + 1)1/2 (2.238) for all x x

1 nh
ta
  ar
2

x
osh
arcoshx ≡ cosh−1 x = ln x + (x2 − 1)1/2 x≥1
0

arc
(2.239) nh x
−1 arsi
 
1 1+x
artanhx ≡ tanh−1 x = ln (2.240) |x| < 1 −2
2 1−x −1 0 1
  x
−1 1 x+1
arcothx ≡ coth x = ln (2.241) |x| > 1
2 x−1
  4
−1 1 (1 − x2 )1/2
arsechx ≡ sech x = ln + 0<x≤1
x x 3

arcoth x
(2.242)
  2
1 (1 + x2 )1/2 arc
arcschx ≡ csch−1 x = ln + x = 0
ar
se
sch
x
x x 1 ch
x
(2.243)
0 1 2
x

2.6 Mensuration

Moiré fringesa
 −1
1 1 dM Moiré fringe spacing
Parallel pattern dM =  −  (2.244) d1,2 grating spacings
d1 d2
d common grating
Rotational d spacing
dM = (2.245)
patternb 2|sin(θ/2)| θ relative rotation angle
(|θ| ≤ π/2)
a From overlapping linear gratings.
b From identical gratings, spacing d, with a relative rotation θ.
 main January 23, 2006 16:6

36 Mathematics

Plane triangles
a b c
Sine formulaa = = (2.246)
sinA sinB sinC

a2 = b2 + c2 − 2bccosA (2.247)
C
Cosine b +c −a
2 2 2 b
formulas cosA = (2.248) a
2bc
a = bcosC + ccosB (2.249) A
B
c
Tangent A−B a−b C
tan = cot (2.250)
formula 2 a+b 2
1
area= absinC (2.251)
2
a2 sinB sinC
= (2.252)
Area 2 sinA
= [s(s − a)(s − b)(s − c)]1/2 (2.253)
1
where s = (a + b + c) (2.254)
2
a The diameter of the circumscribed circle equals a/sinA.

Spherical trianglesa
sina sinb sinc
Sine formula = = (2.255)
sinA sinB sinC

Cosine cosa = cosbcosc + sinbsinccosA (2.256)

formulas cosA = −cosB cosC + sinB sinC cosa (2.257)

Analogue a C b
sinacosB = cosbsinc − sinbcosccosA (2.258)
formula
A
Four-parts B
cosacosC = sinacotb − sinC cotB (2.259)
formula c

Areab E =A+B +C −π (2.260)

a On a unit sphere.
b Also called the “spherical excess.”
 main January 23, 2006 16:6

2.6 Mensuration 37

Perimeter, area, and volume

P perimeter
Perimeter of circle P = 2πr (2.261)
r radius

Area of circle A = πr2 (2.262) A area 2

Surface area of spherea A = 4πR 2 (2.263) R sphere radius

4
Volume of sphere V = πR 3 (2.264) V volume
3
a semi-major axis
P = 4aE(π/2, e) (2.265) b semi-minor axis
Perimeter of ellipse b  2 1/2 E elliptic integral of the
a + b2 second kind (p. 45)
 2π (2.266)
2 e eccentricity
(= 1 − b2 /a2 )

Area of ellipse A = πab (2.267)

abc
Volume of ellipsoidc V = 4π (2.268) c third semi-axis
3
Surface area of
A = 2πr(h + r) (2.269) h height
cylinder

Volume of cylinder V = πr2 h (2.270)

Area of circular coned A = πrl (2.271) l slant height

Volume of cone or V = Ab h/3 (2.272) Ab base area

pyramid
r1 inner radius
Surface area of torus A = π 2 (r1 + r2 )(r2 − r1 ) (2.273)
r2 outer radius

π2 2 2
Volume of torus V= (r − r )(r2 − r1 ) (2.274)
4 2 1
Aread of spherical cap, A = 2πRd (2.275) d cap depth
depth d
  Ω solid angle
Volume of spherical d
V = πd 2
R− (2.276) z distance from centre
cap, depth d 3 α half-angle subtended
 
z
Solid angle of a circle Ω = 2π 1 − 2 2 1/2 (2.277) α r
from a point on its (z + r )
axis, z from centre = 2π(1 − cosα) (2.278) z
a Sphere defined by x2 + y 2 + z 2 = R 2 .
b The approximation is exact when e = 0 and e  0.91, giving a maximum error of 11% at e = 1.
c Ellipsoid defined by x2 /a2 + y 2 /b2 + z 2 /c2 = 1.
d Curved surface only.
 main January 23, 2006 16:6

38 Mathematics

Conic sections
y y y

b
x x x
a a
a

parabola ellipse hyperbola

x2 y 2 x2 y 2
equation y 2 = 4ax + =1 − =1
a2 b2 a2 b2

parametric x = t2 /(4a) x = acost x = ±acosht

form y=t y = bsint y = bsinht
√ √
foci (a,0) (± a2 − b2 ,0) (± a2 + b2 ,0)

√ √
a2 − b2 a2 + b2
eccentricity e=1 e= e=
a a
a a
directrices x = −a x=± x=±
e e

Platonic solidsa
solid volume surface area circumradius inradius
(faces,edges,vertices)
√ √ √
tetrahedron a3 2 √
2 a 6 a 6
a 3
(4,6,4) 12 4 12
√
cube a 3 a
a3 6a2
(6,12,8) 2 2
√
octahedron a3 2 √ a a
2a2 3 √ √
(8,12,6) 3 2 6
√  √

dodecahedron a3 (15 + 7 5) √ a√ √ a 50 + 22 5
3a2 5(5 + 2 5) 3(1 + 5)
(12,30,20) 4 4 4 5
√    
icosahedron 5a3 (3 + 5) √ a √ a √ 5
5a2 3 2(5 + 5) 3+
(20,30,12) 12 4 4 3
a Of side a. Both regular and irregular polyhedra follow the Euler relation, faces − edges + vertices = 2.
 main January 23, 2006 16:6

2.6 Mensuration 39

Curve measure
 b 2 1/2
a start point

Length of plane dy b end point
curve l= 1+ dx (2.279) y(x) plane curve
dx
a
l length 2
   2 1/2
Surface of b
dy
A = 2π y 1+ dx (2.280) A surface area
revolution dx
a
 b
Volume of
revolution V =π y 2 dx (2.281) V volume
a
  2 3/2  −1
Radius of dy d2 y ρ radius of
curvature ρ= 1+ (2.282) curvature
dx dx2

Differential geometrya
τ tangent
ṙ ṙ
Unit tangent τ̂ = = (2.283) r curve parameterised by r(t)
|ṙ| v
v |ṙ(t)|
r̈ −v̇ τ̂
Unit principal normal n̂ = (2.284) n principal normal
|r̈ −v̇ τ̂ |

Unit binormal b̂ = τ̂ × n̂ (2.285) b binormal

|ṙ× r̈|
Curvature κ= (2.286) κ curvature
|ṙ|3
1
Radius of curvature ρ= (2.287) ρ radius of curvature
κ
...
ṙ · (r̈× r )
Torsion λ= (2.288) λ torsion
|ṙ× r̈|2

n̂

τ̂˙ = κv n̂ (2.289) osculating

plane

˙ normal plane
Frenet’s formulas n̂ = −κv τ̂ + λv b̂ (2.290) τ̂

r rectifying
˙
b̂ = −λv n̂ (2.291) b̂ plane
origin
a For a continuous curve in three dimensions, traced by the position vector r(t).
 main January 23, 2006 16:6

40 Mathematics

2.7 Differentiation

Derivatives (general)
d n du n power
Power (u ) = nun−1 (2.292) index
dx dx
d dv du u,v functions
Product (uv) = u +v (2.293) of x
dx dx dx
d u ! 1 du u dv
Quotient = − (2.294)
dx v v dx v 2 dx
Function of a d d du f(u) function of
[f(u)] = [f(u)] · (2.295)
functiona dx du dx u(x)
  n  
dn n d u n dv dn−1 u
[uv] = v + + ··· n
dxn 0 dxn 1 dx dxn−1 binomial
Leibniz theorem   k n−k   n (2.296) k
coefficient
n d v d u n d v
+ k n−k
+ ··· + u
k dx dx n dxn
 q 
d
Differentiation f(x) dx = f(q) (p constant) (2.297)
dq p
under the integral  q 
sign d
f(x) dx = −f(p) (q constant) (2.298)
dp p
 v(x) 
d dv du
General integral f(t) dt = f(v) − f(u) (2.299)
dx u(x) dx dx
d b log base
Logarithm (logb |ax|) = (xlnb)−1 (2.300)
dx a constant

d ax
Exponential (e ) = aeax (2.301)
dx
 −1
dx dy
= (2.302)
dy dx
 −3
d2 x d2 y dy
Inverse functions = − (2.303)
dy 2 dx2 dx
  2  
−5
d3 x d2 y dy d3 y dy
= 3 − (2.304)
dy 3 dx2 dx dx3 dx
a The “chain rule.”
 main January 23, 2006 16:6

2.7 Differentiation 41

Trigonometric derivativesa
d d
(sinax) = acosax (2.305) (cosax) = −asinax (2.306)
dx dx
d
(tanax) = asec2 ax (2.307)
d
(cscax) = −acscax · cotax (2.308)
2
dx dx
d d
(secax) = asecax · tanax (2.309) (cotax) = −acsc2 ax (2.310)
dx dx
d d
(arcsinax) = a(1 − a2 x2 )−1/2 (2.311) (arccosax) = −a(1 − a2 x2 )−1/2 (2.312)
dx dx
d d a 2 2
(arctanax) = a(1 + a2 x2 )−1 (2.313) (arccscax) = − (a x − 1)−1/2 (2.314)
dx dx |ax|
d a 2 2 d
(arcsecax) = (a x − 1)−1/2 (2.315) (arccotax) = −a(a2 x2 + 1)−1 (2.316)
dx |ax| dx
aa is a constant.

Hyperbolic derivativesa
d d
(sinhax) = acoshax (2.317) (coshax) = asinhax (2.318)
dx dx
d d
(tanhax) = asech2 ax (2.319) (cschax) = −acschax · cothax (2.320)
dx dx
d d
(sechax) = −asechax · tanhax (2.321) (cothax) = −acsch2 ax (2.322)
dx dx
d d
(arsinhax) = a(a2 x2 + 1)−1/2 (2.323) (arcoshax) = a(a2 x2 − 1)−1/2 (2.324)
dx dx
d d a
(artanhax) = a(1 − a2 x2 )−1 (2.325) (arcschax) = − (1 + a2 x2 )−1/2 (2.326)
dx dx |ax|
d a
(arsechax) = − (1 − a2 x2 )−1/2 d
dx |ax| (arcothax) = a(1 − a2 x2 )−1 (2.328)
dx
(2.327)
aa is a constant.
 main January 23, 2006 16:6

42 Mathematics

Partial derivatives
Total ∂f ∂f ∂f
df = dx + dy + dz (2.329) f f(x,y,z)
differential ∂x ∂y ∂z
  
∂g  ∂x  ∂y 
Reciprocity = −1 (2.330) g g(x,y)
∂x y ∂y g ∂g x
∂f ∂f ∂x ∂f ∂y ∂f ∂z
Chain rule = + + (2.331)
∂u ∂x ∂u ∂y ∂u ∂z ∂u
 
 ∂x ∂x ∂x 
 
 ∂u ∂v ∂w  J Jacobian

∂(x,y,z)  ∂y ∂y ∂y  u u(x,y,z)
Jacobian J= =  (2.332)
∂(u,v,w)  ∂u ∂v ∂w  v v(x,y,z)
 ∂z ∂z ∂z 
  w w(x,y,z)
 
∂u ∂v ∂w
 
V volume in (x,y,z)
Change of f(x,y,z) dxdy dz = f(u,v,w)J dudv dw
V
V volume in (u,v,w)
variable V
mapped to by V
(2.333)
 b
Euler– if I= F(x,y,y  ) dx
y dy/dx
Lagrange a
  a,b fixed end points
equation ∂F d ∂F
then δI = 0 when = (2.334)
∂y dx ∂y 

Stationary pointsa

saddle point maximum minimum quartic minimum

∂f ∂f
Stationary point if = = 0 at (x0 ,y0 ) (necessary condition) (2.335)
∂x ∂y
Additional sufficient conditions
 2 2
for maximum ∂2 f ∂2 f ∂2 f ∂ f
< 0, and > (2.336)
∂x2 2
∂x ∂y 2 ∂x∂y
 2
for minimum ∂2 f ∂2 f ∂2 f ∂2 f
> 0, and > (2.337)
∂x2 ∂x2 ∂y 2 ∂x∂y
 2 2
for saddle point ∂2 f ∂2 f ∂ f
2 2
< (2.338)
∂x ∂y ∂x∂y
∂2 f 2
a Of a function f(x,y) at the point (x0 ,y0 ). Note that at, for example, a quartic minimum ∂x2
= ∂∂yf2 = 0.
 main January 23, 2006 16:6

2.7 Differentiation 43

Differential equations
Laplace ∇2 f = 0 (2.339) f f(x,y,z)

Diffusiona
∂f
= D∇2 f (2.340)
D diffusion
coefficient
2
∂t

Helmholtz ∇2 f + α2 f = 0 (2.341) α constant

1 ∂2 f
Wave ∇2 f = (2.342) c wave speed
c2 ∂t2
 
d 2 dy
Legendre (1 − x ) + l(l + 1)y = 0 (2.343) l integer
dx dx
   
Associated d 2 dy m2
(1 − x ) + l(l + 1) − y=0 (2.344) m integer
Legendre dx dx 1 − x2
d2 y dy
Bessel x2 2
+x + (x2 − m2 )y = 0 (2.345)
dx dx
d2 y dy
Hermite − 2x + 2αy = 0 (2.346)
dx2 dx
d2 y dy
Laguerre x + (1 − x) + αy = 0 (2.347)
dx2 dx
Associated d2 y dy
Laguerre x + (1 + k − x) + αy = 0 (2.348) k integer
dx2 dx
d2 y dy
Chebyshev (1 − x2 ) −x + n2 y = 0 (2.349) n integer
dx2 dx
Euler (or d2 y dy
Cauchy) x2 2
+ ax + by = f(x) (2.350) a,b constants
dx dx
dy
Bernoulli + p(x)y = q(x)y a (2.351) p,q functions of x
dx
d2 y
Airy = xy (2.352)
dx2
a Also known as the “conduction equation.” For thermal conduction, f ≡ T and D, the thermal diffusivity,

≡ κ ≡ λ/(ρcp ), where T is the temperature distribution, λ the thermal conductivity, ρ the density, and cp the specific
heat capacity of the material.
 main January 23, 2006 16:6

44 Mathematics

2.8 Integration

Standard formsa
      
dv
u dv = [uv] − v du (2.353) uv dx = v u dx − u dx dx (2.354)
dx

 
xn+1 1
n
x dx = (n = −1) (2.355) dx = ln|x| (2.356)
n+1 x

   
1 x 1
e dx = eax
ax
(2.357) ax
xe dx = e ax
− (2.358)
a a a2

 
f  (x)
lnax dx = x(lnax − 1) (2.359) dx = lnf(x) (2.360)
f(x)

   
x2 1 bax
xlnax dx = lnax − (2.361) bax dx = (b > 0) (2.362)
2 2 alnb

 
1 1 1 1 a + bx
dx = ln(a + bx) (2.363) dx = − ln (2.364)
a + bx b x(a + bx) a x

   
1 −1 1 1 bx
2
dx = (2.365) 2 2 2
dx = arctan (2.366)
(a + bx) b(a + bx) a +b x ab a

    
 1 1  x − a 
1 1  xn  dx = ln (2.368)
dx = ln (2.367) x2 − a2 2a  x + a 
x(xn + a) an  xn + a 
 
x 1 x −1
dx = ln|x2 ± a2 | (2.369) dx = (2.370)
x2 ± a2 2 (x2 ± a2 )n 2(n − 1)(x2 ± a2 )n−1

 
1 x! 1
dx = arcsin (2.371) dx = ln|x + (x2 ± a2 )1/2 | (2.372)
(a2 − x2 )1/2 a (x2 ± a2 )1/2

 
x 1 1 x!
dx = (x2 ± a2 )1/2 (2.373) dx = arcsec (2.374)
(x ± a2 )1/2
2 x(x2 − a2 )1/2 a a

aa and b are non-zero constants.

 main January 23, 2006 16:6

2.8 Integration 45

Trigonometric and hyperbolic integrals

 
sinx dx = −cosx (2.375) sinhx dx = coshx (2.376)
 
cosx dx = sinx (2.377) coshx dx = sinhx (2.378)
2
 
tanx dx = −ln|cosx| (2.379) tanhx dx = ln(coshx) (2.380)
  
 x  
 x 
cscx dx = ln tan  (2.381) cschx dx = ln tanh  (2.382)
2 2
 
secx dx = ln|secx + tanx| (2.383) sechx dx = 2arctan(ex ) (2.384)
 
cotx dx = ln|sinx| (2.385) cothx dx = ln|sinhx| (2.386)


sin(m − n)x sin(m + n)x
sinmx · sinnx dx = − (m2 = n2 ) (2.387)
2(m − n) 2(m + n)

cos(m − n)x cos(m + n)x
sinmx · cosnx dx = − − (m2 = n2 ) (2.388)
2(m − n) 2(m + n)

sin(m − n)x sin(m + n)x
cosmx · cosnx dx = + (m2 = n2 ) (2.389)
2(m − n) 2(m + n)

Named integrals
 x
2
Error function erf(x) = exp(−t2 ) dt (2.390)
π 1/2 0
 ∞
Complementary error 2
erfc(x) = 1 − erf(x) = exp(−t2 ) dt (2.391)
function π 1/2 x
 x  x
πt2 πt2
C(x) = dt; S(x) =
cos sin dt (2.392)
0 2 0 2
Fresnel integralsa  1/2 
1+i π
C(x) + i S(x) = erf (1 − i)x (2.393)
2 2
 x t
e
Exponential integral Ei(x) = dt (x > 0) (2.394)
−∞ t
 ∞
Gamma function Γ(x) = tx−1 e−t dt (x > 0) (2.395)
0
 φ
1
F(φ,k) = dθ (first kind) (2.396)
Elliptic integrals 0 (1 − k 2 sin2 θ)1/2
(trigonometric form)  φ
E(φ,k) = (1 − k 2 sin2 θ)1/2 dθ (second kind) (2.397)
0
a See also page 167.
 main January 23, 2006 16:6

46 Mathematics

Definite integrals
 ∞
1 π !1/2
e−ax dx =
2
(a > 0) (2.398)
0 2 a
 ∞
1
xe−ax dx =
2
(a > 0) (2.399)
0 2a
 ∞
n!
xn e−ax dx = (a > 0; n = 0,1,2,...) (2.400)
0 an+1
  2
∞
π !1/2 b
exp(2bx − ax ) dx =2
exp (a > 0) (2.401)
−∞ a a
 ∞ "
n −ax2 1 · 3 · 5 · ... · (n − 1)(2a)−(n+1)/2 (π/2)1/2 n > 0 and even
x e dx = (2.402)
0 2 · 4 · 6 · ... · (n − 1)(2a)−(n+1)/2 n > 1 and odd
 1
p!q!
xp (1 − x)q dx = (p,q integers > 0) (2.403)
0 (p + q + 1)!
 ∞  ∞
2 1 π !1/2
cos(ax ) dx = sin(ax2 ) dx = (a > 0) (2.404)
0 0 2 2a
 ∞  ∞ 2
sinx sin x π
dx = 2
dx = (2.405)
0 x 0 x 2
 ∞
1 π
a
dx = (0 < a < 1) (2.406)
0 (1 + x)x sinaπ

2.9 Special functions and polynomials

Gamma function
 ∞
Definition Γ(z) = tz−1 e−t dt [(z) > 0] (2.407)
0

n! = Γ(n + 1) = nΓ(n) (n = 0,1,2,...) (2.408)

1/2
Relations Γ(1/2) = π (2.409)
 
z z! Γ(z + 1)
= = (2.410)
w w!(z − w)! Γ(w + 1)Γ(z − w + 1)
 
1 1
Γ(z)  e−z z z−(1/2) (2π)1/2 1 + + − ··· (2.411)
12z 288z 2
Stirling’s formulas
(for |z|,n  1) n!  nn+(1/2) e−n (2π)1/2 (2.412)
ln(n!)  nlnn − n (2.413)
 main January 23, 2006 16:6

2.9 Special functions and polynomials 47

Bessel functions
x !ν  (−x2 /4)k
∞ Jν (x) Bessel function of the first
kind
Jν (x) = (2.414)
Series 2 k!Γ(ν + k + 1) Yν (x) Bessel function of the
expansion
Yν (x) =
k=0
Jν (x)cos(πν) − J−ν (x)
(2.415)
second kind
Γ(ν) Gamma function
2
sin(πν) ν order (ν ≥ 0)
Approximations 1 J0
"  x ν 0.5 J1
1
(0 ≤ x  ν)
Jν (x)  Γ(ν+1)
 2 
2 1/2
 (2.416) 0
πx cos x − 12 νπ − π4 (x  ν)
"   −0.5
Y0
−Γ(ν) x −ν
(0 < x  ν) Y1
Yν (x)   π
 2 
2 1/2
 (2.417) −1
πx sin x − 12 νπ − π4 (x  ν) 0 2 4 6 8 10
x
Iν (x) modified Bessel function of
Modified Bessel Iν (x) = (−i)ν Jν (ix) (2.418) the first kind
functions π
Kν (x) = iν+1 [Jν (ix) + iYν (ix)] (2.419) Kν (x) modified Bessel function of
2 the second kind

Spherical Bessel π !1/2 jν (x) spherical Bessel function

jν (x) = Jν+ 1 (x) (2.420) of the first kind [similarly
function 2x 2
for yν (x)]

Legendre polynomialsa
Legendre d2 Pl (x) dPl (x) Pl Legendre
(1 − x2 ) − 2x + l(l + 1)Pl (x) = 0 polynomials
equation dx2 dx
(2.421) l order (l ≥ 0)

Rodrigues’ 1 dl 2
formula Pl (x) = (x − 1)l (2.422)
2l l! dxl
Recurrence
(l + 1)Pl+1 (x) = (2l + 1)xPl (x) − lPl−1 (x) (2.423)
relation
 1
2
Orthogonality Pl (x)Pl  (x) dx = δll  (2.424) δll  Kronecker delta
−1 2l + 1


l/2   
l 2l − 2m l−2m l
Explicit form Pl (x) = 2−l (−1)m x (2.425) m binomial coefficients
m l
m=0
k wavenumber
exp(ikz) = exp(ikr cosθ) (2.426) z propagation axis
Expansion of ∞ z = r cosθ

plane wave = (2l + 1)il jl (kr)Pl (cosθ) (2.427) jl spherical Bessel
function of the first
l=0
kind (order l)

P0 (x) = 1 P2 (x) = (3x2 − 1)/2 P4 (x) = (35x4 − 30x2 + 3)/8

P1 (x) = x P3 (x) = (5x3 − 3x)/2 P5 (x) = (63x5 − 70x3 + 15x)/8
a Of the first kind.
 main January 23, 2006 16:6

48 Mathematics

Associated Legendre functionsa

   
Associated d dP m (x) m2 Plm associated
(1 − x2 ) l + l(l + 1) − P m (x) = 0
Legendre dx dx 1 − x2 l Legendre
equation functions
(2.428)
dm
From Plm (x) = (1 − x2 )m/2 Pl (x), 0 ≤ m ≤ l (2.429)
Legendre dxm Pl Legendre
(l − m)! m polynomials
polynomials Pl−m (x) = (−1)m P (x) (2.430)
(l + m)! l
m
Pm+1 (x) = x(2m + 1)Pmm (x) (2.431)
Recurrence Pmm (x) = (−1)m (2m − 1)!!(1 − x2 )m/2 (2.432) !! 5!! = 5 · 3 · 1 etc.
relations
(l − m + 1)Pl+1
m
(x) = (2l + 1)xPlm (x) − (l + m)Pl−1
m
(x)
(2.433)
 1 δll  Kronecker
(l + m)! 2
Orthogonality Plm (x)Plm (x) dx = δll  (2.434) delta
−1 (l − m)! 2l + 1

P00 (x) = 1 P10 (x) = x P11 (x) = −(1 − x2 )1/2

P20 (x) = (3x2 − 1)/2 P21 (x) = −3x(1 − x2 )1/2 P22 (x) = 3(1 − x2 )
a Of the first kind. Plm (x) can be defined with a (−1)m factor in Equation (2.429) as well as Equation (2.430).

Legendre polynomials associated Legendre functions

3
1
P0
P2
2
2
P1
P2

0.5 P3 P4 P 1
2
P00
P5

1
0 0
P 0 P1
2
0
−0.5
1
P1
−1 −1

−1 −0.5 0 0.5 1 −1 0 1
x x
 main January 23, 2006 16:6

2.9 Special functions and polynomials 49

Spherical harmonics
   
Differential 1 ∂ ∂ 1 ∂2
sinθ + 2 Yl m + l(l + 1)Yl m = 0 Yl m spherical
equation sinθ ∂θ ∂θ sin θ ∂φ2 harmonics

 1/2
(2.435) 2
2l + 1 (l − m)! Plm associated
Definitiona Yl m (θ,φ) = (−1)m Plm (cosθ)eimφ Legendre
4π (l + m)!
functions
(2.436)
  Y∗ complex
2π π
m∗ m  conjugate
Orthogonality Yl (θ,φ)Yl  (θ,φ)sinθ dθ dφ = δmm δll  (2.437) δll  Kronecker
φ=0 θ=0
delta
∞ 
 l
f(θ,φ) = alm Yl m (θ,φ) (2.438)
l=0 m=−l f continuous
Laplace series  2π  π function
where alm = Yl m∗ (θ,φ)f(θ,φ)sinθ dθ dφ
φ=0 θ=0
(2.439)

Solution to if ∇2 ψ(r,θ,φ) = 0, then ψ continuous

Laplace ∞
 
l function
# $
equation ψ(r,θ,φ) = Yl m (θ,φ) · alm rl + blm r−(l+1) (2.440) a,b constants
l=0 m=−l
 
1 3
Y00 (θ,φ) = Y10 (θ,φ) = cosθ
4π 4π
   
3 5 3 1
Y1±1 (θ,φ) = ∓ sinθ e±iφ Y20 (θ,φ) = cos2 θ −
8π 4π 2 2
 
15 15
Y2±1 (θ,φ) = ∓ sinθ cosθ e±iφ Y2±2 (θ,φ) = sin2 θ e±2iφ
8π 32π
 
1 7 ±1 1 21
Y30 (θ,φ) = (5cos2 θ − 3)cosθ Y3 (θ,φ) = ∓ sinθ(5cos2 θ − 1)e±iφ
2 4π 4 4π
 
1 105 2 1 35 3 ±3iφ
Y3±2 (θ,φ) = sin θ cosθ e±2iφ ±3
Y3 (θ,φ) = ∓ sin θ e
4 2π 4 4π
a Defined for −l ≤ m ≤ l, using the sign convention of the Condon–Shortley phase. Other sign conventions are

possible.
 main January 23, 2006 16:6

50 Mathematics

Delta functions
"
1 if i = j
δij = (2.441)
Kronecker delta 0 if i = j δij Kronecker delta
i,j,k,... indices (= 1,2 or 3)
δii = 3 (2.442)

123 = 231 = 312 = 1

Three- 132 = 213 = 321 = −1 (2.443)
dimensional all other ijk = 0
Levi–Civita ijk Levi–Civita symbol
symbol ijk klm = δil δjm − δim δjl (2.444) (see also page 25)
(permutation δij ijk = 0 (2.445)
tensor)a ilm jlm = 2δij (2.446)
ijk ijk = 6 (2.447)

 "
b
1 if a < 0 < b
δ(x) dx = (2.448)
a 0 otherwise
 b
f(x)δ(x − x0 ) dx = f(x0 ) (2.449) δ(x) Dirac delta function
Dirac delta a
f(x) smooth function of x
function δ(x − x0 )f(x) = δ(x − x0 )f(x0 ) (2.450)
a,b constants
δ(−x) = δ(x) (2.451)
δ(ax) = |a|−1 δ(x) (a = 0) (2.452)
−1/2 −n2 x2
δ(x)  nπ e (n  1) (2.453)
a The general symbol ijk... is defined to be +1 for even permutations of the suffices, −1 for odd permutations, and
0 if a suffix is repeated. The sequence (1,2,3,... ,n) is taken to be even. Swapping adjacent suffices an odd (or even)
number of times gives an odd (or even) permutation.

2.10 Roots of quadratic and cubic equations

Quadratic equations
x variable
Equation ax2 + bx + c = 0 (a = 0) (2.454)
a,b,c real constants
√
−b ± b2 − 4ac
x1,2 = (2.455)
Solutions 2a x1 ,x2 quadratic roots
−2c
= √ (2.456)
b ± b2 − 4ac

Solution x1 + x2 = −b/a (2.457)

combinations x1 x2 = c/a (2.458)
 main January 23, 2006 16:6

2.10 Roots of quadratic and cubic equations 51

Cubic equations
x variable
Equation ax3 + bx2 + cx + d = 0 (a = 0) (2.459)
a,b,c,d real constants
 
p=
1 3c b2
− 2 (2.460)
2
3 a a
 3 
Intermediate 1 2b 9bc 27d
definitions q= − 2 + (2.461) D discriminant
27 a3 a a
p ! 3 q ! 2
D= + (2.462)
3 2

If D ≥ 0, also define: If D < 0, also define:

!1/3   −3/2 
−q −q |p|
u= + D1/2 (2.463) φ = arccos (2.467)
2 2 3
−q !1/3
v= − D1/2 (2.464)  1/2
2 |p| φ
y1 = 2 cos (2.468)
y1 = u + v (2.465) 3 3
−(u + v) u − v 1/2  1/2
y2,3 = ±i 3 (2.466) |p| φ±π
2 2 y2,3 = −2 cos (2.469)
3 3
1 real, 2 complex roots
(if D = 0: 3 real roots, at least 2 equal) 3 distinct real roots

b xn cubic roots
Solutionsa xn = yn − (2.470) (n = 1,2,3)
3a

x1 + x2 + x3 = −b/a (2.471)
Solution
x1 x2 + x1 x3 + x2 x3 = c/a (2.472)
combinations
x1 x2 x3 = −d/a (2.473)
ay
n are solutions to the reduced equation y 3 + py + q = 0.
 main January 23, 2006 16:6

52 Mathematics

2.11 Fourier series and transforms

Fourier series
nπx !
∞
a0  nπx
f(x) = + an cos + bn sin (2.474)
2 L L f(x) periodic
n=1
 L function,
Real form 1 nπx period 2L
an = f(x)cos dx (2.475)
L −L L an ,bn Fourier
 coefficients
1 L nπx
bn = f(x)sin dx (2.476)
L −L L
 ∞  
inπx
f(x) = cn exp (2.477) cn complex
Complex n=−∞
L
  Fourier
form  L
1 −inπx coefficient
cn = f(x)exp dx (2.478)
2L −L L
 L ∞
1 a2 1   2 
|f(x)|2 dx = 0 + an + b2n (2.479)
Parseval’s 2L −L 4 2
n=1
|| modulus
theorem ∞

= |cn | 2
(2.480)
n=−∞

Fourier transforma
 ∞
F(s) = f(x)e−2πixs dx (2.481)
f(x) function of x
Definition 1 −∞
∞
2πixs
F(s) Fourier transform of f(x)
f(x) = F(s)e ds (2.482)
−∞
 ∞
F(s) = f(x)e−ixs dx (2.483)
−∞
Definition 2 
1 ∞
f(x) = F(s)eixs ds (2.484)
2π −∞
 ∞
1
F(s) = √ f(x)e−ixs dx (2.485)
2π −∞
Definition 3  ∞
1
f(x) = √ F(s)eixs ds (2.486)
2π −∞
a All three (and more) definitions are used, but definition 1 is probably the best.
 main January 23, 2006 16:6

2.11 Fourier series and transforms 53

Fourier transform theoremsa

 ∞
f,g general functions
Convolution f(x) ∗ g(x) = f(u)g(x − u) du (2.487)
−∞
∗ convolution

Convolution f ∗g =g ∗f (2.488) f f(x) 

 F(s)
2
rules f ∗ (g ∗ h) = (f ∗ g) ∗ h (2.489) g  G(s)
g(x) 

Convolution  Fourier transform

 F(s) ∗ G(s)
f(x)g(x)  (2.490) 
theorem relation
 ∞ correlation
Autocorrela-
f ∗ (x) f(x) = f ∗ (u − x)f(u) du (2.491) f∗ complex
tion −∞ conjugate of f
Wiener–
Khintchine f ∗ (x) f(x) 
 |F(s)|
2
(2.492)
theorem
 ∞
Cross-
f ∗ (x) g(x) = f ∗ (u − x)g(u) du (2.493)
correlation −∞
h,j real functions
Correlation ∗
h(x) j(x) 
 H(s)J (s) (2.494) H H(s) 
 h(x)
theorem
J J(s) 
 j(x)
 ∞  ∞
Parseval’s
f(x)g ∗ (x) dx = F(s)G∗ (s) ds (2.495)
relationb −∞ −∞
 ∞  ∞
Parseval’s
|f(x)| dx =
2
|F(s)|2 ds (2.496)
theoremc −∞ −∞

df(x)

 2πisF(s) (2.497)
dx
Derivatives d df(x) dg(x)
[f(x) ∗ g(x)] = ∗ g(x) = ∗ f(x)
dx dx dx
(2.498)
a Defining
%∞ −2πixs
the Fourier transform as F(s) = −∞ f(x)e dx.
b Also called the “power theorem.”
c Also called “Rayleigh’s theorem.”

Fourier symmetry relationships

f(x) 
 F(s) definitions
even 
 even real: f(x) = f ∗ (x)
odd 
 odd imaginary: f(x) = −f ∗ (x)
real, even 
 real, even even: f(x) = f(−x)
real, odd 
 imaginary, odd odd: f(x) = −f(−x)
imaginary, even 
 imaginary, even Hermitian: f(x) = f ∗ (−x)
complex, even 
 complex, even anti-Hermitian: f(x) = −f ∗ (−x)
complex, odd 
 complex, odd
real, asymmetric 
 complex, Hermitian
imaginary, asymmetric 
 complex, anti-Hermitian
 main January 23, 2006 16:6

54 Mathematics

Fourier transform pairsa

 ∞
f(x) 
 F(s) = f(x)e−2πisx dx (2.499)
−∞
1
f(ax) 
 F(s/a) (a = 0, real) (2.500)
|a|
f(x − a) 
 e−2πias F(s) (a real) (2.501)
n
d
f(x) 
 (2πis)n F(s) (2.502)
dxn
δ(x) 
 1 (2.503)
δ(x − a) 
 e−2πias (2.504)
2a
e−a|x| 
 (a > 0) (2.505)
a2 + 4π 2 s2
8iπas
xe−a|x| 
 (a > 0) (2.506)
(a2 + 4π 2 s2 )2
√
e−x a πe−π a s
2
/a2 
2 2 2
 (2.507)
1 a ! a !
sinax 
 δ s− −δ s+ (2.508)
2i 2π 2π
1  a ! a !
cosax 
 δ s− +δ s+ (2.509)
2 2π 2π
∞
 1 ∞
n !
δ(x − ma) 
 δ s− (2.510)
m=−∞
a n=−∞ a
"
0 x<0 1 i
f(x) = (“step”) 
 δ(s) − (2.511)
1 x>0 2 2πs
"
1 |x| ≤ a sin2πas
f(x) = (“top hat”) 
 = 2asinc2as (2.512)
0 |x| > a πs
 
 1 − |x| |x| ≤ a 1
f(x) = a (“triangle”) 
 (1 − cos2πas) = asinc2 as (2.513)
 2π 2 as2
0 |x| > a
a Equation (2.499) defines the Fourier transform used for these pairs. Note that sincx ≡ (sinπx)/(πx).
 main January 23, 2006 16:6

2.12 Laplace transforms 55

2.12 Laplace transforms

Laplace transform theorems


Definitiona F(s) = L{f(t)} =
∞
f(t)e−st dt (2.514)
L{} Laplace 2
transform
0
& ∞ '
F(s) · G(s) = L f(t − z)g(z) dz (2.515) F(s) L{f(t)}
Convolutionb 0 G(s) L{g(t)}
= L{f(t) ∗ g(t)} (2.516) ∗ convolution

 γ+i∞
1
f(t) = est F(s) ds (2.517)
Inversec 2πi γ−i∞ γ constant

= residues (for t > 0) (2.518)

& '  dr f(t) 

n−1
Transform of dn f(t)
L = sn L{f(t)} − sn−r−1  n integer > 0
derivative dtn dtr t=0
r=0
(2.519)
Derivative of dn F(s)
= L{(−t)n f(t)} (2.520)
transform dsn

Substitution F(s − a) = L{eat f(t)} (2.521) a constant

e−as F(s) = L{u(t − a)f(t − a)} (2.522)

" u(t) unit step
Translation 0 (t < 0) function
where u(t) = (2.523)
1 (t > 0)
a If|e−s0 t f(t)| is finite for sufficiently large t, the Laplace transform exists for s > s0 .
b Also known as the “faltung (or folding) theorem.”
c Also known as the “Bromwich integral.” γ is chosen so that the singularities in F(s) are left of the integral line.
 main January 23, 2006 16:6

56 Mathematics

Laplace transform pairs

 ∞
f(t) =⇒ F(s) = L{f(t)} = f(t)e−st dt (2.524)
0
δ(t) =⇒ 1 (2.525)
1 =⇒ 1/s (s > 0) (2.526)
n!
tn =⇒ (s > 0, n > −1) (2.527)
sn+1

1/2 π
t =⇒ (2.528)
4s3

−1/2 π
t =⇒ (2.529)
s
1
eat =⇒ (s > a) (2.530)
s−a
1
teat =⇒ (s > a) (2.531)
(s − a)2
s
(1 − at)e−at =⇒ (2.532)
(s + a)2
2
t2 e−at =⇒ (2.533)
(s + a)3
a
sinat =⇒ 2 (s > 0) (2.534)
s + a2
s
cosat =⇒ 2 (s > 0) (2.535)
s + a2
a
sinhat =⇒ 2 (s > a) (2.536)
s − a2
s
coshat =⇒ 2 (s > a) (2.537)
s − a2
a
e−bt sinat =⇒ (2.538)
(s + b)2 + a2
s+b
e−bt cosat =⇒ (2.539)
(s + b)2 + a2
e−at f(t) =⇒ F(s + a) (2.540)
 main January 23, 2006 16:6

2.13 Probability and statistics 57

2.13 Probability and statistics

Discrete statistics 2
xi data series
1 
N
Mean x = xi (2.541) N series length
N · mean value
i=1

1 
N
var[·] unbiased
Variancea var[x] = (xi − x )2 (2.542) variance
N −1
i=1
Standard 1/2 σ standard
σ[x] = (var[x]) (2.543)
deviation deviation

N  3
N xi − x
Skewness skew[x] = (2.544)
(N − 1)(N − 2) σ
i=1
 N  4 
1  xi − x
Kurtosis kurt[x]  −3 (2.545)
N σ
i=1
N x,y data series to
Correlation i=1 (xi − x)(y − y )
r =  i (2.546) correlate
coefficientb N N r correlation
i=1 (xi − x ) i=1 (yi − y )
2 2
coefficient
a If x is derived from the data, {x }, the relation is as shown. If x is known independently, then an unbiased
i
estimate is obtained by dividing the right-hand side by N rather than N − 1.
b Also known as “Pearson’s r.”

Discrete probability distributions

distribution pr(x) mean variance domain

n
n binomial
Binomial x px (1 − p)n−x np np(1 − p) (x = 0,1,... ,n) (2.547) x
coefficient

Geometric (1 − p)x−1 p 1/p (1 − p)/p2 (x = 1,2,3,...) (2.548)

Poisson λx exp(−λ)/x! λ λ (x = 0,1,2,...) (2.549)

 main January 23, 2006 16:6

58 Mathematics

Continuous probability distributions

distribution pr(x) mean variance domain

1 a+b (b − a)2
Uniform (a ≤ x ≤ b) (2.550)
b−a 2 12
Exponential λexp(−λx) 1/λ 1/λ2 (x ≥ 0) (2.551)
 
Normal/ 1 −(x − µ)2
√ exp µ σ2 (−∞ < x < ∞) (2.552)
Gaussian σ 2π 2σ 2
e−x/2 x(r/2)−1
Chi-squareda r 2r (x ≥ 0) (2.553)
2r/2 Γ(r/2)
 2
x −x ( π!
Rayleigh exp σ π/2 2σ 2 1 − (x ≥ 0) (2.554)
σ2 2σ 2 4
Cauchy/ a
(none) (none) (−∞ < x < ∞) (2.555)
Lorentzian π(a2 + x2 )
a With r degrees of freedom. Γ is the gamma function.

Multivariate normal distribution

pr probability density
# $
exp − 12 (x − µ)C−1 (x − µ)T k number of dimensions
Density function pr(x) = C covariance matrix
(2π)k/2 [det(C)]1/2
(2.556) x variable (k dimensional)
µ vector of means
T transpose
Mean µ = (µ1 ,µ2 ,... ,µk ) (2.557) det determinant
µi mean of ith variable

Covariance C = σij = xi xj − xi xj (2.558) σij components of C

Correlation σij
r= (2.559) r correlation coefficient
coefficient σi σj

Box–Muller x1 = (−2lny1 )1/2 cos2πy2 (2.560) xi normally distributed deviates

yi deviates distributed
transformation x2 = (−2lny1 )1/2 sin2πy2 (2.561) uniformly between 0 and 1
 main January 23, 2006 16:6

2.13 Probability and statistics 59

Random walk
x displacement after N steps
  (can be positive or negative)
One- 1 −x 2
pr(x) probability density of x
pr(x) = exp %∞
dimensional
2
(2πNl ) 1/2 2Nl 2
(2.562) N
( −∞ pr(x) dx = 1)
number of steps
2
l step length (all equal)
rms 1/2 xrms root-mean-squared
xrms = N l (2.563)
displacement displacement from start point

a !3 r radial distance from start

pr(r) = exp(−a2 r2 ) (2.564) point
Three- π 1/2
 1/2 pr(r) probability
%∞ density of r
dimensional 3 ( 0 4πr2 pr(r) dr = 1)
where a =
2Nl 2 a (most probable distance)−1
 1/2
8 r mean distance from start
Mean distance r = N 1/2 l (2.565) point
3π

rrms = N 1/2 l (2.566) rrms root-mean-squared distance

rms distance from start point

Bayesian inference
 pr(x) probability (density) of x
Conditional
pr(x) = pr(x|y  )pr(y  ) dy  (2.567) pr(x|y  ) conditional probability of x
probability given y 
Joint pr(x,y) = pr(x)pr(y|x) (2.568) pr(x,y) joint probability of x and y
probability
pr(x|y) pr(y)
Bayes’ theorema pr(y|x) = (2.569)
pr(x)
a Inthis expression, pr(y|x) is known as the posterior probability, pr(x|y) the likelihood, and pr(y) the prior
probability.
 main January 23, 2006 16:6

60 Mathematics

2.14 Numerical methods

Straight-line fittinga
 
Data {xi },{yi } n points (2.570)
y
y = mx + c
Weights b {wi } (2.571)
(0,c)
Model y = mx + c (2.572)
(x,y)
Residuals di = yi − mxi − c (2.573)

1   ! x
Weighted (x,y) =  wi xi , wi yi
centre wi
(2.574)
Weighted 
D= wi (xi − x)2 (2.575)
moment
1
m= wi (xi − x)yi (2.576)
D  2
Gradient
1 wi di
var[m]  (2.577)
D n−2

c = y − mx (2.578)
Intercept   2
1 x2 wi di
var[c]   + (2.579)
wi D n−2
a Least-squares fit of data to y = mx + c. Errors on y-values only.
b If the errors on yi are uncorrelated, then wi = 1/var[yi ].

Time series analysisa


M/2 ri response function
Discrete
convolution (r s)j = sj−k rk (2.580) si time series
k=−(M/2)+1 M response function duration
Bartlett  
 j − N/2  wj windowing function
(triangular) wj = 1 −   (2.581)
window N/2  N length of time series

Welch  2 1 Welch Hamming

j − N/2
(quadratic) wj = 1 − (2.582) 0.8
window N/2
   w 0.6
Bartlett
Hanning 1 2πj 0.4
wj = 1 − cos (2.583) Hanning
window 2 N 0.2
  0
Hamming 2πj 0 0.2 0.4 0.6 0.8 1
wj = 0.54 − 0.46cos (2.584)
window N j/N
a The time series runs from j = 0...(N − 1), and the windowing functions peak at j = N/2.
 main January 23, 2006 16:6

2.14 Numerical methods 61

Numerical integration

h
f(x)
2

x
x0 xN

 h = (xN − x0 )/N
xN
h (subinterval
f(x) dx  (f0 + 2f1 + 2f2 + ··· width)
Trapezoidal rule x0 2
fi fi = f(xi )
+ 2fN−1 + fN ) (2.585)
N number of
subintervals
 xN
h
f(x) dx  (f0 + 4f1 + 2f2 + 4f3 + ···
Simpson’s rulea x0 3
+ 4fN−1 + fN ) (2.586)
aN must be even. Simpson’s rule is exact for quadratics and cubics.

Numerical differentiationa
df 1
 [−f(x + 2h) + 8f(x + h) − 8f(x − h) + f(x − 2h)] (2.587)
dx 12h
1
∼ [f(x + h) − f(x − h)] (2.588)
2h
d2 f 1
2
 [−f(x + 2h) + 16f(x + h) − 30f(x) + 16f(x − h) − f(x − 2h)] (2.589)
dx 12h2
1
∼ 2 [f(x + h) − 2f(x) + f(x − h)] (2.590)
h
d3 f 1
∼ [f(x + 2h) − 2f(x + h) + 2f(x − h) − f(x − 2h)] (2.591)
dx3 2h3
a Derivatives of f(x) at x. h is a small interval in x.

Relations containing “” are O(h4 ); those containing “∼” are O(h2 ).

Numerical solutions to f(x) = 0

xn − xn−1 f function of x
Secant method xn+1 = xn − f(xn ) (2.592)
f(xn ) − f(xn−1 ) xn f(x∞ ) = 0

Newton–Raphson f(xn )
xn+1 = xn − (2.593) f = df/dx
method f  (xn )
 main January 23, 2006 16:6

62 Mathematics

Numerical solutions to ordinary differential equationsa

dy
if = f(x,y) (2.594)
dx
Euler’s method and h = xn+1 − xn (2.595)
2
then yn+1 = yn + hf(xn ,yn ) + O(h ) (2.596)

dy
if = f(x,y) (2.597)
dx
and h = xn+1 − xn (2.598)
Runge–Kutta k1 = hf(xn ,yn ) (2.599)
method k2 = hf(xn + h/2,yn + k1 /2) (2.600)
(fourth-order) k3 = hf(xn + h/2,yn + k2 /2) (2.601)
k4 = hf(xn + h,yn + k3 ) (2.602)
k1 k2 k3 k4
then yn+1 = yn + + + + + O(h5 ) (2.603)
6 3 3 6
a Ordinary dy
differential equations (ODEs) of the form dx = f(x,y). Higher order equations should be
reduced to a set of coupled first-order equations and solved in parallel.