App D

APPENDIX D.
VECTOR ANALYSIS 1
Appendix D
Vector Analysis
The following conventions are used in this appendix and throughout the book:
f, g, , are scalar functions of x, t;
A, B, C, D are vector functions of x, t;
A = |A|
A A is the magnitude or length of the vector A;

e
A
A/A is a unit vector in the A direction;
x is the vector from the origin to the point (x, y, z);
T, W, AB, etc., are dyad (second rank tensor) functions of x, t that will
be called simply tensors;
I is the identity tensor or unit dyad;
T
T
is the transpose of tensor T (interchange of indices of the tensor
elements), a tensor;
tr(T) is the trace of the tensor T (sum of its diagonal elements), a scalar;
det(T) T is the determinant of the tensor T (determinant of the
matrix of tensor elements), a scalar.
D.1 Vector Algebra
Basic algebraic relations:
A+B = B+A, commutative addition (D.1)
A+ (B+C) = (A+B) +C, associative addition (D.2)
AB = A+ (B), dierence (D.3)
fA = Af, commutative scalar multiplication (D.4)
(f + g)A = fA+ gA, distributive scalar multiplication (D.5)
f(A+B) = fA+ fB, distributive scalar multiplication (D.6)
f(gA) = (fg)A, associative scalar multiplication (D.7)
DRAFT 11:26
October 11, 2002 c J.D Callen, Fundamentals of Plasma Physics
APPENDIX D. VECTOR ANALYSIS 2
Dot product:
A B = 0 implies A = 0 or B = 0, or A B (D.8)
A B = B A, commutative dot product (D.9)
A (B+C) = A B+A C, distributive dot product (D.10)
(fA) (gB) = fg(A B), associative scalar, dot product (D.11)
Cross product:
AB = 0 implies A = 0 or B = 0, or A B (D.12)
AB = BA, AA = 0, anti-commutative cross product (D.13)
A(B+C) = AB+AC, distributive cross product (D.14)
(fA)(gB) = fg(AB), associative scalar, cross product (D.15)
Scalar relations:
A BC = AB C = (CA) B, dot-cross product (D.16)
(AB) (CD) = (A C)(B D) (A D)(B C) (D.17)
(AB) (CD) + (BC) (AD) + (CA) (BD) = 0 (D.18)
Vector relations:
A(BC) = B(A C) C(A B), bac cab rule
= (CB)A = A (CBBC) (D.19)
A(BC) +B(CA) +C(AB) = 0 (D.20)
(A B)C = A (BC), associative dot product (D.21)
(AB)(CD) = C(AB D) D(AB C)
= B(CD A) A(CD B) (D.22)
Projection of a vector A in directions relative to a vector B:
A = A
(B/B) +A
= A
b +A
(D.23)
b B/B, unit vector in B direction (D.24)

A
B A/B =

b A, component of A along B (D.25)
A
B(BA)/B
2
, component of A perpendicular to B
=
b(
bA) (D.26)
D.2 Tensor Algebra
Scalar relations:
I : AB (I A) B = A B (D.27)
AB: CD A (B C)D = (B C)(A D)
DRAFT 11:26
dotproduct crossproduct
dot-crossproduct
A
B
A
B
A
B
C
{
{
{
Figure D.1: Schematic illustration of dot, cross and dot-cross products of vec-
tors.
DRAFT 11:26
= D AB C = B CD A (D.28)
I : T = tr(T), T: T |T|
2
(D.29)
T: AB = (T A) B = B T A (D.30)
AB: T = A (B T) = B T A (D.31)
BT: W = (T W)
T
: BI (D.32)
Vector relations:
I A = A I = A (D.33)
A T
T
= T A, T
T
A = A T (D.34)
A (CBBC) = A(BC)
= B(A C) C(A B), bac cab rule (D.35)
(AC) T = A (CT) = C (AT) (D.36)
T (AC) = (TA) C = (TC) A (D.37)
A (TC) = (A T)C = C(A T) (D.38)
(AT) C = A(T C) = (T C)A (D.39)
A (TC) C (TA) = [I tr(T) T] (AC) (D.40)
(AT) C(CT) A = (AC) [I tr(T) T] (D.41)
Tensor relations:
I AB = (I A)B = AB, AB I = A(B I) = AB (D.42)
IA = IA (D.43)
A(BC) = (AB)C, (AB)C = A(BC) (D.44)
(AB)I = I(AB) = BAAB (D.45)
(AT)
T
= T
T
A, (TA)
T
= AT
T
(D.46)
(AT) (AT)
T
= I[A tr(T) T A] (D.47)
(TA) (TA)
T
= I[A tr(T) A T] (D.48)
T
S
=
1
2
(T +T
T
), symmetric part of tensor T (D.49)
T
A
=
1
2
(T T
T
), anti-symmetric part of tensor T (D.50)
BT
S
B = B
2
T
S
(BB T
S
+T
S
BB)
(IB
2
BB)(IB
2
BB) T
S
/B
2
BB(BB T
S
)/B
2
(D.51)
D.3 Derivatives
Temporal derivatives:
dA
dt
is a vector tangent to the curve dened byA(t) (D.52)
d
dt
(fA) =
df
dt
A+ f
dA
dt
(D.53)
DRAFT 11:26
d
dt
(A+B) =
dA
dt
+
dB
dt
(D.54)
d
dt
(A B) =
dA
dt
B+A
dB
dt
(D.55)
d
dt
(AB) =
dA
dt
B+A
dB
dt
(D.56)
Denitions of partial derivatives in space ( /x = del or nabla is the
dierential vector operator):
f
f
x
, gradient of scalar function f, a vector vector in direction
of and measure of the greatest rate of spatial change of f (D.57)
A

x
A, divergence of vector function A, a scalar
divergence ( A > 0) or convergence ( A < 0) of A lines (D.58)
A

x
A, curl (or rotation) of vector function A, a vector
1
vorticity of A lines (D.59)
2
f f, del square or Laplacian (divergence of gradient)
derivative of scalar function f, a scalar, which is sometimes
written as f three-dimensional measure of curvature of f
(f is larger where
2
f < 0 and smaller where
2
f > 0) (D.60)
2
A ( )A = ( A) (A), Laplacian derivative
of vector function A, a vector (D.61)
For the general vector coordinate x xe
x
+ye
y
+ze
z
and |x|
_
x
2
+ y
2
+ z
2
:
x = 3, (x/|x|) = 2/|x| (D.62)
x = 0, (x/|x|) = 0 (D.63)
|x| = x/|x|, (1/|x|) = x/|x|
3
(D.64)
x = I (D.65)
(A )(x/|x|) = [A(x A)x/|x|
2
]/|x| A
/|x| (D.66)
2
(1/|x|) (1/|x|) = (x/|x|
3
) = 4(x) (D.67)
1
Rigorously speaking, the cross product of two vectors and the curl of a vector are pseudo-
vectors because they are anti-symmetric contractions of second rank tensors see tensor
references at end of this appendix.
DRAFT 11:26
First derivatives with scalar functions:
(f + g) = f +g (D.68)
(fg) = (f)g + fg = (gf) (D.69)
(fA) = (f)A+ fA (D.70)
fA = f A+ f A (D.71)
fA = fA+ fA (D.72)
fT = f T + f T (D.73)
fT = fT + fT (D.74)
First derivative scalar relations:
(A+B) = A+ B (D.75)
(AB) = B AA B (D.76)
(B )(A C) = C (B )A+A (B )C
CB: A+AB: C (D.77)
A B CC B A (CAAC) : B = (AC) B (D.78)
2A B C 2CA: B = A (B C) +C (B A)
B (A C) + (BC) (A)
+ (BA) (C) + (AC) (B) (D.79)
I : B = B (D.80)
AI : B = A B (D.81)
A T = (A T) A: T = (A T) T: A (D.82)
First derivative vector relations:
(A+B) = A+B (D.83)
(A B) = A(B) +B(A) + (A )B+ (B )A
= (A) B+ (B) A (D.84)
(B
2
/2) (B B/2) = B(B) + (B )B = (B) B (D.85)
(B )(AC) = (B )AC+A(B )C (D.86)
AB = ( A)B+ (A )B = ( A)B+A (B) (D.87)
I = 0 (D.88)
(IA) = A (D.89)
A(B) = (B) AA (B) = (B) A(A )B (D.90)
(AB) = A( B) B( A) + (B )A(A )B
= (BAAB) (D.91)
A BC+CB A = C[A(B)] (D.92)
A BCC BA = [( B)I B] (AC) (D.93)
AB CCB A = (AC) [( B)I B] (D.94)
DRAFT 11:26
First derivative tensor relations:
I B = B, B I = B (D.95)
AB = (A)BAB (D.96)
(AB) = ABBA (D.97)
AB+BA
= I[( B)A(B) A] + [A (B)]I A(B)
= I[( B)AA (B)] + [A (B)]I (B)A (D.98)
BA+ (AB)
T
= [A (B)]I A(B) (D.99)
AB+ (BA)
T
= [A (B)]I (B)A (D.100)
AB(AB)
T
= I[( B)A(B) A] (D.101)
BA(BA)
T
= [( B)AA (B)]I (D.102)
Second derivative relations:
f
2
f (D.103)
f = 0 (D.104)
fg = 0 (D.105)
A
2
A = ( A) (A) (D.106)
A = 0 (D.107)
(B )A = (B )( A) (A) (B) (D.108)
[(A )A]
= (A )(A) + ( A)(A) [(A) ]A (D.109)
Derivatives of projections of A in B direction [
b B/B, A = A
b + A
,
A
b A, A
b(
bA), (
b )
b =
b(
b) ]:
A = (A
/B)( B) + (B )(A
/B) + A
(D.110)
A
= A
[ ln B + (
b )
b] (1/B)
b (BA) (D.111)
b A
b : A = (
b )A
b )
b
= A ln B (1/B)
b (BA) + A(A
/B)( B) (D.112)
For A
= (1/B
2
) Bf,

b (BA
) = (
b f)(
b) (D.113)
D.4 Integrals
For a volume V enclosed by a closed, continuous surface S with dierential
volume element d
3
x and dierential surface element dS ndS where n is the
unit normal outward from the volume V , for well-behaved functions f, g, A, B
and T:
_
V
d
3
xf =
_
_
S
dSf, (D.114)
DRAFT 11:26
_
V
d
3
x A =
_
_
S
dS A, divergence or Gauss theorem, (D.115)
_
V
d
3
x T =
_
_
S
dS T, (D.116)
_
V
d
3
xA =
_
_
S
dSA, (D.117)
_
V
d
3
xf
2
g =
_
V
d
3
xf g +
_
_
S
dS fg,
Greens rst identity, (D.118)
_
V
d
3
x(f
2
g g
2
f) =
_
_
S
dS (fg gf),
Greens second identity, (D.119)
_
V
d
3
x[A (B) B (A)]
=
_
_
dS [B(A) A(B)],
vector form of Greens second identity. (D.120)
The gradient, divergence and curl partial dierential operators can be dened
using integral relations in the limit of small surfaces S encompassing small
volumes V , as follows:
f lim
V 0
_
1
V
_
_
S
dSf
_
gradient, (D.121)
A lim
V 0
_
1
V
_
_
S
dS A
_
divergence, (D.122)
A lim
V 0
_
1
V
_
_
S
dSA
_
curl. (D.123)
For S representing an open surface bounded by a closed, continuous contour C
with line element d which is dened to be positive when the right-hand-rule
sense of the line integral around C points in the dS direction:
__
S
dSf =
_
C
df, (D.124)
__
S
dS A =
_
C
d A, Stokes theorem, (D.125)
__
S
(dS)A =
_
C
dA, (D.126)
__
S
dS (fg) =
_
C
d fg =
_
C
f dg =
_
C
g df,
Greens theorem. (D.127)
The appropriate dierential line element d, surface area dS, and volume d
3
x
can be dened in terms of any three dierential line elements d(i), i = 1, 2, 3
DRAFT 11:26
that are linearly independent [i.e., d(1) d(2)d(3) = 0] by
d = d(i), i = 1, 2, or 3, dierential line element, (D.128)
dS = d(i)d(j), dierential surface area, (D.129)
d
3
x = d(1) d(2)d(3), dierential volume. (D.130)
In exploring properties of uids and plasmas we often want to know how
the dierential line, surface and volume elements change as they move with the
uid ow velocity V. In particular, when taking time derivatives of integrals,
we need to know what the time derivatives of these dierentials are as they
are carried along with a uid. To determine this, note rst that if the ow is
uniform then all points in the uid would be carried along in the same direction
at the same rate; hence, the time derivatives of the dierentials would vanish.
However, if the ow is nonuniform, the dierential line elements and hence all
the dierentials would change in time. To calculate the time derivatives of the
dierentials, consider the motion of two initially close points x
1
, x
2
as they are
carried along with a uid ow velocity V(x, t). Using the Taylor series expansion
V(x
2
, t) = V(x
1
, t)+(x
2
x
1
) V+ and integrating the governing equation
dx/dt = V over time, we obtain
x
2
x
1
= x
2
(t = 0) x
1
(t = 0) +
_
t
0
dt
(x
2
x
1
) V+ (D.131)
in which x
2
(t = 0) and x
1
(t = 0) are the initial positions at t = 0. Taking the
time derivative of this equation and identifying the dierential line element d
as x
2
x
1
in the limit where the points x
2
and x
1
become innetesimally close,
we nd
d

d
dt
(d) = d V. (D.132)
The time derivative of the dierential surface area dS can be calculated by
taking the time derivative of (D.129) and using this last equation to obtain
d
S
d
dt
(dS) = d
(1)d(2) + d(1)d
(2)
= d(1) Vd(2) d(2) Vd(1)
= ( V) dS V dS (D.133)
in which (D.93) and (D.33) have been used in obtaining the last form. Similarly,
the time derivative of the dierential volume element moving with the uid is
d
dt
(d
3
x) = d
(3) dS + d(3) d
S
= d(3) V dS + d(3) ( V)dS d(3) V dS
= ( V) d
3
x, (D.134)
which shows that the dierential volume in a compresssible uid increases or
decreases according to whether the uid is rarefying ( V > 0) or compressing
( V < 0).
DRAFT 11:26
D.5 Vector Field Representations
Any vector eld B can be expressed in terms of a scalar potential
M
and a
vector potential A:
B =
M
+A, potential representation. (D.135)
The
M
part of B represents the longitudinal or irrotational (
M
= 0)
component while the A part represents the transverse or solenoidal compo-
nent ( A = 0). A vector eld B that satises B = 0 is called a lon-
gitudinal or irrotational eld; one that satises B = 0 is called a solenoidal
or transverse eld. For a B(x) that vanishes at innity, the potentials
M
and
A are given by Greens function solutions
M
(x) =
_
d
3
x
( B)
x
4|x x
|
, A(x) =
_
d
3
x
(B)
x
4|x x
|
. (D.136)
When there is symmetry in a coordinate (i.e., a two or less dimensional
system), a solenoidal vector eld B can be written in terms of a stream function
in such a way that it automatically satises the solenoidal condition B = 0:
B = = || e
= , stream function form.

(D.137)
For this situation the vector potential becomes
A = = || e
. (D.138)
For a fully three-dimensional situation with no symmetry, a solenoidal vector
eld B can in general be written as
B = , Clebsch representation, (D.139)
In this representation and are stream functions that are constant along the
vector eld B since B = 0 and B = 0.
D.6 Properties Of Curve Along A Vector Field
The motion of a point x along a vector eld B is described by
dx
d
=
B
B
=

b T, tangent vector (D.140)
in which d is a dierential distance along B. The unit vector

b is tangent to
the vector eld B(x) at the point x and so is often written as T a unit tangent
vector.
The curvature vector of the vector eld B is dened by

d
2
x
d
2
=
d
b
d
= (
b )
b =
b(
b), curvature vector (D.141)

DRAFT 11:26
in which (D.85) has been used in the obtaining the last expression. The unit
vector in the curvature vector direction is dened by
(
b )
b/ |(
b )
b|, curvature unit vector. (D.142)

The local radius of curvature vector R
C
is in the opposite direction from the
curvature vector and is dened by
R
C
/||
2
, = R
C
/R
2
C
, radius of curvature. (D.143)
Hence, |R
C
| R
C
= 1/|| is the magnitude of the local radius of curvature
the radius of the circle tangent to the vector eld B(x) at the point x.
A triad of orthogonal unit vectors (see Fig. D.2) can be constructed from the
tangent unit vector T and an arbitrary unit vector N normal (or perpendicular)
to the vector eld B(x) at the point x:
T
b, N and B TN =

bN, Frenet unit vector triad (D.144)
in which B is the binormal unit vector, the third orthogonal unit vector. The
component of a vector C in the direction of the vector eld B is called the
parallel component: C
T C =

b C. The component in the N direction is
called the normal component: C
N
N C. The component in the B direction,
which is perpendicular to the TN plane, is called the binormal component:
C
B
B C = TN C.
Consider for example the components of the curvature vector . Since
b = 0, the curvature vector has no parallel component (
= 0) the
curvature vector for the vector eld B(x) is perpendicular to it at the point
x. The components of the curvature vector relative to a surface (x) = con-
stant in which the vector eld lies (i.e., B = 0) can be specied as follows.
Dene the normal to be in the direction of the gradient of : N /||.
Then, the components of the curvature vector perpendicular to (normal) and
lying within (geodesic) the surface are given by
n
= N = /||, normal curvature, (D.145)
g
= B = (
b) /|
b|, geodesic curvature. (D.146)

The torsion (twisting) of a vector eld B is dened by

dB
d
= (
b )(
bN), torsion vector. (D.147)

The binormal component of the torsion vector vanishes (
B
B = 0). The
normal component of the torsion vector locally denes the scale length L
along
the vector B over which the vector eld B(x) twists through an angle of one
radian:
L
1/|
N
|,
N
N
dB
d
= N (
b )(
bN), torsion length. (D.148)

DRAFT 11:26
B
B
B
curvature torsion
shear
R
c
Figure D.2: Properties (curvature, torsion, shear) of a spatially inhomogeneous
vector eld B(x). The unit vector

b B/B = dx/d = T is locally tangent to
the vector eld B. The unit normal N is perpendicular to the vector eld B,
shown here in the curvature direction. The binormal B is orthogonal to both

b
and N.
DRAFT 11:26
If the unit normal N is taken to be in the direction, the parallel com-
ponent of the torsion vector is equal to the geodesic curvature [

b =
(
b) (
b )
b/ |
b|
g
].
The local shear (dierential twisting motion, or nonplanar dierential tan-
gential motion in the plane dened by

b = T and N) in a vector eld B is given
by the binormal component of the curl or rotation in the binormal unit vector:
B B = (
bN) (
bN) 1/L
S
, local shear. (D.149)
The shear length L
S
is dened as the scale length over which the vector eld
B(x) shears through an angle of one radian. The parallel component of the
total curl or rotation of a vector eld B is given by a combination of its torsion
and shear, and N N:

b = (
bN) (
bN) 2N (
b )(
bN) +N N
= + 2
N
+N N, total rotation in B eld. (D.150)
If the normal N is taken to be in the direction, N N = 0 and then
=
(
b) (
b)
|
b|
2
1
L
S
, local shear with N /||,
(D.151)
and

b = + 2
N
. (D.152)
In the absence of shear ( = 0), this last relation yields
N
= (1/2)
b
the torsion for rigid body rotation is just half the parallel component of the
rotation in the vector eld B.
In most applied mathematics books the normal N is taken to be in the curva-
ture vector direction (i.e., N ) instead of the direction. Then, the parallel
component of the torsion vector also vanishes [

b =

b (
b )
b =
b = 0] and

N
N, for N . (D.153)
For this case the interrelationships between the triad of unit vectors T, N, B are
given by the Frenet-Serret formulas:
dT
d
=
N
N, T B/B
b,
dN
d
=
N
T +
N
B, N = (
b )
b/ |(
b )
b|,
dB
d
=
N
N, B TN =

b .
(D.154)
The local shear and total rotation in the vector eld B for this case are as
given above in (D.149) and (D.150), respectively, for a general unit normal N.
DRAFT 11:26
D.7 Base Vectors and Vector Components
The three vectors e
1
, e
2
, e
3
, which are not necessarily orthogonal, can be used as
a basis for a three-dimensional coordinate system if they are linearly independent
(i.e., e
1
e
2
e
2
= 0). The three reciprocal base vectors e
1
, e
2
, e
3
are dened by
e
i
e
j
=
i
j
, (D.155)
where
i
j

_
1, i = j,
0, i = j,
Kronecker delta. (D.156)
The reciprocal base vectors can be written in terms of the original base vectors:
e
1
=
e
2
e
3
e
1
(e
2
e
3
)
, e
2
=
e
3
e
1
e
1
(e
2
e
3
)
, e
3
=
e
1
e
2
e
1
(e
2
e
3
)
. (D.157)
Or, in general index notation
e
i
=
ijk
e
j
e
k
e
1
(e
2
e
3
)
, i, j, k = permutations of 1, 2, 3 (D.158)
in which
ijk
=
_
_
_
+1 when i, j, k is an even permutation of 1, 2, 3
1 when i, j, k is an odd permutation of 1, 2, 3
0 when any two indices are equal Levi-Civita symbol.
(D.159)
The reciprocal LeviCivita symbol
ijk
is the same, i.e.,
ijk
=
ijk
. These for-
mulas are also valid if the subscripts and subscripts are reversed. Thus, the
original base vectors could be the reciprocal base vectors e
i
and the recip-
rocal base vectors could be the original base vectors e
i
since both sets of base
vectors are linearly independent. Either set can be used as a basis for repre-
senting three-dimensional vectors.
The identity tensor can be written in terms of the base or reciprocal vectors
as follows:
I

i
e
i
e
i
= e
1
e
1
+e
2
e
2
+e
3
e
3

i
e
i
e
i
= e
1
e
1
+e
2
e
2
+e
3
e
3
.
identity tensor (D.160)
This denition can be used to write any vector or operator in terms of either
its base or reciprocal vector components:
A = A I = (A e
1
)e
1
+ (A e
2
)e
2
+ (A e
3
)e
3
=
i
A
i
e
i
, A
i
A e
i
,
= (A e
1
)e
1
+ (A e
2
)e
2
+ (A e
3
)e
3
=
j
A
j
e
j
, A
j
A e
j
,
(D.161)
I = e
1
(e
1
) +e
2
(e
2
) +e
3
(e
3
)
= e
1
(e
1
) +e
2
(e
2
) +e
3
(e
3
). (D.162)
DRAFT 11:26
The dot product between two vectors A and B is given in terms of their
base and reciprocal vector components by
A B =
i
A
i
B
i
=
i
A
i
B
i
=
ij
(e
i
e
j
)A
i
B
j
=
ij
(e
i
e
j
)A
i
B
j
. (D.163)
Similarly, the cross product between two vectors is given by
AB =
ij
A
i
B
j
e
i
e
j
=
ijk
A
i
B
j
e
k
(e
1
e
2
e
3
)
=
ij
A
i
B
j
e
i
e
j
=
ijk
ijk
A
i
B
j
e
k
(e
1
e
2
e
3
)
= (e
1
e
2
e
3
)
_
_
_
_
_
_
e
1
e
2
e
3
A
1
A
2
A
3
B
1
B
2
B
3
_
_
_
_
_
_
= (e
1
e
2
e
3
)
_
_
_
_
_
_
e
1
e
2
e
3
A
1
A
2
A
3
B
1
B
2
B
3
_
_
_
_
_
_
. (D.164)
The dot-cross product of three vectors is given by
A BC =
ijk
A
i
B
j
C
k
e
i
e
j
e
k
=
ijk
ijk
A
i
B
j
C
k
(e
1
e
2
e
3
)
=
ijk
A
i
B
j
C
k
e
i
e
j
e
k
=
ijk
ijk
A
i
B
j
C
k
(e
1
e
2
e
3
)
= (e
1
e
2
e
3
)
_
_
_
_
_
_
A
1
A
2
A
3
B
1
B
2
B
3
C
1
C
2
C
3
_
_
_
_
_
_
= (e
1
e
2
e
3
)
_
_
_
_
_
_
A
1
A
2
A
3
B
1
B
2
B
3
C
1
C
2
C
3
_
_
_
_
_
_
. (D.165)
For the simplest situation where the three base vectors e
1
, e
2
, e
3
are orthog-
onal (e
1
e
2
= e
2
e
3
= e
1
e
3
= 0), the reciprocal vectors point in the same
directions as the original base vectors. Thus, after normalizing the base and
reciprocal vectors they become equal:
e
1
= e
1
/|e
1
| = e
1
= e
1
/|e
1
| orthogonal
e
2
= e
2
/|e
2
| = e
2
= e
2
/|e
2
| unit
e
3
= e
3
/|e
3
| = e
3
= e
3
/|e
3
| vectors. (D.166)
The simplications of (??)(??) are given in (D.196)(D.201) in the section
(D.9) below on orthogonal coordinate systems.
D.8 Curvilinear Coordinate Systems
Consider transformation from the Cartesian coordinate system x = (x, y, z)
to a curvilinear coordinate system labeled by the three independent functions
u
1
, u
2
, u
3
:
x = x(u
1
, u
2
, u
3
) : x = x(u
1
, u
2
, u
3
), y = y(u
1
, u
2
, u
3
), z = z(u
1
, u
2
, u
3
).
(D.167)
DRAFT 11:26
The transformation is invertible if the partial derivatives x/u
i
for i = 1, 2, 3
are continuous and the Jacobian determinant (i.e., x/u
1
x/u
2
x/u
3
)
formed from these nine partial derivatives does not vanish in the domain of
interest. The inverse transformation is then given by
u
i
= u
i
(x) : u
1
= u
1
(x, y, z), u
2
= u
2
(x, y, z), u
3
= u
3
(x, y, z). (D.168)
In a curvilinear coordinate system there are three coordinate surfaces:
u
1
(x) = c
1
(u
2
, u
3
variable),
u
2
(x) = c
2
(u
1
, u
3
variable),
u
3
(x) = c
3
(u
1
, u
2
variable).
(D.169)
There are also three coordinate curves given by
u
2
(x) = c
2
, u
3
(x) = c
3
(u
1
variable),
u
3
(x) = c
3
, u
1
(x) = c
1
(u
2
variable),
u
1
(x) = c
1
, u
2
(x) = c
2
(u
3
variable).
(D.170)
The direction in which u
i
increases along a coordinate curve is taken to be the
positive direction for u
i
. If the curvilinear coordinate curves intersect at right
angles (i.e., u
i
u
j
= 0 except for i = j), then the system is orthogonal. The
familiar Cartesian, cylindrical and spherical coordinate systems are all orthogo-
nal. They are discussed at the end of the next section which covers orthogonal
coordinates.
A nonorthogonal curvilinear coordinate system can be constructed from an
invertible set of functions u
1
(x), u
2
(x), u
3
(x) as follows. A set of base vectors
e
i
can be dened by
e
i
= u
i
, i = 1, 2, 3 contravariant base vectors. (D.171)
These so-called contravariant (superscript index) base vectors point in the direc-
tion of the gradient of the curvilinear coordinates u
i
, and hence in the directions
perpendicular to the u
i
(x) = c
i
surfaces. The set of reciprocal base vectors e
i
is given by
e
i
=
ijk
e
j
e
k
e
1
e
2
e
3
=

ijk
J
1
u
j
u
k
, covariant base vectors, (D.172)
in which
J
1
u
1
u
2
u
3
= e
1
e
2
e
3
inverse Jacobian (D.173)
is the Jacobian of the inverse transformation from the u
i
curvilinear coordi-
nate system back to the original Cartesian coordinate system.
An alternative form for the reciprocal base vectors can be obtained from the
denition of the derivative of one of the curvilinear coordinates u
i
(x) in terms
of the gradient: du
i
= u
i
dx = u
i

j
(x/u
j
) dx
j
, which becomes an
DRAFT 11:26
identity if and only if u
i
(x/u
j
) =
i
j
. Since this last relation is the same
as the dening relation for reciprocal base vectors (e
i
e
j
=
i
j
), it follows that
e
i
=
x
u
i
, i = 1, 2, 3 covariant base vectors. (D.174)
The so-called covariant (subscript index) base vectors point in the direction
of the local tangent to the u
i
variable coordinate curve (from the x/u
i
def-
inition), i.e., parallel to the u
i
coordinate curve. Alternatively, the covariant
base vectors can be thought of as pointing in the direction of the cross product
of contravariant base vectors for the two coordinate surfaces other than the u
i
coordinate being considered (from the u
j
u
k
denition). That these two
directional denitions coincide follows from the properties of curvilinear sur-
faces and curves. The contravariant base vectors e
i
can also be dened as the
reciprocal base vectors of covariant base vectors e
i
:
e
i
=
ijk
e
j
e
k
e
1
e
2
e
2
=

ijk
J
x
u
j
x
u
k
; i, j, k = permutations of 1, 2, 3
contravariant base vectors
(D.175)
in which
J =
x
u
1

x
u
2
x
u
3
= e
1
e
2
e
3
Jacobian (D.176)
is the Jacobian of the transformation from the Cartesian coordinate system to
the curvilinear coordinate system specied by the functions u
i
.
The geometrical properties of a nonorthogonal curvilinear coordinate system
are characterized by the dot products of the base vectors:
g
ij
e
i
e
j
=
x
u
i

x
u
j
covariant metric elements,
g
ij
e
i
e
j
= u
i
u
j
contravariant metric elements.
(D.177)
These symmetric tensor metric elements can be used to write the covariant
components of a vector in terms of its contravariant components and vice versa:
A
i
A e
i
= A I e
i
=
j
(A e
j
)(e
j
e
i
) =
j
g
ij
A
j
A
i
A e
i
= A I e
i
=
j
(A e
j
)(e
j
e
i
) =
j
g
ij
A
i
.
(D.178)
Similarly, they can also be used to write the covariant base vectors in terms of
the contravariant base vectors and vice versa:
e
i
=
j
g
ij
e
j
, e
i
=
j
g
ij
e
i
. (D.179)
From the dot product between these relations and their respective reciprocal
base vectors it can be shown that
j
g
ij
g
jk
=
j
g
kj
g
ji
=
k
i
. (D.180)
DRAFT 11:26
The determinant of the matrix comprised of the metric coecients is called
g:
g g
ij
=
_
_
g
ij
_
_
1
, (D.181)
in which the second relation follows from interpreting the summation relations
at the end of the preceding paragraph in terms of matrix operations: [g
ij
][g
ik
]
= [I], which yields [g
ij
] = [g
jk
]
1
. Since the determinant of the inner product
of two matrices is given by the product of the determinants of the two matrices,
g = g
ij
=
_
_
_
_
x
u
i

x
u
j
_
_
_
_
=
_
_
_
_
x
u
i
_
_
_
_
_
_
_
_
x
u
j
_
_
_
_
=
_
x
u
1

x
u
2
x
u
3
_
2
= J
2
.
(D.182)
Thus, the determinant of the metric coecients is related to the Jacobian and
inverse Jacobian as follows:
J =
g = e
1
e
2
e
3
=
x
u
1

x
u
2
x
u
3
Jacobian,
J
1
= 1/
g = e
1
e
2
e
3
= u
1
u
2
u
3
inverse Jacobian.
(D.183)
The various partial derivatives in space can be worked out in terms of covari-
ant derivatives (/u
i
) using the properties of the covariant and contravariant
base vectors for a general, nonorthogonal curvilinear coordinate system as fol-
lows:
f =
i
u
i
f
u
i
=
i
e
i
f
u
i
gradient,
(D.184)
A = (A I) =
g (A e
i
)
e
i
g
=
i
e
i
g
(
gA
i
)
=
i
1
u
i
(
g A e
i
) =
i
1
J
u
i
(J A u
i
) divergence,
(D.185)
A = (A I) =
j
(A e
j
)e
j
=
j
A
j
u
j
=
ij
A
j
u
i
u
i
u
j
=
ijk
ijk
g
(A e
j
)
u
i
=
1
g
_
_
_
_
_
_
e
1
e
2
e
3
u
1
u
2
u
3
A
1
A
2
A
3
_
_
_
_
_
_
curl,
(D.186)
DRAFT 11:26
2
f f =
i
1
u
i
(
g e
i
j
e
j
f
u
j
)
=
ij
1
u
i
(
g g
ij
f
u
j
) =
ij
1
J
u
i
(J u
i
u
j
f
u
j
) Laplacian.
(D.187)
Dierential line, surface and volume elements can be written in terms of
dierentials of the coordinates u
i
of a general, nonorthogonal curvilinear coor-
dinate system as follows. Total vector dierential and line elements are:
dx =
i
x
u
i
dx
i
=
i
e
i
dx
i
|d|
dx dx =
_
ij
g
ij
du
i
du
j
metric of coordinates.
(D.188)
Dierential line elements d(i) along curve u
i
(du
j
= du
k
= 0) for i, j, k =
permutations of 1, 2, 3 are
d(i) = e
i
du
i
=

ijk
g
u
j
u
k
du
i
|d(i)| =
e
i
e
i
du
i
=
g
ii
du
i
(D.189)
The dierential surface element dS(i) in the u
i
= c
i
surface (du
i
= 0) for i, j, k
= permutations of 1, 2, 3 is
dS(i) d(j)d(k) =
g
ijk
u
i
du
j
du
k
|dS(i)| =
_
g
jj
g
kk
g
2
jk
du
j
du
k
=
_
g
ii
g du
j
du
k
(D.190)
The dierential volume element is
d
3
x d(1) d(2)d(3) = e
1
(e
2
e
3
) du
1
du
2
du
3
=
g du
1
du
2
du
3
.
(D.191)
D.9 Orthogonal Coordinate Systems
Consider transformation from the Cartesian coordinate system x = (x, y, z)
to an orthogonal curvilinear coordinate system dened by three independent
functions u
i
= u
i
(x, y, z) for i = 1, 2, 3. [Here, the superscripts 1,2,3 are not
powers; rather, they represent labels for the three functions. The functions are
labeled in this way to maintain consistency with the general (nonorthogonal)
curvilinear coordinate literature.] The coordinate surfaces are dened by u
i
=
c
i
, where c
i
are constants. The three orthogonal unit vectors that point in
directions locally perpendicular to the coordinate surfaces are
e
i
u
i
/|u
i
| orthogonal unit vectors. (D.192)
DRAFT 11:26
For the simplest orthogonal coordinate system, the Cartesian coordinate system,
e
1
= x = x, e
2
= y = y, e
3
= z = z.
Because of the normalization and assumed orthogonality of these unit vec-
tors,
e
i
e
j
=
ij

_
1, for i = j,
0, for i = j,
Kronecker delta. (D.193)
The cross products of unit vectors are governed by the right-hand rule which is
embodied in the mathematical relation
e
i
e
j
=
ijk
e
k
(D.194)
in which the Levi-Civita symbol
ijk
is dened by
ijk

_
_
_
+1, for i, j, k = 1, 2, 3 or 2, 3, 1 or 3, 1, 2 (even permutations)
1, for i, j, k = 2, 1, 3 or 1, 3, 2 or 3, 2, 1 (odd permutations)
0, for any two indices the same.
(D.195)
A vector A can be represented in terms of its components in the orthogonal
directions (parallel to u
i
) of the unit vectors e
i
:
A =
i
A
i
e
i
= A
1
e
1
+ A
2
e
2
+ A
3
e
3
, A
i
A e
i
(D.196)
For an orthogonal coordinate system the identity dyad or tensor is
I =
i
e
i
e
i
= e
1
e
1
+e
2
e
2
+e
3
e
3
identity tensor. (D.197)
Thus, the vector dierential operator becomes
= I =
i
e
i
(e
i
) = e
1
( e
1
) + e
2
( e
2
) + e
3
( e
3
)
=
i
u
i

u
i
= u
1

u
1
+u
2

u
2
+u
3

u
3
.
(D.198)
Here and below the sum over i is over the three components 1,2,3.
Using the relations for the dot and cross products of the unit vectors e
i
given
in (D.193) and (D.194) the dot, cross and dot-cross products of vectors become
A B =
i
A
i
B
i
= A
1
B
1
+ A
2
B
2
+ A
3
B
3
, (D.199)
AB =
ij
A
i
B
j
e
i
e
j
=
ijk
ijk
A
i
B
j
e
k
=
_
_
_
_
_
_
e
1
e
2
e
3
A
1
A
2
A
3
B
1
B
2
B
3
_
_
_
_
_
_
= e
1
(A
2
B
3
A
3
B
2
) +e
2
(A
3
B
1
A
1
B
2
) +e
3
(A
1
B
2
A
2
B
1
). (D.200)
A BC =
ijk
ijk
A
i
B
j
C
k
=
_
_
_
_
_
_
A
1
A
2
A
3
B
1
B
2
B
3
C
1
C
2
C
3
_
_
_
_
_
_
. (D.201)
DRAFT 11:26
The dierential line element in the i
th
direction is given by
d(i) = e
i
h
i
du
i
, with h
i
1/|u
i
|, dierential line element. (D.202)
Thus, the dierential surface vector for the u
i
= c
i
surface, which is dened by
dS(i) = d(j)d(k), becomes
dS(i) = e
i
h
j
h
k
du
j
du
k
, for i = j = k, dierential surface area. (D.203)
Since the dierential volume element is d
3
x = d(i) dS(i) = d(1) d(2)d(3)
and the Jacobian of the transformation is given by J = 1/(u
1
u
2
u
3
)
= h
1
h
2
h
3
,
d
3
x = h
1
h
2
h
3
du
1
du
2
du
3
, dierential volume. (D.204)
For orthogonal coordinate systems the various partial derivatives in space
are
f =
i
e
i
h
i
f
u
i
=
i
e
i
(e
i
) f, (D.205)
A =
i
1
J
u
i
_
J
h
i
A e
i
_
=
i
1
h
1
h
2
h
3
u
i
_
h
1
h
2
h
3
h
i
A e
i
_
, (D.206)
A =
ijk
ijk
h
k
e
k
J
u
i
(h
j
A e
j
) =
ijk
ijk
h
k
e
k
h
1
h
2
h
3
u
i
(h
j
A e
j
), (D.207)
2
f =
i
1
J
u
i
_
J
h
2
i
f
u
i
_
=
i
1
h
1
h
2
h
3
u
i
_
h
1
h
2
h
3
h
2
i
f
u
i
_
. (D.208)
The three most common orthogonal coordinate systems are the Cartesian,
cylindrical, and spherical coordinate systems. Their coordinate surfaces and
unit vectors are shown in Fig. D.3. They will be dened in this book by
Cartesian : u
i
= (x, y, z)
h
x
= 1, h
y
= 1, h
z
= 1 = J = 1; (D.209)
cylindrical : u
i
= (r, , z)
r
_
x
2
+ y
2
, arctan(y/x), z z,
x = r cos , y = r sin , z = z,
h
r
= 1, h
= r, h
z
= 1 = J = r; (D.210)
spherical : u
i
= (r, , )
r
_
x
2
+ y
2
+ z
2
, arctan(
_
x
2
+ y
2
/r), arctan(y/x),
x = r sin cos , y = r sin sin , z = r cos ,
h
r
= 1, h
= r, h
= r sin = J = r
2
sin . (D.211)
DRAFT 11:26
x
y
z
r
x=c
1
y=c
2
z=c
3
r=c
1
=c
2
z=c
3
r=c
1
=c
3
x
x
y
z
x
y
z
x
r
z
Cartesian cylindrical
spherical
=c
Figure D.3: Orthogonal unit vectors and constant coordinate surfaces for the
three most common orthogonal coordinate systems.
DRAFT 11:26
Note that with these denitions the cylindrical angle is the same as the az-
imuthal (longitudinal) spherical angle , but that the radial coordinate r is
dierent in the cylindrical and spherical coordinate systems. The spherical an-
gle is a latitude angle see Fig. D.3. Explicit forms for the various partial
derivatives in space, (D.205) (D.208), are given in Appendix Z.
REFERENCES
Intermediate level discussions of vector analysis are provided in
Greenberg, Advanced Engineering Mathematics, Chapters 13-16 (1998) [?]
Kusse and Westwig, Mathematical Physics (1998) [?]
Danielson, Vectors and Tensors in Engineering and Physics, 2nd Ed. (1997) [?]
More advanced treatments are available in
Arfken, Mathematical Methods for Physicists (??) [?]
Greenberg, Foundations of Applied Mathematics, Chapters 8,9 (1978) [?]
Morse and Feshbach, Methods of Theoretical Physics, Part I, Chapter 1 (1953)
[?]
DRAFT 11:26

App D

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

App D

Uploaded by

Copyright:

Available Formats

APPENDIX D.

A A is the magnitude or length of the vector A;

b B/B, unit vector in B direction (D.24)

vorticity of A lines (D.59)

= , stream function form.

b), curvature vector (D.141)

b|, curvature unit vector. (D.142)

b = 0, the curvature vector has no parallel component (

b|, geodesic curvature. (D.146)

bN), torsion vector. (D.147)

bN), torsion length. (D.148)

You might also like