See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/323401526

VECTOR AND TENSOR ANALYSIS Course by E. Kogan

Presentation · February 2018

CITATIONS READS

0 9,865

1 author:

Eugene Kogan
Bar Ilan University
185 PUBLICATIONS   877 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

RKKY in graphene View project

Double exhange View project

All content following this page was uploaded by Eugene Kogan on 10 February 2020.

The user has requested enhancement of the downloaded file.

 VECTOR AND TENSOR CALCULUS
Course by E. Kogan

Contents I. VECTOR ALGEBRA AND CALCULUS IN

CARTESIAN COORDINATES
I. VECTOR ALGEBRA AND CALCULUS
IN CARTESIAN COORDINATES 1 A. Geometric Approach
A. Geometric Approach 1
B. Scalars and Vectors as Tensors 1 Definition A.1 If AB and CD are directed line seg-
C. Tensor of Inertia 2 ments, they are said to define the same geometric vector
−−→ −−→
D. Vector Calculus in Cartesian Coordinates 3 - also designated as AB or CD - provided the segments
have the same length and direction. Alternatively we des-
II. LINEAR OPERATORS (INFORMAL ignate geometric vector by a single boldface letter.
INTRODUCTION INTO TENSORS) 4
A. Linear Operators 4 Definition A.2 If a and b are geometric vectors, to de-
B. Tensors in Electricity 4 termine the sum a + b: place the tail of a segment rep-
C. Transformation of Coordinates 5 resenting b on the head of a segment representing a, and
join the tail of the latter to the head of the former (or
III. FORMAL INTRODUCTION INTO use the parallelogram rule).
TENSORS 6
A. Definition of Tensor 6 Definition A.3 If a is a geometric vector, then ra is a
B. Isotropic Tensors 6 geometric vector that may be represented by a segment |r|
C. Algebra of Tensors 6 times as long and having the same or opposite direction
1. Addition of tensors 6 according as r > 0 or r < 0.
2. Contraction of tensors 7
3. Multiplication of tensors 7 Problem A.1 Represent two geometric vector α and β
4. The Quotient Law 7 by directed line segments drawn at random in the plane,
D. Laboratory System and Rotating System 7 and then construct a representation of rα+sβ, where (a)
E. Differentiation of Tensor Fields 8 r = 1/2, s = 2; (b) r = −2, s = 3.
F. Strain and Stress Tensors 8
Definition A.4 The scalar product of two vectors a and
IV. VECTOR CALCULUS IN b is
ORTHOGONAL CURVILINEAR
COORDINATES 9 a · b = |a||b| cos θ. (1)
A. Orthogonal Curvilinear Coordinates 9
B. Differential Operators 9 Definition A.5 The vector product of two vectors a and
1. Gradient 9 b is a vector in a direction perpendicular to both a and
2. Divergence 10 b with the length given by the equation
3. Curl 10
4. Laplacian 10 |a × b| = |a||b| sin θ. (2)

V. TENSORS IN CURVILINEAR
COORDINATES 11 B. Scalars and Vectors as Tensors
A. Contravariant and Covariant Vectors 11
B. Tensors in Curvilinear Coordinates 12 Definition B.1 A scalar we’ll call a tensor of zero or-
C. Metric Tensor 12 der.

VI. TENSOR CALCULUS IN CURVILINEAR After the Cartesian coordinate system is chosen, a ge-
COORDINATES 13 ometric vector a is a column of three numbers, called
A. Derivatives of Basis Vectors and Christoffel vector components (in a chosen system). From now on
Symbols 13 the vector is these three numbers (a1 , a2 , a3 ). (This def-
B. Covariant Differentiation 13 inition will be completed later.)
C. Vector Operators in Tensor Form 14
D. Affine Geodesics 14 Definition B.2 A vector we’ll call a tensor of the first
E. Metric Geodesics 15 order.
 2

Definition B.3 The operations with vectors are defined or in vector notation
in the following way
f = b(a · c) − c(a · b) (12)
c=a+b ⇐⇒ ci = ai + bi
c = ra ⇐⇒ ci = rai Also
α=a·b ⇐⇒ α = ai bi a·(b × c) = ijk ai bj ck = (a × b)·c (13)
c=a×b ⇐⇒ ci = ijk aj bk , (3)
Problem B.2 Prove that
where ijk is the Levi-Civita symbol
 δik ikm = 0,
 +1 if i, j, k is an even permutation of 1,2,3, iks mks = 2δim ,
ijk = −1 if i, j, k is an odd permutation of 1,2,3, .
 0 otherwise iks iks = 6.
(4)
C. Tensor of Inertia
Pay attention that the same symbol appears in the defi-
nition of the determinant of the matrix 3 × 3
  The connection between angular momentum and an-
 a1 a2 a3  gular velocity
 
 b1 b2 b3  = ijk ai bj ck , (5)
  X X
 c1 c2 c3  J= mr × (ω × r) = m[r2 ω − r(r · ω)], (14)
or more generally in tensor notation is
|A| lmn = ali amj anl ijk . (6) Ji = Iik ωk , (15)
Problem B.1 Prove Eqs. (5) and (6). where
Let us prove a very important relation X
Iik = m (xl xl δik − xi xk ) (16)
ijk klm = δil δjm − δim δjl , (7)
is the moment of inertia tensor (of the second order)
where δij is the Kronecker delta P 
2 2
P P
 P + z ) P− 2mxy 2
m(y − P mxz
1 if i = j, Iik =  − P mxy P + z ) P− 2myz 2
m(x .
δij = . (8)
0 otherwise − mxz − myz m(x + y )
(17)
As each of the three suffixes i, j, l, m can take three val- CM
The tensor of inertia for the center of mass Iik gives
ues, at least two of them should be equal. Because i 6= j
opportunity to calculate tensor of inertia for any other
there is only one value of m which makes ijk different 0
point Iik . Taking into account that x0i = xi − Xi , we
from zero, and hence the only one value of l and m that
obtain
makes klm different from zero is i and k taken in any
order. If the order is the same then both  are either 1
X
0
Iik = m [(xl − Xl )(xl − Xl )δik
or −1 and the product is 1. If the order is different the
CM
product is −1. Hence −(xi − Xi )(xk − Xk )] = Iik + M (Xl Xl δik − Xi X(18)
k ),

ijk klm = 0 (i = j) where M is the mass of the body and R is the radius-
= 0 (l = m) vector of the point relative to the center of mass.
= 1 (i = l, j = m) Problem C.1 Find the principal moments of inertia of
= −1 (i = m, j = l) a uniform solid cube of mass m and edge length 2a (a)
= 0 (i 6= l or m, or j 6= l or m). (9) with respect to the mid-point of an edge, and (b) with
respect to a vertex.
Comparing this with the rhs of Eq. (7) we complete the
proof. For the rotation about the fixed axis only the projec-
Consider tion of the Equation of torque to the axis of rotation is
relevant, ω is always directed along the axis and can be
f = a×(b × c). (10) presented as the derivative of the rotation angle. Thus
we obtain
Using relation (7) we get
d(J · n)
fi = bi aj cj − ci aj bj , (11) = G · n, (19)
dt
 3

where n is the unit vector determining the direction of Problem D.1 By explicitly writing out the terms in
the axis. We get Cartesian coordinates
(a) prove that for r 6= 0, where r is modulus of the radius
J · n = Ji ni = Iik ωk ni = Iik ωnk ni = I ϕ̇, (20)
vector r,
where the moment of inertia about the axis of rotation is
∇·r=3
I = ni nk Iik . (21)
∇×r=0
r
If we chose axis x3 along the rotation axis, Eq. (21)
∇· 3 = 0
becomes just
Z  rr 
∇× 3 = 0
I = I33 = ρ(x2 + y 2 )dV. (22)
 r
1 r
The Eq. (18) has the form ∇ =− 3
 CM  r r
I = ni nk Iik + M (Xl Xl δik − Xi Xk )
h i (b) prove that for any scalar ψ and any vector A
2
= I CM + M R2 − (n · R)
∇ × (∇ψ) = 0
= I CM + M (n × R)2 . (23) ∇·(∇ × A) = 0
Kinetic energy of the rotating body can be written in ∇×(∇ × A) = ∇(∇ · A) − ∇2 A;
the form
(c) show that for any vectors a, b and c
1X 2 1X
T = m (ω × r) = m ijk ωj xk iln ωl xn . (24)
2 2 [c·(b · ∇) − b·(c · ∇)]a = (∇ × a)·(b × c).
Using formula
(d) show, that
ijk iln = δjl δkn − δjn δkl , (25)
∇·(A × r) = r·(∇ × A).
we obtain
Problem D.2 Using coordinate notation prove that
ijk ωj xk iln ωl xn = (δkj xl xl − xj xk )ωj ωk . (26)
Hence ∇(φψ) = φ∇ψ + ψ∇φ;
1 ∇ × (φu) = φ∇ × u + (∇φ) × u;
T =Iik ωi ωk . (27) ∇·(u × v) = v·(∇ × u) − u·(∇ × v);
2
Notice that Eq. (25) in vector notation can be presented ∇×(u × v) = u(∇ · v) − v(∇ · u)
as +(v · ∇)u − (u · ∇)v;
2 2 1
(ω × r) = ω 2 r2 − (ω · r) , (28) 2
∇u = u×(∇ × u) + (u · ∇)u.
2
and proved in the framework of vector algebra.
Hint:
∂(φψ)
D. Vector Calculus in Cartesian Coordinates (∇(φψ))i = ;
∂xi
∂(φuk )
In vector calculus in Cartesian coordinates it is conve- (∇ × (φu))i = ijk ;
nient to introduce vector operator ∂xj
∂ ∂ ∂ ∂(uj vk )
∇ = e1 + e2 + e3 , (29) ∇·(u × v) = ijk ;
∂x1 ∂x2 ∂x3 ∂xi
∂(ul vm )
Applying this operator we obtain (∇×(u × v))i = ijk klm
∂xj
∂ψ
(grad ψ)i = (∇ψ)i = (30) ∂(ul vm )
∂xi = (δil δjm − δim δjl ) ;
∂xj
∂Ai
div A = ∇ · A = (31) ∂um
∂xi (u×(∇ × u)i = ijk klm uj
∂Ak ∂xl
(curl A)i = (∇ × A)i = ijk (32) ∂um
∂xj = (δil δjm − δim δjl )uj ;
∂xl
∂2ψ ∂2ψ ∂2ψ
∇2 ψ = + + (33) Problem D.3 Show that if a is a constant vector
∂x21 ∂x22 ∂x23
∂ 2 Ai
 
a×r (2 − n)a nr(a · r)
(∇2 A)i = (34) curl = + .
∂xj ∂xj rn rn rn+2
 4

II. LINEAR OPERATORS (INFORMAL Theorem A.1 If we apply successively two operators
INTRODUCTION INTO TENSORS) one represented by matrix T (1) and the other by matrix
T (2) the result is represented by the matrix T (1) T (2) . If
A. Linear Operators an operator, represented by a matrix T , has an inverse
operator, the latter is represented by a matrix T −1 .
Linear operator is a vector function of a vector argu-
ment Proof A.1 From

y = T̂ x, (35) (1)
T̂ (1) ei = Tji ej
(2)
which satisfies the following rules. For any vectors x1 , x2 T̂ (2) ej = Tkj ek (44)
and an arbitrary number λ
we get
T̂ (x1 + x2 ) = T̂ x1 + T̂ x2
(2) (1)
T̂ (λx1 ) = λT̂ x̂1 . (36) T̂ (2) T̂ (1) ei = Tkj Tji ek . (45)

Let us introduce in the vector space orthonormal basis

{ei }. Under the action of the operator a basis vector
turns into another vector, the latter can be expanded B. Tensors in Electricity
with respect to the basis
In electrostatics of dielectric media one has
T̂ ei = Tji ej . (37)
P = αE, (46)
Thus introduced matrix T , defines action of the operator
on any radius vector. In fact, vectors can be expanded
where P is the induced polarization, E is the electric
with respect to the basis
field and α is the polarizability. But this is true only in
x = xi ei . (38) isotropic medium. In anisotropic medium we have more
general relation
Thus
Pi = αij Ej . (47)
y = T̂ xi ei = xi Tji ej . (39)
The quantities [αij ] describe linear dependence between
In other words the components of a vector are trans- two vectors. One can show that αij = αji .
formed according to the law Ohm’s law is
yj = Tji xi ⇐⇒ Y = T X. (40) U
I= , (48)
We say that the matrix T introduced in Eq. (37) is the R
representation of the operator T̂ . For example, operator where the resistance R is
acting on plane vectors which stretches the plain in the
direction e1 by a factor of 2, and compresses it in the ρL σ −1 L
direction e2 by a factor of 3 is described by a matrix R= = ; (49)
S S
 
2 0 σ is the conductivity. Locally Ohm’s law can be pre-
T = . (41)
0 1/3 sented as
If an operator is represented by a matrix j = σE. (50)
 
2 1
T = , (42) But this is true only in isotropic medium. In anisotropic
1 2
medium we have more general relation
it means that
ji = σij Ej . (51)
T̂ e1 = 2e1 + e2
T̂ e2 = e1 + 2e2 . (43) The quantities [σij ] describe linear dependence between
two vectors.
Problem A.1 * (b) Find the matrix of transformation
which carries the vectors (1, 1) and (3, −2) into the vec- Definition B.1 (Informal) A (second order) tensor is a
tors (2, 1) and (1, 2) respectively. linear operator acting in a space of geometric vectors.
 5

C. Transformation of Coordinates Problem C.2 * (a) The linear operators acting in a

space of plane geometric vectors are described in an or-
The position of a point in space can be described by thonormal basis by the matrices
three Cartesian coordinates: x1 , x2 , x3 . The coordinates      
depend upon orientation of coordinate axes. If we chose 3 −1 0 1 −1 1
T1 = ; T2 = ; T3 = .
another orientation of axes, the same point is described −1 3 1 0 1 −1
by different set of coordinates x01 , x02 , x03 . The connection
is What is their geometrical meanings?

x0i = Sij xj , (52) After making the transformation of components of a

vector (56), Eq. (47) becomes
where S is an orthogonal matrix. Notice that because
of the orthogonality of S the last equation can also be Pi0 = Sik Sjl αkl Ej0 = αij
0
Ej0 . (60)
presented as
The physics of the problem becomes transparent, if we
xi = Sji x0j . (53) chose appropriate transformation S. For example, every
symmetric tensor can be diagonalized by an appropriate
The same coordinate axes determine the basis for the ex- transformation. (Notice that the law of transformation
pansion of any geometric vector. Change of coordinates (60) can be presented in matrix form as
(52) induces the transformation of basis
α0 = SαS −1 .) (61)
e0i = Sij ej . (54)
Let the electrical conductivity σ in some crystal is mea-
Writing down sured to have the components (the conductivity is mea-
sured in 107 ohm−1 m−1 ).
ai ei = a0i e0i , (55)  √ 
√1 2 0
we see that the components of the vector change accord- σ =  2 3 1 . (62)
ing to the specific law. 0 1 1
Definition C.1 A vector a is a set of quantities ai which
The eigenvalues of [σij ] are given by the equation
under change of coordinates realized by an orthogonal ma-
trix S are transformed according to the law  √ 
 1−σ 2 0 
 √
a0i = Sik ak ⇐⇒ ai = Sji a0j . (56)

 2 3−σ 1  = 0. (63)
 0 1 1−σ 
Consider some linear operator T̂
Hence the diagonal matrix σ 0 will have the form
y = T̂ x. (57)  
4 0 0
Eq. (57) can be presented in two different bases {ei } and σ0 =  0 1 0  . (64)
{e0i } respectively as 0 0 0

Y = TX We see that there is one direction in the crystal along

Y = T 0X 0,
0
(58) which no current can flow, and along one of the two per-
pendicular directions the current flows four times easier
where the matrices T and T 0 are the representations of than along the second. The axis in which the tensor is
the operator T̂ . If the components of a vector are con- diagonal we shall call the principal axes, and the eigen-
nected by Eq. (55), we get values of the tensor the principal values of the tensor.

T 0 = ST S −1 = ST S T . (59) Problem C.3 For the conductivity tensor (62) make

transformation determined by the matrix
Thus given a square matrix T , representing the linear op-  √ 
√1 3 1
√
erator T̂ in a given basis, we may also consider the matrix 2
 6 2 3 
T 0 = ST S −1 for any orthogonal matrix S as representing
q
S= 1 2 .
the same linear operator T̂ , but in a new basis.  − √3 0 3 
√1 − 12 1
2 2
Problem C.1 Can you make transformation of coordi-
nates which makes the meaning of the operator T̂ from Check up that S is an orthogonal matrix. Calculate ex-
Eq. (42) more transparent? plicitly SσS −1 . Compare the result with Eq. (64).
 6

Problem C.4 * Conductivity tensor of some crystal is B. Isotropic Tensors

a basis (e1 , e2 , e3 ) is measured (in some units) to be
  We can prove that introduced earlier symbols δij and
25 0 0√ ijk are tensors of the second and the third order respec-
σ= 0 7√ −3 3  . (65) tively. In fact,
0 −3 3 13
0
δkl = Ski Slj δij = Ski Sli = δkl (69)
Find the directions of the principal axes (e01 , e02 , e03 ) and
the principal values of the conductivity. If we substitute A = S in Eq. (6), we obtain

Similar to Eq. (51) we can introduce resistivity tensor lmn = Sli Smj Snk ijk . (70)
by the Equation
These are the examples of isotropic tensors, whose com-
Ei = ρij jj . (66) ponents are unaltered by rotation of the axes. It is ob-
vious, that δij is an isotropic tensor in the space of any
Problem C.5 If the conductivity tensor is given by Eq. dimensionality. The existence of ijk for d = 3 is a par-
(65), what is the resistivity tensor? ticular case of a general theorem.
Theorem B.1 In a space of the dimensionality d there
Problem C.6 ** The components of two vectors A and
exists an absolute antisymmetric isotropic tensor of order
B and a second-order tensor T are given in one coordi-
d
nate system by 
     √   +1 if i, . . . , p is an even permutation of 1,. . . ,d,
1 0 √2 3 0 ij...p = −1 if i, . . . , p is an odd permutation of 1,. . . ,d, .
A =  0 , B =  1 , T =  3 4 0 .  0 otherwise
0 0 0 0 2 (71)
For example,
In a second coordinate system, obtained from the first by
rotation, the components of A and B are
 
0 1
ij = . (72)
√    −1 0
3 −1
1 1
A0 =  0  , B 0 =  √0  . Proof B.1 If we recall that arbitrary orthogonal matrix
2 1 2 3 2 × 2 can be presented as
 
Find the components of T in this new coordinate system. cos θ sin θ
, (73)
− sin θ cos θ

III. FORMAL INTRODUCTION INTO then Equation

TENSORS      
cos θ sin θ 0 1 cos θ − sin θ 0 1
= (74)
− sin θ cos θ −1 0 sin θ cos θ −1 0
A. Definition of Tensor
proves the Theorem for d = 2.
Definition A.1 A second order Cartesian tensor is a set
of quantities Tij which under change of coordinates real- Hence, in two dimensions there are two isotropic tensors
ized by an orthogonal matrix S are transformed according of the second order: δij and ij
to the law

Tij0 = Sik Sjl Tkl . (67) C. Algebra of Tensors

A Cartesian tensor of general order is a set of quantities 1. Addition of tensors

Tij...k which under such rotation are transformed accord-
ing to the law Let Vi1 ...ir and Wi1 ...ir be two tensors of the same order
0
Tij...k = Sip Sjq . . . Sks Tpq...s (68) r. In every coordinate system we can calculate the sum
of appropriate components
Only the rules which can be written in tensor form can
Si1 ...ir = Vi1 ...ir + Wi1 ...ir . (75)
make sense. It means that both the LHS and the RHS of
the equation should be tensor of the same order. From We claim that Si1 ...ir are components of tensor. The
what we new previously we understand that scalars are same with the difference of two tensors
tensors of zero order and vectors are tensors of the first
order. Si1 ...ir = Vi1 ...ir − Wi1 ...ir . (76)
 7

2. Contraction of tensors D. Laboratory System and Rotating System

When we contract tensor Si1 i2 ...ir of order r with re- Consider two frames of reference: inertial one, which
spect to any two indices, say i1 and i2 , we obtain another we will call the Laboratory System (LS), and another one,
tensor of order n − 2 rotating relative to the first, which we will call the Ro-
tating System (RS). Let us introduce orthonormal basis
Vi3 ...ir = Si1 i1 i3 ...ir . (77) in each of these systems. Thus any vector X has com-
ponents (x1 , x2 , x3 ) in the first system, and components
(x01 , x02 , x03 ) in the second one, which we will present by
3. Multiplication of tensors column vectors X and X 0 respectively. The connection
between the the components can be presented in a matrix
Let Vi1 ...ir and Wk1 ...kp be two tensors of order r and form
p respectively (not necessarily the same ). In every co-
X = SX 0 , (83)
ordinate system we can calculate the (outer) product of
appropriate components where S is some orthogonal (time-dependent) matrix.
The matrix S is by itself a tensor. Differentiating Eq.
Si1 ...ir k1 ...kp = Vi1 ...ir Wk1 ...kp . (78) (83) we obtain
We claim that Si...ir k1 ...kp are components of tensor of the Ẋ = S Ẋ 0 + ṠX 0 , (84)
order equal to the sum of the orders. For example, an
outer product of two vectors ai and bi is a second order or equivalently
tensor ai bj . If we make a contraction of outer product S −1 Ẋ = Ẋ 0 + ΩX 0 , (85)
with respect to a pair (or several pairs) of indices, we
−1
obtain an inner (or contracted) product. where Ω = S S. Let us try to present Eq. (85) in vector
It is worth to look at the scalar and vector products form. The first two terms are obvious. The question is
of vectors from the tensorial point of view. From two what to do with the third term?
vectors ai and bi we can construct a second order tensor
ẊLS = ẊRS + . . . . (86)
cij = ai bj . (79) Let us introduce vector ω
Contracting this tensor we obtain a scalar 1
ωi = ijk Ωjk . (87)
2
α = cii = ai bi , (80) Differentiating Equation
which is called the scalar product of the vectors a and b. ST S = I (88)
Multiplying the third order tensor klm with the second
order tensor cij we obtain the fifth order tensor we obtain
 T
dijklm = ijk al bm . (81) S T Ṡ + Ṡ T S = 0 ⇐⇒ S T Ṡ = − S T Ṡ . (89)

Contracting twice thus obtained tensor we obtain the first Hence, the matrix Ω is antisymmetric ΩT = −Ω, and Eq.
order tensor (87) can be solved for Ω

ei = dijkjk = ijk aj bk . (82) Ωij = ikj ωk , (90)

that is
which is called the vector product of the vectors a and b.  
0 −ω3 ω2
Ω =  ω3 0 −ω1  . (91)
4. The Quotient Law −ω2 ω1 0
Thus introduced vector ω is called the angular velocity
Let in every coordinate system there exists an object of the RS relative to the LS.
Q(j1 . . . jq ). We can define outer and inner product of the
object with a tensor. For the object Q to be the tensor Problem D.1 The matrix of transformation from the
it is necessary and sufficient that for arbitrary tensor X RS to the LS is
of some definite order, either outer product QX or some 
cos ωt sin ωt 0

inner product contracted from QX be a tensor. S =  − sin ωt cos ωt 0  .
Thus looking at the definition of the polarizability (47) 0 0 1
or of the conductivity (51) we see that they are tensors
even without analyzing their transformation properties. Find vector ω.
 8

In vector form Eq. (85) can be presented as which can be contracted into a scalar

ẊLS = ẊRS + ω × X, (92) ∂2ψ

, (102)
∂xi ∂xi
where ẊLS is the rate of change of vector X in the LS,
that is the vector having components (ẋ1 , ẋ2 , ẋ3 ) in this called Laplacian of the scalar. Notice, that in two dimen-
system, and ẊRS is the rate of change of vector X in the sions from the second derivatives of ψ we can construct
RS, that is the vector having components (ẋ01 , ẋ02 , ẋ03 ) in an additional second order tensor
this system. As an important application of Eq. (92) we ∂2ψ 2 !
obtain the connection between the velocity of a particle ∂ ∂ ∂x22
− ∂x∂1 ∂x
ψ
ik jl ψ= 2
∂2ψ
2
. (103)
in the LS vLS and in the RS v ∂xk ∂xl − ∂ ψ ∂x1 ∂x2 ∂x1 ∂x1

vLS = v + ω × r. (93)
This tensor has applications in elasticity.
Differentiating Eq. (93) we obtain

v̇LS = v̇RS + 2ω × v + ω × [ω × r] + ω̇ × r. (94) F. Strain and Stress Tensors

Hence the second law of Newton mr̈LS = F in the RS is

The strain tensor eij is defined by the Equation
presented as
 
mr̈RS = F + mω 2 r⊥ + 2mv × ω + mr × ω̇, (95) 1 ∂ui ∂uj
eij = + , (104)
2 ∂xj ∂xi
where r⊥ = r − ω(ω · r)/ω 2 is the part of r perpendicular
to ω. The terms mω 2 r⊥ and 2mṙ × ω are interpreted as where u is the displacement vector. The element of the
fictitious centrifugal and Coriolis force respectively. stress tensor pij is defined as the xj component of the
stress vector acting across a plane whose normal lies in
the xi direction. It can be proved, that for a plane whose
E. Differentiation of Tensor Fields direction is given by vector n

Differentiating Cartesian tensor of the order n with Fi = pij nj . (105)

respect to coordinate we obtain tensor of the order n + 1
In a liquid, for example, pij = −pδij , so the stress vector
∂Ti1 ...ir is always perpendicular to the surface. A generalized
Ui1 ...ir j = . (96)
∂xj Hooke’s law is

Thus from the scalar field ψ we can construct a vector pij = cijkl ekl , (106)
field
∂ψ where cijkl is an elasticity tensor. For example, consider
, (97) Hooke’s law in isotropic medium.
∂xi
which is called gradient of the scalar. From the vector Theorem F.1 In a 3d space there are three independent
field ai we can construct a second order tensor field tensors of the forth order, namely

∂aj (1)
= bij . (98) cijkl = δij δkl
∂xi (2)
cijkl = δik δjl
From the last one we can construct the scalar field (3)
cijkl = δil δjk . (107)
∂ai
bii = , (99)
∂xi Proof F.1
which is called divergence of the vector, and the vector Hence, the most general forth-order isotropic tensor is
field
∂ak cijkl = λδij δkl + ηδik δjl + νδil δjk , (108)
ijk bjk = ijk , (100)
∂xj
and the Hooke’s law in isotropic medium can be written
which is called curl of the vector. From the second deriva- as
tive of a scalar ψ we can produce a second order tensor
pij = λekk δij + 2µeij , (109)
∂2ψ
, (101)
∂xi ∂xj where η + ν = 2µ. (λ and µ are called Lame constants.)
 9

IV. VECTOR CALCULUS IN ORTHOGONAL B. Differential Operators

CURVILINEAR COORDINATES
In each point of an orthogonal coordinate system there
A. Orthogonal Curvilinear Coordinates is introduced a new basis, the basis vector e1 being tan-
gent to the coordinate line
General curvilinear coordinates are defined by the
u2 = const, u3 = const, (122)
transformation
  and similar to other vectors. For the cylindrical polar
 u1 = u1 (x, y, z)  x = x(u1 , u2 , u3 )
coordinates
u2 = u2 (x, y, z) ⇐⇒ y = y(u1 , u2 , u3 ) . (110)
 u = u (x, y, z)  z = z(u , u , u ) ∂r
3 3 1 2 3 e1 ∼ = cos ϕi + sin ϕj
∂ρ
To represent the system graphically one can plot three ∂r
families of coordinate surfaces e2 ∼ = −ρ sin ϕi + ρ cos ϕj
∂ϕ
ui = const (111) ∂r
e3 ∼ = k. (123)
∂z
and three families of coordinate lines For the spherical polar coordinates

 u2 = const, u3 = const ∂r
u1 = const, u3 = const . (112) e1 ∼ = sin θ cos ϕi + sin θ sin ϕj + cos θk
 u = const, u = const ∂r
1 2 ∂r
e2 ∼ = r cos θ cos ϕi + r cos θ sin ϕj − r sin θk
The radius-vector ∂θ
∂r
e3 ∼ = −r sin θ sin ϕi + r sin θ cos ϕj.(124)
r = xi + yj + zk (113) ∂ϕ
becomes a function of u1 , u2 , u3 . In the cylindrical polar The basis vectors in an orthogonal system are orthog-
coordinates, for example onal to each other. After normalization we obtain the
orthonormal basis
r = ρ cos ϕi + ρ sin ϕj + zk. (114) 1 ∂r
ei = , (125)
In the spherical polar coordinates hi ∂ui
From the chain rule we obtain
r = r sin θ cos ϕi + r sin θ sin ϕj + r cos θk. (115) ei
∇ui = . (126)
Cylindrical and spherical coordinate systems are exam- hi
ples of the orthogonal coordinate system, that is the sys-
tem where all the coordinate lines and all the coordinate
1. Gradient
surfaces intersect at right angles.
It is important to define the scale factors
  The gradient of an arbitrary scalar function may be
 dr  written
hi =  . (116)
dui  ∂f
∇f = ∇ui . (127)
The infinitesimal distance is ∂ui
Hence in orthogonal coordinates
dl2 = h21 du21 + h22 du22 + h23 du23 . (117)
1 ∂f
(∇f )i = . (128)
Volume element hi ∂ui
dV = h1 h2 h3 du1 du2 du3 , (118) In Cartesian system
 
∂f ∂f ∂f
Element of surface perpendicular to the u1 -line ∇f = , , . (129)
∂x1 ∂x2 ∂x3
dS1 = h2 h3 du2 du3 (119) In cylindrical system
 
In cylindrical polar coordinates ∂f 1 ∂f ∂f
∇f = , , . (130)
∂ρ ρ ∂ϕ ∂z
h1 = 1, h2 = ρ, h3 = 1. (120)
In spherical system
In spherical polar coordinates  
∂f 1 ∂f 1 ∂f
∇f = , , . (131)
h1 = 1, h2 = r, h3 = r sin θ. (121) ∂r r ∂θ r sin θ ∂ϕ
 10

2. Divergence Consider, for example, the curl of the first term.

∂(h1 f1 )
To calculate the divergence it is convenient to present ∇ × (h1 f1 ∇u1 ) = ∇u1 × ∇u1
a vector F ∂u1
∂(h1 f1 ) ∂(h1 f1 )
+ ∇u2 × ∇u1 + ∇u3 × ∇u1
F = f1 e1 + f2 e2 + f3 e3 (132) ∂u2 ∂u3
1 ∂(h1 f1 ) 1 ∂(h1 f1 )
as =− e3 + e2 . (143)
h1 h2 ∂u2 h1 h3 ∂u3
F = (h2 h3 f1 )∇u2 × ∇u3 + (h3 h1 f2 )∇u3 × ∇u1 Hence
+(h1 h2 f3 )∇u1 × ∇u2 . (133)  
1 ∂
(∇ × F)i = ijk (hk fk ) . (144)
Taking into account that hj hk ∂uj

∇(Xa × b) = (∇X)a × b In Cartesian system

+X(∇ × a) · b − Xa · [∇ × b], (134) ∂fz ∂fy
(∇ × F)x = −
∂y ∂z
that for any vector c
∂fx ∂az
(∇ × F)y = −
∇(∇ × c) = 0, (135) ∂z ∂x
∂fy ∂fx
(∇ × F)z = − (145)
and relations of the type ∂x ∂y

∇u2 · [∇u2 × ∇u3 ] = 0 (136) In cylindrical system

1 ∂fz ∂fϕ
we obtain (∇ × F)ρ = −
 ρ ∂ϕ ∂z
1 ∂(h2 h3 f1 ) ∂fρ ∂fz
∇·F= (∇ × F)ϕ = −
h1 h2 h3 ∂u1 ∂z ∂ρ

∂(h3 h1 f2 ) ∂(h1 h2 f3 )
 
1 ∂ ∂fρ
+ + , (137) (∇ × F)z = (ρfϕ ) − . (146)
∂u2 ∂u3 ρ ∂z ∂ϕ

where we used the fact that In spherical system

1 1

∂ ∂fθ

∇u1 · [∇u2 × ∇u3 ] = , (138) (∇ × F)r = (sin θfϕ ) −
h1 h2 h3 r sin θ ∂θ ∂ϕ
 
In Cartesian system 1 1 ∂fr ∂
(∇ × F)θ = − (rfϕ )
  r sin θ ∂ϕ ∂r
∂fx ∂fy ∂fz 
1 ∂ ∂fr

∇·F= + + . (139) (∇ × F)ϕ = (rfθ ) − (147)
∂x ∂y ∂z r ∂r ∂θ
In cylindrical system
4. Laplacian
1 ∂ 1 ∂fϕ ∂fz
∇·F= (ρfρ ) + + . (140)
ρ ∂ρ ρ ∂ϕ ∂z
Laplacian of a scalar function f is
In spherical system
∇2 f ≡ ∇ · ∇f. (148)
1 ∂ 1 ∂ 1 ∂fϕ
r 2 fr +


∇·F= (sin θfθ ) + .
r2 ∂r r sin θ ∂θ r sin θ ∂ϕ 1

∂

h2 h3 ∂f

2
(141) ∇ ψ=
h1 h2 h3 ∂u1 h1 ∂u1
   
∂ h1 h3 ∂f ∂ h1 h2 ∂f
+ + (149)
3. Curl ∂u2 h2 ∂u2 ∂u3 h3 ∂u3

In Cartesian coordinates
To find the curl of F we present it as
∂2f ∂2f ∂2f
F = h1 f1 ∇u1 + h2 f2 ∇u2 + h3 f3 ∇u3 . (142) ∇2 f = + + (150)
∂x2 ∂y 2 ∂z 2
 11

In cylindrical coordinates The basis vector e1 is tangent to the coordinate line

1 ∂

∂f

1 ∂2f ∂2f u2 = const, u3 = const, (155)
2
∇ f= ρ + 2 + (151) i
ρ ∂ρ ∂ρ ρ ∂ϕ2 ∂z 2 and similar to other vectors. The basis vector e is normal
to the coordinate surface
In spherical coordinates
ui = const. (156)
 
2 1 ∂ 2 ∂f From the chain rule we obtain
∇ f= 2 r
r ∂r ∂r ei ej = δji . (157)
∂2f
 
1 ∂ ∂f 1
+ 2 sin θ + 2 2 (152) hence ei form a set of vectors reciprocal to the ei . We
r sin θ ∂θ ∂θ r sin θ ∂ϕ2 can expand any vector relative to each of the bases
Problem B.1 For each of the following orthogonal co- a = ai ei = ai ei , (158)
ordinate systems i
1. Rectangular coordinates x, y, z. where ai are covariant components, and a are contravari-
2. Cylinder coordinates r, θ, z. ant components.
3. Spherical coordinates r, θ, ϕ. In orthogonal systems the two bases are the same
4. Elliptic-cylinder coordinates η, ψ, z. (apart from the normalization).
Problem A.1 Find covariant and contravariant bases
x = a cosh η cos ψ in spherical and cylindrical coordinate systems.
y = a sinh η sin ψ Under an arbitrary transformation between the coor-
z=z dinate systems

5. Parabolic-cylinder coordinates η, ψ, z. u0i = u0i (u1 , u2 , u3 ) (159)

the bases vectors are transformed as
x = (µ2 − ν 2 )/2 ∂u0i ∂u0i ∂uj ∂u0i j
y = µν e0i ≡ = = e
∂r ∂uj ∂r ∂uj
z=z ∂r ∂r ∂uj ∂uj
e0i ≡ 0i
= 0i
= ej . (160)
∂u j
∂u ∂u ∂u0i
carry out the following:
(a) Sketch the coordinate surfaces. From the equation
(b) Sketch the coordinate curves. ∂u0i 0
a0i e0i = aj ej = aj e (161)
(c) Obtain hm , the scale factors. ∂uj i
(d) Obtain em , the reference unit vectors. we get that the contravariant components of a vector are
(e) Show that the system is indeed orthogonal. transformed as
(f ) Obtain the infinitesimal distance ds.
∂u0i j
(g) Obtain grad f . a0i = a (162)
(h) Obtain div F. ∂uj
(i) Obtain curl F. (163)
(j) Obtain ∇2 f . The differentials of coordinates are transformed as
(k) Obtain ∇2 F. ∂u0i j
du0i = du , (164)
∂uj
V. TENSORS IN CURVILINEAR hence they are also contravariant components of a vector.
COORDINATES From the equations
∂uj 0i
A. Contravariant and Covariant Vectors a0i e0i = aj ej = aj e (165)
∂u0i
we get that the covariant components of a vector are
Consider general curvilinear coordinate system defined transformed as
by Equation (110). In every point we can introduce two
∂uj
bases. The covariant basis vectors are defined as a0i = aj . (166)
∂u0i
∂r The term contravariant components means the law of
ei = . (153)
∂ui transformation (162), and the term covariant compo-
nents means the law of transformation (166). For Carte-
The contravariant basis vectors are defined as
sian coordinate systems the two sets of vectors ei and
∂ui ei are identical, and hence so are the components of a
ei = ≡ ∇ui . (154) vector with respect to either basis; ai is equivalent to ai .
∂r
 12

B. Tensors in Curvilinear Coordinates To find the volume element dV let us present Eq. (174)
in a matrix form
We define tensors with arbitrary numbers of covariant
G = ST G0 S, (175)
(subscript) and contravariant (superscript) indices by de-
manding where we have introduced matrix
0i 0j 0k d e f
0ij...k ∂u ∂u ∂u ∂u ∂u ∂u ab...c ∂u0i
Tlm...n = ... T (167) sij = . (176)
a
∂u ∂u b ∂uc ∂u0l ∂u0m ∂u0n de...f ∂uj
Let Tki11 ...ir
and Wki11 ...i
...ks be two tensors of the same
r Hence
 ...k s
r
order . They are equal if all the components are g = J 2 g0 . (177)
s
equal in one coordinate system where
Tki11 ...ir i1 ...ir
...ks = Wk1 ...ks . (168) g = det[gij ], (178)

Let Tki11 ...ir

and Wki11 ...i
...ks be two tensors of the same order
r and J is the Jacobian of the transformation
  ...ks
r  0i 
 ∂u 
. We define the sum of these tensors as J = det  j  . (179)
s
∂u
Uki11...i i1 ...ir i1 ...ir
...ks = Tk1 ...ks + Wk1 ...ks .
r
(169) In Cartesian coordinates g 0 = 1. Hence
 
Let Tki11 ...i r j ...j
and Wl11...lqp be two tensors of the arbitrary  ∂xi  √
...ks  det  j  = g. (180)

r
 
p ∂u
orders and . We define the product of these
s q On the other hand, from the Calculus we know that
 
r+p
tensors as the tensor of the order  
 ∂xi 
s+q
dV = det  j  du1 du2 du3 . (181)
∂u
i ...i j ...j j ...j
Uk11 ...krs l11 ...lpq = Tki11 ...i 1 p
...ks Wl1 ...lq .
r
(170)
Hence
√
 
r gdu1 du2 du3 .
Let Tki11 ...i
...ks is a tensor of order
r
. When we contract dV = (182)
s
it with respect to any two indices, one contravariant and The metric tensor defines scalar products of any two
one covariant,
 say i1 and k1 , we obtain another tensor of
 vectors
r−1
order a · b = gij ai bj . (183)
s−1

Uki22...i i1 ...ir Similarly, we can present the scalar product as

...ks = Ti1 ...ks . (171)
r

a · b = g ij ai bj , (184)
C. Metric Tensor where

In general curvilinear coordinates the infinitesimal in- ∂ui ∂uj

g ij = ei · ej = (185)
terval can be written as ∂xk ∂xk

∂xk ∂xk i j are the components of the contravariant metric tensor.

ds2 = dr·dr = du du = gij dui duj , (172) Similarly, we can present the scalar product as
∂ui ∂uj
where a · b = ai bi = ai bi . (186)

∂xk ∂xk From Equation

gij = ei · ej = (173)
∂ui ∂uj
g ij ai bj = aj bj (187)
are the components of the covariant metric tensor. The
fact that gij is the covariant tensor follows from the trans- we see that the metric tensor determines the connec-
formation law. In fact tion between covariant and contravariant components of
a vector
∂xk ∂xk ∂xk ∂u0p ∂xk ∂u0q ∂u0p ∂u0q 0
gij = = = g (174)
. ai = g ij aj . (188)
i
∂u ∂u j 0p i
∂u ∂u ∂u ∂u0q j ∂ui ∂uj pq
 13

Similarly from Equation This can be proved from the Equation

gij ai bj = aj bj (189) ∂ei ∂2r ∂ej

j
= j j
= . (198)
∂u ∂u ∂u ∂ui
we get
To obtain an expression for Γkij we differentiate Equa-
ai = gij aj . (190) tion
Comparing Eqs. (188) and (190) we see that the matrices
gij = ei · ej (199)
of the covariant metric tensor and contravariant metric
tensor are inverse and obtain
ij
g gjk = δki . (191) ∂gij
= Γlik glj + Γljk gil . (200)
∂uk
For the orthogonal systems the metric tensor is diago-
nal By cyclically permuting the free indexes we also get
 2 
h1 0 0 ∂gij
gij =  0 h22 0  . (192) = Γlji glk + Γlki gjl
∂uk
0 0 h23 ∂gki
= Γlkj gli + Γlij gkl . (201)
Problem C.1 Show that Cartesian, cylindrical and ∂uj
spherical systems are orthogonal and for the metric ten- Summing up we obtain
sor in these systems respectively we get gij =
∂gij ∂gki ∂gij

1 0 0
 
1 0 0
 
1 0 0
 + − = 2Γlij gkl , (202)
∂uk ∂uj ∂uk
 0 1 0  ,  0 ρ2 0  ,  0 r 2 0 .
or
0 0 1 0 0 1 0 0 r2 sin2 θ  
1 kl ∂gjl ∂gli ∂gij
Γkij = g i
+ j
− . (203)
2 ∂u ∂u ∂ul
VI. TENSOR CALCULUS IN CURVILINEAR
COORDINATES Problem A.1 Calculate the Christoffel symbols in
spherical and cylindrical coordinate systems. Check up,
A. Derivatives of Basis Vectors and Christoffel that the only three non-zero Christoffel in cylindrical po-
Symbols lar coordinates symbols are
1
We introduce the symbol Γkij to denote the coefficients Γ212 = Γ221 =
in the derivative of the basis vector ρ
Γ122 = −ρ
∂ei
= Γkij ek . (193)
∂uj
Using the reciprocity relation we obtain B. Covariant Differentiation

∂ei
Γkij = ek . (194)
∂uj
∂v i
   i 
∂v i k ∂v i k
By differentiating reciprocity relation (157) we obtain = + Γ kj v e i = + Γ kj v ei . (204)
∂uj ∂uj ∂uj
∂ei
= −Γikj ek . (195) The quantity in brackets is called a covariant derivative
∂uj It can be shown that covariant derivative of a vector is a
The symbol Γkij is called a Christoffel Symbol. The second order tensor.
Christoffel Symbol is not a tensor. In fact, in a new Similarly we can define a covariant derivative of a co-
coordinate system variant vector

0i
∂vi
∂u0k n ∂ − Γkij vk .
 l 
0k ∂e ∂u vi;j ≡ (205)
Γ0k
ij = e = e el (196) ∂uj
∂u0j ∂un ∂u0j ∂u0i
Problem B.1 Show, that in cylindrical polar coordi-
The Christoffel symbol is symmetrical with respect to nates
the interchange of its two subscripts
1 ∂ ∂v ϕ ∂v z
Γkij = Γkji . (197) v;ii = (ρv ρ ) + + .
ρ ∂ρ ∂ϕ ∂z
 14

Problem B.2 Show that where [blk ] = A−1 . Taking into account that the ma-
trix of covariant metric tensor is inverse to the matrix of
∂ √
Γiik = k
ln g, contravariant metric tensor we get
∂u
1 ∂ √ i

where g is the determinant of the matrix [gik ], which has div a = √ ga , (209)
g ∂ui
the covariant components of the metric tensor as its ele-
ments. where g is the determinant of the contravariant metric
tensor.
Problem B.3 (a) Find an expression for the second co-
Curl: In a general space we can define an antisymmet-
variant derivative vi;jk ≡ (vi ; j);k of a vector vi .
ric tensor
(b) We may define the components of Riemann tensor
by ∂vi ∂vj
(curl v)ij = vi;j − vj;i = j
− . (210)
l
∂u ∂ui
vi;jk − vi;kj ≡ Rijk vl .
In three dimensions with every antisymmetric tensor Ajk
Show that these components are given by we can connect the vector
1 1 ijk
l ∂Γlik ∂Γlij pi = √  Ajk . (211)
Rijk = − + Γm l m l
ik Γmj − Γij Γmk . 2 g
∂uj ∂uk
Hence we can define the vector of curl
(c) By first considering Cartesian coordinates, show that
l
all the components Rijk ≡ 0 for any coordinate system 1 1 1 ∂vk
in three-dimensional Euclidean space. (curl v)i = − √ ijk (∇ × v)ij = √ ijk j . (212)
2 g g ∂u
Problem B.4 We may define Christoffel symbols of the Laplacian of a scalar function:
first kind by
∇2 ψ = g ij ψ;ij . (213)
Γijk = gil Γljk .
Substituting Eq. (209) we obtain
(a) Show that these are given by  
2 1 ∂ √ jk ∂ψ
  ∇ ψ=√ gg . (214)
1 ∂gik ∂gjk ∂gij g ∂uj ∂uk
Γijk = + − .
2 ∂uj ∂ui ∂uk
For the case of orthogonal coordinates we get
(b) By permuting indices, verify that ∂ψ
(grad ψ)i =
∂gij ∂ui
= Γijk + Γjik . 1 ∂
∂uk h1 h2 h3 ai


div a = i
h1 h2 h3 ∂u
(c) Using the fact that Γljk = Γlkj , show that 1 ∂vk
i
(curl v) = ijk j
gij;k ≡ 0. h1 h2 h3 ∂u
 
1 ∂ h1 h2 h3 ∂ψ
∇2 ψ = . (215)
h1 h2 h3 ∂uj h2i ∂uk
C. Vector Operators in Tensor Form
Caveat: In vector analysis, in distinction to tensor
analysis, when considering orthogonal coordinates, it is
From the definitions of the vector operators in Carte-
accepted to normalize basis vectors. That is why the
sian system we can obtain formulas valid for any coordi-
there exists the following relation between the compo-
nate system
nents of vectors:
∂ψ 1
(grad ψ)i = ψ;i = (206) avec = ai = hi ai . (216)
∂ui i
hi
Divergence:

∂ai ∂ai 1 ∂glk D. Affine Geodesics

div a = ai;i = i
+ Γ i k
ki a = i
+ g kl i ai . (207)
∂u ∂u 2 ∂u
Consider a curve r = r(t). Derivative of a scalar φ
For any nonsingular matrix A = [aij ] we have along the curve is
∂|A| ∂akl dφ ∂φ duj
i
= blk |A|, (208) = (217)
∂u ∂ui dt ∂uj dt
 15

The vector multiplying by 2L we obtain

duj
ηj = (218) ∂L

d

∂L

dt 2L − = 0, (225)
∂xa du ∂ ẋa
is called the tangent vector.
Derivative of a vector v along the curve is
which can be rewritten as
dv ∂v duj i du
j
= j
= v;j ei . (219)
dt ∂u dt dt ∂L2 ∂L2 ∂L2 dL
 
d
− =2 a . (226)
Derivative of the tangent vector along the curve is du ∂ ẋa ∂xa ∂ ẋ du
j
 i  j
dη i du ∂η i k du
= η;j ei = + Γjk η ei Substituting for L2 , the lhs of Eq. (226) produces
dt dt ∂uj dt
 i j k
dη i k
  
∂η du i du i du
= + Γ e = + Γ ei ∂L2 ∂L2
   
jk i jk d d ∂ b c ∂
∂uj dt dt dt dt − = (gbc ẋ ẋ ) − (gbc ẋb ẋc )
 2 i k
 du ∂ ẋa ∂xa du ∂ ẋa ∂xa
d u i du
= + Γ ei . (220) d
dt2 jk
dt = (2gab ẋb ) − (∂a gbc )ẋb ẋc
du
Geodesic is a curve along which tangent vector does not = 2gab ẍb + 2∂c gab ẋb ẋc − partiala gbc ẋb ẋc
change. Hence, the equation of geodesic is
 
b b c 1
= 2gab ẍ + 2ẋ ẋ (∂c gba + ∂b gca − ∂a gbc )
2
d2 ui j
i du du
k

ds2
+ Γ jk
ds ds
= 0. (221) = 2gab ẍb + 2Γbca ẋb ẋc . (227)
The rhs of Eq. (226) produces

E. Metric Geodesics
∂L2 dL
 
∂ b c d ds
2 = 2 (gbc ẋ ẋ )
∂ ẋa du ∂ ẋa du du
If we introduce the Lagrangian
d2 s
 2 
d s ds
= 2(gbc ẋb ẋc )−1/2 gad ẋd 2 = 2 / gab ẋb .(228)
L(ẋa , xa , u) = [gab (x)ẋa ẋb ]1/2 , (222) du du2 du

where ẋa ≡ dxa /du, the length of the curve between the
points P1 and P2 , given by the functions xa (u) is Finally we obtain
Z P2
s= Ldu. (223) ẍa + Γabc ẋb ẋc = (s̈/ṡ)ẋa . (229)
P1

The metric geodesic is defined as the curve of minimal If we choose the parameter u = s we obtain
length. Hence the functions xa (u) satisfy Euler-Lagrange
equations
ẍa + Γabc ẋb ẋc = 0. (230)
 
∂L d ∂L
− = 0. (224)
∂xa du ∂ ẋa

View publication stats