Chapter 8

PHYSICAL LAWS must be independent of any particular coordinate systems used in describing them
mathematically, if they are to be valid. A study of the consequences of this re-L
quirement leads to tensor analysis, of great use in general relativity theory, differential geometry,
mechanics, elasticity, hydrodynamics, electromagnetic theory and numerous other fields of science
and engineering.

SPACES OF N DIMENSIONS. In three dimensional space a point is a set of three numbers, called
coordinates, determined by specifying a particular coordinate system
or frame of reference. For example (x,y, z), (p, c,z), (r, 8, 55) are coordinates of a point in rectan-
gular, cylindrical and spherical coordinate systems respectively. A point in N dimensional space is,
by analogy, a set of N numbers denoted by (x1, x2, ..., xN) where 1, 2, ..., N are taken not as expo-
nents but as superscripts, a policy which will prove useful.
The fact that we cannot visualize points in spaces of dimension higher than three has of course
nothing whatsoever to do with their existence.

COORDINATE TRANSFORMATIONS. Let (x1, x2, ..., xN) and (x1, x2, ..., RN) be coordinates of a point
in two different frames of reference. Suppose there exists N
independent relations between the coordinates of the two systems having the form

1 _ -X'1 2

2 2 1 2

(1)

xN = zN(x1, x2, ..., xN)

which we can indicate briefly by

(2) xk = xk(x1, x2, ..., xN) k = 1, 2, ..., N

where it is supposed that the functions involved are single-valued, continuous, and have continuous
derivatives. Then conversely to each set of coordinates (x1, x2, ..., xN) there will correspond a
unique set (x1, x2, ..., xN) given by

(3) Xk = x,x,...,x)
1 2 N k = 1, 2,...,N

The relations (2) or (3) define 4 transformation of coordinates from one frame of reference to another.

166
 TENSOR ANALYSIS 167

THE SUMMATION CONVENTION. In writing an expression such as a1x1 + a2x2 + ... + a1yx1 we can
X
use the short notation jZ1 xi. An even shorter notation is sim-
ply to write it as ajxi, where we adopt the convention that whenever an index (subscript or super-
script) is repeated in a given term we are to sum over that index from 1 to N unless otherwise spec-
ified. This is called the summation convention. Clearly, instead of using the index j we could have
used another letter, say p, and the sum could be written aoxO. Any index which is repeated in a giv-
en term, so that the summation convention applies, is called a dummy index or umbral index.
An index occurring only once in a given term is called a free index and can stand for any of the
numbers 1, 2, ..., N such as k in equation (2) or (3), each of which represents N equations.

CONTRAVARIANT AND COVARIANT VECTORS. If N quantities A1, A2, ..., AN in a coordinate sys-
tem (x1, x2, ..., x 1) are related to N other quantities
A1, A2, ..., ff in another coordinate system (x1, x2, ..., xN) by the transformation equations

A _ ax9 Aq p = 1, 2, ..., N
q=1

which by the conventions adopted can simply be written as

A =
axq
axP Aq

they are called components of a contravariant vector or contravariant tensor of the first rank or first
order. To provide motivation for this and later transformations, see Problems 33 and 34 of Chapter 7.
If N quantities A1i A2, ..., AN in a coordinate system (x1, 12 , ..., x1) are related to N other
quantities At, A2, ..., Aff in another coordinate system (x1, x2, ..., xN) by the transformation equations

Ap
= axp Aq p = 1, 2, ..., N
q=1
or

AP
axq A
azp q

they are called components of a covariant vector or covariant tensor of the first rank or first order.
Note that a superscript is used to indicate contravariant components whereas a subscript is
used to indicate covariant components; an exception occurs in the notation for coordinates.
Instead of speaking of a tensor whose components are Ap or AP we shall often refer simply to
the tensor AP or AP A. No confusion should arise from this.

CONTRAVARIANT, COVARIANT AND MIXED TENSORS. If N2 quantities Aqs in a coordinate system

_ (x 1,x2, ..., x1) are related to N2 other quan-
tities A in another coordinate system (x1, x2, ..., x") by the transformation equations

Air
ax9 axs Aqs p,r = 1, 2, ..., N
S=1 q=1 x x
or
168 TENSOR ANALYSIS

axq axrAgs
axq axs
by the adopted conventions, they are called contravariant components of a tensor of the second rank
or rank two.
The N2 quantities Aqs are called covariant components of a tensor of the second rank if
axq axs A
APr ax p -ay r qs

Similarly the N2 quantities AS are called components o f a mixed tensor of the second rank if

AP = ax P axs q
As
axq oxr

THE KRONECKER DELTA, written 8k, is defined by

J0 if jAk
Sk
1 if j = k
As its notation indicates, it is a mixed tensor of the second rank.

TENSORS OF RANK GREATER THAN TWO are easily defined. For example, Akit are the compo-
nents of a mixed tensor of rank 5, contravariant of order
3 and covariant of order 2, if they transform according to the relations
A firm
= axp axr azm axk ax 1. Agst

ti axq axs axt axi ax9 ki

SCALARS OR INVARIANTS. Suppose 0 is a function of the coordinates xk, and let denote the
functional value under a transformation to a new set of coordinates xk
Then cb is called a scalar or invariant with respect to the coordinate transformation if _ . A
scalar or invariant is also called a tensor of rank zero.

TENSOR FIELDS. If to each point of a region in N dimensional space there corresponds a definite
tensor, we say that a tensor field has been defined. This is a vector field or
a scalar field according as the tensor is of rank one or zero. It should be noted that a tensor or
tensor field is not just the set of its components in one special coordinate system but all the possi-
ble sets under any transformation of coordinates.

SYMMETRIC AND SKEW-SYMMETRIC TENSORS. A tensor is called symmetric with respect to two
contravariant or two covariant indices if its com-
ponents remain unaltered upon interchange of the indices. Thus if Aqs r = AQS r the tensor is sym-
metric in m and p. If a tensor is symmetric with respect to any two contravariant and any two co-
variant indices, it is called symmetric.
A tensor is called skew-symmetric with respect to two contravariant or two covariant indices
if its components change sign upon interchange of the indices. Thus if Aqs r= -Aqs r the tensor is
 TENSOR ANALYSIS 169

skew-symmetric in m and p. If a tensor is skew-symmetric with respect to any two contravariant and
any two covariant indices it is called skew-symmetric.

FUNDAMENTAL OPERATIONS WITH TENSORS.

1. Addition. The sum of two or more tensors of the same rank and type (i.e. same number of contra-
variant indices and same number of covariant indices) is also a tensor of the same rank and type.
Thus if AQ 0 and Bq 0 are tensors, then CO = Aq 0 + Br is also a tensor. Addition of tensors
is commutative and associative.
2. Subtraction. The difference of two tensors of the same rank and type is also a tensor of the same
rank and type. Thus if Aq0 and Br are tensors, then Dr = AqO - Bq 0 is also a tensor.
3. Outer Multiplication. The product of two tensors is a tensor whose rank is the sum of the ranks
of the given tensors. This product which involves ordinary multiplication of the components of
the tensor is called the outer product. For example, Aqr BS = Cqs' is the outer product of Alir
and BS . However, note that not every tensor can be written as a product of two tensors of lower
rank. For this reason division of tensors is not always possible.
4. Contraction. If one contravariant and one covariant index of a tensor are set equal, the result in-
dicates that a summation over the equal indices is to be taken according to the summation con-
vention. This resulting sum is a tensor of rank two less than that of the original tensor. The
process is called contraction. For example, in the tensor of rank 5, AgPr, set r=s to obtain
Agrr = Bq " a tensor of rank 3. Further, by setting p = q we obtain 80 = C2 a tensor of rank 1.
5. Inner Multiplication. By the process of outer multiplication of two tensors followed by a contrac-
tion, we obtain a new tensor called an inner product of the given tensors. The process is called
inner multiplication. For example, given the tensors A' O and Bst, the outer product is Aq1 Bst*
r
Letting q = r, we obtain the inner product Ark B . Letting q = r and p = s, another inner product
Ar1' Br is obtained. Inner and outer multiplication of tensors is commutative and associative.
6. Quotient Law. Suppose it is not known whether a quantity X is a tensor or not. If an inner prod-
uct of X with an arbitrary tensor is itself a tensor, then X is also a tensor. This is called the
quotient law.

MATRICES. A matrix of order m by n is an array of quantities apq, called elements, arranged in m

rows and n columns and generally denoted by

all a12 ... aln all a12 ... asn

a21 a22 ... a2 n an a22 a2n
or

a.41 ann a,ns ani2 ... ainn

or in abbreviated form by (a1,q) or [apq] p = 1, ..., m; q= 1, ..., n. If m=n the matrix is a square
matrix of order m by m or simply m; if m = 1 it is a row matrix or row vector; if n = 1 it is a column
matrix or column vector.
The diagonal of a square matrix containing the elements ass, ate, ..., ann is called the princi-
pal or main dia ogT 1. A square matrix whose elements are equal to one in the principal diagonal and
zero else h is called a unit matrix and is denoted by 1. A null matrix, denoted by 0, is a matrix
all of whose elements are zero.
170 TENSOR ANALYSIS

MATRIX ALGEBRA. If A = (apq) and B = (bpq) are matrices having the same order (m by n) then

1. A = B if and only if ap q = b1, q .

2. The sum S and difference D are the matrices defined by

S = A + B = (aj,q + bpq) , D = A - B = (apq- bpq)

3. The product P = AB is defined only when the number n of columns in A equals the number of rows
in B and is then given by
P = AB = (apq) (bpq) = (apr brq)
n
where al,r brq = apr brq by the summation convention. Matrices whose product is defined
r.1
are called conformable.
In general, multiplication of matrices is not commutative, i.e. AB A BA . However the asso-
ciative law for multiplication of matrices holds, i.e.- C) _ (AB)C provided the matrices are
conformable. Also the distributive laws hold, i.e. A(B+C) = AB + AC, (A +B) C = AC + BC.

4. The determinant of a square matrix A = (a,q) is denoted by I A I, det A, I I or det(ajq).

If P=AB then IPI = IAI B.
5. The inverse of a square matrix A is a matrix A-1 such that AA-1 = 1, where I is the unit matrix.
A necessary and sufficient condition that A-1 exist is that det A 0. If det A = 0, A is called
singular.

6. The product of a scalar ?. by a matrix A denoted by X A, is the matrix (Xa pq) where each
element of A is multiplied by X.

7. The transpose of a matrix A is a matrix AT which is formed from A by interchanging its rows and
columns. Thus if A = (apq), then AT = (aqp) . The transpose of A is also denoted by A.

THE LINE ELEMENT AND METRIC TENSOR. In rectangular coordinates (x,y,z) the differential
are length ds is obtained from
By transforming to general curvilinear coordinates (see Problem 17, Chapter 7) this becomes ds
3 3
E I goq dupduq. Such spaces are called three dimensional Euclidean spaces.
P=1 q=1

A generalization to N dimensional space with coordinates (x1, x2, ..., xN) is immediate. We de-
fine the line element ds in this space to be given by the quadratic form, called the metric form or
metric,

In the special case where there exists a transformation of coordinates from xI to xk such that
 TENSOR ANALYSIS 171

the metric form is transformed into (dz 1)2 + (d x2)2 + ... + (d xN)2 or d x kd x k, then the space is call-
ed N dimensional Euclidean space. In the general case, however, the space is called Riemannian.
The quantities gpq are the components of a covariant tensor of rank two called the metric
tensor or fundamental tensor. We can and always will choose this tensor to be symmetric (see Prob-
lem 29).

CONJUGATE OR RECIPROCAL TENSORS. Let g = g q denote the determinant with elements

g and supp e g A 0. Define g pq by-
pq

gpq
cofactor of gpq
g

Then gpq is a symmetric contravariant tensor of rank two called the conjugate or reciprocal tensor
of gpq (see Problem 34). It can be shown (Problem 33) that

sp
gpq grq r

ASSOCIATED TENSORS. Given a tensor, we can derive other tensors by raising or lowering indices.
For example, given the tensor A pq we obtain by raising the index p, the,
tensor A .q , the dot indicating the original position of the moved index. By raising the index q also
we obtain.4'?
A. Where no confusion can arise we shall often omit the dots; thus Apq can be written
Apq . These derived tensors can be obtained by forming inner products of the given tensor with the
metric tensor g pq or its conjugate gpq . Thus, for example

p rp Apq
A.q = g Arq, = grp gsq Ars A rs = grq A-ps
q%ntk _ gpk g grm
A A

These become clear if we interpret multiplication by grp as meaning: let r= p (or p=r) in whatever
follows and raise this index. Similarly we interpret multiplication by grq as meaning: let r= q (or
q = r) in whatever follows and lower this index.
All tensors obtained from a given tensor by forming inner products with the metric tensor and
its conjugate are called associated tensors of the given tensor. For example A'4 and A. are asso-
ciated tensors, the first are contravariant and the second covariant components. The relation be-
tween them is given by

AP = g pq Aq or AP = gpq Aq

For rectangular coordinates g pq = 1 if p = q , and 0 if pA q , so that Ap = Ap, which explains why

no distinction was made between contravariant and covariant components of a vector in earlier chap-
ters.

LENGTH OF A VECTOR, ANGLE BETWEEN VECTORS. The quantity APBP , which is the inner
product of AP and Bq , is a scalar anal-
ogous to the scalar product in rectangular coordinates. We define the length L of the vector AO or
AP as given by
172 TENSOR ANALYSIS

L2 = AP AP = g1gA1'Aq = g1gAPAq

We can define the angle 6 between AP and B1' as given by

A1B1
cos 6 =
(A1A1') (B1B1')

THE PHYSICAL COMPONENTS of a vector A1' or A1' , denoted by Au, AV , and A. are the projec-
tions of the vector on the tangents to the coordinate curves and are
given in the case of orthogonal coordinates by

Au = v Al =
Al
Av = 22 A2 = Aw = , s As = A
V _9_1 1 922 933

Similarly the physical components of a tensor

12 A12 13 A 13
A A g A
Auw = g1g A = etc.
V'9 11922 -11-933
V19

CHRISTOFFEL'S SYMBOLS. The symbols

are called the Christoffel symbols of the first and second kind respectively. Other symbols used in-
stead o and 1 q . The latter symbol suggests however a tensor character, which
is not trut-tf eneral.

TRANSFORMATION LAWS OF CHRISTOFFEL'S SYMBOLS. If we denote by a bar a symbol in a co-

ordinate system x k, then

[ jk m [pq, r] a0 axq axr + ax1' a2 xq

gpq
axk axk a:x' ax's axj axk
n s axn ax1 axq azn a2xq
1k pq axs axq axk axq ax3azk

are the laws of transformation of the Christoffel symbols showing that they are not tensors unless
the second terms on the right are zero.

GEODESICS. The distance s between two points t1 and t2 on a curve xr= x'^(t) in a Riemannian
space is given by
s =
J;1t2/;pq
2dx dt
t at
 TENSOR ANALYSIS 173

That curve in the space which makes the distance a minimum is called a geodesic of the space. By
use of the calculus of variations (see Problems 50 and 51) the geodesics are found from the differen-
tial equation
d2 xr + r dxp dxq = 0
ds2 pq ds ds

where s is the are length parameter. As examples, the geodesics on a plane are straight lines where-
as the geodesics on a sphere are arcs of great circles.

THE COVARIANT DERIVATIVE of a tensor Ap with repect to xq is denoted by Ap,q and is de-
fined by
_ aAp s
_ As
Ap,q axq pq
a covariant tensor of rank two.

The covariant derivative of a tensor Ap with respect to x9 is denoted by Apq and is defined by

Ap q - aAp 1P1),AS
axq qs

a mixed tensor of rank two.

For rectangular systems, the Christoffel symbols are zero and the covariant derivatives are the
usual partial derivatives. Covariant derivatives of tensors are also tensors (see Problem 52).

The above results can be extended to covariant derivatives of higher rank tensors. Thus

APi ... pn
r i ... n, q
_ ri-
aAPl

nC

axq
Api ... pX
s
1,4 Sr2...rn
- r2Sq
Ap. s rap rn ri ... rn_is
iq

sp. .Pln
+ {PiAr2r + P2 Aplsp3...Pin A pi .. pin - i s
n qs i ... rn + ... + Pin
qs 1... rn

pi... pin q
is the covariant derivative of Ari rn with respect to x

The rules of covariant differentiation for sums and products of tensors are the same as those
for ordinary differentiation. In performing the differentiations, the tensors g pq , gpq and 80 maybe
treated as constants since their covariant derivatives are zero (see Problem 54). Since covariant
derivatives express rates of change of physical quantities independent of any frames of reference,
they are of great importance in expressing physical laws.

PERMUTATION SYMBOLS AND TENSORS. Define a pqr by the relations

e123 =em1 =e312 =+1, e213=e132-= e321=-1, e pqr = 0 if two or more indices are equal
epgr
and define in the same manner. The symbols epgr and epgr are called permutation symbols in
174 TENSOR ANALYSIS

three dimensional space.

Further, let us define

It can be shown that E pqr and Epgr are covariant and contravariant tensors respectively, called
permutation tensors in three dimensional space. Generalizations to higher dimensions are possible.

TENSOR FORM OF GRADIENT, DIVERGENCE AND CURL.

1. Gradient. If <J) is a scalar or invariant the gradient of c is defined by

where <D, p is the covariant derivative of

grad cD
'
with respect to xp.
-
=
a CD
axp

2. Divergence. The divergence of AP is the contraction of its covariant derivative with respect to
xg, i.e. the contraction of A1,q. Then
divAp = Ap,p =
ak(gAk)
g

aAp aAg
3 Cur . 1 T he cur o f A p is
1 A p,q - A q,p - a tensor of rank two. The curl is also
axq axp
defined as -- Epgr Ap,q .

4. Laplacian. The Laplacian of is the divergence of grad cP or

v2 div 4>,p = 1 _ (V k
ax ax
In case g < 0, must be replaced by =g . Both cases g > 0 and g < 0 can be included by
writing g in place of V V.

8Ap
THE INTRINSIC OR ABSOLUTE DERIVATIVE of A p along a curve xq = xq(t), denoted by
St
, is
defined as the inner product of the covariant deriva-
q q
tive of AP and dt , i.e. AP, q dt and is given by

SAP dAp r dxg

Ar
st dt pq at
Similarly, we define
8Ap _ dAp p Ar dxq
bt dt qr dt

The vectors AP or Ap are said to move parallelly along a curve if their intrinsic derivatives
along the curve are zero, respectively.
Intrinsic derivatives of higher rank tensors are similarly defined.
 TENSOR ANALYSIS 175

Ap1... p'n
RELATIVE AND ABSOLUTE TENSORS. A tensor r1...rn is called a relative tensor of weight w
if its components transform according to the equation

Ag1...gl _ ax w Ap1.., pm azq i

r1... rn
azg1 ... pix
i ... axrn
s1... sn
ax
axp1
axs1 .sn
ax ax

where J = is the Jacobian of the transformation. If w=0 the tensor is called absolute and is
2z I
the type of tensor with which we have been dealing above. If w= 1 the relative tensor is called a
tensor density. The operations of addition, multiplication, etc., of relative tensors are similar to
those of absolute tensors. See for example Problem 64.

SOLVED PROBLEMS

SUMMATION CONVENTION.

1. Write each of the following using the summation convention.

(a) dW = 4 ax 1 +
'30 dx2 + +
a0
dxN . d% = a0 dxq
axe axN

dz k = a3F k dx1 + ax k dx2 + ax k dxN dz k _ ax k dxy'

(b) ... +
dt -ax1 dt ax2 dt ax' dt dt axrn dt

(c) (x1)2 + (x2)2 + (x3)2 + ... + (xN)2. xk xk

(d) ds 2 = g11(dx1)2 + g., (dx2)2 + gas (dx3)2 . ds2 = gkk dxk dxk , N=3

3 3
(e) }r g dxp dxq g pq dxp dxq, N= 3
p=1 q=1 p9

2. Write the terms in each of the following indicated sums.

N
(a) a , xk . }; a. xk = a x1 + a x2 + ... + ajN xN
jk b=, jk

Ap1A1r + Ap2A2r + +
...
ApNANr

Aqr. Apq Aqr =

(b) Apq q=1

(c) grs a xk , N=3.

dk
ax azs
176 TENSOR ANALYSIS

3 3
axj axk
_
grs j=1 k=1 gk axr axs

axj ax1 ax, ax2 ax, ax3

j=1 j1 ax azs + gj2 axr axs + gj3 axr axs

-ax' -ax' + '6X2 ax1 + _6X3 ax1

_ g11 a`r axs g21 axr azs g31 axr azs
ax1 ax2 ax2 ax2 ax3 ax2
+ g12 axr azS + g22
azr axS +
g32
axr az
ax1 ax3 _6X2 ax3 ax3 ax3
+ 913 -ax' axS + g23 azr azs + g33 a'xr azs

3. If xk, k = 1, 2, ..., N are rectangular coordinates, what locus if any, is represented by each of the
following equations for N=2,3 and Assume that the functions are single-valued, have con-
tinuous derivatives and are independent, when necessary.

(a) ax k = 1 , where ak are constants.

For N=2, a1x1 + a2x2 = 1, a line in two dimensions, i.' e. a line in a plane.
For N=3, a1x1 + a2x2 + a3x3 = 1 , a plane in 3 dimensions.
For N> 4, a1x1 +a2 X2 + ... + aNxN = 1 is a hyperplane.

(b) xkxk = 1.
For N = 2, (x1)2 + (x2)2 = 1 , a circle of unit radius in the plane.
For N = 3, (x1)2 + (x2)2 + (x3)2 = 1 , a sphere of unit radius.
For N >4, (x1)2 + (x2)2 + ... + (xN)2 = 1 , a hypersphere of unit radius.

(c) xk = xk(u) .
For N = 2, x1 = x1(u), x2 = x2(u), a plane curve with parameter u.
For N= 3, x1 = x1(u), x2 = x2(u), x3 = x3(u), a three dimensional space curve.
For N >4, an N dimensional space curve.

(d) xk = xk(u,v).
For N = 2, x1 = x1(u, v), x2 = x2(u, v) is a transformation of coordinates from (u, v) to (x1, x2) .
For N = 3, x1 = x1(u,v), x2 = x2(u,v), x3 = x3(u,v) is a 3 dimensional surface with parameters u and v.
For N>4, a hypersurface.

CONTRAVARIANT AND COVARIANT VECTORS AND TENSORS.

4. Write the law of transformation for the tensors (a) A k, (b) B k, (c) Cm

(a) A = axs axj ax k Ai

qr axi azq axr jk
As an aid for remembering the transformation, note that the relative positions of indices p, q, r on
the left side of the transformation are the same as those on the right side. Since these indices are as-
sociated with the z coordinates and since indices i, j, k are associated respectively with indices p, q, r
the required transformation is easily written.
 TENSOR ANALYSIS 177

pq _ axP axQ axi axj axk inn

(b) B st ax'n - n azr azs axt ijk

(c) Cp Cm
ax"n

5. A quantity A(j, k, 1, m) which is a function of coordinates xx transforms to another coordinate sys-

tem z2 according to the rule
A(p, q, r, s) axj azk axi a. s A(, k 1 m)
axp ax ax ax
(a) Is the quantity a tensor ? (b) If so, write the tensor in suitable notation and (c) give the con-
travariant and covariant order and rank.

(a) Yes. (b) Aj

1'. (c) Contravariant of order 3, covariant of order 1 and rank 3 +1 = 4.

6. Determine whether each of the following quantities is a tensor. If so, state whether it is contra-
s N
a0(axk 'x
variant or covariant and give its rank: (a) dxk, (b)

(a) Assume the transformation of coordinates ..., xN). Then dx1 = ax dxk and so dxk is a
z'1 = x1(x1,
axk
contravariant tensor of rank one or a contravariant vector. Note that the location of the index k is
appropriate.

(b) Under the transformation xk = xk(x1, ..., xT), 0 is a function of xk and hence 0 such that q(x1, ...,J) _
i.e. cP is a scalar or invariant (tensor of rank zero). By the chain rule for partial differ-
entiation,
a4 = a ax k = ak
-ax k k
transforms like A = ax A. Then
= is
axi azj axk ax7 azk axk axk J axi axk
a covariant tensor of rank one or a covariant vector.
Note that in
aO
the index appears in the denominator and thus acts like a subscript which indi-
axk
cates its covariant character. We refer to the tensor or equivalently, the tensor with components
a- , as the gradient of , written grad 0 or VO. axk
ak

7. A covariant tensor has components xy, 2y- z2, xz in rectangular coordinates. Find its covariant
components in spherical coordinates.

Let denote the covariant components in rectangular coordinates x1= x, xZ = y, x3 = z . Then

Al = xy = x1x2, A2 = 2y--z2 = 2x2-(x3)2, A3 = x1x3

where care must be taken to distinguish between superscripts and exponents.

Let Ak denote the covariant components in spherical coordinates x1=r,x2=6,x =0. Then

(1) Ak
178 TENSOR ANALYSIS

The transformation equations between coordinate systems are

x1 = 71 sin x2 Cos x3 , x2 = x1 sin x sin V, x3 = V- Cos -X-2

Then equations (1) yield the required covariant components

Al =
ax1 Al + ax2 A2 + ax3 A3
ax1 ax1 'ax'
= (sin 12 cos z3) (x1x2) + (sin x2 sin x3) (2x2 - (x3)2) + (cos z2) (xlx3)

(sin 6 cos (1b) (r2 sin2 6 sin 0 cos 0)

+ (sin 6 sin m) (2r sin 6 sin (P - r2 cos2 6)
+ (cos 6) (r2 sin 6 cos 6 cos ()

ax1 Al + x2 A2 + x3
A2 = As
ax ax2 az2

(r cos 8 cos 0) (r2 sine 6 sin cos (p)

+ (r cos 6 sin (p) (2r sin 6 sin m - r2 cos2 B)

+ (-r sin 6) (r2 sin 6 cos 6 cos (P)

A ax1
3
=
ax3 A 1 + aax.2
Vx3
A2 + ax3
ax3
A3

(-r sin 6 sin d) (r2 sin20 sin 0 cos gyp)

+ (r sin 6 cos q5) (2r sin 6 sin - r2 cos26)
+ (0) (r2 sin 6 cos 6 cos ()

aAp
8. Show that is not a tensor even though Ap is a covariant tensor of rank one.
axq

By hypothesis, A, -ax, Ap . Differentiating with respect to -k.

axq
aAj dxp aA p C xp
+ A
axk axq axk ax k axq

axp aA p axq a2 xp
+ A
axq axq axk axk DV '
XP
axp axq aAp a2
A
azj axk axq + ax k axk

aAp
Since the second term on the right is present, does not transform as a tensor should. Later we
axq
aAp
shall show how the addition of a suitable quantity to q causes the result to be a tensor (Problem 52).
ax
 TENSOR ANALYSIS 179

9. Show that the velocity of a fluid at any point is a contravariant tensor of rank one.

k
The velocity of a fluid at any point has components dx in the coordinate system xb. In the coor-
dt

dinate system x) the velocity is diJ . But

d'l axl dxk

dt axk dt
by the chain rule, and it follows that the velocity is a contravariant tensor of rank one or a contravariant
vector.

THE KRONECKER DELTA.

10. Evaluate (a) 8q ASS, (b) bq 8q.

Since 8q = 1 if p = q and 0 if p X q, we have

r
(a) 8q Asr = AS. (b)8g8- = 8r

11. Show thataaxq = 8q

x

If p = q , aq
axp = 1 since xp = xq .
x

axe
If p q , q = 0 since x P and xq are independent.
ax

The.. ax4 - 8q
ax

12. Prove that axfi ax = 8r .

azq axr
Coordinates xP are functions of coordinates xq which are in turn functions of coordinates xr. Then
by the chain rule and Problem 11,

axP axp az9

axr ax q az

13. If A = axq Aq prove that Aq = ax - At.

ax ax

Multiply equation A = axpAq by axr

axq azp
180 TENSOR ANALYSIS

Then
-ax-
A =
axr axk' A
q
= 8qrAq = A r by Prob. 12. Placing r = q the result follows. This
az j' az axq
indicates that in the transformation equations for the tensor components the quantities with bars and quan-
tities without bars can be interchanged, a result which can be proved in general.

14. Prove that 8Q is a mixed tensor of the second rank.

If 8q is a mixed tensor of the second rank it must transform accoraing to the rule

axq axq 6p
6k
aXP azk q

The right side equals ax' axk = k by Problem 12. Since 8 kj = 8k =1 if j = k , and 0 if j k , it fol-
ax ax
b
lows that 8q is a mixed tensor of rank two, justifying the notation used.

Note that we sometimes use 8pq = 1 if p = q and 0 if p q, as the Kronecker delta. This is how-
ever not a covariant tensor of the second rank as the notation would seem to indicate.

FUNDAMENTAL OPERATIONS WITH TENSORS.

15. If Apq and Brq are tensors, prove that their sum and difference are tensors.

By hypothesis Arq and Brq are tensors, so that

-3-i axk aPq

jr
Ar
AZ axp axq aTl

ax's azk axr fiq

Bl Br
axp axq azl

jk axk axk axr (APq + Br

Adding, (A + ilk)
l
=
r r
Z
axp axq all
jk azk azk axr (Ar
pq
Subtracting,
Jk
(AZ - Bl) - - Br5q)
axp axq ax" Z

Pq pq Pq Pq
Then 4Y+ Br and Ar - Br are
are tensors of the same rank and type as Ar and Br .

16. If Arq and Bt are tensors, prove that Crt = Arq Bt is also a tensor.
Ps
rts
We must prove that is a tensor whose components are formed by taking the products of compo-
 TENSOR ANALYSIS 181

nents of tensors APg and Bt . Since AIr and are tensors,

are

jk - axJ ark axr APq

l
axp axq arl

arm axt BS
axs am t

-1k Bm ark az k axr azm axt APq Bs

Multiplying, Al
n axp axq azl axS axn
r t

which shows that APgBt is a tensor of rank 5, with contravariant indices p,q,s and covariant indices
r,t , thus warranting the notation Cats. We call Cr r= A,, Bt the outer product of Arg and Bt .

17. Let Ars t be a tensor. (a) Choose p = t and show that Arqp, where the summation convention is
employed, is a tensor. What is its rank ? (b) Choose p = t and q = s and show similarly that Argp
is a tensor. What is its rank ?

(a) Since Arst is a tensor,

A jk axk axk axr axs axt APq

axk' axq ax l arm axn rst
(1) lmn

Pqp
We must show that A is a tensor. Place the corresponding indices j and n equal to each other
and sum over this index. Then
A' axJ ark axr axs at pq
lTnj
axp axg azl 'ax--M axg Arst
axt are axk axr axs vq
A rs t
axg ax 1' axg ax l axm
8t axk axr axs Abq
rst
0 axg art arm

ark axr axs Apq

axg ax l axm
rsp

and so Arse is a tensor of rank 3 and can be denoted by Bqs . The process of placing a contravariant
index equal to a covariant index in a tensor and summing is called contraction. By such a process a
tensor is formed whose rank is two less than the rank of the original tensor.

Pq
(b) We must show that Argp is a tensor. Placing j = n and k = in in equation (1) of part (a) and summing
over j and k , we have
182 TENSOR ANALYSIS

-jk
Alkj
_ axq azk axr axS axt 1'q

axp axq azl azk azk Arst

-ax t ax3 axS azk axr Apq

rst
azk axp azk axq azl

8
t
8
s arxA
Pq
1 q axl rst
axr pq
ax l Arqp

which shows that Arqp is a tensor of rank one and can be denoted by Cr. Note that by contracting
twice, the rank was reduced by 4.

18. Prove that the contraction of the tensor Aq is a scalar or invariant.

We have A = axq axq A p

k axp azk q

Putting j= k and summing, AJ = axq axq AP = S q Ap = Ap

p q p
axp axq q

Then A = Ap and it follows that Al must be an invariant. Since AP is a tensor of rank two and
q to define an invariant as a
contraction with respect to a single index lowers the rank by two, we are led
tensor of rank zero.

19. Show that the contraction of the outer product of the tensors A0 and Bq is an invariant.

Since Ap and Bq are tensors, T j= axq Ap, Bk = axq

Bq . Then
ax ax
- k
axq axq A B
q
axp axk

By contraction (putting j = k and summing)

A B. = azj axq Ap B = 8p q Ap Bq = Ap B
I q p
axp axq
and so A1B1 is an invariant. The process of multiplying tensors (outer multiplication) and then contract-
ing is called inner multiplication and the result is called an inner product. Since ApB p is a scalar, it is
often called the scalar product of the vectors AP and Bq.

p qs
20. Show that any inner product of the tensors Ar and Bt is a tensor of rank three.
Outer product of A p and Bts = A pBgs
t
 TENSOR ANALYSIS 183

Let us contract with respect to indices p and t, i.e. let p = t and sum. We must show that the result-
ing inner product, represented by 0Bgs, is a tensor of rank three.
s
By hypothesis, Ar and Bt are tensors; then

a`xj axr In ax I axn ax t Bqs

A = A Bn =
t
k -a axk axq axs axn

Multiplying, letting j = n and summing, we have

axk axr az l axn axt qs

axP axk axq axs axk Ar Bt
8t axr a31 axn AP
P axk axq axs r Btt
axr ail axn AP Bqs
azk axq axs r p

showing that A Bqs is a tensor of rank three. By contracting with respect to q and r or s and r in the
product A Bts, we can similarly show that any inner product is a tensor of rank three.

Another Method. The outer product of two tensors is a tensor whose rank is the sum of the ranks of
the given tensors. Then APB qs is a tensor of rank 3 + 2 = 5 . Since a contraction results in a tensor
whose rank is two less than that of the given tensor, it follows that any contraction of AP Bqs is a tensor
of rank 5-2=3.

21. If X(p, q, r) is a quantity such that X(p, q, r) Bqn = 0 for an arbitrary tensor Bqn, prove that
X(p, q, r) = 0 identically.
qn
Since Br is an arbitrary tensor, choose one particular component (say the one with q=: 2, r= 3) not
equal to zero, while all other components are zero. Then X(p,2,3) Ban = 0, so that X(p,2,3) = 0 since
Ban 0. By similar reasoning with all possible combinations of q and r, we have X(p,q,r) = 0 and the
result follows.

Brs
22. A quantity A (p, q, r) is such that in the coordinate system x2 A (p, q, r) Brs = C0 where is an
arbitrary tensor and C is a tensor. Prove that A(p,q,r) is a tensor.
kin
In the transformed coordinates x2, A(j, k, 1) B = Cn.
j
s
Then A(j,k,l) axk azn axr Bqs = -67 V, axp
Cl,
axq axs axl r axs axk

or axs axk ax Z A (j, k, 1) - axk A(p, q, r) Bqs - 0

-ax s ax azk
184 TENSOR ANALYSIS

Inner multiplication by ax, (i.e. multiplying by axn and then contracting with t =m) yields
ax ax t

n azk axr
bS .4(j,k,l) -ax' A(p,q,r) Brqs = 0
axq azl

or
axk axr A(j,k,l)
- - axk axA(p,q,r)
Br
qn
= 0.
axq axl
qn
Since Br is an arbitrary tensor, we have by Problem 21,

azk axr k, 1) - ax A(p,q,r) = 0

axq ax l axi

Inner multiplication by a9
x az n yields
ax in axr
k n - axp axq ax`n
b7n 61 A(j,k,l) - 1 A(p,q,r) = 0
az ax's axr

or 4(j,m,n) axp axq azn A(p, q, r)

axj azm axr

which shows that A(p,q,r) is a tensor and justifies use of the notation A q.

In this problem we have established a special case of the quotient law which states that if an inner
product of a quantity X with an arbitrary tensor B is a tensor C, then X is a tensor.

SYMMETRIC AND SKEW-SYMMETRIC TENSORS.

r
23. If a tensor .4Sq is symmetric (skew-symmetric) with respect to indices p and q in one coordinate
system, show that it remains symmetric (skew-symmetric) with respect to p and q in any coordi-
nate system.
Pq
Since only indices p and q are involved we shall prove the results for B
Pq Pq
If B is symmetric, B =B q .
Then

Bjk = -all azk Bpq = azk azk Bqp = kj

axP axq axq axp
and BPq remains symmetric in the z2 coordinate system.

If Bq is skew-symmetric,

B jk = axk axk B?'q =

- azk aacl Bqp _
axp axq axq axk
and BPq remains skew-symmetric in the Ti coordinate system.

The above results are, of course, valid for other symmetric (skew-symmetric) tensors.
 TENSOR ANALYSIS 185

24. Show that every tensor can be expressed as the sum of two tensors, one of which is symmetric
and the other skew-symmetric in a pair of covariant or contravariant indices.

Consider, for example, the tensor B. We have

BP9 (BP9 + B9P) + 1(BP4 - B4P

But Rpq = (Bpq+ Bq') = Rq' is symmetric, and SPq = (BPq- Bq') = -Sq' is skew-symmetric.
By similar reasoning the result is seen to be true for any tensor.

25. If = ajk Ai Ak show that we can always write 4) = b k Aj Ak where b k is symmetric.

(D =
ajk A Ak =
akJ
Ak A =
ak,.
A Ak

Then 2C =
ajk Aj Ak + ak.1
Ai Ak = (ajk + akj ) Aj Ak

and = z(ajk + aki-) AjAk =

bjk
A' A k

where b.k = 2(ajk+ ak') = bkj is symmetric.

MATRICES.

26. Write the sum S =A +B, difference D= A- B, and products P= AB, Q= BA of the matrices
3 1 -2 2 0 -1
A = 4 -2 3 , B= -4 1 2
-2 1 -1 1 -1 0

3+2 1+0 -2-1 5 1 -3

S=A+B= 4-4 -2+1 3+2 = 0 -1 5
-2+1 1-1 -1+0 -1 0 -1
3-2 1-0 -2+1 1 1 -1
D=A -B= 4+4 -2-1 3-2 = 8 -3 1

-2-1 1+1 -1-0 -3 2 -1

(3)(2) + (1)(-4) + (-2)(1) (3)(0) + (1)(1) + (-2)(-1) (3)(-1) + (1)(2) + (-2)(0)
P = AB = (4)(2) + (-2)(-4) + (3)(1) (4)(0) + (-2)(1) + (3)(-1) (4)(-1) + (-2)(2) + (3)(0)
(-2)(2) + (1)(-4) + (-1)(1) (--2)(0) + (1)(1) + (-1)(-1) (-2)(-1) + (1)(2) + (-1)(0)
0 3 -1
19 -5 -8
--9 2 4

8 1 -3
Q= BA = -12 -4 9
1 3 -5

This shows that AB BA, i.e. multiplication of matrices is not commutative in general.
186 TENSOR ANALYSIS

27. If A = 2
and B = (-I 2 show that (A+B)(A-B) A A2- B2.
-1 31 3 -2 ,

A+B = 1 3, A-B = 3 -1 . Then (A+B)(A-B) =

1 3 -2

1 4 5 (_: 5

A2- 2 1 2 1 3 5 B2 + -1 2 -1 2= 7 -6
-1 3 -1 3 y--5 8 ' 3 -2 3 -2 -9 10

11
Then A2-B2 = 1-4
4 -2
Therefore, (A+B)(A-B) A`-B2. However, (A+B)(A-B) = A2-AB+BA-B2.

28. Express in matrix notation the transformation equations for (a) a covariant vector, (b) a contra-
variant tensor of rank two, assuming N = 3 .

(a) The transformation equations AP = a9

x- Aq can be written
ax

axI ax2 ax3

ax1 ox ox

ax1 ax2 ax3

ax2 ax2 ax2

ax1 ax2 ax3

ax3 ax3 ax3

in terms of column vectors, or equivalently in terms of row vectors

ax1 -ax' ax1

ax1 ax2 ax3
ax2 ax2 ax2
(A1 A2 A3) _ (A1 A2 A3)
ay1 az2 -3x3

ax3 ax3 ax3

ax1 ax2 ax3

(b) The transformation equations A pr= ax ax axs

A qs can be written
ax q

-21 -22 23 ' _

A A A

31 A 32 A 33
A

Extensions of these results can be made for N > 3. For higher rank tensors, however, the matrix nota-
tion fails.
 TENSOR ANALYSIS 187

THE LINE ELEMENT AND METRIC TENSOR.

29. If ds2 = gj
dxk dxk is an invariant, show that 9jk is a symmetric covariant tensor of rank two.

By Problem 25, 4) = ds2, AJ = dx and Ak = dxk; it follows that 9,jk can be chosen symmetric. Also
since ds2 is an invariant,

g dx' dxq = g,k dxk dxk = gk axp dTP axe d- = g.k

axe axk d 0 dx`q
pq 9 J ax ax 9 ax ax

Then g`pq = gjk a axk and is a symmetric covariant tensor of rank two, called the metric tensor.
ax ax gjk

30. Determine the metric tensor in (a) cylindrical and (b) spherical coordinates.

(a) As in Problem 7, Chapter 7, ds2 = dp2+ p2d02+ dz2.

If x=p, x2=0,x3=z then g11=1'g22_p2
g33-I'g12Jg21-0'"23-g32^O'g31=g13=O.

1 0 0
911 g12 g13
In matrix form the metric tensor can be written 921 922 g23
0 p2 0

\g21 0 0 1
g32 933

(b) As in Problem 8 (a), Chapter 7, ds2 = dr2+ r2 d82+ r2 sin2 8 dc2.

f1 0 0
If x1= r, x2 = 8, x3 = the metric tensor can be written 0 r2 0
0 0 r2 sin2 8

In general for orthogonal coordinates, gk = 0 for it k .

j

g11 g12 g13

31. (a) Express the determinant g = g21 g22 g23
in terms of the elements in the second row and
g31 g32 g33

their corresponding cofactors. (b) Show that gjk G(j,k) = g where G(j,k) is the cofactor of
gk in g and where summation is over k only.
j
(a) The cofactor of g jk is the determinant obtained from g by (1) deleting the row and column in which
g k appears and (2) associating the sign (-1)j+k to this determinant. Thus,

Cofactor of g21 = (-1)2+1 g12 g12 , Cofactor of g = (--1)2+2 g11 g13 ,

g32 g33 g31 gd3

Cofactor of g _ (-1)2+3 911 912

g31 g32

Denote these cofactors by G(2,1), G(2,2) and G(2,3) respectively. Then by an elementary principle
of determinants
g21 G(2,1) + g22 G(2,2) + gm G(2,3) = g
188 TENSOR ANALYSIS

(b) By applying the result of (a) to any row or column, we have gjk G(j,k) = g where the summation is
over k only. These results hold where g = is an Nth order determinant.
I
I gjk

32. (a) Prove that g21 G(3,l) + g22 G(3,2) + g23 G(3,3) = 0.
(b) Prove that g3k G(p,k) = 0 if j A p .

g11 g12 g13

(a) Consider the determinant which is zero since its last two rows are identical. Ex-
g21 g22 g23
g21 g22 g23

panding according to elements of the last row we have

G(3,1) + g22 G(3,2) + g23 G(3,3) = 0
g21

(b) By setting the corresponding elements of any two rows (or columns) equal we can show, as in part (a),
that g,jk G(p,k) = 0 if j p. This result holds for Nth order determinants as well.

g7k = G(,k)
33. Define where G(j,k) is the cofactor of gjk in the determinant g = gj k A0.
Prove that g-k 9Ok = 8'.

By Problem 31, igk G(g k) = 1 or

jgk
gjk = 1, where summation is over k only.

k) gPk
By Problem 32, j G(g
gk =0 or =0 if p j .

Then g,k gPk(= 1 if p=j, and 0 if pJ j) = b.

We have used the notation gik although we have not yet shown that the notation is warranted, i.e.
that gjk is a contravariant tensor of rank two. This is established in Problem 34. Note that the cofactor
has been written G(j,k) and not Gjk since we can show that it is not a tensor in the usual sense. How-
ever, it can be shown to be a relative tensor of weight two which is contravariant, and with this extension
of the tensor concept the notation Gjk can be justified (see Supplementary Problem 152).

gjk
34. Prove that is a symmetric contravariant tensor of rank two.

Since g is symmetric, G(j,k) is symmetric and so gjk = G(j,k)/g is symmetric.

3.k

If B0 is an arbitrary contravariant vector, Bq =g q B0 is an arbitrary covariant vector. Multiplying

by gjq,

gJq Bq = gJ q g B' = 8i BO = Bi or gJq Bq = B,9

q 0

Since Bq is an arbitrary vector, gIq is a contravariant tensor of rank two, by application of the quotient
law. The tensor gjk is called the conjugate metric tensor.
 TENSOR ANALYSIS 189

35. Determine the conjugate metric tensor in ( ylindrical and (b) spherical coordinates.

1 0 0
(a) From Problem 30 (a), 0 p2 0 p2
g
0 0 1

911
cofactor of g11 1 p2 0
g p2 0 1

cofactor of g22 1 12
_
922 g p2
P
cofactor of g 1 1 0
9W = g = 0 p2
2

cofactor of g12 1 0 0
= 0
g12 = g p2 0 1

3k
Similarly g =0 if j k. In matrix form the conjugate metric tensor can be represented by
1 0 0
0 1/p2 0
0 0 1

(b) From Problem 30 (b), g r4 sin2 &

r2

As in part (a), we find 1 and gjk = 0 for j k, and in matrix form

this can be written r2 sin26

1 0 0

0 1/r2 0
0 0 1/r2 sin2 6

36. Find (a) g and (b) gjk corresponding to ds2 = 5(dx1)2 + 3(dx2)2 + 4(dx3)2 - 6 dx1 dx2 + 4 dx2 dx3 .

5 -3 0
(a) g11=5, g= -3 3 2 = 4.
0. 2 4

(b) The cofactors G(j,k) of gJk are

G(1,1)=8, G(2,2)=20, G(3,3)=6, G(1,2)=G(2,1)=12, G(2,3)=G(3,2)=-10, G(1,3)=G(3,1)=-6

Then g11=2, g22=5, g33=3/2, g12=g2.=3, e=g22=-5/2, g13=g31=-3/2

Note that the product of the matrices (gJ.k) and (gjk) is the unit matrix I, i.e.

5 -3 0 2 3 -3/2 1 0 0
-3 3 2 3 5 -5/2 = 0 1 0

0 2 4 -3/2 -5/2 3/2 0 0 1

190 TENSOR ANALYSIS

ASSOCIATED TENSORS.

37. If A = gjk
Ak, show that Ak = gik A .

Multiply Aj = gJ ,k A k by g" q

Then gJq Aj = gjggjk A"-= 8q Ak = Aq, i.e. Aq = gJq A or Ak = gjk Aj .

The tensors of rank one, Aj and Ak, are called associated. They represent the covariant and contra-
variant components of a vector.

38. (a) Show that L2 = gpq AP Aq is an invariant. (b) Show that L2 = gpg AP Aq .

(a) Let Aj and Ak be the covariant and contravariant components of a vector. Then

Ap = aXA1, q axgAk
axk

and Ap A
p = axk azp A. A k
= k A Ak = A. Al
ax'p axk

so that Aj AI is an invariant which we call L2. Then we can write

L2 = Aj A'l = gjk Ak AI gpq A0 Aq

(b) From (a), L2 = A. AI = AI gkj Ak = gjk Aj Ak = g,q Ap Aq.

The scalar or invariant quantity L = P AP is called the magnitude or length of the vector with
A

covariant components Ap and contravariant components Ap.

39. (a) If AP and Bq are vectors, show that gpq AP Bq is an invariant.

Ap B4
(b) Show that 9pg is an invariant.
(Ap A p) (BgBq )

(a) By Problem 38, A 0 B = AP gpq Bg = gpq AIB' is an invariant.

0
gpgAABq
(b) Since ApAP and BgBq are invariants (ApAp) (BgBq) is an invariant and so is an

invariant. (A'Ap) (Bq Bq)

We define gpgAPBg
cos 6
(APA p) (Bg Bq)

as the cosine of the angle between vectors Ap and Bq. If gpq A1B' = A1B1 = 0, the vectors are
called orthogonal.
 TENSOR ANALYSIS 191

40. Express the relationship between the associated tensors:

Agkr
(c) A. Ajq.sl
(a) Ajkl and A0gr , (b) and q.. t and k

Ajk l gjPgkg glr Apqr jkl

(a) = or A pqr =
g. gkq g1r A
Agkr = g jg91rA'kl
A'kl or
(b) gjq glr Agkr
P.rs.
(c) A .q..t g
j rk
g g tl
...sl
or Ajgk
...sl
= gp j 9rk 9
ti A 5.rs.
. q.. t,
Ajgk

41. Prove that the angles 812, 023 and 631 between the coordinate curves in a three dimensional co-
ordinate system are given by
cos 6 12 = g12
V"g11 g22
cos 6 =
g
923
22 33
cos 6 1
=
--
931
g3

Along the x1 coordinate curve, x2 = constant and x3 = constant.

Then from the metric form, ds2 = g11(dxi2 or dx1 = 1
ds
791,
Thus a unit tangent vector along the x1 curve is Al = 61. Similarly, unit tangent vectors along

the x2 and x3 coordinate curves are Ar2-- ;7182 and Ar3-

= "r br3
33

The cosine of the angle 612 between Al and A2 is given by

p
cos 612 = gpq Al A2q 1 1
81
q
2
g12
= pq VI'g-11 VI-g;
g11 g22

Similarly we obtain the other results.

42. Prove that for an orthogonal coordinate system, g12 = g23 = g31 = 0.
This follows at once from Problem 41 by placing 61 =6 2= 01 = 9 6 0 . From the fact that 91q = gq j
it also follows that g21 = g32 = g13 = D.

43. Prove that for an orthogonal coordinate system, g11 = g11 ' 922 922 ' 933 g33

From Problem 33, gPr9 bq.

rq

If p = q= 1, g1r 9r1 = 1 or 11 + g12 g21 1.

9 911 + 913 931
1
Then using Problem 42, 911 =
g11.

g1
Similarly if p = q = 2, g = ; and if p = q = 3 , 938 = 13
192 TENSOR ANALYSIS

CHRISTOFFEL'S SYMBOLS.

44. Prove (a) [pq,r] _ [qp,r],

,
(c) [pq,r] = grs {;q}
(b) Pq = 4P

aggr
(a) [pq,r] _ i(agPr + _ iaggr + .._P - aggp) _ [qP,r] .
axq axp axr axp axq axr

(b) pq
= g" [Pq,r] = gsr[ qp,r] =
fS
qP

(c) = gksgsr [pq,r] = 8k [pq,r] = [Pq,k]

gks Pq
s
or [Pq, k] = gks i.e. [pq,r] grs
pq Pq

Note that multiplying [ pq, r] by gsr has the effect of replacing r by s, raising this index and re-
placing square brackets by braces to yield {;} . Similarly, multiplying
s
by grs or gsr has the
Pq Fq
effect of replacing s by r, lowering this index and replacing braces by square brackets to yield [pq,r] .

agpq
45. Prove (a) = [Pm, q ] + [qm, P ]
ax
agt Pn q
(c)
P = a lnV g
(b) = g ma g qn p
mn
axm

(a) [ p m ,q ] + [ q m ,p ] = 2
ag + agmg _ agPK) + Z (ag +agm _ aggro) = agfig
axm axp axq axm axq axp 'ax"'

k
(b) axm (g J k gi j) (bi) = 0. Then
axm

pk ag J ag jk agij
+ agaxik gi.. 0 or g
jk
ax gi axm axm

g ir g2J.. ag' k it gjk agtij

axm $ axm
-1 k
8r 19-
j ax = -9tir gik ([irn,j] + [jm,i] )
agrk _ g it k jk r
im g jm
axm

and the result follows on replacing r, k, i, j by p, q, n, n respectively.

(c) From Problem 31, g =gjk G(j, k) (sum over k only).

 TENSOR ANALYSIS 193

Since G(j,k) does not contain gjk explicitly, a 9j -r

= G(j,r). Then, summing over j and r,

ag. ag.
ag
ax's
= ag r=
agxm
G(j, r) jr
axln
jr
Jag.
= 99 ar jr [rm,j])
azj = 99

Thus

i ag
jm
2 g axq

The result follows on replacing j by p and m by q.

46. Derive transformation laws for the Christoffel symbols of (a) the first kind, (b) the second kind.

axp -6.q
(a) Since
'2
gjk azk axk gpq

(1)
agjk = axp axq agpq axr + axj' a2xq g + a2xj' axq
gp9
axq axj a`xk axr axj axk axq axk pq axq azj axk

By cyclic permutation of indices j, k, m and p, q, r

agkln = axq axr agq r axp + axq a2 xr

axr
(2)
axq axk axm axq azj axk axq axq qr
a2xq
axk azk ax1a qr '
(3) - X1
,ax -r axk agr ji axq + axr a2 xP a2xr ax75
g
ax k a'm azj axq axk axq' ark axj azk -3x14 a'xj rr

Subtracting (1) from the sum of (2) and (3) and multiplying by 2, we obtain on using the definition
of the Christoffel symbols of the first kind,
a2 XP
ax axq axr axq
(4) gpq
'3V axk -a-% axj ax k azj

ax n az
(b) Multiply (4) by -nm = to obtain
axs ax t
Pnm [ _ ax, axq axr ax-n azj' st a2xq axq axn azj st
axj axk axm axs axt g [pq,r] + azj ax'k azm axs axt
g gpq
194 TENSOR ANALYSIS

jn axp axq azn r 72x0 azn

Then
jk } azn axk axs t
g
st
[Pq.r]
-61i axk axs
bqgSt
gpg

axp axq azn j s + 72x1' azn

axk axk axs pq axk azk axp

s s
since s gst
[pq,r] gsr [pq,r] 1'q and b g st gpq = g s ggpq = b .
P

72x2
n ax's axp axQ m
47. Prove -
axq axk jk axn ` axk axk pq

n axp axq axn s a2xp axn

From Problem 46 (b),
jk } =
axk axk axs Pq
+
-ay j axk axp .

Multiplying by
axn ax =
axp axk SS
s + a2xp
ax 1 ax dxj ax Pq axk azk
k
a2
_ axp axq ` m + x
azn axk Pq ax ax k
2

Solving for ax , the result follows.

axq axk

48. Evaluate the Christoffel symbols of (a) the first kind, (b) the second kind, for spaces where
g1'9 = 0 if p,- q.

(a) If P=q=r. 1 `ego + agpp agpp

[ Pq r]
2 axp axp axp

If P=qtr. 1 (agpr + agpr _ Ngpp) 1 agpp

[pp,r] 2 axr
2 axp axp ax'_

[ pq. p ] = 1 ('9pp + aggp _ agpg _ 1 agpp

If P=rXq, [Pq,r]
2 axq axp axp 2 axq

If p, q, r are distinct, [ pq, r] = 0.

We have not used the summation convention here.

(b) By Problem 43, gii = g (not summed). Then

1J

= gsr [ pq, r] = 0 if r# s , and = gss [ pq, s ] _ [ s ] (not summed) if r= s .

pq gq,
ss
By (a) :
 TENSOR ANALYSIS

s _ P = [PP' P ] _ 1 agpp In g
Pq = PP axp pp
gpp - 2gpp axp

s s [ PP s ] 1 agpp
Pq PP gss 2gss axs

P
[pq,p] 1 agpp a In g
= =
Pq
S Pq gpp 2gpp axq 2 axg pp

If p, q, s are distinct,
js = 0.
Pq

49. Determine the Christoffel symbols of the second kind in (a) rectangular, (b) cylindrical, and
(c) spherical coordinates.

We can use the results of Problem 48, since for orthogonal coordinates g pq = 0 if p q .

(a) In rectangular coordinates, gpp = 1 so that = 0.

i Pq

(b) In cylindrical coordinates, x1 = p, x2 = 0, x3 = z, we have by Problem 30 (a), g11 = 1, g22 = p2, g33=1.
The only non-zero Christoffel symbols of the second kind can occur where p = 2. These are

22
1 -a a

ax,
g22
-2 (p2)
2g11 p

2 2 g22 1 a 2) - 1
(p
21 12 282, axl 2)02 ap p

(C) In spherical coordinates, x1= r, x2 = 6, x3 = 0, we have by Prob. 30 (b), g11= 1, g22= r2, g,= r2 sin26.
The only non-zero Christoffel symbols of the second kind can occur where p = 2 or 3. These are

22
2g11
1 g22
ax1
= - 2 a (r2) _ -r
2
2 _ 2 _ 1 _ 1 a (r2) = 1

21 12 2g22 ax1 2r2 ar r

1 a
(r2 sin2 6) r sine 6
2 dr

=
20) - sin 6 cos 0
2822 axe 2r2 8 (r2 sin
3 3 __ 1 ag - 1 a (r2 sin2 6) _
1
31 13 2833 axI 2r2 sin2 6 ar r

3 13 1 ag33 _ 1 a (r2 sin26) cot 6

=
32 23 2833 axe 2r2 sin2 6 a6
196 TENSOR ANALYSIS

GEODESICS.

t
50. Prove that a necessary condition that I = ft 2 F(t, x, z) dt be an extremum (maximum or min-
imum) is that aF _ d U
1

ax dt (ax)
Let the curve which makes I an extremum be x = X(t), t1< t< t2. Then x = X(t) + E7](t), where E is
independent of t, is a neighboring curve through t1 and t2 so that 77(t1) = 7)(t2) = 0. The value of I for the
neighboring curve is

F(t, X+E77, X+E7]) dt

1(E)
= ft 1 t2
This is an extremum for E= 0. A necessary condition that this be so is that dl I = 0. But by differ-
entiation under the integral sign, assuming this valid, a=o

dE IE=1ft (e? +
- 77) dt 0

which can be written as

Jt2
1 ax
71dt
+ az
t2
t1
f 1
t2,77 d
dt (ax)
dt =
ft 1
t2
77
aF
ax
d aF
dt (ax )
dt 0

Since 77 is arbitrary, the integrand aF _ d (-F) = 0.

ax dt ax

t2
The result is easily extended to the integral F(t, x1, z1, X 2 , 1 2 , ..., xX ac') dt
J t1

and yields
aF d (aF) _
0
axk dt azk

called Euler's or Lagrange's equations. (See also Problem 73.)

dd
2r r ) dxp d q =
51. Show that the geodesics in a Riemannian space are given by 0
s2 + pq ds ds

rt2 .1p
We must determine the extremum of g x dt using Euler's equations (Problem 50) with
t1 pq
F= g pq zp zq . We have

aF = 1 (g zp zq)-1/2 agpq zp zq
axk 2 pq axk

2 (gpq xp xq)-1/2 2gPk

az
Using = g pg zp zq , Euler's equations can be written
dt
 TENSOR ANALYSIS 197

.p
g x
d -39pq
1
-axk
xp zq = 0
dt S 2s

ag pq p = gpk z s
or zp + agpk x p zq xq
gpk s
axq 2 axk

agpk
Writing xp xq = 1 ( a gpk + a ggk) zp xq this equation becomes
axq 2 axq axp

g kx.p s..
gpk x + [ pq, k ] zp x q =
S

If we use are length as parameter, 1, I S'= 0 and the equation becomes

ad_ x-0 + [pq,k] dxp dxq = 0

g
pk ds2 ds ds

Multiplying by grk , we obtain

2 r
d x + r dxp dxq
ds2 Pq ds ds

THE COVARIANT DERIVATIVE .

aA
52. If Ap and AP are tensors show that (a) axq As
s
-p
and (b) Apq = a + As are tensors.
4S

x
(a) Since A . = Ar
a-xI
aA axr aAr axt a xr
2

(1) _ +
axk -ax! axt -ax,
azk axk Ar

From Problem 47,

a2xr n axr _ axz axl r

axj azk jk
azn azj axk (i t
Substituting in (1),
A. aAr
axr axt n ar
x
Ar
axi ax l r
A.
}
ax j
jk
axk azj azk at axn ax

axp axq aAp n axp axq s

An As
axp azk axq 1k w axk azk Pq
or

__
aAj n
An
axp axq aAp
axk 1k ax azk axg
198 TENSOR ANALYSIS

and
axq
- s
pq
A
s is a covariant tensor of second rank, called the covariant derivative of A with
respect to xq and written

axJ

(b) Since Aj = Ar ,
ax r

(2)
aA J
=
azj aAr axt + a 71 axt Ar
axk axr axt oxk axr oxt axk

From Problem 47, interchanging x and x coordinates,

n azj antioxl j
rt axn axr axt il

Substituting in (2),
J
aA ax`s axt aAr + n axk axt Ar axe oxl oxt
axk
_

axr azk axt rt axn axk axr axt axk it

1
} Ar

axk axt aAr + n axj axt Ar ' a 7

r

axr axk axt rt axn ax k axr k tit

aid axq aAP

axp axk axq
+ {p}i, sq axk axk
s

or

aAk i = az ax (aA + p As
{-T}
ki A
ax axp ax axq qs

aAq + p s
and A is a mixed tensor of second rank, called the covariant derivative of A" with
s

respect to xq and written A,q .

53. Write the covariant derivative with respect to xq of each of the following tensors:
(a)Ajk, (b)AJk, (c)Ak, (d)Akl,

-6.41k
s S
(a) Ajk,q Ask - Ajs
axq 19 kq

aAk
jk q sk + k JS
(b) A
axq
/
qs
A { qs A

(c) j - aAkj - s
A

k, q axq kq } Asj
IIqs
As
k

_ aAkl
(d) Aj
kl,q
_ {:q}4i s

lq
Ai
ks
+ 1j
qs }A
S
kl
axq
 TENSOR ANALYSIS 199

jkl
jkl aAmn s Ajkl _ s Ajkl + j Askl +j k Ajsl +
I
Ajks
(e) Amn,q axq mq sn nq ms qs mn qs sn qs mn

54. Prove that the covariant derivatives of (a) (b) gjk, (c) 8k are zero.

agjk s s
(a)
g1k,q axq j q J gsk - kq is
age
9 - [jq,k] - [kq,j] = 0 by Problem 45(a).
axq

(b)
gjk agjk + {/}gsk +
k
g
is
= 0 by Problem 45(b).
q axq qs qs

J abk s
_
k{q}6s
j + 0
(c) 8k,q ax q

55. Find the covariant derivative of Ak Bum with respect to xq.

I Is
I In a (Ak Bn) _ s Aj Bln m s
kq s
_ nq Ajk
Blm

s
(Ak Bn )q ax q

+ As Blm + l AjBsm + rn Aj Bls

qs
kn qs kn qs kn

I kqs
Aj +
s
j ,4s
k
Blnm
qs

aBlm
n _ s Blms + l Bsm + m Bls
+ Ak
nq qs n qs n
axq

B l in + AI B lin
= A k,q n k n,q

This illustrates the fact that the covariant derivatives of a product of tensors obey rules like those
of ordinary derivatives of products in elementary calculus.

_
56. Prove
km
(gjk An ) ,q gjk nq
Akin

km km km
(jk An m ) ,q g k,q `4 + g k An ,q k An ,q
since gjk = 0 by Prob. 54(a). In covariant differentiation, gjk , gsk and bj can be treated as constants.
q
200 TENSOR ANALYSIS

GRADIENT, DIVERGENCE AND CURL IN TENSOR FORM.

57. Prove that div AP = 1 a (/ A").

axk

The divergence of AP is the contraction of the covariant derivative of AP, i.e. the contraction of
A0,q or A0, p. Then, using Problem 45(c) ,

' 3 A
div A0 AP = A
axk pk

=
aAk
axk +
a In ) Ak
axk
=
aAk
ax k
+
(,
1
-)A
-a vg-

a xk
k

=
1

, ax k
a
( Ak }

V2<p
= 1 a
58. Prove that V"g gkr
axk ax

The gradient of (D is grad _ V4) =axr a covariant tensor of rank one (see Problem 6(b)) de-

r is A
a
fined as the covariant derivative of (1), written (D,, r. The contravariant tensor of rank one associated with
k = gkr
axr
. Then from Problem 57,

ak( -gkr
div (gkr ) V'g
ax ax ax

aAP aAq
59. Prove that App q - Aq,p

LA, _ LA
Ap, q Aq. j, =
axq
s

pq
A
S
(_{:}AS) axq r axp
This tensor of rank two is defined to be the curl of AP.

60. Express the divergence of a vector AP in terms of its physical components for (a) cylindrical,
(b) spherical coordinates.

(a), For cylindrical coordinates x1= P. x2 = 0, x3 = z ,

1 0 0

g 0 p2 0 p2 and ' = p (see Problem 30(a))

0 0 1

The physical components, denoted by A. Ao, A. are given by

Ap = Vg_ Al = Al . A = g- A2 = pA2, Az = V A3 = A3
 TENSOR ANALYSIS 201

Then
div A0 _ (4 Ak)
ax k

(b) For spherical coordinates x1=r, x2=8, x3=0,

1 0 0

g 0 r2 0 r4 sin2 8 and Irg = r2 sin 8 (see Problem 30(b))

0 0 r2 sin2 8

The physical components, denoted by Ar, A8, A4) are given by

Ar= Ae = V -g- A2 = rA2 , g33 A3 = r sin 8 A3

uA1=AI

, AO =
22

Then
div A =
1

-vlg
-6

axk
(v Ak)
(r2 sin 8 Ar) + (r sin 8 AB) + (r A,)]
r2 sin e -6r -66

1 a (r2 Ar) + 1 a (sin 8 A) + 1

e
r2 ar r sine a8 r sine

61. Express the Laplacian of 4>, V2c , in (a) cylindrical coordinates, (b) spherical coordinates.

(a) In cylindrical coordinates g11=1, g22 =1/P2, g =1 (see Problem 35 (a)). Then from Problem 58,

v245
Vg--
axk
(,rg- gkr
a
l aa
P C aP cP aP) + ao c P
a 1a aa cP az ) ]
az

1a a(f) 1D2 A)
P aP cP aP' + p2 22 + az2

(b) In spherical coordinates g11=1, g22=1/r2, g = 1/r2 sin28 (see Problem 35 (b)). Then

v%) _
7g axk''gkra
1 a 2
(r sin 8
M)
ar) + ae (Sin
8
a
ae) + V (sin 4 ]
a 1 a4)
r2 sin 8 -6r

t a a
r2 ar (2 r'
r + 1

r2 sine ae
a
(sin B
a8
) + 1

2 sin2 8
a2
-42
202 TENSOR ANALYSIS

INTRINSIC DERIVATIVES.

62. Calculate the intrinsic derivatives of each of the following tensors, assumed to be differentiable
functions of t: (a) an invariant (l), (b) Ai, (c) Ak , (d) Al kn .

_ ,q =
a =
d4) , th e or di nary der ivative.
(a) b4 d q ax dtq
t
(b) 8AJ =A dxq aAi + 1
As
dxq = aAi dxq + 1 As dxq
8t q dt axq qs dt axq dt qs dt

dAI + 1 As dxq
dt qs dt

SAk
(c )
St
Aj
k,q
dxq
dt
{;q}4+{}A:) dt
q

dAkj
Aj
dz + As dxq
dt kq s dt qs k dt
jk jk
,. Almn , jk dxq l aAlnn s Akn S A jk
lq mq lsn

s Ajk + Ask + k Ajs dxq

nq Ins {i}
qs Inn qs lmn dt

dAA
dt
lmn Is Ajk
s1nn dt
dxq _ s Ajk dxq s Ajk _
dx q
lq lmqf lsn dt nq ins dt

+ j Ask dxq + £kAJS dxq

qs lmn dt qs Inn dt

63. Prove the intrinsic derivatives of gjk , gjk and are zero.
89jk
= (g ) dxq = 0,
8glk
=
gjk dxq
= 0,
S bk -8 dxq
=0 by Problem 54.
8t k ,q
dt St 'q dt bt k,q dt

RELATIVE TENSORS.

64. If Aq and Bts are relative tensors of weights w1 and w2 respectively, show that their inner and
outer products are relative tensors of weight w1 + w2-
By hypothesis,

A jw1 az axq A B
lm _ jw2 13F-' azm axt Brs
k axP axk q' n axr axs axn t

jw, w2 axq axq axl axm axt

The outer product is Ak Bnm Aq Btrs
axp az k axr axs az n
a relative tensor of weight w1 + w2 . Any inner product, which is a contraction of the outer product, is also
a relative tensor of weight w1 + w2.
 TENSOR ANALYSIS 203

65. Prove that Vg- is a relative tensor of weight one, i.e. a tensor density.
-3.p -3..q
The elements of determinant g given by g transform according to gj =
g
q k axe axk oq

Taking determinants of both sides, g = axe axe g = J2g or JV , which shows

azj -3-x

66. Prove that dV = Vrg- Al dx2 ... dx1 is an invariant.

By Problem 65, dV = Vg- dx1 dx ... dxy = vrg- J dx1 dx`2 ... dx'A

_ vg ax dz1 dx ... dxN = dx1 dx2 ... dx" = dV

a)
From this it follows that if is aninvariant, then

f...fdv = f...fdv
V V

for any coordinate systems where the integration is performed over a volume in N dimensional space. A
similar statement can be made for surface integrals.

MISCELLANEOUS APPLICATIONS.

67. Express in tensor form (a) the velocity and (b) the acceleration of a particle.

k
(a) If the particle moves along a curve xk = xk(t) where t is the parameter time, then vk = is its ve-
dt
locity and is a contravariant tensor of rank one (see Problem 9).

k 2k
(b) The quantity dt = t2
is not in general a tensor and so cannot represent the physical quantity
acceleration in all coordinate systems. We define the acceleration ak as the intrinsic derivative of
the velocity, i.e. ak = Stk which is a contravariant tensor of rank one.

68. Write Newton's law in tensor form.

Assume the mass M of the particle to be an invariant independent of time t. Then Mak = Fk a
contravariant tensor of rank one is called the force on the particle. Thus Newton's law can be written
S kk
Fk = Mak = M
204 TENSOR ANALYSIS

69. Prove that ak = 6vk = dx

2
k k dx0 dxq
St dt2 pq dt dt '

Since vk is a contravariant tensor, we have by Problem 62 (b)

2
8vk = dvk + k vs dxq d xk +
k dxq
VP
St dt qs dt dt2 qp dt

2
_ d xk + k dxp dxq
d- i pq dt dt

70. Find the physical components of (a) the velocity and (b) the acceleration of a particle in cylin-
drical coordinates.

(a) From Problem 67(a), the contravariant components of the velocity are
dx1 dp dx2 do dx3 dz
and
dt dt ' dt dt dt dt

Then the physical components of the velocity are

v911
dx1
dt
=
dp
dt 922
dx2
dt ' dt
do
and v 933
dx3
dt
=
dz
dt
911=1,g22=p2.g33=1.
using

(b) From Problems 69 and 49 (b) , the contravariant components of the acceleration are

d2x1 + dx2 dx2 _ d2 p do

al _ p( dt ) 2
dt2 22 dt dt dt2

a
2 d2 x2 2 dxl dx2 2 dx2 dxl do 2 dp do
dt2 + 12 dt dt + 21 dt dt dt2 + p dt dt

and a3 d2 x3 _ d2 z
dt2 dt2

Then the physical components of the acceleration are

11
al = 0 -- p 22 a2
=p + 2pq5 and 33 a 3 =z
where dots denote differentiations with respect to time.

71. If the kinetic energy T of a particle of constant mass M moving with velocity having magnitude v
is given by T = 2Mv2 = 2Mg pq 0 xq, prove that
d (aT) _ aT = Ma
k
dt axk axk
where ak denotes the covariant components of the acceleration.

Since T = 2Mgpq zp zq, we have

 TENSOR ANALYSIS 205

d ('3T
ag j
aT
I agpq zp zq,
aT = Mg zq and
= zq + q,,
kq M(gk 9
axk axk '31k dt ax ax

Then
d ( aT) - aT =
M
+
agk q zjzq - 1 agpq zPzq
dt azk axk (gkq zq axk 2 axk

agkp
M
(gkq
zq + 1 (agkq
2
+ - agpq) xp xq
axp axq axk

M(gkq zq + [pq,k] xp xq)

= Mg
xr + r zj z q = Mg ar = Mak
kr pq kr

using Problem 69. The result can be used to express the acceleration in different coordinate systems.

72. Use Problem 71 to find the physical components of the acceleration of a particle in cylindrical
coordinates.

Since ds2 = dp2+,02dca2+ dz2, v2 = ( )2 = z2 and T = 2Mv2 = e(;;? + Z2) ,

2+

From Problem 71 with x1 = p, x2 = 0, x3 = z we find

a1 = P a2 = dt a3 z

Then the physical components are given by

a, a2 ag

1V,gj-l 22 33

since g11 = 1, g22 = p2 , g33 = 1. Compare with Problem 70.

73. If the covariant force acting on a particle is given by Fk = - a where V (x1..... xj') is the
k
potential energy, show that dt (aLk) - a Lk = 0 where L = T -- V.

From L = T -- V, aL _ aT since V is independent of zk. Then from Problem 71,

axk r axk

d aT aT _
Mak Fk --
aV and d (aL) - aL
= = 0
dt ( azk azk axk dt azk axk

The function L is called the Lagrangean. The equations involving L, called Lagrange's equations,
are important in mechanics. By Problem 50 it follows that the results of this problem are equivalent to the
statement that a particle moves in such a way that f L dt is an extremum. This is called Hamilton's
t1
principle.
206 TENSOR ANALYSIS

74. Express the divergence theorem in tensor form.

Let A k define a tensor field of rank one and let vk denote the outward drawn unit normal to any point
of a closed surface S bounding a volume V. Then the divergence theorem states that

fffA kk dV II A k vk dS
V S

For N dimensional space the triple integral is replaced by an N tuple integral, and the double integral by
an N -1 tuple integral. The invariant A kk is the divergence of Ak (see Problem 57). The invariant
Ak vk is the scalar product of Ak and vk, analogous to A n in the vector notation of Chapter 2.
We have been able to express the theorem in tensor form; hence it is true for all coordinate systems
since it is true for rectangular systems (see Chapter 6). Also see Problem 66.

75. Express Maxwell's equations (a) div B = 0, (b) div D = 47rp, (c) Vx E = - aB , (d) VxH = 4Z f
in tensor form.

Define the tensors Bk, Dk, Ek, Hk, 1 k and suppose that p and c are invariants. Then the equations
can be written

(a) Bkk0
(b) D kk = 47Tp

(c) --- EJkq Ek,q 1 aBj

or EJWq Ek.q
1 aBj
C
at
J
47r1
(d) - Ejkq or Ejkq
Hk , 4 c Hk, q c

These equations form the basis for electromagnetic theory.

76. (a) Prove that Al,gr A0,rq = Rngr An where A0 is an arbitrary covariant tensor of rank one.
(b) Prove that R qr is a tensor. (c) Prove that R' is a tensor. Rpgrs gns
aAM
(a) Ap, gr = (AO.q) r axr
- 1
{Pr
A.
J.q - {'}A
qr O9

J,_ (BA.
aA
i A. - i j
k
- {}Ak) - PS l
Pq Pr qr axq P1

a2 A Aj
k
{;;iq} A. .- -- + A
Pq -ax, Pr axq Pr Iq k

aAP l
1 + 1
A
qr axq qr P1

By interchanging q and r and subtracting, we find

 TENSOR ANALYSIS 207

A p,qr _ i k Ak
a j
JA. - {pJq}{J}A
k
A p rq I r k
pr jq axr P4

- p
r
kq Aj
a ; A
9
_ kj
pq kr Ai
a xr Pq

RpJgr
= A
I

where Rqr = {:r}{q} _ - {:q}{Lr} + pr

axr P4 axq i
Replace j by n and the result follows.

(b) Since Ap,gr - Ap,rq is a tensor, Rqr An is a tensor; and since An is an arbitrary tensor, Rqr is

n n
a tensor by the quotient law. This tensor is called the Riemann-Christoffel tensor, and is sometimes
n
written R, pqr, R pqr , or simply R pqr

(c) Rpgrs = gns Rpgr is an associated tensor of Rpgr and thus is a tensor. It is called the covariant
curvature tensor and is of fundamental importance in Einstein's general theory of relativity.

SUPPLEMENTARY PROBLEMS
Answers to the Supplementary Problems are given at the end of this Chapter.

77. Write each of the following using the summation convention.

(a) a1 x1x3 + a2x2x3 + ... + a) xNx3 (c) Al B1 + A2 B2 + A3 B3 + ... + AjV BN
A21 + g23 g + g24g
(b) B1 + A22 B2 + A' B3 + + By
`42I
(d) g21 g11 + g22 g21 31 41
8221 + 8222
(e) B111 + B 122
12
+
21 22

78. Write the terms in each of the following indicated sums.

az,7 k
, N=2
(a) axk(iAk), N=3 (b) Bp C (c)
xk a'
79. What locus is represented by akxkxk = 1 where zk, k = 1, 2, ..., N are rectangular coordinates, ak are
positive constants and N = 2, 3 or 4 ?

80. If N = 2, write the system of equations represented by apq xq = b p .

k
81. Write the law of transformation for the tensors (a) Ak , (b) B , (c) Can , (d) An .
208 TENSOR ANALYSIS

82. Determine whether the quantities B(j,k,m) and C(j,k,m,n) which transform from a coordinate system
xti to another xti according to the rules
ax ax k axr B(j,k,m) axp axq axq axs C(j,k,m,n)
(a) B(p.q,r) = (b) C(p,q,r.s) =
axp axq axx axq az k azr axn
are tensors. If so, write the tensors in suitable notation and give the rank and the covariant and contra-
variant orders.

83. How many components does a tensor of rank 5 have in a space of 4 dimensions ?

84. Prove that if the components of a tensor are zero in one coordinate system they are zero in all coordinate
systems.

85. Prove that if the components of two tensors are equal in one coordinate system they are equal in all co-
ordinate systems.
k k
86. Show that the velocity = vk of a fluid is a tensor, but that v is not a tensor.
t

87. Find the covariant and contravariant components of a tensor in (a) cylindrical coordinates p, 0, z ,
(b) spherical coordinates r, 6, if its covariant components in rectangular coordinates are 2x --z, x2y,
yz .

88. The contravariant components of a tensor in rectangular coordinates are yz, 3, 2x+y. Find its covariant
components in parabolic cylindrical coordinates.
q 8s
89. Evaluate (a) 8q Bas, (b) Sq Sr Aqs, (c) 8p 8 8s , (d) 8q 8r 8s .

Arq r
90. If is a tensor, show that Ar is a contravariant tensor of rank one.

1 j=k
91. Show that is not a covariant tensor as the notation might indicate.
0 j#k

92. If A0 = a Aq prove that Aq = -axq

AP

93. If A r = azp axs 'IS prove that As = axq axr Ar

axq azr s s axp axs
.
94. If (D is an invariant, determine whether a is a tensor.
axp axq

95. If Aq and Br are tensors, prove that A0 Br and Aq Bq are tensors and determine the rank of each.

Pq
96. Show that if Ars is a tensor, then + ASS is a symmetric tensor and Ars - Asr is a skew-symmetric
tensor.

97. If Apq and Brs are skew-symmetric tensors, show that Cps = Ap4 Brs is symmetric.

98. If a tensor is symmetric (skew-symmetric), are repeated contractions of the tensor also symmetric (skew-
symmetric) ?

99. Prove that Apq xp xq = 0 if Apq is a skew-symmetric tensor.

 TENSOR ANALYSIS 209

100. What is the largest number of different components which a symmetric contravariant tensor of rank two
can have if (a) N = 4. (b) N = 6 ? What is the number for any value of N ?

101. How many distinct non-zero components, apart from a difference in sign, does a skew-symmetric covariant
tensor of the third rank have

102. If AIrs is a tensor, prove that a double contraction yields an invariant.

103. Prove that a necessary and sufficient condition that a tensor of rank R become an invariant by repeated
contraction is that R be even and that the number of covariant and contravariant indices be equal to R/2.

Brs
104. If Apq and are tensors, show that the outer product is a tensor of rank four and that two inner prod-
ucts can be formed of rank two and zero respectively.

105. If A(p, q) Bq = C where Bq is an arbitrary covariant tensor of rank one and C is a contravariant tensor
of rank one, show that A(p,q) must be a contravariant tensor of rank two.

106. Let AP and Bq be arbitrary tensors. Show that if AP Rq C(p, q) is an invariant then C(p,q) is a tensor
which can be written C C.

107. Find the sum S = A+B, difference D = A-B, and products P = AB and Q = BA, where A and B are the
matrices
(a) A = 3 -1 B 4 3
2 4 '
_
(2-1
-
2 0 1 -i
1 2
(b) A -1 -2 2 B= 3 2 -4
(_1 3 -1 -i -2 2

108. Find (3A-2B)(2A-B), where A and B are the matrices in the preceding problem.

109. (a) Verify that det (AB) = {det A } {det B } for the matrices in Problem 107.
(b) Is det (AB) = det (BA) ?

I -3 2 -1
1111. Let A = B= 1 3 -2
2 1 2

Show that (a) AB is defined and find it, (b) BA and A +B are not defined.

2 -1 3 x = 1
111. Find x, y and z such that 1 2 -4 y -.3
-1 3 -2 z 6

A'1
112. The inverse of a square matrix A, written is defined by the equation AA-1 = 1, where 1 is the unit
matrix having ones down the main diagonal and zeros elsewhere.
_ -1 1
Find A-' if (a) A = (_5 (b) A = (12 1 -1 .
4
1 -1 2
A-1
Is A = 1 in these cases ?
210 TENSOR ANALYSIS

2 1 -2
113. Prove that A = 1 -2 3 has no inverse.
4 -3 4

114. Prove that (AB) 71- = B-1 A-1, where A and B are non-singular square matrices.

115. Express in matrix notation the transformation equations for

(a) a contravariant vector (b) a covariant tensor of rank two (c) a mixed tensor of rank two.

1
116. Determine the values of the constant X such that AX = INX, where A = -3 and X is an arbi-

trary matrix. These values of X are called characteristic values or eigenvalues of the matrix A.

117. The equation F(X) = 0 of the previous problem for determining the characteristic values of a matrix A is
called the characteristic equation for A. Show that F(A)=0, where F(A) is the matrix obtained by re-
placing A. by A in the characteristic equation and where the constant term c is replaced by the matrix cl,
and 0 is a matrix whose elements are zero (called the null matrix). The result is a special case of the
Hamilton-Cayley theorem which states that a matrix satisfies its own characteristic equation.

118. Prove that (AB) = BT A_T .

119. Determine the metric tensor and conjugate metric tensor in

(a) parabolic cylindrical and (b) elliptic cylindrical coordinates.

120. Prove that under the affine transformation -'r = as xp + br, where ap and br are constants such that
apaq = bq , there is no distinction between the covariant and contravariant components of a tensor. In
the special case where the transformations are from one rectangular coordinate system to another, the
tensors are called cartesian tensors.

121. Find g and gjk corresponding to ds2 = 3 (dx1)2 + 2 (dx2)2 + 4 (dx3)2 - 6 dx1 dx3.

122. If Ak = gikAi , show that AJ = g Ak and conversely.

J,k

123. Express the relationship between the associated tensors

(a) Apq and q, (b) A qr and Aj ' ql, (c) Apgr and A..,,

124. Show that (a) APq B.s = A1'gBprs , (b) B7 r = AAgr Bpr = B . Hence demonstrate the gen-

eral result that a dummy symbol in a term may be lowered from its upper position and raised from its
lower position without changing the value of the term.
qr
125. Show that if A B Cr A; = B; q
a free index in a tensor equation may be raised or lowered without affecting the validity of the equa-
tion.

126. Show that the tensors g gpq and 89 are associated tensors.
pq'

gpq axe
127. Prove (a) Ejk ax _ g pq axQk (b) gdk axp
, =

axp ax ax ax

128. If AP is a vector field, find the corresponding unit vector.

 TENSOR ANALYSIS 211

129. Show that the cosines of the angles which the 3 dimensional unit vector Uti make with the coordinate
U1 U2 U3
curves are given by .

g11 y g22 933

130. Determine the Christoffel symbols of the first kind in (a) rectangular, (b) cylindrical, and (c) spherical
coordinates.

131. Determine the Christoffel symbols of the first and second kinds in (a) parabolic cylindrical, (b) elliptic
cylindrical coordinates.

132. Find differential equations for the geodesics in (a) cylindrical, (b) spherical coordinates.

133. Show that the geodesics on a plane are straight lines.

134. Show that the geodesics on a sphere are arcs of great circles.

135. Write the Christoffel symbols of the second kind for the metric
ds2 = (dx1)2 + [(x2)2 - (x1)2] (dx2)2
and the corresponding geodesic equations.

136. Write the covariant derivative with respect to xq of each of the following tensors:
AXkl'
(a) Aik, (b) Alm' (c) Ak1X, (d) (e) Ainn

j - jk 1
138. Use the relation A - g Ak to obtain the covariant derivative of A from the covariant derivative of Ak.

139. If 4> is an invariant, prove that ,pq= i.e. the order of covariant differentiation of an invariant
is immaterial.

140. Show that Ejgr and Epgr are covariant and contravariant tensors respectively.

141. Express the divergence of a vector AP in terms of its physical components for (a) parabolic cylindrical,
(b) paraboloidal coordinates.

142. Find the physical components of grad in (a) parabolic cylindrical, (b) elliptic cylindrical coordinates.
2
143. Find V 4) in parabolic cylindrical coordinates.

144. Using the tensor notation, show that (a) div curl Ar = 0, (b) curl grad = 0.

145. Calculate the intrinsic derivatives of each of the following tensor fields, assumed to be differentiable
functions of t : , k
(a) Ak, (b) Al , (c) Aj Bk, (d) OAk where 0 is an invariant.

k
146. Find the intrinsic derivative of (a) gjk A , (b) 8k Aj A. (c) gjk r A .
9 9 r

8A
147. Prove dt (gpq A A A q) = 29Pq A q
8t
212 TENSOR ANALYSIS

148. Show that if no external force acts, a moving particle of constant mass travels along a geodesic given by
p
as(ds) = a.

149. Prove that the sum and difference of two relative tensors of the same weight and type is also a relative
tensor of the same weight and type.

150. if Apq is a relative tensor of weight w, prove that g-V/2 Apq. is an absolute tensor.

151. If A(p,q) Bqs = where Br is an arbitrary relative tensor of weight wl and Cpr is a known relative
tensor of weight w2, prove that A(p,q) is a relative tensor of weight w2---w1. This is an example of
the quotient law for relative tensors.

152. Show that the quantity G(j,k) of Solved Problem 31 is a relative tensor of weight two.

153. Find the physical components of (a) the velocity and (b) the acceleration of a particle in spherical co-
ordinates.

154. Let Ar and Br be two vectors in three dimensional space. Show that if ,\ and,i are constants,then
Cr = X Ar+ LRr is a vector lying in the plane of Ar and Br. What is the interpretation in higher dimen-
sional space ?

155. Show that a vector normal to the surface 0 (xi, x2, x3) = constant is given by AO =9a. Find the
corresponding unit normal.
as
156. The equation of continuity is given by V (0- V) + = 0 where cr is the density and v is the velocity of
ac
a fluid. Express the equation in tensor form.

157. Express the continuity equation in (a) cylindrical and (b) spherical coordinates.

158. Express Stokes' theorem in tensor form.

159. Prove that the covariant curvature tensor Rpqrs is skew-symmetric in (a) p and q, (b) r ands , (c) q ands.

160. Prove Rpqrs = Rrsjiq

161. Prove (a) Rj,grs + Rpsqr + Rjirsq = 0,

(b) R¢grs + Rrgps + Rrspq + R¢srq = 0.

162. Prove that covariant differentiation in a Euclidean space is commutative. Thus show that the Riemann-
Christoffel tensor and curvature tensor are zero in a Euclidean space.

163. Let T 0 = dsP be the tangent vector to curve C whose equation is xP = x '(s) where s is the arc length.
q
(a) Show that g,g TP Tq - 1. (b) Prove that gig TO T = 0 and thus show that Nq = K &sq is a unit

normal to C for suitable K. (c) Prove that Nq is orthogonal to Nq

as

164. With the notation of the previous problem, prove:

TO S N q (SNq
(a) gig T' Nq = 0 (b) gig = - K or gpq T + K Tq) = 0.

r
Hence show that Br = I (6N + K Tr) is a unit vector for suitable T orthogonal to both and Nq .
 TENSOR ANALYSIS 213

165. Prove the Frenet-Serret formulas

$s
K Np, S= TB1'- KT 1',
sB
= - TN1'
where T N1' and BP are the unit tangent, unit normal and unit binormal vectors to C, and K and T are
the curvature and torsion of C .

166. Show that ds2 = c2(dx4)2 - dxk dxk (N=3) is invariant under the linear (affine) transformation
x2 x3 18
x1 = y(x1- vx4) , = x2 , = x3 , z4 = y(x4 -- x1)
2
where 'y,,8, c and v are constants, 8 = v/c and y = (1-,8 )71/2 This is the Lorentz transformation
of special relativity. Physically, an observer at the origin of the xi system sees an event occurring at
position x1,x2,x3 at time x4 while an observer at the origin of the 'xi system sees the same event occur-
ring at position 3F1,`x2,`x3 at time z4. It is assumed that (1) the two systems have the x1 and Z1 axes
coincident, (2) the positive x2 and x3 axes are parallel respectively to the positive x2 and x3 axes,
(3) the xi system moves with velocity v relative to the xi system, and (4) the velocity of light c is a
constant.

167. Show that to an observer fixed in the xi ((i) system, a rod fixed in the xi (xi) system lying parallel to
the x1 (xi) axis and of length L in this system appears to have the reduced length LA - This
phenomena is called the Lorentz-Fitzgerald contraction.

ANSWERS TO SUPPLEMENTARY PROBLEMS.

77. (a) akxkx3 (b) A23B (c) Ak Bk (d) g2q gq1 , N=4 (e) Bar, N= 2

78. (a) 2 1 (vg A1) + ax2 (v g A2) 2x3 (V -g A3) (c) azj ax 1 + axi axe ax ax A
+ + ... +
(b) All BP C, + A2' B1' C2 + A12
B2 C, + A22 B2 C2 axi ax-'IR
ax2 azm -axl -a--n

79. Ellipse for N= 2, ellipsoid for Nj= 3, hyperellipsoid for N= 4.

a11x1 + a12x2 = b1
80.
a21x1 + a22x2 = b2

ij
81. (a) A-pq
r
- ax-p axq ax k Ak (c) C axx axn
Cxn
axi -ax, avr q ax1' axq

(b) B-1'gr = axp azq ax ax's Bijk axi All

S (d) A
axi axi axk aTs = axi

82. (a) B(/, k, m) is a tensor of rank three and is covariant of order two and contravariant of order one. It can
be written (b) C(j, k, m, n) is not a tensor.

83. 45 = 1024

87. (a) 2p cos2 c - z cos 0 + p3 sin2 0 cos2o,

-- 2p2 sin 0 cos 0 + pz sin 0 + p4 sin 0 cos3
pz sin 0.
214 TENSOR ANALYSIS

(b) 2 r sin2 6 cos2 0 - r sin 6 cos 6 cos (p + r3 sin46 sin2 0 cos2 0 + r2 sin 6 cos2 6 sink,
2r 2 sin 6 cos 6 cos2 0 - r2 cos2 6 cos 4 + r4 sin3 6 cos (9 sin2 O cos2 O
- r3 sin26 cos 6 sink,
- 2r2 sin2 6 sin cos ) + r2 sine cos 6 sin + r4 sin4 6 sino cos3 o

88. u2vz + 3v , 3u - uv2z , u2 + uv - v2 89. (a) Bq

r,s (b)
P
(c) bs , (d) N

94. It is not a tensor. 95. Rank 3 and rank 1 respectively. 98. Yes.

100. (a) 10 , (b) 21 , (c) N(N+ 1)/2 101. N(N- 1) (N- 2)/6

107. (a) S
7 2
D
-1 -4 P
14 10 18 8
0 3 ' 4 5 0 2
Q
-8 -2
3 -1 3 1 1 -1 1 -4 6 1 8 -3
(b) S = 2 0 -2 , D= -4 -4 6 , p= -9 -7 10 , Q= 8 -16 11
-2 1 1 0 5 -3 9 9 -16 -2 10 -7

3 -16 20
108. (a)
52
104 -86 ) (b) 9 163 -136 110. -4 5 3
104 17 -2 1
-61 -135 132

1/3 1/3 0
111. x=-1, y=3, z=2 112. (a)
2

5/2 3/2
1
(b) - 5/3 1/3 1 . Yes
-1 0 1

x1

ax3
-ax'

ax1 ax1 ax1

A12 A13
ax1 aaG2 az3

(b) ax2 ax2 ax2

A 22 A 23
ax 1 ax- 2
- az3
ax1 ax2 ax3 ax3 '3X3
A31 A32 A33 A 32 A 33
ax ax ax1 ax2 ax3

ax1 ax1
ax1 ax2

(c) ax ax2
ax1 ax2

Al
3
A2
3
A3
3 ax3 ax
ax1 ax2

u
2 +v2 0 0 U'2
2+v2 0 0

2 + v2 0 ! \
116.\.=4,-1 119. (a) 0
11
, 0
u2 +v
1
2 0

0 0 1 0 0 1
 TENSOR ANALYSIS 215

a2(sinh2u + sin2v) 0 0 1 0 0
(a2(sinh2u + sin2v)
(b) 0 a2(sinh2u + sin2v) 0 0 1 0
a2(sinh2u + sin2v)
0 0 1 0 0 1

4/3 0 1

121. g = 6, 0 1/2 0
1 0 1

Ap9 _ gPi A 9, (b) A. q r

..r rl Jk
123. (a) Ap
q A..1
(c) = gpj gqk g

A A
128. or
AY , gig AP A q

130. (a) They are all zero.

(b) [22,1] =-p, [12,2] _ [21,2] = p. All others are zero.
(c) [22,1] _ -r, [33.1] r sin28, [33,2] _ - r2 sin 8 cos 8
[21,2] [12,2] =r, [31,3] = [13,3] = r sin28
[ 32,31 _ [ 23,3] = r2 sin 8 cos e. All others are zero.

131. (a) [ii,1] = u, [22,2] = v, [11,2] _ - v, [ 22,1] -u,

[12,1]= [21,1]=v, [21,2]= [12,2]=u.
1 _ u 2 v 1 -u 2 -v
11 u2 + v2 ' 22 u2 + v2 ' 22 u2 + v2 11 u2 + v2
1 _ 1 v 2 1 2 u All others are zero.
21 12 u2 + v2 21 12 u2 + v2

(b) [11,1] = 2a2 sinh u Cosh u

, [22,2] = 2a2 sine cosv , [11,2] _ -2a2 sinv cosv
[22.1] =-2a2sinh u cosh u, [12,1] _ [21,1] = 2a2 sinv cosv, [21,2]= [12,2]= 2a2sinhu coshu
51 sinh u coshu 2 _ sinv cosv 1 - sinh u cosh u
11 sinh2u + sin2v ' 22 sinh2u + sin2v ' 122 5 - sinh2u + sin2v
52 -sin v cosv 1 1 sinv cos v
1 11 sinh2u + sin2v 21 12 sinh2u + sin2v '

j2 = 2 sinh u cosh u .
All others are zero.
21 12 y sinh2u + sin2v

d dsP d2 0 + 2 do dq d2
0,
ds2z
132. (a) =0
- p( d2 0.
ds2 p ds ds
d'r d<p
{b) _ r(d8)2 r sing 0( )2 = 0
ds2

d2 B 2 dr !L6 sin6cos8(dO)2 = 0
ds2 r ds ds ds

d24) +2dr d0+ 2cot8d8d(k = 0

ds2 r ds ds ds ds
216 TENSOR ANALYSIS

1 _ 1 2 2 x1 2 x2
135. x All others are zero.
22 J 12 21
(x1)2 - (x2)2 122 (x2)2 - (x 1)2
2x1 dx1 dx2 + x2 dx2 2
d2 x1 + x( dx2 }2 = 0' =
ds2 ds
d 2x2 +
ds2 (XI)2_(X2)2
ds ds (x2)2 - (x1)2 ds )
0

jk
jk aAl s Ask k Ajs
136. (a) A + +
l,q axq lq { qs l
jk
(b)
jk aA im s
Asp
jk s jk {/}sk k
qs Ali
js
Aln,q axq lq mq A is + qs Alin +

(c ) A
klnc,q
a A
k lm. ( s
kq
Aj

lq
s Aj
ksm
- s
mq
Aj
k1s
+ j
qs
As
klm

jkl s jkl jsl l jks

(d) A,n, q - mq
As +
qs qs
axq
jk
jk aAlmn si A
s jk s jk j k
j Ask + k is
(e) Aln,q axq lq Asmn - mq Alsn - nqf Ins
+
qs Inn { qs Alnnn

kj
137. (a) gjk A q , (b) A1q Bk + Al Bk 0 (C) A
,9 j,q

141. (a)
u2 + v2 au
Au ) + aav ( u2 + v2 Av )

1 a2 Az
(b) uv(u2 + v2) au (uv u2 + v2 AU) + ava (uv u2 + v2 AV) +
uv az2

142. (a) 1 a 0 +
1 a ev +
a(1)
ez
uV
2+v2 au w u2 + v2 av az

(b)
1 ( a(p eu + ai ev) a
a sinh2u + sin2v au av +
az e2
where eu, ev and ez are unit vectors in the directions of increasing u, v and z respectiveiy.

143. 1
a2T +
a + (u2 + v2)
u2 + v2 au2 av2

-
145. (a)
8Ak

8t
=
A
k,q
dxq

dt
=
aAk

axq
_ s
kq
AS
A )
dxq

dt
dAk

dt
s

kq
As
A -
dxq

dt

aAjk dAjk + j }ASkdXq +

5
(b) A
St dt qs dt qs dt

k
(c) 8 (A. Bk) = SAC Bk + A bB
bt St bt

dAj - {.}'4s
s
dt
Bk +
dt
+ k
qs }
Bs dxq
dt
 TENSOR ANALYSIS 217

LA
+ 6(t AI
(d) Ak ) k
bt 8t 8t
(dAk
+ 1 AS
dxq -{ S
Ai dxq d'
dt qs k dt kq S dt dt k

8Ak dAk + k dxq

146. (a) AS
gjk 8t - gjk dt qs dt

6Aj - j (dAj

- s dxq _ dAk - s dxq

b 8j 8 A AS
() k 8i k dt { jq S dt dt kq dt

8r
j6A _r _
dAp s I r dxq + r
AS
dxq
(C) gjk AS
8t grk dt pq dt qs P dt

153. (a) r, re, r sine 0

156.
a _ 0 where v4 are the contravariant components of the velocity.
ax 2g ax at

1
157. (a) a (6-0) )
P
+ (ov2) + a (crv3) +P a= 0

a
(b) ar (crvl) +
a ao
(O"v2)
r
+ (0-v3)
'at
where v1, V2 and vs are the contravariant components of the velocity.
+ 0- (Zv1 + v2 cote) +
a 0--
0

158. fC
d
AP dsP ds
II
S
Epgr A dS where
dxq is the unit tangent vector to the closed curve C

and v1, is the positive unit normal to the surface S which has C as boundary.