Let us consider a general coordinate transformation:
x'a =
x'a(x), a =
1….,n |
(1) |
n is the dimension of space.
We have:
In analogy with this we say that Va is
a contravariant vector if under (1) it transforms like (2), i.e.
Similarly, let us consider the gradient of a scalar function. We say
that A(x) is a scalar function
if:A'(x') = A(x). It follows
that:
In analogy with this transformation law, we say that
Ua is a covariant vector if under a
change of variables (1), we have:
Tensor Product
Consider two covariant vectors,
Ua,Va.
We have:
U'a(x')V'b(x')
= Uc(x)Vd(x) |
(6) |
We say that a set of functions Tab that transforms
under (1) as (6) is a covariant tensor of rank 2. That is:
We call Sab =
UaVb
the tensorial product of the two covariant vectors
Ua and
Vb.
Generally we says that a set of functions T is a p-covariant and q-contravariant tensor if under (1)
we have:
T'(x') = ….….Tc1….cp
d1….dq(x) |
(8) |
The tensor product of two arbitrary tensors U,V defined by
is a (p+r) covariant and (q+s) contravariant tensor.
Contraction
Consider a mixed tensor T , then
S = ∑na =
1T
is a scalar.
NOTATION(Einstein): Repeated indices in a monomial are summed from 1 to
n.
Proof:
T
'
(
x
') =
T
(
x
) = δ
c
b
T
(
x
) =
T
(
c
)
In general the contraction of an upper index with a lower index in a
p-covariant and q-contravariant tensor produces a (p-1) covariant and
(q-1) contravariant tensor.
Notice that contraction of different indices, produces different
tensors.
Some important tensors
Kronecker
delta:δab is a
one covariant, one contravariant tensor. Moreover it is an invariant
tensor, because it has the same components in all coordinate systems.
Proof:
Symmetric and antisymmetric tensors.
Consider Tcab a tensor.
Then: Scab =
Tcab +
Tcba and
Acab =
Tcab -
Tcba are tensors, called the
symmetric and antisymmetric component of
Tcab.
Proof:
S'cab(x')
= T'cab(x') +
T'cba(x') =
Tijk(x)
+ Tijk(x)
= (Tijk(x)
+ Tikj(x)) =
Sijk(x) |
(11) |
The proof is similar for the antisymmetric part.Notice that this
operation can be applied to any pair of indices. Iterative application
of the operation will produce tensors that will form representations of
the permutation group Sn
Pseudotensors
We define a pseudotensor of weight r a set of functions
Db1….bqa1….ap
that under (1) transform as follows:
D'(x') =
Jr….….Dc1….cp
d1….dq(x) |
(12) |
where J is the jacobian of the transformation (1). i.e.
r=0, tensor
r=1, pseudotensor o tensor density
r=-1, tensor capacity
The Levi-Civita symbol
Here sgn is the sign of the permutation in brackets. If some of the
indices are repeated, it gives zero.
Use the determinant identity:
J
- 1
sgn(σ) = ∑
λϵSn
sgn(λ)
….
or
J -
1εa1….an
= ….εb1….bn |
(15) |
It follows that
εa1….an
is a pseudotensor of weight 1.
Similarly, we can prove that:
Jεa1….an
= ….εb1….bn |
(16) |
so
εa1….anis
a pseudotensor of weight -1.
An important identity
We have that:
εj1….jnεi1….in
= |
δi1j1….δi1jn |
δi2j1….δi2jn |
………… |
δinj1….δinjn |
| |
(17) |
Proof:
We must have: i1 =
σ(1)….,in =
σ(n) and j1 =
λ(1),….jn =
λ(n)for some permutations σ and λ,
otherwise the determinant vanishes (either two columns or two rows are
equal). Then, we get:
|
δi1j1….δi1jn |
δi2j1….δi2jn |
………… |
δinj1….δinjn |
| = sgn(σ)|
δ1j1….δ1jn |
δ2j1….δ2jn |
………… |
δnj1….δnjn |
| = sgn(σ)sgn(λ)|
δ11….δ1n |
δ21….δ2n |
………… |
δn1….δnn |
| = sgn(σ)sgn(λ) = ε
j1….jn
ε
i1….in
Covariant Derivative
In cartesian coordinates, the partial derivative of a tensor is a
tensor. This is no longer true in a general space. In order to define
parallel transport we must provide for more structure.
We assume the space is provided with a set of n3
functions Γijk which
transformation properties under (1) will be defined below:
Let Ai be a contravariant vector. We
define the covariant derivative of Ai
by:
Ai;j
= Ai,j +
ΓikjAk |
(18) |
NOTATION:
A
i
,j
=
We want the covariant derivative to be a tensor under (1). That is:
A'i;j(x')
= Aa;b(x)
= (Aa,b
+
ΓakbAk) |
|
|
=
A'i,j +
Γ'ikjA'k |
|
|
= (Aa(x)),j
+ Γ'ikj(Ab(x)) |
|
|
comparing LHS and RHS, we get:
The set of quantities Γijk that
transforms under (1) according to (19) is called an affine connection
and the spaces provided with it, spaces with affine connection.
We postulate that the covariant derivative satisfies Leibnitz's rule:
(TS),k =
T;kS +
TS;k |
(20) |
Consider Ai and
Bj, then S =
AiBi
is a scalar. For a scalar, the covariant derivative coincides with the
ordinary partial derivative. We have:
(AiBi);k
=
Ai;kBi
+
AiBi;k
= (Ai,k +
ΓijkAj)Bi
+
AiBi;k
=
Ai,kBi
+
AiBi,k
That is:
Ai(ΓjikBj
+ Bi;k -
Bi,k) = 0
This is true for any Ai. It follows
that:
Bi;k =
Bi,k -
ΓjikBj |
(21) |
In general:
ti1….ipj1….jq;k
=
ti1….ipj1….jq,k
+ ∑Γiahkti1….h….ipj1….jq
- ∑Γhjakti1….ipj1….ja
-
1h…jq |
(22) |
Torsion and Curvature
Let us consider a space with an affine connection Γ
We have:
T is a tensor, called the torsion tensor.
Additionally, consider:
ui;kl -
ui;lk = tensor =
Rijlkuj
-
Thklui;h |
(24) |
where:
Rijlk =
Γijk,l -
Γijl,k +
ΓhjkΓihl
-
ΓhjlΓihk |
(25) |
is the curvature tensor.