Coordinate Transformations and The Action
Coordinate Transformations and The Action
Lecture 17
17.1
Coordinate Transformation
There is one last wrinkle in our classical field Lagrangian. We have seen
that
Z
S = L(, , ) d
(17.1)
is appropriate for Cartesian spatial coordinates and time, but what if we
wanted to transform to some other coordinate system?
Suppose we are in a two dimensional Euclidean space with Cartesian coordinates xj =(x,
Lecture 17
(17.2)
I=
(17.3)
B0
,
x
x
x0k
01
02
x
x
(17.4)
where we view x1 (x01 , x02 ), x2 (x01 , x02 ) as functions of the new coordinates.
There is a connection between the Jacobian and the transformed metric
according to the covariant transformation law
0
g
=
x x
x x
g
=
x0 x0
x0 x0
(17.5)
using the identity-matrix form for g . This means that the new metric can
be viewed as a matrix product of the Jacobian of the transformation
0
g
= J J
(17.6)
2 of 7
17.1.1
Lecture 17
Example
Jk =
=
(17.8)
y
y
sin r cos
r
(17.9)
0 r2
with determinant g = r2 , so it is indeed the case that det Jkj = g.
If we think of the integral over a volume in this two-dimensional space,
Z
Z
dxdy = r dr d,
(17.10)
then we can clearly see the transformation: d = dx dy is related to
d 0 = dr d via the Jacobian:
d = |J| d 0 ,
(17.11)
g d =
|J| p 0 0
g d ,
|J|
(17.13)
a scalar.
17.1.2
There is only one final issue: the signature of the metrics we will be using.
Since our spaces are really spacetimes, we have a Lorentzian signature
metric even with Euclidean spatial coordinates, the determinant of g
for Minkowski, say, is g = 1, so taking the square root introduces a factor
3 of 7
Lecture 17
change anything weve done so far, since for us, g has been one until
now, but in the interest of generality, the action for a scalar field theory is
S[(x )] =
Z
d
, ) ,
g L(,
|
{z
}
(17.14)
d is not the whole point of putting in the g factor was to make the
integration procedure itself a scalar operation.
17.1.3
Transformation of g
Lets look at how the determinant of the metric transforms under arbitrary
coordinate transformation. We have:
0
g
=
x x
g
x0 x0
(17.15)
and if we view the right-hand side as three matrices multiplied together, then
using the multiplicative property of determinants, det(A B) = det(A) det(B),
we have
x 2
0
g = det
g
(17.16)
x0
x
0
and the matrix x
0 is the Jacobian for x x, as above. The Jacobian for
the inverse transformation, taking us from x x0 is just the matrix inverse
x
of x
0 , so that:
0 2
x
0
g = det
g
(17.17)
x
4 of 7
Lecture 17
x
x
g 0 d 0 =
g det
d = g d
(17.20)
det
x
x
and we see that
scalar.
17.1.4
While were at it, its a good idea to set some of the notation for derivatives
of densities, as these come up any time integration is involved. Recall the
covariant derivative of a first rank (zero-weight) tensor:
A; = A, + A .
(17.21)
A; = ( g)p ( g)p A ;
The advantage of
a tensor A1 :
Lecture 17
g A; =
g A
,
(17.23)
R
H
so we can express the Cartesian divergence theorem: V E d = V Eda
in multi-dimensional form:
I
Z
g A da .
(17.24)
d g A; =
Notice also that the covariant divergence of the integrand reduces to the
17.2
Lets see how all of this works for our scalar field. We have the general field
equations
L
L
= 0,
(17.25)
1
1
, g , g = , g ,
2
2
(17.26)
and we recover
1 2
+ 2 = 0.
c2 t2
But suppose we have in mind spherical coordinates? Then
1 0 0
0
0 1 0
g =
0 0 r2
0
2
2
0 0 0 r sin
=
and
g = r2 sin . We have
g g , = 0.
6 of 7
(17.27)
(17.28)
(17.29)
Lecture 17
1 2
+ 2 = 0
c2 dt2
(17.30)
where now the field = (t, r, , ) and the Laplacian refers to the spherical
one.
This is an interesting development we started with a Lagrange density
that made the action a scalar, and now we discover that the field equations
themselves behave as tensors. Appropriately, the field equations have the
same form no matter what coordinate system we use!
17.3
Coordinate Invariance
The coordinate invariance we have just seen follows directly from the action
itself since we have been careful to write the action as a scalar, it does
not change under a change in coordinates. We can take x x0 and S
retains its numerical value (we integrate over all of the coordinates, with
integration region shifted appropriately). That sounds like a transformation
that should lead to some sort of conserved quantity. Next time, we will
carry out the transformation explicitly, but the basic idea is to perform the
coordinate transformation, which induces some transformation in the fields
and their derivatives viewed as an expansion, this gives us a perturbation
to S which then must be zero, enforcing the invariance of the action.
7 of 7