12—Tensors 318Take the case for whichf=yk, then
∂f
∂yj
=
∂yk
∂yj
=δjk
which gives
~e′k=~ei
∂yk
∂xi
(12.53)
The transformation matrices for the direct and the reciprocal basis are inverses of each other, In
the present context, this becomes
e′k.~e′j=δkj=~e`
∂yk
∂x`
.~ei∂x
i∂yj
=δi`
∂yk
∂x`
∂xi
∂yj
=
∂yk
∂xi
∂xi
∂yj
=
∂yk
∂yj
The matrices∂xi
/
∂yjand its inverse matrix,∂yk
/
∂xiare called Jacobian matrices. When you
do multiple integrals and have to change coordinates, the determinant of one or the other of these
matrices will appear as a factor in the integral.
As an example, compute the change from rectangular to polar coordinates
x^1 =x y^1 =r x^2 =y y^2 =φ
x=rcosφ r=
√
x^2 +y^2 y=rsinφ φ= tan−^1 y/x
~e′j=~ei
∂xi
∂yj
~e′ 1 =~e 1
∂x^1
∂y^1
+~e 2
∂x^2
∂y^1
=xˆ
∂x
∂r
+yˆ
∂y
∂r
=xˆcosφ+yˆsinφ=ˆr
~e′ 2 =~e 1
∂x^1
∂y^2
+~e 2
∂x^2
∂y^2
=xˆ
∂x
∂φ
+yˆ
∂y
∂φ
=xˆ(−rsinφ) +ˆy(rcosφ) =rφˆ
Knowing the change in the basis vectors, the change of the components of any tensor is found
by computing it in the new basis and using the linearity of the tensor itself. The new components will
be linear combinations of those from the old basis.
A realistic example using non-orthonormal bases appears in special relativity. Here the manifold
is four dimensional instead of three and the coordinate changes of interest represent Lorentz transfor-
mations. Points in space-time (”events”) can be described by rectangular coordinates(ct, x, y, z),
which are concisely denoted byxi
i= (0, 1 , 2 ,3) where x^0 =ct, x^1 =x, x^2 =y, x^3 =z
The introduction of the factorcintox^0 is merely a question of scaling. It also makes the units the
same on all axes.