12—Tensors 302
is one that is the negative of its transpose. It is easiest to see the significance of this when the tensor
is written in the bilinear functional form:
Tij=T(eˆi,eˆj)
This matrix will equal its transpose if and only if
T(~u, ~v) =T(~v, ~u)
for all~uand~v. Similarly, if for all~uand~v
T(~u, ~v) =−T(~v, ~u)
thenT=−T ̃. Notice that it doesn’t matter whether I speak ofT as a scalar-valued function of two
variables or as a vector-valued function of one; the symmetry properties are the same.
From these definitions, it is possible to take an arbitrary tensor and break it up into its symmetric
part and its antisymmetric part:
T=
1
2
(
T+T ̃
)
+
1
2
(
T−T ̃
)
=TS+TA (12.23)
TS(~u, ~v) =
1
2
[
T(~u, ~v) +T(~v, ~u)
]
TA(~u, ~v) =
1
2
[
T(~u, ~v)−T(~v, ~u)
]
Many of the common tensors such as the tensor of inertia and the dielectric tensor are symmetric.
The magnetic field tensor in contrast, is antisymmetric. The basis of this symmetry in the case of the
dielectric tensor is in the relation for the energy density in an electric field,
∫ ~
E.dD~.* Apply an electric
field in thexdirection, then follow it by adding a field in theydirection; undo the field in thex
direction and then undo the field in theydirection. The condition that the energy density returns to
zero is the condition that the dielectric tensor is symmetric.
All of the above discussions concerning the symmetry properties of tensors were phrased in terms
of second rank tensors. The extensions to tensors of higher rank are quite easy. For example in the
case of a third rank tensor viewed as a 3-linear functional, it would be called completely symmetric if
T(~u, ~v, ~w) =T(~v, ~u, ~w) =T(~u, ~w, ~v) =etc.
for all permutations of~u,~v,~w, and for all values of these vectors. Similarly, if any interchange of two
arguments changed the value by a sign,
T(~u, ~v, ~w) =−T(~v, ~u, ~w) = +T(~v, ~w, ~u) =etc.
then theTis completely antisymmetric. It is possible to have a mixed symmetry, where there is for
example symmetry on interchange of the arguments in the first and second place and antisymmetry
between the second and third.
* This can be proved by considering the energy in a plane parallel plate capacitor, which is, by
definition of potential,
∫
V dq. The Potential differenceV is the magnitude of theE~ field times the
distance between the capacitor plates. [V =Ed.] (E~is perpendicular to the plates by∇×E~= 0.)
The normal component ofD~ related toqby∇.D~=ρ. [AD~.nˆ=q.] Combining these, and dividing
by the volume gives the energy density as