7.4.3 Tensor Component Array Axis Role

Tensor Component Array Axis Role is the specific meaning assigned to each individual axis of a tensor's component array, determined by whether that axis corresponds to an upper or a lower index of the tensor, and by which particular index slot among several of the same variance it represents.

Definition and Scope

Axes Are Not Interchangeable

Although a tensor's component array may look like a generic multi-dimensional grid of numbers, each of its axes carries a distinct role inherited from the tensor's index structure. For a ((2,1)) tensor (T^{ik}_{\ \ j}), the array has three axes: the first corresponds to the upper index (i), the second to the upper index (k), and the third to the lower index (j), and swapping the first and third axes of the raw array would not produce a valid representation of the same tensor.

Upper Versus Lower Axis Roles

An axis tied to an upper index transforms contravariantly under a change of basis, using the change-of-basis matrix directly, while an axis tied to a lower index transforms covariantly, using the inverse of that matrix. This difference in transformation behavior is the fundamental reason axis role cannot be ignored when manipulating the array, even though, viewed purely as a grid of numbers, all axes look alike.

Structural Properties

Role Within a Group of Same-Variance Axes

When a tensor has more than one upper, or more than one lower, index, the axes sharing that variance must still be distinguished from one another unless a symmetry relates them. In a ((0,3)) tensor (T_{ijk}) with no symmetry assumed, the three lower axes each play a distinct role, and permuting them produces, in general, a different tensor; only when symmetry or antisymmetry is explicitly present do two same-variance axes become interchangeable without altering meaning.

Axis Role and Contraction

Contraction is only defined between an axis carrying an upper-index role and an axis carrying a lower-index role; summing over two axes that share the same variance, two upper indices for instance, is not a contraction in the standard sense and does not, by itself, produce a well-defined tensor without first invoking a metric to lower one of the indices.

Axis Role Under Reshaping

Reshaping a raw array, merging two axes into one or splitting one axis into two, is only meaningful for a tensor when the axes involved share a compatible role; merging two axes of matching variance and equal dimension into one longer axis can be done consistently, while merging an upper-index axis with a lower-index axis destroys the transformation properties that made the original object a tensor at all.

Role Within Tensor Algebra

Guiding Correct Implementation

Correctly tracking axis role is what allows a tensor library to apply the right transformation law when changing basis, the right pairing rule when contracting, and the right concatenation order when forming a tensor product, since each of these operations depends on knowing not just the size of an axis but the algebraic role that axis plays.

Documentation in Practice

Explicitly labeling each axis of a stored array with its role, such as recording which axes are contravariant and which are covariant, and in what order they correspond to the tensor's abstract index list, is standard practice in both mathematical notation and software interfaces, preventing silent errors that would otherwise arise from treating a tensor's axes as an undifferentiated grid.