9.2.4 Tensor Basis Change Area
Tensor Basis Change Area explores how tensor bases transform under coordinate changes, revealing algebraic structures and invariance properties in multilinear algebra.
Tensor Basis Change Area is the classification of the distinct settings in which a change of basis for tensor components is carried out — a constant, global linear change of basis on a fixed vector space; a position-dependent, local change of basis arising from a change of coordinates on a manifold; and a restricted change confined to a particular subgroup, such as rotations or unimodular transformations, preserving some additional structure beyond mere linear independence — each area imposing its own characteristic form on the Jacobian matrix governing the change and on what remains invariant under it. It organizes basis change situations by what kind of transformation matrix is involved and what that matrix is permitted, or required, to preserve.
The Global Linear Change of Basis Area
A Single Constant Matrix for the Whole Space
In this area, the same invertible matrix J relates an old basis {eᵢ} to a new basis {ēᵢ} uniformly across the entire vector space, with no dependence on position, since the setting is a fixed vector space rather than a manifold; tensor components transform by this single, constant J and its inverse throughout, and no notion of a chart or coordinate patch enters the picture at all.
Characteristic Simplicity of This Area
Because the transformation matrix is constant, the transformation of any tensor's components is a single matrix multiplication with no differentiation involved, and there is no possibility of the Jacobian degenerating at some points and not others; this area is therefore the simplest setting for basis change, free of the scope and boundary concerns that arise once position-dependence is introduced.
The Local (Coordinate) Change of Basis Area
A Jacobian That Varies From Point to Point
When the basis in question is a coordinate basis on a manifold, changing from one coordinate system to another produces a Jacobian ∂x̄ⁱ/∂xʲ that generally varies from point to point, so the transformation of tensor components must be evaluated separately, using the locally appropriate Jacobian value, at each point under consideration.
Additional Concerns Specific to This Area
Because the Jacobian is position-dependent, this area inherits the full set of scope and boundary concerns discussed for coordinate systems generally — coordinate singularities where the Jacobian degenerates, restriction of validity to chart overlaps, and the need for connection coefficients when differentiating tensor fields, since ordinary partial derivatives of components no longer correctly account for the basis vectors' own point-to-point variation.
Structure-Preserving Restricted Change Areas
Orthogonal Changes Preserving the Metric
A restricted basis change area confines the transformation matrix to a specific subgroup chosen to preserve some additional structure; an orthogonal change of basis, for instance, is restricted to matrices satisfying JᵀJ = I, guaranteeing that an orthonormal basis is carried to another orthonormal basis and that the metric retains the identity form in the new basis as well as the old.
Other Structure-Preserving Subgroups
Depending on the structure to be preserved, other restricted areas arise: unimodular changes (det J = ±1) preserve volume, and symplectic changes preserve a chosen antisymmetric bilinear form; each such restricted area trades the full generality of an arbitrary invertible change of basis for the guarantee that some specific additional piece of structure survives the change unaltered.
Diagram Comparing the Basis Change Areas
Combining Areas Within a Single Calculation
Local and Restricted Areas Can Overlap
A basis change can belong to more than one of these areas simultaneously; a position-dependent rotation of an orthonormal frame field over a curved manifold is both a local (coordinate-adjacent) change, since the rotation angle varies from point to point, and a restricted, orthogonal change, since the transformation matrix at every point individually satisfies the orthogonality condition — the areas classify different aspects of a basis change rather than mutually exclusive categories.
Choosing the Right Area for the Structure Being Preserved
Recognizing which basis change area a given transformation belongs to determines which invariance properties can be relied upon afterward: a global linear change guarantees nothing is lost to positional variation, a local coordinate change requires the machinery of connections and covariant differentiation once derivatives are involved, and a restricted change guarantees that whatever additional structure defines its subgroup — orthogonality, volume, or a symplectic form — survives the change intact.