2.20 Tensor Dual Vector Space Context

Tensor Dual Vector Space Context is the framework that specifies how the dual space of a vector space participates in the construction of tensors, fixing the roles that vectors and covectors play as arguments, index positions, and transformation behavior when the two spaces are combined to build multilinear objects. Where a vector space V supplies the raw directions and magnitudes of a geometric setting, its dual space V*, the space of linear functionals on V, supplies the complementary machinery needed to extract scalars from those directions. The dual vector space context is the discipline of keeping straight which space an index belongs to, how bases of V and V* are paired, and how that pairing propagates through every tensor built on top of V.

The Dual Space as a Partner Structure

Definition of the Dual Space

Given a vector space V over a field F, the dual space V* is the set of all linear maps from V to F, itself a vector space under pointwise addition and scalar multiplication. Elements of V* are called covectors, linear functionals, or one-forms. For finite-dimensional V, V* has the same dimension as V, but the two spaces are not canonically identified with one another; an explicit choice, such as an inner product or a basis, is required to build an isomorphism between them.

The Natural Pairing

The defining relationship between V and V* is the natural pairing, a bilinear evaluation map

sending a covector f and a vector v to the scalar f(v), written ⟨f, v⟩. This pairing is basis-independent and is what allows the dual space to act on V without any auxiliary structure such as a metric.

Dual Basis and Index Placement

Constructing the Dual Basis

Given a basis e_1, ..., e_n of V, the dual basis e^1, ..., e^n of V* is the unique basis satisfying

where δ^i_j is the Kronecker delta, equal to 1 when i = j and 0 otherwise. Each dual basis covector e^i is the functional that reads off the i-th coordinate of a vector expressed in the basis e_1, ..., e_n.

Upper and Lower Index Convention

The dual vector space context is the source of the convention that vector components carry upper indices and covector components carry lower indices. A vector v in V is written v = v^i e_i, with components v^i, while a covector f in V* is written f = f_i e^i, with components f_i. This asymmetry in index placement is not cosmetic: it tracks which space, V or V*, each object belongs to, and it determines how that object transforms when the basis of V changes.

Change of Basis and Contragredient Transformation

Transformation of the Basis of V

If the basis of V changes according to a matrix A, so that a new basis vector is ẽ_j = A^i_j e_i, then the components of a vector transform with the inverse matrix, ṽ^i = (A^{-1})^i_j v^j, since the vector itself must remain fixed regardless of how it is described.

Transformation of the Dual Basis

The dual basis transforms contragrediently, meaning it uses the inverse-transpose of the matrix used for V. If ẽ^i = (A^{-1})^i_j e^j describes the new dual basis, then the components of a covector transform using A directly, f̃_i = A^j_i f_j. This opposite transformation behavior is precisely what keeps the pairing ⟨f, v⟩ invariant under any change of basis, since the matrix and its inverse cancel:

Consequence for Tensor Index Rules

This contragredient behavior is the reason tensor calculus assigns different transformation rules to upper and lower indices in general: every upper index transforms like a V-component, and every lower index transforms like a V*-component. The dual vector space context is what makes this rule consistent across tensors of arbitrary rank, since a tensor of type (p, q) is built from p copies of V and q copies of V*.

Double Duality and Canonical Identification

The Double Dual Space

The dual of the dual space, V** = (V*)*, consists of linear functionals on V*. For finite-dimensional V, there is a canonical isomorphism between V and V** that does not depend on any choice of basis: each vector v in V is identified with the functional on V* given by evaluation, v ↦ (f ↦ f(v)).

Why This Matters for Tensor Contexts

Because this identification is canonical, V can be treated as playing both the role of "vectors" and the role of "functionals on covectors" without ambiguity. This is what licenses the multilinear-map perspective on tensors, in which a (p, q)-tensor is regarded as a map taking p covector arguments and q vector arguments, since the p slots that would otherwise require elements of V** can be filled directly by elements of V.

Distinguishing the Dual Context from a Metric-Induced Identification

No Canonical Isomorphism Between V and V*

Unlike the double dual, there is in general no canonical isomorphism between V and V* itself; any identification of the two requires an additional structure, most commonly a nondegenerate bilinear form such as a metric tensor. Confusing this metric-dependent identification with the basis-independent dual vector space context is a common source of error, since raising and lowering indices via a metric is a choice, while the pairing ⟨f, v⟩ between V* and V is not.

Role in Building Mixed Tensors

The dual vector space context provides the raw material, the pair (V, V*) together with their pairing and dual bases, from which all mixed tensors of type (p, q) are assembled, prior to and independent of any metric structure that a particular application might later introduce.

The diagram shows V and V* linked by the natural pairing in one direction and by contragredient basis transformation in the other, the two features that jointly define the dual vector space context used throughout tensor algebra.