2.19.1 Tensor Coordinate Free Vector Space

Explore how tensor coordinate-free vector spaces generalize linear algebra by encoding multilinear relationships without reliance on coordinate systems.

Tensor Coordinate Free Vector Space is a vector space treated as a geometric or algebraic object in its own right, whose elements, operations, and defining properties are specified without reference to any particular coordinate system, numerical tuple representation, or fixed basis. It is the specific realization of the abstract vector space idea in which a vector is understood as an intrinsic entity, such as a directed displacement or an element of an equivalence class of representations, rather than as a list of numbers indexed by position, and it is the setting in which the definitions used to build tensors are stated so that they remain valid under every possible choice of basis simultaneously, rather than being tied to one choice from the start.

Vectors Without Coordinates

The Coordinate Tuple Versus the Underlying Object

The familiar space R^n presents vectors as ordered tuples of real numbers, (x_1, ..., x_n), but this presentation already commits to a specific basis, the standard basis, whose choice is not forced by the underlying geometric or algebraic content. A coordinate-free vector space instead treats a vector v as an element of an abstract set V closed under addition and scalar multiplication, with no tuple of numbers attached to it until a basis is deliberately introduced for computation.

Geometric Vectors as a Motivating Example

A directed line segment, or arrow, from one point to another in physical space is a natural coordinate-free vector: it has a length and a direction independent of any coordinate grid drawn over the space. Two arrows are regarded as the same vector precisely when they have equal length and direction, an equivalence relation defined purely geometrically, with no numerical coordinates required to state it.

Operations Defined Intrinsically

Addition Without Components

Vector addition in the coordinate-free setting is defined by an intrinsic rule, such as the parallelogram or head-to-tail construction for geometric vectors, or by the abstract vector space axioms for a general V, rather than by adding corresponding numerical entries. Once a basis e_1, ..., e_n is chosen, the familiar componentwise addition rule is recovered as a consequence, not assumed as the definition.

v + w = \sum_{i = 1}^{n} (v^{i} + w^{i}) e_{i}

Scalar Multiplication as an Intrinsic Rescaling

Similarly, scalar multiplication α v is defined as an intrinsic rescaling of the vector's magnitude, with direction preserved or reversed depending on the sign of α, again independent of any coordinate description, with the componentwise formula α v = \sum_i (α v^i) e_i emerging only after a basis is fixed.

Why Coordinate Freedom Matters for Tensor Definitions

Avoiding Basis-Dependent Artifacts

A quantity defined using explicit coordinates risks accidentally depending on the choice of basis used to define it, producing a rule that looks like a tensor in one basis but fails to transform correctly in another. Building tensor definitions, such as the tensor product, contraction, and symmetrization, directly from the coordinate-free vector space avoids this risk entirely, because no basis is referenced anywhere in the construction.

Tensors as Coordinate-Free Objects First

A tensor in this setting is an element of a tensor product space built from copies of V and V*, both understood coordinate-free, so the tensor itself is a single geometric object; its components relative to a basis are a secondary, derived description, computed by pairing the tensor against basis vectors, and different bases give different but equivalent numerical descriptions of the same coordinate-free tensor.

T^{i} = T (e^{i})

where e^i denotes the dual basis covector paired against the coordinate-free tensor T to extract its i-th component.

The Dual Space in Coordinate-Free Terms

Linear Functionals as Intrinsic Objects

The dual space V* of a coordinate-free vector space V is defined intrinsically as the set of all linear functionals on V, that is, all linear maps from V to the underlying field, with no reference to a dual basis until one is needed for computation. This mirrors the treatment of V itself and keeps the pairing between V and V* a coordinate-free bilinear operation.

The Natural Pairing

The evaluation pairing ⟨ ·, · ⟩ : V* × V → F, sending a covector and a vector to the scalar obtained by applying the functional to the vector, is defined without coordinates and is exactly the operation used, coordinate-freely, to contract tensor indices once components are introduced.

Diagrammatic Summary

The diagram shows a single coordinate-free vector v on the left, defined without any numerical tuple, and its representation as a coordinate tuple on the right, obtained only after a basis has been chosen, with the underlying vector remaining the same object throughout.