3.10.2 Tensor Covector Component Basis Dependence

Tensor Covector Component Basis Dependence is the property that the numerical values f_1, ..., f_n describing a covector are tied entirely to the particular basis of V used to build the dual basis through which those values were extracted, so that the identical abstract covector generally produces different component values under a different choice of basis. This dependence is not a defect of the component description; it is an intrinsic and expected feature of representing a basis-free object using basis-relative coordinates, and understanding it precisely is essential for correctly interpreting, comparing, and combining covector components.

Why Components Cannot Be Basis-Free

Components Are Defined Through a Choice

Each component f_i is defined as f(e_i), a value that explicitly depends on which vector e_i was designated as the i-th basis vector. Since this designation is a choice, made once at the outset and not part of the covector f itself, any two different choices of basis for V will, in general, assign different numerical roles to the same underlying functional.

Illustrating the Dependence Directly

Take V = R^2 and the covector f(x, y) = 3x - y. Relative to the standard basis e_1 = (1, 0), e_2 = (0, 1), the components are f_1 = 3, f_2 = -1. Relative to the alternative basis e'_1 = (1, 1), e'_2 = (2, 0), the components become f(e'_1) = 3 - 1 = 2 and f(e'_2) = 6 - 0 = 6, giving an entirely different pair of numbers for the same covector.

The Precise Transformation Behavior

Covariant Transformation Law

If a new basis relates to the old one by e'_j = A^i_j e_i, the components transform as

using the same matrix A that relates the two bases of V themselves, which is exactly why the components are called covariant: they vary together with the basis, rather than against it.

Verifying the Numerical Example

Applying this rule to the example above, with A the matrix whose columns give e'_1 = (1,1) and e'_2 = (2,0) in terms of the standard basis, that is A = [[1, 2], [1, 0]], the transformation gives f̃_1 = A^1_1 f_1 + A^2_1 f_2 = 1(3) + 1(-1) = 2 and f̃_2 = A^1_2 f_1 + A^2_2 f_2 = 2(3) + 0(-1) = 6, matching the directly computed values 2 and 6.

Distinguishing Component Dependence from Covector Identity

The Covector Itself Never Changes

Throughout any change of basis, the covector f remains the same fixed linear functional on V; only the tuple of numbers used to describe it changes. This distinction is analogous to describing the same physical location using different coordinate systems: the location itself does not move, but its coordinates differ depending on the reference frame chosen.

Practical Consequence: Basis Must Always Be Specified

Because of this dependence, quoting a covector's components without specifying the associated basis leaves the description incomplete and, strictly speaking, meaningless as a description of a particular covector, since the same tuple of numbers could correspond to entirely different covectors under different implicit bases.

Basis Dependence Versus Invariant Quantities Built from Components

Evaluation Results Remain Invariant

Although f_i individually depends on the basis, the evaluated scalar f_i v^i, obtained by pairing with a vector's components v^i, does not, since the covariant transformation of f_i and the contravariant transformation of v^i cancel exactly. This is the general phenomenon by which basis-dependent ingredients combine, through proper contraction, into basis-independent results.

Invariants That Can Be Built Directly from Components Alone

Certain combinations of a single covector's components, such as its full component tuple regarded as an element of F^n, remain basis-dependent with no invariant meaning on their own; extracting an invariant, such as a norm, requires additional structure, like a metric on V*, that itself must be specified independent of any particular basis.

Practical Guidance

Tracking Basis Choice Through a Calculation

When performing calculations involving covector components across multiple steps, it is essential to fix a single basis at the start and either maintain it consistently or explicitly apply the transformation rule whenever the basis changes partway through, to avoid inadvertently mixing components described relative to different, incompatible bases.

Recognizing Basis-Dependent Errors

A common source of error in tensor calculations is treating components computed relative to one basis as though they were relative to another; verifying that a claimed relationship between components respects the covariant transformation rule is a reliable way to catch such mistakes before they propagate further.

Diagrammatic Summary

The diagram shows the same covector producing two structurally different component tuples depending on the basis chosen, illustrating basis dependence concretely.