2.19.4 Tensor Abstract Operation Handling

Tensor Abstract Operation Handling is the set of rules governing how structure-preserving maps, such as linear maps, multilinear maps, and the operations built from them like the tensor product, contraction, and composition, are defined and manipulated on abstractly presented vector spaces so that every such operation remains well-defined regardless of how its inputs happen to be represented. Where abstract element handling concerns the objects a tensor construction produces, abstract operation handling concerns the functions acting between and among those objects: it specifies how a map is allowed to be defined on a tensor product space, how maps on the factor spaces induce a corresponding map on the tensor product itself, and what properties, such as compatibility with composition and identity maps, any legitimate tensor operation must preserve.

Operations as Structure-Preserving Maps

Linear Maps Between Abstract Vector Spaces

A linear map f : V → W between abstract vector spaces is an operation satisfying f(α v_1 + β v_2) = α f(v_1) + β f(v_2) for all scalars α, β and vectors v_1, v_2, defined without reference to any basis. Handling such a map abstractly means specifying its action on elements of V directly, or, when a basis is convenient, specifying it on basis vectors and extending by linearity, with the guarantee that the extension is independent of which basis was used.

Multilinear Maps as the Primary Building Operation

Multilinear maps generalize linear maps to several arguments and are the operations from which tensors are built in the first place; abstract operation handling treats a multilinear map T : V_1 × ... × V_k → U as a single well-defined operation, characterized entirely by its behavior on each slot separately, prior to any question of how it might be computed in coordinates.

Defining Operations Through the Universal Property

Factoring an Operation Through the Tensor Product

The defining technique for handling operations abstractly is to use the universal property of the tensor product: given a multilinear map B : V × W → U, there exists a unique linear map L : V ⊗ W → U such that B = L ∘ ⊗. Constructing an operation on a tensor product space is therefore reduced to constructing a multilinear map on the original factor spaces, a strictly simpler task, and then invoking the universal property to obtain the corresponding linear operation automatically.

Well-Definedness on General Elements

Because a general element of V ⊗ W is a sum of elementary tensors and can be written in multiple equivalent ways, the operation L defined this way must give a consistent answer on every such sum; the universal property guarantees this automatically, which is precisely why operations on tensor spaces are always constructed by first defining a multilinear map on the factors rather than by attempting to specify L directly on arbitrary sums.

The Induced Tensor Product of Linear Maps

Constructing f Tensor g

Given linear maps f : V_1 → V_2 and g : W_1 → W_2, abstract operation handling defines the induced map f ⊗ g : V_1 ⊗ W_1 → V_2 ⊗ W_2 on elementary tensors by

and extends this to all of V_1 ⊗ W_1 by linearity, again relying on the universal property to guarantee that this extension is single-valued despite the source space consisting of equivalence classes rather than single formal expressions.

Functorial Properties: Composition and Identity

A central requirement in abstract operation handling is that this induced construction behaves functorially: it must send the identity map on each factor to the identity map on the tensor product, \mathrm{id}_V ⊗ \mathrm{id}_W = \mathrm{id}_{V ⊗ W}, and it must respect composition, (f_2 ∘ f_1) ⊗ (g_2 ∘ g_1) = (f_2 ⊗ g_2) ∘ (f_1 ⊗ g_1). These two properties are what make the tensor product construction compatible with sequences of linear transformations applied to the factor spaces, rather than an operation that must be recomputed from scratch after each transformation.

Contraction as an Abstract Operation

Defining Contraction Without Coordinates

Contraction, ordinarily described in components as summing over a repeated upper and lower index, is handled abstractly as the linear map obtained by applying the natural evaluation pairing V* × V → F to one contravariant and one covariant slot of a tensor, extended by linearity to the whole tensor product space, with all remaining slots left untouched.

for φ ∈ V*, v ∈ V, and x an arbitrary remaining tensor factor, showing contraction as a genuine abstract operation defined uniformly on the whole space, not as an index-bookkeeping trick tied to one basis.

Permutation Operations on Tensor Slots

Handling the action of permuting tensor arguments, used to define symmetric and alternating tensors, follows the same pattern: a permutation σ of k slots induces a linear map on V^{⊗k} sending elementary tensors v_1 ⊗ ... ⊗ v_k to v_{σ(1)} ⊗ ... ⊗ v_{σ(k)}, extended by linearity, with well-definedness again following from the universal property rather than from any coordinate computation.

Diagrammatic Summary

The diagram shows the multilinear map B factoring through the tensor product V ⊗ W as B = L ∘ ⊗, the defining mechanism by which every abstract operation on a tensor product space is constructed and guaranteed to be well-defined.