6.2.4 Tensor Type Classification Area

Tensor Type Classification Area explores how tensors are categorized by their type, defining their structure and operations in algebraic frameworks.

Tensor Type Classification Area is the branch of tensor classification that splits a tensor's total slot count into the number of contravariant (upper) slots p and covariant (lower) slots q, recording the pair (p, q) as the tensor's type and attaching to each variance the transformation law that governs how its components change under a change of basis. This area sits between order classification, which gives only the sum p + q, and valence classification, which gives the full ordered arrangement of the p and q slots, refining the former while remaining coarser than the latter.

Defining Type as a Split of Indices

The Type Pair

For a tensor written T_{j₁...j_q}^{i₁...i_p}, the type is the ordered pair (p, q) where p counts the upper indices and q counts the lower indices:

type (T) = (p, q)

Contravariant Slots

An upper (contravariant) slot is one whose associated component transforms with the inverse of the Jacobian matrix of a change of basis x' = Jx:

V^{i^{'}} = \sum_{j} J_{j}^{i^{'}} - 1 V^{j}

Covariant Slots

A lower (covariant) slot is one whose associated component transforms directly with the Jacobian:

W_{i^{'}} = \sum_{j} J_{i^{'}}^{j} W_{j}

The Standard Named Types

Type (0,0): Scalars

A type (0,0) tensor has no indices of either variance and is a scalar, invariant under every change of basis since there is nothing for a Jacobian to act on.

Type (1,0) and (0,1): Vectors and Covectors

Type (1,0) is a vector, transforming contravariantly; type (0,1) is a covector or linear functional, transforming covariantly; the two are dual to one another, with a covector acting on a vector to produce a scalar in a way that is invariant under the paired transformation laws of the two types.

Type (1,1): Linear Operators

A type (1,1) tensor, once a specific vector space and its dual are both in play, is naturally identified with a linear operator on that space, since it has one contravariant slot to accept output and one covariant slot to accept input.

Type (0,2): Bilinear Forms

A type (0,2) tensor is a bilinear form on V, taking two vector arguments and returning a scalar; a symmetric, positive-definite example of this type is a metric tensor.

Diagram of Named Types by Position

Type Arithmetic Under Tensor Operations

Addition of Type Under the Tensor Product

type (A \otimes B) = (p_{A} + p_{B}, q_{A} + q_{B})

follows because forming a tensor product concatenates the upper indices of the two factors and separately concatenates their lower indices.

Matched Reduction of Type Under Contraction

Contraction is only type-consistent when it pairs one upper index against one lower index; doing so lowers the type from (p, q) to (p-1, q-1), which is the type-level refinement of the general order-level rule that contraction always removes exactly two indices.

Distinguishing Type From Order and From Valence

Type Versus Order

A single order, such as k = 2, covers the three distinct types (2,0), (1,1), and (0,2); type classification is strictly finer than order classification because it additionally records how the total is split between upper and lower.

Type Versus Valence

A single type, such as (2,1), still leaves open how its two upper and one lower slot are interleaved in left-to-right order; valence classification resolves this remaining ambiguity, making it strictly finer than type.

Type Versus Rank of a Matrix Representation

Two type (1,1) tensors, once represented as matrices in a basis, can have different linear-algebraic rank while sharing the exact same type; type says nothing about the numerical rank of the resulting matrix.

Why Type Classification Is Useful

Determining Transformation Behavior Without Computing It

Knowing a tensor's type immediately tells which transformation law — the inverse-Jacobian rule or the direct-Jacobian rule — applies to each of its slots, without needing to inspect the tensor's actual components or perform the transformation explicitly.

Identifying Familiar Objects by Their Type Signature

Recognizing that a vector is type (1,0), a covector is type (0,1), an operator is type (1,1), and a bilinear form is type (0,2) allows type classification to serve as a quick dictionary connecting the abstract tensor formalism to the more familiar objects of elementary linear algebra.