6.16.5 Tensor Zero Two Algebraic Role

Tensor Zero Two Algebraic Role is the function that type (0,2) tensors serve within the broader algebraic structure of tensor algebra, acting as elements of a vector space in their own right, as bilinear forms defined directly on V, and as the algebraic source of the metric structure that connects a vector space to its dual through the operations of raising and lowering indices. This role makes type (0,2) tensors indispensable wherever geometry, distance, or angle needs to be introduced into an otherwise purely linear-algebraic setting.

The Vector Space Role

Addition and Scalar Multiplication

The set of all type (0,2) tensors on V forms a vector space V* ⊗ V*, with addition and scalar multiplication defined componentwise:

This closure allows the space of type (0,2) tensors, including the subspace of symmetric bilinear forms, to be studied with ordinary linear algebra, treating individual metrics or forms as points in an n²-dimensional space, or an n(n+1)/2-dimensional space once symmetry is imposed.

Dimension as an Algebraic Invariant

Since dim(V* ⊗ V*) = n², the algebraic role of type (0,2) tensors includes serving as a concrete space on which further linear structure can be imposed, such as the cone of positive-definite symmetric forms used to define genuine inner products among all possible symmetric type (0,2) tensors.

The Bilinear Form Role

Acting on Pairs of Vectors

Algebraically, a type (0,2) tensor T is identified with a bilinear map V × V → ℝ, via T(v, w) = T_{ij} v^i w^j. This identification is an isomorphism between the space of type (0,2) tensors and the space of bilinear forms on V, so every algebraic fact about bilinear forms, such as classification by rank and signature, transfers directly to type (0,2) tensors.

Signature and Positive Definiteness

When T is symmetric, its algebraic role includes carrying a signature, the count of positive, negative, and zero eigenvalues of its matrix, which is invariant under the congruence transformation T' = A^T T A even though the eigenvalues themselves are not. A symmetric type (0,2) tensor with signature consisting entirely of positive eigenvalues is positive definite and qualifies as a genuine inner product, the algebraic foundation of Euclidean and Riemannian geometry.

The Metric Generating Role

Raising and Lowering Indices

The central algebraic role of a nondegenerate type (0,2) tensor g_{ij} is to define an isomorphism between V and V*, lowering the index of a vector v^i to produce a covector v_i = g_{ij} v^j, and, through the inverse g^{ij}, raising the index of a covector back to a vector. This isomorphism is what allows the type (0,2), (1,1), and (2,0) classifications to be treated as different faces of the same underlying object once a metric has been fixed.

Building Operators from Forms

Given a symmetric type (0,2) tensor T_{ij} and a metric g_{ij}, the algebraic role of T extends to defining a type (1,1) operator S^i_j = g^{ik} T_{kj}, whose eigenvalues, called the eigenvalues of T relative to g, are invariant under changes of basis that preserve the metric, and which underlie constructions such as principal axis decompositions.

Diagram of the Algebraic Connections

Algebraic Consequences of This Role

Congruence Classes

Symmetric type (0,2) tensors are classified up to congruence, T ~ A^T T A, and Sylvester's law of inertia guarantees that the signature is the complete invariant of this classification over the real numbers, meaning any two symmetric type (0,2) tensors with the same signature are congruent, and hence algebraically indistinguishable up to a change of basis.

Interaction with the Exterior and Symmetric Algebras

Just as with the type (2,0) classification, the symmetric part of the type (0,2) construction generates the degree-two piece of the symmetric algebra on V*, while the antisymmetric part generates the degree-two piece of the exterior algebra on V*, making type (0,2) tensors the covariant counterpart of the same algebraic layering described for contravariant tensors, but built from covectors rather than vectors.