6.5 Tensor Valence Classification

Tensor Valence Classification is the classification of tensors by the exact, ordered sequence of contravariant and covariant argument slots they expose, refining type classification's bare totals (p, q) into a full left-to-right arrangement such as "up, down, up" and treating two tensors that share a type but differ in slot arrangement as distinguishable. It is the finest of the three classification schemes built directly on a tensor's index structure — order, type, and valence — and is the level of detail required whenever an operation must refer unambiguously to one specific argument position among several of the same variance.

The Core Definition

Valence as an Ordered Sequence of Variances

For a tensor whose indices are written in a fixed left-to-right order, the valence is the sequence recording the variance, upper or lower, of each position in turn:

Distinguishing Tensors of Equal Type but Different Valence

Type (2,1) admits three distinct valences: (↑,↑,↓), (↑,↓,↑), and (↓,↑,↑); a tensor written T^{ij}{}_{k} has the first valence, while one written T^{i}{}_{k}{}^{j} has the second, and although both are type (2,1), they are classified as different tensors under valence classification.

Counting the Valences Compatible With a Given Type

gives the number of distinct interleavings of p up-slots and q down-slots, growing quickly as p and q increase, which is why valence carries strictly more information than type once either p or q exceeds 1.

Notating Valence

Arrow-Sequence Diagrams

Valence is commonly displayed as a horizontal row of up-arrows and down-arrows, one per slot, giving an immediate visual contrast between tensors of the same type but different arrangement.

Slot-Position Labels

Alongside arrows, valence classification uses explicit ordinal labels — "the first slot," "the third slot" — allowing an operation to reference exactly one argument of a tensor without ambiguity, a precision unavailable from type or order alone.

Valence Under Tensor Operations

Concatenation Under the Tensor Product

reflects that forming a tensor product simply appends the second factor's slots, in order, after the first factor's slots.

Removal of Two Positions Under Contraction

Contracting a specific upper slot at position i against a specific lower slot at position j removes exactly those two positions from the valence sequence and closes the resulting gap, leaving the relative order of all remaining slots unchanged; this is a strictly more precise statement than the corresponding order-level rule, since it specifies which two positions are removed rather than only how many.

How Valence Relates to Order and Type

Valence Is the Finest of the Three Index-Structure Schemes

Order records only the length k of the valence sequence; type records only the count of up-arrows (p) and down-arrows (q) within it; valence records the sequence itself, so each coarser scheme can be recovered from valence by forgetting information, but not the reverse.

When Valence Adds No New Information Beyond Type

Whenever p ≤ 1 and q ≤ 1 (that is, type (0,0), (1,0), (0,1), or (1,1)), there is only one possible interleaving of the upper and lower slots, so valence and type coincide in information content; the distinction between the two schemes becomes meaningful only once a tensor has two or more slots of the same variance.

Operational Necessity of Valence Classification

Ambiguity Without Valence

An instruction to "contract the upper index against the lower index" of a type (2,1) tensor is ambiguous, since there are two upper slots to choose from; only by specifying the valence, and hence which specific slot is meant, does the instruction become well-posed.

Symmetrization and Antisymmetrization Require Valence

Symmetrizing or antisymmetrizing a tensor over a specified subset of its slots — for example, "symmetrize over the first and third upper slots" — is only meaningful once valence has fixed which positions are the first and third upper slots in the first place.

Why Valence Classification Matters

The Natural Endpoint of Purely Combinatorial Classification

Because valence records the complete arrangement of variances and nothing about the numerical values of components, it forms a natural stopping point for classification based purely on a tensor's combinatorial index structure, with any further distinctions necessarily coming from the tensor's actual component values, such as symmetry patterns, rather than from its slot arrangement.

Indispensable for Precise Multi-Index Manipulation

Any tensor with more than one slot of a given variance requires valence, not merely type, to specify unambiguously which slot participates in a given contraction, symmetrization, or basis substitution, making valence classification a practical necessity rather than a purely theoretical refinement whenever multiple same-variance slots are present.