6.7.4 Tensor Type Slot Organization

Tensor Type Slot Organization is the systematic arrangement of a tensor's argument positions, or "slots," into an ordered sequence of contravariant (upper) slots and covariant (lower) slots, together with the convention that fixes where each slot sits relative to the others so that a tensor of type (p, q) is treated as a specific multilinear map with p designated covector arguments and q designated vector arguments occupying determinate, individually addressable positions rather than an unordered bag of indices. This organization is what allows operations such as contraction, raising and lowering, and index substitution to be applied unambiguously to a particular slot instead of to the tensor as an undifferentiated whole.

Why Slots Must Be Organized

From Multilinear Maps to Addressable Positions

A tensor of type (p, q) over a vector space V is a multilinear map

Multilinearity alone says nothing about which argument is "first" or "third"; it only guarantees linearity in each slot separately. Slot organization is the additional bookkeeping layer that assigns each argument position a fixed identity — upper slot 1, upper slot 2, lower slot 1, and so on — so that statements like "contract upper slot 2 with lower slot 1" refer to a well-defined operation rather than an ambiguous one.

The Two Slot Families

Every slot belongs to exactly one of two families: contravariant slots, which accept covectors (elements of V*) and are conventionally written with upper indices, and covariant slots, which accept vectors (elements of V) and are conventionally written with lower indices. A tensor of type (p, q) has p contravariant slots and q covariant slots, and the pair (p, q) is itself a summary of the slot organization rather than a complete description of it, since it records how many slots of each kind exist but not the order in which they are interleaved.

Ordering Conventions

Fixed Left-to-Right Sequencing

The standard convention lists all contravariant slots first, in a fixed left-to-right order, followed by all covariant slots, also in a fixed left-to-right order, giving component notation of the form

Here the position of an index within the upper row or the lower row is exactly the slot's identity: i₁ is upper slot 1, i₂ is upper slot 2, j₁ is lower slot 1, and permuting the written order of the indices without also permuting the underlying arguments changes the tensor being denoted, since component values generally differ across slot permutations for tensors lacking symmetry.

Interleaved Conventions

Some authors interleave upper and lower indices to reflect a specific pairing relevant to a computation, for instance writing a (1,1) tensor as Tⁱⱼ in one context and grouping slots differently when composing with another tensor whose natural pairing runs the opposite way. Interleaving does not change which slots exist or their contravariant/covariant character; it only changes the notational grouping used to make a particular contraction or composition easier to read. The organization that matters operationally is which slot is which, not which order they happen to be printed in for a given calculation.

Slot Organization and Symmetry

Distinguishable Versus Interchangeable Slots

Slot organization becomes essential precisely when slots of the same variance are not symmetric under exchange. For a general (0,2) tensor T, the two covariant slots are distinguishable positions, and T(v, w) need not equal T(w, v); swapping the arguments fed into slot 1 and slot 2 produces a generally different scalar. Symmetric and antisymmetric tensors are special cases in which the value is invariant, or invariant up to sign, under such slot exchanges, but this invariance is a property proved about the organized slots, not a substitute for organizing them.

Slot Identity Under Symmetrization

Symmetrizing or antisymmetrizing a tensor over a subset of its slots requires first knowing which slots belong to that subset. The symmetrization of a (0,3) tensor over its first and third covariant slots, for example,

is only meaningful because "first slot" and "third slot" are fixed labels; without an established slot organization there would be no way to specify which two of the three covariant arguments are being exchanged.

Diagram of Slot Organization

Consequences for Tensor Operations

Contraction Requires a Slot Pair

Contraction acts on a designated contravariant slot and a designated covariant slot by setting their indices equal and summing. Because the operation is defined slot-by-slot, a tensor with several slots of each variance admits several distinct contractions, one for each choice of an upper slot paired with a lower slot, and these contractions generally produce different results; the slot organization is what makes "contract upper slot 2 with lower slot 1" a different, well-posed instruction from "contract upper slot 1 with lower slot 1."

Raising, Lowering, and Slot Reassignment

Raising a covariant slot to a contravariant one, or lowering a contravariant slot to a covariant one, using a metric tensor changes the variance label attached to a specific slot while leaving the other slots and their order untouched. The resulting tensor has an updated type (p, q) and a slot organization that agrees with the original in every slot except the one that was raised or lowered, which illustrates that slot organization is a per-slot attribute, modifiable independently, rather than a single global property of the tensor.

Composition and Slot Matching

When composing tensors, for instance forming a tensor product T ⊗ S or feeding the output of one contraction into another operation, the slots of the result are organized by concatenating the slot sequences of the inputs in a fixed, agreed order. Correct composition depends on tracking which slot of the product came from which input, since a mismatch in slot bookkeeping produces a tensor with the right type (p, q) but the wrong assignment of arguments to positions.