6.25.5 Tensor Order Type Operation Boundary

Tensor Order Type Operation Boundary is the set of type-matching conditions that separate operations on tensors which are always well-defined from operations which are well-defined only under specific type constraints, delineating exactly how much agreement in order and type two or more tensors must share before a given operation — addition, tensor product, or contraction — can be legitimately applied to them. Each of the basic tensor operations sits at a different point along this boundary, admitting different degrees of type mismatch before the operation becomes meaningless.

Addition: The Strictest Boundary

Exact Type Match Required

Tensor addition S + T is defined only when S and T have exactly the same type (p, q), since addition is performed component-by-component in a shared basis, and components of different types transform by different rules; adding a (2,0) array to a (1,1) array entrywise might be numerically possible but produces an object that does not transform consistently under any change of basis, and is therefore not a tensor operation at all.

where R = S + T,

only when both S and T are of type (1,1). This is the tightest boundary among the three basic operations: any type mismatch at all, in either p or q, disqualifies the sum.

Tensor Product: No Boundary at All

Any Types May Be Combined

Tensor product places no restriction whatsoever on the types of the tensors involved: a (2,0) tensor and a (0,3) tensor can always be combined via ⊗ to produce a (2,3) tensor, since tensor product simply concatenates index lists without requiring any of them to match or cancel. This operation sits at the opposite extreme from addition — it is universally defined regardless of type, which is why it is the operation used to build tensors of arbitrarily high and mixed type from simpler pieces.

Contraction: A Selective, Slot-Specific Boundary

Opposite-Variance Pairing Required

Contraction occupies a middle position: it is not blocked by type mismatch in general, but it requires that the two specific slots being contracted have opposite variance — one contravariant, one covariant — drawn from the same underlying space. Contracting an upper index against a lower index is legitimate:

while attempting to "contract" two upper indices or two lower indices together (without an intervening metric to first lower or raise one of them) is not a legitimate tensor contraction, since no cancellation of transformation factors occurs in that case.

Diagram of the Three Boundaries

Consequences of the Differing Boundaries

Why Some Expressions Fail to Type-Check

Any tensor expression can be checked for validity by applying these three boundary rules to each operation it contains: every addition must join equal types, every tensor product is unconditionally allowed, and every contraction must pair opposite-variance indices; an expression fails to type-check the moment any operation within it violates its specific boundary condition, regardless of whether the numerical values involved would produce a computable result.

Building Complex Operations From the Three Boundaries

Because every more elaborate tensor manipulation — matrix multiplication, trace, inner products, raising and lowering indices with a metric — decomposes into some combination of tensor product followed by contraction (and occasionally addition of the results), the three boundary conditions described here are jointly sufficient to validate the type-correctness of essentially any expression built from these operations, making them the complete set of rules needed to check well-formedness in tensor algebra.