8.23.4 Tensor Index Notation Contraction Boundary

Tensor Index Notation Contraction Boundary is the specific subset of the general index notation boundary concerned with where the contraction operation, as expressed by a repeated upper-lower index, stops being valid or well-defined — same-variance pairings that require the metric before they qualify as a contraction, contractions performed on quantities that are not genuine tensors, contractions attempted across incompatible spaces, and contractions that become ill-defined when the metric itself is degenerate. It marks precisely where the convenient shorthand "repeated index means summed" can no longer be applied without additional justification or additional structure being supplied first.

The Boundary at Same-Variance Repeated Indices

Repetition Without a Metric Is Not Yet a Contraction

An index letter repeated twice with the same variance, such as in AᵢBᵢ with both indices lower, does not by itself constitute a valid Einstein-convention contraction; the convention as originally stated applies only to an upper-lower pair, so a same-variance repetition sits right at the notation's boundary — it may represent a legitimate operation once the metric is invoked to raise one of the indices, g^{ij}A_iB_j, but it is not itself, as written, a contraction in the strict sense the notation defines.

before it lies within the domain where the summation convention applies unambiguously.

Conventions That Blur This Boundary

In settings that fix an orthonormal basis throughout, such as elementary Cartesian tensor treatments, upper and lower components coincide numerically, and some presentations write every index as a subscript and sum over any repeated subscript regardless of variance; this convention works only because the metric is implicitly the identity in that fixed basis, and it silently steps past the contraction boundary by suppressing the distinction that the boundary is defined in terms of — the convention remains valid only as long as the basis is never changed to a non-orthonormal one.

The Boundary at Non-Tensorial Quantities

Contracting an Index of a Non-Tensor

Applying the contraction operation to an object that carries indices but does not transform as a tensor — such as the Christoffel symbols Γ^i_{jk} — produces a result whose coordinate-independence is not guaranteed by the ordinary contraction argument, since that argument relies on the Jacobian-cancellation property that only genuine tensors possess. A contraction such as Γ^i_{ij}, sometimes performed in specific derivations, must be justified separately rather than assumed valid purely because it has the surface form of a standard contraction.

Mixed Expressions Straddling the Boundary

Many important formulas in differential geometry combine genuine tensors with non-tensorial connection coefficients in the same contracted expression, such as the components of a covariant derivative; the contraction boundary here is not a prohibition but a signal that extra terms (correction terms built from the connection) are required precisely to compensate for the non-tensorial piece, restoring an overall expression that does transform correctly even though a naive contraction of its individual non-tensorial ingredients would not.

The Boundary at Contraction Across Incompatible Spaces

Indices Must Range Over the Same Space to Be Paired

A contraction is only meaningful when the upper and lower indices being paired range over the same index set, associated with the same underlying vector space (or the appropriate dual); attempting to contract an index belonging to one vector space against an index belonging to an unrelated vector space of possibly different dimension is outside the notation's domain, since there is no canonical pairing between the two spaces' bases that the summation convention could refer to.

Degenerate or Non-Invertible Metrics

When the metric tensor used to convert same-variance indices into a contractable pair is degenerate (not invertible), the raising and lowering operations that would normally supply the missing partner index are not well-defined for at least some vectors, and any contraction relying on the inverse metric g^{ij} sits at the boundary of validity in exactly the directions where the metric degenerates; this is a genuine limitation encountered in certain physical and geometric settings involving degenerate or null structures.

Diagram of Valid Contraction Versus Boundary Cases

Practical Handling at the Boundary

Supplying the Missing Structure

Most boundary cases are resolved, rather than left unresolved, by supplying whatever additional structure the plain contraction rule was missing: the metric to convert a same-variance pair, correction terms to account for non-tensorial connection pieces, an explicit identification map to relate two otherwise unrelated vector spaces, or a restriction to the non-degenerate directions of a degenerate metric. Recognizing that an expression sits at the contraction boundary is therefore usually the first step toward correctly extending the calculation, not a reason to abandon it.

Flagging Boundary Cases in a Derivation

Because contractions at the boundary do not carry the same automatic coordinate-independence guarantee as contractions safely inside the notation's core domain, careful tensor derivations explicitly flag any step that relies on a same-variance pairing, a non-tensorial ingredient, or a degenerate metric, so that the coordinate-independence of the final result can be argued for directly rather than assumed to follow automatically from the mere appearance of a repeated index.