Definition
A locally finite collection of nonnegative smooth functions on a topological manifold (or domain) subordinate to an open cover whose pointwise sum is identically one; used to localize global constructions by weighting and gluing local data.
Principle
Principle
Use of smooth, locally supported cutoff functions that sum to one to transfer and combine local objects (functions, forms, metrics) into a single global object while preserving smoothness and locality.
Demonstration
Demonstration
On a smooth manifold, choose a locally finite open cover, refine it if necessary, and construct smooth bump functions supported in each refined set so that their sum equals one; then glue a family of local sections of a vector bundle into a global section by multiplying each local section by the corresponding bump and summing.
Misapplication
Misapplication
Employing a family of functions that are not locally finite or not smooth (e.g., infinite overlapping supports with nonconvergent sums, or only measurable cutoffs) so the resulting sum fails to be smooth or well-defined, or attempting to subordinate a partition to a cover that is not refined or paracompact.
Consequence
Consequence
Allows extension of local constructions to global ones without losing regularity; enables localization arguments (integration by parts, construction of global metrics or connections) and the patching of solutions to PDEs or sections of bundles.
Reversal
Reversal
Replacing the requirement that the sum equals one by requiring the sum equals zero or an arbitrary function destroys the normalization that ensures a faithful convex combination of local data; without the unit-sum constraint one cannot guarantee invariance of glued objects under changes of local choices.
Boundary
Boundary
Requires a paracompact topological space for existence in the smooth category; the construction presupposes sufficient regularity (smooth, C^k, or continuous) and typically fails on non-paracompact spaces or when only discrete, nonlocally-finite covers are available.
Semantic Tension
Semantic Tension
Differs from partitions of a set into disjoint characteristic functions: partitions of unity use overlapping, smooth weights summing to one rather than mutually exclusive indicator functions, trading exclusivity for regularity and localization control.
Synthesis
Synthesis
A partition of unity is a smooth, locally finite weighted cover whose normalized sum equals one and that lets one coherently glue local smooth data into globally defined smooth objects on paracompact manifolds.