Metric Currents, Flat Chains & Nonsmooth Geometric Measure Theory

Begin C01.01 with the geometric question, not the notation: how can “Oriented cells act rather than merely occupy” retain orientation, multiplicity, boundary, and cancellation when the ambient space may have no coordinates or differential forms? In plain language, an oriented piece returns a signed number when tested, so opposite orientations can cancel even on the same carrier. The learner first identifies the object being acted on and the quantity being returned.

Read the types aloud. X is the declared complete or local metric space; T is a k-dimensional current; (f,π₁,…,π_k) is a Lipschitz test tuple; ||T|| is the mass measure; M(T) is total mass when finite; ∂T is the boundary; φ#T is pushforward; C01.01 is the audit label. A current is a multilinear functional, not a subset, parametrization, positive measure, or ordinary vector. Its support is derived from its mass behavior; its mass forgets orientation; its boundary lowers dimension.

The local teaching meaning is “an oriented piece returns a signed number when tested, so opposite orientations can cancel even on the same carrier” inside the unit focus “signed functionals, Lipschitz probes, dimension, orientation, multiplicity, and the difference between a geometric carrier and an algebraic action.” List the input metric, dimension, coefficient group, orientation convention, multiplicity, test regularity, support condition, and whether finiteness is global or only local.

Keep five ledgers separate. The action ledger evaluates signed test tuples. The mass ledger is positive and bounds action. The boundary ledger records algebraic faces or endpoints. The support ledger records where every neighborhood carries mass. The convergence ledger says weak, flat, intrinsic flat, Hausdorff, or another declared mode.

The finite record is On X={a,b,c}, let the signed 0-chain have weights (1,-2,1) and let f(a,b,c)=(1,2,3). Its declared action is 1·1+-2·2+1·3=0; replacing the signed weights by absolute values would compute mass-like data, not the current action. Recompute the signed action before taking absolute values. Then recompute mass and boundary independently. This order exposes orientation cancellation that a picture of the support cannot show.

The bounded theorem route is: finite oriented Lipschitz chains define multilinear local functionals; in dimension zero the action is a signed weighted evaluation, and orientation reversal changes action without changing mass. Mark the exact step using completeness, finite mass, local compactness, rectifiability, integer multiplicity, boundary-mass control, a cone inequality, tightness, or a common ambient space. None of these hypotheses comes from the symbol T alone.

Test-tuple continuity is sequential and typed: change one Lipschitz coordinate probe while retaining uniform Lipschitz control, and distinguish that operation from weak convergence of a sequence of currents. Locality removes a test when one probe is constant near the support of its coefficient; it is not a vague statement that distant geometry never matters.

Boundary and pushforward are algebraic operations before they become geometric pictures. Verify ∂∂T=0 and ∂(φ#T)=φ#(∂T) only when every action is defined. A nonproper map, lost local finiteness, or unsupported restriction can make the displayed expression unavailable.

Rectifiability says that mass is carried by countably many Lipschitz images up to a null set; integer rectifiability also controls multiplicity. It does not mean the support is a smooth manifold, embedded without overlap, free of branch points, or uniquely parametrized.

Flat comparison permits both a same-dimensional error and a one-higher-dimensional filling. A small flat value can hide thin cancellation or discarded volume, while a small Hausdorff distance can preserve support that currents cancel. Never replace one convergence label by another without a theorem.

Execute the required mutation: replace a current by its support, treat a positive measure as oriented, omit the test coefficient, or infer smooth tangent planes from a finite chain. Stop at the first missing definition or theorem hypothesis, state the strongest valid remainder, and keep an explicit nonclaim about smoothness, uniqueness, optimality, topology, or physical interpretation.

For C01.01, write a dependency card with columns datum, type, operation, sign convention, mass bound, boundary term, convergence mode, compactness input, and first failure. A formula is not understood until each column can be spoken as a complete sentence.

Reconstruct the finite row by a second route: subdivide a chain, reverse one orientation, push through an explicit Lipschitz map, fill by named cells, or reduce coefficients. Compare exact signed action, mass, and boundary rather than only the final scalar.

Separate theorem proof from numerical evidence. A mesh may certify an algebraic boundary, an admissible filling upper bound, or a mutation trace. It cannot by itself prove countable rectifiability, closure, lower semicontinuity, compactness, existence of a minimizing filling, or equality of intrinsic distances.

Finish by changing exactly one item—orientation, multiplicity, coefficient group, ambient embedding, boundary control, compactness, or locality. Recompute all five ledgers and explain which conclusion survives. This one-variable mutation prevents C01.01 from becoming a memorized but untyped recipe.