02.13.07 · analysis / pde

Rectifiable currents

shipped3 tiersLean: none

Anchor (Master): Federer, Geometric Measure Theory (Springer 1969), §4.1; Simon, Lectures on Geometric Measure Theory (ANU 1983), §26-37; Whitney, Geometric Integration Theory (Princeton UP 1957), Ch. V-XI; de Rham, Variétés différentiables (Hermann 1955); Almgren, De Lellis-Spadaro regularity programme

Intuition Beginner

Imagine you want to talk about a curved surface, a wire bent into a loop, or a soap film, all in one language. You also want that language to handle messy objects: a surface with creases, a wire that doubles back on itself, a film that has two sheets sitting on top of each other. The usual idea of a smooth surface is too clean for this. A current is a more flexible kind of object that still behaves like a surface but tolerates the mess.

The trick is to describe a surface not by listing its points, but by what it does. A surface knows how to be integrated over. If you hand it a little measuring gadget that assigns a number to each tiny oriented patch, the surface adds up all those numbers across itself and reports a total. A current is defined to be exactly this reporting machine: give it a gadget, it gives you back a number. Two different-looking surfaces that report the same totals for every gadget count as the same current.

Why bother with such an abstract definition? Because once a surface is just a machine that eats gadgets and returns numbers, you can take limits of surfaces easily. A sequence of crinklier and crinklier surfaces might not settle down to any nice surface, but the machines can still settle down to a limiting machine. That limit is a current, even when it is not a classical surface. This is what lets you prove that a soap film of least area actually exists: you take a sequence of films getting closer to the smallest area, and the limit current is the answer.

The currents we care about most are the rectifiable ones. A rectifiable current is built from a set that, apart from a negligible remainder, is covered by countably many smooth pieces, together with two extra pieces of data at each point: an orientation telling you which way the patch faces, and a multiplicity, a whole number counting how many sheets lie there. A double-layer soap film has multiplicity two. Adding the multiplicity is what lets the machine keep track of cancellation, because two sheets facing opposite ways can cancel to nothing.

Visual Beginner

Picture an oriented disk in space, like a flat circular membrane with a chosen front face marked by little arrows poking out of it. To turn this disk into a current, imagine a tiny windsock placed at every point: a gadget that measures how much of a flow passes through the oriented patch there. The disk-current eats the field of windsocks and returns the total flux through the whole disk. The current is the rule that does this for every possible windsock field.

The middle panel shows the most important feature: every current has a boundary, which is another current one dimension lower. The boundary of an oriented disk is its rim, the oriented circle. The boundary of a curve is its two endpoints, counted with signs. The rule for the boundary is set up so that the classical Stokes theorem, which relates an integral over a region to an integral over its edge, holds automatically. You do not prove Stokes for currents; you build the boundary so that Stokes is true by definition.

The right panel shows multiplicity and cancellation. If you stack two copies of the same oriented disk, you get a current of multiplicity two: every gadget reports twice the total. If instead you stack two copies facing opposite ways, every gadget reports a number and its negative, which sum to zero, so the two sheets cancel to the empty current. Multiplicity is a signed count of sheets, and the sign is the orientation. This signed bookkeeping is exactly what classical surfaces cannot do and what makes currents the right tool for least-area problems, where competing sheets must be allowed to cancel.

Worked example Beginner

We compute the boundary of a simple oriented surface and see the Stokes rule in action, with concrete numbers. Take the flat unit square in the plane with corners at $(0, 0)$ , $(1, 0)$ , $(1, 1)$ , $(0, 1)$ , oriented so that its front face points up out of the page. Call this surface-current $T$ . We will find its boundary current and check the total signed edge length.

Step 1. Identify the edges. The square has four edges: the bottom from $(0, 0)$ to $(1, 0)$ , the right from $(1, 0)$ to $(1, 1)$ , the top from $(1, 1)$ to $(0, 1)$ , and the left from $(0, 1)$ to $(0, 0)$ . Each edge has length $1$ .

Step 2. Orient the boundary consistently. The boundary rule for an upward-facing region gives the counterclockwise direction around the rim. So the boundary runs bottom (rightward), then right (upward), then top (leftward), then left (downward), returning to the start. Each edge is traversed exactly once in this single consistent loop.

Step 3. Read off the boundary current. The boundary is the oriented closed loop of four unit edges, each with multiplicity $1$ , traversed counterclockwise. Its total length is $1 + 1 + 1 + 1 = 4$ . This loop is itself a one-dimensional rectifiable current.

Step 4. Check the cancellation when two squares meet. Place a second identical square to the right, sharing the right edge of the first. The first square's boundary runs up that shared edge; the second square's boundary runs down it. The two contributions on the shared edge are $+ 1$ and $- 1$ , which cancel. The boundary of the combined two-square rectangle is just its outer rim of length $6$ , not $8$ , because the interior shared edge cancelled.

What this tells us: the boundary of a rectifiable current is computed edge by edge with a consistent orientation, and opposite-facing pieces cancel automatically. The cancellation in Step 4 is the whole reason currents use signed multiplicities. When many small oriented pieces are glued, all interior boundaries cancel and only the true outer boundary survives, which is the content of the Stokes rule built into the definition.

Check your understanding Beginner

Exercise (easy, multiple choice).

A current is best described as which kind of object?

A. A finite list of points in space B. A machine that takes a measuring gadget (a differential form) and returns a number C. A single smooth function defined on all of space D. A solid region with positive volume

Hint

The definition describes a surface by what it does when you integrate over it, not by listing its points.

Answer

B. A machine that takes a measuring gadget and returns a number. A current is a linear rule that eats a differential form (the measuring gadget) and reports the total of that form over the current. This is the reporting-machine picture from the intuition. Feedback-correct: describing a surface by what it integrates is exactly what makes limits of surfaces well-behaved. Feedback-wrong: a current is not a point list (A), not a single function (C), and need not be a solid region (D) since it can be a curve or a surface of any dimension.

Formal definition Intermediate+

Fix integers $0 \leq k \leq n$ . Let $D^{k} = D^{k} (R^{n})$ denote the space of smooth differential $k$ -forms with compact support, topologised by uniform convergence of all derivatives on compacta. A $k$ -current is a continuous linear functional $T : D^{k} \to R$ . The space of $k$ -currents is written $D_{k} = D_{k} (R^{n})$ , the topological dual of $D^{k}$ ^{[de Rham 1955]}. The case $k = 0$ recovers the distributions, since $D^{0} = D$ ; currents are the differential-form-valued generalisation of distributions.

At each point the comass of a $k$ -covector $ω \in Λ^{k} (R^{n})^{*}$ is $∥ ω ∥ = sup {⟨ ω, ξ ⟩ : ξ a unit simple k -vector}$ , where a simple $k$ -vector is a wedge $ξ = v_{1} \land \dots \land v_{k}$ and "unit" means the associated parallelepiped has $k$ -volume $1$ . The mass of a current is $M (T) = sup {T (ω) : ω \in D^{k}, ∥ ω (x) ∥ \leq 1 for all x} .$ A current of finite mass extends to a continuous functional on continuous compactly-supported forms and is represented, by Riesz representation, as $T (ω) = \int ⟨ ω, T ⟩ d ∥ T ∥$ for a Radon measure $∥ T ∥$ (the mass measure) and a $∥ T ∥$ -measurable unit simple $k$ -vector field $T$ (the orientation) ^{[Federer 1969 §4.1.7]}.

The boundary of a $k$ -current $T$ (for $k \geq 1$ ) is the $(k - 1)$ -current $\partial T$ defined by $⟨ \partial T, ω ⟩ = ⟨ T, d ω ⟩ for all ω \in D^{k - 1},$ so that the Stokes theorem holds by fiat. From $d \circ d = 0$ it follows that $\partial \circ \partial = 0$ . For $k = 0$ one sets $\partial T = 0$ .

A current $T$ is $k$ -rectifiable if there is an $H^{k}$ -measurable countably $k$ -rectifiable set $M \subseteq R^{n}$ (a set that, up to an $H^{k}$ -null remainder, is contained in a countable union of $C^{1}$ embedded $k$ -submanifolds), an $H^{k}$ -measurable orientation $ξ$ assigning to $H^{k}$ -a.e. $x \in M$ a unit simple $k$ -vector spanning the approximate tangent plane $T_{x} M$ , and a locally $H^{k}$ -integrable nonnegative multiplicity $θ : M \to [0, \infty)$ , such that $T (ω) = \int_{M} ⟨ ω (x), ξ (x)⟩ θ (x) d H^{k} (x) (ω \in D^{k}) .$ Then $M (T) = \int_{M} θ d H^{k}$ and $∥ T ∥ = θ H^{k} └ M$ . The current is integer-multiplicity rectifiable (written $T \in R_{k}$ ) when $θ$ takes values in $Z_{\geq 0}$ at $H^{k}$ -a.e. point ^{[Simon 1983 §27]}.

A current $T$ is integral ( $T \in I_{k}$ ) when $T$ is integer-multiplicity rectifiable, $M (T) < \infty$ , $M (\partial T) < \infty$ , and $\partial T$ is itself integer-multiplicity rectifiable. A current with $M (T) < \infty$ and $M (\partial T) < \infty$ (without the rectifiability or integrality hypotheses) is normal ( $T \in N_{k}$ ). Thus $I_{k} \subseteq R_{k} \cap N_{k}$ .

Counterexamples to common slips Intermediate+

Rectifiable is not the same as integral. A rectifiable current may have a boundary of infinite mass, or a boundary that fails to be rectifiable. Integrality is the stronger condition that both $T$ and $\partial T$ are integer-multiplicity rectifiable with finite mass. The unit disk with real multiplicity $θ = π$ is rectifiable but not integer-multiplicity; a rectifiable current whose boundary spreads over a non-rectifiable set is rectifiable but not integral.
Mass is not the dual norm of the supremum norm on forms; it uses comass. The pointwise bound in the definition of $M$ is $∥ ω (x) ∥ \leq 1$ in the comass norm on covectors, not the Euclidean norm on coefficients. Comass and Euclidean norms differ already for simple wedges in dimension $n \geq 4$ (the Euclidean norm of $ω$ can exceed its comass), so using the wrong norm gives the wrong mass.
The boundary of a rectifiable current need not be rectifiable. Defining $\partial T$ by $⟨ \partial T, ω ⟩ = ⟨ T, d ω ⟩$ always produces a current, but that current can have infinite mass or fail to concentrate on a rectifiable set. The structure theorem is exactly the statement that a finite-mass-boundary plus integer multiplicity forces rectifiability of $\partial T$ ; without finite boundary mass nothing is guaranteed.
Multiplicity $θ$ and orientation $ξ$ are separate data and cannot be merged into a signed scalar in dimension $k \geq 2$ . For curves ( $k = 1$ ) one can fold the sign of the orientation into the multiplicity, but for higher $k$ the orientation is a unit simple $k$ -vector carrying directional information that a scalar cannot encode. Treating multiplicity as a signed scalar loses the tangent-plane orientation.

Key theorem with proof Intermediate+

Theorem (Boundary-rectifiability / structure theorem; Federer-Fleming 1960, Federer 1969 §4.1.28). Let $T \in R_{k}$ be an integer-multiplicity rectifiable $k$ -current with $M (T) < \infty$ , and suppose $M (\partial T) < \infty$ . Then $\partial T$ is integer-multiplicity rectifiable; consequently $T$ is an integral current. Equivalently: a current of finite mass and finite boundary mass that is integer-multiplicity rectifiable has integer-multiplicity rectifiable boundary.

Proof. The proof localises the boundary, slices the current by hyperplanes to reduce the dimension, and applies the one-dimensional case as the base.

Step 1 (reduction to a base case by slicing). For a unit vector $e$ and $t \in R$ , write $u (x) = ⟨ x, e ⟩$ and define the slice $⟨ T, u, t ⟩ = \partial (T └ {u < t}) - (\partial T) └ {u < t}$ , the boundary contribution of $T$ cut at the level set ${u = t}$ . The coarea inequality for the slicing of rectifiable currents gives, for $L^{1}$ -a.e. $t$ , that $⟨ T, u, t ⟩$ is an integer-multiplicity rectifiable $(k - 1)$ -current supported in ${u = t}$ , with $\int_{R} M (⟨ T, u, t ⟩) d t \leq M (T),$ and the multiplicity of the slice equals the restriction of $θ$ to the level set, hence is integer-valued $H^{k - 1}$ -a.e. ^{[Federer 1969 §4.3]}. Slicing thus turns the $k$ -dimensional integrality question into a family of $(k - 1)$ -dimensional ones, and an induction on $k$ reduces the theorem to the case $k = 1$ , where $\partial T$ is a $0$ -current.

Step 2 (the base case $k = 1$ ). Let $T \in R_{1}$ have $M (T) < \infty$ and $M (\partial T) < \infty$ . By the structure theory of one-dimensional rectifiable sets, $T$ decomposes as an at-most-countable sum of oriented Lipschitz curves with integer multiplicities, $T = \sum_{j} θ_{j} [[γ_{j}]]$ , converging in mass. The boundary of each oriented arc $[[γ_{j}]]$ is $[[γ_{j} (1)]] - [[γ_{j} (0)]]$ , a difference of point masses. Hence $\partial T = \sum_{j} θ_{j} ([[γ_{j} (1)]] - [[γ_{j} (0)]])$ is a sum of integer multiples of point masses. Grouping coincident endpoints, $\partial T = \sum_{p} m_{p} [[p]]$ with $m_{p} \in Z$ , and $M (\partial T) = \sum_{p} ∣ m_{p} ∣ < \infty$ forces the sum to have finitely many nonzero terms with integer weights. Thus $\partial T$ is integer-multiplicity rectifiable of dimension $0$ .

Step 3 (lifting the base case through the slices). Returning to general $k$ , fix $T \in R_{k}$ with finite mass and boundary mass. For each coordinate direction $e_{i}$ the slices $⟨ T, u_{i}, t ⟩$ are integer-multiplicity rectifiable for a.e. $t$ , by the inductive hypothesis applied in the level hyperplane. A current is integer-multiplicity rectifiable if and only if almost every slice in every direction is integer-multiplicity rectifiable and the slice multiplicities are consistent — this is the slicing characterisation of rectifiability ^{[Simon 1983 §28]}. Applying the characterisation to $\partial T$ : the slices of $\partial T$ are, up to sign, boundaries of slices of $T$ , namely $⟨ \partial T, u_{i}, t ⟩ = - \partial ⟨ T, u_{i}, t ⟩$ , and each $⟨ T, u_{i}, t ⟩$ is integer-multiplicity rectifiable of dimension $k - 1$ with finite boundary mass for a.e. $t$ (the latter from the coarea bound applied to $\partial T$ ). By the inductive hypothesis at level $k - 1$ , each $\partial ⟨ T, u_{i}, t ⟩$ is integer-multiplicity rectifiable. Therefore almost every slice of $\partial T$ is integer-multiplicity rectifiable, and the slicing characterisation upgrades this to integer-multiplicity rectifiability of $\partial T$ itself.

Step 4 (conclusion). With $\partial T$ shown integer-multiplicity rectifiable and $M (\partial T) < \infty$ by hypothesis, the pair $(T, \partial T)$ satisfies every clause of the definition of an integral current, so $T \in I_{k}$ . $□$

Bridge. The structure theorem builds toward the entire existence-and-regularity edifice of geometric measure theory: it certifies that the class $I_{k}$ is closed under the boundary operator, which is what makes integral currents a chain complex and lets homological methods enter the calculus of variations. The slicing machinery used in the proof appears again in the coarea formula and in the construction of tangent cones at singular points of mass-minimisers. The integer-multiplicity hypothesis is precisely the ingredient that powers the Federer-Fleming compactness theorem stated in the Master tier, because integrality is preserved under flat limits exactly when boundaries stay integral. The deformation theorem of 02.13.05 supplies the approximation by polyhedral chains that turns these slice-by-slice statements into global compactness, and the same circle of ideas appears again in the regularity theory where slices are used to prove that singular sets of area-minimisers have small dimension.

Exercises Intermediate+

Exercise 1 (easy, symbolic).

Let $T$ be an oriented $C^{1}$ embedded $k$ -submanifold $M$ with multiplicity $1$ , so $T (ω) = \int_{M} ω$ . Show directly from the definition $⟨ \partial T, ω ⟩ = ⟨ T, d ω ⟩$ that $\partial T$ is integration over $\partial M$ with the induced orientation, recovering the classical Stokes theorem.

Hint

Apply the classical Stokes theorem for smooth manifolds to $\int_{M} d ω$ .

Answer

By definition $⟨ \partial T, ω ⟩ = ⟨ T, d ω ⟩ = \int_{M} d ω$ . The classical Stokes theorem for the compact-support form $ω$ on the oriented manifold-with-boundary $M$ gives $\int_{M} d ω = \int_{\partial M} ω$ , where $\partial M$ carries the boundary orientation induced by $M$ . Hence $⟨ \partial T, ω ⟩ = \int_{\partial M} ω$ for all $ω \in D^{k - 1}$ , which says exactly that $\partial T$ is the current of integration over $\partial M$ with multiplicity $1$ and the induced orientation. The current boundary is built to make Stokes hold, and on smooth manifolds it reproduces the classical statement.

Exercise 4 (medium, symbolic).

Let $T_{1}$ and $T_{2}$ be the same oriented $k$ -submanifold $M$ but with opposite orientations $ξ$ and $- ξ$ , each multiplicity $1$ . Show $T_{1} + T_{2} = 0$ as currents, and explain why this forces multiplicity and orientation to be tracked together.

Hint

Evaluate $(T_{1} + T_{2}) (ω)$ using $⟨ ω, - ξ ⟩ = - ⟨ ω, ξ ⟩$ .

Answer

For any $ω \in D^{k}$ , $(T_{1} + T_{2}) (ω) = \int_{M} ⟨ ω, ξ ⟩ d H^{k} + \int_{M} ⟨ ω, - ξ ⟩ d H^{k} = \int_{M} (⟨ ω, ξ ⟩ - ⟨ ω, ξ ⟩) d H^{k} = 0.$ So the two oppositely oriented sheets cancel to the zero current. This shows that the orientation field $ξ$ carries a sign that interacts with the multiplicity: a "double sheet" with opposite orientations has total multiplicity $1 + 1 = 2$ as an unsigned count of sheets but represents the zero current. The geometrically meaningful object is the signed combination $θ ξ$ , so multiplicity and orientation must be carried together; neither alone determines the current.

Exercise 5 (medium, symbolic).

Prove that mass is lower semicontinuous under weak-* convergence of currents: if $T_{j} \to T$ in the sense that $T_{j} (ω) \to T (ω)$ for every $ω \in D^{k}$ , then $M (T) \leq lim inf_{j} M (T_{j})$ .

Hint

Fix an admissible $ω$ with $∥ ω ∥ \leq 1$ and pass to the limit in $T_{j} (ω) \leq M (T_{j})$ .

Answer

Fix $ω \in D^{k}$ with $∥ ω (x) ∥ \leq 1$ for all $x$ . For each $j$ , $T_{j} (ω) \leq M (T_{j})$ by definition of mass. Letting $j \to \infty$ , the left side converges to $T (ω)$ and the right side has $lim inf$ equal to $lim inf_{j} M (T_{j})$ , so $T (ω) \leq lim inf_{j} M (T_{j})$ . Taking the supremum over all admissible $ω$ with $∥ ω ∥ \leq 1$ gives $M (T) = sup_{ω} T (ω) \leq lim inf_{j} M (T_{j})$ . Lower semicontinuity of mass is the property that makes the direct method work: a mass-minimising sequence has a limit whose mass does not jump up.

Exercise 6 (hard, symbolic).

Let $T \in R_{k}$ be integer-multiplicity rectifiable with constant multiplicity $θ \equiv 1$ on a closed $C^{1}$ submanifold without boundary, so $\partial T = 0$ . Using the deformation theorem of 02.13.05, show that for every $δ > 0$ there is an integral polyhedral $k$ -current $P$ with $F (T - P) < δ$ , where $F$ is the flat norm.

Hint

Apply the deformation theorem at grid scale $ε$ to $T$ , use the resulting decomposition $T - P = \partial R + Q$ , and bound the flat norm by the masses of $R$ and $Q$ .

Answer

The current $T$ is normal: $M (T) < \infty$ and $M (\partial T) = 0 < \infty$ . The deformation theorem at grid scale $ε$ yields a polyhedral $k$ -chain $P_{ε}$ on the $ε$ -grid skeleton and a decomposition $T - P_{ε} = \partial R_{ε} + Q_{ε}$ with $M (R_{ε}) \leq c ε M (T)$ and $M (Q_{ε}) \leq c ε M (\partial T) = 0$ . Since the multiplicities are integers and the projection preserves integrality, $P_{ε}$ is an integral polyhedral current. Choosing $S = R_{ε}$ in the flat-norm infimum $F (U) = in f_{S} {M (U - \partial S) + M (S)}$ with $U = T - P_{ε}$ gives $F (T - P_{ε}) \leq M (Q_{ε}) + M (R_{ε}) \leq c ε M (T)$ . Taking $ε < δ / (c M (T))$ makes $F (T - P_{ε}) < δ$ . Thus integral polyhedral currents are flat-dense in this class.

Exercise 7 (hard, symbolic).

Establish the constancy theorem in its simplest form: if $T$ is a $k$ -current in an open connected set $U \subseteq R^{k}$ (top dimension, $n = k$ ) with $\partial T = 0$ and $T$ representable by integration against an integer multiplicity, then $θ$ is constant on $U$ .

Hint

In top dimension a current is $T (ω) = \int_{U} θ ω$ for $ω = f d x^{1} \land \dots \land d x^{k}$ . The condition $\partial T = 0$ says $\int_{U} θ d f^{'} = 0$ for test functions, which is a distributional gradient condition on $θ$ .

Answer

In top dimension write $T (ω) = \int_{U} θ g d x$ where $ω = g d x^{1} \land \dots \land d x^{k}$ and $θ \in L_{l oc}^{1}$ is integer-valued. The boundary acts on $(k - 1)$ -forms $η$ ; choosing $η$ so that $d η = (\partial_{i} g) d x^{1} \land \dots \land d x^{k}$ ranges over all forms with $g \in C_{c}^{\infty} (U)$ , the condition $⟨ \partial T, η ⟩ = ⟨ T, d η ⟩ = 0$ becomes $\int_{U} θ \partial_{i} g d x = 0$ for every $g \in C_{c}^{\infty} (U)$ and each $i$ . This says the distributional partial derivatives of $θ$ vanish: $\partial_{i} θ = 0$ in $D^{'} (U)$ for all $i$ . A locally integrable function with vanishing distributional gradient on a connected open set is a.e. equal to a constant (mollify and use connectedness). Since $θ$ is integer-valued, the constant is an integer. Hence $θ$ is a constant integer on $U$ . The constancy theorem is what pins down the multiplicity of a boundaryless top-dimensional current from a single value.

Advanced results Master

The class $I_{k}$ of integral currents is closed under the boundary operator, by the structure theorem, and the resulting chain complex $(I_{∙}, \partial)$ computes the integral homology of reasonable spaces; this is the homological content that lets the calculus of variations be run in a prescribed boundary or homology class. The compactness theorem is the analytic core. Let $K \subseteq R^{n}$ be compact and $Λ < \infty$ . The set ${T \in I_{k} : spt T \subseteq K, M (T) + M (\partial T) \leq Λ}$ is sequentially compact in the flat-norm topology, and the flat limit of integral currents with uniformly bounded mass and boundary mass is again integral ^{[Federer-Fleming 1960]}. The proof couples the deformation theorem of 02.13.05 — which approximates each $T$ to flat-error $O (ε)$ by a polyhedral chain on the finite grid $G_{ε}^{(k)} \cap K$ , a finite-dimensional space where bounded sets are precompact — with the closure theorem, which guarantees the limit is integer-multiplicity rectifiable. The closure theorem in turn rests on the rectifiability criterion: a normal current whose slices are a.e. integer-multiplicity rectifiable is itself integer-multiplicity rectifiable, exactly the slicing characterisation used in the structure-theorem proof ^{[Federer 1969 §4.2.16]}.

Compactness plus lower semicontinuity of mass solves the Plateau problem in the integral-current formulation. Given an integral $(k - 1)$ -cycle $B$ with $\partial B = 0$ and a compact convex $K \supseteq spt B$ , the variational problem $in f {M (T) : T \in I_{k}, \partial T = B, spt T \subseteq K}$ has a minimiser: take a minimising sequence, extract a flat-convergent subsequence by compactness, observe the boundary condition passes to the limit (the boundary operator is flat-continuous), and use lower semicontinuity of mass to conclude the limit realises the infimum ^{[Federer-Fleming 1960]}. The minimiser is a mass-minimising integral current spanning $B$ , the higher-dimensional and higher-codimensional generalisation of the soap film.

The constancy theorem in full generality states that a $k$ -current $T$ in a connected open subset of a $k$ -dimensional $C^{1}$ submanifold $M$ with $\partial T = 0$ (relative to $M$ ) equals $θ_{0} [[M]]$ for a single constant $θ_{0}$ ; it is the rigidity that pins down the multiplicity of a cycle from local data and underlies the uniqueness statements in regularity theory ^{[Simon 1983 §26]}. Slicing refines all of this: for a Lipschitz $f : R^{n} \to R^{m}$ and a.e. $y$ , the slice $⟨ T, f, y ⟩$ is an integral current of dimension $k - m$ , with the coarea bound $\int^{*} M (⟨ T, f, y ⟩) d y \leq (Lip f)^{m} M (T)$ and the compatibility $\partial ⟨ T, f, y ⟩ = (- 1)^{m} ⟨ \partial T, f, y ⟩$ ^{[Federer 1969 §4.3]}.

Currents are not the only weak surfaces. Varifolds — Radon measures on the Grassmann bundle of unoriented tangent planes — discard orientation and therefore admit no boundary operator and no cancellation; they are the natural setting for unoriented or non-cancelling problems such as the Allard regularity theory and Brakke's mean-curvature flow, where two sheets meeting do not annihilate. The choice between currents and varifolds is the choice of whether oriented cancellation is a feature (Plateau, where a minimiser may have lower mass through cancellation) or a defect (soap films that genuinely have triple junctions, where varifolds and the size functional are more faithful). Almgren's interior-regularity theorem, that a mass-minimising integral $k$ -current in $R^{n}$ has singular set of Hausdorff dimension at most $k - 2$ , is the deepest structural result; its modern account is the De Lellis-Spadaro programme, which rebuilt Almgren's $Q$ -valued-function machinery and centre-manifold construction ^{[Almgren 2000]} ^{[De Lellis-Spadaro 2014]}.

Synthesis. Rectifiable and integral currents form the variational completion of the class of oriented submanifolds, and their three defining features — duality with forms, the Stokes-by-definition boundary, and integer multiplicity — each carry a distinct structural load. Duality builds toward the weak-* compactness that the direct method needs; the boundary operator builds toward the chain-complex structure that lets homology constrain minimisers; integer multiplicity builds toward the closure theorem, since integrality is what survives flat limits and prevents minimising sequences from leaking into diffuse non-rectifiable mass. The deformation theorem of 02.13.05 connects to this material as the engine of compactness, converting bounded mass into finite-dimensional grid approximation; slicing connects forward to the coarea formula and to the dimension-reduction arguments that bound singular sets; and the constancy theorem connects to the uniqueness-of-tangent-cone questions at the heart of regularity. The whole apparatus appears again in the regularity theory of minimal surfaces, where the singular set is controlled by Almgren's stratification and the De Lellis-Spadaro centre manifold, and it connects laterally to the varifold formulation whenever orientation and cancellation are the wrong bookkeeping for the geometry at hand.

Full proof set Master

Proposition (mass is the total variation of the mass measure). Let $T \in D_{k}$ have $M (T) < \infty$ . Then there is a Radon measure $∥ T ∥$ and a $∥ T ∥$ -measurable unit simple $k$ -vector field $T$ with $T (ω) = \int ⟨ ω, T ⟩ d ∥ T ∥$ , and $M (T) = ∥ T ∥ (R^{n})$ .

Proof. The functional $T$ is linear on $D^{k}$ and, by $M (T) < \infty$ , bounded in the comass norm: $∣ T (ω) ∣ \leq M (T) sup_{x} ∥ ω (x) ∥$ . Hence $T$ extends by density to a bounded linear functional on the space $C_{c}^{0} (R^{n}; Λ^{k})$ of continuous compactly-supported $k$ -covector fields with the comass-sup norm. By the Riesz representation theorem for vector-valued functionals, there is a Radon measure $μ$ and a $μ$ -measurable field $T$ of $k$ -vectors with $∣ T ∣ \leq 1$ (comass-dual norm) such that $T (ω) = \int ⟨ ω, T ⟩ d μ$ and $μ$ is the total-variation measure, so $μ (R^{n}) = ∥ T ∥_{op} = M (T)$ . A polar-decomposition step replaces $T$ by a unit simple $k$ -vector field and absorbs its length into $μ$ , giving the stated $∥ T ∥ = μ$ . The supremum defining $M (T)$ is attained in the limit by forms aligning with $T$ , so $M (T) = ∥ T ∥ (R^{n})$ . $□$

Proposition (lower semicontinuity of mass). If $T_{j} \to T$ weakly-* (that is, $T_{j} (ω) \to T (ω)$ for all $ω \in D^{k}$ ), then $M (T) \leq lim inf_{j} M (T_{j})$ .

Proof. Let $ω \in D^{k}$ with $sup_{x} ∥ ω (x) ∥ \leq 1$ . For each $j$ , $T_{j} (ω) \leq M (T_{j})$ . Passing to the limit, $T (ω) = lim_{j} T_{j} (ω) \leq lim inf_{j} M (T_{j})$ . Taking the supremum over admissible $ω$ yields $M (T) \leq lim inf_{j} M (T_{j})$ , since the supremum of a family of quantities each bounded by $lim inf_{j} M (T_{j})$ is bounded by the same quantity. $□$

Proposition (boundary is weak- continuous).* The boundary operator $\partial : D_{k} \to D_{k - 1}$ is continuous for the weak-* topology: $T_{j} \to T$ implies $\partial T_{j} \to \partial T$ .

Proof. For any $ω \in D^{k - 1}$ , the form $d ω \in D^{k}$ is a fixed test form. By definition $⟨ \partial T_{j}, ω ⟩ = ⟨ T_{j}, d ω ⟩ \to ⟨ T, d ω ⟩ = ⟨ \partial T, ω ⟩$ , the convergence being the assumed weak-* convergence of $T_{j}$ evaluated at the single form $d ω$ . Since this holds for every $ω$ , $\partial T_{j} \to \partial T$ weakly-*. $□$

Proposition (existence of a mass-minimiser; Plateau). Let $B \in I_{k - 1}$ with $\partial B = 0$ and $spt B \subseteq K$ for a compact convex $K$ . Assume the admissible class $A = {T \in I_{k} : \partial T = B, spt T \subseteq K}$ is nonempty. Then $in f_{T \in A} M (T)$ is attained.

Proof. Let $m = in f_{A} M$ and pick a minimising sequence $T_{j} \in A$ with $M (T_{j}) \to m$ . Each $T_{j}$ has $\partial T_{j} = B$ , so $M (\partial T_{j}) = M (B)$ is a fixed finite constant, and $M (T_{j}) \leq m + 1$ for large $j$ . Thus $M (T_{j}) + M (\partial T_{j}) \leq Λ$ uniformly, and $spt T_{j} \subseteq K$ compact. By the Federer-Fleming compactness theorem there is a subsequence (not relabelled) converging in flat norm to some $T \in I_{k}$ with $spt T \subseteq K$ . Flat convergence implies weak-* convergence, so by weak-* continuity of $\partial$ , $\partial T = lim_{j} \partial T_{j} = B$ , placing $T \in A$ . By lower semicontinuity of mass, $M (T) \leq lim inf_{j} M (T_{j}) = m$ . Since $T \in A$ forces $M (T) \geq m$ , equality holds and $T$ is a minimiser. $□$

Proposition (constancy theorem, top-dimensional case). Let $U \subseteq R^{k}$ be open and connected and let $T \in D_{k} (U)$ be representable by an integer-valued $θ \in L_{l oc}^{1} (U)$ , $T (ω) = \int_{U} θ g d x$ for $ω = g d x^{1} \land \dots \land d x^{k}$ . If $\partial T = 0$ , then $θ$ is a.e. equal to a constant integer.

Proof. The condition $\partial T = 0$ means $⟨ T, d η ⟩ = 0$ for all $η \in D^{k - 1} (U)$ . Take $η = h d x^{1} \land \dots \land d x^{i} \land \dots \land d x^{k}$ with $h \in C_{c}^{\infty} (U)$ ; then $d η = \pm (\partial_{i} h) d x^{1} \land \dots \land d x^{k}$ , so $0 = ⟨ T, d η ⟩ = \pm \int_{U} θ \partial_{i} h d x$ . Hence $\int_{U} θ \partial_{i} h d x = 0$ for all $h \in C_{c}^{\infty} (U)$ and all $i$ , that is, $\partial_{i} θ = 0$ in $D^{'} (U)$ for every $i$ . Mollifying, $θ_{ε} = θ * ρ_{ε}$ satisfies $\nabla θ_{ε} = 0$ on the set where it is defined, so $θ_{ε}$ is locally constant; on the connected set $U$ it is globally constant, and $θ_{ε} \to θ$ in $L_{l oc}^{1}$ forces $θ$ to equal that constant a.e. Because $θ$ is integer-valued, the constant is an integer. $□$

Connections Master

The mass measure $∥ T ∥ = θ H^{k} └ M$ and the Hausdorff-measure representation of a rectifiable current rest directly on the Carathéodory outer-measure construction of 02.07.02: the multiplicity-weighted Hausdorff measure is the mass measure, and the measurability of the orientation field is a Lusin-type consequence of the rectifiable-set structure built there.

The deformation theorem of 02.13.05 is the analytic engine for the compactness theorem stated here: it converts the uniform mass-plus-boundary-mass bound into approximation by polyhedral chains on a finite grid skeleton, and the polyhedral approximants live in a finite-dimensional space where bounded sequences are precompact, which is exactly the precompactness that the direct method needs.

The Stokes-by-definition boundary $⟨ \partial T, ω ⟩ = ⟨ T, d ω ⟩$ and the resulting chain complex are the geometric-measure-theoretic counterpart of the de Rham complex underlying the fundamental-solution and potential theory of 02.13.02: solving $\partial R = T$ for an integral filling $R$ is the homological analogue of solving $Δ u = f$ for a Newtonian potential, with mass-minimisation replacing the energy minimisation that selects the potential.

Historical & philosophical context Master

Georges de Rham introduced currents in his 1955 monograph Variétés différentiables as the duals of compactly-supported differential forms, unifying the chains of algebraic topology and the distributions of Schwartz into a single object that could both bound and be integrated against ^{[de Rham 1955]}. Hassler Whitney's 1957 Geometric Integration Theory developed the parallel theory of flat chains and the flat norm, supplying the polyhedral-approximation technology and the deformation theorem ^{[Whitney 1957]}. The decisive synthesis came in Herbert Federer and Wendell Fleming's 1960 Annals of Mathematics paper, which isolated the classes of normal and integral currents, proved the closure and compactness theorems, and thereby solved the Plateau problem of least-area surfaces in arbitrary dimension and codimension — the problem that had resisted the parametric methods of Douglas and Radó beyond the case of disk-type surfaces ^{[Federer-Fleming 1960]}. Federer's 1969 Geometric Measure Theory gave the encyclopedic treatment, with the structure theorem at §4.1.28 and the slicing apparatus at §4.3 ^{[Federer 1969]}. Leon Simon's 1983 ANU lectures became the standard graduate route into the subject and the regularity theory ^{[Simon 1983]}. The interior regularity of mass-minimisers, that the singular set has codimension at least two, was established by Frederick Almgren in his thousand-page manuscript using $Q$ -valued functions and a centre-manifold construction ^{[Almgren 2000]}; Camillo De Lellis and Emanuele Spadaro rebuilt and clarified this programme in a sequence of papers beginning in 2014 ^{[De Lellis-Spadaro 2014]}.

Bibliography Master

@book{derham1955varietes,
  author    = {de Rham, Georges},
  title     = {Vari\'et\'es diff\'erentiables: formes, courants, formes harmoniques},
  publisher = {Hermann},
  series    = {Actualit\'es Scientifiques et Industrielles},
  number    = {1222},
  year      = {1955},
  address   = {Paris}
}

@book{whitney1957geometric,
  author    = {Whitney, Hassler},
  title     = {Geometric Integration Theory},
  publisher = {Princeton University Press},
  series    = {Princeton Mathematical Series},
  volume    = {21},
  year      = {1957},
  address   = {Princeton, NJ}
}

@article{federerfleming1960normal,
  author  = {Federer, Herbert and Fleming, Wendell H.},
  title   = {Normal and integral currents},
  journal = {Annals of Mathematics},
  volume  = {72},
  number  = {3},
  pages   = {458--520},
  year    = {1960},
  doi     = {10.2307/1970227}
}

@book{federer1969geometric,
  author    = {Federer, Herbert},
  title     = {Geometric Measure Theory},
  publisher = {Springer-Verlag},
  series    = {Die Grundlehren der mathematischen Wissenschaften},
  volume    = {153},
  year      = {1969},
  address   = {Berlin}
}

@book{simon1983lectures,
  author    = {Simon, Leon},
  title     = {Lectures on Geometric Measure Theory},
  publisher = {Centre for Mathematical Analysis, Australian National University},
  series    = {Proceedings of the Centre for Mathematical Analysis},
  volume    = {3},
  year      = {1983},
  address   = {Canberra}
}

@book{morgan2016geometric,
  author    = {Morgan, Frank},
  title     = {Geometric Measure Theory: A Beginner's Guide},
  edition   = {5},
  publisher = {Academic Press},
  year      = {2016},
  address   = {Amsterdam}
}

@book{almgren2000big,
  author    = {Almgren, Frederick J.},
  title     = {Almgren's Big Regularity Paper},
  publisher = {World Scientific},
  series    = {World Scientific Monograph Series in Mathematics},
  volume    = {1},
  year      = {2000},
  address   = {Singapore},
  note      = {Edited by V. Scheffer and J. E. Taylor}
}

@article{delellisspadaro2014multiple,
  author  = {De Lellis, Camillo and Spadaro, Emanuele},
  title   = {Multiple valued functions and integral currents},
  journal = {Geometric and Functional Analysis},
  volume  = {24},
  number  = {6},
  pages   = {1831--1884},
  year    = {2014},
  doi     = {10.1007/s00039-014-0306-3}
}

Prerequisites

02.13.05
02.07.02

Tier anchors

beginner: Morgan, Geometric Measure Theory: A Beginner's Guide, 5e (Academic Press 2016), Ch. 1-4; physics-anchored oriented-surface and weighted-multiplicity pictures
intermediate: Morgan, Geometric Measure Theory: A Beginner's Guide, 5e (Academic Press 2016), Ch. 4-6, 8; Simon, Lectures on Geometric Measure Theory (ANU 1983), §26-31
master: Federer, Geometric Measure Theory (Springer 1969), §4.1; Simon, Lectures on Geometric Measure Theory (ANU 1983), §26-37; Whitney, Geometric Integration Theory (Princeton UP 1957), Ch. V-XI; de Rham, Variétés différentiables (Hermann 1955); Almgren, De Lellis-Spadaro regularity programme

References

Federer — Geometric Measure Theory · Springer Grundlehren 153 (1969), §4.1 (rectifiable and integral currents), §4.1.28 (structure/closure), §4.2.17 (compactness)
Federer-Fleming — Normal and integral currents · Annals of Mathematics 72 (1960), 458-520; integral currents, closure, compactness, Plateau problem
de Rham — Variétés différentiables · Hermann, Paris (1955); currents as the dual of compactly-supported differential forms
Whitney — Geometric Integration Theory · Princeton University Press (1957), Ch. V-XI; chains, cochains, the flat norm, geometric integration
Simon — Lectures on Geometric Measure Theory · Proceedings of the Centre for Mathematical Analysis, ANU 3 (1983), §26-37
Morgan — Geometric Measure Theory: A Beginner's Guide, 5e · Academic Press (2016), Ch. 4-6, 8
Almgren — Almgren's Big Regularity Paper · World Scientific Monograph Series 1 (2000); Q-valued functions and interior regularity of mass-minimisers
De Lellis-Spadaro — Regularity of area-minimizing currents · Geom. Funct. Anal. 24 (2014) and Ann. of Math. 183 (2016); modern revisitation of Almgren's regularity theory

Estimated time

beginner: 25m
intermediate: 65m
master: 120m