Smirnov-Type Decomposition

• Smirnov-Type Decomposition is a framework that represents divergence-free measures and similar conserved objects as superpositions of elementary constituents with exact mass or norm preservation.
• It extends classical decompositions to N-flows on hypergraphs, horizontal currents in the Heisenberg group, and structured factorizations in Hilbert and free function spaces.
• The method leverages approximation, flow invariance, and measure-theoretic identities to link analysis, geometry, and transport theory through conservation laws.

Smirnov-type decomposition denotes, in the sources considered here, a family of representation theorems associated with S. Smirnov and later analogues. In its classical form, it states that a divergence-free vector-valued Radon measure, equivalently a $1$-current of order $0$, can be written as a superposition of elementary curve-currents with an exact equality of masses (Rodríguez-Arenas et al., 2024). Related constructions appear for $N$-flows on countable hypergraphs, where a finite-mass flow splits uniquely into acyclic, solenoidal, and finite-cycle parts (Petrache, 2019); for horizontal vector charges in the Heisenberg group, where divergence-free horizontal currents are decomposed into horizontal curves (Huang et al., 12 May 2026); and in Hilbert function space theory, where “Smirnov-type” takes the form of factorization $h=b/a$ with an outer denominator (Jury et al., 2018). This suggests a unifying template: an object with a conservation law or dense-range property is represented by elementary constituents without loss of mass or norm.

1. Classical decomposition into elementary curve-currents

Let $\Omega\subset \mathbb{R}^n$ be open, and let $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$ be a finite $\mathbb{R}^n$-valued Radon measure. Its action on test fields $\phi\in C_c^\infty(\Omega)^n$ is

$$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$

its total variation is the positive Radon measure

$$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$

and its distributional divergence is

$0$0

The measure is divergence-free when $0$1 in $0$2 (Rodríguez-Arenas et al., 2024).

If $0$3 and $0$4 is Lipschitz, then $0$5 defines the elementary curve-current

$0$6

with

$0$7

Its total variation satisfies

$0$8

and equality occurs exactly in the no-backtracking case: $0$9 This identifies the elementary constituents that appear in Smirnov’s theorem (Rodríguez-Arenas et al., 2024).

For fixed $N$0, let

$N$1

equipped with the topology of uniform convergence. The classical theorem states that if $N$2 is finite and divergence-free, then for every $N$3 there exists a finite positive Borel measure $N$4 on $N$5 such that

$N$6

and for $N$7-almost every $N$8,

$N$9

Here $h=b/a$0 is the Smirnov measure, and the second identity is the equality of masses (Rodríguez-Arenas et al., 2024).

2. Proof architecture and measure-theoretic identities

A streamlined proof proceeds in three steps. First, one approximates $h=b/a$1 by smooth divergence-free fields $h=b/a$2. Second, at each $h=b/a$3 one considers the flow line $h=b/a$4 solving

$h=b/a$5

on $h=b/a$6. Liouville’s theorem for divergence-free flows shows that the push-forward of $h=b/a$7 under $h=b/a$8 is time-independent, yielding

$h=b/a$9

Third, Prokhorov’s theorem, or Banach–Alaoglu on $\Omega\subset \mathbb{R}^n$0, gives a weakly convergent subsequence $\Omega\subset \mathbb{R}^n$1; continuity of $\Omega\subset \mathbb{R}^n$2 in $\Omega\subset \mathbb{R}^n$3 allows passage to the limit and recovery of $\Omega\subset \mathbb{R}^n$4 (Rodríguez-Arenas et al., 2024).

The mass identity is not auxiliary but structural. Testing against vector fields of unit norm supported in open sets yields

$\Omega\subset \mathbb{R}^n$5

from which one deduces both $\Omega\subset \mathbb{R}^n$6 and $\Omega\subset \mathbb{R}^n$7 for $\Omega\subset \mathbb{R}^n$8-almost every $\Omega\subset \mathbb{R}^n$9. In the 2024 note, these equalities are presented alongside the curve-current action, the mass formula

$u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$0

the divergence identity for curves, and the no-backtracking criterion (Rodríguez-Arenas et al., 2024).

The same source also records broader directions: variants for currents of higher dimension, for elementary solenoids parametrized on $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$1, and in metric spaces; and applications to optimal transport and metric currents through the interpretation of $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$2 as a transport plan or as a decomposition in the sense of Ambrosio–Kirchheim (Rodríguez-Arenas et al., 2024).

3. Smirnov-type decomposition for $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$3-flows

In multimarginal transport, the relevant combinatorial object is an $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$4-flow. Let $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$5 be a Polish space and $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$6. One forms the $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$7-partite directed hypergraph

$u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$8

A signed $u=(u_1,\dots,u_n)\in M(\Omega;\mathbb{R}^n)$9-flow is a finitely supported real-valued function $\mathbb{R}^n$0, written formally as

$\mathbb{R}^n$1

with mass

$\mathbb{R}^n$2

and $\mathbb{R}^n$3-boundary

$\mathbb{R}^n$4

The flow is closed, or a cycle, when $\mathbb{R}^n$5 (Petrache, 2019).

The Smirnov-type decomposition theorem for $\mathbb{R}^n$6-flows states that any finite-mass $\mathbb{R}^n$7-flow on a countable hypergraph admits unique subflows

$\mathbb{R}^n$8

such that

$\mathbb{R}^n$9

Here $\phi\in C_c^\infty(\Omega)^n$0 is acyclic, meaning it has no nonzero closed subflows; $\phi\in C_c^\infty(\Omega)^n$1 is a finite or countable superposition of finite closed $\phi\in C_c^\infty(\Omega)^n$2-flows; and $\phi\in C_c^\infty(\Omega)^n$3 is solenoidal, meaning it is closed but contains no finite cycles (Petrache, 2019).

The proof removes finite closed subflows first, always choosing one of largest mass, then removes acyclic subflows in decreasing order of boundary-mass, and identifies the remainder as closed and cycle-free in the finite sense. A diagonal argument yields convergence of the peeled-off pieces in mass, and uniqueness follows because solenoidal and acyclic flows cannot share any nonzero subflow (Petrache, 2019). In the case $\phi\in C_c^\infty(\Omega)^n$4, every finite closed $\phi\in C_c^\infty(\Omega)^n$5-flow is a finite sum of elementary loops, the Euler–Kirchhoff decomposition on a directed graph; for $\phi\in C_c^\infty(\Omega)^n$6, that simple loop decomposition fails, which is precisely why the more elaborate trichotomy is needed (Petrache, 2019).

4. Multimarginal transport and the failure of the $\phi\in C_c^\infty(\Omega)^n$7 intuition

The same framework is used to analyze $\phi\in C_c^\infty(\Omega)^n$8-cyclical monotonicity in $\phi\in C_c^\infty(\Omega)^n$9-marginal transport. A central combinatorial statement is that a set $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$0 is $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$1-monotone if and only if every difference $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$2 of two finite-mass $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$3-flows supported in $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$4 with the same boundary has nonpositive $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$5-weight. Consequently, if the $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$6-graph of $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$7 admits no nonzero finite closed $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$8-flow, then every plan supported in $$\langle u,\phi\rangle = \sum_{k=1}^n \int_\Omega \phi_k\,d u_k,$$9 is trivially $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$0-monotone, because there is no competitor $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$1 to test (Petrache, 2019).

This mechanism underlies a counterexample for $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$2. In the construction, $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$3 and

$$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$4

for a suitably chosen bounded $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$5. The resulting $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$6-graph has vertices $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$7, edges $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$8, and for each $$|u|(E)=\sup\{\langle u,\phi\rangle: \phi\in C_c^\infty(E)^n,\ |\phi(x)|\le 1\},$$9 the three permutations of $0$00. It has no nontrivial finite $0$01-cycles, so every finite $0$02-flow with zero boundary vanishes (Petrache, 2019).

Nevertheless, the paper constructs two symmetric transport plans

$0$03

with the same one-point marginal $0$04 but satisfying

$0$05

Thus $0$06 is nonoptimal despite being $0$07-cyclically monotone (Petrache, 2019). The paper identifies $0$08 as a closed $0$09-flow of infinite support, namely a solenoidal flow. The usual $0$10 strategy of peeling off finite loops therefore cannot detect the cost difference. This directly addresses the misconception that the two-marginal criterion “$0$11-cyclical monotonicity implies optimality” extends naively to $0$12.

5. Horizontal currents in the Heisenberg group

A Heisenberg-group analogue replaces Euclidean divergence-free vector measures by horizontal vector charges. A horizontal vector measure on $0$13 is a $0$14-tuple of signed Radon measures

$0$15

acting on compactly supported horizontal test fields $0$16 by

$0$17

Its total variation identifies $0$18 with an element of $0$19, and its horizontal divergence is

$0$20

The current is divergence-free when $0$21 in the sense of distributions (Huang et al., 12 May 2026).

The key dynamical ingredient is a horizontal Liouville theorem: if $0$22 is a $0$23-smooth horizontal vector field of at most linear growth generating a complete flow $0$24, and if a Radon measure $0$25 satisfies $0$26, then $0$27 for all $0$28. Fixing $0$29, one considers the space $0$30 of absolutely continuous horizontal curves $0$31 with $0$32 a.e.; it is $0$33-compact in the uniform metric (Huang et al., 12 May 2026).

The main theorem states that if $0$34 is a horizontal vector measure on $0$35 with finite total variation and $0$36, then for each $0$37 there exists a positive Borel measure $0$38 on $0$39 such that

$0$40

and $0$41-almost every $0$42 has constant speed $0$43, hence length $0$44, and lies in $0$45 (Huang et al., 12 May 2026). The proof regularizes $0$46, applies the horizontal Liouville theorem to the normalized smooth field, pushes the invariant measure forward to curve space, and then passes to the limit by weak-* compactness. For general finite-divergence currents, the argument embeds $0$47 into $0$48, constructs a divergence-free lift, decomposes there, and projects back (Huang et al., 12 May 2026).

An illustrative corollary is a horizontal analogue of the Havin–Smirnov theorem: if $0$49 is compact and contains no nonconstant horizontal rectifiable curve, then for any continuous $0$50 on $0$51 and $0$52 there exists $0$53 such that

$0$54

This parallels the Euclidean application to free approximation on purely unrectifiable sets (Huang et al., 12 May 2026).

6. Smirnov factorization in complete Pick and free Fock settings

In Hilbert function spaces, “Smirnov-type decomposition” takes a different but closely related form. If $0$55 is a reproducing-kernel Hilbert space on $0$56, with multiplier algebra

$0$57

then $0$58 is $0$59-outer, or cyclic, when $0$60. The Smirnov class is

$0$61

equivalently the set of ratios arising from a contractive column multiplier $0$62 with $0$63 cyclic (Jury et al., 2018).

For a normalized complete Nevanlinna–Pick space $0$64, every $0$65 admits such a factorization: $0$66 with $0$67, $0$68 $0$69-outer, the column $0$70 contractive, and

$0$71

The proof uses the Agler–McCarthy model, a norm-preserving free lift to the full Fock space $0$72, and restriction back to commuting arguments (Jury et al., 2018).

The free version is formulated in terms of the left Smirnov class on the free unit ball. A free holomorphic function $0$73 belongs to $0$74 when

$0$75

for bounded left multipliers $0$76 with $0$77 left-outer. The associated multiplication operator

$0$78

is closed and densely defined precisely in this case, and then there is a unique inner-outer pair $0$79, up to a unimodular constant, with

$0$80

an isometric left multiplier. Moreover every $0$81 satisfies these hypotheses, and

$0$82

Here the decomposition is not into curves or cycles but into a quotient by an outer multiplier with exact norm control (Jury et al., 2018).