# Preferred Frame Writing [🏠 Home](/) - [About](/about.md.html) - [Submissions](/submissions.md.html) - [Policies](/policies.md.html) - [Journals](/journals.md.html) ---

The Physics of Energy Flow - Minimal Propagating Closure of Source-Free Flow

2026-03-13

# 203. Minimal Propagating Closure of Source-Free Flow This appendix gives the mathematical step used structurally in chapter 7. The result is the following. > A single real first-order self-curl evolution of a divergence-free field does > not produce neutral propagating transport. The minimal local propagating > closure in this class requires two coupled divergence-free fields. The point is not to postulate Maxwell's equations, but to show why they appear as the minimal propagating closure of source-free rotational transport. ## 203.1 Source-Free Transport Let $$ \mathbf{F}(\mathbf{r},t) $$ be a vector field on three-dimensional space. Source-free transport means $$ \nabla \cdot \mathbf{F} = 0. $$ This expresses the absence of primitive beginnings or endings of the flow. If a primitive start or end point were enclosed by a closed surface, the net flux through that surface would not vanish. The source-free condition says that this does not happen: the enclosed net flow remains identically zero. ## 203.2 Divergence Preservation Under Evolution Assume a local first-order evolution relation $$ \partial_t \mathbf{F} = \mathcal{D}(\mathbf{F}), $$ where $\mathcal{D}$ is a spatial differential operator. To preserve the source-free condition we require $$ \nabla \cdot (\partial_t \mathbf{F}) = 0. $$ Substituting the evolution relation gives $$ \nabla \cdot \mathcal{D}(\mathbf{F}) = 0 $$ for every divergence-free field $\mathbf{F}$. A natural local first-order differential operator with this property is curl, since $$ \nabla \cdot (\nabla \times \mathbf{A}) = 0 $$ for any vector field $\mathbf{A}$. So a natural divergence-preserving self-update is $$ \partial_t \mathbf{F} = k\,\nabla \times \mathbf{F}. $$ This relation is posed simultaneously for all $\mathbf{r}$ in the extent. It is therefore a whole-field update, not a rule for tracking one individually marked point through space. We now examine whether this relation yields propagating transport. ## 203.3 Failure of the Single Self-Curl Relation The simplest test is to differentiate the self-curl relation once more: $$ \partial_t^2 \mathbf{F} = k\,\nabla\times(\partial_t\mathbf{F}). $$ Substituting $$ \partial_t \mathbf{F} = k\,\nabla \times \mathbf{F} $$ gives $$ \partial_t^2 \mathbf{F} = k^2\,\nabla\times(\nabla\times\mathbf{F}). $$ Using the vector identity $$ \nabla \times (\nabla \times \mathbf{F}) = \nabla(\nabla\cdot\mathbf{F})-\nabla^2\mathbf{F}, $$ and the source-free condition $$ \nabla\cdot\mathbf{F}=0, $$ we obtain $$ \partial_t^2 \mathbf{F} = -k^2\nabla^2\mathbf{F}. $$ Equivalently, $$ \partial_t^2 \mathbf{F} + k^2\nabla^2\mathbf{F}=0. $$ This is not the neutral propagating wave equation, in which the second temporal derivative term and the spatial Laplacian term appear with opposite signs. Here both second-order terms enter with the same sign. A single self-curl evolution therefore does not by itself furnish the propagating closure we seek. It preserves turning, but it does not produce the neutral propagating form. There is also a direct obstruction to bodily transport. Assume, for contradiction, that a nontrivial bounded closure could be carried bodily by the single self-curl relation. Then there would exist a smooth localized profile $\mathbf{G}$ and a constant drift velocity $\mathbf{v}$ such that $$ \mathbf{F}(\mathbf{r},t)=\mathbf{G}(\mathbf{r}-\mathbf{v}t). $$ Differentiating gives $$ \partial_t\mathbf{F} = -(\mathbf{v}\cdot\nabla)\mathbf{G}, \qquad \partial_t^2\mathbf{F} = (\mathbf{v}\cdot\nabla)^2\mathbf{G}. $$ Substituting this translating ansatz into $$ \partial_t^2 \mathbf{F} + k^2\nabla^2\mathbf{F}=0 $$ gives $$ (\mathbf{v}\cdot\nabla)^2\mathbf{G}+k^2\nabla^2\mathbf{G}=0. $$ Now take the Euclidean inner product with $\mathbf{G}$ and integrate over all space. Because $\mathbf{G}$ is localized, the boundary terms vanish under integration by parts. Therefore $$ \int \mathbf{G}\cdot(\mathbf{v}\cdot\nabla)^2\mathbf{G}\,dV = -\int \left|(\mathbf{v}\cdot\nabla)\mathbf{G}\right|^2\,dV $$ and $$ \int \mathbf{G}\cdot\nabla^2\mathbf{G}\,dV = -\int |\nabla\mathbf{G}|^2\,dV. $$ So $$ \int \left|(\mathbf{v}\cdot\nabla)\mathbf{G}\right|^2\,dV + k^2\int |\nabla\mathbf{G}|^2\,dV = 0. $$ Both integrands are nonnegative. Hence both integrals must vanish: $$ (\mathbf{v}\cdot\nabla)\mathbf{G}=0, \qquad \nabla\mathbf{G}=0. $$ So $\mathbf{G}$ is constant. Since $\mathbf{G}$ is localized, that constant must be zero. Therefore the only localized rigidly translating solution of the single self-curl relation is the trivial one. This proves the point needed in the main text: a single self-curl update can turn a structure, but it cannot carry a nontrivial bounded closure bodily from one region to another. The same argument rules out rigid bodily rotation of a localized closure. Assume that a nontrivial bounded closure rotates rigidly about a fixed axis. Let $Q(t)=e^{t\Omega}$ be the corresponding one-parameter family of rotation matrices, with $\Omega$ a constant skew-symmetric matrix, and suppose $$ \mathbf{F}(\mathbf{r},t)=Q(t)\,\mathbf{G}(Q(t)^{-1}\mathbf{r}). $$ Define the linear operator $$ A_\Omega \mathbf{G} := \Omega \mathbf{G}-(\Omega\mathbf{r})\cdot\nabla \mathbf{G}. $$ Then $$ \partial_t\mathbf{F}\big|_{t=0}=A_\Omega\mathbf{G}, \qquad \partial_t^2\mathbf{F}\big|_{t=0}=A_\Omega^2\mathbf{G}. $$ Substituting into $$ \partial_t^2 \mathbf{F} + k^2\nabla^2\mathbf{F}=0 $$ at $t=0$ gives $$ A_\Omega^2\mathbf{G}+k^2\nabla^2\mathbf{G}=0. $$ Now take the $L^2$ inner product with $\mathbf{G}$. The operator $A_\Omega$ is skew-adjoint on compactly supported fields: - $\Omega$ is pointwise skew-symmetric, so $$ \int \mathbf{U}\cdot(\Omega\mathbf{V})\,dV = -\int (\Omega\mathbf{U})\cdot\mathbf{V}\,dV $$ - the vector field $\Omega\mathbf{r}$ has zero divergence because $\mathrm{tr}(\Omega)=0$, so integration by parts gives $$ \int \mathbf{U}\cdot\big((\Omega\mathbf{r})\cdot\nabla\mathbf{V}\big)\,dV = -\int \big((\Omega\mathbf{r})\cdot\nabla\mathbf{U}\big)\cdot\mathbf{V}\,dV $$ Therefore $$ \int \mathbf{G}\cdot A_\Omega^2\mathbf{G}\,dV = -\int |A_\Omega\mathbf{G}|^2\,dV. $$ Together with $$ \int \mathbf{G}\cdot\nabla^2\mathbf{G}\,dV = -\int |\nabla\mathbf{G}|^2\,dV, $$ we obtain $$ \int |A_\Omega\mathbf{G}|^2\,dV + k^2\int |\nabla\mathbf{G}|^2\,dV = 0. $$ Again both terms are nonnegative, so both must vanish. Hence $$ A_\Omega\mathbf{G}=0, \qquad \nabla\mathbf{G}=0. $$ Thus $\mathbf{G}$ is constant, and since it is localized, it must be zero. Therefore the only localized rigidly rotating solution of the single self-curl relation is the trivial one. So the one-field self-curl update does not bodily move a bounded closure, either by translation or by rigid rotation. What remains possible is weaker than rigid-body motion: internal reorientation, internal deformation, or phase progression on a fixed support. Those possibilities are not classified by the present no-go result. This is the precise content of the chapter-7 summary: single curl reorganizes locally. ## 203.4 Coupled Curl Evolution Now introduce two divergence-free fields $$ \mathbf{F}_+, \qquad \mathbf{F}_-. $$ Consider the coupled evolution $$ \partial_t \mathbf{F}_+ = k\,\nabla \times \mathbf{F}_- $$ $$ \partial_t \mathbf{F}_- = -k\,\nabla \times \mathbf{F}_+. $$ Taking a time derivative of the first equation, $$ \partial_t^2 \mathbf{F}_+ = k\,\nabla \times (\partial_t \mathbf{F}_-). $$ Substituting the second equation, $$ \partial_t^2 \mathbf{F}_+ = -k^2\,\nabla \times (\nabla \times \mathbf{F}_+). $$ Using the vector identity $$ \nabla \times (\nabla \times \mathbf{F}) = \nabla(\nabla\cdot\mathbf{F})-\nabla^2\mathbf{F}, $$ and the divergence-free condition $$ \nabla\cdot\mathbf{F}_+ = 0, $$ we obtain $$ \partial_t^2 \mathbf{F}_+ = k^2\nabla^2\mathbf{F}_+. $$ Thus $\mathbf{F}_+$ satisfies the wave equation $$ \partial_t^2\mathbf{F}_+ - k^2\nabla^2\mathbf{F}_+ = 0. $$ The same derivation holds for $\mathbf{F}_-$. There is also an explicit transporting branch. Let $\phi:\mathbb{R}\to\mathbb{R}$ be any smooth scalar profile, and define $$ \mathbf{F}_+(\mathbf{r},t)=\phi(x-kt)\,\mathbf{e}_y, \qquad \mathbf{F}_-(\mathbf{r},t)=\phi(x-kt)\,\mathbf{e}_z. $$ Then $$ \nabla\cdot\mathbf{F}_+=0, \qquad \nabla\cdot\mathbf{F}_-=0, $$ because each field has only one transverse component and depends only on $x$. Now compute the curls: $$ \nabla\times\mathbf{F}_- = \nabla\times(0,0,\phi(x-kt)) = (0,-\partial_x\phi(x-kt),0), $$ and $$ \nabla\times\mathbf{F}_+ = \nabla\times(0,\phi(x-kt),0) = (0,0,\partial_x\phi(x-kt)). $$ Also, $$ \partial_t\mathbf{F}_+ = (0,-k\,\partial_x\phi(x-kt),0), $$ and $$ \partial_t\mathbf{F}_- = (0,0,-k\,\partial_x\phi(x-kt)). $$ Therefore $$ \partial_t\mathbf{F}_+ = k\,\nabla\times\mathbf{F}_-, \qquad \partial_t\mathbf{F}_- = -k\,\nabla\times\mathbf{F}_+. $$ So the coupled curl system admits exact translating solutions. If the initial profile $\phi$ is supported in an interval $[a,b]$, then at time $t$ the transported profile is supported in the shifted interval $$ [a+kt,b+kt]. $$ Thus the doubled structure does what the single self-curl relation cannot do: it carries a profile from one region to another. The transport is explicit. The shape is preserved, and the profile advances rigidly at speed $k$ along the $x$ direction. This establishes the existence of a genuine transport branch. It is a one-direction translating profile embedded in three dimensions. Additional closure is needed later to build bounded self-sustained modes from such transport. ## 203.5 Minimal Propagating Closure The analysis shows: - a single divergence-preserving self-curl evolution makes the temporal and spatial second-order terms enter with the same sign, not the neutral propagating form - single curl reorganizes locally, but does not bodily carry a nontrivial bounded closure - the coupled system has exact translating branches carrying a profile from one region to another - two coupled curl evolutions do yield neutral wave propagation So the minimal propagating closure in this class is $$ \partial_t \mathbf{F}_+ = k\,\nabla \times \mathbf{F}_- $$ $$ \partial_t \mathbf{F}_- = -k\,\nabla \times \mathbf{F}_+. $$ These equations preserve $$ \nabla\cdot\mathbf{F}_+ = 0,\qquad \nabla\cdot\mathbf{F}_- = 0. $$ ## 203.6 Electromagnetic Normalization Now define $$ \mathbf{E} \equiv \mathbf{F}_+,\qquad \mathbf{B} \equiv \mathbf{F}_-/k. $$ Then the coupled equations become $$ \partial_t \mathbf{E} = k^2\nabla \times \mathbf{B} $$ $$ \partial_t \mathbf{B} = -\nabla \times \mathbf{E}. $$ With conventional constants absorbed into the normalization of $k$, these correspond to the source-free Maxwell equations. ## 203.7 Interpretation The two fields are not independent substances. They are two complementary transverse aspects of the same organized source-free transport. Their mutual curl coupling yields the minimal propagating structure compatible with divergence-free flow. ## 203.8 Summary Starting from divergence-free transport: - curl preserves the source-free condition - a single self-curl evolution makes the temporal and spatial second-order terms enter with the same sign, not the neutral propagating form - two coupled curl evolutions yield neutral wave propagation - the resulting equations coincide with the source-free Maxwell system Maxwell dynamics therefore appears here as the minimal propagating closure of source-free rotational transport. Appendix 211 then proves that within the stated real local isotropic first-order two-field class, this closure is also unique up to real field recombination.
--- - [Preferred Frame Writing on GitHub.com](https://github.com/siran/writing) (built: 2026-03-18 13:41 EDT UTC-4)