Pillar I — Structural: arrangement forces the equations

From mass and pairwise interaction, ∂_tρ+∇·(ρuv)=0 and ρ Duv/Dt=∇·σ (registry T≡σ) are exact identities; the Newtonian closure is a located, falsifiable gate. Solver corollaries hold to 3×10⁻⁸ and 4×10⁻⁶, and the compressible corollary carries the substrate sound speed c_s²=B/ρ. [DERIVE]

The first pillar establishes the weakest, most secure claim, and the one on which the others lean: the form of the fluid equations is not postulated; it is forced by the arrangement. We do not need to assume that the medium "obeys Navier–Stokes." We need only A0. The balance laws then follow as exact identities, and the Newtonian closure follows conditionally.

From arrangement to fields

Given the kernel $W_\ell$ (smooth, normalized, $\int W_\ell=1$ ), define

\begin{equation} \rho(\xv,t)=\sum_i m_i\,W_\ell(\xv-\xv_i),\qquad \rho\uv(\xv,t)=\sum_i m_i\vv_i\,W_\ell(\xv-\xv_i). \label{eq:cg_fields} \end{equation}

These are definitions ([LOCK]): the fields are weighted bookkeeping of the arrangement. No dynamical assumption has entered.

Mass and momentum balance are exact

Exact balance laws

[DERIVE]. If the particles move by $\dot\xv_i=\vv_i$ and exchange momentum through pairwise forces, then (eq.) satisfy, identically,

\begin{align} \partial_t\rho+\div(\rho\uv)&=0,\label{eq:cont}\\ \partial_t(\rho\uv)+\div(\rho\uv\otimes\uv)&=\div\,\sigma+\rho\mathbf f, \qquad \sigma=-p\,\Id+\tau.\label{eq:mom} \end{align}

The stress $\sigma$ is given exactly by an Irving–Kirkwood formula: a kinetic part from peculiar-velocity transport and an interaction part from pairwise forces integrated along the segment joining each pair.

The derivation, in full.

Both identities come from differentiating the definitions (eq.) and using nothing but $\dot\xv_i=\vv_i$ and Newton's third law. For continuity, differentiate $\rho$ in time and use $\partial_t W_\ell(\xv-\xv_i)=-\dot\xv_i\!\cdot\!\nabla_\xv W_\ell=-\vv_i\!\cdot\!\nabla_\xv W_\ell$ :

\begin{equation} \partial_t\rho=\sum_i m_i\,\partial_t W_\ell(\xv-\xv_i) =-\nabla_\xv\!\cdot\!\sum_i m_i\vv_i\,W_\ell(\xv-\xv_i) =-\div(\rho\uv), \end{equation}

which is (eq.), exactly, with no closure. For momentum, differentiate $\rho\uv$ :

\begin{equation} \partial_t(\rho\uv)=\sum_i m_i\dot\vv_i\,W_\ell \;-\;\nabla_\xv\!\cdot\!\sum_i m_i\,\vv_i\!\otimes\!\vv_i\,W_\ell . \end{equation}

The second sum is split by writing each particle velocity as the local mean plus a peculiar part, $\vv_i=\uv(\xv_i)+\cv_i$ ; the mean part assembles the convective flux $\rho\,\uv\!\otimes\!\uv$ while the fluctuations leave a kinetic stress $\sigma^{\rm kin}=-\sum_i m_i\,\cv_i\!\otimes\!\cv_i\,W_\ell$ . In the first sum $m_i\dot\vv_i=\mathbf F_i=\sum_{j\neq i}\mathbf F_{ij}+m_i\mathbf f$ . The pairwise piece is converted to a divergence by the Irving–Kirkwood identity: since $\mathbf F_{ij}=-\mathbf F_{ji}$ , pairing terms and writing the difference of two kernels as the integral of a gradient along the bond $\rv_{ij}=\xv_i-\xv_j$ gives

\begin{equation} \sum_i\Big(\sum_{j\neq i}\mathbf F_{ij}\Big)W_\ell(\xv-\xv_i) =\nabla_\xv\!\cdot\!\,\sigma^{\rm int},\qquad \sigma^{\rm int}=-\tfrac12\sum_{i\neq j}\mathbf F_{ij}\!\otimes\!\rv_{ij}\! \int_0^1\! W_\ell\big(\xv-\xv_j-s\,\rv_{ij}\big)\,ds . \label{eq:IK_stress} \end{equation}

Collecting terms yields (eq.) with $\sigma=\sigma^{\rm kin}+\sigma^{\rm int}$ and $\rho\mathbf f=\sum_i m_i\mathbf f\,W_\ell$ . Every step is an identity: the continuum balance laws are not a model of the arrangement; they are the arrangement, re-summed. The pressure is the isotropic part of $\sigma$ and the deviatoric remainder is $\tau$ ; the kinetic part dominates in dilute, fast media and the interaction part in dense, jammed ones, but the form (eq.) is fixed regardless.

Why this matters for the thesis. The continuity and momentum forms in (eq.)–(eq.) are consequences of three facts about the arrangement: there is mass, it moves, and it pushes on itself. There is no separate axiom "the medium is a fluid"—and Section §3 showed none is needed: the marginal arrangement earns fluidity as a measured property. In this sense the shape of fluid dynamics is already configured: it is what the bookkeeping of any finite-mass, interacting arrangement must look like. The only genuinely open object is the stress $\sigma$ , and the rest of the pillar is about when $\sigma$ takes the Newtonian form. (For cross-document searchability we also record the registry form of the two laws, $\partial_t\rho+\nabla\!\cdot\!(\rho\uv)=0$ and $\rho\,D\uv/Dt=\nabla\!\cdot\!T$ with $T\equiv\sigma$ ; the symbol $\sigma$ is used throughout this document.)

The Newtonian closure is conditional, not automatic

Closure as a gate, not an assumption

[GATE] (G-Newton). The interaction stress reduces to the Newtonian form

\begin{equation} \tau=2\mu S+\lambda(\div\uv)\Id,\qquad S=\tfrac12\big(\nabla\uv+(\nabla\uv)^\top\big)-\tfrac13(\div\uv)\Id, \label{eq:newtonian} \end{equation}

if and only if the arrangement's stress responds to deformation by local, isotropic, Markovian linear response in the declared regime. This requires three measurable conditions to pass: (i) stress–strain-rate linearity; (ii) effective isotropy after window averaging; (iii) a short memory (small Deborah number, $\mathrm{De}\ll1$ ). When they pass, (eq.)–(eq.) become the compressible Navier–Stokes equations, and in the incompressible limit

\begin{equation} \partial_t\uv+\uv\!\cdot\!\nabla\uv=-\tfrac{1}{\rho_0}\nabla p+\nu\nabla^2\uv+\mathbf f, \qquad\nu=\mu/\rho_0. \label{eq:incompNS} \end{equation}

Failure mode. If linearity or short memory fails (jammed force chains with long memory, strong anisotropy that survives averaging), the closure is non-Newtonian and (eq.) is simply wrong for that regime. The gate is therefore falsifiable per regime.\\ Where the gate sits (Section §3). The three conditions are not free-floating: they name a position on the distance-to-margin axis $\Delta z=z-2d$ of the substrate. Far from the margin, rearrangements are fast and local (short memory, small Deborah number) and the closure holds; as $\Delta z\to0$ the relaxed shear stiffness and the yield scale vanish while the viscosity proxy diverges (inherited verification, Section §3), and the closure fails in exactly the jammed, force-chain way named above. The gate is thereby located, not merely stated.

The viscosity inherits a definite sign from the same construction: because the dissipative part of the stress is a positive quadratic form in the strain rate (a Green–Kubo integral of a stress autocorrelation, which is nonnegative), one has $\mu\ge0$ and hence $\eps_\nu\ge0$ ([DERIVE]). The value of $\mu$ in a concrete arrangement is a gate item (it must be measured by Green–Kubo or dissipation matching), but its sign is forced.

What the solver confirms

Two corollaries of Pillar I are exact and therefore make sharp numerical targets. First, with no viscosity and no forcing the incompressible flow is a Hamiltonian (Lie–Poisson) system and conserves both energy and enstrophy in two dimensions. Second, with viscosity but no forcing the energy obeys the exact budget

\begin{equation} \frac{dE}{dt}=-\eps_\nu=-2\nu Z,\qquad E=\tfrac12\langle|\uv|^2\rangle,\quad Z=\tfrac12\langle\omega^2\rangle, \label{eq:budget} \end{equation}

which is the bookkeeping backbone of Pillar IV. We verify both with a self-contained pseudo-spectral solver.

Pillar I — exact corollaries of the balance laws (ns2d.py, validate_all.py)

2D incompressible Navier–Stokes in vorticity form, periodic box $L=2\pi$ , integrating-factor RK4, 2/3 dealiasing; band-limited random initial vorticity, fixed seed.

Inviscid invariants ( $\nu=0$ , , ): relative drift $\Delta E/E_0=2.9\times10^{-8}$ , $\Delta Z/Z_0=5.4\times10^{-6}$ . PASS\\ $\Rightarrow$ confirms the spectral discretization and dealiasing realize the Hamiltonian structure of Proposition §6 in the inviscid limit.
Energy-budget identity (eq.) ( $\nu=0.01$ , , ): $\max_t |(-\dot E)-\eps_\nu|/\eps_\nu=4.4\times10^{-6}$ . PASS\\ $\Rightarrow$ the exact budget that Pillar IV will split into viscous and event channels holds to machine-gradient precision.
Enstrophy monotone (no forcing): $dZ/dt\le0$ at every step, so $\max_t Z=Z(0)$ and hence $\max_t\eps_\nu=2\nu Z(0)\propto\nu$ exactly. PASS

Run python validate_all.py; checks 1–3 correspond to the three bullets.

The compressible corollary: the arrangement has a sound speed

The balances (eq.)–(eq.) are not restricted to the incompressible regime in which the solver above runs. Linearize them about rest ( $\rho=\rho_0+\delta\rho$ , $\uv$ small) with a barotropic closure $p=p(\rho)$ : continuity gives $\partial_t\delta\rho=-\rho_0\nabla\!\cdot\!\uv$ , momentum gives $\rho_0\,\partial_t\uv=-c_s^2\nabla\delta\rho$ , and together

\[ \partial_t^2\,\delta\rho=c_s^2\,\nabla^2\delta\rho, \qquad c_s^2=\frac{dp}{d\rho}=\frac{B}{\rho}, \]

with

the bulk modulus. This is textbook acoustics; what the arrangement adds is that

is not a free parameter but a measured property of the packing, and that at the margin it is the only stiffness left: $G_{\rm relaxed}\to0$ while

stays finite (Section §3), so the marginal medium supports exactly one elastic wave—longitudinal sound, the defining acoustic signature of a fluid. The substrate measurement puts numbers on it: $c^2=B_{\rm rel}/\rho$ agrees across system sizes to $0.4\%$ , softened below the affine (Born) prediction by the non-affine relaxation ( $\langle B_{\rm rel}/B_{\rm Born}\rangle=0.711$ here; $\times0.65$ in the companion's larger-

runs). Scope, declared honestly: the pseudo-spectral solver of this pillar is incompressible, so the acoustic face is verified at substrate level (marginal_fluidity.py), not by a compressible flow run ([DERIVE] for the corollary given measured

; the companion's speed-of-light identification is not imported—see Section §14).

Pillar I, in one line. The equations of fluid motion are the necessary bookkeeping of a massive, interacting arrangement ([DERIVE]); their Newtonian specialization is a falsifiable regime property ([GATE]); their inviscid invariants and energy budget reproduce to $10^{-6}$ – $10^{-8}$ .