Method of Planes (MoP): Simulations & Analysis

• Method of Planes (MoP) is a dual-use approach that defines surface fluxes in molecular dynamics and applies reflection comparisons in PDE to derive symmetry properties.
• In molecular simulation, MoP computes local density, momentum, and stress by tracking molecular crossings across finite-area control planes, ensuring conservation laws are met.
• In PDE analysis, the moving planes method leverages maximum principles and reflection techniques to establish symmetry, monotonicity, and rigidity in geometric problems.

Method of Planes (MoP) denotes a plane-based methodology that appears in two distinct technical lineages. In molecular simulation and non-equilibrium molecular dynamics, it is a surface-based procedure for computing density, velocity, momentum flux, and local stress on a finite-area control plane from molecular crossing events and pair interactions (Kusudo et al., 2021). In analysis and geometric PDE, closely related literature uses the reflection-based method of moving planes to derive symmetry, monotonicity, and rigidity by comparing a function or hypersurface with its reflection across a moving hyperplane (Ciraolo et al., 2018). The shared geometric object is a plane, but the operative role of the plane differs: in microscopic transport it is a control surface carrying fluxes, whereas in PDE it is a moving comparison interface.

1. Terminological scope and lineages

Two usages dominate the literature.

Usage	Plane object	Primary purpose
Surface-based MoP in microscopic systems	Finite-area control plane $S_k$	Surface density, fluxes, stress, and conservation-law-consistent local mechanics
Method of moving planes in analysis	Hyperplane $T_\lambda$ moving through space	Symmetry, monotonicity, rigidity, and quantitative approximate symmetry

In the microscopic setting, a control plane $S_k$ is a finite area oriented with outward normal along Cartesian direction $k$. Surface-averaged quantities are defined by counting molecular passages across $S_k$ during a sampling interval and by summing pairwise interaction forces whose bond lines intersect the plane (Kusudo et al., 2021). In the analytical setting, one fixes a direction $e\in S^{n-1}$, defines hyperplanes $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$, reflects points by $x^\lambda=x-2(x\cdot e-\lambda)e$, and compares a solution $u$ with its reflection $u^\lambda(x)=u(x^\lambda)$ as the plane moves (Ciraolo et al., 2018).

The historical trajectories are separate. The reflection method originates in Alexandrov’s reflection principle and soap bubble theorem, Serrin’s symmetry result in potential theory, and the Gidas–Ni–Nirenberg symmetry theory for semilinear elliptic equations, with later quantitative developments and nonlocal extensions (Ciraolo et al., 2018). The surface-flux formulation in molecular simulation is rooted in the Irving–Kirkwood framework and was reformulated for non-equilibrium systems with macroscopic flow through the velocity distribution function, then extended to machine-learning potentials through an explicit pairwise force decomposition compatible with MoP control-volume balances (Kusudo et al., 2021).

2. Surface-based MoP in molecular simulation

In molecular dynamics, the Method of Planes computes local mechanical stress on a surface element rather than assigning stress to atom positions. The stress tensor on a surface with outward normal in direction $T_\lambda$0 is decomposed into kinetic and interaction parts,

$T_\lambda$1

where $T_\lambda$2 is the $T_\lambda$3-direction stress acting on that surface (Kusudo et al., 2021). In the standard equilibrium formulation, the kinetic part is obtained from momentum transfer by molecules crossing the plane, and the interaction part is obtained from pairwise forces whose bond lines cross the plane.

A central construction is the plane-average relation connecting a microscopic particle quantity $T_\lambda$4 with a macroscopic field $T_\lambda$5 defined on $T_\lambda$6: $T_\lambda$7 Here $T_\lambda$8 is the local velocity distribution function. By setting $T_\lambda$9, one obtains the surface density,

$S_k$0

and by setting $S_k$1, one obtains the surface mass flux,

$S_k$2

The streaming velocity on the plane follows as

$S_k$3

Because this estimator is defined for any component $S_k$4, including $S_k$5, any velocity component can be measured on the plane. That point is operationally important: it enables the separation of kinetic momentum flux into advection and stress in non-equilibrium systems with macroscopic flow (Kusudo et al., 2021).

The kinetic contribution to stress is written in terms of peculiar velocity $S_k$6, but the plane formulation makes the advection–stress split explicit. In particular,

$S_k$7

with the advective term reconstructed by

$S_k$8

The interaction part is accumulated from forces $S_k$9 for pairs $k$0 whose bond line crosses $k$1. The resulting formulation defines density, mass flux, momentum flux, and stress directly on a surface. The molecular-dynamics papers emphasize that this surface definition is essential for consistency with the mass and momentum conservation laws in dynamic systems (Kusudo et al., 2021).

3. Conservation laws, implementation, and validation in molecular and ML settings

The surface-based MoP is organized around the macroscopic balance laws. Mass conservation is written as

$k$2

and momentum conservation as

$k$3

Because the convective momentum flux $k$4 and the stress $k$5 appear as separate surface terms, the microscopic passage statistics across $k$6 must be split accordingly. This is the distinctive role of MoP in non-equilibrium molecular dynamics (Kusudo et al., 2021).

In practice, the limit $k$7 is replaced by the MD timestep $k$8. Crossings are detected from particle segments between $k$9 and $S_k$0, and the paper using velocity Verlet evaluates

$S_k$1

to maintain consistency between flux calculations and position updates and to avoid divisions by $S_k$2 (Kusudo et al., 2021). The same source states two practical assumptions: $S_k$3 does not vary appreciably within distances $S_k$4 across the oblique pillar used in the derivation, and the crossing velocities are evaluated consistently with the integration scheme.

Validation in a quasi-1D steady-state Couette flow showed that MoP reproduced volume-average density and velocity profiles to within $S_k$5 in density and $S_k$6 in velocity. In bulk, $S_k$7 and $S_k$8 as expected for laminar Couette flow, and $S_k$9 (approximately $e\in S^{n-1}$0) equaled the imposed external pressure $e\in S^{n-1}$1. The same study reports that $e\in S^{n-1}$2 differed from $e\in S^{n-1}$3 away from walls, highlighting the necessity of subtracting advection before interpreting $e\in S^{n-1}$4 as stress, and that the off-diagonal shear stress $e\in S^{n-1}$5 was constant in the liquid bulk and varied near walls due to solid–liquid friction (Kusudo et al., 2021). In a quasi-2D moving-contact-line system, MoP revealed a clockwise “caterpillar-like” flow pattern from plane-based velocities, heterogeneous $e\in S^{n-1}$6 due to viscosity, and strong tensile stress near liquid–vapor interfaces due to interfacial tension.

A later extension derives MoP stress for the MACE potential within the Irving–Kirkwood framework (Smith, 17 Sep 2025). For a plane with unit normal $e\in S^{n-1}$7 at point $e\in S^{n-1}$8, the MoP stress components are written as

$e\in S^{n-1}$9

where $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$0 is nonzero only if the interaction path crosses the plane. For MACE, the pairwise antisymmetric force decomposition

$T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$1

is used so that $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$2 and $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$3. The same work reports exact per-timestep momentum conservation for control volumes bounded by MoP planes, and for energy flux it distinguishes a preferred per-atom decomposition $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$4 from a naive pairwise alternative that does not converge to exact energy conservation as $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$5 (Smith, 17 Sep 2025).

The water–ZrO$T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$6 testcase in that study uses a triclinic cell of dimensions $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$7 Å, two solid walls of approximately $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$8 Å at top and bottom, $T_\lambda=\{x\in\mathbb{R}^n:x\cdot e=\lambda\}$9 slabs with $x^\lambda=x-2(x\cdot e-\lambda)e$0 Å, and a temperature of $x^\lambda=x-2(x\cdot e-\lambda)e$1 K. MoP yields a flat normal stress profile across $x^\lambda=x-2(x\cdot e-\lambda)e$2 planes, while the local virial profile shows pronounced peaks near the solid–liquid interface; the paper further states that the control-volume momentum balance is verified to machine precision at every timestep (Smith, 17 Sep 2025).

4. Reflection-based moving planes in PDE and geometric analysis

In analysis, the method of moving planes proceeds by fixing a direction $x^\lambda=x-2(x\cdot e-\lambda)e$3, defining the hyperplane

$x^\lambda=x-2(x\cdot e-\lambda)e$4

the reflected point

$x^\lambda=x-2(x\cdot e-\lambda)e$5

the cap

$x^\lambda=x-2(x\cdot e-\lambda)e$6

and the reflected function $x^\lambda=x-2(x\cdot e-\lambda)e$7. One starts with $x^\lambda=x-2(x\cdot e-\lambda)e$8 so large that $x^\lambda=x-2(x\cdot e-\lambda)e$9 does not intersect $u$0, decreases $u$1, and stops at the critical position

$u$2

At $u$3, either an interior tangency or a boundary tangency occurs. The comparison function $u$4 is then analyzed in the reflected region. In the model Serrin problem $u$5 in $u$6, $u$7 on $u$8, one has $u$9 in the reflected region; the strong maximum principle and Hopf lemma handle interior tangency, while Serrin’s corner lemma handles boundary tangency (Ciraolo et al., 2018).

The method yields classical rigidity and symmetry theorems. For Serrin’s overdetermined problem, if $u^\lambda(x)=u(x^\lambda)$0 is bounded with $u^\lambda(x)=u(x^\lambda)$1 and

$u^\lambda(x)=u(x^\lambda)$2

then $u^\lambda(x)=u(x^\lambda)$3 is a ball and $u^\lambda(x)=u(x^\lambda)$4 is radial; if $u^\lambda(x)=u(x^\lambda)$5, then

$u^\lambda(x)=u(x^\lambda)$6

The same reflection mechanism underlies Gidas–Ni–Nirenberg-type radial symmetry for positive solutions of $u^\lambda(x)=u(x^\lambda)$7 in balls or $u^\lambda(x)=u(x^\lambda)$8, and Alexandrov’s theorem that a bounded connected $u^\lambda(x)=u(x^\lambda)$9 hypersurface with constant mean curvature is a sphere (Ciraolo et al., 2018).

The geometric version rewrites the hypersurface locally as graphs over a tangent plane. Constant mean curvature gives

$T_\lambda$00

and the graph difference satisfies a linear uniformly elliptic equation $T_\lambda$01, so maximum principles again force coincidence at the critical plane. The survey also records a nonlocal Alexandrov theorem: for $T_\lambda$02, if the $T_\lambda$03-nonlocal mean curvature $T_\lambda$04 is constant on $T_\lambda$05, then $T_\lambda$06 is a ball (Ciraolo et al., 2018).

5. Quantitative, nonlocal, and direct nonlocal variants

A major development is the quantitative method of moving planes, where exact symmetry is replaced by stability estimates. The survey defines the Fraenkel asymmetry

$T_\lambda$07

the Hausdorff distance between $T_\lambda$08 and a sphere, and the oscillation of boundary data $T_\lambda$09. For the Serrin problem, Aftalion–Busca–Reichel define

$T_\lambda$10

and obtain concentric balls $T_\lambda$11 with

$T_\lambda$12

when the deficit is small. For the discrete Serrin problem, Ciraolo–Magnanini–Sakaguchi obtain the linear bound $T_\lambda$13. For quantitative Alexandrov, Ciraolo–Vezzoni use $T_\lambda$14 and prove $T_\lambda$15, together with $T_\lambda$16-smallness of $T_\lambda$17 relative to a sphere. For quantitative nonlocal Alexandrov, Ciraolo–Figalli–Maggi–Novaga define a Lipschitz deficit of $T_\lambda$18 and prove

$T_\lambda$19

with no touching ball or connectedness assumptions (Ciraolo et al., 2018).

The analytical toolkit behind these estimates includes quantitative Hopf boundary point lemmas, quantitative Serrin corner lemmas, interior and boundary Harnack inequalities, Carleson-type estimates, interior elliptic regularity, and the use of touching ball geometry to convert smallness of $T_\lambda$20 into small geometric asymmetry. The survey organizes the quantitative argument into approximate symmetry in each direction, localization of critical hyperplanes near an approximate center, control of inner and outer radii, and, for Alexandrov, upgrade to $T_\lambda$21-closeness (Ciraolo et al., 2018).

The same reflection framework has been adapted to genuinely nonlocal operators. For the uniformly elliptic nonlocal Bellman operator

$T_\lambda$22

Dai and Qin establish strong maximum principles for antisymmetric functions, maximum principles in bounded and unbounded sets, narrow region principles, decay-at-infinity principles, and sliding methods. These tools yield symmetry, monotonicity, uniqueness, asymptotic results in bounded and unbounded domains, and applications to the uniformly elliptic nonlocal Monge–Ampère operator $T_\lambda$23 of Caffarelli–Charro (Dai et al., 2020).

For the fractional Laplacian, a direct method of moving planes works with the integral operator itself rather than the Caffarelli–Silvestre extension. The key ingredients are again strong maximum principles for antisymmetric functions, narrow region principles, and decay at infinity, leading to radial symmetry and monotonicity in critical cases and nonexistence in subcritical cases for equations such as $T_\lambda$24 (Chen et al., 2014). An analogous direct method has also been developed for the logarithmic Schrödinger operator $T_\lambda$25, with kernel

$T_\lambda$26

where $T_\lambda$27. Under the asymptotic condition $T_\lambda$28 and $T_\lambda$29, nonnegative bounded Dini continuous solutions of

$T_\lambda$30

are radially symmetric and radially nonincreasing (Zhang et al., 2022).

6. Limitations, comparisons, and recurring points of confusion

A recurrent source of confusion is terminological. In molecular simulation, MoP is a surface-flux construction tied to control surfaces and conservation laws. In PDE, the moving-planes method is a reflection and comparison method whose decisive ingredients are maximum principles and the behavior of reflected caps. The acronym may overlap in informal usage, but the technical content is distinct.

Within molecular simulation, MoP is repeatedly contrasted with volume-average and local virial approaches. The 2021 formulation states that density, mass, and momentum fluxes are defined on a surface rather than in a volume, and presents this as essential for consistency with mass and momentum conservation in dynamic systems (Kusudo et al., 2021). The MACE study makes the comparison sharper: the local virial formula is described as strictly valid for whole periodic domains, not for local inhomogeneous bins, and near walls it produces peaks aligned with density oscillations, whereas plane-based MoP maintains a flat normal stress profile under mechanical equilibrium (Smith, 17 Sep 2025). The same paper also identifies limitations: long-range electrostatics of Ewald type are not included in the presented MoP, and accurate energy flux for many-body ML models requires a per-atom energy decomposition rather than a naive pairwise one.

Within analysis, the scope of moving planes is broad but conditional. The survey states that typical PDE formulations require $T_\lambda$31 or comparable touching-ball assumptions, together with validity of maximum principles for the operator. Convexity is not required for the sharp quantitative Alexandrov result, although in qualitative settings it prevents bubbling; for quantitative results, touching-ball conditions are used except in the nonlocal Alexandrov theorem, where nonlocality precludes bubbling (Ciraolo et al., 2018). Related techniques include moving spheres, which is especially useful in conformal or scale-invariant problems, and the sliding method, which uses translations rather than reflections and is employed for monotonicity and symmetry in unbounded domains (Ciraolo et al., 2018).

Taken together, these lineages show how a plane can function either as a geometric boundary carrying fluxes or as a moving mirror enforcing comparison. In molecular mechanics, the plane localizes traction and momentum transfer; in PDE and geometry, the plane organizes symmetry and rigidity arguments. The persistence of the term across both areas reflects a shared dependence on planar structure, but the mathematical and physical content is determined by the surrounding theory rather than by the name alone.