High-Discretization Method of Moments for Capacitance Calculation: A Cube and a Hollow Cylinder

This paper employs the method of moments (MOM) to calculate the capacitances of a cube and a hollow cylinder. For the cube, each face was divided into a maximum of 600 x 600 sub-areas. By fully exploiting the geometric symmetry between sub-areas and incorporating parallel computing, computational resources were significantly conserved. Our results show that the calculated capacitance of the cube first increases and then decreases as the number of sub-areas increases. When each face was divided into 90 x 90 sub-areas, the capacitance of the unit cube (with an edge length of 1 m) reached a maximum reference value of 73.519014 pF. This indicates that higher accuracy cannot be achieved merely by indefinitely increasing the number of discretized sub-areas. Subsequently, the method was applied to compute the capacitance of a hollow cylinder. The results were compared with numerical solutions based on Lekner’s theoretical formula and Cavendish’s experimental values, showing good agreement among the three.

💡 Research Summary

The paper presents a high‑resolution implementation of the Method of Moments (MOM) for calculating the capacitance of two canonical electrostatic problems that lack closed‑form solutions: a unit cube and a hollow cylindrical conductor. The authors first review the extensive literature on the cube’s capacitance, noting that previous numerical studies have employed relatively coarse surface discretizations (e.g., 6 × 6, 20 × 20, 48 × 48 patches) and have reported values ranging from 0.62211 a to 0.71055 a (in CGS units) or roughly 72.9 pF to 73.5 pF for a 1 m edge.

In the cube section, each face is subdivided into N × N rectangular patches, with N ranging up to 600. By exploiting the geometric symmetry of the cube, the authors reduce the full set of unknown surface charge densities to ten distinct groups. They formulate a 10 × 10 linear system φ = P σ, where φ is the vector of potentials (all set to 1 V) and σ contains the representative charge densities. The self‑potential of a patch is approximated by 3.52549 σ/(4π ε₀) (derived from an exact integral over a square), while the interaction between distinct patches is approximated by σ/(4π ε₀ d), with d the center‑to‑center distance. Solving the linear system yields σ, from which the total charge Q is assembled by counting the multiplicities of each symmetry class. The capacitance follows from C = Q/φ.

A systematic sweep of N shows that the computed capacitance rises with increasing resolution, reaches a maximum of 73.519014 pF at N = 90, and then declines for larger N. The authors attribute the rise to the approach toward electrostatic equilibrium (Thomson’s theorem) and the decline to discretization errors inherent in the interaction approximation (A2). They argue that the peak at N = 90 represents the best attainable value with their formulation, and that the true capacitance must be slightly higher. Their results are compared with a broad set of prior methods (MOM, Brownian dynamics, boundary element, finite‑difference, etc.) in Table I, showing that the 90 × 90 value sits between the most accurate previous numerical estimates and the refined Brownian dynamics result.

The second part addresses a hollow cylinder of length L and radius R. The authors review analytical limits (Maxwell, Landau) and introduce Lekner’s matrix‑based formulation, where the capacitance C₀ is obtained from the (0,0) element of the inverse of an infinite matrix Kₘₙ expressed via Meijer‑G functions. To apply MOM, the cylinder surface is discretized into L annular rings of width 1 m, each further divided into K square patches of side 1 m, exploiting axial symmetry. The same self‑potential and interaction approximations as for the cube are used, and each ring’s potential is fixed at 1 V to solve for its uniform charge density. The total charge yields C = Q/φ, and the authors report the capacitance‑to‑radius ratio C/R for various L/K combinations.

Table II compares these ratios with Lekner’s 6th‑order truncated matrix results, showing agreement within 0.01 % for most aspect ratios; larger discrepancies appear for extreme L/K values where computational limits (maximum L = 5000) restrict accuracy. Table III adds experimental data from Cavendish, noting that the theoretical and numerical values are consistently higher, which the authors explain by historical measurement uncertainties (visual reading errors, charge leakage, imperfect insulation). Figure 5 visualizes the charge density along the cylinder axis, confirming higher densities near the ends and a smoother distribution in the middle, as expected from edge effects.

The discussion emphasizes that while high‑resolution MOM can achieve sub‑percent accuracy for complex geometries, the dominant source of error remains the approximation of inter‑patch potentials (A2). The authors suggest that more sophisticated Green’s function evaluations or higher‑order multipole expansions could mitigate this limitation. They also highlight the computational efficiency gained by symmetry reduction (10 × 10 matrix for the cube) and parallel implementation in Python, which makes even 600 × 600 discretizations tractable on modest hardware.

In conclusion, the paper demonstrates that a carefully symmetrized, high‑resolution MOM, combined with modern parallel computing, can deliver highly accurate capacitance predictions for geometries where analytical solutions are unavailable. It validates the method against both established numerical techniques and historical experimental data, and it provides a clear roadmap for future improvements, such as adaptive meshing, refined interaction kernels, and automated symmetry detection.