Substrate Geometry
Publications · five works in the core sequence

Papers

Four papers in the core Substrate Geometry sequence plus a methodology paper. Three are live on arXiv, one is in moderation, the methodology paper awaits endorsement for math.NA submission. A chemistry preprint on hERG channel QSAR for iboga alkaloids is drafted in parallel and queued for ChemRxiv.

Paper I  ·  Oloid as Local Optimum

Live · v2 arXiv:2604.12238 April 2026

Introduces the Contact Distribution Score (CDS), an area-weighted variance of contact time across a rolling body's surface. CDS is the program's first formal operational invariant expressible as a computable function on a watertight mesh. The oloid scores CDS = 8.2 × 10⁻⁷ at 1,198 faces against the cylinder's 4.75 × 10⁻⁵, a 58× discrimination, and is shown to occupy a local minimum within the developable roller family across a 1,430-genome parametric sweep varying angle, offset, and radius ratio.

Contact distribution comparison between oloid and cylinder showing the 58× variance difference
Figure 1. Contact distribution comparison between the oloid and the cylinder at matched mesh resolution. The variance of contact time across surface elements is the load-bearing quantity; the oloid's tighter distribution corresponds to more uniform wear, stress, and thermal loading under rolling.

The two-tier invariant transfer structure is the load-bearing finding beyond CDS itself. The four linear variance metrics — CDS, Hertz SDS, thermal TDS, wear WDS — cluster within measurement precision at approximately 8 × 10⁻⁷, while the nonlinear Basquin fatigue metric FDS diverges to 2.42 × 10⁻⁶ due to S–N amplification. The structural distinction between linear and nonlinear invariants is methodological as well as descriptive: it tells the framework what to expect from a candidate primitive across operational regimes without requiring each regime to be measured independently.

The v2 revision applies the trajectory-coupling correction from the oracle hardening program. The original v1 SDS/CDS = 0.98 transfer was partly an artifact of shared rolling trajectory between oracles; geometry-only SDS reframes to 4.8 × 10⁻⁸ while preserving the 58× discrimination headline. The correction is documented in the methodology paper and traced in the v2 revision history. Read on arXiv.

Paper II  ·  Sloan's Catalog Extension

Live arXiv:2604.17120 April 2026

The thirteen-member mono-monostatic catalog. The mathematical finding underneath: surface critical points are necessary but not sufficient for mono-monostatic equilibrium. The center-of-mass height function on the unit sphere exhibits four to eleven local minima despite Sloan's analytical phase function having exactly two surface critical points. Sloan's parameterization is extended with a Fourier term and optimized via differential evolution, producing the first openly available catalog of computationally verified instances.

Visualization of the thirteen catalog members across two perturbation families
Figure 2. The thirteen catalog members across two perturbation families. The verified instance at β = 0.023149, a₁ = 0.234433 anchors the family; the remaining twelve span a 7.2× range of surface asymmetry from 0.41% to 2.96% deviation from spherical.
Gentleness-robustness trade-off curve across catalog members
Figure 3. The Pearson r = 0.9993 gentleness–robustness trade-off correlation across the catalog. Members further from spherical achieve faster self-righting but are more sensitive to perturbation; the trade-off is monotone and tightly constrained.

The catalog is positioned as a sharpening of Domokos and Várkonyi 2006, not a refutation. Surface critical points remain the foundational mathematical object; what the catalog adds is the empirical observation that the center-of-mass height landscape on the unit sphere can carry equilibrium structure that the surface itself does not predict. Closing this gap requires the Fourier extension; the catalog is the family of solutions in which the gap is closed.

Read on arXiv.

Paper III  ·  Engineering Applications

Live arXiv:2604.17095 April 2026

Measures the engineering payoff of computationally verified mono-monostatic geometries across four application domains. The headline finding is the IMU calibration housing: 349× precision improvement over conventional cylindrical housings, 0.006° versus 2–3° initial orientation uncertainty under simulated MEMS-IMU deployment.

Center-of-mass height landscape on the unit sphere for a verified catalog instance
Figure 4. The center-of-mass height landscape on the unit sphere for a verified catalog instance. The single minimum is the load-bearing visual: under rest the body settles to one orientation regardless of starting pose. This is the mechanism behind the IMU calibration improvement.

The Gömböc baseplate eliminates orientation uncertainty mechanically, before any software calibration stage. The mechanism is structurally analogous to ZUPT for velocity in inertial navigation: a physical pre-fixing that removes a degree of freedom from the calibration problem before algorithmic correction begins. Conventional cylindrical housings cannot achieve the same mechanism through ballast alone; the paper demonstrates this by exhaustive bottom-weighted mass-concentration sweep up to 30%, none of which produces ECS = 1.

Density perturbation analysis showing cylindrical housing failure modes
Figure 5. Density perturbation analysis across the cylindrical housing family. ECS remains greater than one for all tested mass-distribution configurations, including extreme bottom-weighting. The cylinder is a category, not a tunable parameter, for the mono-monostatic invariant.

Three additional application domains are characterized in less detail. Aerial seed pods achieve 100% correct orientation against an 83.5% cylindrical baseline. Marine buoy self-righting is scoped but not measured experimentally. Standalone capsule comparison gives BOA = 1.000 with SRE = 0.028, a 27× gentler self-righting trajectory than the capsule baseline. Read on arXiv.

Paper IV  ·  TPMS Electrodes

In moderation May 2026 · primary math.MG, cross-listed math-ph

A 240-run FEniCS surface-PDE regeneration across eight triply-periodic minimal surface geometries — gyroid, diamond, primitive, IWP, FRD, Neovius, plus conventional reference geometries — at three cooling regimes spanning three orders of magnitude in heat transfer coefficient. The methodology pivots from peak-of-time-averaged-temperature to expected-peak-temperature across arc positions, addressing a 41.6% coefficient of variation on peak temperature in the original framing that the single-arc Monte Carlo evaluation resolves.

Surface-area times temperature constancy bar chart
Figure 6. The within-category 1.16× constancy and cross-category 1.4× TPMS-vs-conventional topology offset. The headline mechanism finding made visible: thermal advantage is consistent with surface-area-mediated heat dissipation, with the topological structure of triply-connected channels contributing the residual factor.

Diamond edges out gyroid on mean temperature at all three cooling regimes, inverting the gyroid-default framing common in the TPMS thermal literature. The 1.16× within-category constancy and 1.4× cross-category TPMS-versus-conventional topology offset are consistent with surface-area-mediated thermal advantage, with three-dimensional channel-network heat spreading contributing the remaining factor. The face-normal-distribution-uniformity hypothesis remains open as an alternative explanation for the topological residual; discriminating between mechanisms is preserved as future work.

Neovius peak temperature distribution histogram showing the tungsten melting threshold outlier
Figure 7. Neovius peak temperature distribution. One of ten arc positions exceeds the tungsten melting threshold at h = 10³ W/m²K. The topological bottleneck failure mode is the honest negative result of the paper, surfaced in the main body rather than buried in limitations.

Methodology Paper  ·  Oracle Hardening

Awaiting endorsement May 2026 · primary math.NA

Documents the oracle hardening program that ran in parallel with Paper I. The methodology paper makes five load-bearing claims, each supported by two or more independent empirical data points, and includes six explicit "does NOT claim" statements plus six deferred findings preserved for theoretical follow-up. The most striking single result is the discrete Gaussian curvature failure: |Kf| = 1.07 on the oloid mesh at all tested resolutions, while the analytical Gaussian curvature of the oloid is identically zero. The standard discrete K estimator has a topology-mismatched failure mode on developable surfaces; this is documented here for the first time in this specific form.

Discrete curvature convergence failure on oloid mesh
Figure 8. Discrete Gaussian curvature does not converge to zero on the oloid mesh as resolution increases. The standard angle-deficit estimator produces |Kf| ≈ 1.07 at all tested resolutions; analytical K = 0. SDS shifts 17% when the analytical value is substituted.

See the methodology page for full treatment of the five claims and the operating principles they underpin.