betaFeedback

WorksComputer Science

Sloan's Analytical Gömböc at Published ββ: A Strict-Convexity-Constrained Reanalysis

Abstract

Varkonyi and Domokos (2006) proved that convex homogeneous bodies with exactly one stable and one unstable equilibrium point exist. Sloan (2023) gave the first analytical parameterization, with radial function R(θ,ϕ)R(θ,ϕ) having exactly two critical points on S2S^2. This is the v2 amendment-of-record of arXiv:2604.17120. v1 claimed Sloan's parameterization does not produce mono-monostatic bodies and reported a 13-member catalog of Fourier/radial extensions certified at ECS=1 via mesh-vertex drainage-basin analysis. Following correspondence with P. L. Varkonyi (BME), an analytical verification suite was built around the Varkonyi-Gauss identity. Finding 1: Sloan's parameterization does produce mono-monostatic bodies in a strictly-convex sub-regime (β≲0.036β\lesssim 0.036), where Kmin⁡>0K_{\min} > 0 and the identity certifies ECS=1. v1 missed this because its mesh-vertex oracle over-counted on shallow COM-height landscapes. At Sloan's published β=0.05β=0.05, strict convexity is lost (Kmin⁡=−0.569K_{\min}=-0.569; K<0K<0 over 4.01% of surface); the identity's precondition fails. v1's "global surface information" mechanism is replaced by the strict-convexity precondition. Finding 2: Of v1's 13 catalog instances only Phase-1 (β=0.023149β=0.023149, a1=0.234433a_1=0.234433, k=1k=1) survives identity-based verification; the remaining twelve were per-kk optimizer extrema overshooting the strict-convex boundary. Probing the regime interior verifies further mono-monostatic bodies in k=2k=2 and k=3k=3 sub-families: the verified set is an open regime in (β,a1,k)(β, a_1, k), not a discrete list. Finding 3: v1's ECS=1 readings for the 9 radial-family members reflected drainage-basin merging; the r=0.9993r=0.9993 gentleness-robustness correlation is retracted.

they were drawn from. Probing the interior of the Fourier-phase (β,a1,k)(\beta, a_1, k) parameter space verifies additional mono-monostatic bodies in both the k=2k = 2 and k=3k = 3 sub-families (see § 6 and Table 5): the verified set is an open regime in (β,a1,k)(\beta, a_1, k) parameter space, with nontrivial extent in each kk family probed. v1’s twelve non-surviving instances were per-kk optimization extrema that overshot the strict-convexity boundary; the interior of the regime is rich and the identity-based methodology probes it directly. v2 does not enumerate the regime; its contribution is the classification methodology that makes the regime accessible.

kkβ\betaa1a_1Kmin⁡K_{\min}#RR-critECSNotes
10.0231490.234433+0.09121Phase-1 (v1 catalog, survives)
20.03210.025+0.02821Interior, near-boundary in a1a_1
20.0250.050+0.21021Interior, deeper
30.025−0.025+0.34421Interior
30.020−0.050+0.42821Interior, deeper

Table 5.

Finding 3 (mesh-vertex ECS oracle is unreliable for near-spherical bodies). v1’s ECS = 1 measurements for the 9 radial-family members resulted from drainage-basin merging of features with gap below threshold, masking real multi-equilibrium structure. The mesh-vertex h(d)h(d) recipe is replaced in v2 by the identity-based classifier, which uses only the analytical quantities K(θ,ϕ)K(\theta, \phi) and ∇R(θ,ϕ)\nabla R(\theta, \phi). The gentleness–robustness r=0.9993r = 0.9993 correlation in v1, computed across the contaminated catalog, is retracted.

This v2 retains Phase-1 as a verified mono-monostatic body in the Sloan-extension family, demonstrates that the verified mono-monostatic set in the Fourier-phase family is an open regime extending into k=2k = 2 and k=3k = 3 sub-families beyond Phase-1’s k=1k = 1 slice, retains the Sloan negative finding at his published β\beta (now attributed to strict-convexity loss), and contributes a verification methodology (strict-convexity gate by analytical Gaussian curvature; identity-based ECS classification via analytical RR-critical-point counting) that is more rigorous than the mesh-vertex basin counter of v1.

Keywords: mono-monostatic body, Gömböc, equilibrium stability, Várkonyi-Domokos identity, Gauss map, strict convexity, Gaussian curvature, support function, Sloan parameterization

Introduction

A convex homogeneous body with exactly one stable and one unstable equilibrium point under gravity is called mono-monostatic. The existence of such bodies was conjectured by V. I. Arnold in 1995 and proven by Várkonyi and Domokos [1, 2], who named the shape the Gömböc. Physical specimens have been manufactured and are commercially available, but the exact geometric parameters have never been published.

Sloan [3] made significant progress by providing the first analytical parameterization of Gömböc surfaces, expressing the boundary as

r4=1+4βsin⁡θcos⁡(ϕ−P(θ))(1)r^4 = 1 + 4\beta\sin\theta\cos(\phi- P(\theta)) \tag*{(1)}

in spherical coordinates centered at the center of mass, with specific choices of the phase function P(θ)P(\theta) satisfying a center-of-mass constraint. Sloan proved that the radial function R(θ,ϕ)=(r4)1/4R(\theta, \phi) = (r^4)^{1/4} has exactly two critical points on S2S^2 for these parameterizations.

v2 Amendment of Record

This paper is the v2 amendment of arXiv:2604.17120 (v1 published 2026-04-20). v1 made three principal claims: (i) Sloan’s parameterization does not produce mono-monostatic bodies at any tested parameter value; (ii) a thirteen-member catalog of bodies constructed by Fourier-extending Sloan’s phase function and by radial perturbation, all verified at ECS = 1 via a mesh-vertex drainage-basin oracle; (iii) a mechanism explanation for (i) in terms of the COM height function “integrating global surface information beyond what the surface function predicts.”

Following technical correspondence from P. L. Várkonyi (BME) on 2026-05-15, an analytical verification suite was constructed (§ 4). v1’s claim (i) is preserved but for a different mechanism than v1 stated. v1’s claim (ii) is largely retracted: only one of thirteen v1 instances survives identity-based verification. v1’s claim (iii) is incorrect for smooth strictly convex bodies and is replaced with a strict-convexity precondition analysis (§ 2).

§ 8 enumerates the full v1→v2 amendments. The full v1 text and verification scripts are retained in the public repository for reproducibility.

Methodology Note: From Mesh-Vertex Oracle to Identity-Based Classification

v1 measured ECS via drainage-basin analysis on the mesh-vertex discretization of h(d)h(\mathbf{d}). This method has a known threshold-sensitivity problem for near-spherical bodies: when the gap between adjacent basin sinks is comparable to numerical noise, the basin count depends on the merge threshold. v1 absorbed this sensitivity into a 0.5–10% threshold robustness check, but the analytical verification suite in v2 reveals that several v1 catalog members had real multi-basin structure with sink-height gap below v1’s merge threshold; these were absorbed into “ECS=1” incorrectly.

v2 replaces the mesh-vertex oracle with an identity-based classifier (§ 4.3) that uses only analytical quantities K(θ,ϕ)K(\theta,\phi) (Gaussian curvature) and the critical-point structure of R(θ,ϕ)R(\theta,\phi). This is rigorous because, under the Gauss-map identity of § 2, these analytical quantities uniquely determine the equilibrium structure of a smooth strictly convex body.

The Várkonyi-Gauss Identity and its Strict-Convexity Precondition

For a smooth convex body B⊂R3B \subset\mathbb{R}^{3} with center of mass at the origin, the radial function R:S2→R+R:S^{2}\to\mathbb{R}_{+} and the support function H:S2→R+H:S^{2}\to\mathbb{R}_{+} are defined by

R(u)=max⁡{t>0:tu∈B},H(n)=max⁡v∈Bv⋅n.(2)R(\mathbf{u})=\max\{t>0:t\mathbf{u}\in B\},\qquad H(\mathbf{n})=\max_{\mathbf{v}\in B}\mathbf{v}\cdot\mathbf{n}. \tag*{(2)}

The Gauss map G:∂B→S2G:\partial B\to S^{2} assigns each boundary point its outward unit normal. For strictly convex smooth bodies, GG is a diffeomorphism. In this regime the two representations are related by a classical identity of convex geometry (see, e.g., [9 §1.7]):

∇S2H(n)=−R(u)∇S2R(u)R(u)2+∥∇S2R(u)∥2(3)\nabla_{S^{2}}H(\mathbf{n})=-\frac{R(\mathbf{u})\nabla_{S^{2}}R(\mathbf{u})}{\sqrt{R(\mathbf{u})^{2}+\lVert\nabla_{S^{2}}R(\mathbf{u})\rVert^{2}}} \tag*{(3)}

with u\mathbf{u} and n\mathbf{n} related by the Gauss map. Since R>0R>0 on a body containing the origin, equation (3) yields the bijection

∇S2H(n)=0⟺∇S2R(u)=0.(4)\nabla_{S^{2}}H(\mathbf{n})=0\Longleftrightarrow\nabla_{S^{2}}R(\mathbf{u})=0. \tag*{(4)}

At a critical point u∗\mathbf{u}^{*} of RR, the outward normal direction equals the radial direction, so n=u∗\mathbf{n}=\mathbf{u}^{*}; local maxima of RR correspond to local maxima of HH in the same direction.

Theorem 1 (ECS via RR-critical-point count, under strict convexity). Let BB be smooth, strictly convex, with COM at the origin. Let ECS⁡(B)\operatorname{ECS}(B) denote the number of local minima of the COM-height function h(d)=H(−d)h(\mathbf{d})=H(-\mathbf{d}) on S2S^{2}. Then ECS⁡(B)\operatorname{ECS}(B) equals the number of local minima of RR on S2S^{2}, which equals the number of local maxima of RR on S2S^{2} when the total count is ≤2\le2 (more generally, #{min⁡R}\#\{\min R\} under the standard sign convention).

The theorem follows from (4) combined with the orientation-of-extrema correspondence under the Gauss map. ECS = 1 thus reduces to “RR has exactly one local minimum on S2S^{2},” which (combined with the Morse-Euler relation #min⁡−#saddle+#max⁡=2\#\min-\#\text{saddle}+\#\max=2 for smooth functions on S2S^{2}) forces #R\#R-critical-points = 2 with no saddles.

Precondition Failure: Sloan's Published β\beta Values

The identity (3) requires strict convexity. Sloan’s parameterization is strictly convex for small β\beta but loses strict convexity above a threshold. We compute Gaussian curvature K(θ,ϕ)K(\theta,\phi) analytically from R(θ,ϕ)R(\theta,\phi) using symbolic partial differentiation (sympy), giving the first and second fundamental forms EE, FF, GG, LL, MM, NN and

K(θ,ϕ)=LN−M2EG−F2.(5)K(\theta,\phi)=\frac{LN-M^{2}}{EG-F^{2}}. \tag*{(5)}

The body is strictly convex iff K(θ,ϕ)>0K(\theta,\phi)>0 everywhere on the surface. (This is a strictly stronger criterion than the mesh-volume/convex-hull-volume ratio used in v1, which is satisfied by bodies with small K<0K<0 regions.)

Applying (5) to Sloan’s parameterization with P(θ)=η(θ)=3π2(cos⁡θ−13cos⁡3θ)P(\theta)=\eta(\theta)=\frac{3\pi}{2}\left(\cos\theta-\frac{1}{3}\cos^{3}\theta\right):

At β=0.05\beta=0.05 (where v1 observed minimum ECS = 2 on the mesh-vertex oracle), the precondition fails over 4% of surface area, and Kmin⁡=−0.569K_{\min}=-0.569 lies well below the threshold: Sloan’s published β=0.05\beta=0.05 is not a marginal precondition failure but a substantial one (the convexity-loss transition is near β≈0.036\beta\approx0.036, with Kmin⁡K_{\min} already negative at $\beta=0.040). For these bodies, the Gauss map is not bijective: multiple surface points share normal directions, and the support function HH acquires critical points beyond those predicted by ∇R=0\nabla R=0 alone. This is the correct mechanism for Sloan’s published instances failing mono-monostatic; it is not a generic property of smooth convex bodies, but a specific consequence of strict-convexity loss.

Behavioral Confirmation: Multiple Maximizers

A second behavioral demonstration of strict-convexity loss: for a strictly convex body, the support point achieving H(n)H(\mathbf{n}) in any direction is unique. For Sloan β=0.05\beta=0.05 at n=(−1,0,0)\mathbf{n}=(-1,0,0), we find by Powell-tight optimization from many starts that two distinct surface points both achieve H(n)=0.9507H(\mathbf{n})=0.9507: at (θ,ϕ)=(1.36,3.10)(\theta,\phi)=(1.36,3.10) and at (1.78,3.19)(1.78,3.19), with the supposed-critical-point (π/2,π)(\pi/2,\pi) giving only H=0.9457H=0.9457 (a saddle, not the extremum). The support-point map is multi-valued; strict convexity is violated.

Sloan’s Parameterization: A Strict-Convexity-Bounded Mono-Monostatic Regime

This section is the most substantively revised portion of v2. v1’s central empirical claim was that Sloan’s parameterization “does not produce mono-monostatic bodies at any tested parameter value.” The identity-based reanalysis revises this to a more precise statement: Sloan’s parameterization produces mono-monostatic bodies in a β\beta-bounded strictly-convex regime, and fails outside this regime through loss of strict convexity.

Sloan Produces Mono-Monostatic Bodies at Small β\beta

Sloan [3] proves that the radial function R(θ,ϕ)=[1+4βsin⁡θcos⁡(ϕ−η(θ))]1/4R(\theta,\phi)=\left[1+4\beta\sin\theta\cos(\phi-\eta(\theta))\right]^{1/4} has exactly two critical points on S2S^{2} at all β\beta. Combined with Theorem 1, this proves that whenever Sloan’s body is strictly convex, it is mono-monostatic. The strictly-convex regime is identified by the analytical Gaussian curvature gate (§ 4.1). A dense β\beta-sweep localizes the convexity-loss transition to β≈0.036\beta\approx0.036: at β=0.035\beta= 0.035, Kmin⁡=+0.0178K_{\min} = +0.0178 (still strictly convex, no K<0K < 0 region at 640×1280 resolution); at β=0.040\beta= 0.040, Kmin⁡=−0.164K_{\min} = -0.164 with K<0K < 0 over 1.45% of surface (strict convexity lost). At all tested β≤0.035\beta\leq0.035, RR retains exactly two critical points (1 maximum + 1 minimum, no saddles, Morse-Euler M−S+m=2M - S + m = 2 verified analytically), so the entire interval β≲0.036\beta\lesssim0.036 is verified mono-monostatic via the identity.

At β=0.02\beta= 0.02 and β=0.03\beta= 0.03, Sloan’s body is mono-monostatic via the identity-based method. v1 missed this: v1’s mesh-vertex oracle reported ECS = 3 at β=0.02\beta= 0.02 and ECS = 7 at β=0.005\beta= 0.005, treating these counts as real and concluding that “Sloan’s analytical Gömböc does not produce mono-monostatic bodies.” Those counts are drainage-basin over-counts on the shallow COM-height landscapes of near-spherical bodies, not real equilibria of the smooth surface. The #RR-critpts = 2 structure plus Kmin⁡>0K_{\min} > 0 is dispositive under Theorem 1.

This is a substantive correction to v1’s headline. The corrected statement: Sloan’s analytical parameterization produces mono-monostatic bodies for β\beta in the strictly-convex regime, β≲0.036\beta\lesssim0.036. The mono-monostatic property is certified by the identity-based method (strict-convexity gate + RR-critical-point count) and does not require mesh-vertex basin counting.

Sloan's Published Instances Lie Outside the Strict-Convex Regime

Sloan [3] published two specific Gömböc instances: Gömböc 1 at P(θ)=5θP(\theta) = 5\theta, β≤0.15\beta\leq0.15, and Gömböc 2 at P(θ)=η(θ)P(\theta) = \eta(\theta), β≤0.17\beta\leq0.17. Both published β\beta ranges include values where strict convexity fails. Specifically for Gömböc 2 at β=0.05\beta= 0.05, the curvature analysis (§ 2.1, Table 1) shows Kmin⁡=−0.569K_{\min} = -0.569 with K<0K < 0 over 4.01% of surface area.

β\betaKmin⁡K_{\min}Frac. K<0K<0Strictly convex?
0.02+0.4940.00%Yes
0.03+0.1880.00%Yes
0.05−0.5694.01%No
0.08−2.2248.68%No

Table 1. Analytical Gaussian curvature on Sloan Gömböc 2, computed by symbolic partial differentiation of RR and evaluation of (5) on 640 × 1280 grid (820k points). The strict-convexity precondition of identity (3) holds at β≤0.03\beta\le0.03 and fails by β=0.05\beta=0.05; the crossover lies in the range where Sloan’s published instances sit.

The behavioral confirmation of § 2.1 applies dispositively here: at β=0.05\beta= 0.05 and direction n=(−1,0,0)\mathbf{n} = (-1, 0, 0), two distinct surface points share the outward normal, both giving H=0.9507H = 0.9507, while the radial-critical point at (π/2,π)(\pi/2, \pi) gives only H=0.9457H = 0.9457 (a saddle of v⋅n\mathbf{v} \cdot\mathbf{n}, not its extremum). The Gauss map is not bijective at β=0.05\beta= 0.05. Sloan’s argument relating ∇R=0\nabla R = 0 to mono-monostatic structure presupposes the Gauss map’s bijectivity; at his published β\beta values, this presupposition fails.

For these instances, ECS via the identity-based method is indeterminate. The mesh-vertex oracle’s reports of 2 (at β=0.05\beta= 0.05) and 7 (at β=0.08\beta= 0.08) are also not reliable: at non-strictly-convex bodies the drainage-basin oracle measures features of a discretized multi-valued Gauss map rather than smooth-surface equilibria. Whether Sloan’s published instances are mono-monostatic by some other measure (e.g., a direct analysis of the multi-valued Gauss map or a regularization argument) is an open question.

Reconciliation with v1’s Headline Claim

v1’s title was “Sloan’s Analytical Gömböc Does Not Produce Mono-Monostatic Bodies.” The corrected statement is: Sloan’s analytical parameterization does produce mono-monostatic bodies in the strictly-convex sub-regime (β≲0.036\beta\lesssim0.036), but the published Sloan Gömböc 2 at β=0.05\beta= 0.05 does not, because strict convexity is lost (Kmin⁡=−0.569K_{\min} = -0.569, well below the threshold). v1 missed both halves of this picture: it over-counted at small β\beta (false ECS>1> 1 on strictly-convex mono-monostatic bodies) and correctly identified ECS>1> 1 at published β\beta but for the wrong mechanism (drainage-basin multi-counting on a non-convex surface rather than the precondition failure of the identity).

The Analytical Verification Suite

The verification methodology of v2 consists of three analytical components, all implemented in the public repository and reproducible from a fresh clone.

Strict-Convexity Gate via Analytical Gaussian Curvature

Script: scripts/verify_K_analytical.py. For a radial parameterization R(θ,ϕ)R(\theta,\phi), we construct the surface map x(θ,ϕ)=R(θ,ϕ)u(θ,ϕ)\mathbf{x}(\theta,\phi) = R(\theta,\phi)\mathbf{u}(\theta,\phi) symbolically using sympy. Partial derivatives xθ\mathbf{x}_{\theta}, xϕ\mathbf{x}_{\phi}, xθθ\mathbf{x}_{\theta\theta}, xθϕ\mathbf{x}_{\theta\phi}, xϕϕ\mathbf{x}_{\phi\phi} are computed symbolically, giving the first fundamental form (E,F,G)(E,F,G) and second fundamental form (L,M,N)(L,M,N) in closed form. Gaussian curvature K=(LN−M2)/(EG−F2)K = (LN - M^2)/(EG - F^2) is then lambdified to a numpy callable and evaluated on a dense (θ,ϕ)(\theta,\phi) grid. We use a 640×1280640 \times1280 grid by default (820k points); convergence checks at 160×320160 \times320 and 320×640320 \times640 confirm that Kmin⁡K_{\min} is stable to four decimal places across resolutions. The gate decision is Kmin⁡>0K_{\min} > 0; bodies passing are strictly convex, bodies failing have one or more surface regions where the second fundamental form is degenerate or indefinite.

Analytical RR-Critical-Point Counting with Hessian Classification

Script: scripts/verify_R_critpoints_analytical.py. Symbolic partial derivatives ∂R/∂θ\partial R/\partial\theta and ∂R/∂ϕ\partial R/\partial\phi are computed via sympy; the 2D gradient system is lambdified to a numpy callable. The system ∇R(θ,ϕ)=0\nabla R(\theta,\phi) = 0 is solved by scipy.optimize.root (hybrid Powell) from a 24×4824 \times48 grid of (θ,ϕ)(\theta,\phi) starts (1152 total). Solutions are deduplicated by spherical distance (<0.02< 0.02 rad considered the same point). Each critical point is classified as maximum/minimum/saddle/degenerate by the sign of the Hessian determinant and trace. The Morse-Euler relation #min−#saddle+#max=χ(S2)=2\#\mathrm{min} - \#\mathrm{saddle} + \#\mathrm{max} = \chi(S^2) = 2 is checked as a self-consistency condition that all critical points have been found.

Identity-Based ECS Classifier

Script: scripts/identity_based_classifier.py. For each candidate body, the pipeline is:

  1. Run 4.1 to check Kmin⁡>0K_{\min} > 0.

  2. If yes, run 4.2 to count RR critical points and classify each. By Theorem 1, ECS =#R= \#R-minima, and mono-monostatic ⇔#R\Leftrightarrow\#R-critpts =2= 2.

  3. If no, the identity precondition fails; ECS is indeterminate by this method.

This pipeline uses no mesh-vertex computation. All quantities are analytical functions of R(θ,ϕ)R(\theta,\phi), which is itself an elementary closed-form expression in β\beta, perturbation coefficients, and (θ,ϕ)(\theta,\phi).

Methodology Discipline Artifacts

Two artifacts at the v2 sandbox root encode discipline contributions that emerged through the v1→v2 amendment process:

preconditions_check.md: enumerates the preconditions of every theorem v2 applies (Sloan’s RR-critical-point structure, the Várkonyi-Gauss identity, the Várkonyi-Domokos existence proof). For each precondition, the file points to the script that verifies it on every body the paper analyzes. The strict-convexity precondition silently inherited from Sloan in v1 (and surfaced by Várkonyi correspondence) is the originating incident; the artifact prevents recurrence.

claims_to_scripts.md: maps every quantitative claim in this main.tex to the script in scripts/ that produces the claimed number, plus the JSON output anchoring the specific value cited. Consistency-pass grep through main.tex for “we comput,” “analytical,” “convergence,” “Kmin⁡K_{\min},” “#R-crit” etc. verifies each phrase maps to an existing script. v1’s phantom “analytical confirmation at 2000 directions” paragraph (no backing script) is the originating incident; the artifact prevents recurrence.

Both artifacts are mandatory for future Substrate Geometry papers per standard practices brain S2.14 (verify preconditions of borrowed theorems) and S2.15 (every quantitative claim backed by reproducible script). The artifacts emerged from the Várkonyi pushback incident of 2026-05-15 and are described in detail in the standard practices brain.

The Phase-1 Verified Mono-Monostatic Body

We extend Sloan’s phase function with Fourier terms:

P(θ)=η(θ)+a1sin⁡(η(θ))(6)P(\theta) = \eta(\theta) + a_1 \sin(\eta(\theta)) \tag*{(6)}

At β=0.023149\beta= 0.023149, a1=0.234433a_1 = 0.234433, the resulting body (“Phase-1”) passes the verification suite.

ParameterValue
β\beta0.023149
a1a_10.234433
Analytical Kmin⁡K_{\min} (640 ×\times 1280 grid)+0.0907
Fraction K<0K < 00.0000
Strict-convexity gatePASS
#RR-critical-points2
RR-max location (θ,ϕ)(\theta, \phi)(π/2,0)(\pi/2, 0)
RR-min location (θ,ϕ)(\theta, \phi)(π/2,π)(\pi/2, \pi)
RR-max value1.0224
RR-min value0.9760
ECS via Theorem 11
Mono-monostaticYes

Table 3. Phase-1 verification report.

This is the unique catalog member from v1 that passes the identity-based gate. It is the verified contribution of this paper: an analytical, openly-published, strictly-convex mono-monostatic body in the Sloan-extension family, with parameters (β,a1)(\beta, a_1) computable to arbitrary precision from the closed-form R(θ,ϕ)R(\theta, \phi) and machine-checkable strict-convexity and equilibrium counts.

Honest Reclassification of the v1 Catalog and the Verified Regime Interior

We apply the verification suite to all thirteen v1 catalog instances. Results in Table 4. The reclassification establishes that v1’s specific instance picks do not survive in their published form. § 6.2 then demonstrates that this is not a statement about the underlying families: probing the interior of the Fourier-phase parameter space verifies additional mono-monostatic bodies in the k=2k = 2 and k=3k = 3 sub-families. The non-survival of v1’s twelve specific points is a property of where v1’s optimizer sampled (per-kk asymmetry extrema near the boundary of the strict-convex regime), not of the families themselves.

#v1 Nameβ\betaKmin⁡K_{\min}#RR-critECSv2 Classification
1Phase-1 (sin⁡η\sin\eta)0.0231+0.09121MM (verified)
2Phase-2 (sin⁡2η\sin2\eta)0.0321−0.431——Indeterminate (not strictly convex)
3Phase-3 (sin⁡3η\sin3\eta)0.0517−0.889——Indeterminate (not strictly convex)
4Radial-f3f_3 primary0.0231+0.4382Polystatic
5Radial-f4f_4 primary0.0231+0.3862Polystatic
6Radial-f3f_3 scan0.0231+0.4382Polystatic
7Radial-f4f_4 scan0.0231+0.3962Polystatic
8Radial-f4f_4 low-β\beta0.0150+0.6362Polystatic
9Radial-f4f_4 high-β\beta0.0350−0.02——Indeterminate (not strictly convex)
10Radial-f3f_3 low-β\beta0.0150+0.6582Polystatic
11Min-asym β=0.01\beta= 0.010.0100+0.7782Polystatic
12Min-asym β=0.008\beta= 0.0080.0080+0.8282Polystatic
13Min-asym β=0.005\beta= 0.0050.0050+0.8582Polystatic

Table 4. Identity-based reclassification of v1 catalog members. “MM” = mono-monostatic (passes identity-based gate: strictly convex AND RR has exactly 2 critical points). “Polystatic” indicates the body is strictly convex but RR has >2> 2 critical points, yielding multiple stable equilibria (ECS via Theorem 1 is the RR-minima count, listed in the ECS column). “Indeterminate” indicates strict-convexity precondition fails; ECS via the identity-based method is not certified.

Why the v1 Mesh-Vertex Oracle Reported ECS=1 for the Polystatic Instances

v1 evaluated h(d)h(\mathbf{d}) as c⋅d−min⁡v∈vertsv⋅d\mathbf{c}\cdot\mathbf{d} - \min_{\mathbf{v}\in\mathrm{verts}} \mathbf{v}\cdot\mathbf{d}, i.e., the support function approximated by the maximum over the finite vertex set, then identified drainage basins on a Fibonacci direction grid, merging basins with sink-height gap below 1% of total hh-range.

For the polystatic catalog members, the actual gap between adjacent basin sinks is of order 10−310^{-3} to 10−410^{-4} in absolute height, which is roughly 1–5% of total hh-range for the near-spherical instances. v1’s 0.5–10% threshold sweep merged these into a single basin, yielding ECS = 1. The threshold-robustness check in v1 (“ECS=1 stable across 0.5–10%”) was satisfied because the smallest threshold tested (0.5%) was above the gap; below this threshold, additional basins would have appeared.

The identity-based classifier does not have this failure mode: it counts critical points of an analytical scalar field, not basins of a discretized one. The Morse-Euler check #min − #saddle + #max = 2 is self-verifying.

The Verified Regime Interior: k=2k = 2 and k=3k = 3 Sub-Families Beyond Phase-1

The reclassification of v1’s twelve non-surviving instances raised a question that v1 did not pose and v2 must answer: are the Fourier-phase sub-families k=2k = 2 and k=3k = 3 empty of verified mono-monostatic bodies, or did v1’s optimizer simply sample at points that overshot the strict-convexity boundary in those families?

v1’s Phase-2 picked (β,a1,k)=(0.0321,0.137606,2)(\beta, a_1, k) = (0.0321, 0.137606, 2) and Phase-3 picked (β,a1,k)=(0.0517,−0.055157,3)(\beta, a_1, k) = (0.0517, -0.055157, 3). Both were outputs of an optimizer maximizing an asymmetry-like objective subject to v1’s then-incorrect mesh-vertex mono-monostatic gate; both lie outside the strict-convex regime. Backing a1a_1 off toward zero in each family produces points that pass the strict-convexity gate and retain RR-critical-point count = 2. Table 5 reports analytical verifications at three interior points per non-trivial sub-family.

Table 5: Interior points of the verified mono-monostatic regime in the Sloan Fourier-phase family. All entries pass the identity-based gate: Kmin⁡>0K_{\min} > 0 analytically (sympy on 640×1280640 \times1280 grid) and RR has exactly two critical points (analytical gradient + scipy root-finding + Hessian classification, Morse-Euler M−S+m=2M - S + m = 2). Each is a distinct mono-monostatic body in the r4=1+4βsin⁡θcos⁡(ϕ−η(θ)−a1sin⁡(kη(θ)))r^4 = 1 + 4\beta\sin\theta\cos(\phi- \eta(\theta) - a_1\sin(k\eta(\theta))) family.

Three observations from Table 5:

  1. The verified regime is non-empty in each kk sub-family tested. v1’s Phase-2 and Phase-3 failures to verify (Table 4) reflect their specific optimization-extremum coordinates, not emptiness of the underlying families.

  1. The verified regime extends beyond a one-dimensional curve in (β,a1)(\beta, a_1) within each kk slice. The two k=2k = 2 interior points and the two k=3k = 3 interior points lie at distinct (β,a1)(\beta, a_1) coordinates and both verify, consistent with the verified set having non-trivial extent in the parameter plane. We do not claim a positive-measure result; characterizing the regime’s extent is open work.

  1. v1’s twelve non-surviving instances were boundary-pushing picks: per-kk asymmetry extrema sit at or just beyond the strict-convexity boundary in a1a_1 (for k=2k = 2) or in β\beta (for k=3k = 3). The optimizer pushed each instance to the boundary of v1’s incorrect gate, which is approximately the boundary of the actual identity-based gate plus a margin; backing off restores the verification.

This v2 does not enumerate the regime exhaustively. The classification methodology — strict-convexity gate plus analytical RR-critical-point counting — probes the regime directly at any specified (β,a1,k)(\beta, a_1, k), and the regime is open in parameter space, so any reader can construct additional verified instances at will. The methodology contribution of v2 is therefore not a catalog of points but a classification procedure that converts the question of mono-monostatic membership into a finite, reproducible analytical computation.

Retraction of the Gentleness-Robustness Trade-off

v1 reported a near-perfect correlation r=0.9993r = 0.9993 between COM height range and self-righting energy, computed across the thirteen catalog instances. Twelve of those thirteen instances are reclassified as polystatic or indeterminate. The correlation was a property of the contaminated sample, not of the mono-monostatic family. This claim is retracted.

The polystatic instances may still exhibit a similar correlation within their class; this is left for future work. They are not Gömböcs.

Discussion

What v2 Establishes

Four contributions:

  1. Mechanism correction. Sloan’s analytical parameterization does not produce mono-monostatic bodies at the published β\beta values because those bodies are not strictly convex (analytical Kmin⁡=−0.57K_{\min} = -0.57 at β=0.05\beta= 0.05). The Várkonyi-Gauss identity (3) requires strict convexity; at Sloan’s parameters it does not apply. v1’s claimed mechanism (“support function integrates global surface information”) is corrected.

  1. Phase-1 verified. A single analytical mono-monostatic body in the Sloan-extension family is verified rigorously: strict convexity by analytical Gaussian curvature, mono-monostatic property by analytical RR-critical-point counting with Morse-Euler check.

  1. Catalog retraction. Twelve of v1’s thirteen verified instances are reclassified: nine as polystatic (strictly convex but #R\#R-critpts >2> 2), three as indeterminate (not strictly convex; identity precondition fails). The “thirteen verified mono-monostatic bodies” framing is retracted. The gentleness-robustness r=0.9993r = 0.9993 correlation is retracted.

  1. Verification methodology. An identity-based ECS classifier replaces the mesh-vertex drainage-basin oracle. The classifier uses analytical Gaussian curvature for the strict-convexity gate and analytical RR-critical-point counting for the equilibrium structure. Scripts in the repository are reproducible from fresh clone.

Relationship to Várkonyi-Domokos and to the Sloan Parameterization

Várkonyi-Domokos [1] establishes existence of mono-monostatic convex bodies. v2 verifies a specific instance (Phase-1) in the Sloan-extension family. The verification is analytical (via the Várkonyi-Gauss identity (3), which is implicit in the Várkonyi-Domokos proof and explicit in standard convex-body geometry textbooks [9]). The analytically certified body Phase-1 is, to our knowledge, the first openly published smooth mono-monostatic body with a verified analytical parameterization and machine-checkable strict-convexity and equilibrium counts.

Sloan’s analytical parameterization [3] provides a two-critical-point radial function; this is the same condition identified by Theorem 1 as necessary for mono-monostatic structure under strict convexity. Sloan’s argument implicitly relies on strict convexity (so that the Gauss map is bijective and surface critical points correspond to height-function critical points). At Sloan’s published β\beta values, this implicit assumption fails by 4–9% of surface area. v2 makes the assumption explicit (strict-convexity gate) and identifies the regime (β≲0.036\beta\lesssim0.036, localized by dense β\beta-sweep in Table 2) where the construction satisfies it.

β\betaKmin⁡K_{\min}#RR-critECS via identityv1 mesh-vertex ECS
0.010+0.76321 (mono-monostatic)not in v1 sweep
0.020+0.49421 (mono-monostatic)3 (over-count artifact)
0.025+0.34621 (mono-monostatic)not in v1 sweep
0.030+0.18821 (mono-monostatic)not in v1 sweep
0.035+0.01821 (mono-monostatic)not in v1 sweep
0.040−0.1642— (precondition fails)not in v1 sweep
0.045−0.3592— (precondition fails)not in v1 sweep
0.050−0.5692— (precondition fails)2
0.080−2.2242— (precondition fails)7

Table 2. Identity-based ECS classification of Sloan Gömböc 2 across β\beta. ECS via Theorem 1: count RR-critical-points analytically, classify by Hessian. “–” indicates indeterminate ECS via identity (strict-convexity precondition fails).

Open Questions

  1. Whether Sloan’s published β\beta values are mono-monostatic via a non-Gauss-map argument (e.g., direct analysis of the multi-valued normal at K<0K < 0 regions) is unknown.

  1. Whether other phase-function families (beyond aksin⁡(kη)a_k \sin(k\eta)) produce additional analytical mono-monostatic bodies satisfying the identity-based gate is a constructive question. Optimization should target #R\#R-critpts =2= 2 directly, not basin-counting.

  1. The polystatic catalog members may have engineering applications distinct from the Gömböc; characterizing them rigorously as members of a different geometric family is open work.

Summary of v1 to v2 Amendments

This section provides the full enumeration of v1 to v2 amendments for reference.

  1. Title and subtitle. Title changed from “Sloan’s Analytical Gömböc Does Not Produce Mono-Monostatic Bodies” to “Sloan’s Analytical Gömböc at Published β\beta: A Strict-Convexity-Constrained Reanalysis,” reflecting the small-β\beta rescue of Sloan’s construction (the negative finding applies specifically to the published β\beta values). Subtitle changed from “Computational Verification, Extended Construction, and a Thirteen-Member Verified Catalog” to “Identity-Based Verification, Strict-Convexity Gating, and a Verified Mono-Monostatic Regime.”

  1. Abstract. Rewritten to reflect findings, retractions, and methodology change. Three numbered findings now structure the abstract.

  1. v1 ECS Oracle definition (v1 §2). The mesh-vertex drainage-basin definition of ECS is retained for reference but is no longer the primary metric. v2 uses identity-based ECS classification (§ 4.3). The threshold-robustness check in v1 is shown to have a known failure mode for near-spherical bodies (§ 6).

  1. v1 §3.2 “Analytical Confirmation” paragraph. v1 stated: “we computed h(d)h(\mathbf{d}) analytically from (1) at 2000 directions using multi-start optimization for the support point at each direction. The analytical computation confirms 9–21 distinct basins for the two Sloan instances.” This claim was not backed by any script in the v1 repository. The paragraph is removed in v2 and replaced by the analytical K+RK + R-critical-point measurements (§ 4), which are backed by scripts.

  1. v1 §3.3 “The Gap” framing. v1 stated that “the COM height function integrates global surface information through the support-point trajectory and can exhibit local minima that the surface function does not predict.” For smooth strictly convex bodies, this is false: the Várkonyi-Gauss identity bijectively links critical points of RR and of HH. v2 replaces this section with § 2 and § 2.1, identifying strict-convexity loss as the actual mechanism for Sloan’s published instances.

  1. Convexity metric. v1 used mesh_vol/convex_hull_vol > 0.999 as the convexity check. v2 uses analytical Kmin⁡>0K_{\min} > 0 on a 640 × 1280 grid (§ 4.1). The v1 metric passes bodies with several percent K<0K < 0; the v2 metric does not.

  1. v1 catalog (Table 2). Reclassified (§ 6): of v1’s thirteen specific instances, 1 mono-monostatic (Phase-1), 9 polystatic, 3 indeterminate. The “thirteen verified mono-monostatic bodies” framing is retracted. v2 establishes (§ 6.2, Table 5) that this is a statement about v1’s specific picks (per-kk asymmetry extrema that overshot the strict-convexity boundary), not about the underlying families: the verified mono-monostatic set is an open regime in (β,a1,k)(\beta, a_1, k) parameter space, with verified interior bodies confirmed analytically in the k=2k = 2 and k=3k = 3 sub-families beyond Phase-1.

  1. Gentleness–robustness correlation (r=0.9993r = 0.9993). Retracted (§ 6). Computed across the contaminated catalog; not a property of mono-monostatic structure.

  1. S–U angle and feasible region topology discussion. v1 §5.4 and §5.5 discussed stable-unstable angle and feasible-region topology across the thirteen catalog members. Since twelve of those members are now reclassified, these discussions are retracted to the polystatic subset and deferred for future work.

  1. Scripts and reproducibility. v2 adds (a) verify_K_analytical.py (analytical KK via sympy), (b) verify_R_critpoints_analytical.py (analytical ∇R=0\nabla R = 0 root-finding with Hessian classification), (c) identity_based_classifier.py (full pipeline), (d) preconditions_check.md (theorem-precondition audit), and (e) claims_to_scripts.md (mapping of every quantita񎟙

tive claim in main.tex to the producing script). All scripts run from a fresh clone.

Honest Scope and Future Work

Computational, Not Mathematical

We do not provide a formal mathematical proof of Phase-1’s mono-monostatic property; we provide analytical verification via the Várkonyi-Gauss identity (3), applied to a body whose strict-convexity precondition is itself verified analytically. Formal proof would require either a direct construction following the Várkonyi-Domokos [1] existence-proof scaffold for the specific Phase-1 parameters, or a characterization of the parameter region within (β,a1)(\beta, a_1)-space where the construction yields mono-monostatic bodies. Both are open mathematical work.

Identity-Based Method Limitations

The identity-based classifier is rigorous for smooth strictly convex bodies, where the Gauss map is a diffeomorphism. It does not apply directly to bodies that lose strict convexity. For the three indeterminate v1 catalog members (Phase-2, Phase-3, Radial-f4f_4 high-β\beta), determining ECS requires either (a) direct analysis of the multi-valued Gauss map at K≤0K \le0 regions, or (b) a different verification methodology (physical experiment, e.g.). Both are open work.

Future Directions

  1. Larger verified catalog under the identity-based gate. Optimize (β,a1,…)(\beta, a_1, \ldots) directly against the identity-based criterion (Kmin⁡>0K_{\min} > 0 AND #RR-critpts =2= 2), rather than against the threshold-sensitive basin counter of v1. The phase-extension family (§ 5) is a natural starting point.

  2. Polystatic catalog. The 9 reclassified polystatic instances are interesting geometric objects in their own right; characterizing their stable-equilibrium structure analytically and exploring engineering applications distinct from Gömböcs is open work.

  3. Physical validation of Phase-1. 3D-printing or CNC-machining Phase-1 at the <0.01%< 0.01\% tolerance required to preserve the mono-monostatic property remains future work.

  4. Mathematical proof. As noted in § 9.1.

Reproducibility

All quantitative claims in this paper are reproducible from a fresh clone of the public repository. The v2 sandbox is at gomboc_family/v2/; relevant paths below are relative to this directory.

Software stack. Python 3.12+, sympy 1.14.0 (symbolic differentiation), scipy 1.16+ (root-finding via scipy.optimize.root with method hybr; Hessian classification via numpy.linalg), numpy 2.0+, scipy.spatial.cKDTree (Fibonacci-sphere nearest-neighbor graphs in legacy diagnostics). MiKTeX or any TeX Live distribution for the LaTeX build (pdflatex main.tex twice for cross-refs).

Reproducing the verification suite. From gomboc_family/v2/:

  • Sloan β\beta-sweep Kmin⁡K_{\min}: python scripts/verify_K_analytical.py -family sloan -beta $BETA -out results/V_sloan_$BETA.json for β∈{0.02,0.03,0.05,0.08}\beta\in\{0.02, 0.03, 0.05, 0.08\}.

  • Phase-1 Kmin⁡K_{\min}: python scripts/verify_K_analytical.py -family sloan-phase -beta 0.023149 -coef 0.234433 -k 1 -out results/V2_analytical_K_phase1.json.

  • Phase-1 RR-critical-points: python scripts/verify_R_critpoints_analytical.py -family sloan-phase -beta 0.023149 -coef 0.234433 -k 1.

  • Radial-family RR-critical-points (e.g. Min-asym β=0.005\beta= 0.005, f3f_3): `python scripts/verify_R_critpoints_analytical.py -family sloan-radial -beta 0.005 -coef 0.05102 -f-type f3 -out results/V4_R_critpoints_mina

  • Two-maximizer test on Sloan β=0.05\beta= 0.05: python scripts/two_maximizer_test.py -family sloan -beta 0.05 -out results/two_maximizer_sloan_b005.json.

Runtime. The KK analytical sweep at 640×1280640 \times1280 grid runs in ∼100\sim100 seconds per body on a single CPU. The RR-critical-point root-finding from a 24×4824 \times48 grid of starts runs in ∼1\sim1 second per body. The full identity-based catalog reclassification of 13 v1 members runs in approximately 45 minutes on a single CPU.

Output anchoring. Each table in this paper has an accompanying entry in claims_to_scripts.md mapping the cited numbers to specific JSON files in results/. The consistency-pass discipline (S2.15) verifies the mapping before arXiv submission.

Data Availability

All verification scripts, the Phase-1 mesh (STL), and v1 mesh artifacts for the reclassified catalog members are openly available at https://github.com/gyapaganda-a11y/substrate-geometry. The v1 paper text and codebase are retained in the repository for reproducibility of the original claims; this v2 supersedes v1 as the amendment-of-record. The v2 sandbox at https://github.com/gyapaganda-a11y/substrate-geometry/tree/main/gomboc_family/v2 contains all v2-specific artifacts (main.tex, preconditions_check.md, claims_to_scripts.md, scripts/, results/).

References

  1. [1]P. L. Várkonyi and G. Domokos, “Mono-monostatic bodies: the answer to Arnold’s question,” The Mathematical Intelligencer, vol. 28, no. 4, pp. 34–38, 2006.DOI
  2. [2]P. L. Várkonyi and G. Domokos, “Static equilibria of rigid bodies: dice, pebbles, and the Poincaré-Hopf theorem,” Journal of Nonlinear Science, vol. 16, pp. 255–281, 2006.DOI
  3. [3]M. L. Sloan, “An analytical Gomboc,” arXiv:2306.14914, 2023.arxiv.org/abs/2306.14914
  4. [9]R. Schneider, Convex Bodies: The Brunn-Minkowski Theory, 2nd expanded ed., Cambridge University Press, 2014.

Paper details

vlnc:W_nsgemrcu
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents