Alec Kriebel

A complete negatively curved observation Gramian with two-state ambiguity

Can a system respond to every small change of state and still hide which state it occupies? This note gives a geometric example. Its observation metric is positive, complete and negatively curved, yet each possible output history comes from two distinct states. The ambiguity survives every observation time. The construction combines classical geometry with a deliberately simple dynamical system.

Original abstract

We give a smooth autonomous observed system on a connected closed genus-three surface whose exact finite-time differential observation Gramian is complete and has constant negative sectional curvature, although every attained output history corresponds to exactly two initial states. The system has zero dynamics and a Euclidean-valued output obtained by composing a connected double cover of a closed hyperbolic surface with a smooth Nash isometric embedding into ℝ¹⁷. For each fixed T > 0 its Gramian is P_T = Tπ*g and its curvature is −1/T. Uniform positive upper and lower bounds hold in a fixed finite normalized atlas and relative to any fixed smooth background metric. This refutes the manifold implication printed in Krener's 2009 open problem under those coherent meanings of its coordinate bounds. The construction is an application of classical constructions. A bounded primary-source audit found no earlier explicit complete construction in the inspected corpus; no absolute historical firstness is asserted.

The result

There exist a connected closed oriented genus-three surface M and a smooth output h:M → ℝ¹⁷, with zero dynamics f=0, having the following properties for every fixed T > 0. The exact differential observation Gramian is P_T = Tπ*g, a smooth positive definite, geodesically complete metric of constant sectional curvature −1/T. The finite-history map is locally injective, while every attained history has exactly two initial states. The same paired states remain indistinguishable for all real times.

The bounds are explicit. For any fixed smooth background metric q, positive finite constants c,C give Tcq ≤ P_T ≤ TCq. One fixed finite normalized atlas also satisfies 4T I₂ ≤ [P_T] ≤ (64T/9) I₂. These statements are uniform over states. They concern the actual differential observation metric, rather than a separately selected auxiliary metric.

PT = Tπ*g,   K(PT) = −1/T.

Two sheets, one output

Start with a closed hyperbolic surface S of genus two and curvature −1. A connected unramified double cover π:M → S has genus three: every point of S has two distinct lifts. Classical covering-space theory supplies this connected cover; it is more than two disconnected copies of the same surface. Pulling the hyperbolic metric back to M preserves its local curvature.

Nash’s smooth isometric embedding theorem places S in ℝ¹⁷ by a globally injective map e that preserves the metric. Define the output by h=e∘π. Because the state does not move, its output history is constant, and the Gramian integral becomes T times the pullback metric. Nearby states on a single sheet have distinct outputs. The two lifts of any base point have the same output because π identifies them, and injectivity of e rules out additional collisions.

This separates local sensitivity from global identification. A lower bound on infinitesimal changes does not become a lower bound on output differences between all distant states. Negative curvature and completeness do not eliminate the two-sheeted ambiguity. The surface is genuinely two-dimensional, so its sectional curvature condition has real content.

What the counterexample addresses

The target is Krener’s manifold implication printed in Oberwolfach Report 11/2009, pages 674–675. The note refutes that implication under its stated coherent interpretations of unspecified coordinate bounds: a fixed finite normalized atlas, or comparison with a fixed background metric. It does not settle a strengthened problem requiring one global Euclidean state domain. Bounds under arbitrary rescalings of every possible chart are not asserted.

Observation time is also fixed when uniform bounds are stated. At T=0 the Gramian vanishes, and as T grows its curvature approaches zero. The theorem gives neither one positive lower bound nor one negative curvature constant valid over all observation times. Increasing T still does not distinguish the paired states.

Classical foundations and context

This is a classical synthesis of hyperbolic surface geometry, Hatcher’s covering classification and Nash’s embedding theorem. Older local/global observability ambiguity appears in Hermann–Krener, and Gauthier treats covering realizations under different controlled analytic assumptions. Other global-observer results use additional separation, injective-coordinate or system-structure hypotheses. These distinctions matter when comparing the exact Gramian here with empirical or auxiliary metrics. Historical firstness and worldwide present-day open status are not established.

Proof and supplementary checks

The supplied exact verifier supports elementary polynomial, rational, monodromy and matrix identities, with a separate finite-precision diagnostic. It does not construct a Nash embedding or machine-certify global topology, smoothness or completeness. Those parts rest on the written proof and its classical theorem inputs.

Read the paper and materials

Open the original PDF · Download PDF · Zenodo record · BibTeX

The unchanged manuscript prints Version 1.0, 5 October 2026. All original files follow in the native record’s order:

  1. negative_gramian_cover.tex (15,912 bytes) · Zenodo original
  2. verification_results.json (2,832 bytes) · Zenodo original
  3. SOURCE_AUDIT.md (21,292 bytes) · Zenodo original
  4. SHA256SUMS.txt (697 bytes) · Zenodo original
  5. LICENSE.txt (682 bytes) · Zenodo original
  6. negative_gramian_cover_sources.zip (25,192 bytes) · Zenodo original
  7. verify_cover.py (10,231 bytes) · Zenodo original
  8. negative_gramian_cover.pdf (79,846 bytes) · Zenodo original
  9. README.md (5,018 bytes) · Zenodo original

Original manuscript, verification code, results and authored documentation are CC BY 4.0. Cited third-party works retain their own rights.

Citation

Alec Kriebel, A complete negatively curved observation Gramian with two-state ambiguity, version 1.0, 5 October 2026. doi:10.5281/zenodo.23153134. Download BibTeX.

Status and collaboration

AI tools were used extensively in solving, drafting, reproduction and adversarial review. This is an unrefereed preprint without conventional human peer review. Internal AI-assisted checking does not represent journal acceptance or formal proof-assistant certification.

Independent verification, substantive mathematical extensions and journal preparation would be useful contributions. Collaboration and authorship would depend on the work actually contributed.

References

  1. A. J. Krener, Open Problem: Global Observability of Spaces of Negative Curvature, Oberwolfach Report 11/2009, 674–675. DOI.
  2. J. Nash, The Imbedding Problem for Riemannian Manifolds, Annals of Mathematics 63 (1956), 20–63, Theorem 2, p. 59. Original paper.
  3. A. Hatcher, Algebraic Topology (2002), 51, 61, 66–67. Author’s online version.
  4. R. Hermann and A. J. Krener, Nonlinear Controllability and Observability, IEEE TAC 22 (1977), 728–740. DOI.
  5. J.-P. Gauthier, Observability (deterministic systems) and realization theory, author manuscript dated 29 June 2009, §7.1; original public posting date unverified. Author manuscript.
  6. G. Dirr, U. Helmke and J. Jordan, Global observability of real analytic systems, ECC (2009), 4582–4586. DOI.
  7. W. Kang, A. J. Krener, M. Xiao and L. Xu, A Survey of Observers for Nonlinear Dynamical Systems (2013), 1–25. DOI.
  8. H. J. Sussmann, On quotients of manifolds: a generalization of the closed subgroup theorem, Bull. AMS 80 (1974), 573–575. DOI.
  9. P. S. Krishnaprasad, Geometry of Parametric Models: Some Probabilistic Questions, Allerton (1977), 661–670. Author-hosted original.