Complete Riemannian Metrics

• Complete Riemannian metrics are defined by the property that every maximal geodesic extends infinitely, ensuring metric and geodesic completeness.
• Metric completion techniques, such as the Gordon construction using barrier functions, are used to augment incomplete metrics in various geometric contexts.
• These metrics are crucial in both finite and infinite-dimensional settings, supporting advances in shape analysis, control theory, and numerical modeling.

A complete Riemannian metric is one for which the associated geodesic flow is defined for all time, equivalently, every maximal geodesic can be extended to the entire real line. Completeness plays a central role in Riemannian geometry, topology, variational calculus, and the geometric analysis of both finite- and infinite-dimensional manifolds. The following article synthesizes key definitions, construction principles, analytic criteria, and contemporary advances with emphasis on new results for submanifolds, shape spaces, and structure-preserving metrics.

1. Definition and Fundamental Criteria

Let $(M,g)$ be a connected Riemannian manifold, finite- or infinite-dimensional. The metric $g$ is complete if every maximal geodesic $\gamma:[a,b)\to M$ (solution to the geodesic equation $\nabla_{\dot\gamma}\dot\gamma=0$) with parameter domain $[a,b)$ can be extended to $\mathbb R$. In metric terms, $(M,g)$ is complete as a length space: every Cauchy sequence with respect to the Riemannian (geodesic) distance $d_g$ converges in $M$.

In finite dimensions, the Hopf–Rinow theorem equates geodesic completeness, metric completeness, and the property that all closed, $d_g$-bounded sets are compact. In infinite dimensions, these equivalences break down; completeness, existence of minimal geodesics, and the topological compactness of metric balls require separate verification.

A conformal deformation, $g \mapsto \tilde g=f^2g$, preserves completeness if and only if for every $g$-geodesic ray $\gamma:[0,\infty)\to M$ parametrized by arclength, $\int_0^{\infty} f(\gamma(s))\,ds=+\infty$; at most linear growth $f$ ensures $\tilde g$ is complete, whereas superlinear growth can destroy completeness (see (Dirmeier, 2012)).

2. Metric Completion Techniques and the Gordon Construction

A general procedure for guaranteeing completeness, central in both geometry and applications to control theory, is metric augmentation by a proper barrier function (Gordon's construction). Given an open manifold or region $M\subset Q$ equipped with a Riemannian metric $g$ (possibly incomplete due to boundary effects or degeneration), and a smooth proper function $f:M\to\mathbb R$, the augmented metric

$\tilde{g} = g + df\otimes df$

is complete, provided $f$ is proper ($f\to\infty$ along any sequence escaping every compact subset of $M$) (Acosta et al., 2023). The classical theorem of Gordon (1973) establishes that $\tilde{g}$-balls are compact and geodesics never reach the boundary in finite time.

This "barrier augmentation" is systematically used to patch incomplete ends, enforce state-space constraints in controlled mechanical systems, and convert "hard" forbidden regions into points at infinite distance. The approach is robust with respect to metric equivalence away from the boundary and extends to time-dependent and certain nonholonomic systems.

3. Completeness Criteria in Shape Spaces and Function Spaces

Infinite-dimensional shape spaces, such as Hilbert or Sobolev manifolds of curves and surfaces, present nontrivial challenges. In "On a Complete Riemannian Metric on the Space of Embedded Curves" (Döhrer et al., 28 Jan 2025), a strong Sobolev-type metric $G$ on the manifold of $H^s$-regular embeddings, $s\in(3/2,2)$, is proven to be complete in four distinct senses:

1. Sequential weak compactness: $G$-bounded nets are relatively compact in the weak $H^s$ topology.
2. Metric completeness: Every $d_G$-Cauchy sequence converges.
3. Geodesic completeness: Geodesics exist for all time; ODE techniques and "escape lemmas" are used to exclude finite-time singularity formation.
4. Existence of minimizing geodesics: Any two embeddings in the same path component (knot class) can be joined by a length-minimizing geodesic.

For reparametrization-invariant Sobolev metrics of order $k\geq3$ on the space of immersed surfaces $\Imm^k(M,\mathbb R^3)$, similar completeness results are established (Bauer et al., 1 Dec 2025). The core analytic tool is the Michael–Simon–Sobolev inequality, furnishing $L^\infty$ control over geometric quantities and enabling the extension of the Hopf–Rinow paradigm to the infinite-dimensional setting.

4. Structure-Preserving and Quotient Metrics

On finite-dimensional submanifolds defined by geometric or algebraic constraints (e.g., Stiefel manifolds, positive semidefinite matrices of fixed rank), the construction of complete Riemannian metrics is essential for optimization, manifold learning, and numerical shape analysis. In (Nguyen, 2020), a new family of complete quotient metrics is constructed on $S_{+}(n,p)$, the manifold of positive semidefinite $n\times n$ matrices of fixed rank $p$. The core construction leverages a Riemannian submersion model:

$$S_{+}(n, p) \cong (St(p, n) \times S_{++}(p)) / O(p),$$

with parameters $(\alpha_0,\alpha_1,\beta)$ controlling the Stiefel and positive-definite parts to guarantee geodesic completeness and allow efficient gradient/Hessian computation. The induced metric admits closed-form geodesic expressions and is strictly stronger than traditional non-complete choices (ambient or canonical). By varying these parameters, curvature can be tuned across horizontal and vertical directions.

5. Completeness in Discrete and Geometric Modeling

In discrete geometric modeling, it is critical to endow spaces of planar meshes or shapes with Riemannian metrics that guarantee qualitative preservation of structure under deformation. The manifold of planar triangular meshes with fixed connectivity is an example (Herzog et al., 2020). Here, the metric is constructed by augmenting the Euclidean metric on configuration space $\mathbb R^{2\times N_V}$ by the gradient of a proper "barrier" function penalizing small triangle heights and vertex-to-nonincident-edge distances:

$$g_{ab}(Q) = \delta_{ab} + \left(\frac{\partial f}{\partial q^a}(Q)\frac{\partial f}{\partial q^b}(Q)\right),$$

where $$f(Q)=\sum_{T\in\Delta_2}\sum_{\ell=0}^2 \beta_1/h_\ell(T;Q) + (\beta_3/2)\|Q-Q_{\text{ref}}\|_F^2$$, and $h_\ell$ are triangle heights. Completeness is established by showing $f$ blows up at the boundary of the manifold of admissible meshes, preventing degeneration in finite time.

Numerical schemes, e.g. symplectic Störmer–Verlet integrators for the geodesic Hamiltonian, preserve energy and structure, and empirical evidence confirms that the mesh aspect ratio remains bounded away from zero along all geodesics, with no triangle flipping or collapsing permitted for arbitrarily long geodesic flows.

6. Analytical and Geometric Characterizations

Classical questions of completeness under conformal deformations or submanifold metrics depend on integrability criteria along escaping geodesics. For a conformal change $g \mapsto \tilde{g} = f^2 g$, the transformed metric is complete iff $\int_0^\infty f(\gamma(s))ds = \infty$ for every $g$-arclength ray $\gamma$ (Dirmeier, 2012). If $f$ exhibits at most linear growth in $g$-distance, completeness is preserved; faster growth may render the metric incomplete.

On infinite-dimensional mapping spaces, completeness criteria are typically reduced to conditions on the control of analytic norms (e.g., Michael–Simon–Sobolev bounds), closedness under extensions in the background topology, and the behavior of geometric invariants under deformation. This framework supports robust existence and regularity results for variational problems in shape analysis, surface matching, and geometric evolution.

7. Implications, Applications, and Limitations

Complete Riemannian metrics underpin the well-posedness of geodesic boundary and initial value problems, the existence of minimizing paths in variational calculus of submanifolds, state constraint enforcement in geometric control, and the development of stable numerical schemes for shape deformation. The completeness-enforcing constructions—barrier augmentation, Sobolev-type metrics, and parameterized quotient models—provide tools to extend results from finite to infinite dimensions, across discrete and continuous spaces, and for both unconstrained and constrained scenarios.

Nonetheless, technical limitations arise at low regularity, under inappropriate weighting of geometric invariants, or for certain non-invariant or low-order metrics where key analytic control is lost. Each construction's scope is therefore determined by the interplay of topological, analytic, and geometric constraints intrinsic to the manifold and metric under consideration.