Weak Frenet Frame: Analysis & Applications

• Weak Frenet frames generalize the classical Frenet–Serret apparatus to non-smooth curves by leveraging weak regularity frameworks such as W²,² and rectifiable approximations.
• They maintain key geometric invariants like curvature and torsion through constructions including Bishop frames and projective, piecewise-linear methods.
• Applications include variational analysis, computational geometry, and physical modeling, enabling derivative‐free reconstruction from sampled data and robust elastic rod modeling.

A weak Frenet frame generalizes the classical Frenet–Serret apparatus to the analysis of non-smooth curves—curves with lower regularity or sampled curves where classical derivatives and moving frames are ill-defined. Several rigorous constructions exist, relating weak Frenet frames to the Bishop (RPAF) frame for $W^{2,2}$-regular curves, to piecewise-linear and projective-geometric approaches for rectifiable curves of finite total curvature and torsion, and to derivative-free algorithms based on point samples or tangent approximations. These frameworks maintain the essential geometric structure of a moving orthonormal frame and associated invariants (curvature, torsion) without requiring classical differentiability, enabling extensions to variational analysis, discrete geometry, and geometry processing.

1. Classical and Weak Frenet Frames: Smooth vs. Non-smooth Settings

The classical Serret–Frenet frame for a spatial curve $x:[0, L] \to \mathbb{R}^3$ requires $C^3$-regularity and everywhere non-vanishing curvature ($|x''|\ne 0$). Given arclength parametrization ($|x'| \equiv 1$), the frame is constructed as

• $t = x'$, the unit tangent;
• $n = t'/|t'|$, the unit normal;
• $b = t \times n$, the binormal.

The curvature and torsion are given by

• $\kappa = |t'|$,
• $\tau = b' \cdot n$.

These satisfy the Serret–Frenet ODE system: $$\begin{aligned} t' &= \kappa n,\ n' &= -\kappa t - \tau b,\ b' &= \tau n. \end{aligned}$$ This framework fails for non-smooth curves due to the lack of well-defined higher derivatives and possible points with vanishing curvature.

The weak Frenet frame addresses these limitations by:

• Replacing classical derivatives with frameworks compatible with weak regularity (e.g., $W^{2,2}$ Sobolev spaces or rectifiable curves),
• Preserving geometric invariance,
• Ensuring consistency with the classical case when smoothness is present (Bevilacqua et al., 2023).

2. Relatively Parallel Adapted Frames (Bishop Frames) and Weak Geometric Invariants

For $W^{2,2}$-regular curves, the Bishop or Relatively Parallel Adapted Frame (RPAF) replaces the classical frame. Given $x \in W^{2,2}([0,L];\mathbb{R}^3)$ and $t=x' \in W^{1,2}$ with $|t|=1$, an RPAF consists of an orthonormal triple $\{t(s), d_1(s), d_2(s)\}$ where $d_1, d_2 \in W^{1,2}$ and $d_i'(s)$ is parallel to $t(s)$. Equivalently, there exist $u_1,u_2 \in L^2(0, L)$, called flexural densities, such that

$$\begin{aligned} t' &= u_2 d_1 - u_1 d_2,\ d_1' &= -u_2 t,\ d_2' &= u_1 t. \end{aligned}$$

The complex-valued normal development $u(s) = u_2(s) + i u_1(s)$ allows the curvature and torsion to be defined by

$$\kappa(s) := |u(s)| \in L^2, \qquad \tau(s) := \theta'(s)$$

where $\theta(s) = \arg u(s)$, with the derivative taken in the sense of distributions. This construction is independent of the initial choice of normal directions up to constant rotation, and recovers the classical invariants when additional smoothness is present (Bevilacqua et al., 2023).

3. Projective and Piecewise-linear Weak Frenet Frame for Rectifiable Curves

For rectifiable curves $c:[0,L] \to \mathbb{R}^3$ of finite total curvature $\mathrm{TC}(c)$ and total absolute torsion $\mathrm{TAT}(c)$, a geometric construction based on polygonal approximations and projective geometry yields weak analogues of the tangent, normal, and binormal:

• The weak tangent indicatrix $t_c: [0, \mathrm{TC}(c)] \to S^2$ is the pointwise limit of reparametrized tangent indicatrices of finer and finer inscribed polylines.
• The weak binormal $b_c: [0, \mathrm{TAT}(c)] \to \mathbb{R}P^2$ arises as the uniform limit of the polar of the tangent indicatrix (in the projective plane), with length equal to total absolute torsion.
• The weak principal normal $n_c(s) := b_c(s) \times t_c(s) \in \mathbb{R}P^2$ is defined via the projective cross product and is rectifiable of length $\mathrm{TC}(c) + \mathrm{TAT}(c)$.

For $C^3$-regular curves with positive curvature, these weak objects lift to the classical Frenet frame under suitable reparameterizations (Mucci et al., 2019).

4. Discrete and Derivative-free Algorithms: Severi–Bouligand Tangents and Sample-based Frenet k-frames

For point clouds or sets $X \subset \mathbb{R}^n$, or sampled curves, the weak Frenet frame can be defined via discrete sequences:

• A Severi–Bouligand tangent at $x \in X$ is a nonzero vector $u$ approached by normalized difference vectors of a convergent sequence.
• The discrete Frenet $k$-frame is inductively constructed by orthonormalizing the directions of displacements and their residuals up to $k$-th order, purely from the data.

If $X$ is a smooth curve, the discrete process converges to the classical Frenet frame as the sampling density increases. In general, existence of an outgoing SB-tangent frame at a point is equivalent to certain algebraic properties of the function space $R(X)$, specifically the failure of strong semisimplicity (Cabrer et al., 2013).

This computation is practical for applications in mesh processing and geometric modeling, requiring no derivatives and converging at linear rate for $C^{k+1}$ underlying manifolds.

5. Existence, Uniqueness, and Invariance Results for Weak Frames

In the RPAF formalism, given $t \in W^{1,2}$ with $|t| \equiv 1$ and any initial orthonormal triple, the associated Volterra integral equations for $u_1, u_2$ have unique $L^2$ solutions, yielding a unique adapted frame modulo rotations in the normal plane (Bevilacqua et al., 2023). For polygonal/projective constructions, the weak binormal and normal admit unique limits as sequences of inscribed polygonals refine, with Lipschitz and bounded variation arguments ensuring compactness and convergence (Mucci et al., 2019).

All these constructions are invariant under reparametrization and congruence, and different choices of initial frame or coordinates affect only a constant rotation in the normal plane, hence the invariants $\kappa$ and the distributional derivative for $\tau$ are truly geometric.

6. Applications in Variational Problems and Computational Geometry

The weak Frenet frame underpins the analysis of variational problems on non-smooth curves:

• Elastic rods and ribbons can be treated by prescribing $L^2$ flexural densities and reconstructing frames/curves in $W^{2,2}$; when the twist density is zero, the RPAF provides an appropriate frame for defining bending and torsional energies.
• The Kirchhoff–Plateau analytical model utilizes $W^{2,2}$ centerlines and Bishop frames to define well-posed energies for soap films spanned by elastic links, even when only distributional torsion is available.
• In the analysis of elastic ribbons (Sadowsky functional), energies depending on curvature and torsion can be rigorously defined and relaxed using the RPAF and weak invariants, facilitating mathematical analysis of physical ribbons with low regularity centerlines (Bevilacqua et al., 2023).

In computational geometry and vision, derivative-free estimation of tangent and normal directions at sample points enables robust detection of features, ridge lines, and curvature-driven geometric processing (Cabrer et al., 2013).

7. Summary Table: Weak Frenet Frame Constructions

Approach	Applicable Regularity	Key Output
Bishop/RPAF	$W^{2,2}$	$\{t(s), d_1(s), d_2(s)\}$, $L^2$-curvature and distributional torsion
Polygonal/Projective	Rectifiable, finite total curvature	Rectifiable indicatrices $t_c$, $b_c$, $n_c$ in $S^2$, $\mathbb{R}P^2$
Discrete/SB-tangent	Compact subsets, point clouds	Orthonormal $k$-frame from sample sequences

The development of weak Frenet frames enables a unified geometric theory bridging smooth and non-smooth scenarios, provides geometric invariants consistent across regularity classes, and underlies modern approaches in analysis, discrete differential geometry, and variational modeling [(Bevilacqua et al., 2023); (Mucci et al., 2019); (Cabrer et al., 2013)].