Generic Cuspidal Points
- Generic cuspidal points are isolated cusp-type singularities identified in smooth surfaces, singular interfaces, algebraic curves, and matrix families via specific nondegeneracy conditions.
- They are characterized by precise local invariants such as tangency of asymptotic lines, curvature constraints, and square-root branches in eigenvalue structures.
- Their analysis reveals deep links between local monodromy, topological restrictions, and enumerative geometry, enhancing classification and deformation theories.
The literature suggests that “generic cuspidal point” does not have a single universal meaning. In different research programs it denotes several nondegenerate cusp-type phenomena: a cusp of Gauss or godron on a generic smooth surface in , a generic cuspidal edge or generalized cuspidal edge on a singular surface in , an ordinary cusp occurring as the unique non-nodal singularity in an equiclassical family of algebraic curves, and a transversal eigenvalue coalescence in a two-parameter complex matrix family (Kazarian et al., 2020, Naokawa et al., 2014, Ganor et al., 2018, Dieci et al., 21 Jul 2025).
1. Terminological scope and the role of genericity
Across these settings, “generic” refers to a nondegeneracy condition that stabilizes a cusp-type configuration under perturbation. The specific condition depends on the ambient category: projective surface theory uses tangency of asymptotic and parabolic data; singular surface theory uses nonvanishing limiting normal curvature; algebraic-curve theory uses a codimension-one cusp condition inside a nodal family; matrix theory uses a transversal zero of the discriminant map.
| Setting | Cuspidal object | Genericity condition |
|---|---|---|
| Smooth surfaces in | Godron / cusp of Gauss | Parabolic point where the unique double asymptotic line is tangent to the parabolic curve |
| Singular surfaces in | Generic cuspidal edge | Osculating plane of the singular image is not orthogonal to the unit normal; equivalently |
| Generalized cuspidal edges | Generic point | , equivalently |
| Toric or plane algebraic curves | One ordinary cusp in an equiclassical family | The cusp condition cuts the nodal Severi family by one additional condition |
| complex matrix families | Generic cuspidal point | and invertible |
This suggests that the term functions less as a single definition than as a family of parallel notions: an isolated cusp-type singularity together with the transversality needed for local normal forms, index theory, or enumerative stability (Honda et al., 2019, Ganor et al., 2018, Dieci et al., 21 Jul 2025).
2. Cusps of Gauss on generic smooth projective surfaces
In projective differential geometry, the relevant generic cuspidal points are the godrons, explicitly identified with cusps of Gauss on generic smooth surfaces in 0. A godron is a parabolic point at which the unique double asymptotic line is tangent to the parabolic curve; equivalently, it is a point of tangency of the flecnodal and parabolic curves, and it maps to a swallowtail on the dual surface (Kazarian et al., 2020).
The local geometry is organized by the quadratic form 1, the Hessian/discriminant
2
and the fundamental cubic form
3
At a parabolic point different from a godron, 4 has a double zero line equal to the asymptotic line and a simple zero line tangent to the parabolic curve; at a godron, those directions merge, and 5 has a triple zero line tangent to the parabolic curve. This is the paper’s sharpest local signature of the cuspidal degeneration (Kazarian et al., 2020).
For godrons the Landis–Platonova normal form is
6
The sign convention is 7 for a positive godron and 8 for a negative godron. The same paper assigns local indices to godrons and derives global Euler-characteristic relations. For a hyperbolic component 9 and an elliptic component 0 of a generic smooth compact surface,
1
These formulas constrain the coexistence of projective umbilics and cusps of Gauss, so the local cusp-of-Gauss geometry is tied directly to the topology of the elliptic/hyperbolic decomposition (Kazarian et al., 2020).
3. Generic cuspidal edges and generalized cuspidal edges in 2
For singular surfaces in Euclidean 3-space, a cuspidal edge is a map-germ right-left equivalent to
4
A cuspidal edge singular point is called generic when the osculating plane of the singular image curve is not orthogonal to the unit normal vector field 5. Equivalently, the limiting normal curvature satisfies
6
The same work shows that a generic real analytic cuspidal edge can be isometrically deformed, preserving 7, into a cuspidal edge whose singular set lies in a plane, and that the resulting planar model is uniquely determined up to congruence (Naokawa et al., 2014).
The later generalized theory enlarges the class from ordinary cuspidal edges to generalized cuspidal edges, including cuspidal cross caps and 8-cuspidal edges. In this setting the genericity condition is
9
equivalently 0. The paper constructs an involution 1 on generic real analytic generalized cuspidal edges along a fixed edge image 2, sending 3 to an isometric dual 4 with the same first fundamental form and the same singular set image, but with opposite cuspidal angle. A second involution 5 produces orientation-reversing isomers, so that in the admissible, no-symmetry case the four realizations
6
represent four distinct congruence classes (Honda et al., 2019).
The flat geometry of cuspidal edges is described through height functions and orthogonal projections. The contact with planes is measured by singularities of height functions, and the contact with lines by singularities of orthogonal projections. For a generic cuspidal edge, only low-codimension 7- and 8-types occur, yielding a controlled list of plane-contact and profile bifurcation phenomena (Sinha et al., 2016). A related construction folds a cuspidal edge by
9
producing cuspidal 0-singularities. In particular, the cuspidal cross-cap is the folded 1 case, and a cuspidal 2-singularity appears when the original cuspidal edge curve has 3-point contact with the folding plane (Sinha et al., 2017).
4. Ordinary cusps on algebraic curves
In algebraic-curve theory, the local cusp is usually the ordinary plane cusp 4 or, in the global rational setting, a locally unibranch singular point encoded by Newton or Puiseux data. For rational unicuspidal plane curves, the object is an irreducible curve 5 whose normalization is 6 and which has exactly one cuspidal singular point. If 7 has degree 8 and cusp 9, then
0
In the one-Newton-pair case, the local model is
1
with local equation
2
The degree-3 classification is complete in terms of Newton pairs, with separate behavior in prime and composite degree (DeVleming et al., 2023).
On toric surfaces, the equiclassical viewpoint is more explicit. For a nondegenerate lattice polygon 4, the family 5 of irreducible genus-6 curves with one cusp has pure dimension
7
and a generic member has one ordinary cusp and
8
nodes. In this sense, the cusp condition is codimension one relative to the nodal Severi family. Tropicalization shows that the corresponding generic tropical curves carry exactly one localized cuspidal fragment and are otherwise trivalent; in the 9-transverse case, the generic cuspidal fragments reduce to three types: a rational four-valent vertex, a rational flat trivalent vertex, or an elliptic edge (Ganor et al., 2018).
Taken together, these results indicate that algebraic “generic cuspidal points” are not generic points of the full linear system. They occur as the distinguished singularities of a codimension-one equiclassical stratum, and their local analytic type is strongly constrained by degree, genus, semigroup data, and global topology (DeVleming et al., 2023, Ganor et al., 2018).
5. Enumerative counts and topological obstructions
In enumerative geometry on del-Pezzo surfaces, a cusp is imposed as a first-order vanishing condition on the derivative of a stable map. For a class 0, the paper counts rational cuspidal curves in class 1 through
2
generic points. The cusp is encoded by a section
3
of
4
and the count is obtained as an Euler class computation plus an explicit boundary contribution from two-component splittings with a ghost bubble (Biswas et al., 2015).
A different line of work studies which cusp configurations can occur at all. For odd-degree rational cuspidal plane curves in 5, the connected sum of the cusp links must be involutively simple, yielding strong semigroup restrictions. In the two-cusp case, the paper proves
6
and in particular at least one of the two cusp links must be an even L-space knot (Borodzik et al., 2016). On Hirzebruch surfaces 7, rational cuspidal curves of type 8 satisfy a semigroup-counting inequality
9
whenever
0
and they are also subject to a spectrum-at-infinity inequality comparing the spectra of local cusps with the spectrum of the link at infinity (Borodzik et al., 2014).
These results do not classify generic cusps in a moduli-theoretic sense. Rather, they show that global geometry sharply limits which local cusp types are realizable: enumeratively, the cusp is a codimension-one condition; topologically, many candidate cusp configurations are excluded before any genericity question inside a family arises (Biswas et al., 2015, Borodzik et al., 2016, Borodzik et al., 2014).
6. Generic cuspidal points in matrix families
A recent spectral reinterpretation uses the term for smooth complex matrix families depending on two real parameters. For
1
define
2
A parameter value 3 is a generic cuspidal point if
4
This is a transversal coalescence of two eigenvalues. The local eigenvalue structure is
5
so the spectrum has a square-root branch singularity. The paper treats these points as closely related to exceptional points, but reserves “generic cuspidal point” for the transversal two-parameter case (Dieci et al., 21 Jul 2025).
Loops in parameter space reveal the local monodromy. If a 6-periodic loop encloses a single generic cuspidal point, then any continuous labeling of the two active eigenvalues satisfies
7
Under the paper’s parallel-transport-type gauge, the associated eigenvectors also accumulate phase: after one loop the phase constants satisfy
8
and for shrinking loops one obtains
9
If no generic cuspidal point lies inside the loop, the phase is instead
0
This gives a localization principle: periodicity defects of eigenvalues and nontrivial phase accumulation of eigenvectors detect enclosed generic cuspidal points. The same paper also shows the limitation of this method: two enclosed generic cuspidal points may cancel in the eigenvalue permutation, so localization may require both monodromy and phase data (Dieci et al., 21 Jul 2025).
This suggests a broad structural analogy with the geometric cases above. The cusp is again an isolated singular event selected by a nondegenerate Jacobian condition, and its principal signature is monodromy—of asymptotic directions, singular set geometry, tropical branches, or eigenvalue sheets—around the singular locus (Dieci et al., 21 Jul 2025).