On the geometry of holomorphic curves and complex

surface
arXiv:2505.24719v1 [math.DG] 30 May 2025

Amanda Dias Falqueto and Farid Tari

June 13, 2025

Abstract
We investigate the geometry of holomorphic curves and complex surfaces
from the perspective of singularity theory. We show that, with a suitable choice
of a complex bilinear symmetric form, the families of functions and mappings
that measure the contact between curves or surfaces and model objects be-
come holomorphic. This allows the application of singularity theory, yielding
analogues of classical results from the real case. Our approach enables the def-
inition of geometric invariants of curves, which we call the C-curvature and
C-torsion, as well as surface invariants such as the C-principal curvature and
C-Gaussian curvature. It also gives geometric meaning to the complexification
of of the families measuring contact of analytic surfaces in R3 with lines, planes
and spheres.

Key–words: holomorphic curves, complex surfaces, contact, curvature, singularities.

2020 Mathematics Subject Classification: 57R45, 53B99.

1 Introduction
We study in this paper the local differential geometry of holomorphic curves in C2 and
C3 , and of complex surfaces in C3 from a singularity theory viewpoint.
The geometry of submanifolds of Cn relies on the Hermitian inner product (see for
example [11]). However, with this inner product, the families of functions and map-
pings of interest are not holomorphic. For instance, the Gauss map of a hypersurface
is not holomorphic.
We remedy this by choosing a complex bilinear symmetric form in Cn (n = 2, 3
here). With this setting, the Gauss map of a hypersurface is the complex conjugate
of a normal vector (with respect to the Hermitian inner product), resulting in a holo-
morphic map. In all the paper, the analogues of real geometric concepts are denoted

1
by the same names, with the prefix C attached (where C stands for conjugate and
reflects the way the Hermitian inner product is used).
Extensive studies of the geometry of submanifolds of the Euclidean space Rn and
the Minkowski space Rn1 from a singularity theory viewpoint resulted in furthering our
understanding of the geometry of these objects and lead to valuable and numerous
applications in areas such as robotics, computer vision and shape recognition (see, for
example, [3, 6, 7, 9, 10, 12])
In the case of curves and surfaces in R3 , the study is carried out by considering their
contact with lines, circles, planes, spheres and cylinders. For example, the contact of
a surface with spheres is measured by the R-singularity types of the distance squared
function. Each singularity type at a given point on the surface reveals some aspect
of the geometry of the surface at that point. For generic surfaces, the family of
distance squared functions, obtained by varying the centres of the contact spheres,
is an R+ -versal family of the singularities of its members. This allows one to obtain
the bifurcations of non-stable singularities of a given member of the family. Also, the
bifurcation set of the family is the focal set (caustic) of the surface, and singularity
theory tools of versal deformations are used to deduce the generic local models of the
focal set.
We show in this paper that, with the right approach, one can have analogous results
to the real case for holomorphic curves in C2 and in C3 , and complex surfaces in C3 .
We deal in §3 with plane curves where we define the concept of curvature and obtain
the usual results about vertices and the evolute. In §4 we consider space curves, define
the C-curvature and C-torsion, the C-Frenet-Serret frame and its moving formula, all
analogous to those in the real case. We deal in §5 with surfaces in C3 . We define the
C-Gauss map and show that its derivative is a symmetric operator. We define the C-
first and C-second fundamental forms, and with that, all the concepts on the geometry
of surfaces in the Euclidean 3-space can be carried over to the complex case. For both
the curve and surface cases, we define the family of C-height functions, C-distance
squared functions and C-orthogonal projections. All these families are holomorphic
and the results in the real case are carried over to the complex case. These families
can be considered as complexifications of their real counterpart.
In §3, we show that the envelope of the normal lines to a plane curve, with respect
to the Hermitian inner product, is empty, whereas the envelope of the C-normal lines
(with respect to the bilinear forms chosen here) is not. This indicates that the approach
taken in this paper is a natural one.
Usually, when calculating invariants of analytic curves and surfaces in the Euclidean
space, such as the number of inflections and vertices stored at a singularity of a curve,
or the number of umbilics stored at a degenerate umbilic point, one complexifies and
obtains an upper bound for their real counterpart. The approach in this paper gives
a geometric meaning to the complexified objects.

2
2 Preliminaries
The standard Hermitian inner product in Cn is defined as
n
X
⟨v, w⟩ = vi w i ,
i=1

for v = (v1 , . . . , vn ), w = (w1 , . . . , wn ).

The modulus squared function v 7→ ⟨v, v⟩ is not holomorphic, the orthogonal pro-
jection along a given vector is not holomorphic, and the Gauss map of a complex
hypersurface which associate at each point a unit orthogonal vector is also not holo-
morphic. Therefore, the singularity theory techniques applied to differential geometry
(see for example [3, 9]) do not apply directly if the hypersurface is endowed with the
Hermitian inner product.
We consider instead the following complex bilinear, symmetric and non-degenerate
form n
X
g(v, w) = ⟨v, w⟩ = vi wi . (1)
i=1

The bilinear form g is a holomorphic function, and its restriction to Rn is the

standard inner product in Rn .
Two vectors v and w are said to be C-orthogonal if g(v, w) = 0, that is v is
orthogonal to w with respect to the Hermitian inner product. A vector v is a C-unit
vector if g(v, v) = 1.
The bilinear form has a property similar to Lorentzian inner product: there are
non-zero vectors that satisfy g(v, v) = 0. We call such vectors C-lightlike vectors.
A complex sphere of centre p and radius r ∈ C is defined as the set of points z ∈ Cn
that satisfy
⟨z − p, z − p⟩ = r2 .
The complex unit sphere CS n−1 is the set

CS n−1 = {z ∈ Cn : ⟨z, z⟩ = 1},

so every point in CS n−1 represents a C-unit vector.

When n = 2, CS 1 is the complex unit circle and has equation z12 + z22 = 1, where
z = (z1 , z2 ) ∈ C2 . Complex circles are conics that passes through the circular points
(1 : i : 0) and (1 : −i : 0) at infinity in CP 2 . These represent the two lines z2 = ±iz1 .
These lines, with the origin removed, represent the C-lightlike vectors in C2 .
For v = (v1 , v2 ) ∈ C2 , we denote by v ⊥ = (−v2 , v1 ) and orthogonal vector with
respect to the Hermitian inner product.
We use the usual singularity theory concepts and notation about the actions of the
Mather groups on the set of germs of holomorphic mappings, see for example [15].

3
 Two germs of families of holomorphic functions F, G : (Cn × Cm , (0, 0)) → (C, 0)
are said to be R+ -equivalent if

G(z, v) = F (ϕ(z, v), ψ(v)) + c(v),

where (ϕ, ψ) : (Cn × Cm , (0, 0)) → (C × Cm , (0, 0)) is a germ of a holomorphic diffeo-
morphism and c : (Cm , 0) → (C, 0) is a germ of a holomorphic function.

3 Plane curves
We define some basic concepts of holomorphic curves C, including curvature, vertices
and evolute of such curves. We then consider the contact of C with lines and circles,
as is done for real plane case in [3]. All the concepts treated here are local in nature,
so we take a local parametrisation γ : D → C2 of C at a given point, where ϕ is a
holomorphic function and D is a connected and simply connected open set, taken as
small as necessary.

Definition 3.1 Let γ : D → C2 be a local parametrisation of a regular holomorphic

curve. Suppose that γ ′ (t) is not a C-lightlike vector for all t ∈ D. Then, we call the
vector
γ ′ (t)
T (t) = 1
⟨γ ′ (t), γ ′ (t)⟩ 2
the C-unit tangent vector to γ at t.
The vector
γ ′ (t)⊥
N (t) = T (t)⊥ = 1
⟨γ ′ (t), γ ′ (t)⟩ 2
is called the C-unit normal vector to γ at t.
The map N : D → CS 1 is called the C-Gauss map of the curve γ.
We say that γ is a C-unit speed parametrisation if γ ′ (t) ∈ CS 1 for all t ∈ D,
that is, ⟨γ ′ (t), γ ′ (t)⟩ = 1, for all t in D.
A point γ(t) is a C-lightlike point if γ ′ (t) is a C-lightlike vector.

If we write γ(t) = (z1 (t), z2 (t)), then

(z1′ (t), z2′ (t)) (−z2′ (t), z1′ (t))

T (t) = 1 , N (t) = 1 .
(z1′ (t)2 + z2′ (t)2 ) 2 (z1′ (t)2 + z2′ (t)2 ) 2

Remarks 3.2 1. In the expressions for T and N , we take the principal part of the
logarithm function when computing the square root. As the concepts here are local in
nature, this can be done, shrinking D if necessary.

4
 2. Observe that the Gauss map of a regular holomorphic curve with respect to the
Hermitian inner product is given by
(−z2′ (t), z1′ (t))
t 7→ 1 ,
(z1′ (t)z1′ (t) + z2′ (t)z2′ (t)) 2
and is not holomorphic. The C-Gauss map N is holomorphic.
3. As the bilinear form g is holomorphic, the C-lightlike points are isolated in D
or are the whole D. In the latter case, the curve is parallel to one of the C-lightlike
lines z2 = ±iz1 .
Theorem 3.3 Let γ : D → C2 be a local parametrisation of a regular holomorphic
curve. Suppose that t0 ∈ D is not a C-lightlike point. Then there is a simply connected
open subset D′ of D containing t0 such that γ|D′ can be reparametrised by C-unit speed.
Proof The proof follows the same steps as that for curves in the real plane; however
some necessary adjustment need to be made. Let
Z t
1
l(t) = ⟨γ ′ (u), γ ′ (u)⟩ 2 du,
t0

where the integral is taken along any path in D joining t0 and t. The function l : D → C
is holomorphic in some neighbourhood of t0 where ⟨γ ′ (t), γ ′ (t)⟩ = ̸ 0.
1
′ ′ ′
Since l (t0 ) = ⟨γ (t0 ), γ (t0 )⟩ ̸= 0, it follows by the inverse function theorem that
2

there exist a connected, simply connected open set D′ ⊂ D containing t0 such that
l : D′ → l(D′ ) is biholomorphic.
The local reparametrisation β(s) = γ(l−1 (s)) of the curve γ from l(D′ ) → C2 is
C-unit speed since
γ ′ (l−1 (s)) γ ′ (l−1 (s))
β ′ (s) = = 1 .
l′ (l−1 (s)) ⟨γ ′ (l−1 (s)), γ ′ (l−1 (s))⟩ 2
2
Remarks 3.4 1. In the real case, l′ (u) ̸= 0 for all u in the interval of definition of γ,
so l is strictly monotonous. Therefore, it has an inverse l−1 : l(I) → I. In the complex
case, the biholomorphicity of l is valid only locally at t0 .
2. Theorem 3.3 also holds for any regular holomorphic curve γ : D → Cn at points
t0 that are not C-lightlike, that is, points where ⟨γ ′ (t0 ), γ ′ (t0 )⟩ =
̸ 0.
We take now γ : D → C2 to be a C-unit speed local parametrisation of a holomor-
phic plane curve. Differentiation ⟨T (s), T (s)⟩ = 1 gives ⟨T ′ (s), T (s)⟩ = 0. This means
that T ′ (s) is parallel to N (s), so there exist a scalar function κ(s) such that
T ′ (s) = κ(s)N (s).
We call κ(s) the C-curvature of γ at s.

5
Proposition 3.5 Let γ : D → C2 be a C-unit speed local parametrisation of a holo-
morphic plane curve, with γ(t) = (z1 (t), z2 (t)). Then,
(1) κ(s) = ⟨T ′ (s), N (s)⟩ = (z1′ z2′′ − z2′ z1′′ )(s).
(2) N ′ (s) = −κ(s)T (s).

Proof Statement (1) is immediate.

For (2), differentiating ⟨T, T ⟩ = z1′2 + z2′2 = 1 gives z1′ z1′′ + z2′ z2′′ = 0. That, with
z1′ z2′′ − z2′ z1′′ = κ yields z1′′ = −κz2′ and z2′′ = κz1′ .
Differentiating N = (−z2′ , z1′ ), we get N ′ = (−z2′′ , z1′′ ) = −κ(z1′ , z2′ ) = −κT. 2

Remarks 3.6 1. For a general parametrisation γ, one can reparametrise by C-unit

speed, and show that the C-curvature has the same expression as in the real case,
namely,
z ′ z ′′ − z2′ z1′′
κ= 1 2 3 .
(z1′2 + z2′2 ) 2
2. The curve γ is a part of a line if and only if κ ≡ 0. It is a part of a complex
circle if and only if κ ≡ const ̸= 0.

Definition 3.7 A point t on γ is called an inflection if κ(t) = 0. It is called a

vertex if κ′ (t) = 0

Remarks 3.8 1. Inflections of a curve are affine invariant and are defined as points
where the curve has at least 3-point contact with its tangent line, equivalently, the
intersection multiplicity of the curve with its tangent line is greater than 2. It is not
difficult to show that this agrees with Definition 3.7 above (see §3.1).
2. Vertices of plane curves can be defined in a similar way using the contact of the
curve with complex circle, see [2, 14] and §3.2.

We have the following obvious observation as a corollary of Proposition 3.5.

Corollary 3.9 The C-Gauss map is a local diffeomorphism away from inflection
points. It has an A1 -singularity at ordinary inflection points (where κ′ (t) ̸= 0).

Inflections and vertices can be captured, as in the real case, by considering the
contact of the curve with, respectively, lines and circles.

3.1 Contact with lines

A line in C2 , with a C-orthogonal vector v, has equation

⟨z, v⟩ = c.

6
 We define the family of C-height functions H : D × CS 1 → C on γ by

H(t, v) = ⟨γ(t), v⟩.

The function H is holomorphic. For v fixed, the function Hv (t) = H(t, v) is the
C-height function along v. Its singularities measure the contact of the curve γ with
the lines C-orthogonal to v.
We have the following result, analogous to the real case ([3]).

Theorem 3.10 Let γ : D → C2 be a local parametrisation of a regular holomorphic

curve without C-lightlike points. Then,
(1) The function Hv is singular at t0 if and only if v = ±N (t0 ).
(2) For v = ±N (t0 ), the singularity of Hv at t0 is of type A1 if and only if κ(t0 ) ̸= 0.
It is of type A2 if and only if κ(t0 ) = 0 and κ′ (t0 ) ̸= 0, that is, t0 is an ordinary
inflection point of the curve.
(3) For generic curves, Hv can only have an A1 or A2 singularity and these are
+
R -versally unfolded by the family H.

3.2 Contact with circles

The contact of the curve γ at t0 with the circle of centre c passing through γ(t0 ) is
measured by the singularities of the (contact) function

dc (t) = ⟨γ(t) − c, γ(t) − c⟩. (2)

The function dc is holomorphic, and we call it the C-distance squared function.

Varying c ∈ C2 , yields the family C-distance squared functions d : D × C2 →
C, given by
d(t, c) = dc (t), (3)
which is also a holomorphic function.
We take γ to be a C-unit speed local parametrisation. Then, the function dc is
singular at s0 (i.e., has an A≥1 -singularity) if and only if
1 ′
d (s0 ) = ⟨T ′ (s0 ), γ(t0 ) − c⟩ = ⟨γ(t0 ) − c, T ′ (s0 )⟩ = 0,
2 c
that is, c = γ(s0 ) + λN (s), for some λ ∈ C. This means that the circle is tangent to
the curve at s0 if and only if its centre c belongs to the line through γ(s0 ) and parallel
to N (s0 ). We call this line the C-normal line to γ.
Differentiating twice, gives
1 ′′
d (s0 ) = κ(s0 )⟨N (s0 ), γ(s0 ) − c⟩ + 1.
2 c

7
 Suppose that s0 is not an inflection point. Then, the singularity of dc at s0 is of
type A≥2 if and only if d′c (s0 ) = d′′c (s0 ) = 0, equivalently,
1
c = e(s0 ) = γ(s0 ) + N (s0 ). (4)
κ(s0 )
Varying s0 in D, we get a parametrised plane curve, called the C-evolute of γ. The
circle of centre e(s0 ) and radius 1/κ(s0 ) is called the C-osculating circle of γ at s0 .
With c = e(s0 ), we have 12 d′′′ ′
c (s0 ) = 0 if and only if κ (s0 ) = 0. Thus, the singularity
of dc at t0 is of type A≥3 if and only if c = e(s0 ) and t0 is a vertex of γ.
We have the following result, in a similar way to the real case ([3]).

Theorem 3.11 Let γ : D → C2 be a local parametrisation of a regular holomorphic

curve without C-lightlike points. Then, the C-distance squared functions dc has a
singularity of type
A1 ⇐⇒ c ̸= e(t0 ) is on the C-normal line of γ at t0 .
A2 ⇐⇒ c = e(t0 ) and κ′ (t0 ) ̸= 0.
A3 ⇐⇒ c = e(t0 ), κ′ (t0 ) = 0 and κ′′ (t0 ) ̸= 0.
For a generic γ, the function dc has only the above singularities and these are
R+ -versally unfolded by the family d.
The bifurcation set of the family d is precisely the C-evolute of γ. Consequently,
the C-evolutes of a generic curve can only have cusp singularities.
′
κ (t)
As e′ (t) = − κ(t)2 N (t), we have the following consequence.

Corollary 3.12 The C-evolute of a regular curve γ is singular at a point t0 if and

only if t0 is a vertex of γ. Away from vertices, the tangent line to the C-evolute is the
C-normal line to the curve at the associated point. Consequently, the C-evolute is the
envelope of the C-normal lines of γ.

In the real case, the evolute of a plane curve is the envelope of the normal lines of
the curve. In the complex case, the equation of the normal line of the curve γ at t,
with respect to the Hermitian inner product, is given by

Gt (p) = ⟨γ(t) − p, γ ′ (t)⟩ = 0.

The family of functions G(t, p) = Gt (p) is not holomorphic, so we cannot define

the envelope as a discriminant ([1]). We identify C2 with R4 and use the following
definition of an envelope from [13]; see also [5]. A germ of a q-parameter family of
submanifolds of codimension m − n + q in Rm is a diagram of smooth map-germs of
the form
π f
Rq , 0 ←
− Rn , 0 →
− Rm , 0
that satisfies the following conditions:

8
 (a) π is a germ of a fibration;
(b) f restricted to each fibre π −1 (q) is a germ of a one to one immersion.
If S(f ) = {p ∈ Rn : dfp is not surjective}, then E(f (S(f ))) is the envelope of the
family.

Theorem 3.13 Let γ : D → C2 be a parametrisation of a regular holomorphic curve.

Then, the envelope of the family of the normal lines to γ, with respect to the Hermitian
inner product, is empty.

Proof We take γ a C-unit speed local parametrisation (away from C-lightlike points)
and consider the following diagram
π f
C← − C2 ,
−C×C→

where π(s, v) = v and f (s, v) = γ(s) + vγ ′ (s)⊥ . Then, the image of the fibre π −1 (s) is
the normal line to γ at s.
We consider the family of the normal lines to γ as a family of planes in R4 and
show that their envelope is empty.
We write s = s1 + is2 , v = v1 + iv2 , and γ(s) = (z1 (s), z2 (s)). Using the Wirtinger
derivatives, we get
∂f ∂f
∂s1
= (z1′ , z2′ ) + v(−z2′′ , z1′′ ), ∂v1
= (−z2′ , z1′ ),
∂f ∂f
∂s2
= (z1′ , z2′ ) + v(iz2′′ , −iz1′′ ), ∂v2
= i(−z2′ , z1′ ).

Identifying Ck with R2k , the determinant of the Jacobian matrix of df(s,v) is 1 +

||vκ(s)||2 > 0. Therefore, S(f ) is empty. Consequently, the normal lines to γ do not
have an envelope. 2

4 Space curves
In this section, we deal with holomorphic curves in C3 , without lightlike points. For
a given curve, we take γ : D → C3 a C-unit speed local parametrisation, with D as
in §3. We also assume that T ′ (s) is not C-lightlike, that is, ⟨T ′ (s), T ′ (s)⟩ ̸= 0 for all
s ∈ D, where T (s) = γ ′ (s).
With the above assumptions, there exists a scalar function κ and a C-unit vector
N such that
T ′ (s) = κ(s)N (s).
Observe that κ = ⟨T ′ , N ⟩ is a holomorphic function, and κ(s) ̸= 0 for all s ∈ D.
With ⟨T (s), T (s)⟩ = 1 we get ⟨T ′ (s), T (s)⟩ = 0, so κ(s)⟨N (s), T (s)⟩ = 0. Therefore,

⟨N (s), T (s)⟩ = 0.

9
 It follows that T and N are C-orthogonal. (Consequently, they are linearly inde-
pendent.)
We call the vector N (s) the C-unit normal vector to γ at s and call κ(s) the
C-curvature of γ at s.
Let e1 , e2 , e3 be the standard basis in C3 . The cross product of two vectors z =
(z1 , z2 , z3 ) and w = (w1 , w2 , w3 ) is the vector
e1 e2 e3
z×w = z1 z2 z3 = (z2 w3 − z3 w2 , z3 w1 − z1 w3 , z1 w2 − z2 w1 ).
w1 w2 w3
We define the C-binormal vector B(s) of γ at s as the vector
B(s) = T (s) × N (s).
It follows from the properties of the determinant that
⟨B(s), T (s)⟩ = 0 and ⟨B(s), N (s)⟩ = 0.
Proposition 4.1 The C-binormal vector B(s) is a C-unit vector.
The frame {T (s), N (s), B(s)} is C-orthogonal.
We call {T, N, B} the C-Frenet-Serret frame of γ
Proof The claim that B(s) ∈ CS 2 follows from the fact that T (s), N (s) ∈ CS 2 and
⟨N (s), T (s)⟩ = 0.
The determinant of the matrix of the coefficients of B(s), T (s), N (s) is equal to
⟨B(s), B(s)⟩ = 1, so the vectors form a basis of C3 . We showed above that the vectors
are pairwise C-orthogonal. 2
We have B ′ = T ′ × N + T × N ′ = T × N ′ , so ⟨B ′ , T ⟩ = ⟨T × N ′ , T ⟩ = 0.
Writing B ′ = λ1 T + λ2 N + λ3 B (by Proposition 4.1), it follows from ⟨B ′ , T ⟩ = 0,
and from the C-orthogonality of the vectors of the frame, that λ1 = 0.
From ⟨B(s), B(s)⟩ = 1, we get ⟨B ′ (s), B(s)⟩ = 0. It follows that ⟨λ2 N +λ3 B, B(s)⟩ =
λ3 = 0. Therefore, B ′ is parallel to N . We write
B ′ (s) = τ (s)N (s),
and call τ (s) the C-torsion of γ at s. Clearly, τ = ⟨B ′ , N ⟩ is a holomorphic function.
One can show that T = N × B and N = B × T . Then,
N ′ = B ′ × T + B × T ′ = −τ B − κT.
Theorem 4.2 The C-Frenet-Serret formulae for the moving frame T, N, B are:
T ′ = κN,
N ′ = −κT − τ B,
B ′ = τ N.

10
 Of course one can definite the C-Frenet-Serret frame, the C-curvature and C-
torsion for regular holomorphic curves not parametrised by C-unit speed (without
1
lightline points and assuming that (γ ′ (t)/⟨γ ′ (t), γ ′ (t)⟩ 2 )′ is not C-lightlike). Then, one
gets
1
⟨γ ′ × γ ′′ , γ ′ × γ ′′ ⟩ 2 ⟨γ ′ × γ ′′ , γ ′′′ ⟩
κ= 3 and τ = − ′ .
⟨γ ′ , γ ′ ⟩ 2 ⟨γ × γ ′′ , γ ′ × γ ′′ ⟩

4.1 Contact with planes

We define, as for plane curves, the family of C-height functions on a space curve with
a local parametrisation γ : D → C3 as the holomorphic function H : D × CS 2 → C,
where
H(t, v) = ⟨γ(t), v⟩.
We take γ a C-unit speed local parametrisation, with T (s) not a lightlike vector
for all s ∈ D. Then, as shown above, κ(s) ̸= 0 in D and T, N, B are C-orthogonal in
D.
The C-height function along v is defined as Hv (s) = H(s, v) and measures the
contact of the curve γ with planes C-orthogonal to v.
We have Hv′ = ⟨T, v⟩, and it vanishes at s0 if and only if ⟨v, T (s0 )⟩ = 0, that is,
v = λN (s0 ) + µB(s0 ), for some λ, µ ∈ C, with λ2 + µ2 = 1.
Now Hv′′ = κ⟨N, v⟩, so Hv′ (s0 ) = Hv′′ (s0 ) = 0 if and only if v = ±B(s0 ).
Differentiating again, we get Hv′′′ = κ′ ⟨N, v⟩ − κ⟨κT + τ B, v⟩. Thus, Hv′ (s0 ) =
Hv′′ (s0 ) = Hv′′′ (s0 ) = 0 if and only if v = ±B(s0 ) and τ (s0 ) = 0. With these conditions,
(4)
one can show that Hv (s0 ) ̸= 0 if and only if τ ′ (s0 ) ̸= 0
We call the plane C-orthogonal to T (s) the C-normal plane of γ at γ(s), and the
plane C-orthogonal to B(s) the C-osculating plane of γ at γ(s).
We have the following result, analogous to that for real space curves ([3]).

Theorem 4.3 Let γ : D → C3 be a local parametrisation of a generic regular holo-

1
morphic curve without lightlike points. Suppose that (γ ′ (t)/⟨γ ′ (t), γ ′ (t)⟩ 2 )′ is not a
C-lightlike vector for all t ∈ D. Then, the C-height function Hv can only have singu-
larities of type A1 , A2 , A3 and these occur at a given point t0 when:
A1 ⇐⇒ v ̸= ±B(t0 ) belongs to the C-normal plane of γ at t0 .
A2 ⇐⇒ v = ±B(t0 ) and τ (t0 ) ̸= 0.
A3 ⇐⇒ v = ±B(t0 ), τ (t0 ) = 0 and τ ′ (t0 ) ̸= 0.
The above singularities of Hv are R+ -versally unfolded by the family H.

4.2 Contact with lines

Here too one has to adapt to the complex framework as the usual family of orthogonal
projections is not holomorphic.

11
 We defined the C-orthogonal projection along v ∈ CS 2 as the parallel projection
along v to the plane C-orthogonal to v (i.e., orthogonal to v with respect to the
Hermitian inner product).
The projection Pv (p) = p + λv of a point p ∈ C3 is on the C-orthogonal plane to
v. Therefore, ⟨p + λv, v⟩ = 0, which gives λ = −⟨p, v⟩.
Observe that the C-orthogonal plane to v can be identified with the tangent plane to
the unit circle CS 2 at v. By varying v in CS 2 , we obtain the family of C-orthogonal
projections
P : C3 × CS 2 → T CS 2
(p, v) 7→ P (p, v) = (v, Pv (p)) = (v, p − ⟨p, v⟩v),

where T CS 2 is the tangent bundle of CS 2 . Clearly, P is a holomorphic map.

The family of C-orthogonal projections of a space curve γ is the restriction of P
to γ.

Remark 4.4 Any vector v ∈ CS 2 is transverse to the plane orthogonal to v with

respect to the Hermitian inner product as ⟨v, v⟩ = 1 ̸= 0. Therefore, for v fixed, Pv is
A-equivalent to the orthogonal projection with respect to the Hermitian inner product.
However, the orthogonal projection, with respect to the Hermitian inner product, is
given by P̃v (p) = p − ⟨p, v⟩v is not holomorphic with respect to v.

We take γ : D → C3 a C-unit speed parametrisation. The C-orthogonal projection

Pv (s) = γ(s) − ⟨γ(s), v⟩v of the curve γ along v at a given s0 ∈ D can be considered
as a map-germ (C, s0 ) → (C2 , Pv (γ(s0 ))).
We have
Pv′ (s) = T (s) − ⟨T (s), v⟩v.
This means that Pv is singular at s0 if and only if v = ±T (s0 ). Differentiating
again gives
Pv′′ (s) = κ(s)(N (s) − ⟨N (s), v⟩v).
Therefore, at a singularity s0 of Pv , we have Pv′′ (s0 ) = κ(s0 )N (s0 ). It follows that
the singularity of the defining equation of Pv (γ) is of type Ak (since κ(s0 ) ̸= 0).
We have

Pv′′′ (s) = κ′ (s)[N (s)−⟨N (s), v⟩v]−κ(s)[κ(s)T (s)+τ (s)B(s)−⟨κ(s)T (s)+τ (s)B(s), v⟩v],

so at a singularity s0 of Pv we have

Pv′′′ (s0 ) = κ′ (s0 )N (s0 ) − κ(s0 )τ (s0 )B(s0 ).

It follows that Pv′′ (s0 ) and Pv′′′ (s0 ) are linearly independent if and only if τ (s0 ) ̸= 0.
In that case, the singularity of the projection is of type A2 , that is, its germ at s0 is
A-equivalent to the cusp t 7→ (t2 , t3 ).

12
 With v = ±T (s0 ) and τ (s0 ) = 0, we have
(4)
Pv (s0 ) = (κ′′ (s0 ) − κ3 (s0 ))N (s0 ) − κ(s0 )τ ′ (s0 )B(s0 ),
(5)
Pv (s0 ) = (κ′′′ (s0 ) − 6κ′ (s0 )κ2 (s0 ))N (s0 ) − (3κ′ (s0 )τ ′ (s0 ) + κ(s0 )τ ′′ (s0 ))B(s0 ).

In that case, the singularity of Pv is of type A4 , that is, Pv ∼A (t2 , t5 ), if and only
if (κ′ τ ′ − 3κτ ′′ )(s0 ) ̸= 0.
We have the following results, analogous to the real case ([4]).

Theorem 4.5 Let γ : D → C3 be a local parametrisation of a generic regular holo-

1
morphic curve without lightlike points. Suppose that (γ ′ (t)/⟨γ ′ (t), γ ′ (t)⟩ 2 )′ is not a
C-lightlike vector for all t ∈ D. Then, the C-orthogonal projection Pv along v can only
have local singularities of type A2 or A4 . We get, at t0 , a singularity of type
A2 ⇐⇒ v is tangent to γ at t0 and τ (t0 ) ̸= 0.
A4 ⇐⇒ v is tangent to γ at t0 , τ (t0 ) = 0 and (κ′ τ ′ − 3κτ ′′ )(t0 ) ̸= 0.
The above singularities are Ae -versally unfolded by the family P .

5 Surfaces in C3
Let M be surface in C3 , that is, a complex 2-dimensional sub-manifold of C3 . We
study the C-extrinsic geometry of M and work locally at a given point on M . We
take a local parametrisation ϕ : U → C3 of M at that point. The map ϕ is a regular
holomorphic map, and we write M = ϕ(U ). We take U to be sufficiently small so that
the principal branch of the square root of any non-vanishing holomorphic function on
U used here is well defined (and is holomorphic).
We define the C-first fundamental form of M at p = ϕ(q) ∈ M as the quadratic
form Ip (v) = g(v, v) = ⟨v, v⟩, for all v ∈ Tp M . We denote

E = ⟨ϕz1 , ϕz1 ⟩, F = ⟨ϕz1 , ϕz2 ⟩, G = ⟨ϕz2 , ϕz2 ⟩,

the coefficients of the C-first fundamental form I. Then, for any v = aϕz1 (q)+bϕz2 (q) ∈
Tp M , we have
Ip (v) = a2 E + 2abF + b2 G.
The bilinear symmetric form g restricted to M may be degenerate on some subset
of M , that is, there may exist points p on M and non-zero tangent vectors v ∈ Tp M ,
such that Ip (v) = g(v, v) = 0. Equivalently, Tp M may contain C-lightlike vectors.
This happens if and only if δ(q) = (EG − F 2 )(q) = 0.
We call the set of points p = ϕ(q) on M where δ(q) = 0 the C-lightlike locus of
M and identify it with its pre-image on U . We write

LD = {(z1 , z2 ) ∈ U : (EG − F 2 )(z1 , z2 ) = 0}.

13
 Assumption: We assume that the LD on M is either empty or is a curve on M ,
so its complement in M is an open and dense subset of M .
At points on M \ LD, ϕz1 and ϕz2 are non-lightlike and the C-unit vector
ϕz1 × ϕz2
N (z1 , z2 ) = 1 (z1 , z2 )
⟨ϕz1 × ϕz2 , ϕz1 × ϕz2 ⟩ 2
is well defined. We have ⟨ϕz1 , N ⟩ = ⟨ϕz2 , N ⟩ = 0, so N (z1 , z2 ) is C-orthogonal to Tp M ,
where p = ϕ(z1 , z2 ). (The vector N (z1 , z2 ) is parallel to the conjugate of the normal
vector to M at p with respect to the Hermitian inner product.)
We call the map N : M \ LD → CS 2 , with N (z1 , z2 ) as above, the C-Gauss map
of M . It is worth observing that N is a holomorphic map (unlike the Gauss map with
respect to the Hermitian inner product).
Differentiating the identity ⟨N, N ⟩ = 1, gives ⟨Nz1 , N ⟩ = ⟨Nz2 , N ⟩ = 0. It follows
that dNp : Tp M → TN (p) CS 2 ≃ Tp M is a linear operator on Tp M .
From ⟨ϕz1 , N ⟩ = ⟨ϕz2 , N ⟩ = 0, we get
⟨ϕz1 z2 , N ⟩ + ⟨ϕz1 , Nz2 ⟩ = 0,
⟨ϕz2 z1 , N ⟩ + ⟨ϕz2 , Nz1 ⟩ = 0.

It follows that ⟨ϕz1 , Nz2 ⟩ = ⟨ϕz2 , Nz1 ⟩, equivalently, ⟨Nz2 , ϕz1 ⟩ = ⟨Nz1 , ϕz2 ⟩. This
means that g(dNp (ϕz1 ), ϕz2 ) = g(dNp (ϕz2 ), ϕz1 ), with g as in (1). Therefore, the bilin-
ear form on Tp M , given by
IIp (u, v) = g(−dNp (u), v) = ⟨−dNp (u), v⟩,
is symmetric. We call the quadratic form IIp (v, v) the C-second fundamental form
of M and denote it IIp (v). We denote
l = −⟨Nz1 , ϕz1 ⟩ = ⟨N, ϕz1 z1 ⟩,
m = −⟨Nz1 , ϕz2 ⟩ = ⟨N, ϕz1 z2 ⟩,
n = −⟨Nz2 , ϕz2 ⟩ = ⟨N, ϕz2 z2 ⟩,
the coefficients of the C-second fundamental form on M \ LD. We call the linear
operator −dNp the C-shape operator.
For any v = aϕz1 (q) + bϕz2 (q) ∈ Tp M , we have
IIp (v) = a2 l + 2abm + b2 n.
A direction v ∈ Tp M is called a C-asymptotic direction if IIp (v) = 0.
Following the same calculations as those for surfaces in the Euclidean space (see,
for example, [8]), the matrix of the C-shape operator −dNp , with respect to the basis
ϕz1 , ϕz2 of Tp M is
 −1  
E F l m
Ap = .
F G m n

14
 Since IIp is a symmetric bilinear form, its matrix Ap has two eigenvalues κ1 , κ2
and two C-orthogonal eigenvectors e1 , e2 . We call κ1 , κ2 the C-principal curvatures
and e1 , e2 the C-principal directions. We denote by K = κ1 κ2 = det(Ap ) the
C-Gaussian curvature of M . We have
ln − m2
K= .
EG − F 2
The set of points on where K(p) = 0 is called the C-parabolic set. A point p is
called C-umbilic if κ1 (p) = κ2 (p).
With this framework, we can study the contact of the surface M \ LD with planes
(these have zero C-Gaussian curvature), complex sphere (which have constant C-
Gaussian curvature) and lines. The resulting families of mappings that measure the
contact between M with these objects are holomorphic.

5.1 Contact with planes

We define, as for plane curves, the family C-height functions on the image of a local
parametrization ϕ : U → C3 of a surface as the holomorphic function H : U ×CS 2 → C,
where
H(q, v) = ⟨ϕ(q), v⟩.
The C-height function Hv (q) = H(q, v) is singular at q if and only if v = ±N (q).
The singularity is of type A1 if and only if ln − m2 ̸= 0, that is, q is not a C-parabolic
point. We have the following result, analogous to that for surfaces in the Euclidean
3-space (see [9, Chapter 6]).
Theorem 5.1 For a generic complex surface M in C3 , the height function Hv on M \
LD can have local singularities of type A1 , A2 , A3 and these are R+ -versally unfolded
by the family of C-height functions. The singularities occur at a point p ∈ M \ LD
when:
A1 ⇐⇒ v = ±N (p) and p is not a C-parabolic point.
A2 ⇐⇒ v = ±N (p), p is C-parabolic point and the unique C-asymptotic direction
at p is transverse to the parabolic curve.
A3 ⇐⇒ v = ±N (p), p is C-parabolic point and the unique C-asymptotic direction
at p is tangent to the parabolic curve at p. We call such points the C-cusps of Gauss.
We can choose a suitable coordinate system in C3 parametrised the surface M
locally at a given point p0 ∈ M in Monge form ϕ(z1 , z2 ) = (z1 , z2 , f (z1 , z2 )), for
some holomorphic function f with zero 1-jet at the origin. The C-normal vector
to M at the origin is (0, 0, 1) and nearby vectors in CS 2 can be parametrised by
(v1 , v2 , 1) with (v1 , v2 ) ∈ (C2 , 0). We get the germ of the family of C-height functions
H : (C2 × C2 , (0, 0)) → (C, 0), given by
H(z1 , z2 , v1 , v2 ) = z1 v1 + z2 v2 + f (z1 , z2 ). (5)

15
 The algebraic conditions on the Taylor expansion of f at the origin for H(0,0) to
have one of the singularities listed in Theorem 5.1, and for these singularities to be
R+ -versally unfolded by the family H in (5), are as given in [9, Chapter 6] (interpreting
the coefficients in [9] as complex numbers).

5.2 Contact with spheres

The contact of the surface M = ϕ(U ) at p with the complex sphere of centre c passing
through p is measured by the singularities of the (contact) function

dc (q) = ⟨ϕ(q) − c, ϕ(q) − c⟩. (6)

The function dc is holomorphic, and we call it the C-distance squared function.

The family of C-distance squared functions d : U × C3 → C is given by

d(q, c) = dc (q) = ⟨ϕ(q) − c, ϕ(q) − c⟩. (7)

The function dc measures the contact of M with the complex spheres of centre c.
We have the following result, analogous to that for surfaces in the Euclidean 3-space
(see [9, Chapter 6]).

Theorem 5.2 For a generic complex surface M in C3 , the C-distance squared func-
tion dc on M \ LD can have local singularities of type A1 , A2 , A3 , A4 , D4 and these are
R+ -versally unfolded by the family of C-distance squared functions.
The singularities occur at a point p = ϕ(q) ∈ M \ LD when:
A1 : c is on the C-normal line of M at p.
A2 : c is a C-focal point, that is, c = ϕ(q) + κi1(q) N (q), i = 1 or i = 2.
A3 : c is a focal point and is a generic point on the C-ridge curve, where one of
the C-principal curvature is extremal along its associated line of curvature.
A3 : c is a special point on the C-ridge curve.
D4 : κ1 (q) = κ2 (q), that is, p is a C-umbilic, and c = 1/κ1 (p).

If we take M locally in Monge form (as in §5.1), we get the germ of the family of
C-distance squared functions d : (C2 × C3 , (0, (0, 0, c0 ))) → (C, 0), given by

d(z1 , z2 , (a, b, c)) = (z1 − a)2 + (z1 − a)2 + (f (z1 , z2 ) − (c + c0 ))2 . (8)

The algebraic conditions on the Taylor expansion of f at the origin for d(0,0,c0 ) to
have one of the singularities listed in Theorem 5.2, and for these singularities to be
R+ -versally unfolded by the family d in (8), are as given in [9, Chapter 6].

16
5.3 Contact with lines
The family of C-orthogonal projections in C3 is as given in §4.2 and is as follows:

P : C3 × CS 2 → T CS 2
(p, v) 7→ P (p, v) = (v, Pv (p)) = (v, p − ⟨p, v⟩v),

Theorem 5.3 For a generic complex surface M in C3 , the C-orthogonal projection Pv

on M \LD can have local singularities of Ae -codimension ≤ 2 and these are Ae -versally
unfolded by the family P of C-orthogonal projections on M \ LD.

Take M locally in Monge form (as in §5.1) and project along directions close to
v0 = (0, 1, 0) to a fixed plane C-orthogonal to v0 . The directions are parametrised
by (v1 , 1, v2 ) and the germ of the family of C-orthogonal projections can be taken as
P : (C2 × C2 , (0, 0)) → (C2 , 0), and given, after a change of coordinate in the source,
by
P (z1 , z2 , v1 , v2 ) = (z1 , f (z1 + v1 z2 , z2 ) − v2 z2 ). (9)
Here too, the algebraic conditions on the Taylor expansion of f at the origin for
P(v1 ,v2 ) to have one of Ae -codimension ≤ 2 singularities of map-germs from the plane
to the plane, and for these singularities to be Ae -versally unfolded by the family P in
(8), are as given in [9, Chapter 6].

Remark 5.4 The families (5), (8), (9) can be seen as the complexification of their
real counterpart on analytic surfaces in R3 . The framework highlighted in this paper
gives a geometric interpretation of the singularities of the members of the complexified
families of these maps.

Acknowledgement: ADF was supported by the Fundação de Amparo à Pesquisa do

Estado de São Paulo (FAPESP) doctoral grant 2023/11669-0. FT was supported by
the FAPESP Thematic project grant 2019/07316-0.

References
[1] J. W. Bruce, Envelopes and characteristics. Math. Proc. Cambridge Philos. Soc.
100 (1986), 475–492.

[2] J. W. Bruce, M. A. C. Fernandes and F. Tari, Geometric invariants of singular

plane curves. Preprint.

[3] J. W. Bruce and P. J. Giblin, Curves and Singularities. Cambridge University

Press, 1984. Second Edition, 1992.

17
 [4] J. M. S. David, Projection-generic curves. J. London Math. Soc. 27 (1983), 552–
562.

[5] M. J. Dias Carneiro, Singularities of envelopes of families of submanifolds in Rn .

Ann. Sci. École Norm. Sup. 16 (1983), 173–192.

[6] R. Cipolla and P. Giblin, Visual motion of curves and surfaces. Cambridge Uni-
versity Press, 2000.

[7] J. Damon, P. Giblin and G. Haslinger, Local feature in natural images via singu-
larity theory. Lecture Notes in Mathematics 2165, Springer 2016.

[8] M. P. do Carmo, Differential Geometry of Curves and Surfaces. Prentice-Hall,

Inc., Englewood Cliffs, NJ, 1976,

[9] S. Izumiya, M. C. Romero Fuster, M. A. S. Ruas and F. Tari, Differential Geom-

etry from a Singularity Theory Viewpoint. World Scientific Publishing Company,
2015.

[10] S. Izumiya, D-H. Pei and T. Sano, Singularities of hyperbolic Gauss maps. Proc.
London Math. Soc. 86 (2003), 485–512.

[11] K. Nomizu and B. Smyth, Differential geometry of complex hypersurfaces II*. J.

Math. Soc. Japan, 20 (1968), 498–521.

[12] I. R. Porteous, Geometric differentiation for the intelligence of curves and sur-
faces. Cambridge University Press, Cambridge, 1994.

[13] R. Thom, Sur la théorie des enveloppes. J. Math. Pures Appl. 41 (1962), 177–192.

[14] A. O. Viro, Differential geometry “in the large” of plane algebraic curves, and
integral formulas for invariants of singularities Zap. Nauchn. Sem. S.-Peterburg.
Otdel. Mat. Inst. Steklov. (POMI) 231 (1995), 255–268, 327. J. Math. Sci. (New
York) 91 (1998), 3499–3507.

[15] C. T. C. Wall, Finite determinacy of smooth map-germs. Bull. Lond. Math. Soc.
13 (1981), 481–539.

Instituto de Ciências Matemáticas e de Computação - USP, Avenida Trabalhador são-

carlense, 400, Centro, CEP: 13566-590, São Carlos - SP, Brazil.
E-mails:
[email protected]
[email protected]

18