Preface

The present book is about differential geometry of space

curves and surfaces. The formulation and presentation are
largely based on a tensor calculus approach, which is the
dominant trend in the modern mathematical literature of this
subject, rather than the geometric approach which is usually
found in some old style books. The book is prepared, to some
extent, as part of tutorials about topics and applications related
to tensor calculus. It can therefore be used as part of a course
on tensor calculus as well as a textbook or a reference for an
intermediate-level course on differential geometry of curves
and surfaces.
Apart from general background knowledge in a number of
mathematical branches such as calculus, geometry and algebra,
an important requirement for the reader and user of this book is
familiarity with the terminology, notation and concepts of
tensor calculus at reasonable level since many of the notations
and concepts of differential geometry in its modern style are
based on tensor calculus.
The book contains a mathematical background section in the
first chapter to outline some important pre-required
mathematical issues. However, this section is restricted to
materials related directly to the contents of differential
geometry of the book and hence the reader and user should not
expect this mathematical background section to be
comprehensive in any way. General mathematical knowledge,
plus possible consultation of mathematical textbooks related to
other disciplines of mathematics when needed, should
therefore be considered.
The book is furnished with an index in the end of the book as
well as sets of exercises in the end of each chapter to provide
useful revisions and practice. To facilitate linking related
concepts and parts, and hence ensure better understanding of
the provided materials, cross referencing is used extensively
throughout the book where these referrals are hyperlinked in
the electronic version of the book for the convenience of the
ebook users. The book also contains a considerable number of
graphic illustrations to help the readers and users to visualize
the ideas and understand the abstract concepts.
The materials of differential geometry are strongly interlinked
and hence any text about the subject, like the present one, will
face the problem of arranging the materials in a natural order to
ensure gradual development of concepts. In this book we
largely followed such a scheme. However, this is not always
possible and hence in some cases references are provided to
materials in later parts of the book for concepts needed in
earlier parts. Nevertheless, in most cases brief definitions of
the main concepts are provided in the first chapter in
anticipation of more detailed definitions and investigations in
the subsequent chapters.
Regarding the preparation of the book, everything is made by
the author including all the graphic illustrations, indexing,
typesetting, book cover, as well as overall design. In this
regard, I should acknowledge the use of LaTeX typesetting
package and the LaTeX based document preparation package
LyX which facilitated the typesetting and design of the book
substantially.
Taha Sochi
London, March 2017
Table of Contents
Preface
Nomenclature
1: Preliminaries
1.1: Differential Geometry
1.2: General Remarks, Conventions and Notations
1.3: Classifying the Properties of Curves and Surfaces
1.3.1: Local versus Global Properties
1.3.2: Intrinsic versus Extrinsic Properties
1.4: General Mathematical Background
1.4.1: Geometry and Topology
1.4.2: Functions
1.4.3: Coordinates, Transformations and Mappings
1.4.4: Intrinsic Distance
1.4.5: Basis Vectors
1.4.6: Flat and Curved Spaces
1.4.7: Homogeneous Coordinate Systems
1.4.8: Geodesic Coordinates
1.4.9: Christoffel Symbols for Curves and Surfaces
1.4.10: Riemann-Christoffel Curvature Tensor
1.4.11: Ricci Curvature Tensor and Scalar
1.5: Exercises
2: Curves in Space
2.1: General Background about Curves
2.2: Mathematical Description of Curves
2.3: Curvature and Torsion of Space Curves
2.3.1: Curvature
2.3.2: Torsion
2.4: Geodesic Torsion
2.5: Relationship between Curve Basis Vectors and their
Derivatives
2.6: Osculating Circle and Sphere
2.7: Parallelism and Parallel Propagation
2.8: Exercises
3: Surfaces in Space
3.1: General Background about Surfaces
3.2: Mathematical Description of Surfaces
3.3: Surface Metric Tensor
3.3.1: Arc Length
3.3.2: Surface Area
3.3.3: Angle Between Two Surface Curves
3.4: Surface Curvature Tensor
3.5: First Fundamental Form
3.6: Second Fundamental Form
3.6.1: Dupin Indicatrix
3.7: Third Fundamental Form
3.8: Fundamental Forms
3.9: Relationship between Surface Basis Vectors and their
Derivatives
3.9.1: Codazzi-Mainardi Equations
3.10: Sphere Mapping
3.11: Global Surface Theorems
3.12: Exercises
4: Curvature
4.1: Curvature Vector
4.2: Normal Curvature
4.2.1: Meusnier Theorem
4.3: Geodesic Curvature
4.4: Principal Curvatures and Directions
4.5: Gaussian Curvature
4.6: Mean Curvature
4.7: Theorema Egregium
4.8: Gauss-Bonnet Theorem
4.9: Local Shape of Surface
4.10: Umbilical Point
4.11: Exercises
5: Special Curves
5.1: Straight Line
5.2: Plane Curve
5.3: Involute and Evolute
5.4: Bertrand Curve
5.5: Spherical Indicatrix
5.6: Spherical Curve
5.7: Geodesic Curve
5.8: Line of Curvature
5.9: Asymptotic Line
5.10: Conjugate Direction
5.11: Exercises
6: Special Surfaces
6.1: Plane Surface
6.2: Quadratic Surface
6.3: Ruled Surface
6.4: Developable Surface
6.5: Isometric Surface
6.6: Tangent Surface
6.7: Minimal Surface
6.8: Exercises
7: Tensor Differentiation over Surfaces
7.1: Exercises
References
Author Notes
8: Footnotes
Nomenclature
In the following table, we define some of the common
symbols, notations and abbreviations which are used in the
book to provide easy access to the reader.
∇ nabla differential operator
∇ 2 Laplacian operator
~ isometric to
, subscript partial derivative with respect to the following
index(es)
; subscript covariant derivative with respect to the
following index(es)
1D, 2D, 3D, one-, two-, three-, n-dimensional
nD
overdot (e.g. derivative with respect to general parameter t
r)
prime (e.g. derivative with respect to natural parameter s
r’)
δ ⁄ δt absolute derivative with respect to t
∂α, ∂i partial derivative with respect to αth and ith
variables
a determinant of surface covariant metric tensor
a surface covariant metric tensor
a11, a12, a22 coefficients of surface covariant metric tensor
a11, a12, a22 coefficients of surface contravariant metric
tensor
aαβ, aαβ, aβα surface metric tensor or its components
b determinant of surface covariant curvature
tensor
b surface covariant curvature tensor
B binormal unit vector of space curve
b11, b12, b22 coefficients of surface covariant curvature tensor
bαβ, bαβ, bβα surface curvature tensor or its components
C curve
C̃B, C̃N, C̃T spherical indicatrices of curve C
Ce, Ci evolute and involute curves
Cn of class n
cαβ, cαβ, cβα tensor of third fundamental form or its
components
d Darboux vector
d1, d2 unit vectors in Darboux frame
det determinant of matrix
ds length of infinitesimal element of curve
dsB, dsN, dsT length of line element in binormal, normal,
tangent directions
dσ area of infinitesimal element of surface
e, f, g coefficients of second fundamental form
E, F, G coefficients of first fundamental form
ℰ, ℱ, V number of edges, faces and vertices of
polyhedron
Ei, Ej covariant and contravariant space basis vectors
Eα, Eβ covariant and contravariant surface basis vectors
Eq./Eqs. Equation/Equations
f function
Fig./Figs. Figure/Figures
g topological genus of closed surface
gij, gij space metric tensor or its components
H mean curvature
IS, IIS, IIIS first, second and third fundamental forms
IS, IIS tensors of first and second fundamental forms
iff if and only if
J Jacobian of transformation between two
coordinate systems
J Jacobian matrix
K Gaussian curvature
Kt total curvature
L length of curve
n normal unit vector to surface
N principal normal unit vector to curve
P point
r, R radius
ℛ Ricci curvature scalar
r position vector
rα, rαβ 1st and 2nd partial derivative of r with
subscripted variables
R1, R2 principal radii of curvature
ℝn n-dimensional space (usually Euclidean)
Rij, Ri j Ricci curvature tensor of 1st and 2nd kind for
space
Rαβ, Rα β Ricci curvature tensor of 1st and 2nd kind for
surface
Rijkl Riemann-Christoffel curvature tensor of 1st kind
for space
Rαβγδ Riemann-Christoffel curvature tensor of 1st kind
for surface
Ri jkl Riemann-Christoffel curvature tensor of 2nd kind
for space
Rα βγδ Riemann-Christoffel curvature tensor of 2nd kind
for surface
Rκ radius of curvature
Rτ radius of torsion
s natural parameter of curve representing arc
length
S surface
ST tangent surface of space curve
t general parameter of curve
T function period
T tangent unit vector of space curve
TPS tangent space of surface S at point P
tr trace of matrix
u geodesic normal vector
u1, u2 surface coordinates
uα surface coordinate
u, v surface coordinates
xi space coordinate
xiα surface basis vector in full tensor notation
x, y, z coordinates in 3D space (usually Cartesian)
[ij, k] Christoffel symbol of 1st kind for space
[αβ, γ] Christoffel symbol of 1st kind for surface
Γkij Christoffel symbol of 2nd kind for space
Γγαβ Christoffel symbol of 2nd kind for surface
δij, δij, δji covariant, contravariant and mixed Kronecker
delta
δijkl generalized Kronecker delta
Δ discriminant of quadratic equation
ϵi1…in, ϵi1…in covariant, contravariant relative permutation
tensor in nD space
εi1…in, εi1…in covariant, contravariant absolute permutation
tensor in nD space
θ angle or parameter
θs sum of interior angles of polygon
κ curvature of curve
κ1, κ2 principal curvatures of surface at a given point
κB, κT curvature of binormal and tangent spherical
indicatrices
κg, κn geodesic and normal curvatures
κgu, κgv geodesic curvatures of u and v coordinate curves
κnu, κnv normal curvatures of u and v coordinate curves
K curvature vector
Kg, Kn geodesic and normal components of curvature
vector
λ direction parameter of surface
ξ real parameter
ρ pseudo-radius of pseudo-sphere
ρ, φ polar coordinates of plane
ρ, φ, z cylindrical coordinates of 3D space
σ area of surface patch
τ torsion of curve
τB, τT torsion of binormal and tangent spherical
indicatrices
τg geodesic torsion
φ angle or parameter
χ Euler characteristic
ω real parameter

Note: due to the restrictions on the availability and visibility of

symbols in the mobi format, as well as similar formatting
issues, we should draw the attention of the ebook readers to the
following points:
1. Bars over symbols, which are used in the printed version,
were replaced by tildes. However, for convenience we kept
using the terms “barred” and “unbarred” in the text to refer to
the symbols with and without tildes.
2. The square root symbol in mobi is √( ) where the argument
is contained inside the parentheses. For example, the square
root of a is symbolized as √(a).
3. In the mobi format, superscripts are automatically displayed
before subscripts unless certain measures are taken to force the
opposite which may distort the look of the symbol and may not
even be the required format when the superscripts and
subscripts should be side by side which is not possible in the
mobi text and live equations. Therefore, for convenience and
aesthetic reasons we only forced the required order of the
subscripts and superscripts or used imaged symbols when it is
necessary; otherwise we left the symbols to be displayed
according to the mobi choice although this may not be ideal
like displaying the Christoffel symbol of the second kind as:
Γγαβ or the generalized Kronecker delta as: δαδβω instead of

their normal look as: and .

Chapter 1
Preliminaries
In this chapter, we provide preliminary materials in the form of
a general introduction about differential geometry, remarks
about the conventions and notations used in this book,
classification of the properties of curves and surfaces, and a
general mathematical background related to differential
geometry of curves and surfaces.
1.1 Differential Geometry
Differential geometry is a branch of mathematics that largely
employs methods and techniques of other branches of
mathematics such as differential and integral calculus,
topology and tensor analysis to investigate geometrical issues
related to abstract objects, such as space curves and surfaces,
and their properties where these investigations are mostly
focused on these properties at small scales. The investigations
of differential geometry also include characterizing categories
of these objects. There is also a close link between differential
geometry and the disciplines of differential topology and
differential equations. Differential geometry may be contrasted
with “algebraic geometry” which is another branch of
geometry that uses algebraic tools to investigate geometric
issues mainly of global nature.
The investigation of the properties of curves and surfaces in
differential geometry are closely linked. For instance,
investigating the characteristics of space curves is extensively
exploited in the investigation of surfaces since common
properties of surfaces are defined and quantified in terms of the
properties of curves embedded in these surfaces. For example,
several aspects of the surface curvature at a point are defined
and quantified in terms of the parameters of the surface curves
passing through that point.
1.2 General Remarks, Conventions and
Notations
First, we should remark that the present book is largely based
on investigating curves and surfaces embedded in a 3D flat
space coordinated by a rectangular Cartesian system. In most
cases, “surface” and “space” in the present book mean 2D and
3D manifolds respectively.
Another remark is that twisted curves can reside in a 2D
manifold (surface) or in a higher dimensionality manifold
(usually 3D space). Hence we usually use “surface curves” and
“space curves” to refer to the type of the manifold of residence.
However, in most cases a single curve can be viewed as
resident of more than one manifold and hence it is a surface
and space curve at the same time. For example, a curve
embedded in a surface which in its turn is embedded in a 3D
space is a surface curve and a space curve at the same time.
Consequently, in this book these terms should be interpreted
flexibly. Many statements formulated in terms of a particular
type of manifold can be correctly and easily extended to
another type with minimal adjustments of dimensionality and
symbolism. Moreover, “space” in some statements should be
understood in its general meaning as a manifold embracing the
curve not as opposite to “surface” and hence it can include a
2D space, i.e. surface.
Following the convention of several authors, when discussing
issues related to 2D and 3D manifolds the Greek indices range
over 1, 2 while the Latin indices range over 1, 2, 3. Therefore,
the use of Greek and Latin indices should in general indicate
the type of the intended manifold unless it is stated otherwise.
We use u indexed with superscript Greek letters (e.g. uα and
uβ) to symbolize surface coordinates (see Footnote 1 in § 8↓)
while we largely use x indexed with superscript Latin letters
(e.g. xi and xj) to represent space Cartesian coordinates
although they are sometimes used to represent space general
curvilinear coordinates. Comments are usually added to clarify
the situation if necessary.
A related issue is that the indexed E are mostly used for the
surface, rather than the space, basis vectors where they are
subscripted or superscripted with Greek indices, e.g. Eα and
Eβ. However, in some cases indexed E are used for space basis
vectors; in which case they are distinguished by using Latin
indices, e.g. Ei and Ej. When the basis vectors are indexed
numerically rather than symbolically, the distinction between
surface and space bases should be obvious from the context if
it is not stated explicitly.
Regarding the Christoffel symbols of the first and second kind
of various manifolds, they may be based on the space metric or
the surface metric. Hence, when a number of Christoffel
symbols in a certain context or equation are based on more
than one metric, the type of indices, i.e. Greek or Latin, can be
used as an indicator to the underlying metric where the Greek
indices represent the surface (e.g. [αβ, γ] and Γγαβ) while the
Latin indices represent the space (e.g. [ij, k] and Γkij).
Nevertheless, comments are generally added to account for
potential oversight. In particular, when the Christoffel symbols
are numbered (e.g. Γ122), instead of being indexed
symbolically, comments will be added to clarify the situation.
For brevity, convenience and clean notation in certain contexts,
we use an overdot (e.g. r) to indicate derivative with respect to
a general parameter t while we use a prime (e.g. r’) to indicate
derivative with respect to a natural parameter s representing arc
length. For the same reasons, subscripts are also used
occasionally to symbolize partial derivative with respect to the
subscript variables, e.g. fv and rαβ. We should also remark that
we follow the summation convention, which is largely used in
tensor calculus. Comments are added in a few exceptional
cases where this convention does not apply.
We deliberately use a variety of notations for the same
concepts for the purpose of convenience and to familiarize the
reader with different notations all of which are in common use
in the literature of differential geometry and tensor calculus.
Having proficiency in these subjects requires familiarity with
these various, and sometimes conflicting, notations. Moreover,
in some situations the use of one of these notations or the other
is either necessary or advantageous depending on the location
and context. An important example of using different notations
is that related to surface coordinates where we use both u, v
and u1, u2 to represent these coordinates since each one of
these notations has advantages over the other depending on the
context; moreover the latter is necessary for expressing the
equations of differential geometry in indicial tensor forms.
Another important example related to the use of a variety of
notations is the employment of different symbols for the
coefficients of the first and second fundamental forms (i.e. E,
F, G, e, f, g) on one hand and the coefficients of the surface
covariant metric tensor and the surface covariant curvature
tensor (i.e. a11, a12, a22, b11, b12, b22) on the other despite the
equivalence of these coefficients, that is
(E, F, G, e, f, g) = (a11, a12, a22, b11, b12, b22),
and hence these different notations can be replaced by just one.
However, we keep both notations as they are both in common
use in the literature of differential geometry and tensor
calculus; moreover in many situations the use of one of these
notations in particular is either necessary or advantageous, as
indicated earlier.
We should also mention that we use bold face symbols (e.g. A)
to mark tensors of rank > 0 (including matrices which
represent such tensors) in their symbolic notation, while we
use normal face symbols (e.g. c) for labeling scalars as well as
tensors of rank > 0 and their components in their indicial form
where in the latter case the symbols are indexed (e.g. Ai and
bβα). Also, square brackets containing arrays are used to
represent matrices while vertical bars are used to represent
determinants.
1.3 Classifying the Properties of Curves
and Surfaces
There are two main classifications for the properties of curves
and surfaces embedded in higher dimensionality spaces; these
classifications are local properties versus global properties, and
intrinsic properties versus extrinsic properties. In the following
subsections we briefly investigate these overlapping
classifications.

1.3.1 Local versus Global Properties

The properties of curves and surfaces may be categorized into

two main groups: local and global where these properties
describe the geometry of the curves and surfaces in the small
and in the large respectively. The local properties correspond
to the characteristics of the object in the immediate
neighborhood of a point on the object such as the curvature of
a curve or surface at that point, while the global properties
correspond to the characteristics of the object on a large scale
and over extended parts of the object such as the number of
stationary points of a curve or a surface or being a one-side
surface like Mobius strip (Fig. 1↓) which is locally a double-
side surface.
Figure 1 Mobius strip.

As indicated earlier, differential geometry of space curves and

surfaces is mainly concerned with the local properties. The
investigation of global properties normally involve topological
treatments which are beyond the common tools and methods
that are usually employed in differential geometry. In fact,
there is a special branch of differential geometry dedicated to
the investigation of global (or in the large) properties.
Regarding the present book, it is mostly limited to differential
geometry in the small although a number of global differential
geometric issues are investigated following the tradition
pursued in the common textbooks of differential geometry.

1.3.2 Intrinsic versus Extrinsic Properties

Another classification of the properties of curves and surfaces,

which is based on their relation to the embedding external
space which they reside in, may be made where the properties
are divided into intrinsic and extrinsic. The first category
corresponds to those properties which are independent in their
existence and definition from the ambient space which
embraces the object such as the distance along a given curve or
the Gaussian curvature of a surface at a given point (see §
4.5↓), while the second category is related to those properties
which depend in their existence and definition on the external
embedding space such as having a normal vector at a point on
the curve or the surface.
More technically, the intrinsic properties (see Footnote 2 in §
8↓) are defined and expressed in terms of the metric tensor
which is formulated in differential geometry as the first
fundamental form (see § 3.5↓) while the extrinsic properties
are expressed in terms of the surface curvature tensor which is
formulated in differential geometry as the second fundamental
form (see § 3.6↓). As we will see in several places of this book,
some quantities can be expressed once in terms of the
coefficients of the first fundamental form exclusively and once
in terms of expressions involving the coefficients of the second
fundamental form as well. In this regard, a quantity is
classified as intrinsic if it can be expressed as a function of the
coefficients of the first fundamental form only even if it can
also be expressed in terms involving the coefficients of the
second fundamental form. We should also remark that extrinsic
properties are normally expressed in terms of both forms
although these properties are characterized by being expressed
necessarily in terms of the second form since this is the feature
that distinguishes them from intrinsic properties.
The idea of intrinsic and extrinsic properties may be illustrated
by an inhabitant of a surface with a 2D perception (hereafter
this creature will be called “2D inhabitant”) where he can
detect and measure intrinsic properties but not extrinsic
properties as the former do not require appealing to an external
embedding 3D space in which the surface is immersed while
the latter do. Hence, in simple terms all the properties that can
be detected and measured by a 2D inhabitant are intrinsic to
the surface while all the other properties are extrinsic. A 1D
inhabitant of a curve may also be used, to a lesser extent,
analogously to distinguish between intrinsic and extrinsic
properties of surface and space curves (refer for example to §
2.3↓).
The so-called “intrinsic geometry” of the surface comprises the
collection of all the intrinsic properties of the surface. When
two surfaces can have a coordinate system on each such that
the first fundamental forms of the two surfaces are identical at
each pair of corresponding points on the two surfaces then the
two surfaces have identical intrinsic geometry. Such surfaces
are isometric (see § 6.5↓) and can be mapped on each other by
a transformation that preserves the curve lengths, the angles
and the surface areas. Moreover, if these surfaces are subjected
to identical coordinate transformations, they remain identical
in their intrinsic properties.
1.4 General Mathematical Background
In this section, we provide a general mathematical background
outlining concepts and definitions needed in general to
understand the forthcoming materials of differential geometry.
However, it should be remarked that some of the materials
provided in the present section which are related to concepts
from other subjects of mathematics, such as calculus and
topology, are elementary because of the limits on the text size;
moreover the book is not prepared as a text about these
subjects. The purpose of these outlines and definitions is to
provide a basic understanding of the related ideas. The readers
are, therefore, advised to refer to textbooks on those subjects
for more technical and extensive revelations. This section also
contains some materials that may not be needed in the
forthcoming chapters but they are usually discussed and
investigated in differential geometry texts. Our objective of
including such materials is for the present book to be more
comprehensive and compatible with similar differential
geometry texts.

1.4.1 Geometry and Topology

Here, we define some geometric shapes which are used as

examples and prototypes in the forthcoming sections and
chapters and they may not be familiar to some readers.
Common geometric shapes (like straight line, circle and
sphere) are common knowledge at this level and hence they
will not be defined. We also introduce some basic topological
concepts which will be needed in the forthcoming parts of the
book.
A surface of revolution is an axially symmetric surface
generated by a plane curve C (see § 5.2↓) revolving around a
straight line L contained in the plane of the curve but not
intersecting the curve. The curve C is called the profile of the
surface and the line L is called the axis of revolution which is
also the axis of symmetry of the surface. Meridians of a
surface of revolution are plane curves on the surface formed by
the intersection of a plane containing the axis of revolution
with the surface, and hence the meridians are identical versions
of the profile curve C. Parallels of a surface of revolution are
circles generated by intersecting the surface by planes
perpendicular to the axis of revolution, and hence they
represent the paths of specific points on the profile curve C.
For spheres, these curves are called meridians of longitude and
parallels of latitude. On any surface of revolution, meridians
and parallels intersect at right angles.
A surface of revolution may be generated in its simplest form
by revolving a curve x = f(z) around the z-axis of a Cartesian
coordinate system. The surface can then be parameterized as:
(1) x = f cosφ
(2) y = f sinφ
(3) z = z
where (f, φ, z) represent the standard cylindrical coordinates
(ρ, φ, z).
The helix (Fig. 2↓) is a space curve characterized by having a
tangent vector that forms a constant angle with a specified
direction which is the direction defined by its axis of rotation.
The helix can be defined parametrically by:
(4) x = a cos(θ)
(5) y = a sin(θ)
(6) z = b θ
where a, b are non-zero real constants and θ is a real parameter
with − ∞ < θ < + ∞. The circle may be regarded as a
degenerate form of helix corresponding to b = 0, while a
straight line may be regarded as another degenerate form of
helix corresponding to a = 0.
Figure 2 Helix and its parameters.

The torus (Fig. 3↓) is a surface of revolution whose profile

curve C is a circle. It can be defined parametrically by:
(7) x = (R + r cosφ) cosθ
(8) y = (R + r cosφ) sinθ
(9) z = r sinφ
where R is the torus radius (i.e. the distance between the center
of the generating circle and the center of symmetry of the torus
which is the perpendicular projection of the circle center on the
axis of revolution), r is the radius of the generating circle (r <
∈ ∈
R), φ [0, 2π) is the angle of variation of r, and θ [0, 2π) is
the angle of variation of R.
Figure 3 Torus and its parameters.

The ellipsoid (Fig. 4↓) is a quadratic surface (see § 6.2↓) that

can be defined parametrically by:
(10) x = a sinθ cosφ
(11) y = b sinθ sinφ
(12) z = c cosθ
where a, b, c are non-zero real constants and θ, φ are real
parameters with 0 ≤ θ ≤ π and 0 ≤ φ < 2π.
Figure 4 Ellipsoid.

The hyperboloid of one sheet (Fig. 5↓) is a quadratic surface

that can be defined parametrically by:
(13) x = a coshξ cosθ
(14) y = b coshξ sinθ
(15) z = c sinhξ
where a, b, c are non-zero real constants and ξ, θ are real
parameters with − ∞ < ξ < + ∞ and 0 ≤ θ < 2π.
Figure 5 Hyperboloid of one sheet.

The hyperboloid of two sheets (Fig. 6↓) is a quadratic surface

that can be defined parametrically by:
(16) x = a sinhξ cosθ
(17) y = b sinhξ sinθ
(18) z = c coshξ
where a, b, c are non-zero real constants and ξ, θ are real
parameters with 0 ≤ ξ < ∞ and 0 ≤ θ < 2π.
Figure 6 Hyperboloid of two sheets.

The elliptic paraboloid (Fig. 7↓) is a quadratic surface that

can be defined parametrically by:
(19) x = a √(ξ) cosθ
(20) y = b √(ξ) sinθ
(21) z = c ξ
where a, b, c are non-zero real constants and ξ, θ are real
parameters with 0 ≤ ξ < ∞ and 0 ≤ θ < 2π.
Figure 7 Elliptic paraboloid.

The hyperbolic paraboloid (Fig. 8↓) is a quadratic surface

that can be defined parametrically by:
(22) x = a ξ
(23) y = b ω
(24) z = c ξ ω
where a, b, c are non-zero real constants and ξ, ω are real
parameters with − ∞ < ξ < + ∞ and − ∞ < ω < + ∞.
Figure 8 Hyperbolic paraboloid.

The parabolic cylinder (Fig. 9↓) is a quadratic surface

generated in its simplest form by translating a parabola
(expressed in y as a quadratic function of x, or the other way
around, in a canonical form) along the z-direction. Hence, it
can be parameterized by the following equations:
(25) x = ξ
(26) y = a ξ2
(27) z = b ω
where a, b are non-zero real constants and ξ, ω are real
parameters with − ∞ < ξ < + ∞ and − ∞ < ω < + ∞.
Figure 9 Parabolic cylinder.

The catenary is a plane curve (see § 5.2↓) with a hyperbolic

cosine shape. It can be defined parametrically by:
(28) x = a cosh(ξ ⁄ a)
(29) z = ξ
where a is a non-zero real constant and − ∞ < ξ < + ∞ is a real
parameter. The catenoid (Fig. 10↓) is a surface of revolution
generated by revolving a catenary around its directrix, which is
the z-axis in the above formulation, and hence it can be defined
parametrically by:
(30) x = a cosh(ξ ⁄ a) cosθ
(31) y = a cosh(ξ ⁄ a) sinθ
(32) z = ξ
where a is a non-zero real constant and ξ, θ are real parameters
with − ∞ < ξ < + ∞ and 0 ≤ θ < 2π.
Figure 10 Catenoid.

The helicoid (Fig. 11↓) is a ruled surface (see § 6.3↓) with the
property that for each point P on the surface there is a helix
passing through P and contained entirely in the surface. It can
be defined parametrically by:
(33) x = a ξ cosθ
(34) y = a ξ sinθ
(35) z = b θ
where a, b are non-zero real constants and ξ, θ are real
parameters with − ∞ ≤ ξ ≤ + ∞ and − ∞ ≤ θ < + ∞.
Figure 11 Helicoid.

The monkey saddle (Fig. 12↓) is a saddle surface that can be

defined parametrically by:
(36) x = ξ
(37) y = ω
(38) z = ξ3 − 3 ξ ω2
where ξ, ω are real parameters with − ∞ < ξ < + ∞ and − ∞
< ω < + ∞.
Figure 12 Monkey saddle.

The enneper (Fig. 13↓) is a self-intersecting surface that can

be defined parametrically by:
(39) x = − (ξ3 ⁄ 3) + ξ + ξ ω2
(40) y = − ξ2ω − ω + (ω3 ⁄ 3)
(41) z = ξ2 − ω2
where ξ, ω are real parameters with − ∞ < ξ < + ∞ and − ∞
< ω < + ∞.
Figure 13 Enneper.

The tractrix (Fig. 14↓) is a plane curve starting (see Footnote

3 in § 8↓) from the point (ρ, 0) on the x-axis with the property
that the length of the line segment of its tangent between the
tangency point and the point of intersection with the z-axis is
equal to ρ where ρ is a real constant (ρ > 0). Hence, the tractrix
is a solution of the following differential equation:
(42) dz ⁄ dx = ±√(ρ2 − x2) ⁄ x
where 0 < x ≤ ρ and with the condition z(ρ) = 0. The plus sign
in this equation corresponds to the lower part (z < 0) while the
minus sign corresponds to the upper part. The Beltrami
pseudo-sphere (Fig. 14↓) is a surface of revolution generated
by revolving a tractrix around its asymptote which is the z-axis
in the above formulation. The pseudo-sphere can be defined
parametrically by:
(43) x = a sinθ cosφ
(44) y = a sinθ sinφ
(45) z = a [cosθ + ln(tan(θ ⁄ 2))]
where a is a real constant and θ, φ are real parameters with 0
< θ < π and 0 ≤ φ < 2π. The constant ρ of the tractrix which
generates the pseudo-sphere is called the pseudo-radius of the
pseudo-sphere.
Figure 14 Tractrix (left frame) and pseudo-sphere (right
frame).

It should be remarked that the geometric shapes that we

defined parametrically in this subsection (e.g. torus, ellipsoid,
and hyperbolic paraboloid) can also be defined by other
parametric forms as well as by explicit Cartesian and non-
Cartesian forms (see e.g. § 6.2↓).
The Euler characteristic, or Euler-Poincare characteristic, is a
topological parameter of closed surfaces (see § 3.1↓) which,
for polyhedral surfaces, is given by:
(46) χ = V + ℱ − ℰ
where χ is the Euler characteristic of the surface, and V, ℱ, ℰ
are the numbers of vertices, faces and edges of the polyhedron.
Examples of Euler characteristic of some common polyhedrons
are given in Fig. 15↓.
Figure 15 Polyhedral surfaces: tetrahedron (left frame), cube
(middle frame) and octahedron (right frame). The Euler
characteristic for these surfaces is given respectively by: χ = 4
+ 4 − 6, χ = 8 + 6 − 12, and χ = 6 + 8 − 12, which in all cases is
equal to 2.

The Euler characteristic can also be defined for more general

types of surface. The Euler characteristic of a compact
orientable (see § 3.1↓) non-polyhedral surface, like sphere and
torus, can be obtained by polygonal decomposition based on
dividing the entire surface into a finite number of non-
overlapping curvilinear polygons which share at most edges or
vertices (see Fig. 16↓). The above formula (Eq. 46↑) is then
used, as for polyhedral surfaces, to determine the Euler
characteristic of the surface.
Figure 16 Polygonal decomposition of a sphere into four
non-overlapping curvilinear polygons.

In simple terms, the topological genus of a surface is the

number of handles or topological holes on the surface. For
example, the ellipsoid (Fig. 4↑) is a surface of genus 0 while
the torus (Fig. 3↑) is a surface of genus 1. Similarly, the
surfaces in Fig. 17↓ are of genus 2 and 3. For an orientable
surface of genus g the Euler characteristic χ is related to the
genus by:
(47) χ = 2(1 − g)
Figure 17 Examples of surfaces of genus 2 (left frame) and
genus 3 (right frame).

The length of a straight line segment connecting two points in

a Euclidean space is a measure of the distance (in its common
sense) between the two points (see Footnote 4 in § 8↓). The
length of a polygonal arc (Fig. 18↓ a) is the sum of the lengths
of its straight segments. The length of an arc of a generalized
twisted space curve is the limit of the length of an asymptotic
polygonal arc (Fig. 18↓ b) as the length of the longest straight
line segment of the asymptotic polygonal arc tends to zero.
Figure 18 (a) Polygonal arc and (b) twisted curve (solid) with
two of its asymptotic polygonal arcs (dashed and dotted) where
the dashed represents a better approximation to the length of
the twisted curve than the dotted.

The area of a polygonal plane fragment (e.g. triangle or square)

is a measure of its two-dimensional expansion (see Footnote 5
in § 8↓). The area of a surface patch consisting entirely of
polygonal plane fragments is the sum of the areas of its
polygonal plane fragments. The area of a patch of a
generalized twisted space surface is the limit of the area of its
asymptotic polygonal patch as the area of the largest of its
polygonal fragments tends to zero (refer to Fig. 19↓).
Figure 19 A smooth surface (bottom) approximated by a
coarse asymptotic polygonal surface (top) and a fine
asymptotic polygonal surface (middle) where the fine
represents a better approximation to the area of the smooth
surface than the coarse.

1.4.2 Functions

The domain of a functional mapping: f:ℝm → ℝn is the largest

set of ℝm on which the mapping is defined. A bicontinuous
function or mapping is a continuous function with a continuous
inverse. A scalar function is of class Cn if the function and all
of its first n (but not n + 1) partial derivatives do exist and are
continuous. A vector function (e.g. a position vector
representing a space curve or surface) is of class Cn if one of
its components is of this class while all the other components
are of this class or higher. A curve or a surface is of class Cn if
it is mathematically represented by a function of this class.
In this context we should remark that in most cases the purpose
of imposing the condition of having a function of class Cn is to
have a function which is at least of this class and hence the
condition is met by having a function of this class or higher. In
fact, this is generally the case in this book when this condition
is imposed unless there is an obvious indicator that the
function is strictly of this class. Another remark is that there
should be no confusion between the bare C symbol and the
superscripted C symbol, like C2, as the bare C is usually used
in this book to symbolize a curve while the superscripted C
stands for the above differentiability condition.
In gross terms, a “smooth” or “sufficiently smooth” curve or
surface means that the functional relation that represents the
object is sufficiently differentiable for the intended objective,
being of class Cn at least where n is the minimum requirement
for the differentiability index to satisfy the required conditions.
A deleted neighborhood of a point P on a 1D interval on the
∈
real line is defined as the set of all points x ℝ in the interval
such that:
(48) 0 < |x − xP| < ϵ
where xP is the coordinate of P on the real line and ϵ is a
positive real number. Hence, the deleted neighborhood
includes all the points in the open interval (xP − ϵ, xP + ϵ)
excluding xP itself. For a space curve (which is not straight in
general) represented by r = r(t), where r is the spatial
representation of the curve and t is a general parameter in the
curve representation, the definition applies to the neighborhood
of tP where tP is the value of t corresponding to the point P on
the curve. Deleted neighborhood of a twisted curve may also
be identified by being confined in a circle of radius ϵ centered
at P, however this applies only to plane curves. Therefore, it
may be identified by being confined in a sphere of radius ϵ
centered at P to be more general.
A deleted neighborhood of a point P on a 2D flat surface is
∈
defined as the set of all points (x, y) ℝ2 on the surface such
that:
(49) 0 < √([x − xP]2 + [y − yP]2) < ϵ
where (xP, yP) are the coordinates of P on the plane and ϵ is a
positive real number. Hence, the deleted neighborhood
includes all the points inside a circular disc of radius ϵ and
center (xP, yP) excluding the center itself. For a space surface
(which is not flat in general) represented by r = r(u, v), where r
is the spatial representation of the surface and u, v are the
surface coordinates on the uv plane (see § 1.4.3↓) that map on
the surface, the definition applies to the neighborhood of (uP,
vP) where (uP, vP) are the coordinates on the 2D uv plane
corresponding to the point P on the surface. Similar to twisted
curves, deleted neighborhood of a twisted surface may also be
identified by being confined in a sphere of radius ϵ centered at
P. The above definitions of deleted neighborhoods of curves
and surfaces can be easily extended to spaces of higher
dimensionality. For instance, in a 3D space the neighborhood is
contained in a spherical surface with radius ϵ. However, this is
not needed in this book whose focus is space curves and
surfaces.
A quadratic expression:
(50) Q(x, y) = a1x2 + 2a2xy + a3y2
of real coefficients a1, a2, a3 and real variables x, y is described
as “positive definite” if it possesses positive values ( > 0) for
all pairs (x, y) ≠ (0, 0). The sufficient and necessary condition
for Q to be positive definite is that:
(51)
a1 > 0
and
(a1a3 − a2a2) > 0
These two conditions necessitate a third condition that is: a3 >
0 since these coefficients are real.
The first variation of a functional F may be defined by the
Gateaux derivative of the functional as:
(52)
where x and h are variable functions and ξ is a real scalar
parameter. The Euler-Lagrange variational principle is a
mathematical rule whose objective is to minimize or maximize
a certain functional F(f) which depends on a function f. It is
represented mathematically by a partial differential equation
whose solutions optimize the particular functional F. The
Euler-Lagrange principle in its generic, simple and most
common form is given mathematically by:
(53)
where f(x, y, yx) is a function of the given variables that
optimizes the functional F, y is a function of x, and yx is the
derivative of y with respect to x.

1.4.3 Coordinates, Transformations and Mappings

The Jacobian J of a transformation between two coordinate

systems, labeled as unbarred and barred, is the determinant of
the Jacobian matrix J of the transformation between these
systems, that is:
(54)
where the indexed x and x̃ are the coordinates in the unbarred
and barred coordinate systems in an nD space. An admissible
coordinate transformation may be defined generically as a
mapping represented by a sufficiently differentiable set of
equations and it is invertible by having a non-vanishing
Jacobian (J ≠ 0) (see Footnote 6 in § 8↓). An invariant
property of a curve or a surface is a property which is
independent of admissible coordinate transformations and
parameterizations. It should be noted that an invariant property
may be invariant with respect to certain types of transformation
or parameterization but not with respect to others and hence it
is sometimes used generically where the context should be
taken into consideration for sensible interpretation.
An orthogonal coordinate transformation is a combination of
translation, rotation and reflection of axes. The Jacobian of
orthogonal transformations is unity, that is J = ±1. The
orthogonal transformation is described as positive iff J = + 1
and negative iff J = − 1. Positive orthogonal transformations
consist solely of translation and rotation (possibly trivial ones
as in the case of the identity transformation) while negative
orthogonal transformations include reflection, by applying an
odd number of axes reversal, as well. Positive transformations
can be decomposed into an infinite number of continuously
varying infinitesimal positive transformations each one of
which imitates an identity transformation. Such a
decomposition is not possible in the case of negative
orthogonal transformations because the shift from the identity
transformation to reflection is impossible by a continuous
process.
A surface S in a 3D space may be described directly by the
three spatial coordinates of a given 3D coordinate system (e.g.
x, y, z of a Cartesian system) or by the two surface coordinates
(e.g. u, v). In the former case, the description is rather familiar
where each point on the space surface is identified directly by
three spatial coordinates linked by a given functional relation
(see § 3.1↓ and 6.2↓). In the latter case, a 2D domain over
which the surface S is defined is introduced, where this domain
is identified by two independent variables (usually u, v or u1,
u2) . The domain is usually assumed, for simplicity, to be a
plane (which may be called the uv plane or parameters plane)
where a grid of u and v curves are defined over this plane.
Again, for simplicity the curves in these u and v families are
usually defined as straight lines; moreover, the curves in each
family are regularly spaced with the two families of u and v
curves being mutually orthogonal (see Fig. 20↓). Although
these assumptions about the domain and about the grid and
parameter curves can be violated (e.g. by using a 2D polar or
curvilinear coordinate system over the parameters plane
instead of the above-described 2D rectangular Cartesian), in
most cases there is no advantage in doing so. Along each curve
of the u and v families only one variable (either u or v) varies
while the other variable (v or u) is held constant. So, along any
curve of the u family v is constant and along any curve of the v
family u is constant.
Figure 20 The uv parameters plane corresponding to a space
surface and the mapping of the uv grid.

The surface S is then defined by a functional mapping from the

uv plane onto the space surface S. This functional mapping
identifies any point on the surface in the 3D space
corresponding to any point in its domain in the uv plane.
Hence, the uv grid on the uv plane is projected by this
functional mapping on a corresponding uv grid on the surface
S. As we are assuming that the surface is embedded in a 3D
space, the functional mapping consists of three independent
relations where each relation correlates the u and v coordinates
of a given point in the domain to one of the three spatial
coordinates (x1, x2, x3) of a corresponding point of the surface
in the embedding space, that is: x1(u, v), x2(u, v) and x3(u, v).
While the uv grid on the uv plane is usually a planar,
orthogonal and regularly spaced grid, the corresponding grid
on the surface S is not necessarily so because it is generally a
3D grid that follows the bends and variations of the twisted
space surface. In this context, “coordinate curves” (which are
also called parametric curves or parametric lines) on a surface
are defined as the map or projection of the uv curves of the uv
grid of the parameters plane onto the space surface, as
described above, and hence they are curves along which only
one coordinate variable (u or v) varies while the other
coordinate variable (v or u) remains constant. So, along the u
coordinate curves v is held constant while along the v
coordinate curves u is held constant (see Fig. 20↑).
It is noteworthy that although the surface coordinates u, v (or
u1, u2) primarily represent the coordinates on the uv parameters
plane which maps on the surface, in the literature of
differential geometry they are sometimes used (to ease the
notation) as labels for the curves on the surface. Hence,
vigilance is required to avoid confusion.
A regular representation of class Cm (m > 0) of a surface patch
S in a 3D Euclidean space is defined as a functional mapping
of an open set Ω in the uv plane onto S that satisfies the
following two conditions:
1. The functional mapping relation is of class Cm over the
entire Ω.
2. The Jacobian matrix of the transformation between the
representation of the surface in the 3D space and its 2D domain
is of rank 2 for all the points in Ω.
We remark that for a functional mapping of the form
S(u, v) = (S1(u, v), S2(u, v), S3(u, v)),
the aforementioned Jacobian matrix is given by:
(55)
In fact, having a Jacobian matrix of rank 2 for the
transformation, according to the above condition, is equivalent
to the condition that E1 × E2 ≠ 0 where E1 = ∂ur and E2 = ∂vr
are the surface basis vectors (see § 1.4.5↓, 3.1↓ and 3.2↓),
which are the tangents to the u and v coordinate curves
respectively, and r = r(u, v) is the 3D spatial representation of
the curves. Having a Jacobian matrix of rank 2 is also
equivalent to having a well-defined tangent plane to the surface
at the related point. In this regard, we note that “rank” here
refers to its meaning in linear algebra and should not be
confused with the rank of a tensor which is related to the
number of its free indices although there is a connection
between the two.
As we will see, a regular point on a surface is a point that
satisfies the above condition about the basis vectors. A point
on a surface which is not regular is called singular. Singularity
occurs either because of a geometric reason, which is the case
for instance for the apex of a cone, or because of the particular
parametric representation of the surface. While the first type of
singularity is inherent and hence it cannot be removed, the
second type can be removed by changing the representation.
Corresponding points on two curves refer to two points, one on
each curve, with a common value of a common parameter of
the two curves. When the two curves have two different
parameterizations then a one-to-one correspondence between
the two parameters should be established and the
corresponding points then refer to two points with
corresponding values of the two parameters. Corresponding
points on two surfaces can be defined in a similar manner
taking into account that surfaces require two independent
parameters in their identification.
In many cases of theoretical and practical significance, a mixed
tensor Aiα, which is contravariant with respect to
transformations in space coordinates xi and covariant with
respect to transformations in surface coordinates uα, may be
defined. Following a coordinate transformation in which both
the space and surface coordinates change, the tensor Aiα will be
given in the new (barred) system by:
(56) Ãiα = Ajβ(∂x̃i ⁄ ∂xj)(∂uβ ⁄ ∂ũα)
More generally, tensors with space and surface contravariant
indices and space and surface covariant indices (e.g. Aiαjβ) can
also be defined in a similar manner. The extension of the above
transformation rule to include such tensors can be easily
achieved by following the obvious pattern seen in the last
equation.

1.4.4 Intrinsic Distance

The intrinsic distance between two points on a surface is the

greatest lower bound (or infimum) of the lengths of all regular
arcs connecting the two points on the surface. The intrinsic
distance is an intrinsic property of the surface. The intrinsic
distance d between two points is invariant under a local
isometric mapping (see § 3.1↓), that is:
(57) d(f(P1), f(P2)) = d(P1, P2)
where f is an isometric mapping from a surface S1 to a surface
S2 (see § 3.1↓ and 6.5↓), P1 and P2 are the two points on S1 and
f(P1) and f(P2) are their images on S2. In fact, this may be
taken as the definition of isometric mapping, i.e. it is the
mapping that preserves intrinsic distance.
The following conditions apply to the intrinsic distance d
between points P1, P2 and P3:
1. Symmetry: d(P1, P2) = d(P2, P1).
2. Triangle inequality: d(P1, P3) ≤ d(P1, P2) + d(P2, P3).
3. Positive definiteness: d(P1, P2) > 0 with d(P1, P2) = 0 iff P1
and P2 are the same point.
An arc C connecting two points, P1 and P2, on a surface is
described as an arc of minimum length between P1 and P2 if
the length of C is equal to the intrinsic distance between P1 and
P2. The existence and uniqueness of an arc of minimum length
between two specific points on a surface is not guaranteed, i.e.
it may not exit and if it does exist it may not be unique (refer to
§ 5.7↓ for examples). Yes, for certain types of surface such an
arc does exist and it is unique. For example, on a simply
connected (see § 3.1↓) plane surface there exists an arc of
minimum length between any two points on the plane and it is
unique; this arc is the straight line segment connecting the two
points.

1.4.5 Basis Vectors

The set of basis vectors in a given manifold plays a pivotal role

in the theoretical construction of the geometry of the manifold,
and this applies to the basis vectors in differential geometry
where these vectors are used in the definition and construction
of essential concepts and objects such as the metric tensor of
the surface. The set of basis vectors may also be employed to
serve as a moving coordinate frame for the enveloping space of
their underlying constructions (see § 2.2↓ and § 3.2↓).
The differential geometry of curves and surfaces employs two
main sets of basis vectors:
1. One set is constructed on space curves and consists of three
unit vectors: the tangent T, the normal N and the binormal B to
the curve.
2. Another set is constructed on surfaces and consists of two
linearly independent vectors which are the tangents to the
coordinate curves of the surface, E1 = (∂ r ⁄ ∂u1) and E2 = (∂
r ⁄ ∂u2), plus the unit normal vector to the surface n, where
r(u1, u2) is the spatial representation of a surface coordinate
curve, and u1 and u2 are the surface coordinates as discussed
earlier and as will be investigated further later in the book
(refer to § 1.4.3↑ and 3.2↓).
Each one of the above basis sets is defined on each regular
point of the curve or the surface and hence in general the
vectors in each one of these basis sets vary from one point to
another, i.e. they are position dependent. The vectors E1 and
E2, which are not necessarily of unit length, vary in magnitude
and direction while the unit vectors (which are the rest) vary in
direction. Also, while the vectors of the T, N, B set are
mutually orthogonal, the vectors of the E1, E2, n set is not
necessarily so since E1 and E2 are not orthogonal in general
although n is orthogonal to both E1 and E2.
The surface basis vectors, E1 and E2, are given in full tensor
notation by (∂xi ⁄ ∂uα) (i = 1, 2, 3 and α = 1, 2) which is usually
abbreviated as xiα. These vectors can be seen as contravariant
space vectors or as covariant surface vectors. Further details
about this will be given in § 3.3↓. We remark that other sets of
basis vectors are also defined and employed in differential
geometry, as we will see in the future (refer to § 4.1↓ and
4.4↓).

1.4.6 Flat and Curved Spaces

A manifold, such as a 2D surface or a 3D space, is called “flat”

if it is possible to find a coordinate system for the manifold
with a diagonal metric tensor whose all diagonal elements are
±1; the space is called “curved” otherwise. More formally, an
nD space is described as a flat space iff it is possible to find a
coordinate system for which the line element ds is given by:
(58)
(ds)2 =
ζ1(dx1)2 + ζ2(dx2)2 + … + ζn(dxn)2 =
Σni = 1 ζi(dxi)2
where the indexed ζ are ±1 while the indexed x are the
coordinates of the space. For the space to be flat (i.e. globally
not just locally), the condition given by Eq. 58↑ should apply
all over the space and not just at certain points or regions.
An example of flat space is the 3D Euclidean space which can
be coordinated by a rectangular Cartesian system whose metric
tensor is diagonal with all the diagonal elements being + 1.
Another example is the 4D Minkowski space-time manifold
whose metric is diagonal with elements of ±1. All 1D spaces
are Euclidean and hence they cannot be curved intrinsically, so
twisted curves are curved only when viewed externally from
the embedding space which they reside in, e.g. the 2D space of
a surface curve or the 3D space of a space curve. An example
of curved space is the 2D surface of a sphere or an ellipsoid.
A necessary and sufficient condition for an nD space to be
intrinsically flat is that the Riemann-Christoffel curvature
tensor (see § 1.4.10↓) of the space vanishes identically. Hence,
cylinders are intrinsically flat, since their Riemann-Christoffel
curvature tensor vanishes identically, although they are curved
as seen extrinsically from the embedding 3D space. On the
other hand, planes are intrinsically and extrinsically flat. In
brief, a space is intrinsically flat iff the Riemann-Christoffel
curvature tensor vanishes identically over the space, and it is
extrinsically (as well as intrinsically) flat iff the curvature
tensor (see § 3.4↓) vanishes identically over the space. This is
because the Riemann-Christoffel curvature tensor characterizes
the space curvature from an intrinsic perspective while the
curvature tensor characterizes the space curvature from an
extrinsic perspective.
Due to the strong connection between the Gaussian curvature
(see § 4.5↓) and the Riemann-Christoffel curvature tensor
which implies that each one of these will vanish if the other
does (see for example Eq. 92↓), we see that having an
identically vanishing Gaussian curvature is another sufficient
and necessary condition for a 2D space to be flat. It should be
remarked that the Gaussian curvature in differential geometry
is defined for 2D spaces although the concept may be extended
to higher dimensionality manifolds in the form of Riemannian
curvature. This issue will be discussed further later in the book
(see § 4.4↓).
A curved space may have constant non-vanishing curvature all
over the space, or have variable curvature and hence the
curvature is position dependent. An example of a space of
constant curvature is the surface of a sphere of radius R whose
curvature (i.e. Gaussian curvature) is (1 ⁄ R2) at each point of
the surface. Torus (Fig. 3↑), ellipsoid (Fig. 4↑) and paraboloids
(Figs. 7↑ and 8↑) are simple examples of surfaces with variable
curvature. As we will see, there are various characterizations
and quantifications for the curvature; hence in the present
context “curvature” may be a generic term. For 2D surfaces,
curvature usually refers to the Gaussian curvature (see § 4.5↓)
which is strongly linked to the Riemannian curvature, as
indicated above.
Schur theorem related to nD spaces (n > 2) of constant
curvature states that: if the Riemann-Christoffel curvature
tensor at each point of a space is a function of the coordinates
only, then the curvature is constant all over the space. Schur
theorem may also be stated as: the Riemannian curvature is
constant over an isotropic region of an nD (n > 2) Riemannian
space. The focus of the book, however, is limited to spaces of
lower dimensionality.
The geometry of curved spaces is usually described as the
Riemannian geometry. One approach for investigating the
Riemannian geometry of a curved manifold is to embed the
manifold in a Euclidean space of higher dimensionality and
inspect the properties of the manifold from this perspective.
This approach is largely followed in the present book where
the geometry of curved 2D spaces (twisted surfaces) is
investigated by immersing the surfaces in a 3D Euclidean
space and examining their properties as viewed from this
external enveloping 3D space. Such an external view is
necessary for examining the extrinsic geometry of the surface
but not its intrinsic geometry. A similar approach is also
followed in the investigation of surface and space curves.
A surface with positive/negative Gaussian curvature (see §
4.5↓) at each point is described as a surface of
positive/negative curvature. Ellipsoids (Fig. 4↑), hyperboloids
of two sheets (Fig. 6↑) and elliptic paraboloids (Fig. 7↑) are
examples of surfaces of positive curvature while hyperboloids
of one sheet (Fig. 5↑) and hyperbolic paraboloids (Fig. 8↑) are
examples of surfaces of negative curvature. A surface with
positive/negative Gaussian curvature at each point may also be
described as a surface of constant curvature since its sign is
constant all over the surface although its magnitude may be
variable. In general, a twisted surface possesses coordinate-
dependent curvature which varies in magnitude and sign and
may take all possible signs (i.e. < 0, 0 and > 0) over different
points or regions.
The geometric description and quantification of flat spaces are
simpler than those of curved spaces, and hence in general the
differential geometry of flat spaces is more motivating and less
challenging than that of curved spaces. However, as there is a
subjective element in this type of statements, it may not apply
to everyone.

1.4.7 Homogeneous Coordinate Systems

When all the diagonal elements of a diagonal metric tensor of a
flat space are + 1, the coordinate system is described as
homogeneous. In this case the line element ds of Eq. 58↑
becomes:
(59) (ds)2 = dxidxi
An example of homogeneous coordinate systems is the
orthonormal Cartesian system (x, y, z) of a 3D Euclidean space
(Fig. 21↓). A homogeneous coordinate system can be
transformed to another homogeneous coordinate system only
by linear transformations. Any coordinate system obtained
from a homogeneous coordinate system by an orthogonal
transformation is also homogeneous. As a consequence of the
last statements, infinitely many homogeneous coordinate
systems can be constructed in any flat space.
Figure 21 Orthonormal Cartesian coordinate system and its
basis vectors e1, e2 and e3 in a 3D space.

A coordinate system of a flat space can always be

homogenized by allowing the coordinates to be imaginary.
This is done by redefining the coordinates as:
(60) Xi = √(ζi)xi
where the new coordinates Xi are imaginary when ζi = − 1.
Consequently, the line element will be given by:
(61) (ds)2 = dXidXi
which is of the same form as Eq. 59↑. An example of a
homogeneous coordinate system with some real and some
imaginary coordinates is the coordinate system of a Minkowski
4D space-time related to the mechanics of Lorentz
transformations.

1.4.8 Geodesic Coordinates

It is always possible to introduce coordinates at particular

points in a multi-dimensional manifold so that the Christoffel
symbols (see § 1.4.9↓) vanish at these points. These
coordinates are called geodesic coordinates. Geodesic
coordinates are employed as local coordinate systems mainly
for the purpose of achieving certain advantages, as will be
outlined next. These geodesic systems for these particular
points are also described as locally Cartesian coordinates. The
allocated points at which the Christoffel symbols are made to
vanish in these coordinates are described as the poles. Based
on the above and what we will see later in the book, geodesic
coordinates of space surfaces are normally taken as the
equivalent of the Cartesian coordinates of a flat space. Hence,
non-geodesic coordinates on 2D spaces may be compared to
general curvilinear coordinates in general nD spaces.
The main reason for the use of geodesic coordinates is that the
covariant and absolute derivatives (see § 7↓) in such systems
become respectively partial and total derivatives at the poles
since the Christoffel symbol terms in the covariant and
absolute derivative expressions vanish at these points. Any
tensor property can then be easily proved in the geodesic
system at the pole and consequently generalized to other
systems due to the invariance of the zero tensor under
permissible coordinate transformations. If the allocated pole is
a general point in the space, the property is then established
over the whole space.
In any Riemannian space it is always possible to find a
coordinate system for which the coordinates are geodesic at
every point of a given analytic curve. Moreover, there is an
infinite number of ways by which geodesic coordinates can be
defined over a coordinate patch (see § 3.1↓). We should remark
that some authors define geodesic coordinates on a coordinate
patch of a surface as a coordinate system whose u and v
coordinate curve families are orthogonal, with one of these
families (u or v) being a family of geodesic curves (refer to §
5.7↓). Hence, “geodesic coordinates” may appear to have
multiple usage although they are essentially the same. More
details about this should be looked for in more advanced
textbooks on differential geometry.

1.4.9 Christoffel Symbols for Curves and Surfaces

The Christoffel symbols of the first kind for a general nD space

(n > 1) are defined by:
(62) [ij, l] = (∂jgil + ∂igjl − ∂lgij) ⁄ 2
where the indexed g is the covariant form of the metric tensor
of the given space. We remark that the Latin indices used in the
previous equation, as well as the next four equations, do not
imply a 3D space and hence the equations and the metric are
not specific to such a space as they apply to any nD space (n ≥
2).
The Christoffel symbols of the second kind, which may also be
called affine connections, are obtained by raising the third
index of the Christoffel symbols of the first kind, that is:
(63)
Γkij =
gkl[ij, l] =
(gkl ⁄ 2)(∂jgil + ∂igjl − ∂lgij)
where the indexed g is the metric tensor of the given space in
its contravariant and covariant forms with implied summation
over l. Similarly, the Christoffel symbols of the first kind can
be obtained from the Christoffel symbols of the second kind by
reversing the above process through lowering the upper index,
that is:
(64)
gkmΓkij =
gkmgkl[ij, l] =
δlm[ij, l] = [ij, m]
where δlm is the mixed form of the Kronecker delta tensor for
the space. It is noteworthy that the Christoffel symbols of the
first and second kind are symmetric in their paired indices, that
is:
(65) [ij, k] = [ji, k]
(66) Γkij = Γkji
For 1D spaces, the Christoffel symbols are not defined. The
Christoffel symbols of the first kind for a 2D surface are given
by:
(67) [11, 1] = (∂ua11) ⁄ 2 = Eu ⁄ 2
(68) [11, 2] = ∂ua12 − (∂va11) ⁄ 2 = Fu − (Ev ⁄ 2)
(69) [12, 1] = (∂va11) ⁄ 2 = Ev ⁄ 2 = [21, 1]
(70) [12, 2] = (∂ua22) ⁄ 2 = Gu ⁄ 2 = [21, 2]
(71) [22, 1] = ∂va12 − (∂ua22) ⁄ 2 = Fv − (Gu ⁄ 2)
(72) [22, 2] = (∂va22) ⁄ 2 = Gv ⁄ 2
where the indexed a are the elements of the surface covariant
metric tensor (refer to 3.3↓) and E, F, G are the coefficients of
the first fundamental form (refer to 3.5↓). The subscripts u and
v which suffix the coefficients stand for partial derivatives of
these coefficients with respect to these variables (i.e. ∂ ⁄ ∂u and
∂ ⁄ ∂v). As indicated before and will be seen in the future, E =
a11, F = a12 = a21 and G = a22, and hence the equalities in the
previous equations are justified. The above relations can be
obtained from the definition of the Christoffel symbols of the
first kind (Eq. 62↑) using the coefficients of the metric tensor
and first fundamental form of the surface.
In orthogonal coordinate systems, F = a12 = a21 = 0 identically
and hence the above formulae will be simplified accordingly
by dropping any term involving the derivatives of these
coefficients. We note that “orthogonal coordinate system” here
and in similar contexts means a system whose coordinate
curves are orthogonal everywhere. Therefore, the above
condition is fully justified since E1 and E2 basis vectors, which
are the tangents to the u and v coordinate curves, will be
orthogonal in such a system and hence the dot product, seen in
Eq. 235↓, will vanish accordingly.
The Christoffel symbols of the second kind for a 2D surface
are given by:
(73) Γ111 = (a22∂ua11 − 2a12∂ua12 + a12∂va11) ⁄ (2a) = (GEu −
2FFu + FEv) ⁄ (2a)
(74) Γ211 = (2a11∂ua12 − a11∂va11 − a12∂ua11) ⁄ (2a) = (2EFu −
EEv − FEu) ⁄ (2a)
(75) Γ112 = (a22∂va11 − a12∂ua22) ⁄ (2a) = (GEv − FGu) ⁄ (2a) =
Γ121
(76) Γ212 = (a11∂ua22 − a12∂va11) ⁄ (2a) = (EGu − FEv) ⁄ (2a) =
Γ221
(77) Γ122 = (2a22∂va12 − a22∂ua22 − a12∂va22) ⁄ (2a) = (2GFv −
GGu − FGv) ⁄ (2a)
(78) Γ222 = (a11∂va22 − 2a12∂va12 + a12∂ua22) ⁄ (2a) = (EGv −
2FFv + FGu) ⁄ (2a)
where a ( = a11a22 − a12a21 = EG − F2) is the determinant of
the surface covariant metric tensor, and the other symbols are
as explained above. These relations can be obtained from the
definition of the Christoffel symbols of the second kind (Eq.
63↑) using the coefficients of the metric tensor and first
fundamental form of the surface. The above formulae will also
be simplified in orthogonal coordinate systems, where F = a12
= a21 = 0 identically, by dropping the vanishing terms that
involve these coefficients or their derivatives.
The Christoffel symbols of the second kind for a 2D surface
may also be given by:
(79) Γ111 = − [( E2 × ∂1 E1)⋅ n] ⁄ √(a)
(80) Γ211 = + [( E1 × ∂1 E1)⋅ n] ⁄ √(a)
(81) Γ112 = − [( E2 × ∂2 E1)⋅ n] ⁄ √(a) = Γ121
(82) Γ212 = + [( E1 × ∂2 E1)⋅ n] ⁄ √(a) = Γ221
(83) Γ122 = − [( E2 × ∂2 E2)⋅ n] ⁄ √(a)
(84) Γ222 = + [( E1 × ∂2 E2)⋅ n] ⁄ √(a)
where the indexed E are the surface covariant basis vectors
(see § 1.4.5↑ and 3.2↓), n is the unit normal vector to the
surface and a is the determinant of the surface covariant metric
tensor as defined above. It is worth noting that: √(a) = (E1 ×
E2)⋅ n, as we will see later in the book.
Since the Christoffel symbols of both kinds are dependent on
the metric only, as can be seen from the above equations (Eqs.
67↑-72↑ and Eqs. 73↑-78↑) as well as from the definitions of
these symbols (see Eqs. 62↑ and 63↑), they represent intrinsic
properties of the surface geometry and hence they are part of
its intrinsic geometry. The involvement of extrinsic parameters
in the definitions given by Eqs. 79↑-84↑ does not affect their
intrinsic qualification, as explained earlier.
The Christoffel symbols of the first kind are linked to the
surface covariant basis vectors by the following relation:
(85) [αβ, γ] = (∂Eα ⁄ ∂uβ)⋅Eγ
where α, β, γ = 1, 2. This relation may also be written as:
(86) [αβ, γ] = rαβ⋅rγ
where α, β, γ = 1, 2 and the subscripts represent partial
derivatives with respect to the variables represented by these
coordinate indices. The last equation may provide an easier
form to remember these formulae.
On applying the index raising operator to Eq. 85↑, we obtain a
similar expression for the Christoffel symbols of the second
kind, that is:
(87) Γγαβ = (∂Eα ⁄ ∂uβ)⋅Eγ
where α, β, γ = 1, 2 and Eγ is the contravariant form of the
surface basis vectors (see § 3.2↓).

1.4.10 Riemann-Christoffel Curvature Tensor

The Riemann-Christoffel curvature tensor is an absolute rank-4
tensor that characterizes important properties of spaces,
including 2D surfaces, and hence it plays an important role in
differential geometry. The tensor is used, for instance, to test
for the space flatness (see § 1.4.6↑). There are two kinds of
Riemann-Christoffel curvature tensor: first and second. The
Riemann-Christoffel curvature tensor of the first kind is a type
(0, 4) tensor while the Riemann-Christoffel curvature tensor of
the second kind is a type (1, 3) tensor. Shifting from one kind
to the other is achieved by using the index shifting operator.
The first and second kinds of the Riemann-Christoffel
curvature tensor are given respectively by:
(88) Rijkl = ∂k[jl, i] − ∂l[jk, i] + [il, r]Γrjk − [ik, r]Γrjl
(89) Rijkl = ∂kΓijl − ∂lΓijk + ΓrjlΓirk − ΓrjkΓirl
where the indexed square brackets and Γ are the Christoffel
symbols of the first and second kind of the given space, as
defined in § 1.4.9↑.
We remark that in the above two equations, as well as in the
following equations in the present and the next subsection, the
Latin indices do not necessarily range over 1, 2, 3 as these
equations are valid for general nD manifolds (n ≥ 2) including
surfaces (n = 2) and spaces of higher dimensionality (n > 3).
Another remark is that the Riemann-Christoffel curvature
tensor of the first kind is anti-symmetric in its first two indices
and in its last two indices and block symmetric with respect to
these two sets of indices, that is:
(90)
Rijkl = − Rjikl
Rijkl = − Rijlk
Rijkl = + Rklij
From Eqs. 88↑ and 89↑, it can be seen that the Riemann-
Christoffel curvature tensor depends exclusively on the
Christoffel symbols of the first and second kind which are both
dependent on the metric tensor (or the first fundamental form,
see § 3.5↓) and its partial derivatives only. Hence, the
Riemann-Christoffel curvature, as represented by the Riemann-
Christoffel curvature tensor, is an intrinsic property of the
manifold. Since the Riemann-Christoffel curvature tensor
depends on the metric which, in general curvilinear
coordinates, is a function of position, the Riemann-Christoffel
curvature tensor follows this dependency on position.
The Riemann-Christoffel curvature tensor vanishes identically
iff the space is globally flat from intrinsic view; otherwise the
space is curved (see § 1.4.6↑). Hence, the Riemann-Christoffel
curvature tensor vanishes identically over 1D manifolds as
represented by surface and space curves. As we will see in §
2.3↓, surface and space curves are intrinsically Euclidean.
Similarly, the Riemann-Christoffel curvature tensor vanishes
identically over plane surfaces. More generally, a surface is
isometric to the Euclidean plane iff the Riemann-Christoffel
curvature tensor is zero at each point on the surface since it is
intrinsically flat. The last statement also applies if the Gaussian
curvature of the surface vanishes identically due to the link
between the Riemann-Christoffel curvature tensor and the
Gaussian curvature (refer to § 4.5↓ and see Eq. 92↓).
The 2D Riemann-Christoffel curvature tensor has only one
degree of freedom and hence it possesses a single independent
non-vanishing component which is represented by R1212.
Hence, for a 2D Riemannian space we have:
(91) R1212 = R2121 = − R1221 = − R2112
while all the other components of the tensor are identically
zero. The signs in the equalities of Eq. 91↑ are justified by the
aforementioned anti-symmetric relations of the Riemann-
Christoffel tensor in its indices (see Eq. 90↑). The vanishing of
the other components (e.g. R1111 and R1121) can also be
explained by the anti-symmetric relations since these
components contain two identical anti-symmetric indices.
Based on the above facts, the Riemann-Christoffel curvature
tensor can be expressed in tensor notation by:
(92)
Rαβγδ =
R1212 ϵαβ ϵγδ =
(R1212 ⁄ a) εαβ εγδ =
K εαβ εγδ
where the indexed ϵ are the relative permutation tensors, the
indexed ε are the absolute permutation tensors, and K is the
Gaussian curvature (see § 4.5↓) whose expression is obtained
from Eq. 356↓. The non-vanishing component of the 2D
Riemann-Christoffel curvature tensor, R1212, may be given in
expanded form by:
(93) R1212 = (1 ⁄ 2)(2∂12a12 − ∂22a11 − ∂11a22) + aαβ(Γα12Γβ12 −
Γα11Γβ22)
where ∂αβ ≡ ∂2 ⁄ (∂uα∂uβ) with α, β = 1, 2, the indexed a are the
coefficients of the surface covariant metric tensor, and the
Christoffel symbols are based on the surface metric.

1.4.11 Ricci Curvature Tensor and Scalar

The Ricci curvature tensor of the first kind, which is an

absolute rank-2 covariant symmetric tensor, is obtained by
contracting the contravariant index with the last covariant
index of the Riemann-Christoffel curvature tensor of the
second kind, that is:
(94)
Rij =
Raija =
∂jΓaia − ∂aΓaij + ΓabjΓbia − ΓabaΓbij
The Ricci tensor of the second kind is obtained by raising the
first index of the Ricci tensor of the first kind using the index
raising operator.
For 2D spaces, the Ricci curvature tensor of the first kind is
related to the Riemann-Christoffel curvature tensor by the
following relations:
(95)
(R11 ⁄ a11) =
(R12 ⁄ a12) =
(R21 ⁄ a21) =
(R22 ⁄ a22) = − (R1212 ⁄ a)
where the indexed a are the elements of the 2D covariant
metric tensor (see § 3.3↓) and the bare a is its determinant. As
the 2D Riemann-Christoffel curvature tensor has only one
independent non-vanishing component, the last equation
provides a full link between the Ricci curvature tensor and the
Riemann-Christoffel curvature tensor. Since K = (R1212 ⁄ a)
(see Eq. 356↓), the above relations also link the Ricci tensor to
the Gaussian curvature.
The Ricci scalar ℛ, which is also called the curvature scalar
and the curvature invariant, is the result of contracting the
indices of the Ricci curvature tensor of the second kind, that is:
(96) ℛ = Rii
Hence, it is a rank-0 tensor. As seen from the above relations,
the Ricci curvature tensor and curvature scalar are part of the
intrinsic geometry of the manifold.
1.5 Exercises
Exercise 1. Give a brief definition of differential geometry
indicating the other disciplines of mathematics to which
differential geometry is intimately linked.
Exercise 2. A surface embedded in a 3D space can be regarded
as a 2D and as a 3D object at the same time. Discuss this
briefly. From the same perspective, discuss also the state of a
curve embedded in a surface which in its turn is embedded in a
3D space.
Exercise 3. What are the following symbols:
[αβ, γ], [ij, k], Γγαβ and Γkij?
What is the difference between those with Greek indices and
those with Latin indices?
Exercise 4. What is the relation between the coefficients of the
surface covariant metric tensor and the surface covariant
curvature tensor on one hand and the coefficients of the first
and second fundamental forms on the other? What are the
symbols representing all these coefficients?
Exercise 5. What is the difference between the local and global
properties of a manifold? Give an example for each. What are
the colloquial terms used to label these two categories?
Exercise 6. What is the meaning of “intrinsic” and “extrinsic”
properties of a manifold? Give an example for each.
Exercise 7. Explain the concept of “2D inhabitant” and how it
is used to classify the properties of a space surface.
Exercise 8. Find the equation of a plane passing through the
points:
(1, 2, 0), (0, − 3, 1.5) and (1, 0, − 1).
What is the normal unit vector to this plane?
Exercise 9. Is the normal unit vector of a plane surface unique?
Exercise 10. Define briefly each one of the following terms:
surface of revolution, meridians and parallels.
Exercise 11. Prove that the meridians and parallels of a surface
of revolution are mutually perpendicular at their points of
intersection.
Exercise 12. State the parametric equations of the following
geometric shapes: torus, hyperboloid of one sheet, and
hyperbolic paraboloid.
Exercise 13. Write down the parametric equations of a circle in
the uv plane centered at point (a, b) with radius r = c where a,
b, c are real constants and c > 0.
Exercise 14. Find the parametric equations of an ellipse in the
xy plane centered at the origin of coordinates with A = 5 and B
= 3 where A, B are the semi-major and semi-minor axes.
Exercise 15. A surface of revolution may be represented
locally in a 3D space by the following form:
r(u, v) = (u cosv, u sinv, f(u, v))
where f is a continuous function. Determine the equations
representing the parallels and meridians of this surface.
Exercise 16. Find the parametric equations of a curve formed
by the intersection of the surfaces represented by:
r1(u, v) = (u, u2, v) and
r2(u, v) = (u, v, u2)
where − ∞ < u, v < + ∞.
Exercise 17. Write down the general form of the parametric
equations of each of the following surfaces: hyperboloid of two
sheets, parabolic cylinder, catenoid, monkey saddle and
pseudo-sphere.
Exercise 18. Sketch the following (using a 3D computer
graphic package if available):
(a) a straight line passing through the point (11, − 5, 6.3) and
parallel to the vector ( − 3, − 1.8, 6.5)
(b) a plane passing through the point (6, − 8.2, − 7) with a
normal vector (3, − 1.6, − 2.5).
Exercise 19. A surface is parameterized by:
r(u, v) = (a sinhu cosv, b sinhu sinv, c coshu).
What is the name of this surface? What is the condition for this
surface to be a surface of revolution around the third spatial
axis?
Exercise 20. Find the equation of the straight line passing
through the point ( − 6, 3.1, 8.4) and the point (1, 0, − 3).
Exercise 21. Classify the following as curves or surfaces:
ellipsoid, elliptic paraboloid, catenary, helicoid, enneper and
tractrix.
Exercise 22. Make a simple sketch for each one of the
geometric shapes in the previous question. Use a computer
graphic package if convenient.
Exercise 23. Prove that
v × (dv ⁄ dt) = 0
iff the direction of the vector v(t) is constant.
Exercise 24. Define “Euler characteristic” stating the equation
that links it to the number of vertices, faces and edges of a
polyhedron.
Exercise 25. Explain how the Euler characteristic is defined
for non-polyhedral compact orientable surfaces such as
ellipsoids.
Exercise 26. What is the topological meaning of “genus of a
surface”?
Exercise 27. Give examples for surfaces of genus 0, 1, 2 and 3
from common geometric shapes other than those given in the
text.
Exercise 28. By using the Euler formula, calculate the Euler
characteristic χ of the following surfaces:
(a) parallelepiped
(b) dodecahedron
(c) icosahedron.
Show your work in detail.
Exercise 29. By using polygonal decomposition, calculate the
Euler characteristic χ of the following surfaces:
(a) sphere
(b) ellipsoid
(c) torus.
Show your work in detail with simple sketches to demonstrate
the polygonal decomposition in each case.
Exercise 30. What is the genus of the surfaces in the previous
question?
Exercise 31. What is the Cartesian form of the equation of a
sphere centered at point (a, b, c) with radius r = d where a, b,
c, d are real constants and d > 0?
Exercise 32. Explain briefly the meaning of the following
terms: bicontinuous function, surface of class Cn, and
sufficiently smooth curve.
Exercise 33. Explain in detail, using equations and simple
sketches, the concept of “deleted neighborhood” in 1D and 2D
flat and curved spaces as seen from the ambient space.
Exercise 34. How can we extend the concept of “deleted
neighborhood” to spaces of dimensionality higher than 2?
Exercise 35. Find the equation of a cone generated by rotating
the line z = − 2x around the z-axis.
Exercise 36. Derive the parametric equations of a helix
rotating around the z-axis, passing through the point (3, 0, 0)
and climbing (or descending) 5.3 units in the z-direction as it
makes a 4π turn around the z-axis.
Exercise 37. Define “positive definite” in words and by stating
the mathematical conditions for a quadratic expression to be
positive definite.
Exercise 38. Describe orthogonal coordinate transformations
and how they are characterized by their Jacobian.
Exercise 39. State the difference between positive and
negative orthogonal transformations.
Exercise 40. Find a set of parametric equations representing a
cylinder generated by rotating a straight line parallel to the z-
axis and passing through the point (2.5, 0, 0) around the z-axis.
Exercise 41. What is the unit normal vector to the surface of a
sphere, centered on the origin of coordinates with radius r, at a
point on its surface with coordinates (xP, yP, zP)? Consider the
possibility of having more than one normal vector at that point.
Exercise 42. What “coordinate curves” means? What are the
other names given to these curves?
Exercise 43. Define regular representation of a class Cn
surface patch in a 3D Euclidean space stating its mathematical
conditions.
Exercise 44. Using a parametric representation of the elliptic
paraboloid, show that it is a regular surface.
Exercise 45. What are the reasons for having a singular point
on a space surface?
Exercise 46. A sphere centered at the origin of coordinates can
be represented parametrically by:
r(θ, φ) = a(sinθ cosφ, sinθ sinφ, cosθ)
where a > 0 is a constant, 0 ≤ θ ≤ π and 0 ≤ φ < 2π.
At what points, if any, this representation is not regular?
Exercise 47. How a mathematical correspondence can be
established between points on two different curves and two
different surfaces?
Exercise 48. State the mathematical conditions which are
satisfied by the intrinsic distance between two points on a
smooth connected surface.
Exercise 49. Is it guaranteed that an arc of minimum length
between two specific points on a surface does exist and it is
unique?
Exercise 50. Prove the three properties of intrinsic distance,
i.e. symmetry, triangle inequality, and positive definiteness.
Exercise 51. Show that the intrinsic distance d between two
points is invariant under a local isometric mapping f, i.e.
d(f(P1), f(P2)) = d(P1, P2).
Exercise 52. Find the intrinsic distance on the unit sphere
centered at the origin of coordinates between the point (1, 0, 0)
and the point (1 ⁄ √(3), 1 ⁄ √(3), 1 ⁄ √(3)).
Exercise 53. What is the intrinsic distance between the two
points of the last question in a 3D Euclidean space that
encloses the sphere?
Exercise 54. Define the basis vectors and state their roles.
Exercise 55. Describe in detail the two main sets of space
basis vectors in differential geometry related to curves and
surfaces. Are there any other sets of basis vectors?
Exercise 56. Are the basis vectors necessarily of unit length
and/or mutually orthogonal? If not, give examples of basis
vectors which are not of unit length and/or mutually
orthogonal.
Exercise 57. Define, in mathematical terms, flat and curved
spaces giving examples for each.
Exercise 58. State a sufficient and necessary condition for an
nD space to be flat.
Exercise 59. Is it necessary that an nD curved space possesses
universally constant curvature? If not, give an example of a
space with variable curvature in sign and magnitude.
Exercise 60. What is the locus of the points (if any) which are
shared between the xy plane and the following surfaces:
(a) a sphere centered at (0, 0, 5) with radius r = 6
(b) a sphere centered at (1, 1, 1) with radius r = 1.5
(c) a plane passing through the point (5, − 9.6, 0) with a unit
normal vector (0, 0, − 1)?
Exercise 61. Describe the commonly used approach for
investigating the Riemannian geometry of curved manifolds.
Exercise 62. Give a brief definition of homogeneous
coordinate systems giving a common example of such systems.
Exercise 63. What is the relation between the Christoffel
symbols of the first kind and the Christoffel symbols of the
second kind?
Exercise 64. Write the mathematical expressions for the
symbols [12, 1] and Γ122 of a surface in terms of the
coefficients of the surface metric tensor.
Exercise 65. Using the definition of the Christoffel symbols of
the first kind and the rules of tensors, derive Eqs. 68↑ and 71↑.
Exercise 66. Using the definition of the Christoffel symbols of
the second kind and the rules of tensors, derive Eqs. 76↑ and
78↑.
Exercise 67. State the mathematical relations correlating the
Christoffel symbols of the first and second kind to the surface
basis vectors and their derivatives.
Exercise 68. What is the relation between the Riemann-
Christoffel curvature tensor and the Gaussian curvature of a
surface?
Exercise 69. How many independent non-vanishing
components the 2D Riemann-Christoffel curvature tensor
possesses?
Exercise 70. What is the significance of having an identically
vanishing Riemann-Christoffel curvature tensor on a 2D
surface?
Exercise 71. Show that the Riemann-Christoffel curvature
tensor is anti-symmetric in its first two indices and in its last
two indices and block symmetric with respect to these two sets
of indices.
Exercise 72. Prove Eqs. 85↑ and 87↑.
Exercise 73. What is the rank of the Ricci curvature tensor?
Exercise 74. State the mathematical relation that links the
Ricci curvature tensor of the first kind to the Christoffel
symbols of the second kind and their partial derivatives.
Exercise 75. What is the relation between the elements of the
Ricci curvature tensor of the first kind and the Gaussian
curvature?
Exercise 76. How do you obtain the Ricci scalar from the
Riemann-Christoffel curvature tensor of the first kind? Explain
your answer step by step.
Chapter 2
Curves in Space
In this chapter, we investigate curves residing in a higher
dimensionality space and how they are characterized. We
should first remark that “space” in this title is general and
hence it includes surface since it is a 2D space, as explained in
§ 1.2↑.
2.1 General Background about Curves
In simple terms, a space curve is a set of connected points (see
Footnote 7 in § 8↓) in the embedding space such that any
totally connected subset of it can be twisted into a straight line
segment without affecting the neighborhood of any point.
More technically, a curve is defined as a differentiable
parameterized mapping between an interval of the real line and
a connected subset of the embedding space, that is C(t):I→ℝn
where C represents a space curve defined on the interval I ℝ⊆
∈
and parameterized by the variable t I. Hence, different
parameterizations of the same “geometric curve” will lead to
different “mapping curves”. The image of the mapping in the
embedding space is known as the trace of the curve; hence
different mapping curves can share the same trace. The curve
may also be defined as a topological image of a real interval
and may be linked to the concept of Jordan arc. We note that
Jordan arcs or Jordan curves may be defined as injective
mappings with no self intersection.
Space curves can be defined symbolically in different ways;
the most common of these is parametrically where the three
space coordinates of the curve points are given as functions of
∈
a real valued parameter, e.g. xi = xi(t) where t ℝ is the curve
parameter and i = 1, 2, 3. The parameter t may represent time
or arc length or even an arbitrarily defined quantity. Similarly,
surface curves are defined parametrically where the two
surface coordinates are given as functions of a real valued
parameter, e.g. uα = uα(t) with α = 1, 2. The surface coordinates
can then be mapped, through another mapping relation, onto
the spatial representation of the surface in the enveloping
space. Parameterized curves are oriented objects as they can be
traversed in one direction or the other depending on the sense
of increase of their parameter.
There are two main types of curve parameterization:
parameterization by arc length and parameterization by
something else such as time. The condition for a space curve
∈ ⊆
C(t):I→ℝ3, where t I is the curve parameter and I ℝ is an
interval over which the curve is defined, to be parameterized
by arc length is that: for all t we have |dr ⁄ dt| = 1 where r(t) is
the position vector representing the curve in the ambient space.
As a consequence of this, parameterization by arc length is
equivalent to traversing the curve with a constant unity speed.
Hence, using a parameterization by something other than arc
length may be considered as traversing the curve with varying
or non-unity speed. The advantage of parameterization by arc
length is that it confines the attention on the geometry of the
curve rather than other factors, which are usually irrelevant to
the geometric investigation, such as the temporal rate of
traversing the curve. Furthermore, it usually results in a more
simple mathematical formulation, as we will see later in the
book.
The parameter symbol which is used normally for
parameterization by arc length is s, while t is used to represent
a general parameter which could be arc length or something
else. This notation is followed in the present book. For curves
parameterized by arc length, the length L of a segment between
two points on the curve corresponding to s1 and s2 is given by
the simple formula:
(97)
Parameterization by arc length s may be called natural
parameterization of the curve and hence s is called natural
parameter.
Natural parameterization is not unique; however any other
natural parameter š is related to a given natural parameter s by
the following relation:
(98) š = ±s + c
where c is a real constant and hence the above-stated condition
|dr ⁄ dt| = 1 remains valid (see Footnote 8 in § 8↓). This may
be stated in a different way by saying that natural
parameterization with arc length s is unique apart from the
possibility of having a different sense of orientation and an
additive constant to s.
Natural parameterization may also be used for
parameterization by a parameter which is proportional to s and
hence the transformation relation between two natural
parameters becomes:
(99) š = ±ms + c
where m is another real constant. The two parameterizations
then differ, apart from the sense of orientation and the constant
shift, by the length scale which can be chosen arbitrarily.
Consequently, natural parameterization will be equivalent to
traversing the curve with a constant speed not necessarily of
unity magnitude. This may be based on the vision of extending
the aforementioned benefits of natural parameterization by
scaling, i.e. by choosing a different length scale a natural
parameterization will be obtained spontaneously. In this book,
natural parameterization is restricted to parameterization with
arc length s.
In a general nD space, the tangent vector to a space curve,
represented parametrically by the spatial representation r(t)
where t is a general parameter, is given by (dr ⁄ dt). A vector
tangent to a space curve at P is a non-trivial scalar multiple of
(dr ⁄ dt) and hence it can differ in magnitude and direction
from (dr ⁄ dt) as it can be parallel or anti-parallel to (dr ⁄ dt).
The tangent vector to a surface curve, represented
parametrically by: C(u(t), v(t)) where u and v are the surface
coordinates and t is a general parameter, is given by:
(100) dr ⁄ dt = (∂r ⁄ ∂u)(du ⁄ dt) + (∂r ⁄ ∂v)(dv ⁄ dt)
where r(u(t), v(t)) is the spatial representation of C and where
all these quantities are defined and evaluated at a particular
point on the curve. The last equation can be cast compactly,
using tensor notation, as:
(101)
dxi ⁄ dt =
(∂xi ⁄ ∂uα)(duα ⁄ dt) =
xiα(duα ⁄ dt)
where i = 1, 2, 3, α = 1, 2 and (u1, u2) ≡ (u, v).
⊆ ∈
A space curve C(t):I → ℝ3, where I R and t I is a general
parameter, is “regular at point t0” iff ṙ(t0) exists and ṙ(t0) ≠ 0
where r(t) is the spatial representation of C and the overdot
stands for differentiation with respect to the general parameter
t. The curve is “regular” iff it is regular at each interior point in
I. On a regular parameterized curve there is a neighborhood to
each point in its domain in which the curve is injective. On
transforming a surface S by a differentiable regular mapping f
of class Cn to a surface S̃ , a regular curve C of class Cn on S
will be mapped on a regular curve C̃ of class Cn on S̃ by the
same functional mapping relation, that is r(̃ t) = f(r(t)) where
the barred and unbarred r(t) are the spatial parametric
representations of the two curves on the barred and unbarred
surfaces.
The tangent line to a sufficiently smooth curve at one of its
regular points P is a straight line passing through P but not
through any point in a deleted neighborhood of P. More
technically, the tangent line to a curve C at a given regular
point P is a straight line passing through P and having the
same orientation as the tangent vector (dr ⁄ dt) of C at P. The
tangent line to a curve at a given point P on the curve may also
be defined geometrically as the limit of a secant line passing
through P and another neighboring point on the curve as the
other point converges, while staying on the curve, to the
tangent point. These different definitions are equivalent as they
represent the same entity. It is noteworthy that the tangent line
of a smooth curve, where such a tangent does exit, is unique.
We also note that the geometric definition may be useful in
some cases where the analytical definition does not apply.
A non-trivial vector v is said to be tangent to a regular surface
S (see § 1.4.3↑) at a given point P on S if there is a regular
curve C on S passing through P such that
v = dr ⁄ dt
where r(t) is the spatial representation of C and (dr ⁄ dt) is
evaluated at P (also see 3.1↓ for further details). In fact, any
vector
v = c(dr ⁄ dt)
where c ≠ 0 is a real number, is a tangent although it may not
be the tangent.
A periodic curve C is a curve that can be represented
parametrically by a continuous function of the form r(t + T) =
r(t) where r is the spatial representation of C, t is a real general
parameter and T is a real constant called the function period.
Circles and ellipses (Fig. 22↓) are prominent examples of
periodic curves where they can be represented parametrically
by:
(102) r(t) = (a cost, a sint) (circle)
(103) r(t) = (a cost, b sint) (ellipse)
∈
where a and b are real positive constants and t ℝ. Due to the
periodicity of the trigonometric functions, these equations
satisfy the condition: r(t + 2π) = r(t). Hence, circles and
ellipses are periodic curves with a period of 2π.
Figure 22 Circle (left frame) and ellipse (right frame) and
their main parameters.

A closed curve is a continuous periodic curve defined over a

minimum of one period. We note that periodicity is not a
necessary requirement for the definition of closed curves as the
curves can be defined over a single period without being
considered as such or by functions of non-periodic nature.
Closed curves may be regarded as topological images of
circles.
A curve is described as a plane curve if it can be embedded
entirely in a plane with no distortion. Orthogonal trajectories of
a given family of curves is a family of curves that intersect the
given family perpendicularly at their intersection points. Any
curve can be mapped isometrically to a straight line segment
where both are naturally parameterized by arc length. From the
last statement plus the fact that isometric transformation is an
equivalence relation (see § 6.5↓), it can be concluded that any
two space and surface curves can be connected by an isometric
relation.
2.2 Mathematical Description of
Curves
Let have a space curve of class C2 in a 3D Riemannian
manifold with a given metric gij (i, j = 1, 2, 3). The curve is
parameterized naturally by s representing arc length. As stated
earlier, we choose to parameterize the curve by s to have
simpler formulae although, for the sake of completeness and
generality, other formulae based on a more general
parameterization will also be given. The curve can therefore be
represented by:
(104) xi = xi(s)
where i = 1, 2, 3 and the indexed x represent the space
coordinates. This is equivalent to:
(105) r(s) = xi(s)Ei
where r is the spatial representation of the space curve and Ei
are the space basis vectors.
Three mutually perpendicular vectors each of unit length can
be defined at each regular point of the above-described space
curve: tangent T, normal N and binormal B (see Fig. 23↓). As
well as characterizing the curve, these vectors can serve as a
moving coordinate system for the embedding space as
indicated earlier. For simplicity, clarity and potential lack of
familiarity with tensor differentiation (see § 7↓) at this stage,
the following is mostly based on assuming a Euclidean space
coordinated by a rectangular Cartesian system although
supplementary remarks related to more general space and
coordinates are added when necessary. We also use a mix of
tensor and symbolic notations as each has certain advantages
and to familiarize the reader with both notations since different
authors use different notations.
Figure 23 The vectors T, N, B and their associated planes
(osculating plane OP, normal plane NP, and rectifying plane
RP) of a curve C embedded in a Euclidean space with a
rectangular Cartesian coordinate system (see also Fig. 24↓).

The unit vector tangent to the curve at a given regular point P

on the curve is given by (see Footnote 9 in § 8↓):
(106) [T]i = Ti = dxi ⁄ ds
For a t-parameterized curve, where t is not necessarily the arc
length, the tangent vector is given by:
(107) T = r(t) ⁄ |r(t)|
where the overdot represents differentiation with respect to t.
The unit vector normal to the tangent Ti, and hence to the
curve, at the point P is given by:
(108)
[N]i =
Ni =
(dTi ⁄ ds) ⁄ |dTi ⁄ ds| =
(dTi ⁄ ds) ⁄ κ
where κ is a scalar called the “curvature” of the curve at the
point P and is defined, according to the normalization
condition, by:
(109) κ = √([dTi ⁄ ds][dTi ⁄ ds])
For a t-parameterized curve, the normal unit vector is given by:
(110) N = [r(t) × (r(t) × r(t))] ⁄ [|r(t)||r(t) × r(t)|]
The vector N is also called the principal normal vector. Based
on Eq. 108↑, this vector is defined only on points of the curve
where the curvature κ ≠ 0. Also, since T is a unit vector then its
derivative is orthogonal to it, so the above-stated facts are
consistent. We note that Eq. 109↑ is based on an underlying
Cartesian coordinate system. For general curvilinear
coordinates, the formula becomes:
(111) κ = √(gij[δTi ⁄ δs][δTj ⁄ δs])
where gij is the space covariant metric tensor and the notation
of absolute derivative is in use (see § 7↓).
The binormal unit vector is defined as:
(112) [B]i = Bi = [κTi + (dNi ⁄ ds)] ⁄ τ
which is a linear combination of two vectors both of which are
perpendicular to Ni and hence it is perpendicular to Ni. In the
last equation, the normalization scalar factor τ is the “torsion”
whose sign is chosen to make T, N, B a right handed system
satisfying the condition:
(113) εijk Ti Nj Bk = 1
where εijk is the covariant absolute permutation tensor for the
3D space. We should also impose the condition τ ≠ 0 on Eq.
112↑. For plane curves, where the torsion vanishes identically
(see § 2.3.2↓), and at the points with τ = 0 of twisted curves,
the binormal unit vector B may be defined geometrically or as
the cross product of T and N, i.e. B = T × N.
For a t-parameterized curve, the binormal vector is given by:
(114) B = [r(t) × r(t)] ⁄ |r(t) × r(t)|
Inline with making T, N, B a right handed system, there is a
geometric significance for the sign of the torsion as it affects
the orientation of the space curve. It should be remarked that
some authors reverse the sign in the definition of τ and this
reversal affects the signs in the forthcoming Frenet-Serret
formulae (see § 2.5↓). The convention that we follow in this
book may have certain advantages.
At any point on the space curve, the triad T, N, B represent a
mutually perpendicular right handed system fulfilling the
condition:
(115) Bi = [T × N]i = εijkTjNk
where εijk is the contravariant absolute permutation tensor for
the 3D space. Since the vectors in the triad T, N, B are
mutually perpendicular, they satisfy the conditions:
(116) T⋅N = T⋅B = N⋅B = 0
Moreover, because they are unit vectors they also satisfy the
conditions:
(117) T⋅T = N⋅N = B⋅B = 1
It should be remarked that the triad T, N, B form what is called
the Frenet frame which represents a set of orthonormal basis
vectors for the embedding space. This frame serves as a basis
for a moving orthogonal coordinate system on the points of the
curve. The Frenet frame varies in general as it moves along the
curve and hence it is a function of the position on the curve.
The triad T, N, B may also be called the Frenet trihedron or the
moving trihedron of the curve. The Frenet frame can suffer
from problems or become undefined, e.g. at non-regular points
where T is undefined or at inflection points where dT ⁄ ds = 0.
The tangent line of a curve C at a given point P on the curve is
a straight line passing through P and is parallel to the tangent
vector, T, of C at P. The principal normal line of a curve C at a
given point P on the curve is a straight line passing through P
and is parallel to the principal normal vector, N, of C at P. The
binormal line of a curve C at a given point P on the curve is a
straight line passing through P and is parallel to the binormal
vector, B, of C at P. As a consequence, the equations of the
three lines can be given by the following generic form:
(118) r = rP + kVP
where r is the position vector of an arbitrary point on the line,
rP is the position vector of the point P, k is a real variable ( −
∞ < k < ∞) and the vector VP is the vector corresponding to the
particular line, that is VP ≡ T for the tangent line, VP ≡ N for
the principal normal line, and VP ≡ B for the binormal line.
At any point P on a space curve where the Frenet frame is
defined, the triad T, N, B define three mutually perpendicular
planes where each one of these planes passes through the point
P and is formed by a linear combination of two of these
vectors in turn. These planes are: the “osculating plane” which
is the span of T and N, the “rectifying plane” which is the span
of T and B, and the “normal plane” which is the span of N and
B and is orthogonal to the curve at P (Fig. 24↓). As a result, the
equations of the three planes can be given by the following
generic form:
(119) (r − rP)⋅VP = 0
where r is the position vector of an arbitrary point on the plane,
rP is the position vector of the point P, and where for each
plane the vector VP is the perpendicular vector to the plane at
P, that is VP ≡ B for the osculating plane, VP ≡ N for the
rectifying plane, and VP ≡ T for the normal plane. Following
the style of the definition of the tangent line of a curve as the
limit of the secant line (see § 2.1↑), the osculating plane may
also be defined as the limiting position of a plane passing
through P and two other points on the curve as the two points
converge simultaneously along the curve to P.
Figure 24 Frenet planes (osculating plane OP, normal plane
NP, and rectifying plane RP) and basis vectors (T, N, B) at a
point on a space curve C.

It is noteworthy that the positive sense of a parameterized

curve, which corresponds to the direction in which the
parameter increases and hence defines the orientation of the
curve, can be determined in two opposite ways. While the
sense of the tangent T and the binormal B is dependent on the
curve orientation and hence they are in opposite directions in
these two ways, the principal normal N is the same as it
remains parallel to the normal plane in the direction in which
the curve is turning. This is consistent with the fact that the
triad T, N, B form a right handed system.
2.3 Curvature and Torsion of Space
Curves
The curvature and torsion of space curves may also be called
the first and second curvatures respectively, and hence a
twisted curve with non-vanishing curvature and non-vanishing
torsion is described as double-curvature curve. The expression
√([dsT]2 + [dsB]2),
where dsT and dsB are respectively the lengths of the line
element components in the tangent and binormal directions,
may be described as the total or the third curvature of the
curve. The equation of Lancret states that:
(120) (dsN)2 = (dsT)2 + (dsB)2
where dsN is the length of the line element component in the
principal normal direction. We note that the term “total
curvature” is also used for surfaces (see § 4.4↓ and 4.8↓) but
the meaning is obviously different.
According to the fundamental theorem of space curves in
differential geometry, a space curve is completely determined
by its curvature and torsion. More technically, given a real
interval I⊆ ℝ and two differentiable real functions: κ(s) > 0
∈
and τ(s) where s I, there is a uniquely defined parameterized
regular space curve C(s): I → ℝ3 of class C2 with κ(s) and τ(s)
being the curvature and torsion of C respectively and s is its
arc length. Hence, any other curve meeting these conditions
will be different from C only by a rigid motion transformation
(i.e. translation and rotation) which determines its position and
orientation in space. On the other hand, any curve with the
above-described properties possesses uniquely defined κ(s) and
τ(s). As a consequence of the last statements, the fundamental
theorem of space curves provides the existence and uniqueness
conditions for curves. We note in this context that in rigid
motion transformation, which may also be called Euclidean
motion, the distance between any two points on the image is
the same as the distance between the corresponding points on
the inverse image. Hence, rigid motion transformation is a
form of isometric mapping.
The equations: κ = κ(s) and τ = τ(s), where s is the arc length,
are called the intrinsic or natural equations of the curve. The
curvature and torsion are invariants of the space curve and
hence they do not depend in magnitude on the employed
coordinate system or the type of parameterization. While the
curvature is always non-negative (κ ≥ 0), as it represents the
magnitude of a vector according to the above-stated definition
(see Eqs. 108↑, 109↑ and 111↑), the torsion can be negative as
well as zero or positive. It is worth mentioning that some
authors define the curvature vector (see § 4.1↓) and the
principal normal vector of space curves in such a way that it is
possible for the curvature to be negative.
The following are some examples of the curvature and torsion
of a number of commonly-occurring simple curves:
1. Straight line: κ = 0 and τ = 0.
2. Circle of radius R: κ = (1 ⁄ R) and τ = 0. Hence, the radius of
curvature (see § 2.3.1↓) of a circle is its own radius
3. Helix parameterized by r(t) = (acos(t), asin(t), bt): κ = a ⁄
(a2 + b2) and τ = b ⁄ (a2 + b2). It is worth noting that a space
curve of class C3 with non-vanishing curvature is a helix iff the
ratio of its torsion to curvature is constant.
In the above three examples, the curvature and torsion are
constant along the whole curve. However, in general the
curvature and torsion of space curves are position dependent
and hence they vary from point to point.
Following the example of 2D surfaces, a 1D inhabitant of a
space curve can detect all the properties related to the arc
length. Hence, the curvature and torsion, κ and τ, of the curve
are extrinsic properties for such a 1D inhabitant. This fact may
be expressed by saying that curves are intrinsically Euclidean,
and hence their Riemann-Christoffel curvature tensor vanishes
identically and they naturally admit 1D Cartesian systems
represented by their natural parameterization of arc length.
This should be obvious when considering that any curve can be
mapped isometrically to a straight line where both are naturally
parameterized by arc length. Another demonstration of their
intrinsic 1D nature is represented by the forthcoming Frenet-
Serret formulae (see § 2.5↓).
It is noteworthy that some authors resemble the role of κ and τ
in curve theory to the surface curvature tensor bαβ in surface
theory (see § 3.4↓) and describe κ and τ as the curve theoretic
analogues of bαβ in surface theory. In another context, κ and τ
may be compared (non-respectively!) with the first and second
fundamental forms of surfaces in their roles in defining the
curve and surface in the fundamental theorems of these
structures (compare the above with what is coming in § 3.6↓).
The curvature and torsion also play in the Frenet-Serret
formulae for space curves a similar role to the role played by
the coefficients of the first and second fundamental forms in
the Gauss-Weingarten equations for space surfaces (see §
3.9↓). Another useful remark in this context is that from the
first and the last of the Frenet-Serret formulae (see Eqs. 136↓
and 138↓), we have:
(121) |κτ| = |T’⋅B’|
where the prime stands for derivative with respect to a natural
parameter s of the curve.

2.3.1 Curvature

The curvature κ of a space curve is a measure of how much the

curve bends as it progresses in the tangent direction at a
particular point. The curvature represents the magnitude of the
rate of change of the direction of the tangent vector with
respect to the arc length and hence it is a measure for the
departure of the curve from the orientation of the straight line
passing through that point and oriented in the tangent direction.
Consequently, the curvature vanishes identically for straight
lines (see § 5.1↓). In fact, having an identically vanishing
curvature is a necessary and sufficient condition for a curve of
class C2 to be a straight line. From the first of the Frenet-Serret
formulae (Eq. 136↓) and the fact that:
(122) (N⋅T)’ = (0)’ = 0 ⇒ N⋅T’ = − N’⋅T
which is based on the orthogonality of N and T and the product
rule of differentiation, the curvature κ can be expressed as:
(123) κ = N⋅T’ = − N’⋅T
where the prime represents differentiation with respect to the
arc length s of the curve. The minus sign in the second equality
is consistent with the fact that κ is non-negative since the
component of N’ in the tangential direction is anti-parallel to
T. As for the first equality, N and T’ are parallel (see Eqs.
108↑ and 110↑) and hence the dot product is non-negative as it
should be.
The “radius of curvature”, which is the radius of the osculating
circle (see § 2.6↓), is defined at each point of a space curve at
which κ ≠ 0 as the reciprocal of the curvature, that is:
(124) Rκ = 1 ⁄ κ
A different way for introducing these concepts, which is
followed by some authors, is to define first the radius of
curvature as the reciprocal of the magnitude of the acceleration
vector, that is Rκ = (1 ⁄ | r’’(s)|) where r(s) is the spatial
representation of an s-parameterized curve; the curvature is
then defined as the reciprocal of the radius of curvature.
Hence, the radius of curvature may be described as the
reciprocal of the norm of the acceleration vector where
acceleration means the second derivative of the spatial
representation of the curve with respect to its natural
parameter.
There may be an advantage in using the concept of “curvature”
as the principal concept instead of “radius of curvature”, that is
the curvature can be defined at all points of a smooth curve
where a tangent vector is defined, including those with
vanishing curvature, while the radius of curvature is defined
only at those points with non-vanishing curvature.
As indicated above, if C is a space curve of class C2 which is
⊆
defined on a real interval I ℝ and is parameterized by arc
∈
length s I, that is C(s):I→ℝ3, then the curvature of C at a
given point P on the curve is defined by:
(125) κ = |r’’(s)|
where r(s) is the spatial representation of the curve, the double
prime represents the second derivative with respect to s, and r’’
is evaluated at P.
For a space curve represented parametrically by r(t), where t is
a general parameter, we have:
(126)
κ=
|T| ⁄ |r| =
|r × r| ⁄ |r|3 =
√([r⋅r][r⋅r] − [r⋅r]2) ⁄ ( r⋅r)3 ⁄ 2
where all the quantities, which are functions of t, are evaluated
at a given point corresponding to a given value of t, and the
overdot represents derivative with respect to t. It is noteworthy
that all surface curves passing through a point P on a surface S
and have the same osculating plane at P have identical
curvature κ at P if the osculating plane is not tangent to S at P.

2.3.2 Torsion

The torsion τ represents the rate of change of the osculating

plane, and hence it quantifies the twisting, in magnitude and
sense, of the space curve out of the plane of curvature and its
deviation from being a plane curve (see § 5.2↓). The torsion
therefore vanishes identically for plane curves. In fact, having
an identically vanishing torsion is a necessary and sufficient
condition for a curve of class C2 to be a plane curve. If C is a
space curve of class C2 which is defined on a real interval I ⊆
∈
ℝ and it is parameterized by arc length s I, that is
C(s):I→ℝ3, then the torsion of C at a given point P on the
curve is given by:
(127) τ = N’ · B
where N’ and B are evaluated at P and the prime represents
differentiation with respect to s. This equation can be obtained
from the second of the Frenet-Serret formulae (Eq. 137↓) by
dot producting both sides with B. The formula may also be
given as:
(128) τ = − N · B’
for the same reason as that given for the alternative formulae of
κ (see Eq. 123↑ and related text) or by dot producting both
sides of the third of the Frenet-Serret formulae (Eq. 138↓) with
N.
For a space curve represented parametrically by r(t), where t is
a general parameter, we have:
(129)
τ=
[r⋅(r × r)] ⁄ |r × r|2 =
[r⋅(r × r)] ⁄ [(r × r)⋅(r × r)] =
[r⋅(r × r)] ⁄ [(r⋅r)(r⋅r) − (r⋅r)2]
where all the quantities, which are functions of t, are evaluated
at a given point P corresponding to a given value of t, and the
overdot represents derivative with respect to t. The curve
should have non-vanishing curvature κ at P.
For rectangular Cartesian coordinates, the torsion of an s-
parameterized curve is given in tensor notation by:
(130) τ = (ϵijkxi’xj’’xk’’’) ⁄ κ2
where κ is the curvature of the curve as defined previously. The
last formula is based on its predecessor. For general curvilinear
coordinates, the torsion of an s-parameterized curve is given in
tensor notation by:
(131) τ = εijkTiNj(δNk ⁄ δs)
The magnitude of torsion is independent of the nature of the
curve parameterization and orientation as determined by the
sense of increase of its parameter. It is also invariant under
permissible coordinate transformations. Finally, the “radius of
torsion” is defined at each point of a space curve for which τ ≠
0 as the absolute value of the reciprocal of the torsion, that is:
(132) Rτ = |1 ⁄ τ|
We note that some authors do not take the absolute value and
accordingly the radius of torsion can be negative.
2.4 Geodesic Torsion
Geodesic torsion, which is also known as the relative torsion,
is an attribute of a curve embedded in a surface. The geodesic
torsion of a surface curve C at a given point P is the torsion of
the geodesic curve (see § 5.7↓) that passes through P in the
tangent direction of C at P. As we will see in § 5.7↓, in the
neighborhood of a given point P on a smooth surface and for
any specified direction there is one and only one geodesic
curve passing through P in that direction.
The geodesic torsion τg of a surface curve represented spatially
by r(s) is given by the following scalar triple product:
(133) τg = n⋅(n’ × r’)
where n is the unit normal vector to the surface, the primes
represent differentiation with respect to the natural parameter s,
and all these quantities are evaluated at a given point on the
curve corresponding to a given value of s.
The geodesic torsion of a curve C at a non-umbilical point P
(see § 4.10↓) is given in terms of the principal curvatures κ1
and κ2 (see § 4.4↓) by:
(134) τg = (κ1 − κ2)sinθcosθ
where θ is the angle between the tangent vector T to the curve
C at P and the first principal direction d1 (see Darboux frame
in § 4.4↓).
The geodesic torsion of a surface curve C parameterized by arc
length s at a given point P is also given in terms of the curve
torsion by:
(135) τg = τ − (dφ ⁄ ds)
where τ is the torsion of C at P, and φ is the angle between the
unit normal vector n to the surface and the principal normal
vector N of C at P, i.e. φ = arccos(n⋅N). Also, the curve C
should not be asymptotic (see § 5.9↓). This formula (Eq. 135↑)
which is known as the Bonnet formula, demonstrates that when
n and N are collinear along the curve, the geodesic torsion and
the torsion are equal (i.e. τg = τ). As we will see in § 5.7↓,
when n and N are collinear, the geodesic component of the
curvature vector (see § 4.1↓) will vanish. In this case, the
geodesic curvature (see § 4.3↓) will vanish and the curve
becomes a geodesic. So in brief, on a geodesic curve we have:
τg = τ which is consistent with the above statement at the
beginning of this section.
The geodesic torsion of a surface curve C at a given point P is
zero iff C is tangent to a line of curvature at P (see § 5.8↓).
Hence, on a line of curvature the geodesic torsion vanishes
identically. This can be seen from Eq. 134↑ where either sinθ
or cosθ vanishes. The geodesic torsions of two orthogonal
surface curves at their point of intersection are equal in
magnitude and opposite in sign.
2.5 Relationship between Curve Basis
Vectors and their Derivatives
The three basis vectors T, N, B of a space curve are connected
to their derivatives by the Frenet-Serret formulae which are
given in rectangular Cartesian coordinates by:
(136) dTi ⁄ ds = κNi
(137) dNi ⁄ ds = τBi − κTi
(138) dBi ⁄ ds = − τNi
As indicated previously (see § 2.2↑), the sign of the terms
involving τ depends on the convention about the torsion and
hence these equations differ between different authors. The
above equations are also known as Frenet formulae.
The Frenet-Serret formulae can be cast in the following matrix
form using symbolic notation:
(139)
where all the quantities in this equation are functions of arc
length s and the prime represents derivative with respect to s.
As seen, the coefficient matrix of this system is anti-
symmetric.
The Frenet-Serret formulae can also be given in the following
form:
(140) T’ = d × T
(141) N’ = d × N
(142) B’ = d × B
where d is the “Darboux vector” which is given by:
(143) d = τT + κB
This form of the Frenet-Serret formulae is more memorable
apart from the expression of d. We note that some authors
define d as a scalar multiple of what is given in Eq. 143↑. The
above three equations (i.e. Eqs. 140↑-142↑) may be merged in
a single equation as:
(144) (T’, N’, B’) = d × (T, N, B)
In general curvilinear coordinates, the Frenet-Serret formulae
are given in terms of the absolute derivatives of the three
vectors by:
(145) δTi ⁄ δs = (dTi ⁄ ds) + ΓijkTj(dxk ⁄ ds) = κNi
(146) δNi ⁄ δs = (dNi ⁄ ds) + ΓijkNj(dxk ⁄ ds) = τBi − κTi
(147) δBi ⁄ δs = (dBi ⁄ ds) + ΓijkBj(dxk ⁄ ds) = − τNi
where the indexed x represent general spatial coordinates and s
is a natural parameter while the other symbols are as defined
earlier.
According to the fundamental theorem of space curves, which
is outlined previously in § 2.3↑, a curve does exist and it is
unique iff its curvature and torsion as functions of arc length
are given. Now, it is natural to expect that such a solution can
be obtained from the system of differential equations given by
the Frenet-Serret formulae. However, such a solution cannot be
obtained in general by direct integration of these equations.
More elaborate methods (e.g. methods based on the Riccati
equation for reducing a system of simultaneous differential
equations to a first order differential equation) may be used to
obtain the solution. Nevertheless, a solution can be obtained by
direct integration of the Frenet-Serret formulae for plane
curves (see § 5.2↓) where the torsion vanishes identically. A
solution by direct integration of the Frenet-Serret formulae can
also be obtained in other simple cases such as when the
curvature and torsion are constants.
2.6 Osculating Circle and Sphere
At any point P with non-zero curvature of a smooth space
curve C, an “osculating circle” (Fig. 25↓), which may also be
called the circle of curvature or the kissing circle, can be
defined where this circle is characterized by:
1. It is tangent to C at P, i.e. the circle and the curve have a
common tangent vector at P.
2. It lies in the osculating plane of C at P.
3. Its radius Rκ is equal to (1 ⁄ κ) where κ is the curvature of C
at P.
4. Its center Cc is located at rC which is given by:
(148)
rC =
rP + (1 ⁄ κ)N =
rP + RκN
where rP is the position vector of P and N is the principal
normal vector of C at P.
The center of curvature of a curve at a point on the curve is
defined as the center of the osculating circle at that point, as
given above. If the curve C is a circle, then the center of
curvature at any point is the center of the circle itself, so the
circle is its own osculating circle.
Figure 25 The osculating circle Co of a space curve C at
point P with the principal normal vector N, center of curvature
Cc and radius of curvature Rκ.

The osculating circle provides a good approximation to the

curve in the neighborhood of its points where the osculating
circle is defined. Following the manner of defining the tangent
line to a curve as a limit of the secant line (see § 2.1↑), the
osculating circle to a curve at a given point P may be defined
geometrically as the limit of a circle passing through P and two
other points on the curve as these two points converge to P
while staying on the curve (Fig 26↓). It should be remarked
that in some cases the osculating circle and its parameters may
be defined geometrically but not analytically when the second
derivative of the curve at the given point is not properly
defined to determine the radius of curvature (see Eq. § 125↑).
Figure 26 The osculating circle (small) of a curve C at a
point P as a limit of another circle (big) passing through P and
two other points, P1 and P2.

Following the manner of defining the osculating circle as a

limit, the “osculating sphere” of a curve C at a given point P
may be defined similarly as the limit of a sphere passing
through P and three neighboring points on the curve as these
three points converge to P. The position of the center CS of the
osculating sphere at P, which is called the center of spherical
curvature of C at P, is given by:
(149)
rS =
rP + (1 ⁄ κ)N − [κ’ ⁄ (τκ2)] B =
rP + RκN + sgn(τ)RτRκ’B
where rS and rP are the position vectors of CS and P, B and N
are the binormal and principal normal vectors, κ and τ are the
curvature and torsion, Rκ and Rτ are the radii of curvature and
torsion, sgn(τ) is the sign function of τ(s), and the prime
represents derivative with respect to a natural parameter s of C.
All these quantities belong to C at P which should have non-
vanishing curvature and torsion, i.e. κ, τ ≠ 0. From Eq. 149↑, it
can be seen that the radius of the osculating sphere is given by:
(150) |rS − rP| = √(R2κ + [RτRκ’]2)
2.7 Parallelism and Parallel
Propagation
In flat spaces, parallelism is an absolute property as it is
defined without reference to a peripheral object. However, in
Riemannian spaces the idea of parallelism is defined in
reference to a prescribed curve and hence it is different from
the idea of parallelism in the Euclidean sense. A vector field Aα
is described as being parallel along the surface curve uβ = uβ(t)
iff its absolute derivative (see § 7↓) along the curve vanishes,
that is:
(151)
δAα ⁄ δt ≡
Aα;β(duβ ⁄ dt) ≡
(dAα ⁄ dt) + ΓαβγAγ(duβ ⁄ dt) = 0
This means that the sufficient and necessary condition for a
vector field to be parallel along a surface curve is that the
covariant derivative of the field is normal to the surface.
All the vectors of a field of parallel vectors have the same
constant magnitude. A field of absolutely parallel unit vectors
on a surface do exist iff there is an isometric correspondence
between the plane and the surface. When two surfaces are
tangent to each other along a given curve C, then a vector field
which is parallel along C with respect to one of these surfaces
will also be parallel along C with respect to the other surface.
When two non-trivial vectors experience parallel propagation
along a particular curve the angle between them stays constant.
As a consequence of the definition of parallelism in
Riemannian spaces and the previous statements, we have:
1. Parallel propagation is path dependent. Hence, a surface
vector field parallelly propagated along a given curve between
two points P1 and P2 on the curve does not necessarily
coincide with another vector field parallelly propagated along
another curve connecting P1 and P2.
2. Since parallel propagation depends on the path of
propagation, then given two points P1 and P2 on a surface, the
vector obtained at P2 by parallel propagation of a vector from
P1 along a given surface curve C connecting P1 to P2 depends
on the curve C.
3. Starting from a given point P on a closed surface curve C
enclosing a simply connected region (see § 3.1↓) on the
surface, parallel propagation of a vector field around C starting
from P does not necessarily result in the same vector field
when arriving at P. We note that the angle between the initial
and final vectors in this situation is a measure of the Gaussian
curvature on the surface (see § 4.5↓).
4. If C:I → S is a regular curve on a surface S defined on the
⊆
interval I ℝ, and v1 and v2 are parallel vector fields over C,
then the dot product v1⋅ v2 which is associated with the metric
tensor, the norm of the vector fields |v1| and |v2|, and the angle
between v1 and v2 are constants.
2.8 Exercises
Exercise 1. State the technical definition of space curve
outlining the difference between a curve and its trace.
Exercise 2. What is the most common way of defining space
curves mathematically? Give an example from simple curves
like circle and ellipse.
Exercise 3. Make a clear distinction between general and
natural parameterization of space curves.
Exercise 4. State a mathematical condition for a space curve to
be parameterized naturally.
Exercise 5. What is the relation between two natural
parameters of a given space curve?
Exercise 6. The following equation:
r(t) = (t ⁄ 2, − t ⁄ 2, t ⁄ √(2))
is a parametric representation of a space curve. Is t a natural
parameter or not? Justify your answer.
Exercise 7. Prove that two natural parameters, s and š, of a
curve are related by the equation
š = ±s + c
where c is a real constant.
Exercise 8. Using tensor notation, write down the equation of
the tangent vector to a surface curve represented parametrically
by C(u1(t), u2(t)) where t is a general parameter.
Exercise 9. What is the meaning of having “regular curve at a
specific point”? What “regular curve” means?
Exercise 10. Prove that the parametric representation:
r(t) = (t2, et, t + 1)
of a space curve is regular for all t.
Exercise 11. State the condition for a vector to be tangent to a
regular surface at a given point on the surface.
Exercise 12. Find the unit tangent vector, T(t), for a space
curve represented by:
r(t) = (t2, t, sint).
Exercise 13. Find the arc length of a space curve given by:
r(t) = (5t, 7cosht, 2sinht)
for 1 ≤ t ≤ 4.
Exercise 14. Find the curvature, as a function of t, of a space
curve represented by:
r(t) = (cost − 1, sint + t, t2).
Exercise 15. Define “periodic curve” giving two common
examples other than those given in the text.
Exercise 16. Should a periodic curve be a plane curve?
Exercise 17. Should a periodic continuous curve be a closed
curve? Should a periodic smooth curve be a closed curve?
Exercise 18. A plane curve called cissoid of Diocles is given in
polar coordinates by:
ρ = 2 sinφ tanφ.
Find the equation of the curve in a rectangular Cartesian
coordinate system.
Exercise 19. Sketch the curve of the previous question for 0 ≤ |
φ| ≤ (π ⁄ 4) (notice the two branches).
Exercise 20. Show that for a curve represented spatially by r
and parameterized naturally by s and generally by t, the
relation between s and t is given by:
|ds ⁄ dt| = |dr ⁄ dt|.
Exercise 21. Define Frenet frame with a simple sketch
showing the basis vectors at a given point on an arbitrary space
curve.
Exercise 22. Write down the mathematical equations of the
unit vectors T, N, B for a curve parameterized by a general
parameter t and a natural parameter s.
Exercise 23. State the mathematical definition of the curvature
κ and the torsion τ of an s-parameterized space curve.
Exercise 24. Show that the sufficient and necessary condition
for a space curve to be a plane curve is that its torsion vanishes
identically.
Exercise 25. What are the curvature and torsion of
(a) a straight line
(b) a circle with radius R = 3.2
(c) a curve parameterized by: r(t) = (3cos(t), 3sin(t), 1.9t)?
Exercise 26. Find the torsion, as a function of t, of
(a) a curve represented by: r(t) = (sint + t, cost − 3, t + 2)
(b) a curve represented by: r(t) = (t, 2t2, t3)
(c) a curve represented by: r(t) = (at, bt3, ct2) where a, b, c are
non-vanishing real constants.
Exercise 27. What are the mathematical conditions that
represent the fact that the vectors T, N, B are mutually
orthogonal and of unit length?
Exercise 28. Prove the theorem of Lancret which states that a
space curve of class C3 with non-vanishing curvature is a helix
iff the ratio of its torsion to its curvature is constant along the
curve.
Exercise 29. Show that if two space curves, which have an
injective association, possess parallel tangent vectors at their
corresponding points, then their normal and binormal vectors
at these points are parallel as well.
Exercise 30. Prove Eq. 121↑ (i.e. |κτ| = |T’⋅B’|).
Exercise 31. Give a parametric representation of a circular
helix using a natural parameter s.
Exercise 32. For a plane curve in a 3D space given by the
equation:
y = 2x2 − x + 3 (0 ≤ x ≤ 10),
find the equations of the osculating, normal and rectifying
planes at point (1, 4).
Exercise 33. For a space curve parameterized as:
(x, y, z) = (t, t3, 3t2),
find the equations of the tangent, principal normal and
binormal lines passing through the point (1, 1, 3) on the curve.
Exercise 34. Give an example of a non-planar space curve
whose principal normal vectors at all points of the curve are
parallel to a particular plane.
Exercise 35. For a space curve parameterized as:
(x, y, z) = (cost, sint, 5t),
find the equations of the osculating, rectifying and normal
planes at the point on the curve with t = 1.3.
Exercise 36. Which of the three vectors T, N, B is not affected
by reversing the sense of traversing the space curve?
Exercise 37. What are the principal normal vector N and the
binormal vector B of a helix represented by:
r(t) = (cos3t, sin3t, 3t)?
Exercise 38. Find the three vectors T, N, B along a curve
represented by:
r(t) = (2t − 3, t3 + t, 5 − t2).
Exercise 39. Make a simple sketch for the osculating,
rectifying and normal planes at a point on an arbitrary space
curve. Use a computer graphic package if convenient.
Exercise 40. Write down the equation of Lancret related to the
third curvature of space curves and discuss its significance.
Exercise 41. Discuss, in detail, the fundamental theorem of
space curves in differential geometry outlining its application
and significance.
Exercise 42. Given that the curvature of a plane curve is given
by:
κ = 1 ⁄ (3s + 5)
where s > 0 is a natural parameter, find the equation of this
curve.
Exercise 43. Write down the Frenet-Serret formulae of a space
curve in a rectangular Cartesian coordinate system explaining
all the symbols involved.
Exercise 44. By integrating the Frenet-Serret formulae, obtain
the solution of a space curve with κ = a and τ = b where a, b >
0 are real constants.
Exercise 45. Identify the type of the curve in the previous
question.
Exercise 46. What are the “intrinsic” or “natural” equations of
a curve?
Exercise 47. Find the intrinsic equations of the catenary
defined by Eqs. 28↑-29↑.
Exercise 48. Find the parametric representation of a curve with
the following intrinsic equations:
κ = √(c ⁄ s) and τ = 0
where c > 0 is a real constant and s > 0 is a natural parameter.
Exercise 49. Prove that the curvature of a t-parameterized
space curve is given by:
κ = |r × r| ⁄ |r|3
Exercise 50. Prove that the torsion of a t-parameterized space
curve is given by:
τ = [r⋅(r × r)] ⁄ |r × r|2
Exercise 51. Which of the two main curve parameters, κ and τ,
is necessarily non-negative and why?
Exercise 52. Show that along an s-parameterized curve r(s),
the following relation holds true:
r’⋅(r’’ × r’’’) = τκ2.
Exercise 53. Can we obtain the curvature and torsion of circle
as a special case of the curvature and torsion of helix? If yes,
how? Is this consistent with the definition of helix as given in §
1.4.1↑ (see Eqs. 4↑-6↑ and Fig. 2↑)?
Exercise 54. Find the curvature and torsion of a space curve
represented by:
r(t) = (t3, t + 1, − t2)
at the point with t = 2.4.
Exercise 55. Give an example of a space curve whose
curvature and torsion are variables.
Exercise 56. Investigate the relation between the curvature and
torsion at corresponding points of two space curves which are
mirror-reflection of each other with respect to a given plane.
Exercise 57. Investigate the relation between the curvature and
torsion at corresponding points of two space curves which are
symmetric with respect to a given point.
Exercise 58. Discuss, in detail, the concept of “1D inhabitant”
in the context of classifying the properties of space curves.
Exercise 59. Establish a correspondence between the two main
parameters of space curve (i.e. curvature and torsion) and the
first and second fundamental forms of space surface.
Exercise 60. Discuss the resemblance between κ and τ of space
curve and the curvature tensor of space surface.
Exercise 61. State, in words, the mathematical relation:
τ = εijkTiNj(δNk ⁄ δs)
using technical terms for all the notations and symbols used in
this equation.
Exercise 62. What is the significance of the curvature and
torsion of space curves as measures of their variation in the
embedding space?
Exercise 63. What is the relation between the curvature and
the radius of curvature of a space curve? Is there an advantage
in using one of these or the other as the main concept?
Exercise 64. Outline the physical significance of the relation:
κ = |r’’|.
Exercise 65. Express κ in terms of r and its first and second
derivatives where r is a spatial representation of a curve
parameterized by a general parameter t.
Exercise 66. Express τ in terms of N’ and B of a naturally
parameterized curve.
Exercise 67. Express τ in terms of r and its first, second and
third derivatives where r is a spatial representation of a curve
parameterized by a general parameter t.
Exercise 68. What is the relation between the torsion and the
radius of torsion of a space curve?
Exercise 69. Define geodesic torsion in words and state its
mathematical relation to r and n and their derivatives.
Exercise 70. What is the significance of the relation:
τg = τ − (dφ ⁄ ds)
and what the symbols in this relation mean?
Exercise 71. What is the condition for the torsion and the
geodesic torsion of a space curve to be equal?
Exercise 72. Prove that along a sufficiently smooth curve
represented by r(t), the vector r at a given point P on the curve
is parallel to the osculating plane at P.
Exercise 73. Obtain the equation of the osculating plane of a
curve represented parametrically by:
r(t) = (3cost, 2sint, cost + 5sint)
at a general point on the curve.
Exercise 74. Derive the second of the Frenet-Serret formulae
using the other two formulae.
Exercise 75. Define Darboux vector d and hence verify that
the relations given by Eqs. 140↑-142↑ are valid.
Exercise 76. Explain all the symbols and notations used in the
following relation:
τ = [r⋅(r × r)] ⁄ [(r⋅r)(r⋅r) − (r⋅r)2]
Exercise 77. Prove that for a curve with helical shape, the
Darboux vector is constant along the curve.
Exercise 78. State the Frenet-Serret formulae in a single
equation using the Darboux vector.
Exercise 79. Write down the Frenet-Serret formulae assuming
a general curvilinear coordinate system.
Exercise 80. Give a brief explanation of why the solution of a
space curve cannot be obtained in general by a direct
integration of the Frenet-Serret equations.
Exercise 81. Give a brief definition of the osculating circle and
the osculating sphere of a space curve.
Exercise 82. How the osculating circle and the osculating
sphere of a space curve can be defined using the concept of
limit?
Exercise 83. Prove that for a given space curve C, the
binormal lines of C and the tangent lines to the locus of the
centers of spherical curvature of C are parallel at their
corresponding points.
Exercise 84. Derive an expression for the position of the center
of curvature of a t-parameterized twisted curve represented
spatially by r(t) in terms of r and its derivatives.
Exercise 85. Find the spatial coordinates of the center of the
osculating circle of a space curve represented by:
r(t) = (t, √(t), t2)
at the point with t = 2.6.
Exercise 86. What is the relation between the osculating circle
and the osculating plane at a given point of a space curve?
Exercise 87. Explain all the symbols involved in the equation:
rS = rP + RκN + sgn(τ)RτRκ’B
which is related to the center of the osculating sphere.
Exercise 88. Write down the formula for the radius of the
osculating sphere explaining all the symbols used.
Exercise 89. What is the difference between the concept of
parallelism in Euclidean and non-Euclidean spaces?
Exercise 90. State the mathematical condition for a vector field
to be parallel along a given surface curve in terms of its
absolute derivative.
Exercise 91. What is the meaning of describing parallel
propagation as path dependent? What are the consequences of
this dependency?
Chapter 3
Surfaces in Space
Here, we examine sufficiently smooth surfaces embedded in a
3D Euclidean space using a Cartesian coordinate system (x, y,
z) for the most parts. Some parts are based on a more general
Riemannian space with a curvilinear coordinate system.
3.1 General Background about
Surfaces
A 2D surface embedded in a 3D space may be defined loosely
as a set of connected points in the space such that the
immediate neighborhood of each point on the surface can be
deformed continuously to form a flat disk. Technically, a
surface in a 3D manifold is a mapping from a subset of the
parameters plane to a 3D space, that is S:Ω → ℝ3, where Ω is a
subset of ℝ2 plane and S is a sufficiently smooth injective
function. Similar conditions may also be imposed to ensure the
existence of a tangent plane and a normal vector at each point
of the surface. In particular, the condition
∂ur × ∂vr ≠ 0
at all points on the surface is usually imposed to ensure
regularity of the surface.
Like space and surface curves, the image of the mapping in ℝ3
is known as the trace of the surface. For convenience, we use
curve and surface in the present book for trace as well as for
mapping; the meaning should be obvious from the context. We
should remark that the trace of a curve or a surface should not
be confused with the trace of a matrix which is the sum of its
diagonal elements.
A 2D surface embedded in a 3D space can be defined
explicitly by: z = f(x, y), or implicitly by: F(x, y, z) = 0, or
parametrically by: x(u1, u2), y(u1, u2), z(u1, u2) where u1 and
u2 (or u and v) are the surface coordinates on the parameters
plane, as defined previously (see § 1.4.3↑), which are mutually
independent parameters. By substitution, elimination and
algebraic manipulation these forms can be transformed
interchangeably.
A coordinate patch of a surface is an injective, bicontinuous,
regular, parametric representation of a part of the surface. In
more technical terms, a coordinate patch of class Cn (n > 0) on
a space surface S is a functional mapping of an open set Ω in
the uv parameters plane onto S that satisfies the following
conditions:
1. The functional mapping relation is of class Cn over Ω.
2. The mapping is one-to-one and bicontinuous over Ω.
3. E1 × E2 ≠ 0 at any point in Ω where E1 and E2 are the
surface basis vectors (see § 1.4.5↑ and 1.4.3↑).
As indicated previously, a vector v is described as a tangent
vector to the surface S at a given point P on the surface if there
is a regular curve C embedded in S and passing through P such
that v is a tangent to the curve at P, i.e.
∈
v = c(dr ⁄ dt) (c ℝ, c ≠ 0)
where r is a t-parameterized position vector representing C.
The set of all tangent vectors to the surface S at point P forms a
tangent plane to S at P (Fig. 27↓). This set is called the tangent
space of S at P and it is usually notated with TPS. It is obvious
that there exist infinitely many tangent vectors, with varying
magnitude and direction, to a smooth surface at its regular
points.
Figure 27 Tangent plane (solid) of a surface (outlined by a
grid) at a given point alongside the surface basis vectors, E1
and E2, and the unit normal vector, n, at that point.

As we will see (also refer to § 1.4.5↑), the tangent space of a

regular surface at a given point on the surface is the span of the
two linearly independent basis vectors defined as:
(152)
E1 = ∂ r ⁄ ∂u1
E2 = ∂ r ⁄ ∂u2
where r(u1, u2) is the spatial representation of the surface
coordinate curves and u1 and u2 are the surface coordinates, as
explained before. The tangent space, therefore, is the plane
passing through P and is perpendicular to the vector E1 × E2.
As indicated previously, every vector tangent to a regular
surface S at a given point P on S can be expressed as a linear
combination of the surface basis vectors E1 and E2 at P. The
reverse is also true, that is every linear combination of E1 and
E2 at P is a tangent vector to a regular curve embedded in S
and passing through P and hence it is a tangent to S at P.
Based on the previous statements, we see that the tangent plane
of a surface at a given point P on the surface can be given by:
(153) r = rP + pE1 + qE2
where r is the position vector of an arbitrary point on the
∈
tangent plane, rP is the position vector of the point P, p, q (
− ∞, ∞) are real variables, and E1 and E2 are the surface basis
vectors at P. Alternatively, the tangent plane can be expressed
in terms of the normal vector n to the surface at P by:
(154) (r − rP)⋅n = 0
The tangent space at a specific point P of a surface is a
property of the surface at P and hence it is independent of the
patch that contains P and the particular parameterization of the
surface. For any non-trivial vector v which is parallel to the
tangent plane of a simple and smooth surface S at a given point
P on S, there is a curve in S passing through P and represented
parametrically by r(t) such that v = c(dr ⁄ dt) where c ≠ 0 is a
real constant. As a result of the last and the previous
statements, a non-trivial vector is parallel to the tangent plane
of a surface S at a given point P iff it is equal to a tangent
vector to S at P.
The straight line passing through a given point P on a surface S
in the direction of the normal vector of S at P is called the
normal line to S at P. The equation of this normal line is given
by:
(155) r = rP + kn
where r is the position vector of an arbitrary point on the
∈
normal line, rP is the position vector of the point P, k ( − ∞,
∞) is a real variable, and n is the unit normal vector to the
surface at P. We remark that this normal line should not be
confused with the normal line of a surface curve C that passes
through P which is usually called the principal normal line of
C (see § 2.2↑). Anyway, the two should be easily distinguished
by noticing their affiliation to surface or curve.
As stated before, a surface is regular at a given point P iff E1
× E2 ≠ 0 at P where E1 = ∂ur and E2 = ∂vr are the tangent
vectors to the coordinate curves at P. A surface is regular iff
E1 × E2 ≠ 0 at any point on the surface. A regular curve (see §
2.1↑) of class Cn on a sufficiently smooth surface is an image
of a unique regular plane curve of class Cn in the parameters
plane, where in this statement we are considering each
connected part of the curve being embedded in a coordinate
patch if there is no single patch that contains the entire curve.
A “Monge patch” is a coordinate patch in a 3D space defined
by a function in one of the following three forms:
(156) r(u, v) = (f(u, v), u, v)
(157) r(u, v) = (u, f(u, v), v)
(158) r(u, v) = (u, v, f(u, v))
where f is a differentiable function of the surface coordinates u
and v. When f is of class Cn then the coordinate patch is of this
class.
A simply connected region on a surface means that a closed
curve contained in the region can be shrunk down continuously
onto any point in the region without leaving the region. In
more simple terms, it means that the region contains no holes
or gaps that separate its parts. A simple surface is a
continuously deformed plane by compression, stretching and
bending (see Footnote 10 in § 8↓). Examples of simple surface
are cylinders, cones and elliptic and hyperbolic paraboloids. A
connected surface S is a simple surface which cannot be
entirely represented by the union of two disjoint open point
sets in ℝ3 where these sets have non-empty intersection with S.
Hence, for any two arbitrary points, P1 and P2, on S there is a
continuous curve which is totally embedded in S with P1 and
P2 being its end points. Examples of connected surface are
planes, ellipsoids and cylinders.
A closed surface is a simple surface with no open edges.
Examples of closed surface are spheres and ellipsoids. A
bounded surface is a surface that can be contained entirely in a
sphere of a finite radius such as ellipsoid and torus. A compact
surface is a surface which is bounded and closed like a torus or
a Klein bottle (Fig. 28↓). If f is a differentiable regular
mapping from a surface S to a surface S̃ , then if S is compact
then S̃ is compact. If S1 and S2 are two simple surfaces where
S1 is connected and S2 is closed and contained in S1, then the
two surfaces are equal as point sets. As a result, a simple
closed surface cannot be a proper subset of a simple connected
surface.
Figure 28 Klein bottle.

An orientable surface is a simple surface over which a

continuously-varying normal vector can be defined. Hence,
spheres, cylinders and tori are orientable surfaces while the
Mobius strip (Fig. 1↑) is a non-orientable surface since a
normal vector moved continuously around the strip from a
given point will return, following a complete round, to the
point in the opposite direction. Similarly, Klein bottle (Fig.
28↑) is another example of a non-orientable surface. An
oriented surface is an orientable surface over which the
direction of the normal vector is determined. An orientable
surface which is connected can be oriented in only one of two
possible ways.
An elementary surface is a simple surface which possesses a
single coordinate patch basis, and hence it is an orientable
surface which can be mapped bicontinuously to an open set in
the plane. Examples of elementary surface are planes, cones
and elliptic paraboloids. A developable surface is a surface
that can be flattened into a plane by unfolding without local
distortion by compression or stretching. It is called developable
because it can be developed into a plane by rolling the surface
out on a plane without compression or stretching. A
characteristic feature of developable surface is that, like the
plane, its Riemann-Christoffel curvature tensor (see § 1.4.10↑)
and Gaussian curvature (see § 4.5↓) are zero at every point on
the surface. Cylinders and cones are obvious examples of
developable surface.
A topological property of a surface is a property which is
invariant with respect to injective bicontinuous mappings. An
example of a topological property is compactness. A
differentiable regular mapping from a surface S to a surface S̃
is called conformal if it preserves angles between oriented
intersecting curves on the surface. The mapping is described as
direct if it preserves the sense of the angles and inverse if it
reverses it. Technically, the mapping is conformal if there is a
function q(u, v) > 0 that applies to all patches on the surface
such that:
(159) aαβ = q ãαβ
where α, β = 1, 2 and the unbarred and barred indexed a are the
coefficients of the surface covariant metric tensor in S and S̃
respectively. The above condition of conformal mapping may
be stated by saying that the coefficients of the first fundamental
forms of the two surfaces are proportional at their
corresponding points. An example of conformal mapping is the
stereographic projection (Fig. 29↓) from the Riemann sphere to
a plane. We remark that stereographic projection is a mapping
of the unit sphere onto a plane where each point of the sphere
is projected, through the line connecting this point to the north
pole of the sphere, onto the point of intersection of that line
with the plane. This plane is the tangent plane to the sphere at
its south pole.
Figure 29 Stereographic projection where the points P1 and
P2 on the unit sphere are projected respectively on the points
P̃1 and P̃2 on the plane with N representing the north pole of
the sphere. The plane is touching the sphere at its south pole.

An isometry or isometric mapping of two surfaces is a one-to-

one mapping from a surface S to a surface S̃ that preserves
distances. Hence, any arc in S is mapped onto an arc in S̃ with
equal length. The two surfaces S and S̃ are described as
isometric surfaces. An example of isometric mapping is the
deformation of a rectangular plane sheet into a cylinder with
no local distortion by compression or stretching and hence the
two surfaces are isometric since all distances are preserved
during this process. Isometry is a symmetric relation and hence
the inverse of an isometric mapping is an isometric mapping,
that is if f is an isometry from S to S̃ , then f − 1 is an isometry
from S̃ to S (refer to § 6.5↓ for more details).
An injective mapping from a surface S onto a surface S̃ is an
isometry iff the coefficients of the first fundamental form (see
§ 3.5↓) for any patch on S are identical to the corresponding
coefficients of the first fundamental form of its image on S̃ ,
that is:
(160)
E=Ẽ
F = F̃
G = G̃
where the unbarred and barred E, F, G are the coefficients of
the first fundamental form of the two surfaces at their
corresponding points. As seen and will be seen, the coefficients
of the first fundamental form are the same as the coefficients of
the surface covariant metric tensor, that is:
(161)
a11 = E
a12 = a21 = F
a22 = G
and hence these conditions mean that the two surfaces have the
same metric at their corresponding points. The mapping that
preserves distances but it is not injective is described as local
isometry. The statement about the equality of the coefficients
of the first fundamental form on the two surfaces also applies
to local isometry.
Since intrinsic properties are dependent only on the
coefficients of the first fundamental form of the surface,
intrinsic properties of the surface are invariant with respect to
isometric mappings. As a consequence of the equality of
corresponding lengths of two isometric surfaces, the
corresponding angles are also equal. However, the reverse is
not true, that is a mapping that attains the equality of
corresponding angles does not necessarily ensue the equality of
corresponding lengths. Hence, isometric mapping is more
restrictive than conformal mapping. This means that every
isometric mapping is conformal but not every conformal
mapping is isometric, so isometric mapping is a subset of
conformal mapping. This can be seen by comparing Eqs. 159↑
and 160↑ where the latter corresponds to the former with q = 1.
In fact, conformal mapping can be set up between any two
surfaces and in many different ways but this is not always
possible for isometric mapping. Isometric mapping also
preserves areas of mapped surfaces since it preserves lengths
and angles.
A surface generated by the collection of all the tangent lines to
a given space curve is called the “tangent surface” of the curve
while the tangent lines are called the generators or the rulings
of the surface. Similarly, a “branch” of the tangent surface of a
curve C at a given point P on the curve refers to the tangent
line of C at P. In this context, we should remark that the
“tangent surface” of a curve should not be confused with the
aforementioned “tangent plane” of a surface at a given point.
The tangent surface of a curve may be demonstrated visually
by a taut flexible string connected to the curve where it scans
the surface while being directed tangentially at each point of
the curve at its base. However, the taut string visualization
should extend to both tangential directions to give the full
extent of the tangent surface (see § 6.6↓ for more details).
If Ce is a space curve with a tangent surface ST and Ci is a
curve embedded in ST and is orthogonal to all the tangent lines
of Ce at their intersection points, then Ci is called an involute
of Ce while Ce is called an evolute of Ci (see § 5.3↓).
3.2 Mathematical Description of
Surfaces
We start by assuming a parametric representation for the
surface, where each one of the space coordinates (x, y, z) on the
surface is a real differentiable function of the two surface
coordinates (u, v). The position vector of a point P on the
surface as a function of the surface coordinates is then given
by:
(162) r(u, v) = x(u, v)e1 + y(u, v)e2 + z(u, v)e3
where the indexed e are the Cartesian orthonormal basis
vectors in the three directions.
It is also assumed that ∂ur and ∂vr are linearly independent and
hence they are not parallel or anti-parallel, that is:
(163) (∂r ⁄ ∂u) × (∂r ⁄ ∂v) ≠ 0
As seen before, this is a sufficient and necessary condition for
the surface to be “regular” at a given point. The point is also
described as “regular”; otherwise it is “singular” if the
condition is violated. The surface is regular on Ω, a closed
subset of ℝ2, if it is regular at each interior point of Ω. The
regularity condition guarantees that the surface mapping is
one-to-one and possesses a continuous inverse. It also ensures
the existence of a tangent plane and a normal vector where this
condition is satisfied.
To express the position vector of P in tensor notation, we re-
label the space and surface coordinates as:
(164)
(x, y, z) ≡ (x1, x2, x3)
(u, v) ≡ (u1, u2)
and hence the position vector of Eq. 162↑ becomes:
(165) r(u1, u2) = xi(u1, u2) ei
where i = 1, 2, 3. We note that relabeling the surface
coordinates is not necessary in this notation but it will be
useful in the future for other tensor notations. To define a
surface grid serving as a curvilinear positioning system for the
surface, one of the surface coordinate variables is held fixed in
turn while the other is varied (Figs. 20↑ and 30↓). Hence, each
one of the following two surface functions:
(166)
r(u1, c2)
r(c1, u2)
defines a coordinate curve for the surface, where c1 and c2 are
given real constants. These two coordinate curves meet at the
common surface point (c1, c2). The grid is then generated by
varying c1 and c2 uniformly to obtain coordinate curves at
regular intervals in its domain.
Figure 30 Surface coordinate grid with the surface covariant
basis vectors, E1 and E2, and the normal vector to the surface n
at a particular point on the surface where “CC” stands for
“coordinate curves”.

Corresponding to each one of the surface coordinate curves in

the above order, a tangent vector to the curve at a given point
on the curve is defined by:
(167)
Eα =
∂r ⁄ ∂uα =
(∂xi ⁄ ∂uα)ei =
xiαei
where the derivatives are evaluated at that point, and α = 1, 2
and i = 1, 2, 3. So in brief, E1 is tangent to the r(u1, c2)
coordinate curve and E2 is tangent to the r(c1, u2) coordinate
curve. These tangent vectors serve as a set of basis vectors for
the surface, and for each regular point on the surface they
generate, by their linear combination, any vector in the surface
at that point (see Footnote 11 in § 8↓). They also define, by
their linear combination, a plane tangent to the surface at that
point. The plane generated by the linear combination of E1 and
E2 is the aforementioned tangent space, TPS, to the surface at
point P as described earlier, and hence E1(u1P, u2P) and
E2(u1P, u2P) form a basis for this space where the subscript P is
a reference to the point P.
We should remark that the surface basis vectors, E1 and E2, are
defined on all points of the surface and not only on the points
of the above-described regularly spaced grid of coordinate
curves which is presented in that way for pedagogical reasons.
Also, a coordinate curve is any surface curve along which only
one coordinate variable, u1 or u2, varies regardless of being
part of the above grid or not. Another important remark is that
the surface coordinate curves are orthogonal at every point on
the surface iff the surface metric tensor (see § 3.3↓) is diagonal
everywhere on the surface. This is equivalent to having an
identically vanishing F, which is the coefficient of the first
fundamental form, as can be seen for example from Eq. 235↓
where the dot product will vanish due to the orthogonality of
E1 and E2 which are the tangents to the coordinate curves.
When this condition is satisfied, the coordinate system of the
surface is described as orthogonal. Also, the surface coordinate
curves are orthogonal at any particular point on the surface iff
F = 0 at that point even if this condition is not satisfied over
the entire surface.
Following the definition of the surface basis vectors, E1 and
E2, a normal vector to the surface at the given point is then
defined as the cross product of these tangent basis vectors: E1
× E2. This normal vector, like E1 and E2, is a function of
position on the surface and hence in general it varies
continuously, in magnitude and direction, as it moves around
the surface. This normal vector can be scaled by its magnitude
to produce a unit normal vector n to the surface at that point,
that is (see Footnote 12 in § 8↓):
(168)
n=
(E1 × E2) ⁄ | E1 × E2| =
(E1 × E2) ⁄ √(a)
where a is the determinant of the surface covariant metric
tensor (see § 3.3↓). Based on the cross product rule, the vectors
of the triad E1, E2, n form a right handed system (see Fig.
30↑).
On dot producting both sides of Eq. 168↑ with n, which is a
unit vector, we obtain:
(169) n⋅(E1 × E2) = √(a)
We may also take the modulus of both sides of Eq. 168↑ (or
just compare the denominators of the second equality of Eq.
168↑) to obtain:
(170) |E1 × E2| = √(a)
The surface basis vectors, Eα, are symbolized in full tensor
notation by:
(171) [Eα]i ≡ Eiα = ∂xi ⁄ ∂uα = xiα
(i = 1, 2, 3 and α = 1, 2) and hence they can be regarded as 3D
contravariant space vectors or as 2D covariant surface vectors
(refer to § 3.3↓ for further details).
It can be shown that the covariant form of the unit normal
vector n to the surface is given in full tensor notation by:
(172) ni = (1 ⁄ 2)εαβεijkxjαxkβ
where xjα = (∂xj ⁄ ∂uα) and similarly for xkβ, and εαβ and εijk are
the absolute permutation tensors for the surface and space. The
implication of this equation, which defines n in terms of the
surface basis vectors xjα and xkβ, is that n is a space vector
which is independent of the choice of the surface coordinates,
u1 and u2, in support of the geometric intuition. Since n is
normal to the surface, we have:
(173) gijnixjα = 0
which is a statement, in tensor notation, that n is orthogonal to
every vector in the tangent space of the surface at the given
point. In this equation, gij is the space metric tensor.
Although E1 and E2 are linearly independent they are not
necessarily orthogonal or of unit length. However, they can be
orthonormalized as follows:
(174)
Ê1 = E1 ⁄ | E1| = E1 ⁄ √(a11)
Ê2 = (a11 E2 − a12 E1) ⁄ √(a11a)
where a is the determinant of the surface covariant metric
tensor (see § 3.3↓), the indexed a are the coefficients of this
tensor, and the hatted vectors are orthonormal basis vectors,
that is:
(175)
Ê1⋅ Ê1 = 1
Ê2⋅ Ê2 = 1
Ê1⋅ Ê2 = 0
This can be verified by conducting the dot products of the last
equation using the vectors defined in Eq. 174↑.
The transformation rules from one surface coordinate system
to another surface coordinate system, notated with unbarred
(u1, u2) and barred (ũ1, ũ2) symbols respectively, where:
(176)
u1 = u1(ũ1, ũ2)
u2 = u2(ũ1, ũ2)
(177)
ũ1 = ũ1(u1, u2)
ũ2 = ũ2(u1, u2)
are similar to the general rules for the transformation between
coordinate systems in a general nD space, as explained in the
textbooks of tensor analysis.
Following a transformation from an unbarred surface
coordinate system to a barred surface coordinate system, the
surface becomes a function of the barred coordinates, and a
new set of basis vectors for the surface, which are the tangents
to the coordinate curves of the barred system, are defined by
the following equations:
(178)
Ẽ1 =
∂r ⁄ ∂ũ1 =
(∂r ⁄ ∂u1)(∂u1 ⁄ ∂ũ1) + (∂ r ⁄ ∂u2)(∂u2 ⁄ ∂ũ1) =
E1(∂u1 ⁄ ∂ũ1) + E2(∂u2 ⁄ ∂ũ1)
(179)
Ẽ2 =
∂r ⁄ ∂ũ2 =
(∂r ⁄ ∂u1)(∂u1 ⁄ ∂ũ2) + (∂ r ⁄ ∂u2)(∂u2 ⁄ ∂ũ2) =
E1(∂u1 ⁄ ∂ũ2) + E2(∂u2 ⁄ ∂ũ2)
These equations, which correlate the surface basis vectors in
the barred and unbarred surface coordinate systems, can be
compactly presented in tensor notation as:
(180) ∂xi ⁄ ∂ũα = (∂xi ⁄ ∂uβ)(∂uβ ⁄ ∂ũα)
where i = 1, 2, 3 and α, β = 1, 2.
A set of contravariant basis vectors for the surface may also be
defined as the gradient of the surface coordinate curves, that is:
∇
(181) Eα = uα
In tensor notation, this basis may be given by:
(182) [Eα]i ≡ Eαi = ∂uα ⁄ ∂xi = xαi
Hence, these basis vectors can be regarded as 2D contravariant
surface vectors or as 3D covariant space vectors.
The contravariant and covariant forms of the surface basis
vectors, Eα and Eα, are obtained from each other by the index
shifting operator for the surface, that is:
(183)
Eα = aαβEβ
Eα = aαβEβ
where the indexed a are the covariant and contravariant forms
of the surface metric tensor (see § 3.3↓). The contravariant and
covariant forms of the surface basis vectors, Eα and Eα, are
reciprocal systems and hence they satisfy the following
reciprocity relations:
(184)
Eα⋅Eβ = δαβ ≡ aαβ
Eα⋅Eβ = δαβ ≡ aαβ
where the indexed δ and a are the mixed type of the Kronecker
delta and surface metric tensors.
The surface basis vectors in their covariant and contravariant
forms, xiα and xβj, and the unit normal vector to the surface nk
are linked by the following relation:
(185) xiα = εijk εαβ xβj nk
This equation means that the given product (which looks like a
vector cross product) of the surface contravariant basis vector
xβj and the unit normal vector nk produces a surface covariant
basis vector xiα and hence it is perpendicular to both. Being a
surface basis vector implies orthogonality to the unit normal
vector, while being a covariant surface basis vector implies
here orthogonality to the contravariant surface basis vector.
3.3 Surface Metric Tensor
The surface metric tensor is an absolute, rank-2, 2 × 2 tensor.
As we will see, all forms of this tensor (i.e. covariant,
contravariant and mixed) are symmetric. In differential
geometry, the surface metric tensor aαβ may be called the first
groundform or the fundamental surface tensor. We remark that
the coefficients of the metric tensor are real numbers.
Following the example of the metric in general nD spaces, the
surface metric tensor of a 2D surface embedded in a 3D
Euclidean flat space with metric gij = δij is given in its
covariant form by:
(186)
aαβ =
Eα⋅Eβ =
(∂r ⁄ ∂uα)⋅(∂r ⁄ ∂uβ) =
(∂xi ⁄ ∂uα)(∂xi ⁄ ∂uβ)
where the indexed x and u are the space Cartesian coordinates
and the surface coordinates respectively, and i = 1, 2, 3 and α,
β = 1, 2.
The surface and space metric tensors in a general Riemannian
space with general metric gij are related by:
(187)
aαβ =
Eα⋅Eβ =
(∂r ⁄ ∂uα)⋅(∂r ⁄ ∂uβ) =
gij(∂xi ⁄ ∂uα)(∂xj ⁄ ∂uβ) =
gijxiαxjβ
where aαβ and gij are respectively the surface and space
covariant metric tensors, the indexed x and u are general
coordinates of the space and surface respectively, and i, j = 1,
2, 3 and α, β = 1, 2. It is obvious that Eq. 186↑ is a special
instance of Eq. 187↑ for the case of a flat space with a
Cartesian system where the space metric is the unity tensor.
Eq. 187↑ is the fundamental relation that provides the crucial
link between the surface and its enveloping space. As indicated
before, the partial derivatives in this relation, (∂xi ⁄ ∂uα) and
(∂xj ⁄ ∂uβ), may be considered as contravariant rank-1 3D space
tensors or as covariant rank-1 2D surface tensors. Hence, a
tensor like (∂xi ⁄ ∂uα) is usually labeled as xiα to indicate that it
represents two surface vectors which are contravariantly-
transformed with respect to the three space coordinates xi:
(188)
xi1 = (∂x1 ⁄ ∂u1, ∂x2 ⁄ ∂u1, ∂x3 ⁄ ∂u1)
xi2 = (∂x1 ⁄ ∂u2, ∂x2 ⁄ ∂u2, ∂x3 ⁄ ∂u2)
or three space vectors which are covariantly-transformed with
respect to the two surface coordinates uα:
(189)
x1α = (∂x1 ⁄ ∂u1, ∂x1 ⁄ ∂u2)
x2α = (∂x2 ⁄ ∂u1, ∂x2 ⁄ ∂u2)
x3α = (∂x3 ⁄ ∂u1, ∂x3 ⁄ ∂u2)
Any surface vector Aα (α = 1, 2), defined as a linear
combination of the surface basis vectors E1 and E2, can also be
considered as a space vector Ai (i = 1, 2, 3) where the two
representations are linked through the relation:
(190) Ai = (∂xi ⁄ ∂uα)Aα = xiαAα
where i = 1, 2, 3 and α = 1, 2. Now, since we have:
(191) aαβ Aα Aβ = gij xiα xjβ Aα Aβ
(Eq. 187↑)
= gij xiα Aα xjβ Aβ
= gij Ai Aj
(Eq. 190↑)
then we see that the two representations are equivalent, that is
they define a vector of the same magnitude and direction.
The contravariant form of the surface metric tensor is defined
as the inverse of the surface covariant metric tensor, that is:
(192)
aαγ aγβ = δαβ
aαγ aγβ = δαβ
Since the first fundamental form is positive definite (see §
3.5↓), and hence a > 0, the existence of an inverse is
guaranteed. Similar to the metric tensor in general nD spaces,
the covariant and contravariant forms of the surface metric
tensor, aαβ and aαβ, are used for lowering and raising indices
and related tensor operations.
The covariant type of the surface metric tensor aαβ is given in
matrix form by:
(193)
where the coefficients aαβ are as defined above (see Eq. 187↑),
and E, F, G are the coefficients of the first fundamental form
(refer to § 3.5↓). Because the contravariant form of the surface
metric tensor aαβ is the inverse of its covariant form, it is given
by:
(194)
where the symbols are as defined previously. As seen in the
above equation, this tensor is symmetric and hence a12 = a21.
As indicated before, the mixed form of the surface metric
tensor aαβ is the identity tensor, that is:
(195)

Following the style of the space metric tensor, the surface

metric tensor transforms between the barred and unbarred
surface coordinate systems as:
(196) ãαβ = aγδ(∂uγ ⁄ ∂ũα)(∂uδ ⁄ ∂ũβ)
where the indexed ã and a are the surface covariant metric
tensors in the barred and unbarred systems respectively. The
contravariant and mixed forms of the surface metric tensor also
follow similar transformation rules to their counterparts of the
space metric tensor.
Similar to the determinants of the space metric, the
determinants of the surface metric in the barred and unbarred
coordinate systems are linked through the Jacobian of
transformation by the following relation:
(197) ã = J2a
where ã and a are the determinants of the covariant form of the
surface metric tensor in the barred and unbarred systems
respectively and J ( = |∂u ⁄ ∂ũ|) is the Jacobian of the
transformation between the two surface systems (refer to §
1.4.3↑). This relation can be obtained directly from Eq. 196↑
by taking the determinant of the two sides of that equation.
The Christoffel symbols of the first kind [αβ, γ] for the surface
are linked to the surface basis vectors and their partial
derivatives by the following relation:
(198) [αβ, γ] = (∂Eα ⁄ ∂uβ)⋅Eγ
Hence, the relation between the partial derivative of the surface
covariant metric tensor and the Christoffel symbols of the first
kind is given by:
(199)
∂aαβ ⁄ ∂uγ =
∂(Eα⋅Eβ) ⁄ ∂uγ =
(∂Eα ⁄ ∂uγ)⋅Eβ + Eα⋅(∂Eβ ⁄ ∂uγ) =
[αγ, β] + [βγ, α]
A similar relation between the derivative of the surface metric
tensor and the Christoffel symbols of the second kind can be
obtained from the previous equation by using the index shifting
operator of the surface:
(200) ∂aαβ ⁄ ∂uγ = aδβΓδαγ + aδαΓδβγ
For a Monge patch of the form r(u, v) = (u, v, f(u, v)), the
surface covariant metric tensor a is given by:
(201)
where the subscripts u and v stand for partial derivatives of f
with respect to these surface coordinates, and IS is the tensor of
the first fundamental form of the surface (see § 3.5↓). It is
noteworthy that scaling a surface up or down by a constant
factor c > 0, which is equivalent to scaling all the distances on
the surface by that factor, can be done by multiplying the
surface metric tensor by c2. This can be seen, for example,
from Eq. 187↑.
In the following subsections, we investigate arc length, area
and angle between two curves on a surface. All these entities
depend in their definition and quantification on the metric
tensor. We will see that all these entities, due to their exclusive
dependence on the metric tensor, are invariant under isometric
transformations. We will also see that their geometric and
tensor formulations are identical to those in a general nD space
with the use of the surface metric tensor and the surface
representation of the involved quantities.

3.3.1 Arc Length

The length of an infinitesimal line element of a surface curve

represents the resultant growth along the element in the u and v
directions between its two end points (see Fig. 31↓). Following
the example of the length of an element of arc of a curve
embedded in a general nD space, the length of an element of
arc of a curve on a 2D surface, ds, is given in its general form
by (see Eqs. 167↑ and 187↑):
(202)
(ds)2 =
dr⋅dr =
(∂r ⁄ ∂uα)⋅(∂r ⁄ ∂uβ) duαduβ =
Eα⋅Eβ duαduβ =
aαβ duαduβ
where aαβ is the covariant type of the surface metric tensor, r is
the spatial representation of the surface curve and α, β = 1, 2.
Figure 31 The length of an infinitesimal line element, ds, of a
surface curve C where du and dv are used to label infinitesimal
element growths on the coordinate curves (CC).

From the above formula we have the following identity which

is valid at each point of a naturally parameterized surface
curve:
(203) aαβ(duα ⁄ ds)(duβ ⁄ ds) = 1
Based on the above formula (Eq. 202↑), the length of a
segment of a t-parameterized curve between a starting point
corresponding to t = t1 and a terminal point corresponding to t
= t2 is given by:
(204)
 ⊂
where I ℝ is an interval on the real line and E, F, G are the
coefficients of the first fundamental form. The fourth equality
is based on the equality: a12 = a21 because the metric tensor is
symmetric, as seen earlier.
For a Monge patch of the form r(u, v) = (u, v, f(u, v)), the
length of an element of arc of a surface curve is given by:
(205) ds = √([1 + fu2]dudu + 2fufvdudv + [1 + fv2]dvdv)
where the subscripts u and v stand for partial derivatives of f
with respect to these surface coordinates. The last equation can
be obtained directly from Eq. 202↑ using the coefficients of the
metric tensor from Eq. 201↑.
It is noteworthy that the length of a surface curve is an intrinsic
property since it depends on the metric tensor only. Also, the
length of a surface curve is invariant with respect to the type of
parameterization and isometric transformations.

3.3.2 Surface Area

The area of an infinitesimal surface element of a space surface

represents the resultant growth in the element as a result of the
growth in the u and v directions along the boundary curves that
define the element (see Fig. 32↓).
Figure 32 The area of an infinitesimal surface element, dσ, of
a space surface where du and dv are used to label infinitesimal
element growths on the coordinate curves and “CC” stands for
“coordinate curve”.

The area of an infinitesimal element of a surface, dσ, in the

neighborhood of a point P on the surface is given by (see Eqs.
170↑ and 187↑):
(206)
dσ =
|dr1 × dr2| =
|E1 × E2|du1du2 =
√(|E1|2| E2|2 − [ E1⋅ E2]2) du1du2 =
√(a11a22 − [a12]2) du1du2 =
√(EG − F2) du1du2 =
√(a) du1du2
where E1 and E2 are the surface covariant basis vectors, E, F,
G are the coefficients of the first fundamental form, a ( =
a11a22 − a12a21) is the determinant of the surface covariant
metric tensor and the indexed a are its elements. All the
quantities in these expressions belong to the point P. We also
assume that du1du2 is positive.
The area of a surface patch S:Ω → ℝ3, where Ω is a proper
subset of the ℝ2 parameters plane, is given by:
(207)
σ=
∬ Ω dσ =
∬ Ω √(a11a22 − [a12]2) du1du2 =
∬ Ω √(EG − F2) du1du2 =
∬ Ω √(a) du1du2
where the symbols are as defined above. The patch S should be
injective, sufficiently differentiable, and regular on the interior
of Ω. The above formulae for the area are reminder of the
volume formulae in a 3D space, so the area can be regraded as
a volume in a 2D space.
As stated before, the areas of two corresponding surface
elements and surface patches on two isometric surfaces are
equal. This can be seen from the above formulae since the
metric tensor is identical at the corresponding points of two
isometric surfaces. For a Monge patch of the form r(u, v) = (u,
v, f(u, v)), the surface area is given by:
∬
(208) σ = Ω √(1 + fu2 + fv2) dudv
where the subscripts u and v stand for partial derivatives of f
with respect to these surface coordinates. The last equation can
be obtained directly from Eq. 207↑ using the coefficients of the
metric tensor from Eq. 201↑.

3.3.3 Angle Between Two Surface Curves

The angle between two sufficiently smooth surface curves

intersecting at a given point on the surface is defined as the
angle between their tangent vectors at that point (Fig. 33↓). As
there are two opposite directions for each curve, corresponding
to the two senses of traversing the curve, there are two main
angles θ1 and θ2 such that θ1 + θ2 = π. The principal angle
between the two curves is usually taken as the smaller of the
two angles and hence the directions are determined
accordingly. In fact, there are still two possibilities for the
directions but this has no significance as far as the angle
between the two curves is concerned.
Figure 33 The angle between two surface curves, C1 and C2,
as the angle θ between their tangents, t1 and t2, at the point of
intersection P.

The angle θ between two surface curves passing through a

given point P on the surface with tangent vectors A and B at P
is given by:
(209)
cosθ =
(A⋅B) ⁄ (|A||B|) =
(aαβAαBβ) ⁄ [√(aγδAγAδ) √(aϵζBϵBζ)]
where the indexed a are the elements of the surface covariant
metric tensor and the Greek indices run over 1, 2. If A and B
are two unit surface vectors with surface representations Aα
and Bβ and space representations Ai and Bj then the angle θ
between the two vectors is given by:
(210) cosθ = aαβ Aα Bβ
(Eq. 209↑)
= gij xiα xjβ Aα Bβ
(Eq. 187↑)
= gij Ai Bj
(Eq. 190↑)
where α, β = 1, 2 and i, j = 1, 2, 3. Hence, the surface and
space representations of the two vectors define the same angle.
The vectors A and B (whether unit vectors or not) are
orthogonal iff aαβAαBβ = 0 at P.
As seen earlier, the coordinate curves at a given point P on a
surface are orthogonal iff a12 ≡ F = 0 at P. This can be seen
from Eqs. 187↑ and 235↓ where the dot products will vanish
due to the orthogonality of E1 and E2 which are the tangents to
the coordinate curves. The corresponding angles of two
isometric surfaces, like the corresponding lengths and areas,
are equal as can be seen from the above formulae considering
that the metric tensor is identical at the corresponding points.
However, the reverse is not true in general, that is the equality
of angles on two surfaces related by a given mapping, as in the
conformal mapping, does not lead to the equality of the
corresponding lengths on the two mapped surfaces, as
explained before. This is due to the fact that the equality of
angles requires the proportionality of the two metric tensors at
the corresponding points but not necessarily the equality (see
Eq. 159↑). This may be concluded from Eq. 209↑ where any
proportionality factor will be canceled out.
The sine of the angle θ between two surface unit vectors, A
and B, is given by:
(211) sinθ = εαβAαBβ
where εαβ is the 2D absolute permutation tensor. The sine in
this formula is numerically equal in magnitude to the area of
the parallelogram with sides A and B. Based on the last
formula, the sufficient and necessary condition for A and B to
be orthogonal is that:
(212) |εαβAαBβ| = 1
3.4 Surface Curvature Tensor
The surface curvature tensor is an absolute, rank-2, 2 × 2
tensor. As we will see, the covariant and contravariant forms of
this tensor are symmetric. The surface curvature tensor bαβ
may also be called the second groundform. We note that the
coefficients of the curvature tensor are real numbers. The
elements of the surface covariant curvature tensor, bαβ, are
given by:
(213)
bαβ =
− (∂r ⁄ ∂uα)⋅(∂n ⁄ ∂uβ) =
− Eα⋅(∂n ⁄ ∂uβ) =
− [(∂r ⁄ ∂uα)⋅(∂n ⁄ ∂uβ) + (∂r ⁄ ∂uβ)⋅(∂n ⁄ ∂uα)] ⁄ 2
and also by:
(214)
bαβ =
(∂2 r ⁄ ∂uα∂uβ)⋅n =
(∂Eα ⁄ ∂uβ)⋅n =
[∂Eα ⁄ ∂uβ]⋅[(E1 × E2) ⁄ √(a)] =
[(∂Eα ⁄ ∂uβ)⋅(E1 × E2)] ⁄ √(a)
where Eq. 168↑ is used in the last two steps. We note that the
equality:
(215) (∂Eα ⁄ ∂uβ)⋅n = − Eα⋅(∂n ⁄ ∂uβ)
which is seen by comparing Eqs. 213↑ and 214↑ is based on
the equality:
(216) ∂(Eα⋅n) ⁄ ∂uβ = ∂(0) ⁄ ∂uβ = 0
plus the product rule of differentiation. Considering Eq. 214↑,
the symmetry of the surface covariant curvature tensor (i.e. b12
= b21) follows from the fact that:
(217)
∂Eα ⁄ ∂uβ =
∂2 r ⁄ ∂uβ∂uα =
∂2 r ⁄ ∂uα∂uβ =
∂Eβ ⁄ ∂uα
In full tensor notation, the surface covariant curvature tensor is
given by:
(218)
bαβ =
(εγδ εijk xiα;β xjγ xkδ) ⁄ 2 =
(εijk xiα;β xj1 xk2) ⁄ √(a)
where a is the determinant of the surface covariant metric
tensor and the indexed ε’s are the absolute permutation tensors
of 2D and 3D spaces in their contravariant and covariant
forms. This formula will simplify to the following when the
space coordinates are rectangular Cartesian:
(219) bαβ = [ϵijk(∂2xi ⁄ ∂uα∂uβ)xj1xk2] ⁄ √(a)
The surface curvature tensor obeys the same transformation
rules as the surface metric tensor. For example, for the
transformation between the barred and unbarred surface
coordinate systems we have (see Eq. 196↑):
(220) b̃αβ = bγδ(∂uγ ⁄ ∂ũα)(∂uδ ⁄ ∂ũβ)
where the indexed b̃ and b are the coefficients of the covariant
curvature tensor in the barred and unbarred systems
respectively. Similarly, we have (see Eq. 197↑):
(221) b̃ = J2b

̃
where b̃ and b are the determinants of the surface covariant
curvature tensor in the barred and unbarred systems
respectively and J ( = |∂u ⁄ ∂ũ|) is the Jacobian of the
transformation between the two surface coordinate systems.
The surface covariant curvature tensor is given in matrix form
by:
(222)
where e, f, g are the coefficients of the second fundamental
form of the surface (refer to § 3.6↓). This is a symmetric
matrix as indicated earlier and as seen in the last part of the
equation.
The mixed form of the surface curvature tensor bα β is given
by:
(223)
where E, F, G, e, f, g are the coefficients of the first and second
fundamental forms and a = EG − F2 is the determinant of the
surface covariant metric tensor. As seen, the coefficients of bαβ
depend on the coefficients of both the first and second
fundamental forms. The mixed form bαβ = bαγaγβ is the
transpose of the above form.
The contravariant form of the surface curvature tensor is given
by:
(224)
where
A = eG2 − 2fFG + gF2
B = fEG − eFG − gEF + fF2
C = fEG − eFG − gEF + fF2
D = gE2 − 2fEF + eF2.
Like the covariant form, the contravariant form is a symmetric
tensor as indicated before and as seen above.
As we will see, the trace of the surface mixed curvature tensor
bαβ is twice the mean curvature H (see § 4.6↓), while its
determinant is the Gaussian curvature K (see § 4.5↓), that is:
(225)
H = tr(bαβ) ⁄ 2
K = det(bαβ)
Because the trace and the determinant of a tensor are its main
two invariants under permissible coordinate transformations
(refer to similarity transformations of linear algebra), then H
and K are invariant, as will be established later in the book.
The surface covariant curvature tensor of a Monge patch of the
form r(u, v) = (u, v, f(u, v)) is given by:
(226)
where the subscripts u and v stand for partial derivatives of f
with respect to these surface coordinates. Since fuv = fvu, the
tensor is symmetric as it should be. The equality: fuv = fvu is
based on the well known continuity condition which is fully
explained in any standard textbook on calculus.
The covariant curvature tensor of a surface and the Riemann-
Christoffel curvature tensor of the first kind are linked by the
following relation:
(227) Rαβγδ = bαγbβδ − bαδbβγ
This relation may also be given in terms of the Riemann-
Christoffel curvature tensor of the second kind using the index
raising operator for the surface, that is:
(228) aαωRωβγδ = Rαβγδ = bαγbβδ − bαδbβγ
From Eqs. 88↑ and 227↑, it can be concluded that the surface
covariant curvature tensor and the surface Christoffel symbols
of the first and second kind are related by:
(229) bαγbβδ − bαδbβγ = (∂[βδ, α] ⁄ ∂γ) − (∂[βγ, α] ⁄ ∂δ) + [αδ,
ω]Γωβγ − [αγ, ω]Γωβδ
Similarly, from Eqs. 89↑ and 228↑, it can be concluded that the
surface curvature tensor and the surface Christoffel symbols of
the second kind are related by:
(230) bαγbβδ − bαδbβγ = (∂Γαβδ ⁄ ∂γ) − (∂Γαβγ ⁄ ∂δ) + ΓωβδΓαωγ −
ΓωβγΓαωδ
Considering the fact that the 2D Riemann-Christoffel curvature
tensor has only one degree of freedom (see § 1.4.10↑) and
hence it possesses a single independent non-vanishing
component which is represented by R1212, we see that Eq. 227↑
has only one independent component which is given by:
(231) R1212 = b11b22 − b12b21 = b
where b is the determinant of the surface covariant curvature
tensor.
Equation 227↑ also shows that each one of the following
provisions:
(232)
Rαβγδ = 0
and
bαβ = 0
(α, β, γ, δ = 1, 2) if satisfied identically is a sufficient and
necessary condition for having a flat 2D space. However, for a
plane surface all the coefficients of the Riemann-Christoffel
curvature tensor and the coefficients of the surface curvature
tensor vanish identically throughout the surface, but for a
surface which is isometric to plane the coefficients of the
Riemann-Christoffel curvature tensor vanish identically but not
necessarily the coefficients of the surface curvature tensor.
This is based on the fact that having a zero determinant does
not imply having a zero tensor (see Eq. 231↑). So, as stated
previously, the provision Rαβγδ = 0 is a sufficient and necessary
condition for having an intrinsically flat space, while the
provision bαβ = 0 is a sufficient and necessary condition for
having an extrinsically flat space. Having an extrinsically flat
space implies having an intrinsically flat space since a
curvature that cannot be observed from outside the space
cannot be seen from inside the space.
From Eqs. 88↑ and 227↑, we see that the Riemann-Christoffel
curvature tensor can be expressed in terms of the coefficients
of the surface curvature tensor as well as in terms of the
coefficients of the surface metric tensor where the two sets of
coefficients are linked through Eq. 229↑. This does not mean
that Riemann-Christoffel curvature is an extrinsic property but
it means that some intrinsic properties can also be defined in
terms of extrinsic parameters. It should be noticed that the sign
of the surface curvature tensor (i.e. the sign of its coefficients)
is dependent on the choice of the direction of the unit normal
vector to the surface, n, and hence the sign of the coefficients
will be reversed if the direction of n is reversed. This can be
seen, for example, from Eq. 214↑.
3.5 First Fundamental Form
As indicated previously, the first fundamental form, which is
based on the metric, encompasses all the intrinsic information
about the surface that a 2D inhabitant of the surface can obtain
from measurements performed on the surface without
appealing to an external dimension. In the old books, the first
fundamental form may be labeled as the first fundamental
quadratic form.
The first fundamental form IS of the length of an element of arc
of a curve on a surface is a quadratic expression given by:
(233)
IS =
(ds)2 =
dr⋅dr =
(∂r ⁄ ∂uα)⋅(∂r ⁄ ∂uβ)duαduβ =
Eα⋅Eβ duαduβ =
aαβ duαduβ =
E(du1)2 + 2F du1du2 + G(du2)2
where E, F, G, which in general are continuous variable
functions of the surface coordinates u1 and u2, are given by:
(234)
E=
a11 =
E1⋅ E1 =
(∂r ⁄ ∂u1)⋅(∂ r ⁄ ∂u1) =
gij(∂xi ⁄ ∂u1)(∂xj ⁄ ∂u1)
(235)
F=
a12 =
E1⋅ E2 =
(∂r ⁄ ∂u1)⋅(∂ r ⁄ ∂u2) =
gij(∂xi ⁄ ∂u1)(∂xj ⁄ ∂u2) =
E2⋅ E1 = a21
(236)
G=
a22 =
E2⋅ E2 =
(∂r ⁄ ∂u2)⋅(∂ r ⁄ ∂u2) =
gij(∂xi ⁄ ∂u2)(∂xj ⁄ ∂u2)
where the indexed a are the elements of the surface covariant
metric tensor, the indexed x are the general coordinates of the
enveloping space and gij is its covariant metric tensor.
For a flat space with a Cartesian coordinate system xi, the
space metric is gij = δij and hence the above equations become:
(237) E = (∂xi ⁄ ∂u1)(∂xi ⁄ ∂u1)
(238) F = (∂xi ⁄ ∂u1)(∂xi ⁄ ∂u2)
(239) G = (∂xi ⁄ ∂u2)(∂xi ⁄ ∂u2)
The first fundamental form can be cast in the following matrix
form:
(240)
where v is a direction vector, vT is its transpose, and IS is the
first fundamental form tensor which is equal to the surface
covariant metric tensor. Hence, the matrix associated with the
first fundamental form is the covariant metric tensor of the
surface. We remark that the notation regarding the dot product
operation between two matrices in the first line of Eq. 240↑
may not be a standard one. We also note that while the
coefficients of the first fundamental form depend on the
position on the surface and hence they are functions of the
surface coordinates but not the direction, the first fundamental
form is a function of both position and direction.
The first fundamental form is not a unique characteristic of the
surface and hence two geometrically different surfaces as seen
from the enveloping space, such as plane and cylinder, can
have the same first fundamental form. Such surfaces are
different extrinsically as seen from the embedding space
although they are identical intrinsically as viewed internally
from the surface by a 2D inhabitant.
The first fundamental form is positive definite at regular points
of 2D surfaces; hence its coefficients are subject to the
following conditions:
(241)
E>0
and
det(IS) = EG − F2 > 0
As indicated earlier, the above conditions imply G > 0 since
these coefficients are real. It is noteworthy that the condition of
positive definiteness may be amended to allow for metrics
based on coordinate systems with imaginary coordinates as it is
the case in the coordinate systems of the Minkowski space (see
§ 1.4.7↑), but this is out of the scope of the present book whose
focus is the geometry of space curves and surfaces in a static
sense and hence all metrics considered in this book are positive
definite.
As indicated above, the first fundamental form encompasses
the intrinsic properties of the surface geometry. Hence, as seen
in § 3.3↑, the first fundamental form is used to define and
quantify things like arc length, area and angle between curves
on a surface based on its qualification as a metric. For
example, Eq. 204↑ shows that the length of a curve segment on
a surface is obtained by integrating the square root of the first
fundamental form of the surface along the segment.
If a surface S1 can be mapped isometrically (see § 3.1↑ and §
6.5↓) onto another surface S2, then the two surfaces have
identical first fundamental forms and identical first
fundamental form coefficients at their corresponding points,
that is (see Footnote 13 in § 8↓):
(242)
E1 = E2
F1 = F2
G1 = G2
where the subscripts are labels for the two surfaces. Based on
Eq. 201↑, for a Monge patch of the form r(u, v) = (u, v, f(u, v)),
the first fundamental form is given by:
(243) IS = (1 + fu2)dudu + 2fufvdudv + (1 + fv2)dvdv
where the subscripts u and v stand for partial derivatives of f
with respect to these surface coordinates.
3.6 Second Fundamental Form
The mathematical entity that characterizes the extrinsic
geometry of a surface is the normal vector to the surface. This
entity can only be observed externally from outside the surface
by an observer in a reference frame in the space that envelops
the surface. Hence, the normal vector and all its subsidiaries
are strange to a 2D inhabitant to the surface who can only
access the intrinsic attributes of the surface as represented by
and contained in the first fundamental form. In brief, the 2D
inhabitant has no notion of the extra dimension which
embraces the normal vector. As a consequence, the variation of
the normal vector as it moves around the surface can be used
as an indicator to characterize the surface shape from an
external point of view and that is how this indicator is
exploited in the second fundamental form to represent the
extrinsic geometry of the surface as will be seen from the
forthcoming formulations of the second fundamental form.
The second fundamental form IIS of a surface, which in the old
books may be labeled as the second fundamental quadratic
form, is defined by the following quadratic expression:
(244)
IIS =
− dr⋅dn =
− [(∂r ⁄ ∂uα)duα]⋅[(∂n ⁄ ∂uβ)duβ] =
− [(∂r ⁄ ∂u1)du1 + (∂ r ⁄ ∂u2)du2]⋅[(∂ n ⁄ ∂u1)du1 + (∂ n ⁄
∂u2)du2] =
e(du1)2 + 2f du1du2 + g(du2)2
where the coefficients of the second fundamental form e, f, g
are given by:
(245)
e=
− (∂r ⁄ ∂u1)⋅(∂ n ⁄ ∂u1) =
− E1⋅(∂ n ⁄ ∂u1)
(246)
f=
− [(∂r ⁄ ∂u1)⋅(∂ n ⁄ ∂u2) + (∂ r ⁄ ∂u2)⋅(∂ n ⁄ ∂u1)] ⁄ 2 =
− [E1⋅(∂ n ⁄ ∂u2) + E2⋅(∂ n ⁄ ∂u1)] ⁄ 2
(247)
g=
− (∂r ⁄ ∂u2)⋅(∂ n ⁄ ∂u2) =
− E2⋅(∂ n ⁄ ∂u2)
In the above equations, r(u1, u2) denotes the spatial
representation of the surface, n(u1, u2) is the unit normal vector
to the surface and α, β = 1, 2.
The second fundamental form may also be given by:
(248)
where d2 r is the second order differential of the position
vector r of an arbitrary point on the surface in the direction
(du1, du2). As a result, the coefficients of the second
fundamental form can also be given by the following
alternative expressions:
(249)
e=
n⋅(∂E1 ⁄ ∂u1) =
− (∂n ⁄ ∂u1)⋅ E1
(250)
f=
n⋅(∂E1 ⁄ ∂u2) =
n⋅(∂E2 ⁄ ∂u1) =
− (∂n ⁄ ∂u2)⋅ E1 =
− (∂n ⁄ ∂u1)⋅ E2
(251)
g=
n⋅(∂E2 ⁄ ∂u2) =
− (∂n ⁄ ∂u2)⋅ E2
The two forms of each formula in the last equations are based
on the fact that n⋅Eα = 0, since n is perpendicular to the surface
basis vectors, plus the product rule of differentiation, that is:
(252) ∂(n⋅Eα) ⁄ ∂uβ = ∂(0) ⁄ ∂uβ = n⋅(∂Eα ⁄ ∂uβ) + (∂n ⁄ ∂uβ)⋅Eα
=0 ⇒ n⋅(∂Eα ⁄ ∂uβ) = − (∂n ⁄ ∂uβ)⋅Eα
The equality:
n⋅(∂2 E1) = n⋅(∂1 E2)
which is seen in Eq. 250↑ is based on the equality:
∂αEβ = ∂βEα
as stated before (see Eq. 217↑).
From Eqs. 249↑-251↑ plus Eq. 168↑, we can see that the
coefficients of the second fundamental form may also be given
by the following expressions:
(253) e = [(E1 × E2)⋅(∂ E1 ⁄ ∂u1)] ⁄ √(a)
(254) f = [(E1 × E2)⋅(∂ E1 ⁄ ∂u2)] ⁄ √(a)
(255) g = [(E1 × E2)⋅(∂ E2 ⁄ ∂u2)] ⁄ √(a)
where a ( = a11a22 − a12a21 = EG − F2) is the determinant of
the surface covariant metric tensor. Like the coefficients of the
first fundamental form, the coefficients of the second
fundamental form are, in general, continuous variable
functions of the surface coordinates u1 and u2.
The second fundamental form can also be cast in the following
matrix form:
(256)
where v is a direction vector, vT is its transpose, and IIS is the
second fundamental form tensor. As indicated before, the
notation regarding the dot product operation may not be
standard in the commonly employed matrix notation. Also,
although the coefficients of the second fundamental form
depend on the position but not the direction, the second
fundamental form depends on both.
As seen before, the coefficients of the second fundamental
form satisfy the following relations:
(257)
e = b11
f = b12 = b21
g = b22
Hence, the second fundamental form tensor IIS is the same as
the surface covariant curvature tensor b, that is:
(258)
As a consequence of the previous statements, we see that the
second fundamental form can also be given in a compact
tensor notation form by:
(259) IIS = bαβ duα duβ
where the indexed b are the elements of the surface covariant
curvature tensor and α, β = 1, 2.
The second fundamental form of the surface at a given point
and in a given direction can also be expressed in terms of the
first fundamental form IS and the normal curvature κn of the
surface at that point and in that direction (see § 4.2↓) as:
(260) IIS = κnIS = κn(ds)2
Based on Eqs. 259↑ and 226↑, for a Monge patch of the form
r(u, v) = (u, v, f(u, v)), the second fundamental form is given
by:
(261) IIS = (fuududu + 2fuvdudv + fvvdvdv) ⁄ √(1 + fu2 + fv2)
where the subscripts u and v stand for partial derivatives of f
with respect to these surface coordinates.
While the first fundamental form is invariant under isometric
transformations, the second fundamental form is invariant
under rigid motion transformations. The second fundamental
form is also invariant (considering spatially-fixed directions)
under permissible surface re-parameterizations that maintain
the sense of the normal vector to the surface, n. However, the
second fundamental form changes its sign if the sense of n is
reversed. As indicated earlier, while the first fundamental form
encompasses the intrinsic geometry of the surface, the second
fundamental form encompasses its extrinsic geometry.
Consequently, the first fundamental form is associated with the
surface covariant metric tensor, while the second fundamental
form is associated with the surface covariant curvature tensor.
A major difference between the two fundamental forms is that
while the first fundamental form is positive definite, as stated
previously, the second fundamental form is not necessarily
positive or definite.
Unlike space curves which are completely defined by specified
curvature and torsion, κ and τ, providing arbitrary first and
second fundamental forms is not a sufficient condition for the
existence and determination of a surface with these forms. This
is due to the fact that the first and second fundamental forms
when independently defined do not provide acceptable
determination for the surface unless they satisfy extra
compatibility conditions to ensure the consistency of these
forms and secure the existence of a surface associated with
these forms. In more technical terms, defining six functions E,
F, G, e, f, g of class C3 on a subset of ℝ2 where these functions
satisfy the conditions for the coefficients of the first and
second fundamental forms (in particular E, G > 0 and EG − F2
> 0) does not guarantee the existence of a surface over the
given subset with a first fundamental form:
E(du1)2 + 2Fdu1du2 + G(du2)2
and a second fundamental form:
e(du1)2 + 2fdu1du2 + g(du2)2.
Further compatibility conditions relating the first and second
fundamental forms are required to fully identify the surface
and secure its existence.
The above difference between curves and surfaces may be
linked to the fact that the conditions for the curves are
established based on the existence theorem for ordinary
differential equations where these equations generally have a
solution, while the conditions for the surfaces are established
based on the existence theorem for partial differential
equations which have solutions only when they meet additional
integrability conditions. The details should be sought in more
extensive books on differential geometry.
Following the statements in the last paragraphs, the required
compatibility conditions for the existence of a surface with
predefined first and second fundamental forms are given by the
Codazzi-Mainardi equations (Eqs. 294↓-295↓) plus the
following equation:
(262) eg − f2 = F(∂uΓ222 − ∂vΓ212 + Γ122Γ211 − Γ112Γ212) +
E(∂uΓ122 − ∂vΓ112 + Γ122Γ111 + Γ222Γ112 − Γ112Γ112 − Γ212Γ122)
where all the symbols are as defined previously and the
Christoffel symbols belong to the surface.
From the above statements, the fundamental theorem of space
surfaces in differential geometry, which is the equivalent of the
fundamental theorem of curves (see § 2.3↑), emerges. The
theorem states that: given six sufficiently smooth functions E,
F, G, e, f, g on a subset of ℝ2 satisfying the following
conditions:
1. E, G > 0 and EG − F2 > 0, and
2. E, F, G, e, f, g satisfy Eqs. 262↑ and 294↓-295↓,
then there is a unique surface with E, F, G as its first
fundamental form coefficients and e, f, g as its second
fundamental form coefficients. Hence, if two surfaces meet all
these conditions, then they are identical within a rigid motion
transformation in space. As a result, the fundamental theorem
of surfaces, like the fundamental theorem of curves, provides
the existence and uniqueness conditions for surfaces.
On the other hand, according to one of the theorems of Bonnet,
if there are two surfaces of class C3, S1:Ω → ℝ3 and S2:Ω →
⊆
ℝ3, which are defined over a connected set Ω ℝ2 and they
have identical first and second fundamental forms, then the
two surfaces can be mapped on each other by a purely rigid
motion transformation. The existence of these surfaces
guarantees the compatibility of their first and second
fundamental forms and hence they are identical within a rigid
motion transformation according to the fundamental theorem
of surfaces.
It is noteworthy that two surfaces having identical first
fundamental forms but different second fundamental forms
may be described as applicable. An example of applicable
surfaces is plane and cylinder. Although all applicable
surfaces, according to this definition, are isometric since they
have identical first fundamental forms, not all isometric
surfaces are applicable since two isometric surfaces may also
have identical second fundamental forms as it is the case of
two planes.
It should be remarked that some authors use L, M, N instead of
e, f, g to symbolize the coefficients of the second fundamental
form. However, the use of e, f, g is advantageous since they
correspond nicely to the coefficients of the first fundamental
form E, F, G making the formulae involving the first and
second fundamental forms more symmetric and memorable.
On the other hand, the use of L, M, N is also advantageous
when reciting formulae especially if the formulae contain the
coefficients of both fundamental forms; moreover, it is less
susceptible to errors in the writing and typing of these
formulae.
Another remark is that the coefficient g of the second
fundamental form should not be confused with the symbol g of
the determinant of the space covariant metric tensor which is
commonly used in the literature of tensor calculus and
differential geometry but we do not use it in the present book.
The coefficient f of the second fundamental form should also
be distinguished easily from the symbol f which is widely used
in the mathematical literature to symbolize mathematical
functions and hence we kept its use in this book for the sake of
readability while making some effort to avoid potential
confusion.

3.6.1 Dupin Indicatrix

On displacing the tangent plane of a surface S at a given point

P infinitesimally toward the surface in the orientation of the
normal vector of S at P, the intersection curves of S with the
displaced tangent plane will take a particular and distinctive
shape depending on the local shape of S in the neighborhood of
P (refer to Figs. 34↓, 35↓ and 36↓). This idea is the base for
characterizing and quantifying the local shape of a surface at a
particular point using the concept of Dupin indicatrix (see
Footnote 14 in § 8↓).
Dupin indicatrix at a given point on a sufficiently smooth
surface is a function of the coefficients of the second
fundamental form at the point and hence it is a function of the
surface coordinates. Dupin indicatrix is an indicator of the
departure of the surface from the tangent plane in the close
proximity of the point of tangency. Accordingly, the second
fundamental form coefficients are used in Dupin indicatrix to
measure this departure. In quantitative terms, Dupin indicatrix
is the family of conic sections given by the following quadratic
equation:
(263) eu2 + 2fuv + gv2 = ±1
where e, f, g are the coefficients of the second fundamental
form at the point with the coordinates u and v. As seen, Dupin
indicatrix, which is represented by an expression similar in
form to the second fundamental form, depends on the
coefficients of the second fundamental form and the
coordinates of the particular point on the surface and hence it is
a function of position but, unlike the second fundamental form,
it does not depend on the direction.
Figure 34 Contour curves on a typical surface in the
neighborhood of an elliptic point.
Figure 35 Contour curves on a typical surface in the
neighborhood of a parabolic point.
Figure 36 Contour curves on a typical surface in the
neighborhood of a hyperbolic point.

As a consequence of the previous statements, Dupin indicatrix

can be used to classify the surface points with respect to the
local shape of the surface as elliptic, parabolic, hyperbolic or
flat (see § 4.9↓). At an elliptic point the Dupin indicatrix is an
ellipse or circle (see Footnote 15 in § 8↓), at a parabolic point
the Dupin indicatrix becomes two parallel lines, while at a
hyperbolic point the Dupin indicatrix becomes two conjugate
hyperbolas (see Fig. 37↓). At a flat point the Dupin indicatrix is
not defined. More details about Dupin indicatrix and the local
shape of surface will be given in § 4.9↓. It is noteworthy that in
all cases where the Dupin indicatrix is defined, the principal
directions (see § 4.4↓) at the point coincide with the two
perpendicular axes of the indicatrix as seen in Fig. 37↓.
Figure 37 Dupin indicatrix at (a) elliptic point, (b) parabolic
point and (c) hyperbolic point. The principal directions (refer
to § 4.4↓) are labeled as d1 and d2.
3.7 Third Fundamental Form
The third fundamental form IIIS of a space surface is defined
by:
(264) IIIS = dn⋅dn = cαβduαduβ
where n is the unit normal vector to the surface at a given point
P, cαβ are the coefficients of the third fundamental form at P
and α, β = 1, 2.
The coefficients of the third fundamental form are given in full
tensor notation by:
(265) cαβ = gij ni, α nj, β
where gij is the space covariant metric tensor and the indexed n
is the unit normal vector to the surface. We note that these
coefficients are real numbers. The coefficients of the third
fundamental form may also be given by:
(266) cαβ = aγδ bαγ bβδ
where aγδ is the surface contravariant metric tensor and the
indexed b are the coefficients of the surface covariant
curvature tensor.
3.8 Relationship between First, Second
and Third Fundamental Forms
The first, second and third fundamental forms are linked,
through the Gaussian curvature K and the mean curvature H
(see § 4.5↓ and 4.6↓), by the following relation:
(267) K IS − 2H IIS + IIIS = 0
Accordingly, the coefficients of the first, second and third
fundamental forms are correlated, through the Gaussian
curvature and the mean curvature, by the following relation:
(268) Kaαβ − 2Hbαβ + cαβ = 0
In fact, Eq. 267↑ can be obtained from Eq. 268↑ by
multiplying the latter by duαduβ and applying the summation
convention. On multiplying both sides of Eq. 268↑ by aαβ and
shifting the indices we obtain:
(269) Kaαα − 2Hbαα + cαα = 0
that is:
(270) tr(cβα) = 4H2 − 2K
The transition form Eq. 269↑ to Eq. 270↑ is justified by the
fact that: aαα = δαα = δ11 + δ22 = 2 and H = (bαα ⁄ 2) (see Eq.
383↓).
3.9 Relationship between Surface Basis
Vectors and their Derivatives
The focus of this section is the equations of Gauss and
Weingarten which, for surfaces, are the analogue of the
equations of Frenet-Serret for curves. While the Frenet-Serret
formulae express the derivatives of T, N, B as combinations of
these vectors using κ and τ as coefficients, the equations of
Gauss and Weingarten express the derivatives of E1, E2, n as
combinations of these vectors with coefficients based on the
first and second fundamental forms.
As shown earlier (see § 2.2↑), three unit vectors can be
constructed on each point at which the curvature does not
vanish of a class C2 space curve: the tangent T, the principal
normal N and the binormal B. These mutually orthogonal
vectors (i.e. T⋅N = T⋅B = N⋅B = 0) can serve as a set of basis
vectors for the embedding 3D space. Hence, the derivatives of
these vectors with respect to the distance traversed along the
curve, s, can be expressed as combinations of this set, since
these derivatives are 3D vectors that reside in this space, as
demonstrated by the Frenet-Serret formulae (refer to § 2.5↑).
Similarly, the surface vectors:
E1 = (∂ r ⁄ ∂u1), E2 = (∂ r ⁄ ∂u2)
and the unit normal vector to the surface, n, at each regular
point on a class C2 surface form a basis set for the embedding
3D space and hence their partial derivatives with respect to the
surface coordinates, u1 and u2, can be expressed as
combinations of this set. The equations of Gauss and
Weingarten demonstrate this fact.
The equations of Gauss express the partial derivatives of the
surface vectors, E1 and E2, with respect to the surface
coordinates as combinations of the E1, E2, n basis set, that is:
(271) ∂E1 ⁄ ∂u1 = Γ111 E1 + Γ211 E2 + en
(272) ∂E1 ⁄ ∂u2 = Γ112 E1 + Γ212 E2 + fn = ∂E2 ⁄ ∂u1
(273) ∂E2 ⁄ ∂u2 = Γ122 E1 + Γ222 E2 + gn
where e, f, g are the coefficients of the second fundamental
form. These equations can be expressed compactly, with partial
use of tensor notation, as:
(274) ∂Eα ⁄ ∂uβ = ΓγαβEγ + bαβn
where α, β = 1, 2, the Christoffel symbol of the second kind
Γγαβ is based on the surface metric, as given by Eqs. 73↑-78↑,
and bαβ is the surface covariant curvature tensor. The last
equation can be expressed in full tensor notation as:
(275) xiα, β = Γγαβ xiγ + bαβ ni
where the symbols and notations are as defined previously.
These equations may also be expressed as:
(276) xiα;β = bαβ ni
where the covariant derivative notation is in use and a
rectangular Cartesian coordinate system for the space is
assumed (refer to Eq. 459↓).
Likewise, the equations of Weingarten express the partial
derivatives of the unit normal vector to the surface, n, with
respect to the surface coordinates as combinations of the
surface vectors, E1 and E2, that is:
(277) ∂n ⁄ ∂u1 = [(fF − eG) ⁄ a]E1 + [(eF − fE) ⁄ a]E2
(278) ∂n ⁄ ∂u2 = [(gF − fG) ⁄ a]E1 + [(fF − gE) ⁄ a]E2
where E, F, G, e, f, g are the coefficients of the first and second
fundamental forms and a = EG − F2 is the determinant of the
surface covariant metric tensor. We note that these expressions
are also combinations of the E1, E2, n basis set but with
vanishing normal components.
The above equations of Weingarten can be expressed
compactly, with partial use of tensor notation, as:
(279) ∂n ⁄ ∂uα = − bαβEβ
where bαβ ( = bαγaγβ) is the mixed type of the surface curvature
tensor, bαγ is the surface covariant curvature tensor and aγβ is
the surface contravariant metric tensor (refer to Eq. 223↑ and
surrounding text). They can also be expressed with full use of
tensor notation as:
(280) ni, α = − bαγ aγβ xiβ = − bαβ xiβ
To make sense of this equation, the vector ni, α is orthogonal to
ni and hence it is parallel to the tangent space of the surface, so
it can be expressed as a linear combination of the surface basis
vectors xiβ, that is:
ni, α = dαβ xiβ
for a certain set of coefficients
dαβ = − bαγaγβ,
as seen above.
The Weingarten equations may also be expressed in matrix
form as:
(281)
where IIS is the surface covariant curvature tensor and IS − 1 is
the surface contravariant metric tensor. In fact, this is the
matrix form of Eq. 279↑.
The partial derivatives of the unit normal vector to the surface,
n, with respect to the surface coordinates, u1 and u2, are linked
to the Gaussian curvature K (see § 4.5↓) and the surface basis
vectors, E1 and E2, as well as the coefficients of the first and
second fundamental forms E, F, G, e, f, g by the following
relation:
(282)
(∂n ⁄ ∂u1) × (∂ n ⁄ ∂u2) =
[(eg − f2) ⁄ (EG − F2)]( E1 × E2) =
K(E1 × E2)
where Eq. 356↓ is used in the last step.
The partial derivatives of the unit normal vector to the surface,
n, with respect to the surface coordinates, u1 and u2, are also
linked to the Gaussian and mean curvatures, K and H, and the
coefficients of the first and second fundamental forms E, F, G,
e, f, g by the following relations:
(283) (∂n ⁄ ∂u1)⋅(∂ n ⁄ ∂u1) = 2eH − EK
(284) (∂n ⁄ ∂u1)⋅(∂ n ⁄ ∂u2) = 2fH − FK
(285) (∂n ⁄ ∂u2)⋅(∂ n ⁄ ∂u2) = 2gH − GK
The above equations of Weingarten (Eqs. 277↑-278↑) can be
solved for the surface basis vectors, E1 and E2, and hence these
vectors can be expressed as combinations of the partial
derivatives of the normal vector, n, that is:
(286) E1 = [(fF − gE) ⁄ b](∂n ⁄ ∂u1) + [(fE − eF) ⁄ b](∂n ⁄ ∂u2)
(287) E2 = [(fG − gF) ⁄ b](∂n ⁄ ∂u1) + [(fF − eG) ⁄ b](∂n ⁄ ∂u2)
where b = eg − f2 is the determinant of the surface covariant
curvature tensor.
From Eq. 274↑, it can be seen that the coefficients of the
surface covariant curvature tensor, bαβ, are the projections of
the partial derivative of the surface basis vectors, ∂Eα ⁄ ∂uβ, in
the direction of the unit normal vector to the surface, n, that is:
(288) bαβ = (∂Eα ⁄ ∂uβ)⋅n
This can also be seen directly from the definition of the
coefficients of the second fundamental form, i.e. Eqs.
249↑-251↑.
As indicated above, the essence of the above equations of
Gauss and Weingarten is that the partial derivatives of E1, E2
and n can be represented as combinations of these vectors with
coefficients obtained from the coefficients of the first and
second fundamental forms and their partial derivatives. The
above equations are various demonstrations of this fact.
For a Monge patch of the form r(u, v) = (u, v, f(u, v)), the
Gauss equations are given by:
(289) ∂E1 ⁄ ∂u = [fufuuE1 + fvfuuE2 + fuu√(1 + fu2 + fv2) n] ⁄ (1 +
fu2 + fv2)
(290) ∂E1 ⁄ ∂v = [fufuvE1 + fvfuvE2 + fuv√(1 + fu2 + fv2) n] ⁄ (1 +
fu2 + fv2) = ∂ E2 ⁄ ∂u
(291) ∂E2 ⁄ ∂v = [fufvvE1 + fvfvvE2 + fvv√(1 + fu2 + fv2) n] ⁄ (1 +
fu2 + fv2)
where the subscripts u and v represent partial derivatives of f
with respect to the surface coordinates u and v.
Similarly, for a Monge patch of the form r(u, v) = (u, v, f(u, v)),
the Weingarten equations are given by:
(292) ∂n ⁄ ∂u = [(fufvfuv − fuufv2 − fuu) ⁄ (1 + fu2 + fv2)3 ⁄ 2] E1 +
[(fufvfuu − fu2fuv − fuv) ⁄ (1 + fu2 + fv2)3 ⁄ 2] E2
(293) ∂n ⁄ ∂v = [(fufvfvv − fuvfv2 − fuv) ⁄ (1 + fu2 + fv2)3 ⁄ 2] E1 +
[(fufvfuv − fu2fvv − fvv) ⁄ (1 + fu2 + fv2)3 ⁄ 2] E2

3.9.1 Codazzi-Mainardi Equations

From the aforementioned equations of Gauss and Weingarten,

supported by further compatibility conditions, the following
equations, called Codazzi or Codazzi-Mainardi equations, can
be derived:
(294) (∂b12 ⁄ ∂u1) − (∂b11 ⁄ ∂u2) = b22Γ211 − b12(Γ212 − Γ111) −
b11Γ112
(295) (∂b22 ⁄ ∂u1) − (∂b21 ⁄ ∂u2) = b22Γ212 − b12(Γ222 − Γ112) −
b11Γ122
where the Christoffel symbols are based on the surface metric.
These equations can be expressed compactly in tensor notation
as:
(296) (∂bαβ ⁄ ∂uγ) − (∂bαγ ⁄ ∂uβ) = bδβΓδαγ − bδγΓδαβ
Now, if we arrange the terms of the last equation and subtract
the term bαδΓδγβ from both sides we obtain:
(297) (∂bαβ ⁄ ∂uγ) − bδβΓδαγ − bαδΓδγβ = (∂bαγ ⁄ ∂uβ) − bδγΓδαβ −
bαδΓδγβ
which can be expressed compactly, using the covariant
derivative notation (see § 7↓), as:
(298) bαβ;γ = bαγ;β
The Codazzi-Mainardi equations in the form given by Eq.
298↑ reveal that there are only two independent components
for these equations because, adding to the fact that all the
indices range over 1 and 2 and hence we have only eight
components, the covariant derivative according to Eq. 298↑ is
symmetric in its last two indices (i.e. β and γ), and the
covariant curvature tensor is symmetric in its two indices (i.e.
bαβ = bβα). These two independent components are given by:
(299) bαα;β = bαβ;α
where α ≠ β and there is no summation over α. On writing
these equations in full, using the covariant derivative
expression [i.e. bαβ;γ = (∂bαβ ⁄ ∂uγ) − bδβΓδαγ − bαδΓδβγ] and
noting that one term of the covariant derivative expression is
the same on both sides and hence it drops away, we have:
(300) (∂bαα ⁄ ∂uβ) − bαδΓδαβ = (∂bαβ ⁄ ∂uα) − bδβΓδαα
where α ≠ β with no sum on α. It should be remarked that as a
consequence of the aforementioned two symmetries, the
covariant derivative of the surface covariant curvature tensor,
bαβ;γ, is fully symmetric in all of its indices.
There is also another more general equation called (according
to some authors) the Gauss-Codazzi equation which is given
by:
(301) Rδαβγxiδ = xiδbδβbαγ − xiδbδγbαβ + nibαβ;γ − nibαγ;β
The tangential component of this equation represents
Theorema Egregium (in the form given by Eq. 228↑) while its
normal component represents the Codazzi equation (in the
form given by Eq. 298↑). The reader is advised to refer to §
4.7↓ about the essence of Theorema Egregium as an expression
of the fact that certain types of curvature are intrinsic
properties to the surface and hence they can be expressed in
terms of purely intrinsic parameters obtained from the first
fundamental form.
3.10 Sphere Mapping
Sphere mapping or Gauss mapping is a correlation between the
points of a surface and the unit sphere where each point on the
surface is projected onto its unit normal as a point on the unit
sphere which is centered at the origin of coordinates. This sort
of mapping for surfaces is similar to the spherical indicatrix
mapping (see § 5.5↓) for space curves. In technical terms, let S
be a surface embedded in an ℝ3 space and S1 represents the
origin-centered unit sphere in this space, then Gauss mapping
is given by:
(302) {N:S → S1, N(P) = P}
where the point P(x, y, z) on the trace of S is mapped by N onto
the point P(x, y, ž) on the trace of the unit sphere with x, y, z
being the coordinates of P and x, y, ž being the coordinates of
the origin-based position vector of the normal vector to the
surface, n, at P. To have a single-valued sphere mapping, the
functional relation representing the surface S should be one-to-
one.
The image D̃ on the unit sphere of a Gauss mapping of a patch
D on a surface S is called the spherical image of D. The limit
of the ratio of the area of a region Z̃ on the spherical image to
the area of the corresponding region Z on the surface S in the
neighborhood of a given point P on S equals the absolute value
of the Gaussian curvature |K| at P as Z shrinks to the point P,
that is:
(303)
where σ stands for area, and KP is the Gaussian curvature at P.
The tendency of Z to P should be understood in the given
sense.
At a given point P on a surface, where the Gaussian curvature
is non-zero, there exists a neighborhood N of P where an
injective mapping can be established between N and its
spherical image Ñ. A conformal correspondence can be
established between a surface and its spherical image iff the
surface is a sphere or a minimal surface (see § 6.7↓). We
should remark that although the term “spherical image” is
related to surfaces in the context of sphere mapping, it can also
be used for curves since curves can also have spherical images
in a well defined sense.
3.11 Global Surface Theorems
In this small section we state, within the following bullet
points, a few global theorems related to surfaces to have a taste
of this field of differential geometry of surfaces whose
investigation is not the main objective of the present book.
• Planes are the only connected surfaces of class C2 whose all
points are flat.
• Spheres are the only connected closed surfaces of class C3
whose all points are spherical umbilical (see § 4.10↓).
• Spheres are the only connected compact surfaces of class C3
with constant Gaussian curvature.
• Spheres are the only connected compact surfaces with
constant mean curvature and positive Gaussian curvature.
• The tangent plane to a cylinder or a cone is constant along
their generators.
• Any compact surface in ℝ3 should have points with positive
Gaussian curvature.
• Any compact surface in ℝ3, excluding sphere, should have
points with negative Gaussian curvature.
• The reader is also referred to § 4.8↓ for the global form of the
Gauss-Bonnet theorem.
3.12 Exercises
Exercise 1. Give the mathematical definition of space surface
and explain the difference between a surface and its trace
according to this definition.
Exercise 2. What is the mathematical condition for a surface to
be regular at a particular point in terms of its basis vectors?
Exercise 3. State the three main mathematical methods for
defining a space surface and compare them explaining any
advantages or disadvantages in using one of these methods or
the others in various contexts.
Exercise 4. Classify the three methods of the last question into
two main categories and discuss these categories (see §
1.4.3↑).
Exercise 5. What “coordinate patch of class Cn” means? What
are the mathematical conditions that should be satisfied by
such a patch?
Exercise 6. Show that a Monge patch of the form r(u, v) = (u,
v, f(u, v)) is regular of class Cn if f is of this class.
Exercise 7. Give a rigorous definition of “tangent vector” of a
surface curve at a particular point on the surface.
Exercise 8. How the tangent plane of a surface at a particular
point is related to the basis vectors of the surface at that point?
Exercise 9. Find the equation of the tangent plane to the
ellipsoid which is represented parametrically by:
r(θ, φ) = (2sinθ cosφ, 1.5sinθ sinφ, 0.5cosθ)
at the point with θ = 1.3 and φ = 0.72.
Exercise 10. Find the equation of the tangent plane of a surface
represented parametrically by:
r(u, v) = (6v, 2u, 1.4u2 + 6)
at the point with u = 1.2 and v = 3.6.
Exercise 11. Show that for a circular cone the tangent plane at
all points of any one of its generators is the same.
Exercise 12. Discuss the following statement explaining its
meaning in simple words: “The tangent space at a specific
point P of a surface is a property of the surface at P and hence
it is independent of the patch that contains P”.
Exercise 13. Does the tangent space of a surface at a given
point depend on the particular parameterization of the surface?
Exercise 14. Find the equation of the plane passing through the
point ( − 1, 3, − 9) and spanned by the two vectors (3, 0.5, 1.2)
and (0.9, 3, 6.8).
Exercise 15. Define, mathematically, the normal unit vector n
of a surface at a given point in terms of the two basis vectors of
the surface at that point.
Exercise 16. Calculate, symbolically, the normal unit vector n
at a general point of a surface defined by: S(x, y) = (x, y, f)
where f = f(x, y) is a differentiable function.
Exercise 17. Find the equation of the normal line of a
hyperboloid of one sheet given by:
r(ξ, θ) = (1.6coshξ cosθ, 2.1coshξ sinθ, 0.4sinhξ)
at the point with ξ = 3.2 and θ = 1.5.
Exercise 18. Find the equation of the normal line of a surface
represented by:
r(u, v) = (3u, u2 + v, 5v)
at the point with u = 2.5 and v = − 1.8.
Exercise 19. Define “Monge patch” giving its three forms.
Which of these forms is the most common in use?
Exercise 20. Define, briefly, the following terms with some
examples representing these concepts: simple surface, simply
connected region on a surface, closed surface, compact surface,
elementary surface, oriented surface and developable surface.
Exercise 21. Which of the following is a simple surface and
which is not: cylinder, hyperboloid of two sheets, torus, Klein
bottle, and elliptic paraboloid?
Exercise 22. Is the surface represented by the equation:
x2 + y2 + z2 = 9
compact? What about the surface represented by:
x2 − y2 − z2 = 4?
Exercise 23. Why the Mobius strip is not an orientable
surface? Give another example of a non-orientable surface.
Exercise 24. What is conformal mapping? What is direct and
inverse conformal mapping?
Exercise 25. Describe stereographic mapping making a simple
sketch representing this type of mapping.
Exercise 26. Prove that stereographic mapping is conformal.
Exercise 27. Define isometric mapping giving an example of
such mapping between two types of surface.
Exercise 28. What are the mathematical conditions for two
surfaces to be isometric? Does isometry relate to the intrinsic
or extrinsic properties of the surface and why?
Exercise 29. What is the distinctive property of an isometric
relation between two surfaces in terms of angles, arc lengths
and areas defined on the two surfaces?
Exercise 30. What is the relation between conformal mapping
and isometric mapping?
Exercise 31. What is local isometry? What is the difference
between local isometry and global isometry?
Exercise 32. Show that Eq. 160↑ applies to local isometric
mapping.
Exercise 33. A surface S1 is mapped isometrically onto another
surface S2. How the intrinsic properties of S1 will be affected
by this mapping?
Exercise 34. Make a clear distinction between the tangent
surface of a curve and the tangent plane of a surface describing
each of these briefly.
Exercise 35. What is the meaning of “branch of the tangent
surface of a curve C at a given point P on the curve”?
Exercise 36. Give a brief definition of involute and evolute.
Exercise 37. Write down a general mathematical relation
representing the position vector of a point P on a space surface
as a function of the surface coordinates, u1 and u2, where the
surface is embedded in a 3D Euclidean space coordinated by a
rectangular Cartesian system.
Exercise 38. Describe, in detail, how a coordinate grid is
constructed on a space surface with a clear definition of the
coordinate curves used to build this grid.
Exercise 39. Make a simple and fully labeled sketch of a space
surface coordinated by a curvilinear grid with the two
covariant surface basis vectors and the unit normal vector to
the surface at one point.
Exercise 40. A surface is represented spatially by:
r(u, v) = (u, 5, 2v).
Discuss the type and the main properties of this surface.
Exercise 41. Define, symbolically, the covariant basis vectors
of a space surface in terms of the coordinates of the ambient
space xi and the coordinates of the surface uα.
Exercise 42. Give the symbol used in tensor notation to
represent the surface basis vectors Eα. From analyzing this
symbol, describe how the surface basis vectors can be regarded
as covariant and contravariant vectors at the same time.
Exercise 43. Write, in full tensor notation, the equation
representing the covariant form of the unit normal vector to a
space surface.
Exercise 44. Although E1 and E2 are linearly independent at
the regular points of the surface they are not necessarily
orthogonal or of unit length. Is it possible to construct an
orthonormal set of basis vectors from E1 and E2? If so, how?
Exercise 45. Following a transformation from an unbarred
surface coordinate system to a barred surface coordinate
system, what are the mathematical expressions representing the
barred basis set, Ẽ1 and Ẽ2, in terms of the unbarred basis set,
E1 and E2? Give these expressions in vector and tensor
notations.
Exercise 46. Define, descriptively and mathematically, the set
of contravariant basis vectors for a space surface discussing
how these vectors can be regarded as covariant and
contravariant vectors simultaneously.
Exercise 47. How the covariant and contravariant basis sets of
a space surface can be obtained from each other? State this in
words and mathematically defining all the symbols involved.
Exercise 48. What is the significance of the following relation
involving the covariant and contravariant surface basis vectors
and the Kronecker delta: Eα⋅Eβ = δαβ?
Exercise 49. Define, mathematically, the coefficients of the
surface metric tensor in terms of the surface covariant basis
vectors in Euclidean and Riemannian spaces.
Exercise 50. Give the fundamental relation that provides the
important link between the metric tensor of a surface and the
metric tensor of its enveloping space.
Exercise 51. Find the surface basis vectors, E1 and E2, and the
coefficients of the first fundamental form of a surface
parameterized by:
r(u, v) = (a u cosv, b u sinv, c u2)
where a, b, c are constants.
Exercise 52. Find, symbolically, the first fundamental form of
a cylinder represented by:
r(u, v) = (f1(u), f2(u), v)
where f1 and f2 are continuous functions of the given
coordinate.
Exercise 53. Given the fact that for 3D Cartesian systems:
IS = (dx1)2 + (dx2)2 + (dx3)2
plus the transformation equations from spherical to Cartesian
coordinates in 3D, derive IS for spherical coordinate systems.
Exercise 54. Given the fact that for 3D Cartesian systems:
IS = (dx1)2 + (dx2)2 + (dx3)2
plus the transformation equations between general curvilinear
and Cartesian coordinate systems in 3D, prove that for general
curvilinear systems:
IS = aαβduαduβ.
Exercise 55. Define, mathematically, each of the following
vectors: x1α, x2α, x3α, xi1, xi2. Also, discuss their attributes as
space and surface vectors and their variance type.
Exercise 56. Discuss how a surface vector can also be
considered as a space vector stating the mathematical link
between the surface and space representations. Are these
representations equivalent? If so, how?
Exercise 57. Write down, using tensor notation, the
mathematical relation that correlates the surface basis vectors
to the unit normal vector to the surface.
Exercise 58. What is the significance of the following relation
which involves the surface contravariant and covariant metric
tensors and the Kronecker delta:
aαγ aγβ = δαβ?
Exercise 59. Give the matrix [aαβ] that represents the
contravariant form of the surface metric tensor in terms of the
coefficients of the first fundamental form.
Exercise 60. Give the matrix [aαβ] that represents the mixed
form of the surface metric tensor.
Exercise 61. Write down the relations that represent the
transformation between unbarred and barred surface coordinate
systems.
Exercise 62. How the determinants of the surface metric tensor
of two transformed coordinate systems of a given surface are
linked?
Exercise 63. Express the Christoffel symbols of the first kind
[αβ, γ] for a surface in terms of the surface covariant basis
vectors and their partial derivatives.
Exercise 64. Derive the mathematical relation between the
partial derivative of the surface metric tensor ∂αaβγ and the
Christoffel symbols of the first kind for the surface.
Exercise 65. How the coefficients of the surface metric tensor
will be affected by scaling the surface up or down by a
constant positive scalar factor?
Exercise 66. Give the covariant and contravariant types of the
surface metric tensor for a Monge patch of the form r(u, v) =
(u, v, f(u, v)).
Exercise 67. Discuss how the concept of “length of straight
segment” is extended to the length of a polygonal arc. Also
discuss how the concept of “length of polygonal arc” is
extended to the length of an arc of a twisted space curve.
Exercise 68. Derive the following relation which links the
length of an element of arc of a curve residing on a 2D surface
to the covariant metric tensor of the surface:
(ds)2 = aαβduαduβ.
Exercise 69. Is the length of a surface curve an intrinsic or
extrinsic property and why?
Exercise 70. Give the formula for the length, L, of a segment
of a t-parameterized surface curve in terms of the coefficients
of the first fundamental form of the surface.
Exercise 71. Using the metric tensor, verify the following
relation where f represents a Monge patch of the form r(u, v) =
(u, v, f(u, v)):
ds = √([1 + fu2]dudu + 2fufvdudv + [1 + fv2]dvdv)
Exercise 72. Develop an analytical expression for the length of
an element of arc, ds, of the catenary parameterized by Eqs.
28↑-29↑.
Exercise 73. Discuss how the concept of “area of polygonal
plane fragment” is extended to the area of a surface consisting
of polygonal plane fragments. Also discuss how the concept of
“area of surface made of polygonal plane fragments” is
extended to the area of a generalized twisted space surface.
Exercise 74. Derive the mathematical expression for the area
of an infinitesimal element of a surface and the expression for
the area of a surface patch.
Exercise 75. Derive Eq. 208↑ for the surface area of a Monge
patch of the form r(u, v) = (u, v, f(u, v)).
Exercise 76. A cone is represented parametrically in a 3D
space by:
r(ρ, φ) = (ρ cosφ, ρ sinφ, c ρ)
where ρ, φ are polar coordinates (ρ ≥ 0 and 0 ≤ φ < 2π) and c is
a positive constant. Find the area of the part of the cone
corresponding to 0 ≤ ρ ≤ A where A is a given positive
constant.
Exercise 77. Derive the mathematical expression: cosθ =
gijAiBj for the angle θ between two unit surface vectors, A and
B, where gij is the space covariant metric tensor. Also give the
mathematical expression of sinθ for the angle between A and
B.
Exercise 78. Give two mathematical expressions for the
coefficients of the surface covariant curvature tensor bαβ.
Exercise 79. Using a spherical coordinates representation, find
the determinant of the surface curvature tensor of a sphere.
Exercise 80. Show that ∂βEα = ∂αEβ.
Exercise 81. Prove that if two surfaces, S1 and S2, are mapped
isometrically one on the other then they have identical first
fundamental form coefficients at their corresponding points,
i.e. E1 = E2, F1 = F2 and G1 = G2.
Exercise 82. Give the matrix [bαβ] which represents the
contravariant form of the surface curvature tensor in terms of
the coefficients of the first and second fundamental forms.
Exercise 83. Discuss and justify the relation b̃ = J2b explaining
all the symbols involved.
Exercise 84. What are the other symbols used by some authors
to label the coefficients of the second fundamental form e, f, g?
Discuss the advantages and disadvantages of using each one of
these sets of symbols. Also, write the mathematical expression
for the second fundamental form using the alternative symbols.
Exercise 85. How can we obtain the mixed form of the surface
curvature tensor bαβ from the covariant form of this tensor bαβ?
Exercise 86. Express the mean curvature H and the Gaussian
curvature K of a surface in terms of the mixed form of the
surface curvature tensor bαβ.
Exercise 87. Explain, in details, all the symbols and notations
involved in the following relation:
bαγbβδ − bαδbβγ = (∂Γαβδ ⁄ ∂γ) − (∂Γαβγ ⁄ ∂δ) + ΓωβδΓαωγ −
ΓωβγΓαωδ
Exercise 88. Write down the matrix form of the surface
covariant curvature tensor for a Monge patch of the form r(u,
v) = (u, v, f(u, v)).
Exercise 89. What is the relation between the Riemann-
Christoffel curvature tensor of the second kind and the
curvature tensor of a surface?
Exercise 90. Give all the coefficients of the Riemann-
Christoffel curvature tensor and the curvature tensor for a
plane surface.
Exercise 91. On a space surface, how many independent non-
vanishing components the Riemann-Christoffel curvature
tensor possesses?
Exercise 92. Explain the relation between the coefficients of
the metric tensor of a surface and its first fundamental form
stating the necessary equations.
Exercise 93. Derive the mathematical formula for IS in terms
of the coefficients of the first fundamental form.
Exercise 94. Express the determinant of the surface covariant
metric tensor as a function of the coefficients of the first
fundamental form.
Exercise 95. Express E, F, G as dot products of the covariant
basis vectors of the surface and relate this to the space metric
tensor.
Exercise 96. Does the first fundamental form provide a unique
characterization of the space surface as seen internally by a 2D
inhabitant? Explain why.
Exercise 97. Does the first fundamental form provide a unique
characterization of the space surface as seen from the external
ambient space? Explain why.
Exercise 98. State the mathematical conditions for the
provision of positive definiteness of the first fundamental form.
Exercise 99. Give the mathematical conditions that apply to
the coefficients of the covariant metric tensor of two isometric
surfaces at their corresponding points.
Exercise 100. Explain how the second fundamental form
characterizes the surface from the ambient space perspective
and how the unit normal vector to the surface is employed in
this characterization.
Exercise 101. Express the determinant of the surface covariant
curvature tensor as a function of the coefficients of the second
fundamental form.
Exercise 102. Derive the mathematical relation of the second
fundamental form IIS in terms of the coefficients e, f, g. Also
provide the main mathematical definitions for the coefficients
e, f, g in terms of the surface basis vectors and the unit normal
vector to the surface.
Exercise 103. What is the relation between the second
fundamental form IIS and the second order differential of the
position vector d2 r of a surface?
Exercise 104. State the mathematical relations between the
coefficients of the second fundamental form and the
coefficients of the surface covariant curvature tensor.
Exercise 105. Express, in full tensor notation, the second
fundamental form in terms of the coefficients of the surface
covariant curvature tensor and link this to the expression
involving the coefficients e, f, g.
Exercise 106. Explain how the following relation provides a
bridge between the first and second fundamental forms: IIS =
κnIS.
Exercise 107. Derive the following relation where f is a
functional representation of a Monge patch of the form r(u, v)
= (u, v, f(u, v)):
IIS = (fuududu + 2fuvdudv + fvvdvdv) ⁄ √(1 + fu2 + fv2)
Exercise 108. Discuss how the first and second fundamental
forms represent the intrinsic and extrinsic geometry of the
surface.
Exercise 109. If two surfaces have identical first and second
fundamental forms, should they be congruent?
Exercise 110. What are the compatibility conditions linking
the first and second fundamental forms which are needed to
fully identify a surface associated with specific first and second
fundamental forms and secure its existence?
Exercise 111. State, using mathematical technical terms, the
fundamental theorem of space surfaces.
Exercise 112. Give a brief definition of Dupin indicatrix and
state its significance and usage in differential geometry.
Exercise 113. What is the shape of Dupin indicatrix at elliptic,
parabolic and hyperbolic points on a smooth surface? What is
the shape of Dupin indicatrix at flat points?
Exercise 114. Make a simple sketch to illustrate Dupin
indicatrix at an elliptic point, a parabolic point and a
hyperbolic point on a surface marking the two principal
directions in each case.
Exercise 115. Write down the mathematical expression for the
third fundamental form IIIS in terms of the unit normal vector
to the surface and in terms of the coefficients cαβ.
Exercise 116. Express the coefficients of the third fundamental
form as a function of the coefficients of the surface metric and
curvature tensors.
Exercise 117. Explain all the symbols involved in the
following equation:
K IS − 2H IIS + IIIS = 0.
Exercise 118. Derive the equation in the last question using the
Weingarten equations.
Exercise 119. Starting from the equation:
Kaαβ − 2Hbαβ + cαβ = 0,
derive, with full explanation, the following relation:
tr(cβα) = 4H2 − 2K.
Exercise 120. Explain the correspondence between the Frenet-
Serret formulae for space curves and the equations of Gauss
and Weingarten for space surfaces.
Exercise 121. Why the partial derivatives of the surface basis
vectors, E1 and E2, and the unit normal vector to the surface, n,
with respect to the surface coordinates, u1 and u2, can be
expressed as combinations of these vectors?
Exercise 122. Prove Eq. 282↑ using the Weingarten equations.
Exercise 123. Write the derivatives of the surface basis vectors
(i.e. ∂βEα) in terms of the surface vectors (i.e. Eγ and n) in their
vector and tensor forms.
Exercise 124. What is the essence of the equations of
Weingarten? Provide qualitative and quantitative descriptions
of these equations. Also write these equations in a matrix form
involving the covariant curvature tensor and the contravariant
metric tensor of the surface.
Exercise 125. Derive Weingarten equations for a Monge patch
of the form r(u, v) = (u, v, f(u, v)).
Exercise 126. Express the partial derivatives of n with respect
to the surface coordinates in terms of the Gaussian and mean
curvatures and the coefficients of the first and second
fundamental forms.
Exercise 127. Give the equations of Gauss for a Monge patch
of the form r(u, v) = (u, v, f(u, v)).
Exercise 128. State, using full tensor notation, the equations of
Codazzi-Mainardi explaining all the symbols involved.
Exercise 129. Derive, using the Codazzi-Mainardi equations,
the following relation: bαβ;γ = bαγ;β.
Exercise 130. Explain how the relation in the previous
question indicates that there are only two independent
components for the Codazzi-Mainardi equations. What are
these two independent components?
Exercise 131. Describe sphere mapping in qualitative and
technical terms. Also, explain the meaning of the following
equation:

Exercise 132. Prove the theorem represented by the equation

in the previous question.
Exercise 133. Prove that at a given point P on a surface with
K ≠ 0, there exists a neighborhood N of P where an injective
mapping can be established between N and its spherical image
Ñ.
Exercise 134. State one of the global theorems of space
surface and explain why it is global.
Chapter 4
Curvature
“Curvature” is a property of both curves and surfaces at a
given point which is determined by the shape of the curve or
surface at that point. There are also global characteristics of
curvature like total curvature Kt (see § 4.8↓) of a surface but
they are based in general on the local characterization of
curvature at individual points. The curvature also has intrinsic
as well as extrinsic attributes and hence it characterizes the
manifold internally as seen by an inhabitant of the manifold
and externally as seen by an outsider. In this chapter, we
investigate this property in its general meaning and examine
the main parameters used to describe curvature and quantify it
focusing on space surfaces and curves embedded in such
surfaces. The materials are largely based on a 3D flat ambient
space coordinated by a Cartesian orthonormal system.
4.1 Curvature Vector
At a given point P on a surface S, a plane containing the vector
n, which is the normal unit vector to the surface at P, intersects
the surface in a surface curve C having a tangent vector t at P.
The curve C is called the normal section of S at P in the
direction of t. The principal normal vector N of a normal
section at P is collinear with the unit normal vector n at P. We
note that for a normal section, N and n can be parallel or anti-
parallel since on an orientable surface the vector n can have
one of two possible directions. On the other hand, a surface
curve passing through P in the direction of t may not be a
normal section and hence the vectors N and n at P have
different orientations. In this context we remark that we are
considering here the part of the curve in the neighborhood of
the point P as part of a normal section or not, and not
necessarily the whole curve.
As explained before, a space curve C can be parameterized by
s, representing the distance traversed along C, and hence the
curve is defined by the position vector r(s). At a given point P
on C, the vector T = (dr ⁄ ds) is a unit vector tangent to C at P
in the direction of increasing s. When C is embedded in a
surface, T will be contained in the tangent plane to the surface
at P, as explained previously in § 2.2↑ and 3.1↑. We note that
the vector T can be parallel or anti-parallel to the
aforementioned vector t. We chose to introduce t and define it
in this way to be more general since the curve orientation can
be in one direction or the other and hence T and t can be
parallel or anti-parallel. Moreover, t is not necessarily of unit
length or based on a natural parameterization of the curve.
The curvature vector of C at P, which is orthogonal to T, is
defined by:
(304) K = dT ⁄ ds
where K, which is the uppercase Greek letter kappa,
symbolizes the curvature vector. The curvature κ of C at P
(which is defined previously in § 2.2↑) is the magnitude of the
curvature vector, that is: κ = |K|, and the radius of curvature
when κ ≠ 0 is its reciprocal, i.e. Rκ = 1 ⁄ κ, which is the radius
of the osculating circle of C at P (see § 2.6↑). The curvature
vector can therefore be expressed as:
(305) K = |K| (K ⁄ |K|) = κN
where N is the principal normal vector of the curve C at P as
defined previously in § 2.2↑.
The curvature vector of a surface curve is independent of the
orientation and parameterization of the surface and the curve.
However, the curvature vector at a particular point is
determined by the local shape of the curve, which is partly
determined by the position on the surface and the tangential
direction to the surface at that position, and hence the curvature
vector depends on the surface point and tangential direction, as
well as on other factors.
A point on the curve at which the curvature vector K vanishes,
and hence κ = 0, is called inflection point. At such a point, the
radius of curvature is infinite and the principal normal vector N
and the osculating circle are not defined. However, since it is
usually assumed that the curve is of class C2, the curvature
vector varies smoothly and hence at isolated points of
inflection on such a curve, N may be defined in such a way to
ensure continuity when this is possible, which is not always the
case.
On introducing a new unit vector which is orthogonal to both n
and T and defined by the following cross product:
(306) u = n × T
the curvature vector, which lies in the plane spanned by n and
u, can then be resolved in the n and u directions, which
represent the normal and tangential directions to the surface at
the given point, as:
(307) K = Kn + Kg = κnn + κgu
where Kn and Kg are the normal and geodesic components of
the curvature vector K, while κn and κg are the normal and
geodesic curvatures of the curve at the given point
respectively.
On dot producting both sides of Eq. 307↑ with n and u in turn,
noting that n and u are orthogonal unit vectors, the normal and
geodesic curvatures can be obtained, that is:
(308) κn = n⋅K = − T⋅(dn ⁄ ds) = − (dr ⁄ ds)⋅(dn ⁄ ds)
(309) κg = u⋅K = u⋅(dT ⁄ ds) = (n × T)⋅(dT ⁄ ds)
The second equality of Eq. 308↑ is based on the fact that T and
n are orthogonal (since T is tangent to the surface while n is
normal to the surface); therefore by the product rule of
differentiation we have:
(310)
d(n⋅T) ⁄ ds =
d(0) ⁄ ds =
0=
n⋅(dT ⁄ ds) + (dn ⁄ ds)⋅T =
n⋅K + T⋅(dn ⁄ ds)
and hence: n⋅K = − T⋅(dn ⁄ ds). The other equalities in Eqs.
308↑ and 309↑ are based on definitions which have been given
previously.
The vector u, which is a unit vector normal to the curve C, is
called the geodesic normal vector. This vector is the
normalized projection of K onto the tangent space of the
surface and hence it is contained in the tangent plane of the
surface at the given point. Also, because the vector u is
orthogonal to the curve C at P (since u is orthogonal to T by
the cross product), it is contained in the normal plane of C at P.
Hence, u occurs at the intersection of the tangent plane of the
surface and the normal plane of the curve which correspond to
the given point.
While the normal curvature κn is an extrinsic property, since it
depends on the first and second fundamental form coefficients,
as seen in § 3.6↑ (Eq. 260↑) and as will be seen in 4.2↓, the
geodesic curvature κg in an intrinsic property as it depends
only on the first fundamental form coefficients and their
derivatives (see § 4.3↓). We note that the triad (n, T, u) is
another moving frame which is in common use in differential
geometry in addition to the curve-based Frenet frame (T, N, B)
and the surface-based frame (E1, E2, n).
Regarding the relation between the principal normal vector of
the curve N, and the unit normal vector to the surface n, let C
be a curve on a sufficiently smooth surface S. If φ is the angle
between the vector N of C at a given point P and the vector n
of S at P then we have:
(311) cosφ = n⋅N
Accordingly, the normal and geodesic curvatures, κn and κg, of
C at P are given by:
(312) κn = κ cosφ
(313) κg = κ sinφ
where κ is the curvature of C at P as defined previously (see §
2.2↑ and the previous parts of the present section). In fact, Eq.
312↑ can be easily obtained by combining Eq. 308↑ with Eq.
305↑, while Eq. 313↑ can be similarly obtained by combining
Eq. 309↑ with Eq. 305↑, that is:
(314)
κn =
n⋅K =
n⋅(κN) =
κ(n⋅N) =
κ cosφ
(315)
κg =
u⋅K =
u⋅(κN) =
κ(u⋅N) =
κ cos[(π ⁄ 2) − φ] =
κ sinφ
According to the theorem of Meusnier, if P is a given point on
a sufficiently smooth surface S, then all curves on S that pass
through P with the same tangent direction at P have the same
normal curvature at P. The theorem of Meusnier may also be
stated in this context as: the curvature of any surface curve at a
given point P on the curve is equal in magnitude to the
curvature of the normal section which is tangent to the curve at
P divided by the cosine of the angle between the principal
normal vector to the curve at P and the normal vector to the
surface at P. This version of the theorem is based on Eq. 312↑
plus the fact that the normal curvature κn of a normal section is
equal in magnitude to its curvature κ. Also, we should exclude
from this version the case of having cosφ = 0 when the
curvature vector of the curve is tangent to the surface. More
details about Meusnier theorem will be given in § 4.2.1↓.
4.2 Normal Curvature
Using the first and second fundamental forms, given by Eqs.
233↑ and 244↑, the normal curvature at a given point on the
surface in the (du2 ⁄ du1) direction can be expressed as the
following quotient of the second fundamental form involving
the coefficients of the surface covariant curvature tensor to the
first fundamental form involving the coefficients of the surface
covariant metric tensor (see Eq. 259↑):
(316)
κn =
IIS ⁄ IS =
[e(du1)2 + 2fdu1du2 + g(du2)2] ⁄ [E(du1)2 + 2Fdu1du2 +
G(du2)2] =
(bαβduαduβ) ⁄ (aγδduγduδ)
where the symbols are as explained before, and the last part is
to be interpreted as the sum of the terms in the numerator
divided by the sum of the terms in the denominator. This
equation can be obtained as follows:
(317)
κn = − T⋅(dn ⁄ ds)
(Eq. 308↑)
= − [(∂r ⁄ ∂uα)(duα ⁄ ds)]⋅[(∂n ⁄ ∂uβ)(duβ ⁄ ds)]
(Eq. 106↑)
= − [(∂r ⁄ ∂uα)⋅(∂n ⁄ ∂uβ)](duα ⁄ ds)(duβ ⁄ ds)
= − {[(∂r ⁄ ∂uα)⋅(∂n ⁄ ∂uβ)]duαduβ} ⁄ (dsds)
= (bαβduαduβ) ⁄ (ds)2
(Eq. 213↑)
= (bαβduαduβ) ⁄ (aγδduγduδ)
(Eq. 233↑).
From the previous statements plus Eq. 307↑, it can be seen that
the normal component of the curvature vector may also be
given by:
(318) Kn = [e(du1 ⁄ ds)2 + 2f(du1 ⁄ ds)(du2 ⁄ ds) + g(du2 ⁄ ds)2] n
Also, from Eq. 316↑ it can be seen that the sign of the normal
curvature κn (i.e. being greater than, less than or equal to zero)
is determined solely by the sign of the second fundamental
form since the first fundamental form is positive definite. As
seen before (see § 3.6↑), the sign of the second fundamental
form depends on the surface orientation which is determined
by the direction of n. Therefore, the sign of κn depends on the
surface orientation. Apart from this, the sign of the second
fundamental form is related to the sign of the determinant of
the surface covariant curvature tensor b, and hence the sign of
κn is also related to the sign of b.
As given earlier, all surface curves passing through a given
point P on a surface and have the same tangent line at P have
identical normal curvature at P. Hence, the normal curvature is
a property of the surface at a given point and in a given
direction and not only a property of the curve. The normal
curvature κn of a given normal section C of a surface at a
particular point P is equal in magnitude to the curvature κ of C
at P, i.e. |κn| = κ. This can be explained by the fact that the
normal vector n to the surface at P is collinear with the
principal normal vector N of C at P so there is only a normal
component to the curvature vector with no tangential geodesic
component.
The normal curvature of a surface at a given point and in a
given spatially-fixed tangential direction is an invariant
property with respect to change of parameterization and
representation of the surface apart from its sign which is
dependent on the choice of the direction of the unit normal
vector n to the surface, as seen earlier. The normal curvature is
an extrinsic property, since it necessarily depends on the
coefficients of the second fundamental form, and hence it
cannot be expressed purely in terms of the coefficients of the
first fundamental form.
At flat points on a surface, κn = 0 in all directions. At elliptic
points, κn ≠ 0 in any direction and it has the same sign in all
directions. At parabolic points, κn has the same sign in all
directions except the direction in which the second
fundamental form vanishes where κn = 0. At hyperbolic points,
κn is negative, positive and zero depending on the direction.
For the definition of flat, elliptic, parabolic and hyperbolic
points and their significance, the reader is referred to § 4.9↓.
In any two orthogonal tangential directions at a given point P
on a sufficiently smooth surface, the sum of the normal
curvatures corresponding to these directions at P is constant.
At any point P of a sufficiently smooth surface S, there exists a
paraboloid (elliptic or hyperbolic) which is tangent at its vertex
to the tangent plane of S at P such that the normal curvature of
the paraboloid in a given direction at P is equal to the normal
curvature of S at P in that direction. This paraboloid takes the
degenerate form of a plane at flat points and a parabolic
cylinder at parabolic points (see § 4.9↓).
The normal curvatures of the surface curves at a given point P
in the directions of the u and v coordinate curves are given
respectively by:
(319) κnu = b11 ⁄ a11 = e ⁄ E
(320) κnv = b22 ⁄ a22 = g ⁄ G
where κnu and κnv are the normal curvatures in the directions of
the u and v coordinate curves, the indexed a and b are the
coefficients of the covariant metric tensor and the covariant
curvature tensor and where these are evaluated at P. This can
be seen, for example, from Eq. 316↑ where the last two terms
in the sums will vanish for the u1 coordinate curve since du2 =
0 while the first two terms in the sums will vanish for the u2
coordinate curve since du1 = 0. We note that du1 ≠ 0 on the u1
coordinate curve and du2 ≠ 0 on the u2 coordinate curve. As we
will see in § 4.4↓, at each non-umbilical point P (refer to §
4.10↓) of a sufficiently smooth surface there are two
perpendicular directions along which the normal curvature of
the surface at P takes its maximum and minimum values of all
the normal curvature values at P.
The necessary and sufficient condition for a given point P on a
sufficiently smooth surface S to be umbilical point is that the
coefficients of the first and second fundamental forms of the
surface at P are proportional, that is:
(321) e ⁄ E = f ⁄ F = g ⁄ G( = κn)
where E, F, G, e, f, g are the coefficients of the first and second
fundamental forms at P, and κn is the normal curvature of S at
P in any direction. This can be seen from Eq. 316↑ where κn in
this case becomes independent of the direction since by taking
out the common proportionality factor the expression will be
reduced to: κn = c where c is the proportionality factor. More
explicitly, according to Eq. 321↑: e = cE, f = cF and g = cG
where c is the common proportionality factor, and hence from
Eq. 316↑ we obtain: κn = c × 1 = c.
As seen before, the curvature κ and the normal curvature κn of
a surface curve at a given point P on the curve are related by:
(322) κn = κ cosφ
where φ is the angle between the principal normal vector N of
the curve at P and the unit normal vector n of the surface at P.
At every point on a sphere and in any direction, the normal
curvature is constant given by: |κn| = (1 ⁄ R) where R is the
sphere radius. At any point P on a sphere, any surface curve C
passing through P in any direction is a normal section iff C is a
great circle. All these great circles have constant curvature κ
and normal curvature κn which are both equal in magnitude to
(1 ⁄ R) where R is the sphere radius. We remark that the great
circles of a sphere are the plane sections formed by the
intersection of the sphere with the planes passing through the
center of the sphere.

4.2.1 Meusnier Theorem

According to the theorem of Meusnier, all surface curves

passing through a given point P on a surface and have the same
tangential non-asymptotic direction (see § 5.9↓) at P have
identical normal curvature which is the normal curvature κn
(and the curvature κ considering the magnitude) of the normal
section at P in the given direction. Moreover, the osculating
circles of these curves lie on a sphere Ss with radius (1 ⁄ κ) and
with center at rC = rP + (N ⁄ κ) where κ (which is equal in
magnitude to κn) is the curvature of the normal section at P, N
is the principal normal vector of the normal section at P and rP
is the position vector of P. As a consequence of this theorem,
we have:
1. The center of the sphere Ss is the center of curvature (see §
2.6↑) of the normal section at P in the given direction.
2. These curves are characterized by being tangent to the
normal section at P in the given direction and by being plane
sections of the surface with shared tangent direction at P (see
Footnote 16 in § 8↓).
3. The osculating circles of these curves are the intersection of
the sphere Ss with the osculating planes of these curves at P.
4. The sphere Ss is tangent to the tangent plane of the surface at
P.
The theorem of Meusnier may also be stated as follows: the
center of curvature of a surface curve at a given point P on the
curve is obtained by orthogonal projection of the center of
curvature of the normal section, which is tangent to the curve
at P, on the osculating plane of the curve.
4.3 Geodesic Curvature
As described earlier (see § 4.1↑), the curvature vector K of a
surface curve lies in a plane perpendicular to the tangent vector
T and it can be resolved into a normal component Kn = κnn
and a geodesic component Kg = κgu where the normal and
geodesic curvatures, κn and κg, are given by Eqs. 308↑ and
309↑. The geodesic component Kg of the curvature vector K of
a surface curve at a given point P and in a given direction is
the projection of K onto the tangent space TPS of the surface at
P. This geodesic component of the curvature vector of a
surface curve is given by:
(323)
Kg =
κgu =
[(d2u1 ⁄ ds2) + Γ1αβ(duα ⁄ ds)(duβ ⁄ ds)]E1 + [(d2u2 ⁄ ds2) +
Γ2αβ(duα ⁄ ds)(duβ ⁄ ds)]E2
where the Christoffel symbols are derived from the surface
metric. Since u = n × T, the sense of the geodesic curvature
vector depends on the orientation of the surface and the
orientation of the curve. The geodesic component of the
curvature vector may also be given by the following
expression:
(324)
where the symbols are as explained before.
As well as the previously developed expressions for κg (see
Eqs. 309↑ and 313↑), it can be shown that for a naturally
parameterized curve the geodesic curvature κg is also given by:
(325)
where the Christoffel symbols are derived from the surface
metric and a = EG − F2 is the determinant of the surface
covariant metric tensor. While the curvature κ and the normal
curvature κn are extrinsic properties of the surface, the
geodesic curvature κg is an intrinsic property. This can be seen,
for example, from Eq. 325↑.
On the u1 coordinate curves, du2 ⁄ ds = 0 and du1 ⁄ ds = 1 ⁄ √(E)
(see Footnote 17 in § 8↓). Hence, Eq. 325↑ will simplify to:
(326)
κgu =
√(a) Γ211(du1 ⁄ ds)3 =
[√(a) ⁄ E3 ⁄ 2] Γ211
where κgu is the geodesic curvature of the u1 coordinate curve.
The last formula will be simplified further if the u1 and u2
coordinate curves are orthogonal, since in this case F = 0 and
Γ211 = − Ev ⁄ (2G) (see Eq. 74↑), and the formula will become:
(327) κgu = − Ev ⁄ [2E√(G)]
Similarly, on the u2 coordinate curves, du1 ⁄ ds = 0 and du2 ⁄ ds
= 1 ⁄ √(G) (see Footnote 18 in § 8↓). Hence, Eq. 325↑ will
simplify to:
(328)
κgv =
− √(a) Γ122(du2 ⁄ ds)3 =
− [√(a) ⁄ G3 ⁄ 2] Γ122
where κgv is the geodesic curvature of the u2 coordinate curve.
The last formula will be simplified further if the u1 and u2
coordinate curves are orthogonal, since in this case F = 0 and
Γ122 = − Gu ⁄ (2E) (see Eq. 77↑), and the formula then
becomes:
√
(329) κgv = Gu ⁄ [2G√(E)]
Although the geodesic curvature is an intrinsic property, as can
be seen from the above equations, it can also be calculated
extrinsically by:
(330) κg = [r⋅(n × r)] ⁄ (r⋅r)
where the overdots stand for differentiation with respect to a
general parameter t of the curve. As discussed previously,
some intrinsic properties can also be defined in terms of
extrinsic parameters.
As seen before, the curvature κ and the geodesic curvature κg
of a surface curve at a given point P on the curve are related
by:
(331) κg = κ sinφ
where φ is the angle between the principal normal vector N of
the curve at P and the unit normal vector n of the surface at P.
It should be remarked that the geodesic curvature can take any
real value: positive, negative or zero. However, there are some
details related to the definition of the angle φ in the last
equation and the geodesic normal vector u that should be
considered.
On a surface patch of class C2 with orthogonal coordinate
curves, an s-parameterized curve C of class C2 has a geodesic
curvature given by:
(332) κg = (dθ ⁄ ds) + κgucosθ + κgvsinθ
where κgu and κgv are the geodesic curvatures of the u and v
coordinate curves and θ is the angle such that:
(333) T = (E1 ⁄ | E1|)cosθ + (E2 ⁄ | E2|)sinθ
where T is the tangent unit vector of C, and E1 and E2 are the
surface basis vectors, and where all the given quantities are
evaluated at a given point on the curve. We note that Eq. 332↑
is known as Liouville formula.
4.4 Principal Curvatures and
Directions
On rotating the plane containing n (i.e. the unit normal vector
to the surface at a given point P on the surface) around n, the
normal section and hence its curvature κ and normal curvature
κn at P will vary in general (see Footnote 19 in § 8↓). Based
on the previous findings (see § 4.2↑), the normal curvature κn
(which, for a normal section, is equal in magnitude to its
curvature κ) of the surface at P in a given direction λ can be
given by:
(334) κn = (e + 2fλ + gλ2) ⁄ (E + 2Fλ + Gλ2)
where E, F, G, e, f, g are the coefficients of the first and second
fundamental forms and λ = du2 ⁄ du1. In fact, Eq. 334↑ is a
variant of Eq. 316↑ obtained by dividing the numerator and
denominator of Eq. 316↑ by (du1)2. The directions represented
by (du2 ⁄ du1) are the directions of the tangents to the normal
sections at P. We note that for ease of notation and expression,
symbols like (du2 ⁄ du1) and du:dv are laxly labeled as
directions since a pair like (du1, du2) or (du, dv) represents a
direction.
The two principal curvatures of the surface at P, κ1 and κ2,
which represent respectively the maximum and minimum
values of the normal curvature κn of the surface at P as given
by Eq. 334↑, correspond to the two λ roots of the following
quadratic equation:
(335) (gF − fG)λ2 + (gE − eG)λ + (fE − eF) = 0
where (gF − fG) ≠ 0. The last equation is obtained by equating
the derivative of κn (as given by Eq. 334↑) with respect to λ to
zero to obtain the extremum values.
Eq. 335↑ possesses two roots, λ1 and λ2, which according to the
rules of polynomials are linked by the following relations:
(336)
λ1 + λ2 = (gE − eG) ⁄ (gF − fG)
λ1λ2 = (fE − eF) ⁄ (gF − fG)
where (gF − fG) ≠ 0. These roots represent the two directions
corresponding to the two principal curvatures, κ1 and κ2, of the
surface at the given point, as indicated above.
The following two vectors on the surface, which are defined in
terms of the two λ roots of the above quadratic equation, define
the spatial directions corresponding to the principal curvatures:
(337)
(dr ⁄ du1)1 =
(∂r ⁄ ∂u1) + λ1(∂ r ⁄ ∂u2) =
E1 + λ1 E2
(338)
(dr ⁄ du1)2 =
(∂r ⁄ ∂u1) + λ2(∂ r ⁄ ∂u2) =
E1 + λ2 E2
These directions, which are called the principal directions or
the curvature directions of the surface at point P, are
orthogonal at non-umbilical points where κ1 ≠ κ2. At umbilical
points (see § 4.10↓), the normal curvature is the same in all
directions and hence there are no principal directions to be
orthogonal or every direction is a principal direction and hence
there is no sensible meaning for being orthogonal.
Consequently, at any point on a plane surface all directions are
principal directions or, alternatively, there is no principal
direction (depending on allowing more than two principal
directions or not). Similarly, at any point on a sphere all
directions are principal directions or there is no principal
direction.
It is noteworthy that the principal directions are invariant with
respect to permissible changes in surface representation and
parameterization. However, we remark that although the
principal directions in a given coordinate system of the
ambient space are fixed and hence their position and
orientation relative to the surface are invariant, their position
and orientation with respect to the surface coordinates depend
on the coordinate system employed to represent the surface.
The positions of the centers of curvature (see § 2.6↑) of the
normal sections corresponding to the two principal curvatures
at a given point P on a surface S are given in tensor notation
by:
(339) xi1 = xiP + (Ni1 ⁄ |κ1|)
(340) xi2 = xiP + (Ni2 ⁄ |κ2|)
where xi1 and xi2 are the spatial coordinates of the first and
second center of curvature corresponding to the two principal
curvatures, xiP are the spatial coordinates of P, Ni1 and Ni2 are
the principal normal vectors of the two normal sections
corresponding to the two principal curvatures, κ1 and κ2 are the
principal curvatures of S at P, and i = 1, 2, 3.
We note that the principal normal vector N of a normal section
at a given point P on the surface is collinear with the unit
normal vector n to the surface at P and hence the principal
normal vectors are in the same orientation for all normal
sections at P. However, the two principal normal vectors
corresponding to the two principal curvatures may be parallel
or anti-parallel and hence we labeled them differently to be
general (see Footnote 20 in § 8↓). We also remark that the two
normal sections in the principal directions at P are called the
principal normal sections of the surface at P, while the centers
of curvature of these principal normal sections are described as
the principal centers of curvature of the surface at P.
According to one of the Euler theorems, the normal curvature
κn at a given point P on a surface of class C2 in a given
direction can be expressed as a combination of the principal
curvatures, κ1 and κ2, at P as:
(341) κn = κ1cos2θ + κ2sin2θ
where θ is the angle between the principal direction of κ1 at P
and the given direction. Since the principal directions at non-
umbilical points are orthogonal, θ could represent the angle
with the other principal direction but with relabeling of the two
kappas.
There are a number of invariant parameters of the surface at a
given point P on the surface which are defined in terms of the
principal curvatures at P; these include:
1. The principal radii of curvature: R1 = |1 ⁄ κ1| and R2 = |1 ⁄ κ2|.
2. The Gaussian curvature: K = κ1κ2.
3. The mean curvature: H = (κ1 + κ2) ⁄ 2.
Table 1↓ shows the restricting conditions on the principal
curvatures, κ1 and κ2, for a number of common surfaces with
simple geometric shapes (plane, cylinder, sphere, ellipsoid and
hyperboloid of one sheet) and the effect on the Gaussian
curvature K and the mean curvature H.

Table 1 The limiting conditions on the principal curvatures,

κ1 and κ2, for a number of surfaces of simple geometric shapes
(“Hyp. 1S” stands for hyperboloid of one sheet) alongside the
corresponding mean curvature H and Gaussian curvature K.
Apart from the plane, the unit normal vector to the surface, n,
is assumed to be in the outside direction.
__________________
κ1 κ2 H K
Plane 0 0 0 0
Cylinder κ1 = 0 κ2 < 0 H<0 0
Sphere κ1 = κ2 < 0 κ2 = κ1 < 0 H<0 K>0
Ellipsoid κ1 < 0 κ2 < 0 H<0 K>0
Hyp. 1S κ1 > 0 κ2 < 0 --- K<0

The Gaussian curvature may also be called the “Riemannian

curvature”. However, the “Riemannian curvature” is usually
used to label this type of curvature for general nD spaces while
the “Gaussian curvature” is being used to label the special
instance of it that applies to 2D spaces, and hence the Gaussian
curvature is the Riemannian curvature of surfaces.
It should be remarked that some authors use “total curvature”
for the “Gaussian curvature” and hence these two terms are
synonym, while others use “total curvature” for the area
integral
∬ K dσ
as used, for example, in the Gauss-Bonnet theorem (refer to §
4.8↓). In the present book, we use total curvature strictly for
the integral and hence we label the Gaussian curvature with K
and the total curvature with Kt. Another remark is that some
authors define H as the sum of κ1 and κ2, that is: H = κ1 + κ2,
rather than the average as defined above. In fact each one of
these conventions has its merit. However, in the present book
we define H as the average, not the sum, of the two principal
curvatures.
In the neighborhood of a given point on a surface, the surface
can be approximated by a quadratic expression involving the
principal curvatures at that point. More formally, let P be a
point on a sufficiently smooth surface S embedded in a 3D
space coordinated by a rectangular Cartesian system (x, y, z)
with P being above the origin, the tangent plane of S(x, y) at P
being parallel to the xy plane, and the principal directions being
along the x and y coordinate lines. The equation of S in the
neighborhood of P can then be expressed, up and including the
quadratic terms, in the following form:
≃
(342) S(x, y) S(0, 0) + [(κ1x2) ⁄ 2] + [(κ2y2) ⁄ 2]
where κ1 and κ2 are the principal curvatures of S at P. This
means that in the immediate neighborhood of P, S resembles a
quadratic surface (see § 6.2↓) of the given form. The above
form includes umbilical points (see § 4.10↓) where the
principal directions can be arbitrarily chosen as the directions
of the x and y coordinate lines.
∈
The necessary and sufficient condition for a number κ ℝ to
be a principal curvature of a smooth surface S at a given point
P and in a given direction (dv ⁄ du), where (du)2 + (dv)2 ≠ 0, is
that the following equations are satisfied:
(343) (e − κE)du + (f − κF)dv = 0
(344) (f − κF)du + (g − κG)dv = 0
where E, F, G, e, f, g are the coefficients of the first and second
fundamental forms at P. It is worth noting that for simplicity in
notation, we use κ in these and the following equations to
represent principal curvature. This use should not be confused
with the curve curvature which is also symbolized by κ.
However, for this case the curvature is equal (in magnitude at
least) to the principal curvature since the latter is the curvature
of a normal section and hence the use of κ is justified.
The above equations can be cast in a matrix form as:
(345)

This system of homogeneous linear equations has a non-trivial

solution (du, dv) iff the determinant of the coefficient matrix is
zero, that is:
(346)
(EG − F2)κ2 − (gE − 2fF + eG)κ + (eg − f2) = 0
Based on the given conditions, this quadratic equation in κ has
a non-negative discriminant and hence it possesses either two
distinct real roots or a repeated real root. In the former case
there are two distinct principal curvatures at P corresponding
to two orthogonal principal directions, while in the latter case
the point is umbilical where all the normal sections at the point
have the same “principal curvature” although there is no
specific principal direction since each direction can be a
principal direction. So in brief, a given real number κ is a
principal curvature of S at P iff it is a solution of Eq. 346↑.
From Eq. 346↑, it can be seen that the principal curvatures of a
surface at a given point P are the solutions of this quadratic
equation and hence they are given by:
(347)
On dividing Eq. 346↑ by a = EG − F2 (which is positive
definite as established before) we obtain:
(348) κ2 − 2Hκ + K = 0
where H and K are the mean and Gaussian curvatures whose
expressions in terms of the coefficients of the first and second
fundamental forms are taken from Eqs. 383↓ and 356↓. Hence,
Eq. 347↑ can be expressed compactly as:
(349) κ1, 2 = H±√(H2 − K)
In fact, this formula can be obtained directly from Eq. 348↑
using the quadratic formula.
The above conditions about the principal curvatures may be
stated rather differently in terms of the principal directions that
is, for a non-umbilical point P on a sufficiently smooth surface
S, a direction (du2 ⁄ du1) is a principal direction of S at P iff the
following condition is true:
(350) (fE − eF)du1du1 + (gE − eG)du1du2 + (gF − fG)du2du2 =
0
The last equation, which is obtained from Eq. 335↑ by
multiplying both sides with (du1)2, can be factored into two
linear equations each of the form:
A du1 + B du2 = 0
(with A and B being real constants) where these equations
represent the two orthogonal principal directions.
Similarly, at a given non-umbilical point P on a sufficiently
smooth surface S, a direction (dv ⁄ du) is a principal direction iff
for a real number κ the following relation holds true:
(351) dn = − κ dr
where:
(352)
dn = (∂n ⁄ ∂u)du + (∂n ⁄ ∂v)dv
dr = (∂r ⁄ ∂u)du + (∂r ⁄ ∂v)dv
If this condition is satisfied, then κ is the principal curvature of
S at P corresponding to the principal direction (dv ⁄ du). Eq.
351↑ is known as the Rodrigues curvature formula. The
obvious interpretation of the Rodrigues formula is that in any
principal direction the two vectors dn and dr have the same
orientation where the principal curvature κ in that direction is
the scale factor between the two vectors. From the Rodrigues
curvature formula, the following subsidiary equations
corresponding to the surface coordinate curves can be easily
obtained:
(353)
∂n ⁄ ∂u = − κE1
∂n ⁄ ∂v = − κE2
On each non-umbilical point P of a smooth surface S an
orthonormal moving “Darboux frame” can be defined. This
frame consists of the vector triad (d1, d2, n) where d1 and d2
are the unit vectors corresponding to the principal directions at
P, and n = d1 × d2 is the unit normal vector to the surface at P.
This is another moving frame in use in differential geometry in
addition to the three previously-described frames: the (T,N,B)
frame, the (E1, E2, n) frame and the (n, T, u) frame (see §
1.4.5↑, 2.5↑, 3.2↑ and 4.1↑). The first of these frames, i.e.
(T,N,B), is associated with curves while the remaining three
are associated with surfaces. What is common to all these four
frames is that they are moving frames whose vectors can be
used as basis sets for the embedding 3D space since each one
of these sets consists of three linearly independent vectors.
Also, all these sets, except (E1, E2, n), are orthonormal.
When the u and v coordinate curves of a surface at a given
point P are aligned along the principal directions at P, the
principal curvatures at P will be given by (see § 4.2↑):
(354)
κ1 = b11 ⁄ a11 = e ⁄ E
κ2 = b22 ⁄ a22 = g ⁄ G
where the indexed a and b are the coefficients of the surface
covariant metric and covariant curvature tensors, and E, G, e, g
are the coefficients of the first and second fundamental forms
at P. This may be obtained from Eq. 316↑ where the last two
terms in the sums will vanish for the u coordinate curve since
dv = 0 while the first two terms in the sums will vanish for the
v coordinate curve since du = 0. We note that we are assuming
here a particular labeling of the u and v coordinate curves for
the labeling of the two kappas to be appropriate, i.e. the u
coordinate curve is aligned along the first principal direction
and the v coordinate curve is aligned along the second
principal direction.
It should be remarked that on an oriented and sufficiently
smooth surface, the principal curvatures, κ1 and κ2, are
continuous functions of the surface coordinates. Another
remark is that the principal curvatures are the eigenvalues of
the mixed type surface curvature tensor bαβ.
4.5 Gaussian Curvature
The Gaussian curvature, which may also be called the
Riemannian curvature of the surface, represents a
generalization of curve curvature to surfaces since it is the
product of two curvatures of curves embedded in the surface
and hence in this sense it is a 2D curvature. As given earlier,
the Gaussian curvature K at a given point P on a surface is
defined as the product of the two principal curvatures, κ1 and
κ2, of the surface at P that is:
(355) K ≡ κ1κ2
The Gaussian curvature of a surface at a given point P on the
surface is given by:
(356)
K=
(eg − f2) ⁄ (EG − F2) =
b ⁄ a = R1212 ⁄ a
where E, F, G, e, f, g are the coefficients of the first and second
fundamental forms at P, a and b are the determinants of the
surface covariant metric and covariant curvature tensors, and
R1212 is the component of the 2D covariant Riemann-
Christoffel curvature tensor. From Eq. 231↑ we have:
(357)
R1212 =
b11b22 − b12b21 =
eg − f2 = b
where the indexed b are the coefficients of the surface
covariant curvature tensor, and hence the equalities in Eq. 356↑
are fully justified. As discussed previously (see § 1.4.10↑), the
2D Riemann-Christoffel curvature tensor has only one
independent non-vanishing component which is represented by
R1212. Therefore, Eq. 356↑ provides a full link between the
Gaussian curvature and the Riemann-Christoffel curvature
tensor.
The above formulae (Eq. 356↑) are also based on the fact that
the Gaussian curvature K is the determinant of the mixed
curvature tensor bαβ of the surface, that is:
(358)
K=
det(bαβ) =
det(aαγbγβ) =
det(aαγ)det(bγβ) =
det(bγβ) ⁄ det(aαγ) = b ⁄ a
where the symbols are as defined previously. From Eq. 356↑,
we see that the sign of K (i.e. K > 0, K < 0 or K = 0) is the
same as the sign of b and the sign of R1212 since a is positive
definite. We note that being the determinant of a tensor
establishes the status of K as an invariant under permissible
coordinate transformations. We also note that the chain of
formulae in Eq. 358↑ may be taken in the opposite direction
starting primarily from K = (b ⁄ a) or K = (R1212 ⁄ a) as a
definition or as a derived result from other arguments, and
hence the statement K = det(bαβ) will be obtained as a
secondary result.
Since both R1212 (see Eq. 88↑) and a depend exclusively on the
surface metric tensor, Eq. 356↑ reveals that K depends only on
the first fundamental form coefficients and hence it is an
intrinsic property of the surface (refer to § 4.7↓). The
dependence of K on the second fundamental form coefficients
in Eq. 356↑ or Eq. 358↑ does not affect its qualification as an
intrinsic property since this dependency is not indispensable as
K can be expressed in terms of the first fundamental form
coefficients exclusively. In fact, according to Eq. 262↑ even b
can be expressed exclusively in terms of the first fundamental
form coefficients.
Because the Gaussian curvature is an invariant with respect to
permissible coordinate transformations in 2D manifolds, we
have:
(359) K = R1212 ⁄ a = R̃1212 ⁄ ã
where the barred and unbarred symbols represent the quantities
in the barred and unbarred coordinate systems. The Gaussian
curvature is also invariant with respect to the type of
representation and parameterization of the surface. In
particular, the Gaussian curvature is independent, in sign and
magnitude, of the orientation of the surface which is based on
the choice of the direction of the normal vector n to the
surface. This is because a change in the direction of n will
change the sign of the principal curvatures but not their
absolute value and hence the magnitude is preserved.
Furthermore, this change of sign will not affect the sign of the
Gaussian curvature since both signs will be changed by the
reversal of n direction and hence their product will not be
affected. Therefore, the sign and magnitude of the Gaussian
curvature are both preserved under this reversal.
From Table 1↑ we see that the Gaussian curvature of planes
and cylinders are both identically zero. At the root of this is the
fact that the Gaussian curvature is an intrinsic property and the
cylinder is a developable surface obtained by wrapping a plane
with no localized distortion by stretching or compression.
Hence, the planes and cylinders possess identical first
fundamental forms, as indicated previously in § 3.5↑, and
consequently they have identical Gaussian curvature (also see
§ 4.7↓).
Since the magnitude of the normal curvature of a sphere of
radius R is |κn| = (1 ⁄ R) at any point on its surface and for any
normal section in any direction, its Gaussian curvature is a
constant given by K = (1 ⁄ R2). For a Monge patch of the form
r(u, v) = (u, v, f(u, v)), the Gaussian curvature is given by:
(360) K = (fuufvv − fuv2) ⁄ (1 + fu2 + fv2)2
where the subscripts u and v stand for partial derivatives of f
with respect to these surface coordinates. The last equation can
be obtained by combining Eq. 356↑ (or Eq. 358↑) with Eqs.
201↑ and 226↑.
The Gaussian curvature of a surface of revolution generated by
revolving a plane curve of class C2 having the form y = f(x)
around the x-axis is given by:
(361) K = − fxx ⁄ [f(1 + fx2)2]
where the subscript x represents derivative of f with respect to
this variable.
At any point on a sufficiently smooth surface the Gaussian
curvature satisfies the following relation:
(362) ∂un × ∂vn = K(E1 × E2)
On dot producting both sides with n we obtain:
(363) n⋅(∂un × ∂vn) = Kn⋅(E1 × E2) = K√(a)
where the last equality is based on Eq. 169↑. Hence:
(364) K = [n⋅(∂un × ∂vn)] ⁄ √(a)
In the last equation the Gaussian curvature, which is an
intrinsic property, is expressed in terms of the normal vector n,
which is an extrinsic entity, and its derivatives as well as the
metric tensor.
There are surfaces with constant zero Gaussian curvature K = 0
(e.g. planes, cylinders and cones excluding the apex), surfaces
with constant positive Gaussian curvature K > 0 (e.g. spheres
with K = (1 ⁄ R2) where R is the sphere radius), and surfaces
with constant negative Gaussian curvature K < 0 (e.g. Beltrami
pseudo-spheres, seen in Fig. 14↑, with K = ( − 1 ⁄ ρ2) where ρ is
the pseudo-radius of the pseudo-sphere). However, in general
the Gaussian curvature is a variable function, in sign and
magnitude, of the surface coordinates and hence a single
surface can have Gaussian curvature of different signs and
magnitudes. It is noteworthy that surfaces with constant non-
zero Gaussian curvature K may be described as spherical if K
> 0 and pseudo-spherical if K < 0.
On scaling a surface up or down by a constant factor c > 0, the
Gaussian curvature K will scale by a factor of 1 ⁄ c2. This is
based on the fact that scaling the surface by a constant factor c
> 0 is equivalent to scaling the coefficients of the surface
metric tensor by c2 (see Eq. 187↑) and scaling the coefficients
of the surface curvature tensor by c (see Eq. 214↑), and hence
according to Eq. 356↑ or Eq. 358↑, K will be scaled by a factor
of 1 ⁄ c2. This leads to the conclusion that when the surface
curvature is a non-zero constant, the surface can be scaled up
or down to make its Gaussian curvature 1 or − 1 and hence
simplify the formulations and calculations.
Based on the previous statements plus the fact that the
Gaussian curvature is an intrinsic property, the Gaussian
curvature is invariant with respect to all isometric
transformations since, intrinsically, these transformations
correspond to scaling the surface with unity even though the
shape of the surface may have been deformed extrinsically.
Hence, two isometric surfaces have identical Gaussian
curvature at each pair of their corresponding points. However,
two surfaces with equal Gaussian curvature at their
corresponding points are not necessarily isometric. Yes, in the
case of two sufficiently smooth surfaces with equal constant
Gaussian curvature the two surfaces have local isometry. The
details can be found in more advanced books on differential
geometry.
In 3D manifolds, there is no compact surface of class C2 with
non-positive Gaussian curvature (i.e. K ≤ 0) over the whole
surface. Also, any compact surface, excluding the sphere,
should have points with negative Gaussian curvature. In fact,
the sphere is the only connected, compact and sufficiently
smooth surface with constant Gaussian curvature. According to
the Hilbert lemma, if P is a point on a sufficiently smooth
surface S with κ1 and κ2 being the principal curvatures of S at P
such that: κ1 > κ2, κ1 is a local maximum, and κ2 is a local
minimum, then the Gaussian curvature of S at P is non-
positive, that is K ≤ 0.
At a given point P on a spherically-mapped (see § 3.10↑) and
sufficiently smooth surface S, the ratio of the area of the
spherical image Z̃ of a mapped region Z surrounding P on S to
the area of Z converges to the absolute value of the Gaussian
curvature at P as Z shrinks to P (see Eq. 303↑). From the
Gauss-Bonnet theorem (see § 4.8↓), it can be shown that a
surface will have identically-vanishing Gaussian curvature if at
any point P on the surface there are two families of geodesic
curves (see § 5.7↓) in the neighborhood of P intersecting at a
constant angle.
From Eqs. 93↑ and 356↑, it can be seen that the Gaussian
curvature K of a sufficiently smooth surface represented by r
= r(u, v) = r(u1, u2) can also be given by:
(365) K = [Fuv − (Evv ⁄ 2) − (Guu ⁄ 2) + aαβ(Γα12Γβ12 −
Γα11Γβ22)] ⁄ a
where E, F, G are the coefficients of the first fundamental
form, the subscripts u and v stand for partial derivatives with
respect to these surface coordinates, a = EG − F2 is the
determinant of the surface covariant metric tensor, the indexed
a are its coefficients, and α, β = 1, 2. The Christoffel symbols
in the last equation are based on the surface metric.
The Gaussian curvature of a smooth surface of class C3
represented by r(u, v) may also be given by:
(366)

where the symbols are as defined above. Accordingly, the

Gaussian curvature of a surface of class C3 represented by r(u,
v) with orthogonal surface coordinate curves is given by:
(367)

This formula is obtained from the previous formula by setting

F = 0 identically due to the orthogonality of the surface
coordinate curves. The last formula will simplify to:
(368) K = − [∂uu√(G)] ⁄ √(G)
when the surface r(u, v) is represented by geodesic coordinates
(see § 1.4.8↑) with the u coordinate curves being geodesics and
u is a natural parameter (see Footnote 21 in § 8↓).
The Gaussian curvature K can also be expressed in terms of the
mean curvature H (see § 4.6↓), that is:
(369) K = (H + C)(H − C) = H2 − C2
where C is given by:
(370) C = √([e2G2 + E2g2] − 4fF[eG + Eg] + 4[f2EG + F2eg] −
2egEG) ⁄ [2(EG − F2)]
and E, F, G, e, f, g are the coefficients of the first and second
fundamental forms. This can be verified by transforming Eq.
369↑ to the following form: C2 = H2 − K and substituting for H
and K from Eqs. 383↓ and 356↑.
The Gaussian curvature of a surface S at a given point P on the
surface is positive if all the surface points in a deleted
neighborhood of P on S are on the same side of the tangent
plane to S at P. The Gaussian curvature is negative if for all
deleted neighborhoods of P on S some points are on one side
of the tangent plane and some are on the other side. The
Gaussian curvature is zero if, in a deleted neighborhood, either
all the points lie in the tangent plane or all the points are on
one side except some which lie on a curve in the tangent plane.
Hence:
1. A sphere has positive Gaussian curvature at all points.
2. A hyperbolic paraboloid (Fig. 8↑) has negative Gaussian
curvature at all points. Similarly, the monkey saddle (Fig. 12↑)
has negative Gaussian curvature at all points except the origin
(x, y, z) = (0, 0, 0) which is an umbilical point (see § 4.10↓)
with zero Gaussian curvature.
3. A plane has zero Gaussian curvature at all points.
4. A cylinder has zero Gaussian curvature at all points.
5. A torus (Fig. 38↓) has points with positive Gaussian
curvature (outer half), points with zero Gaussian curvature (top
and bottom circles) and points with negative Gaussian
curvature (inner half).
Figure 38 Points of torus with positive Gaussian curvature
(outer blue), points with zero Gaussian curvature (middle
yellow) and points with negative Gaussian curvature (inner
red).

Based on the above statements, the Gaussian curvature of a

developable surface (see § 6.4↓) is identically zero. Hence,
beside the plane, there are other surfaces with constant zero
Gaussian curvature such as cones, cylinders and tangent
surfaces of space curves (refer to § 6.6↓).
Examples of the Gaussian curvature, K, for a number of simple
surfaces are:
1. Plane: K = 0.
2. Sphere of radius R: K = 1 ⁄ R2.
3. Torus parameterized by
x = (R + rcosφ)cosθ,
y = (R + rcosφ)sinθ and
z = rsinφ:
K = cosφ ⁄ [r(R + rcosφ)].
The total curvature Kt is defined as the area integral of the
Gaussian curvature K over a surface or a patch of a surface, S,
that is:
∬
(371) Kt = S K dσ
where dσ symbolizes infinitesimal area element on the surface
and where K is a function of the surface coordinates in general.
From Eqs. 206↑ and 282↑, it can be seen that the total
curvature Kt may be given by:
(372)
Kt ≡
∬ S Kdσ =
∬ S K|E1 × E2|dudv =
∬ S sgn(K)|∂un × ∂vn|dudv
where sgn(K) is the sign function of K as a function of the
surface coordinates, u and v.
The Riemann-Christoffel curvature tensor is related to the
Gaussian curvature through the absolute permutation tensor of
the surface by the following relation:
(373) Rαβγδ = K εαβ εγδ
where the indexed ε are the 2D covariant absolute permutation
tensors and all the indices range over 1 and 2. On multiplying
both sides of the last equation by εαβεγδ we get:
(374) εαβ εγδ Rαβγδ = K εαβ εγδ εαβ εγδ
Now, since εαβ εαβ = εγδ εγδ = 2, the last equation becomes:
(375)
K=
(1 ⁄ 4) εαβ εγδ Rαβγδ =
(1 ⁄ 4) εαβ εγδ (bαγbβδ − bαδbβγ)
where the indexed b are the components of the surface
covariant curvature tensor, and where the last step is based on
Eq. 227↑. The last equation is a demonstration of the fact that
K is an absolute rank-0 tensor since it is represented in both
equalities by a combination of absolute tensors with all the
indices of these tensors being consumed by contraction.
The Gaussian curvature is also linked to the Riemann-
Christoffel curvature tensor, through the surface metric tensor,
by the following relation:
(376) Rαβγδ = K(aαγaβδ − aαδaβγ)
In fact, Eq. 356↑ is an instance of the last equation with α = γ
= 1 and β = δ = 2. The other combinations of index values
provide the link between K and the other elements of Rαβγδ. We
note that Eq. 376↑ may be extended to nD spaces for n > 2 and
with K (representing Riemannian curvature) being constant but
this is out of the scope of this book.
The Gaussian curvature K may also be given by the following
relation:
(377) K = (1 ⁄ 2) εαβ εγδ bγα bδβ
where the indexed ε are the 2D contravariant absolute
permutation tensors. From Eqs. 227↑ and 373↑, it can be seen
that the Gaussian curvature and the surface curvature tensor are
also related by:
(378) K εαβ εγδ = bαγ bβδ − bαδ bβγ
Other formulae for the Gaussian curvature (in terms of the
surface basis vectors, their derivatives and the coefficients of
the first fundamental form) may also be obtained from the
formula K = (b ⁄ a) by manipulating b as follows:
(379)
b=
eg − f2 =
[(∂uE1⋅ E1 × E2)(∂vE2⋅ E1 × E2) − (∂vE1⋅ E1 × E2)2] ⁄ a =
[(∂uE1⋅ E1 × E2)(∂vE2⋅ E1 × E2) − (∂vE1⋅ E1 × E2)2] ⁄ (EG −
F2) =
[(∂uE1⋅ E1 × E2)(∂vE2⋅ E1 × E2) − (∂vE1⋅ E1 × E2)2] ⁄ | E1 ×
E2|2
where these steps are based on Eqs. 253↑-255↑ and Eq. 170↑
as well as the obvious fact that a = EG − F2. Hence:
(380)
K=
b⁄a=
[(∂uE1⋅ E1 × E2)(∂vE2⋅ E1 × E2) − (∂vE1⋅ E1 × E2)2] ⁄ a2 =
[(∂uE1⋅ E1 × E2)(∂vE2⋅ E1 × E2) − (∂vE1⋅ E1 × E2)2] ⁄ (EG −
F2)2 =
[(∂uE1⋅ E1 × E2)(∂vE2⋅ E1 × E2) − (∂vE1⋅ E1 × E2)2] ⁄ | E1 ×
E2|4
Finally, on a 2D surface, the Gaussian curvature K is related to
the Ricci curvature scalar ℛ (see § 1.4.11↑) by the following
relation:
(381) |K| = |ℛ| ⁄ 2
As seen, the Gaussian curvature is just a constant multiple of
the Ricci curvature scalar of the surface and hence they are
essentially the same. Details about the signs of these curvature
parameters should be sought in more expanded textbooks on
differential geometry and tensor analysis.
4.6 Mean Curvature
The mean curvature of a surface at a given point P is a measure
of the rate of change of area of the surface elements in the
neighborhood of P with respect to the surface coordinates. As
given earlier, the mean curvature H is defined as the average
(see Footnote 22 in § 8↓) of the two principal curvatures, κ1
and κ2, that is:
(382) H ≡ (κ1 + κ2) ⁄ 2
The mean curvature is given by the following formula:
(383)
H=
(eG − 2fF + gE) ⁄ [2(EG − F2)] =
tr(bαβ) ⁄ 2 =
bαα ⁄ 2
where E, F, G, e, f, g are the coefficients of the first and second
fundamental forms, the indexed b represent the surface mixed
curvature tensor, tr stands for the trace of matrix, and α, β = 1,
2. The first equality can be obtained by combining Eq. 382↑
with Eq. 347↑, while the second equality can be verified by
taking the trace of bαβ as given by Eq. 223↑. The third equality
is just a matter of different symbolism according to the matrix
and tensor notations.
Unlike the Gaussian curvature, the sign of the mean curvature
H is dependent on the choice of the direction of the unit normal
vector to the surface, n. This can be seen from Eq. 382↑ where
the signs of both kappas will be reversed by the change of n
direction although the magnitude of kappas, and hence the
magnitude of H, will not be affected by this change. Like the
Gaussian curvature, the mean curvature is invariant under
permissible coordinate transformations and representations as
long as the surface orientation is preserved. Being half the
trace of a tensor establishes the status of H as an invariant
under permissible coordinate transformations.
Examples of the mean curvature, H, for a number of simple
surfaces are:
1. Plane: H = 0.
2. Sphere of radius R: |H| = 1 ⁄ R. As stated above, the sign of
H depends on the choice of n direction being inward or
outward.
3. Torus parameterized by
x = (R + r cosφ) cosθ,
y = (R + r cosφ) sinθ and
z = r sinφ:
|H| = |(R + 2rcosφ) ⁄ (2r[R + rcosφ])|.
The sign of H in this case depends on the location of the point
on the surface as well as the choice of n direction.
For a Monge patch of the form r(u, v) = (u, v, f(u, v)), the mean
curvature is given by:
(384) H = [(1 + fv2)fuu − 2fufvfuv + (1 + fu2)fvv] ⁄ [2(1 + fu2 +
fv2)3 ⁄ 2]
where the subscripts u and v stand for partial derivatives of f
with respect to these surface coordinates. This equation can be
obtained from the first equality of Eq. 383↑ where the
coefficients of the first and second fundamental forms of
Monge patch are obtained from Eqs. 201↑ and 226↑.
The mean curvature may be considered as the 2D equivalent of
the geodesic curvature in 1D. The equivalence can be
understood in the sense that the mean curvature is a measure
for extremizing surface area while the geodesic curvature is a
measure for extremizing curve length. Accordingly, the 2D
minimal surfaces (see § 6.7↓) correspond to the 1D geodesic
curves (see § 5.7↓).
4.7 Theorema Egregium
The essence of Gauss Theorema Egregium or Remarkable
Theorem is that the Gaussian curvature K of a surface is an
intrinsic property of the surface and hence it can be expressed
as a function of the coefficients of the first fundamental form
and their partial derivatives alone with no involvement of the
coefficients of the second fundamental form. This can be
guessed for example from the last part of Eq. 356↑. In fact,
even the first part of Eq. 356↑ can be used in this argument
since b can be expressed purely in terms of the coefficients of
the first fundamental form and their derivatives according to
Eq. 262↑.
The essence of Theorema Egregium, as a statement of the fact
that certain types of curvature are intrinsic to the surface, is
contained in several forms and equations; some of which are
indicated in this book when they occur. For example, Eq. 227↑
which links the surface curvature tensor to the Riemann-
Christoffel curvature tensor (which is an intrinsic property of
the surface and is related to the Gaussian curvature by Eq.
373↑ for instance) can be regarded as a statement of Theorema
Egregium since it expresses a form of surface curvature
represented by a certain combination of the coefficients of the
curvature tensor in terms of a combination of purely intrinsic
surface parameters.
An example may be given to demonstrate the significance of
Theorema Egregium that is, if a piece of plane is rolled into a
cylinder of radius R, then κ1, κ2, H will change from
0, 0, 0
to
(1 ⁄ R), 0, (1 ⁄ [2R])
where, for the cylinder, we are assuming a normal unit vector n
in the inner direction. However, as a consequence of Theorema
Egregium, K will not change since K is dependent exclusively
on the first fundamental form which is the same for planes and
cylinders as stated previously.
According to Theorema Egregium, the Gaussian curvature of a
sufficiently smooth surface of class C3 at a given point P can
be represented by the following function of the coefficients of
the first fundamental form and their partial derivatives at P:
(385)
where
C = ( − Evv + 2Fuv − Guu) ⁄ 2
and the subscripts u and v stand for partial derivatives with
respect to these surface coordinates. The other symbols and
notations are as defined previously.
4.8 Gauss-Bonnet Theorem
This theorem ties the geometry of surfaces to their topology.
There are several variants of this theorem; some of which are
local while others are global. Due to the importance and
subtlety of this theorem we give two variants of the theorem
and several examples from both plane and twisted surfaces.
According to the Gauss-Bonnet theorem, if D is a simply
connected region on a surface of class C3 where D is bordered
by a finite number m of piecewise regular curves Cj that meet
in n corners then we have:
(386)

where the first sum is over the curves while the second sum is
over the corners, κg is the geodesic curvature of the curves Cj
as a function of their coordinates, φk are the exterior angles of
the corners and K is the Gaussian curvature of D as a function
of the coordinates over D. The geodesic and Gaussian
curvatures in the above formulation should be continuous and
finite over their domain. As indicated previously, the term
∬ DKdσ, which represents the area integral of the Gaussian
curvature over the region D of the above-described surface, is
called the total curvature Kt of D.
We note that the corners indicated in the last paragraph can be
defined as the points of discontinuity of the tangents of the
boundary curves. The angles of these corners are therefore
defined as the angles between the tangent vectors at the points
of discontinuity when traversing the boundary curves in a
predefined sense. As indicated above, these angles are exterior
to the region surrounded by the curves. Sometimes, “artificial
corners” at regular points are introduced for convenience to
establish an argument; in which case the exterior angle is zero.
Several examples related to artificial corners will be given in
the forthcoming parts of the book.
It should be remarked that the form of the Gauss-Bonnet
theorem given by Eq. 386↑ may be labeled as a local variant of
the theorem although its locality may not be obvious.
However, it can be justified by its application in principle to a
part of the surface in comparison to the forthcoming global
variant of the theorem which applies to the whole surface and
involves the Euler characteristic and the genus of the surface
which are global features of the surface. Anyway, these labels
are not of crucial importance as long as the theorem and its
significance are understood and appreciated.
Some examples for the application of the above form of the
Gauss-Bonnet theorem are given below:
1. A disc in a plane with radius R where Eq. 386↑ becomes:
(387)
(1 ⁄ R)2πR + 0 + 0 =
2π + 0 + 0 ≡ 2π
which is an identity.
2. A semi-circular disc in a plane with radius R where Eq. 386↑
becomes:
(388)
[(1 ⁄ R)πR + 0 × 2R] + 2(π ⁄ 2) + 0 =
π + π + 0 ≡ 2π
which is an identity again.
3. A spherical triangle (Fig. 39↓) on a sphere of radius R whose
sides are two half meridians connecting a pole to the equator
and one quarter of an equatorial parallel and all of its three
corners are right angles where Eq. 386↑ becomes:
(389)
[0(3{πR ⁄ 2})] + 3(π ⁄ 2) + (1 ⁄ R2)(4πR2 ⁄ 8) =
0 + (3π ⁄ 2) + (π ⁄ 2) ≡ 2π
4. The upper half of a sphere (or a hemisphere in general) of
radius R where Eq. 386↑ becomes:
(390)
0(2πR) + 0 + (1 ⁄ R2)2πR2 =
0 + 0 + 2π ≡ 2π
Figure 39 A spherical triangle with three right angles on the
surface of a sphere. The three sides of this spherical triangle
are arcs of great circles.

The fact that the sum of the interior angles of a planar triangle
is equal to π can also be regarded as an instance of the Gauss-
Bonnet theorem since for a planar triangle Eq. 386↑ becomes:
(391)
0 + Σ3i = 1 (π − θi) + 0 =
3π − Σ3i = 1 θi = 2π
where θi are the interior angles of the triangle and hence Σ3i =
1θi = π as it should be.
By a similar argument, we can obtain the sum of the interior
angles of a planar polygon of n sides (n > 2) using the Gauss-
Bonnet theorem, that is:
(392)
0 + Σni = 1 (π − θi) + 0 =
nπ − Σni = 1 θi = 2π
where θi are the interior angles of the n-polygon and hence Σni
= 1θi = (n − 2)π as it should be.
The fact that the perimeter of a planar circle of radius R is 2πR
can be regarded as another instance of the Gauss-Bonnet
theorem since for such a planar circle Eq. 386↑ becomes:
(393) (1 ⁄ R)L + 0 + 0 = 2π
where L is the length of the circle perimeter and hence L = 2πR
which is the required result.
As a result of the Gauss-Bonnet theorem, the sum θs of the
interior angles of a geodesic triangle on a surface with
Gaussian curvature of constant sign is:
1. θs < π iff K < 0.
2. θs = π iff K = 0.
3. θs > π iff K > 0.
Figure 40 Geodesic triangles on a surface with negative
Gaussian curvature (left frame), a surface with zero Gaussian
curvature (middle frame), and a surface with positive Gaussian
curvature (right frame).

These cases are depicted in Fig. 40↑. This shows that the total
curvature provides the excess over π for the sum when K > 0
on the surface and the deficit when K < 0. The vanishing total
curvature in the case of K = 0 is the intermediate case where
the total curvature term has no contribution to the sum. This
can be seen from Eq. 386↑ which, for a geodesic triangle, will
reduce to:
∬
(394) 0 + (3π − θs) + DKdσ = 2π ⇒ ∬
θs = π + DKdσ
We remark that “geodesic triangle” is a triangle with geodesic
sides and hence κg = 0 identically over its boundary (see §
5.7↓). Also, “triangle” here and in the spherical triangle
example related to Fig. 39↑ is more general than a three-side
planar polygon with three straight segments as it can be on a
curved surface with curved non-planar sides.
As a consequence of the findings in the last paragraph, two
geodesic curves on a simply connected patch of a surface with
negative Gaussian curvature cannot intersect at two points
because on introducing a vertex at a regular point on one curve
we will have an artificial corner with zero exterior angle and
hence π interior angle. We will then have a geodesic triangle
with θs > π on a surface over which K < 0, in violation of the
above-stated condition. By a similar argument to the argument
in the previous paragraph, the area of a geodesic polygon on a
surface with constant non-zero Gaussian curvature is
determined by the sum of the polygon interior angles θs. This
can also be seen from Eq. 386↑ which in this case will be
reduced to:
 ∬
(395) 0 + (nπ − θs) + K Ddσ = 2π ⇒ ∬ Ddσ = [θs + (2
− n)π] ⁄ K
where n > 2 is the number of sides of the polygon. As for
geodesic triangle, “geodesic polygon” is a polygon with
geodesic sides and hence κg = 0 identically over its boundary.
Again, “polygon” here is general and hence it includes
curvilinear polygon on curved surface with curved non-planar
sides.
It is worth noting that because the geodesic curvature is an
intrinsic property, as discussed in § 4.3↑, the Gauss-Bonnet
theorem, as given by Eq. 386↑, is another indication to the fact
that the Gaussian curvature (as well as the total curvature) is an
intrinsic property and hence it is another demonstration of
Theorema Egregium (see § 4.7↑).
The Gauss-Bonnet theorem has also a global variant which
links the Euler characteristic χ, which is a topological invariant
of the surface, to the Gaussian curvature K, which is a
geometric invariant of the surface. This global form of the
Gauss-Bonnet theorem states that: on a compact orientable
surface S of class C3 these two invariants are linked through
the following equation:
∬
(396) S Kdσ = 2πχ
Now, since χ is a topological invariant of the surface, Eq. 396↑
reveals that the total curvature is also a topological invariant of
the surface.
The global Gauss-Bonnet theorem can be used to determine the
total curvature Kt of a surface. For example, the Euler
characteristic of a sphere is 2 and hence from Eq. 396↑ it can
be concluded that its total curvature is Kt = 4π with no need for
evaluating the area integral. Similarly, the Euler characteristic
of a torus is 0 and hence it can be concluded immediately that
its total curvature is Kt = 0 with no need for evaluating the
integral. The Euler characteristics of the sphere and torus in
these examples can be obtained easily by polygonal
decomposition, as described in § 1.4.1↑. For example, the
Euler characteristic of the sphere can be calculated by dividing
the surface of the sphere to 4 curved polygonal faces with 4
vertices and 6 edges and hence the Euler characteristic is:
(397) χ = V + ℱ − ℰ = 4 + 4 − 6 = 2
as seen in Fig. 16↑.
The global Gauss-Bonnet theorem can also be used in the
opposite direction, that is it may be used for determining the
Euler characteristic of a surface knowing its Gaussian, and
hence total, curvature although in most cases this may be of
little use practically. For instance, the Gaussian curvature of a
sphere of radius R is (1 ⁄ R2) at every point on the sphere and
hence its total curvature is
Kt = Kσ = (1 ⁄ R2)4πR2 = 4π,
therefore from Eq. 396↑ its Euler characteristic is
χ = (4π) ⁄ (2π) = 2.
The Gauss-Bonnet theorem can also be used to find the total
curvature of a smooth surface which is topologically-
equivalent (i.e. homeomorphic) to another surface with known
total curvature without need for any calculation. For example,
the ellipsoid (Fig. 4↑) is homeomorphic to the sphere and
hence they have the same Euler characteristic. Therefore,
according to Eq. 396↑ they have the same total curvature
which is 4π as known from the aforementioned sphere
example. As a consequence, the total curvature Kt of a smooth
surface with a complex shape can be obtained from the Gauss-
Bonnet theorem by reducing the surface to a topologically-
equivalent simpler surface whose total curvature can be
evaluated promptly.
As seen before (refer to § 1.4.1↑), for an orientable surface of
genus g the Euler characteristic is given by:
χ = 2(1 − g),
and hence its total curvature is given by:
(398) Kt = 2πχ = 4π(1 − g)
So, for a compact orientable complexly-shaped surface which
can be reduced to a sphere with 2 handles the total curvature is
Kt = 4π(1 − 2) = − 4π.
Similarly, the genus of a torus is g = 1 and hence its total
curvature is
Kt = 4π(1 − 1) = 0,
as found earlier by another method. Hence, the total curvature
of any complexly-shaped surface that can be reduced to a torus
is zero.
The important and obvious implication of the global variant of
the Gauss-Bonnet theorem that can be concluded from the
previous discussion is that the total curvature of a closed
surface is dependent on its genus and Euler characteristic and
not on its geometric shape and hence it is a topological
parameter of the surface as stated before.
4.9 Local Shape of Surface
Using the principal curvatures, κ1 and κ2, a point P on a
surface is classified according to the shape of the surface in the
close proximity of P as:
1. Flat when κ1 = κ2 = 0, and hence K = H = 0 (see Eqs. 355↑
and 382↑).
2. Parabolic when either κ1 = 0 and κ2 ≠ 0 or κ2 = 0 and κ1 ≠ 0,
and hence K = 0 and H ≠ 0.
3. Elliptic when either κ1 > 0 and κ2 > 0 or κ1 < 0 and κ2 < 0,
and hence K > 0.
4. Hyperbolic when κ1 > 0 and κ2 < 0, and hence K < 0.
These constraints on κ1 and κ2, and hence on K and H, are
sufficient and necessary conditions for determining the type of
the surface point as described above.
The following are some examples for the above classification:
1. The points of plane are flat.
2. The points of cone (excluding the apex) and the points of
cylinder are parabolic.
3. The points of ellipsoid (Fig. 4↑) are elliptic.
4. The points of catenoid (Fig. 10↑) are hyperbolic.
Surfaces normally contain points of different shapes. For
example, the torus has elliptic points on its outside half,
parabolic points on its top and bottom parallels (see Footnote
23 in § 8↓), and hyperbolic points on its inside half, as seen in
Fig. 38↑. However, there are some types of surface whose all
points are of the same shape; e.g. all points of planes are flat,
all points of spheres are elliptic, all points of catenoids are
hyperbolic, and all points of cylinders are parabolic.
The above classification regarding the local shape can also be
based on the determinant b of the covariant curvature tensor
and the coefficients e, f, g of the second fundamental form of
the surface where:
1. b = eg − f2 = 0 and e = f = g = 0 for flat points.
2. b = eg − f2 = 0 and e2 + f2 + g2 ≠ 0 for parabolic points.
3. b = eg − f2 > 0 for elliptic points.
4. b = eg − f2 < 0 for hyperbolic points.
Considering Eqs. 356↑ and 383↑ plus the fact that the first
fundamental form is positive definite and hence a > 0, this
classification which is based on b and e, f, g is equivalent to the
previous classification which is based on K and H.
The above classification of the shape of a surface in the
immediate neighborhood of a point (i.e. being flat, parabolic,
elliptic or hyperbolic) is an invariant property with respect to
permissible coordinate transformations. This can be concluded
from the dependence of the classification on the sign of b as
explained above, plus Eq. 221↑ where the square of the
Jacobian (which is real) is positive and hence the sign of b and
b̃ is the same. The classification is also independent of the
representation and parameterization of the surface since these
point types are real geometric properties of the surface in their
local definitions.
The invariance of the shape type of the surface points, as
explained in the previous statements, holds true even for the
transformations that reverse the direction of the normal vector
to the surface, n, because the classification depends on the
Gaussian curvature which is invariant even under this type of
transformations (refer to § 4.5↑). Regarding the distinction
between the flat and parabolic points which involves H as well,
the distinction is not affected since it depends on the
magnitude of H (i.e. being zero or not) and not on its sign and
the magnitude is not affected by such transformations.
In the immediate neighborhood of an elliptic point P of a
surface S, the surface lies completely on one side of the tangent
plane to S at P (Fig. 41↓), while at a hyperbolic point the
tangent plane cuts through S and hence some parts of S are on
one side of the tangent plane while other parts are on the other
side (Fig. 42↓). In the neighborhood of a parabolic point, the
surface lies entirely on one side of the tangent plane except for
some points on a curve which lies in the tangent plane itself
(Fig. 43↓) (see Footnote 24 in § 8↓). As for planar points, the
neighborhood of the point lies in the tangent plane.
Figure 41 Tangent plane at an elliptic point.
Figure 42 Tangent plane at a hyperbolic point.
Figure 43 Tangent plane at a parabolic point.

The surface points can also be classified according to the

geometric shape of Dupin indicatrix (refer to § 3.6.1↑) as
follows:
1. If eg−f2 = 0 and e = f = g = 0, then the point is flat and the
Dupin indicatrix is not defined. Hence, having undefined
Dupin indicatrix is a characteristic for planar points. The
normal curvature at the point is zero in all directions.
2. If eg−f2 = 0 and e2 + f2 + g2 > 0, then either κ1 = 0 and κ2 ≠ 0
or κ2 = 0 and κ1 ≠ 0; hence the point is parabolic and the Dupin
indicatrix becomes two parallel lines. The point is
characterized by having a vanishing normal curvature along
the direction of these lines while it has the same sign in all
other directions.
3. If eg−f2 > 0 then κ1 and κ2 have the same sign; hence the
point is elliptic and the Dupin indicatrix is an ellipse or circle.
The normal curvature at the point has the same sign in all
directions.
4. If eg−f2 < 0 then κ1 and κ2 have opposite signs; hence the
point is hyperbolic and the Dupin indicatrix becomes two
conjugate hyperbolas. The normal curvature at the point is
positive along the directions corresponding to one of these
hyperbolas and negative along the directions corresponding to
the other hyperbola, while along the common asymptotes of
these hyperbolas the normal curvature is zero.
In brief, because of these correlations between the type of point
and the shape of its Dupin indicatrix, the Dupin indicatrix can
be used to classify the point as flat, parabolic, elliptic or
hyperbolic. It is worth noting that the relation between eg−f2
and κ1 and κ2 as stated in the above bullet points can be
concluded from the fact that (see Eqs. 355↑ and 356↑):
(399) K = κ1κ2 = (eg−f2) ⁄ a
since a is positive definite.
We remark that in the immediate neighborhood of a point on a
surface, the surface may be approximated by:
1. A plane at a flat point.
2. A parabolic cylinder (Fig. 9↑) at a parabolic point.
3. An elliptic paraboloid (Fig. 7↑) at an elliptic point.
4. A hyperbolic paraboloid (Fig. 8↑) at a hyperbolic point.
Another remark is that in the neighborhood of a parabolic point
P on a surface S, the tangent plane of S at P meets S in a single
line passing through P, while in the neighborhood of a
hyperbolic point P on a surface S, the tangent plane meets S in
two lines intersecting at P where these two lines divide S
alternatively into regions above the tangent plane and regions
below the tangent plane. Finally, the following function:
(400) IIS ⁄ 2 = (e dudu + 2f dudv + g dvdv) ⁄ 2
evaluated at a given point P of a class C2 surface may be called
the osculating paraboloid of P. This osculating paraboloid,
represented by half the second fundamental form, is used to
determine the shape of the surface at P (also see § 4.2↑).
4.10 Umbilical Point
A point on a surface is called “umbilical” or “umbilic” or
“navel” if all the normal sections of the surface at the point
have the same normal curvature κn. Hence, at umbilical points
we have the following condition:
(401) κ1 = κ2
As stated before, for normal sections the normal curvature κn is
equal in magnitude to the curvature κ. Therefore, the curvature
of all the normal sections at umbilical points is also equal. The
condition of Eq. 401↑ implies that an umbilical point cannot be
a hyperbolic point because at hyperbolic points we should have
κ1 > 0 and κ2 < 0 (refer to § 4.9↑). Hence, at umbilical points
the Gaussian curvature should satisfy the necessary (but not
sufficient) condition: K ≥ 0.
The following are some examples of umbilical points on
common surfaces:
1. All points of planes are umbilical. However, some authors
impose the condition K > 0 at umbilical points and hence the
points of planes are not umbilical according to these authors.
2. All points of spheres are umbilical. Hence, umbilical points
may be called spherical points.
3. The vertex of an elliptic paraboloid of revolution is an
umbilical point.
4. The two vertices of an ellipsoid of revolution are umbilical
points.
If all points of a surface of class C3 are umbilical then the
surface must be a sphere. The plane is a special case of sphere
as it can be regarded as a sphere with an infinite radius.
A sufficient and necessary condition for a given point P to be
umbilical is that the coefficients of the curvature tensor bαβ at
P are proportional to the corresponding coefficients of the
metric tensor aαβ at P, that is:
(402) bαβ = c aαβ
where c, which is a proportionality factor, is independent of the
direction of the tangent to the normal section at the umbilical
point and α, β = 1, 2. In fact, this condition is the same as the
previously stated condition of Eq. 321↑, and hence the same
justification of Eq. 321↑ will apply here. We also note that c =
κn, as seen there.
As a result of Eq. 402↑, at umbilical points the determinants of
the two tensors, a and b, satisfy the following relation:
(403) b = c2a
where c is squared because aαβ is a tensor represented by a 2 ×
2 matrix. Now, since the first fundamental form is positive
definite, and hence a > 0, then if at the umbilical point c = 0
then b = 0 according to Eq. 403↑ and the point is a flat umbilic;
otherwise b > 0 (since c is real) and the point is an elliptic
umbilic (see § 4.9↑) (see Footnote 25 in § 8↓). On a plane
surface all points are flat umbilic, while on a sphere all points
are elliptic umbilic.
As seen earlier, a hyperbolic point cannot be an umbilical point
since at the umbilical point we should have identical normal
curvatures in all directions, both in sign and in magnitude, and
this cannot happen at a hyperbolic point whose κ1 and κ2
should be of opposite signs. This is inline with the fact that at
an umbilical point b cannot be negative since c is real. In fact,
Eq. 403↑ can be recast into the following form:
(404) K = b ⁄ a = c2 = (κn)2
where the equation: c = κn and Eq. 356↑ are used. Hence, all
the above-stated facts about the nature of the umbilical point
and the impossibility of being hyperbolic, as represented by the
condition K ≥ 0, are justified.
Because at umbilical points κ1 = κ2, we have:
(405)
K=
κ1κ2 =
κ1κ1 =
(κ1)2 =
[(2κ1) ⁄ 2]2 =
[(κ1 + κ2) ⁄ 2]2 = H2
where K and H are the Gaussian and mean curvatures at the
point (see § 4.5↑ and 4.6↑). This can also be obtained from Eq.
348↑ where the discriminant of this quadratic equation
becomes zero (i.e. 4H2 − 4K = 0), since at an umbilical point
the two roots are equal, and hence H2 = K.
It should be remarked that the provision H2 = K is a sufficient
and necessary condition for a point at which this condition is
satisfied to be umbilical. This can be concluded from the stated
requirements in the last paragraph. Another remark is that the
relation between K and H at umbilical points, as expressed by
Eq. 405↑, may be stated by some authors in the following
disguised form:
(406) (aαβbαβ)2 = (4 ⁄ a)(b11b22 − b212)
where Eqs. 356↑ and 383↑ are employed in this form and α, β
= 1, 2.
4.11 Exercises
Exercise 1. Discuss the similarities and differences between
the curvature of curves and the curvature of surfaces.
Exercise 2. Define, descriptively and mathematically, the
curvature vector K of surface curves and its relation to the
principal normal vector N of the curve.
Exercise 3. Compare the vectors n and N at a point on a
surface curve outlining their similarities and differences.
Exercise 4. Discuss the dependency of the curvature vector of
a surface curve at a given point of the curve on the following
parameters: curve orientation, curve parameterization, surface
orientation as indicated by the direction of n, surface
parameterization, tangential direction and position of the point
on the surface.
Exercise 5. What “inflection point” on a surface curve means?
Exercise 6. What is the radius of curvature at a point of
inflection?
Exercise 7. Resolve the curvature vector of a surface curve
into its tangential and normal components and name these
components. Express these components in terms of the unit
vectors n and u explaining all the symbols involved in this
expression.
Exercise 8. Find the curvature vector, K, of a space curve
represented by:
r(t) = (3t2, t, 2sint).
Exercise 9. Define, descriptively and quantitatively, the normal
and geodesic curvatures κn and κg.
Exercise 10. Which of κn and κg is an intrinsic property and
which is an extrinsic property? Explain why.
Exercise 11. Compare the following four moving frames:
(T, N, B),
(E1, E2, n),
(n, T, u) and
(d1, d2, n)
outlining their similarities and dissimilarities.
Exercise 12. Which of the frames in the previous question
employ both surface and curve vectors? Which of these frames
are orthonormal by definition and which are not?
Exercise 13. How the curvature κ of a surface curve at a given
point is related to its normal and geodesic curvatures κn and κg
at that point? Can you make sense of this considering the
normal and tangential components of the curvature vector K?
Exercise 14. Prove that the geodesic curvature of a naturally
parameterized curve is given by Eq. 325↑.
Exercise 15. Show that in any two orthogonal directions at a
given point P on a sufficiently smooth surface, the sum of the
normal curvatures corresponding to these directions at P is
constant.
Exercise 16. Give a brief statement of the theorem of Meusnier
outlining its significance. State this theorem in a second
alternative form.
Exercise 17. Show that the osculating circles of all curves on a
surface that pass through a given point and in a specific
direction are on a sphere.
Exercise 18. Define, descriptively and mathematically, the
normal component Kn of the curvature vector K of a surface
curve outlining its relation to the curvature vector and the
normal vector to the surface, n.
Exercise 19. Define, descriptively and mathematically, the
geodesic component Kg of the curvature vector K of a surface
curve outlining its relation to the curvature vector and the
surface basis vectors E1 and E2.
Exercise 20. Derive the formula for the normal curvature κn as
a ratio of the second fundamental form to the first fundamental
form.
Exercise 21. Why the sign of the normal curvature κn is
determined only by the sign of the second fundamental form?
Exercise 22. Show that at any point P of a smooth surface S
there exists a paraboloid tangent to S at P such that the normal
curvature of the paraboloid in any direction is equal to the
normal curvature of S at P in that direction.
Exercise 23. Discuss, in detail, the following statement: “The
normal curvature at a given point on a surface and in a given
tangential direction to the surface is a property of the surface”.
Can you link this to the Meusnier theorem?
Exercise 24. For what type of surface curve the following
relation is true: |κn| = κ? Explain why this is so.
Exercise 25. What is the significance of having a paraboloid at
the points of a smooth surface whose normal curvature in a
given direction is equal to the normal curvature of the surface
in that direction?
Exercise 26. What is the sign of b (i.e. being greater than, less
than or equal to zero) at flat, elliptic, parabolic and hyperbolic
points on a surface, where b is the determinant of the surface
covariant curvature tensor?
Exercise 27. Classify the local shape of a surface at a given
point P according to the values of K and H at P.
Exercise 28. Using one of the mathematical definitions of the
geodesic curvature κg, explain why κg should be classified as
an intrinsic or extrinsic property.
Exercise 29. At what type of surface points the following
relation is true:
e⁄E=f⁄F=g⁄G=c
where c is constant for all directions? What c stands for?
Exercise 30. Express the equalities in the previous question in
terms of the coefficients of the covariant metric and covariant
curvature tensors, aαβ and bαβ, of the surface.
Exercise 31. Outline two direct consequences of Meusnier
theorem.
Exercise 32. Write a mathematical relation linking the
geodesic component Kg of the curvature vector to the surface
basis vectors E1 and E2.
Exercise 33. What is the relation between Kg and the tangent
space TPS of the surface at a given point?
Exercise 34. Give the formulae of the geodesic curvature κg of
the coordinate curves. Simplify these formulae in the case of
having orthogonal coordinate curves.
Exercise 35. State a mathematical relation between the
curvature κ and the geodesic curvature κg of a surface curve at
a given point on the curve explaining all the symbols involved.
Exercise 36. Give a formula for the geodesic curvature κg in
which extrinsic entities are involved. Does this mean that κg is
an extrinsic property of the surface?
Exercise 37. Explain, in detail, all the symbols used in the
following formula:
κg = (dθ ⁄ ds) + κgucosθ + κgvsinθ
What is the name of this formula?
Exercise 38. Prove the relation given in the last question.
Exercise 39. Give a mathematical formula in which κn is
expressed in terms of the coefficients of the first and second
fundamental forms E, F, G, e, f, g.
Exercise 40. What the two “principal curvatures” of a surface
at a given point mean?
Exercise 41. Find analytical expressions for the principal
curvatures on a surface represented by the equation:
ξ2cosξ3 − ξ1sinξ3 = 0
where ξ1, ξ2, ξ3 are real variables.
Exercise 42. The principal curvatures of a surface at a given
point correspond to the two directions represented by λ1 and λ2
which are the roots of the following quadratic equation:
(gF − fG)λ2 + (gE − eG)λ + (fE − eF) = 0
From the rules of polynomial equations, find the sum and
product of these roots.
Exercise 43. Define the “principal directions” descriptively
and mathematically.
Exercise 44. Show that κ is a principal curvature with a
principal direction (dv ⁄ du) iff the following conditions are
satisfied:
(e − κE)du + (f − κF)dv = 0
(f − κF)du + (g − κG)dv = 0
Exercise 45. Find the principal curvatures and the principal
directions on a surface represented parametrically by:
r(u, v) = (u, v, 2u2 + 5v2)
at the point with (u, v) = (2.3, 1.6).
Exercise 46. Prove Euler theorem (see Eq. 341↑ and the
surrounding text).
Exercise 47. What is Darboux frame? Are the vectors of this
frame orthonormal? Is this frame defined at umbilical points on
the surface? Fully justify your answer related to the last two
parts of the question.
Exercise 48. Write the formulae for the positions of the centers
of curvature of the normal sections corresponding to the two
principal curvatures at a given point on a surface.
Exercise 49. Correlate, mathematically with full explanation of
all the symbols involved, the normal curvature κn at a given
point and in a given direction on a smooth surface to the two
principal curvatures at that point.
Exercise 50. Define, mathematically in terms of the principal
curvatures, the following terms: principal radii, mean curvature
and Gaussian curvature.
Exercise 51. Distinguish between the “total curvature” of a
curve and the “total curvature” of a surface. For surface, what
are the two meanings of this term?
Exercise 52. Find the Gaussian and mean curvatures of a
surface given by:
r(u, v) = (3u − v, u + 2v, 1.5uv)
at the point with (u, v) = (3, 1).
Exercise 53. State the limiting conditions on the principal
curvatures, and hence deduce the conditions on the mean and
Gaussian curvatures, on the surface of sphere and on the
surface of hyperboloid of one sheet.
Exercise 54. Prove that there is no compact surface of class C2
with non-positive Gaussian curvature over the whole surface.
Exercise 55. Analyze the following equation outlining its
significance:
≃
S(x, y) S(0, 0) + [(κ1x2) ⁄ 2] + [(κ2y2) ⁄ 2]
Exercise 56. State the necessary and sufficient condition for a
real number to be a principal curvature of a surface at a given
point.
Exercise 57. Investigate the number of roots of the following
quadratic equation and the impact of this on the number of
principal curvatures of the surface at the point where this
equation applies:
(EG − F2)κ2 − (gE − 2fF + eG)κ + (eg − f2) = 0
Exercise 58. From the equation in the previous question,
obtain an analytical expression for the principal curvatures of
the surface at the point where this equation applies.
Exercise 59. From the equation in the last two questions,
obtain the equation:
κ2 − 2Hκ + K = 0
and hence verify that the principal curvatures are given by:
κ1, 2 = H±√(H2 − K).
Exercise 60. Write down the equations of the principal
curvatures when the u1 and u2 coordinate curves are aligned
along the principal directions.
Exercise 61. Obtain Eq. 367↑ for the Gaussian curvature of a
surface with orthogonal coordinate curves by using Eq. 366↑.
Exercise 62. Show that the spheres are the only connected,
compact and sufficiently smooth surfaces with constant
Gaussian curvature.
Exercise 63. State the curvature formula of Rodrigues defining
all the symbols involved and discussing its significance.
Exercise 64. Test the validity of the Rodrigues formula for the
principal directions at the point with (u, v) = (1.4, 3.9) on a
surface parameterized by:
r(u, v) = (u, v, u2 + 3v2).
Exercise 65. Use the Rodrigues curvature formula to prove
that spheres are the only connected closed surfaces of class C3
whose all points are spherical umbilical.
Exercise 66. Give a mathematical expression for the Gaussian
curvature in terms of the coefficients of the surface metric and
curvature tensors.
Exercise 67. What is the significance of having an intrinsic
surface curvature, represented usually by the Gaussian
curvature, as a way for a 2D inhabitant to have some
perception of the nature of the surface and its shape as seen
from the ambient space by a 3D inhabitant?
Exercise 68. Discuss the following statement: “The Gaussian
curvature along any parallel line of a surface of revolution is
constant”.
Exercise 69. Starting from the following relation: K = (b ⁄ a),
derive the relation: K = det(bαβ).
Exercise 70. Give a mathematical relation correlating the
Gaussian curvature to the following coefficients of the 2D
Riemann-Christoffel curvature tensor: R1212, R1221, R1121 and
R2112.
Exercise 71. What is the Gaussian curvature of a Monge patch
of the form r(u, v) = (u, v, f(u, v))?
Exercise 72. Why the Gaussian curvature is independent of the
orientation of the surface (where orientation is based on the
choice of the direction of the unit normal vector to the
surface)?
Exercise 73. Which of the following geometric shapes have
identical Gaussian curvatures at their corresponding points and
why: plane, sphere, cylinder, catenoid, ellipsoid, hyperbolic
paraboloid, helicoid, and cone? Compare, in your answer, each
pair of these shapes.
Exercise 74. Write down an expression for the Gaussian
curvature of a surface of revolution generated by revolving a
sufficiently differentiable plane curve of the form x = f(y)
around the y-axis.
Exercise 75. Explain all the symbols of the following equation
with discussion of its significance in relation to the intrinsic
and extrinsic geometries of the surface:
∂un × ∂vn = K(E1 × E2)
Exercise 76. State the mathematical expression that correlates
the Gaussian curvature to the Ricci curvature scalar of a
surface.
Exercise 77. Using Eq. 367↑ and the parametric equations of
Beltrami pseudo-sphere (Eqs. 43↑-45↑), show that the pseudo-
sphere has a negative constant Gaussian curvature and find this
curvature.
Exercise 78. Classify surfaces with regard to their Gaussian
curvature as having constant or variable curvature giving two
examples for each.
Exercise 79. What is the impact of scaling a surface up or
down by a constant positive factor on its Gaussian curvature?
Exercise 80. Discuss the effect of an isometric mapping of a
surface on its Gaussian curvature.
Exercise 81. State the Hilbert lemma giving examples for its
applications from common types of surface.
Exercise 82. Give the conditions for the validity of the
following equation:
Also, give its simplified form in the case of representing the
surface by geodesic coordinates stating the other conditions
required for this simplification.
Exercise 83. Express the mean curvature as a function of the
Gaussian curvature taking care of the signs.
Exercise 84. Show that spheres are the only connected
compact surfaces with constant mean curvature and positive
Gaussian curvature.
Exercise 85. Explain how the position of the surface in a
deleted neighborhood of a given point P relative to the tangent
plane of the surface at P is used to classify the nature of the
Gaussian curvature at P. From this perspective, discuss the
sign of the Gaussian curvature on the points of the following
surfaces: hyperbolic paraboloid, sphere, torus and cylinder.
Exercise 86. The Gaussian curvature of a developable surface
is identically zero. Why?
Exercise 87. What is the Gaussian curvature of a surface
parameterized by:
x = (5 + cosφ) cosθ,
y = (5 + cosφ) sinθ and
z = sinφ?
Exercise 88. Provide a mathematical definition for the total
curvature of a surface explaining all the symbols used in the
definition.
Exercise 89. Define all the symbols used in the following
equation:
εαβ εγδ Rαβγδ = K εαβ εγδ εαβ εγδ
Exercise 90. Explain in detail how the following equation
implies that the Gaussian curvature is a rank-0 tensor:
K = (1 ⁄ 4) εαβ εγδ Rαβγδ.
Exercise 91. Write the Gaussian curvature in terms of the
surface curvature tensor using the most simple form.
Exercise 92. Algebraically manipulate the relation K = (b ⁄ a)
to obtain the following relation:
K = [(∂uE1⋅ E1 × E2)(∂vE2⋅ E1 × E2) − (∂vE1⋅ E1 × E2)2] ⁄
(EG − F2)2
Exercise 93. Express the mean curvature H in terms of the
coefficients of the first and second fundamental forms.
Exercise 94. What is the relation between the mean curvature
H and the mixed type surface curvature tensor bβα?
Exercise 95. Compare the sign of the mean curvature to the
sign of the Gaussian curvature with regard to their dependency
on the direction of the unit normal vector to the surface.
Exercise 96. Give two examples of common types of surface
over which the mean curvature is constant. Also, give an
example of a surface with variable mean curvature.
Exercise 97. What is the mean curvature of a Monge patch of
the form r(u, v) = (u, v, f(u, v))?
Exercise 98. What is the essence of Gauss Theorema
Egregium? Give an example of an equation or a theorem that
demonstrates this theorem.
Exercise 99. Derive Eq. 385↑ using Eq. 380↑.
Exercise 100. Write down the mathematical equation
representing the local form of the Gauss-Bonnet theorem
explaining all the symbols involved.
Exercise 101. Give an example for the application of the local
Gauss-Bonnet theorem using a planar geometric shape and
another example using a non-planar shape.
Exercise 102. Explain why two geodesic curves on a patch of a
surface with negative Gaussian curvature cannot intersect at
two points.
Exercise 103. Apply the Gauss-Bonnet theorem on the
spherical triangle of Fig. 39↑ giving detailed explanations for
each step.
Exercise 104. Use a circular flat disc to demonstrate the
application of the local form of the Gauss-Bonnet theorem
giving detailed explanations for each step.
Exercise 105. What is the global form of the Gauss-Bonnet
theorem and what is its significance geometrically and
topologically?
Exercise 106. Find the total curvature of the surfaces depicted
in Fig. 17↑.
Exercise 107. Show, mathematically, that the area of a
geodesic polygon on a surface with constant non-vanishing
Gaussian curvature is determined by the sum of the internal
angles of the polygon.
Exercise 108. Verify that the total curvatures of ellipsoid and
torus are respectively 4π and 0 by performing detailed surface
integral calculations.
Exercise 109. Outline the usefulness of the global form of the
Gauss-Bonnet theorem in obtaining the total curvature of a
surface with known topological properties without performing
detailed calculations.
Exercise 110. Using the Gauss-Bonnet theorem, prove that the
Gaussian curvature is identically zero on a surface S if at any
point P on S there are two families of geodesic curves in the
neighborhood of P intersecting at a constant angle.
Exercise 111. Write down the mathematical relation that links
the Euler characteristic of a surface to its topological genus.
Exercise 112. Use the principal curvatures and the mean and
Gaussian curvatures to classify the points with regard to the
local shape of the surface as flat, elliptic, parabolic and
hyperbolic giving examples of common geometric shapes for
each case.
Exercise 113. Repeat the classification of the previous
question using this time the coefficients of the second
fundamental form of the surface.
Exercise 114. Prove that on a circular cylinder all points are
parabolic.
Exercise 115. Show that in the neighborhood of an elliptic
point on a surface, the surface lies on one side of its tangent
plane at that point.
Exercise 116. A surface is represented parametrically by:
r(u, v) = (u, v, u2 + v3).
Determine the conditions that identify the parabolic,
hyperbolic and elliptic points on the surface.
Exercise 117. Give an example of a surface having elliptic,
parabolic and hyperbolic points at different locations.
Exercise 118. Why the point type (i.e. being flat, elliptic,
hyperbolic or parabolic) on a surface is an invariant property
with respect to changes in the surface representation and
parameterization?
Exercise 119. Why the point type is invariant with respect to a
change of the surface orientation by reversing the direction of
the normal vector to the surface?
Exercise 120. Make a simple sketch outlining the position of a
surface relative to the tangent plane at elliptic, parabolic and
hyperbolic tangency points.
Exercise 121. Demonstrate that the surface represented
parametrically by:
r(u, v) = (u, v, u2 + v3)
lies on both sides of its tangent plane at the point (u, v) = (0, 0).
Exercise 122. For a surface represented parametrically by:
r = (u, v, v4),
find the equation of a curve on the surface whose points have a
common tangent plane.
Exercise 123. Describe how Dupin indicatrix can be used to
classify the points of a surface with regard to the local shape
(i.e. flat, elliptic, parabolic and hyperbolic).
Exercise 124. What are the prototypical geometric shapes that
provide the best approximation for the local shape of a
sufficiently smooth surface at its: flat, elliptic, hyperbolic and
parabolic points?
Exercise 125. What “umbilical point” means? What are the
other terms used to label such a point?
Exercise 126. What are the characteristic features of umbilical
points?
Exercise 127. Give five examples of umbilical points on
common geometric surfaces such as spheres and paraboloids.
Exercise 128. State the mathematical relation between the
coefficients of the metric and curvature tensors at umbilical
points.
Exercise 129. Demonstrate that at an umbilical point of a
surface we have: K = H2 where K and H are the Gaussian and
mean curvatures at the point.
Exercise 130. Show that the relation: K = H2 can also be
written as:
(aαβ bαβ)2 = (4 ⁄ a)(b11b22 − b212)
Exercise 131. Explain why at umbilical points we have b = c2a
where a and b are the determinants of the covariant metric and
covariant curvature tensors and c is a proportionality factor.
Exercise 132. Give two examples of surfaces whose all points
are umbilical, and two other examples of surfaces with no
umbilical point at all. Also, give an example of a surface with
only one umbilical point, and another example of a surface
with only two umbilical points.
Chapter 5
Special Curves
There are many classifications to space curves depending on
their properties and their relations with each other. In the
following sections of this chapter, we briefly investigate a few
of these categories.
5.1 Straight Line
A necessary and sufficient condition for a curve of class C2 to
be a straight line is that its curvature is zero at every point on
the curve. Hence, another criterion for a curve to be a straight
line is that all the tangents of the curve are parallel, where
“parallel” here is used in its absolute Euclidean sense (see §
2.7↑). Another criterion for a curve C(t):I → ℝ3 where t I ∈
⊆ ℝ to be a straight line is that for all points t in the domain of
the curve, r and r are linearly dependent where r(t) is the
spatial representation of the curve and the overdots represent
derivative with respect to the general parameter t of the curve.
A straight line lying on a surface has the same tangent plane at
each of its points, and hence the line is contained in this unique
tangent plane. Any straight line on any surface is a geodesic
curve (see § 5.7↓) and an asymptotic line (see § 5.9↓).
5.2 Plane Curve
A curve is described as a plane curve if the whole curve can be
contained in a plane with no distortion (Fig. 44↓). A necessary
and sufficient condition for a curve parameterized by a general
parameter t to be a plane curve is that the relation
r⋅(r × r) = 0
holds identically where r(t) is the spatial representation of the
curve and the overdots represent differentiation with respect to
t. Plane curves are characterized by having identically
vanishing torsion. In fact, having identically vanishing torsion
is a necessary and sufficient condition for a regular curve of
class C2 to be a plane curve. For plane curves, the osculating
plane at each regular point on the curve contains the entire
curve. Therefore, the plane curve may be characterized by
having a common intersection point for all of its osculating
planes. It also implies that the curve has the same osculating
plane at all of its points. Two curves are plane curves if they
have the same binormal lines at each pair of their
corresponding points. The locus of the centers of curvature of a
curve C is an evolute (see § 5.3↓) of C iff C is a plane curve.
On a smooth surface, a geodesic curve (see § 5.7↓) which is
also a line of curvature (see § 5.8↓) is a plane curve. A plane
curve has always a Bertrand curve associate (see § 5.4↓).
Figure 44 Plane curve.
5.3 Involute and Evolute
If Ce is a space curve with a tangent surface ST (see § 6.6↓) and
Ci is a curve embedded in ST and it is orthogonal to all the
tangent lines of Ce at their intersection points, then Ci is called
an involute of Ce while Ce is called an evolute of Ci (see Fig.
45↓). Hence, the involute is an orthogonal trajectory of the
generators of the tangent surface of its evolute. Accordingly,
the equation of an involute Ci to a curve Ce is given by:
(407) ri = re + (c − s)Te
where ri is an arbitrary point on the involute, re is the point on
the curve Ce corresponding to ri, c is a given constant, s is a
natural parameter of Ce and Te is the unit vector tangent to Ce
at re.
Figure 45 Evolute Ce, involute Ci, tangent lines (dashed) and
tangent surface (shaded).

A visual demonstration of how to generate an involute Ci of a

curve Ce, when (c − s) in Eq. 407↑ is positive, may be given by
detaching a taut string attached to Ce where the string is kept in
the tangent direction as it is detached. A fixed point P on the
string, where the distance between P and the point of contact
of the string with Ce represents a natural parameter of Ce, then
traces an involute of Ce.
A curve has infinitely many involutes corresponding to
different values of c in Eq. 407↑. Therefore, the involutes may
be described as parallel curves on the tangent surface.
Similarly, an involute has an infinite number of evolutes
corresponding to different values of c. For any tangent of a
given curve, the length of the line segment confined between
two given involutes is constant which is the difference between
the two c’s in Eq. 407↑ of the two involutes.
If Ce is an evolute of Ci, then for a given point Pe on Ce and the
corresponding point Pi on Ci the principal normal line of Ce at
Pe is parallel to the tangent line of Ci at Pi. A curve Ci is a
plane curve iff the locus of the centers of curvature of Ci is an
evolute of Ci. The involutes of a circle are congruent. The
evolutes of plane curves are helices.
5.4 Bertrand Curve
Bertrand curves are two associated space curves with common
principal normal lines at their corresponding points. Associated
Bertrand curves are characterized by the following properties:
1. The product of the torsions of their corresponding points is
constant, that is:
(408) τ1(so)τ2(so) = constant
where τ1 and τ2 are the torsions of the two curves and so is a
given value of their common parameter.
2. The distance between their corresponding points is constant.
3. The angle between their corresponding tangent lines is
constant.
For a plane curve C1, there is always a curve C2 such that C1
and C2 are associated Bertrand curves. If C1 is a curve with
non-vanishing torsion such that C1 has more than one Bertrand
curve associate, then C1 is a circular helix. The reverse is also
true. If C1 is a curve with non-vanishing torsion then a
necessary and sufficient condition for C1 to be a Bertrand
curve (i.e. it possesses an associate curve C2 such that C1 and
C2 are Bertrand curves) is that there are two constants c1 and c2
such that:
(409) κ = c1τ + c2
where κ and τ are the curvature and torsion of the curve C1. If
C1 and C2 are two involutes of a plane curve C, then C1 and C2
are Bertrand curves.
5.5 Spherical Indicatrix
A spherical indicatrix of a continuously-varying unit vector is a
continuous curve C̃ on the origin-based unit sphere generated
by mapping the unit vector (e.g. T or N or B) of a particular
space curve C on an equal unit vector represented by a point on
the origin-based unit sphere. Hence, we have C̃T, C̃N and C̃B as
the spherical indicatrices of C corresponding respectively to
the tangent, principal normal and binormal vectors of C (see
Footnote 26 in § 8↓). Figure 46↓ is a simple demonstration of
the spherical indicatrix C̃T of a space curve C.
Figure 46 The spherical tangent indicatrix C̃T of a space
curve C where the numbers indicate the correspondence
between the unit tangent vectors of C and their map on C̃T.

If C(s) is a naturally parameterized curve then s will not

necessarily be a natural parameter for the tangent indicatrix
C̃T. A necessary and sufficient condition for s to be a natural
parameter for C̃T is that κ(s) = 1 identically where κ is the
curvature of C. The tangent to the curve C̃T of a curve C is
parallel to the normal vector N of C at the corresponding points
of the two curves. The tangent to the curve C̃T of a curve C is
also parallel to the tangent to the curve C̃B of C at the
corresponding points of the two curves. The necessary and
sufficient condition for the curve C̃T of a curve C to be a circle
is that C is a helix.
The curvature of the curve C̃T of a curve C is related to the
curvature and torsion of C by:
(410) (κT)2 = (κ2 + τ2) ⁄ κ2
where κT is the curvature of C̃T while κ and τ are the curvature
and torsion of C respectively. The torsion of the curve C̃T of a
naturally parameterized curve C is given by:
(411) τT = (κ’τ − κτ’) ⁄ [κ(κ2 + τ2)]
where τT is the torsion of C̃T, κ and τ are the curvature and
torsion of C respectively, and the prime stands for derivative
with respect to the natural parameter s of C.
The curvature of the curve C̃B of a curve C is given by:
(412) κB = (κ2 + τ2) ⁄ κ2
where κB is the curvature of C̃B while the other symbols are as
explained before. The torsion of the curve C̃B of a naturally
parameterized curve C is given by:
(413) τB = (κ’τ − κτ’) ⁄ [τ(κ2 + τ2)]
where τB is the torsion of C̃B.
5.6 Spherical Curve
A spherical curve is a curve that lies completely on the surface
of a sphere. Spherical indicatrices are common examples of
spherical curves (see § 5.5↑). Circles are the only spherical
curves with constant curvature. At all points of a spherical
curve, the normal plane of the curve passes through the center
of the embedding sphere. Conversely, if all the normal planes
of a curve meet in a common point, then the curve is spherical
with the common point being the center of the sphere that
envelops the curve.
The sufficient and necessary condition that should be satisfied
by a spherical curve is given by:
(414)

where Rκ and Rτ are the radii of curvature and torsion and s is a

natural parameter of the curve. The center of curvature of a
twisted spherical curve C at a given point P on C is the
projection of the center of the enveloping sphere on the
osculating plane of C at P.
5.7 Geodesic Curve
The characteristic feature of a geodesic curve is that it has
vanishing geodesic curvature κg at every point on the curve.
This is a necessary and sufficient condition for a surface curve
to be geodesic. In more technical terms, let S:Ω → ℝ3 be a
⊆
surface defined on a set Ω ℝ2 and let C(t):I → ℝ3, where I
⊆ ℝ, be a regular curve on S, then C is a geodesic curve iff
∈
κg(t) = 0 on all points t I in its domain. The path of the
shortest distance connecting two points in a Riemannian space
is a geodesic. The length of arc, as given by Eq. 204↑, is used
in the definition of geodesic in this sense.
A physical interpretation may be given to the geodesic curve
that a free particle restricted to move on the surface will follow
a geodesic path. Another physical interpretation is that a
geodesic path minimizes the total kinetic energy spent by a
massive object in moving between two points when the path is
traversed with constant speed. These two physical
interpretations may rest on the same physical principle.
The geodesic is a straight line in a Euclidean space, but it is a
generalized curved path in a general Riemannian space. If a
geodesic surface curve is not a straight line then its principal
normal vector N is collinear with the normal vector n to the
surface at each point on the curve with non-vanishing
curvature; the opposite is also true. In fact, a curve on a surface
is geodesic iff it is either a straight line or its principal normal
vector is collinear with n over the whole curve. As stated
before, collinearity of N and n is equivalent to the condition
that n lies in the osculating plane of the curve at the given
point.
Another sufficient and necessary condition for a curve to be a
geodesic curve is that the first variation (see § 1.4.2↑) of its
length is zero. In fact, this may be taken as the basis for the
definition of geodesic as the curve connecting two fixed points,
P1 and P2, whose length possesses a stationary value with
regard to small variations in its neighborhood, that is:
(415)

It can be shown that a geodesic curve satisfies the Euler-

Lagrange variational principle (see § 1.4.2↑) which is a
necessary and sufficient condition for extremizing the arc
length. Figure 47↓ is an illustration of how the length of the
geodesic curve between two given points is subject to the
variational principle.
Figure 47 The length of a geodesic curve (solid) connecting
two points, P1 and P2, as an extremum with respect to the
length of other curves (dashed) connecting these points that
result from small perturbations in its neighborhood.

Examples of geodesic curves on simple surfaces are the arcs of

great circles on spheres. In fact, being an arc of a great circle is
a sufficient and necessary condition for being a geodesic curve
on a sphere. Other examples of geodesic curves are the arcs of
helices, the generating straight lines and the circles on
cylinders (Fig. 48↓). The generating straight lines and the
circles on cylinders may be considered as degenerate helices.
The meridians of a surface of revolution are also geodesics.
The arcs of parallel circles on a surface of revolution
corresponding to stationary points on the generating curve of
the surface are also geodesic curves. All straight lines on any
surface are geodesic curves. For plane surfaces in particular,
being a straight line on a plane is a sufficient and necessary
condition for being a geodesic. The lines of curvature (see §
5.8↓) are also geodesic curves.
Figure 48 The three types of geodesic curves on cylinders (a)
circular arcs (b) generating lines and (c) helical arcs.

Intrinsically, the geodesic curves are straight lines in the sense

that a 2D inhabitant will see them straight since he cannot
detect their curvature. This is due to the fact that only the
geodesic part of the curvature is an intrinsic property and
hence it can be detected by a 2D inhabitant, therefore if this
part of the curvature vanished the 2D inhabitant will fail to
detect any curvature to the curve which is equivalent for him to
having a straight line. Any deviation from such “straight lines”
within the surface is therefore a geodesic curvature and hence
it can be detected intrinsically by a 2D inhabitant.
Although a geodesic curve is frequently the curve of the
shortest distance between two points on the surface it is not
necessarily so. For instance, the largest of the two arcs forming
a great circle on a sphere is a geodesic curve but it is not the
curve of the shortest distance on the sphere between its two
end points; in fact it is the curve of the longest distance among
the circular arcs connecting the two points (Fig. 49↓). A similar
example is the two arcs of a parallel circle on a circular
cylinder connecting two points where the two arcs are different
in length. Anyway, if on a surface S there is exactly one
geodesic curve connecting two given points, P1 and P2, then
the length of the geodesic curve segment between P1 and P2 is
the shortest distance on S between these points.
Figure 49 Two geodesic curves connecting two points, P1
and P2, on the surface of a sphere: a short one between P1 and
P2 directly, and a long one between P1 and P2 through P3. Both
of these geodesics are arcs of a great circle on the sphere.

Based on the previous statements, being a shortest path is a

sufficient but not necessary condition for being a geodesic, that
is all shortest paths connecting two given points are geodesics
but not all geodesics are shortest paths. A constraint may be
imposed to make the criterion of minimal length apply to all
geodesics by stating that geodesics minimize distance locally
but not necessarily globally where an infinitesimal element of
arc is considered in this constraint. Anyway, the universal
criterion that should be adopted to identify geodesic curves is
the vanishing of the geodesic curvature over the whole curve,
as stated at the start of this section.
The geodesic, even in its restricted sense as the curve of the
shortest distance, is not necessarily unique; for example all
semi-circular meridians of longitude connecting the two poles
(or in fact all semi-circular arcs connecting any two antipodal
points) of a sphere are geodesics even in that sense and there is
an infinite number of them. In fact, even the existence, not
only uniqueness, of a geodesic connecting two points on a
surface is not guaranteed. An example is the xy plane
excluding the origin of coordinates with two points on a
straight line lying in the plane and passing through the origin
where there is a lower limit for the length of any curve
connecting the two points (Fig. 50↓). This limit is the straight
line segment connecting the two points but this segment cannot
be a geodesic on the plane since it includes the origin which is
not on the plane. Any curve C (other than the straight line
segment) on the plane connecting the two points cannot be a
curve of shortest length, and hence a geodesic, since there is
always another curve on the plane connecting the two points
which is shorter than C. In this context, we note that on a plane
surface all geodesic curves are straight lines and hence of
shortest length.
Figure 50 Non-existence of a geodesic curve connecting the
two points P1 and P2 on the shown xy plane which does not
include the origin of coordinates O. The dashed line represents
the straight line segment connecting P1 and P2 while the solid
line C represents other curves on the plane between the two
points.

In the neighborhood of a given point P on a surface and for any

specific direction, there is exactly one geodesic curve passing
through P in that direction. More technically, for any specific
point P on a surface S of class C3, and for any tangent vector v
in the tangent space of S at P, there exists a geodesic curve on
the surface in the direction of v that passes through P. In fact,
this is based on the existence of a unique solution to the
geodesic differential equations (Eqs. 416↓-417↓) when initial
values of a point on the curve and its derivative (which
represents its tangent direction) at that point are given. An
obvious example of the previous statement is the plane where a
straight line passes through any point and in any direction.
Another example is the sphere where a great circle passes
through any point and in any direction. A less obvious example
is the cylinder where a helix (including the straight line
generators and the circles which can be regarded as degenerate
forms of helix) passes through any point and in any direction.
Similarly, there is exactly one geodesic curve passing through
two sufficiently close points on a smooth surface (see
Footnote 27 in § 8↓).
As indicated before, geodesics in curved spaces are the
equivalent of straight lines in flat spaces. For planes (or in fact
for any Euclidean nD manifold) there exists a unique geodesic
passing between any two points (whether the two points are
close or not) which is the straight line segment connecting the
two points.
The necessary and sufficient condition that should be satisfied
by a naturally parameterized curve on a surface, both of class
C2, to be a geodesic curve is given by the following set of
second order non-linear differential equations:
(416) (d2u1 ⁄ ds2) + Γ111(du1 ⁄ ds)2 + 2Γ112(du1 ⁄ ds)(du2 ⁄ ds) +
Γ122(du2 ⁄ ds)2 = 0
(417) (d2u2 ⁄ ds2) + Γ211(du1 ⁄ ds)2 + 2Γ212(du1 ⁄ ds)(du2 ⁄ ds) +
Γ222(du2 ⁄ ds)2 = 0
where s is the arc length, and the Christoffel symbols are
derived from the surface metric. The last equations can be
merged in a single equation using tensor notation, that is:
(418) δ(duα ⁄ ds) ⁄ δs ≡ (d2uα ⁄ ds2) + Γαβγ(duβ ⁄ ds)(duγ ⁄ ds) = 0
where α, β, γ = 1, 2 and the standard notation of absolute
derivative is in use (see § 7↓). These equations, which can be
obtained from Eq. 323↑ by setting the two components of the
geodesic curvature vector to zero, have no closed form explicit
solutions in general because of their non-linearity. Similar
equations are used to identify the geodesic curves in general
nD spaces.
From Eq. 418↑, it can be seen that being a geodesic is an
intrinsic property since the conditions represented by this
equation depend exclusively on the Christoffel symbols which
depend only on the coefficients of the first fundamental form
and their partial derivatives. Hence, geodesic curves can be
detected and measured by a 2D inhabitant. From Eq. 418↑, it
can also be seen that for planes (or indeed for any Euclidean
nD manifold) the geodesic is a straight line since in this case
the Christoffel symbols vanish identically and Eq. 418↑ will be
reduced to (d2uα ⁄ ds2) = 0 which has a straight line solution.
From Eq. 326↑, we see that the u1 coordinate curves on a
sufficiently smooth surface are geodesics iff Γ211 = 0.
Similarly, from Eq. 328↑, we see that the u2 coordinate curves
are geodesics iff Γ122 = 0. We also see from Eqs. 327↑ and
329↑ that for coordinate systems with orthogonal coordinate
curves, the coordinate curves are geodesics iff E is independent
of v and G is independent of u.
For a Monge patch of the form r(u, v) = (u, v, f(u, v)), the
geodesic differential equations are given by:
(419) (1 + fu2 + fv2)u’’ + fufuu(u’)2 + 2fufuvu’v’ + fufvv(v’)2 = 0
(420) (1 + fu2 + fv2)v’’ + fvfuu(u’)2 + 2fvfuvu’v’ + fvfvv(v’)2 = 0
where the subscripts u and v represent partial derivatives of f
with respect to the surface coordinates u and v, and the prime
represents derivatives with respect to a natural parameter.
Based on what we have seen so far, it can be concluded that
each one of the following provisions is a necessary and
sufficient condition for a curve C on a surface S to be a
geodesic curve:
1. The geodesic component of the curvature vector is zero at
each point on the curve, that is Kg = 0 identically. This is based
on the definition of geodesic curve which we stated earlier.
2. The osculating plane of the curve at each point of the curve
is orthogonal to the tangent plane of S at that point. The reason
is that on geodesic curves κg = 0 and hence n and N are in the
same orientation (parallel or anti-parallel) on all points along
the curve (refer to Eqs. 305↑ and 307↑) and hence n lies in the
osculating plane and the osculating plane will be orthogonal to
the tangent plane.
3. The normal vector n to the surface at any point on the curve
lies in the osculating plane. This is because for a geodesic
curve, Kg vanishes identically and hence K = κN = κnn = Kn.
4. The principal normal vector N of C is normal to the surface
at each point on C since N is collinear with n.
5. The curvature vector K of the curve is normal to the tangent
plane of the surface at each point on the curve.
Being a geodesic is independent of the choice of the coordinate
system and hence it is invariant under permissible
transformations. It is also independent of the type of
representation and parameterization and hence it is invariant in
this sense.
Geodesic curves can be open or closed curves and may be self-
intersecting. In this statement, we are considering the totality
of the geodesic path as characterized by having identically
vanishing geodesic curvature and not as a connecting arc
between two distinct points (see Footnote 28 in § 8↓).
Examples of open geodesics are the straight lines on planes
and the helices on cylinders while examples of closed
geodesics are the geodesics of spheres which are great circles
and the parallel circles on cylinders. In fact, all the geodesics
on sphere are closed curves as they are great circles, while
circles on cylinder is the only case of closed geodesics on this
type of surface.
As a result of the Gauss-Bonnet theorem (see § 4.8↑), on a
surface with negative Gaussian curvature two geodesics cannot
intersect at more than one point if the geodesics enclose a
simply-connected region. The reason is that on introducing an
artificial vertex at a regular point on one of these curves we
will have a new corner with π interior angle and hence the sum
of the angles of the geodesic triangle will exceed π which is
impossible on a surface with K < 0 . Also, on introducing an
artificial vertex at a regular point on each one of these curves
we will have a geodesic quadrilateral whose internal angles
add up to more than 2π on a surface with K < 0 which is not
possible (see § 4.8↑).
Another result of the Gauss-Bonnet theorem is that a surface
with negative Gaussian curvature cannot have a geodesic that
intersects itself. This may be established by a similar argument
to the previous one that is: on introducing two artificial corners
at two regular points on the curve, we will have a geodesic
triangle whose interior angles add up to more than π which is
not possible on a surface with K < 0 (see § 4.8↑).
On a patch of a surface of class C2 with orthogonal coordinate
curves and with the first fundamental form coefficients being
dependent on only the u coordinate variable (i.e. E = E(u), F =
0 and G = G(u)) the following statements apply:
1. The u coordinate curves are geodesics.
2. The v coordinate curves are geodesics iff ∂uG = 0 along
these curves.
3. A curve C represented by r = r(u, v(u)) is a geodesic iff:
(421)

where k is a constant.
The case of dependence on only the v coordinate variable can
be obtained by re-labeling the coordinate variables and
coefficients. The second of the above statements may be
generalized by saying: on a surface with orthogonal coordinate
curves, the curves of constant uα are geodesics iff aββ (β ≠ α) is
a function of uβ only.
As indicated before, geodesics in curved spaces represent a
generalization of straight lines in flat spaces. Hence, geodesics
may be described as the straightest curves in the space. In fact,
geodesic curves on a developable surface become straight lines
when the surface is developed into a plane by unrolling. This
may be demonstrated by the perception of a 2D inhabitant of
the surface who will fail to observe any difference to the
geodesic curve when the surface is developed into a plane and
the geodesic curve necessarily becomes a straight line on the
plane. More generally, a geodesic curve will be mapped onto a
geodesic curve by any isometric transformation due to the
invariance of the geodesic curvature under this type of
transformations since geodesic curvature is an intrinsic
property. Therefore, isometric surfaces possess identical
geodesic equations.
Another sufficient and necessary condition for a surface curve
to be geodesic is being a tangent to a parallel vector field. A
vector attained by parallel propagation (see § 2.7↑) of a tangent
vector to a geodesic curve stays always tangent to the geodesic
curve. As a result, a vector field attained by parallel
propagation along a geodesic makes a constant angle with the
geodesic.
5.8 Line of Curvature
A “line of curvature” is a curve C on a surface S defined on an
⊆
interval I ℝ as C:I → S with the condition that the tangent
of C at each point on C is collinear with one of the principal
directions (see § 4.4↑) of the surface at that point. We note that
“line” here does not mean straight. Since the definition of the
line of curvature is seemingly based on the existence of distinct
principal directions, umbilical points (see § 4.10↑) may be
excluded from the above definition of the line of curvature due
to the absence of distinct principal directions at these points
although there seems to be no harm in including isolated
umbilical points (at least) over the path of the line of curvature
(see Footnote 29 in § 8↓). Referring to Eq. 134↑, on a line of
curvature either sinθ = 0 or cosθ = 0 and hence the lines of
curvature are characterized by having identically vanishing
geodesic torsion (i.e. τg = 0).
The condition that should be satisfied by a line of curvature is
usually given by the following relation:
(422) (a12b11 − a11b12)du1du1 + (a22b11 − a11b22)du1du2 +
(a22b12 − a12b22)du2du2 = 0
where the indexed a and b are the coefficients of the surface
covariant metric and covariant curvature tensors respectively.
In fact, this is the same as the condition given by Eq. 350↑ for
the principal directions, which is consistent with the fact that
the line of curvature is aligned along a principal direction at
each of its points. The condition that should be satisfied by a
line of curvature on a surface may be given in tensor notation
by:
(423) εγδ aαγ bβδ duα duβ = 0
where εγδ is the 2D absolute permutation tensor.
Examples of lines of curvature are meridians and parallels of
surface of revolution of class C2. For a developable surface,
the lines of curvature consist of its generators and their
orthogonal trajectories. On a sufficiently smooth surface, any
geodesic which is a plane curve is a line of curvature.
Similarly, on a sufficiently smooth surface, if a geodesic curve
C is a line of curvature then C is a plane curve. The lines of
intersection of each pair of a triply orthogonal system are also
lines of curvature. We remark that three families of surfaces in
a subset V of a 3D space form a triply orthogonal system if at
each point P of V there is a single surface of each family
passing through P such that each pair of these surfaces
intersect orthogonally at their curve of intersection.
At a non-umbilical point P on a sufficiently smooth surface S,
the u1 and u2 coordinate curves are aligned with the principal
directions iff f = F = 0 at P. The “if” part can be seen, for
example, from Eq. 350↑ which in this case (i.e. f = F = 0) will
reduce to (gE − eG)du1du2 = 0 and hence it will be satisfied on
the coordinate curves, since du2 will vanish on the u1
coordinate curve while du1 will vanish on the u2 coordinate
curve, and these curves become aligned with the principal
directions. The “only if” part can also be seen from Eq. 350↑
because if the coordinate curves are aligned with the principal
directions then this equation should be satisfied where it
becomes (fE − eF)du1du1 = 0 for the u1 coordinate curve and
(gF − fG)du2du2 = 0 for the u2 coordinate curve and both of
these equations imply f = F = 0 since du1 ≠ 0 on the u1
coordinate curve and du2 ≠ 0 on the u2 coordinate curve. We
note that by considering the stated conditions and the
definitions of the coefficients of the first and second
fundamental forms as well as the Rodrigues curvature formula
(see § 4.4↑), it can be concluded that the coefficients E, e and
g, G cannot vanish.
As a consequence of the last paragraph, the coordinate curves
on the surface S, excluding the umbilical points, are lines of
curvature iff f = F = 0 over the entire surface. This may also be
stated by saying that on a smooth surface, excluding planes and
spheres (whose all points are umbilical), if the lines of
curvature are selected as the net of coordinate curves then a12
= b12 = 0 over the entire surface excluding the umbilical
points. We remark that when the u and v coordinate curves of a
surface patch are lines of curvature, the principal curvatures, κ1
and κ2, over the entire patch will be given by:
(424)
κ1 = e ⁄ E
κ2 = g ⁄ G
where E, G, e, g are the coefficients of the first and second
fundamental forms at the points of the patch. The reader is
referred to § 4.4↑ for justification (see Eq. 354↑ and related
text).
On a surface of class C3, there are two perpendicular families
of lines of curvature in the neighborhood of any non-umbilical
point. If the curve of intersection of two surfaces is a line of
curvature for one surface then it is a line of curvature for the
other surface when the two surfaces are intersecting each other
at a constant angle.
The lines of curvature form a real orthogonal grid over the
surface. A curve is a line of curvature iff the tangent to the
curve and the tangent to its spherical image (see § 3.10↑) at
their corresponding points are parallel. The lines of curvature
on a surface, which is not a sphere or minimal surface (see §
6.7↓), are represented by an orthogonal net on its spherical
image.
As indicated above, in the neighborhood of a non-umbilical
point on a sufficiently smooth surface there are two orthogonal
families of lines of curvature. Hence, at each point P on such a
surface a coordinate patch including P can be introduced in the
neighborhood of P where the coordinate curves at P are
aligned with the principal directions. On a surface patch where
the Gaussian curvature does not vanish, the angles between the
asymptotic lines (see § 5.9↓) are bisected by the lines of
curvature.
5.9 Asymptotic Line
An asymptotic direction of a surface at a given point P is a
direction for which the normal curvature vanishes, i.e. κn = 0.
Hence, in an asymptotic direction at a point on a surface we
have (see Eq. 307↑):
(425) K = Kg = κgu
As a consequence of Eq. 316↑, κn is zero in the directions for
which the second fundamental form is zero. Hence, the
necessary and sufficient condition for the asymptotic directions
is that:
(426) bαβduαduβ = b11(du1)2 + 2b12du1du2 + b22(du2)2 = 0
We note that asymptotic directions are defined only at points
for which the Gaussian curvature is non-positive (K ≤ 0) and
hence it is not defined at elliptic points (see § 4.9↑) where K >
0. This is because at elliptic points either κ1 > 0 and κ2 > 0 or
κ1 < 0 and κ2 < 0 and hence κn cannot take the value zero at
these points.
As a result, the number of asymptotic directions at elliptic,
parabolic and hyperbolic points is 0, 1 and 2 respectively,
while at flat points all directions are asymptotic. The two
asymptotic directions of a hyperbolic point separate the
directions of positive normal curvature from the directions of
negative normal curvature. The sign of the normal curvature at
elliptic and parabolic points is the same in all directions,
excluding the asymptotic direction of the parabolic point.
Similarly, at flat points the normal curvature is zero in all
directions.
⊆
A t-parameterized surface curve C(t):I → S, where I ℝ is an
open interval and S represents the surface, is described as an
 ∈
asymptotic line if at each point t I the vector T, which is the
tangent to the curve, is collinear with an asymptotic direction
at that point. It should be remarked that “line” here is not
required to be straight; hence asymptotic lines are also called
asymptotic curves.
From the above statements, it can be seen that the asymptotic
lines are characterized by the following features:
1. The normal component of the curvature vector is zero at
each point on the curve, that is Kn = 0 identically. This is based
on the definition of asymptotic line as stated above.
2. The tangent plane to the surface at each point of the curve
coincides with the osculating plane of the curve at that point.
This is a consequence of having identically vanishing normal
curvature, since the curvature vector will then have only a
tangential component and hence the osculating plane at each
point of an asymptotic line becomes tangent to the surface at
that point.
The differential equation representing asymptotic lines can be
obtained from the condition that the normal curvature vanishes
identically over the line, that is:
(427) e(du1 ⁄ ds)2 + 2f(du1 ⁄ ds)(du2 ⁄ ds) + g(du2 ⁄ ds)2 = 0
which is based on Eqs. 316↑ and 233↑ or on Eq. 318↑. The
necessary and sufficient condition for the u1 and u2 coordinate
curves to become asymptotic lines is that e = 0 identically on
the u1 coordinate curve and g = 0 identically on the u2
coordinate curve (see Footnote 30 in § 8↓). This is based on
Eqs. 319↑ and 320↑ which are fully justified there.
According to Eq. 308↑, κn = n⋅K where n and K are
respectively the normal vector to the surface and the curve
curvature vector. Hence, a curve on a sufficiently smooth
surface is an asymptotic line iff n⋅K = 0 identically. Now, since
the vector n cannot vanish on the regular points of the surface,
then this condition is realized if at each point on the curve
either K = 0 or K and n are orthogonal vectors. In the former
case the point is an inflection point while in the latter case the
osculating plane is tangent to the surface at the point.
Therefore, all points on an asymptotic line should be one of
these types or the other. The reverse is also true, i.e. a curve
whose all points are one of these types or the other is an
asymptotic line. As a result of the last statements, any straight
line on a surface is an asymptotic line since the curve curvature
vector K vanishes identically on such a line.
According to the theorem of Beltrami-Enneper, along an
asymptotic non-straight line on a sufficiently smooth surface
the square of the torsion τ is equal to the negative of the
Gaussian curvature K, that is:
(428) τ2 = − K
where τ and K are evaluated at each individual point along the
curve. Since asymptotic directions are defined only at points
for which K ≤ 0, the square of the torsion in the above equation
is equal to the absolute value of the Gaussian curvature at the
point, that is: τ2 = |K| and hence τ is real as it should be. The
torsions of two asymptotic lines passing through a given point
on a sufficiently smooth surface are equal in magnitude and
opposite in sign.
As we will see (refer to § 5.10↓), asymptotic directions are
self-conjugate. In fact, some authors take self-conjugation as
the defining characteristic for being asymptotic. From the
definition of the asymptotic direction plus the Euler equation
(Eq. 341↑), we see that the angle θ which an asymptotic
direction makes with the principal direction of κ1 at a given
non-umbilical point P on a sufficiently smooth surface S is
given by:
(429) tan2θ = − κ1 ⁄ κ2
where κ1 and κ2 are the principal curvatures of S at P. When
κ2 = 0, the reciprocal of this relation should be taken. Again,
tanθ is real since at points with K < 0, the two kappas should
have opposite signs (see Eq. 355↑). The situation is similar
when κ1 = 0 and κ2 < 0. However, when κ1 > 0 and κ2 = 0 the
reciprocal will be taken and the cotangent will be real. The
possibility of κ1 = κ2 = 0 is already excluded by the non-
umbilical condition since a point with κ1 = κ2 = 0 is a flat
umbilic.
Because Eq. 426↑ is quadratic, it possesses two solutions
which are real and distinct, or real and coincident, or conjugate
imaginary depending on its discriminant Δ which is opposite in
sign to the determinant b of the surface covariant curvature
tensor (see Footnote 31 in § 8↓). Hence, the asymptotic
directions at a given point on a surface can be classified
according to the determinant b at the point as:
1. Real and distinct for Δ > 0 and hence b < 0.
2. Real and coincident for Δ = 0 and hence b = 0.
3. Conjugate imaginary for Δ < 0 and hence b > 0.
This is inline with the above statement that the number of
asymptotic directions at elliptic, parabolic and hyperbolic
points is 0, 1 and 2 respectively because, as seen in § 4.9↑, the
local shape of a surface at a given point is determined by the
sign of b at the point where b < 0 at hyperbolic point, b = 0 at
parabolic and flat points, and b > 0 at elliptic point. We also
remark that from Eq. 356↑ we can see that the sign of the
Gaussian curvature K is the same as the sign of b due to the
fact that a > 0 since the first fundamental form is positive
definite, as established earlier. Hence, the above-described
classification of the asymptotic directions can also be based on
K, as stated for b in the previous points. Again, as seen in §
4.9↑, the sign of K is used to determine the local shape at a
point, and hence the number of the asymptotic directions is
determined accordingly.
It is worth noting that on a sufficiently smooth surface with
orthogonal families of asymptotic lines the mean curvature H
is zero. This can be seen from Eq. 383↑ where by aligning the
coordinate curves along the asymptotic directions with a
proper labeling, we will get: F = 0 since the coordinate curves
are orthogonal, and e = g = 0 according to the above condition
which we stated after Eq. 427↑, and hence H = 0. Now since H
is invariant, then this will remain valid under permissible
transformations to other surface coordinates. Another note is
that the principal directions at a given point on a smooth
surface bisect the asymptotic directions at the point, as
indicated before. Also, on a smooth surface of class C3, there
are two distinct families of asymptotic directions in the
neighborhood of any hyperbolic point.
As seen before, any straight line contained in a surface is an
asymptotic line. As well as the previously stated explanation,
this may also be justified by the fact that such a line is wholly
contained in a plane, which is the tangent space of each of its
points, and hence the line is an asymptotic line with an
identically vanishing normal curvature.
5.10 Conjugate Direction
A direction (δv ⁄ δu) at a point on a sufficiently smooth surface
is described as conjugate to the direction (dv ⁄ du) iff the
following relation holds true (see Footnote 32 in § 8↓):
(430) dr⋅δn = 0
where:
(431)
dr = E1du + E2dv
δn = ∂unδu + ∂vnδv
From Eq. 244↑, it can be seen that the condition given by Eq.
430↑ is equivalent to the following condition:
(432) e duδu + f (duδv + dvδu) + g dvδv = 0
The last equation may also be obtained directly by substituting
from Eq. 431↑ into Eq. 430↑ and performing the dot product
with the use of Eqs. 249↑-251↑ to obtain the coefficients e, f, g.
Due to the symmetry in the above relations, (dv ⁄ du) is also
conjugate to (δv ⁄ δu), and hence the two directions are
described as conjugate directions. At a hyperbolic or an elliptic
point on a sufficiently smooth surface, each direction has a
unique conjugate direction. As stated earlier, an asymptotic
direction is a self-conjugate direction.
Two families of curves on a sufficiently smooth surface are
described as conjugate families if the directions of their
tangents at each intersection point of the curves are conjugate
directions. The u and v coordinate curves on a smooth surface
are conjugate families of curves iff f, which is the coefficient of
the second fundamental form, vanishes identically. This can be
seen from Eq. 432↑ where by a proper labeling of the u and v
coordinates in the two directions to make duδu = dvδv = 0 on
the coordinate curves the first and last terms of Eq. 432↑ will
vanish on the coordinate curves and hence the curves will be
conjugate families by satisfying the reduced condition f(duδv
+ dvδu) = 0 which leads to the condition f = 0 since duδv +
dvδu ≠ 0 according to this labeling. Similarly, if f = 0 then the
coordinate curves will satisfy the reduced condition f(duδv +
dvδu) = 0 and hence they are conjugate families.
5.11 Exercises
Exercise 1. State two criteria for a space curve to be straight.
Exercise 2. Prove that a curve represented by r(t) is a straight
line if r and r are linearly dependent over the whole curve.
Exercise 3. Show that a space curve whose all tangent lines are
parallel is a straight line.
Exercise 4. Correct, if necessary, the following statement: “All
straight lines on a surface are geodesic curves and vice versa”.
Exercise 5. What is the characteristic feature of plane curves?
From this, explain why the torsion of plane curves is
identically zero.
Exercise 6. Prove that a curve is a plane curve if its osculating
planes have a common intersection point.
Exercise 7. Show that a space curve represented by r(t) is a
plane curve iff
r⋅(r × r)
vanishes identically.
Exercise 8. Prove that having an identically vanishing torsion
is a necessary and sufficient condition for a curve to be a plane
curve.
Exercise 9. Show that two curves are plane curves if they have
the same binormal lines at each pair of their corresponding
points.
Exercise 10. Define, rigorously, involute and evolute curves
making a simple plot to outline their relation. Also explain the
role of the tangent surface of the evolute in this context.
Exercise 11. Explain all the symbols used in the following
equation which is related to involute curves:
ri = re + (c − s)Te.
Make sense of this equation using your plot in the previous
exercise.
Exercise 12. Outline the visual demonstration which is
commonly used to explain the relation between an involute and
its evolute. Use the plot mentioned in the last two questions in
your explanation.
Exercise 13. Show that the tangent line of a curve and the
principal normal line of its evolute are parallel at their
corresponding points.
Exercise 14. Prove that the evolutes of plane curves are
helices.
Exercise 15. Prove that for a plane curve C the locus of the
centers of curvature of C is an evolute of C.
Exercise 16. Derive the parametric equation of the involute of
a circle represented by:
r(θ) = (5cosθ, 5sinθ)
where 0 ≤ θ < 2π.
Exercise 17. Prove that any two involutes of a plane curve are
associated Bertrand curves.
Exercise 18. How many involutes a given curve can have?
How these involutes are related to each other through the
constant c (see Eq. 407↑)?
Exercise 19. How many evolutes a given curve can have? How
these evolutes are related to each other through the constant c
(see Eq. 407↑)?
Exercise 20. Justify the fact that the involutes of a circle are
congruent with a clear explanation of how these involutes are
related to each other.
Exercise 21. Define Bertrand curves outlining two of their
main characteristic features.
Exercise 22. Show that a helix has an infinite number of
Bertrand associates and identify these associates.
Exercise 23. Prove that on a pair of Bertrand curves, the angle
between their tangents at corresponding points is constant.
Exercise 24. State a sufficient and necessary condition for a
curve to be a Bertrand curve by having an associate Bertrand
curve.
Exercise 25. Show that a plane curve has always a Bertrand
associate.
Exercise 26. Prove that the product of torsions at the
corresponding points of a pair of associated Bertrand curves is
constant (see Eq. 408↑ and related text for explanation).
Exercise 27. Prove that on a pair of Bertrand curves, the
distance between their corresponding points is constant.
Exercise 28. Give a brief definition of spherical indicatrix with
a simple sketch of the spherical normal indicatrix C̃N of a
space curve to illustrate this concept.
Exercise 29. Prove the following equation (Eq. 410↑):
(κT)2 = (κ2 + τ2) ⁄ κ2.
Exercise 30. Discuss the similarities and differences between
Gauss mapping (see § 3.10↑) and spherical indicatrix mapping.
Exercise 31. Justify, using a simple fact about helices, that the
spherical images of T, N, B of a helix rotating around the z-
axis are circles centered around the z-axis.
Exercise 32. Justify the following statement: “The binormal
indicatrix is a single point for a plane curve, and the tangent
indicatrix is a single point for a straight line”.
Exercise 33. Prove, rigorously, that the tangent indicatrix of a
helix is a circle.
Exercise 34. Prove that the tangent to the spherical indicatrix
of the tangent to a given space curve C and the principal
normal of C are parallel.
Exercise 35. Write down the mathematical formula for the
torsion of the spherical binormal indicatrix of a space curve
explaining all the symbols used in the formula.
Exercise 36. What “spherical curve” means? give a common
example of a spherical curve.
Exercise 37. State, mathematically, the sufficient and
necessary condition for a curve to be a spherical curve
explaining all the symbols involved.
Exercise 38. Investigate if the curve represented parametrically
by:
r(t) = (5cost, 5cost sint, 5sin2t)
is a spherical curve or not.
Exercise 39. Show that the spherical image of a curve C(t) is a
closed curve when the vector that generates the spherical
image is a periodic function of t although C may not be
periodic. Discuss in this context the helix as an example.
Exercise 40. What is the characteristic feature of geodesic
curves?
Exercise 41. Give a rigorous mathematical definition of
geodesic curve.
Exercise 42. Give examples of geodesic curves on the
following surfaces: plane, sphere and cylinder.
Exercise 43. On a surface of revolution, what type of curve is
necessarily geodesic and what type is potentially geodesic?
Exercise 44. Define geodesic curve variationally using the
concepts of calculus of variations.
Exercise 45. Show that any helix on a circular cylinder is a
geodesic curve.
Exercise 46. Outline the relation between the concept of
geodesic curve and the concept of curve of shortest distance
between two points.
Exercise 47. Prove that Eq. 418↑ is a sufficient and necessary
condition for a curve to be geodesic.
Exercise 48. Find an analytical expression representing the
geodesic curves on a circular cone using one of its parametric
representations.
Exercise 49. Outline the concept of geodesic curve on a
surface as perceived by a 2D inhabitant of the surface.
Exercise 50. Prove that all the geodesic curves on a plane are
straight lines.
Exercise 51. Does a geodesic curve necessarily exist between
two given points on a space surface, and if it does exist is the
geodesic curve necessarily unique? Support your answer with
illustrating examples for both cases.
Exercise 52. Discuss the following statement and its
implications: “Being a shortest path is a sufficient but not
necessary condition for being a geodesic curve”.
Exercise 53. Correct, if necessary, the following statement: “In
the neighborhood of a given point P on a surface, there is
exactly one geodesic curve that passes through P”.
Exercise 54. Write down, with full explanation, the differential
equations which provide the necessary and sufficient
conditions for a naturally parameterized curve on a surface to
be geodesic.
Exercise 55. Correct the following relations which represent
the geodesic differential equations for a Monge patch of the
form r(u, v) = (u, v, f(u, v)):
(1 + fu2 + fv2)u’ + fufuu(u’)2 + 2fufuvu’v’ + fufvv(v’)2 = 0
(1 + fu2 + fv2)v’’ + fvfu(u’)2 + 2fvfuvu’v’ + fvfvv(v’)2 = 0
Exercise 56. Why the normal vector to the surface at any point
on a geodesic curve should be contained in the osculating
plane of the curve at that point? Give a clear technical
justification.
Exercise 57. Using Gauss-Bonnet theorem, explain why a
surface with negative Gaussian curvature cannot have a
geodesic that intersects itself.
Exercise 58. Using Eq. 418↑, prove that all meridians of a
surface of revolution are geodesic curves.
Exercise 59. Discuss the following statement in the context of
the perception of geodesic curves by a 2D inhabitant:
“Geodesic curves on a developable surface become straight
lines when the surface is developed into a plane”.
Exercise 60. Give an example of a surface curve whose normal
curvature and geodesic curvature are identically zero over the
whole curve.
Exercise 61. What is the relation between a line of curvature
on a surface and the principal directions at the points of the
curve?
Exercise 62. Prove that if a plane and a surface are intersecting
at a constant angle then their curve of intersection is a line of
curvature.
Exercise 63. Repeat the previous exercise replacing plane with
sphere.
Exercise 64. Can a line of curvature include umbilical points?
Discuss this issue considering the question of allowing more
than two principal directions at a point or not.
Exercise 65. Prove that for any sufficiently smooth surface of
revolution, the parallels and meridians are lines of curvature.
Exercise 66. Show that for a given non-umbilical point P on a
sufficiently smooth surface there is a patch that contains P
where the directions of the coordinate curves at P are principal
directions.
Exercise 67. Prove Hilbert lemma using the proposal that the
coordinate curves on a patch can coincide with the lines of
curvature in the neighborhood of a non-umbilical point.
Exercise 68. Prove that if the curve of intersection of two
surfaces is a line of curvature for one surface then it is a line of
curvature for the other surface when the two surfaces are
intersecting each other at a constant angle.
Exercise 69. Outline the role of geodesic torsion in
characterizing the line of curvature employing a mathematical
formulation in this context.
Exercise 70. Give two examples for the line of curvature on
specific types of surface discussing in each case why the
described curve should be a line of curvature.
Exercise 71. Give the formulae for the principal curvatures, κ1
and κ2, when the u1 and u2 coordinate curves of a surface patch
are lines of curvature.
Exercise 72. Using tensor notation, state the mathematical
condition that should be met by a line of curvature on a space
surface with full explanations of all the symbols involved.
Exercise 73. Which types of surface should be excluded from
the following statement: “The lines of curvature on a surface
are represented by an orthogonal net on its spherical image”?
Exercise 74. Prove that on a Monge patch of the form r(u, v)
= (u, v, f(u, v)), the coordinate curves are orthogonal family iff
fufv = 0 identically.
Exercise 75. Give a mathematical condition for a direction on
a surface at a given point to be asymptotic.
Exercise 76. Prove that the asymptotic directions are bisected
by the lines of curvature.
Exercise 77. Give a rigorous technical definition of asymptotic
line.
Exercise 78. Why the second fundamental form at a point of a
surface should vanish in the asymptotic direction?
Exercise 79. Prove Beltrami-Enneper theorem (see Eq. 428↑
and surrounding text).
Exercise 80. One of the characteristic features of asymptotic
line is that the tangent plane to the surface at each point of the
line coincides with the osculating plane of the line at that point.
Why?
Exercise 81. Show that on a smooth surface with orthogonal
families of asymptotic lines the mean curvature is zero.
Exercise 82. Justify the following statement: “The necessary
and sufficient condition for the u1 and u2 coordinate curves to
be asymptotic lines is that e = 0 identically on the u1
coordinate curves and g = 0 identically on the u2 coordinate
curves”.
Exercise 83. Using Eq. 426↑, prove that the generators of a
circular cylinder are asymptotic lines.
Exercise 84. According to the theorem of Beltrami-Enneper
we have: τ2 = − K where τ and K stand for torsion and
Gaussian curvature. Does this mean that τ is imaginary? Fully
justify your answer.
Exercise 85. The angle θ which an asymptotic direction makes
with the principal direction of κ1 is given by:
tan2θ = − (κ1 ⁄ κ2)
where κ1 and κ2 are the principal curvatures. Derive this
equation.
Exercise 86. Classify the asymptotic directions at a given point
on a surface as real and distinct, or real and coincident, or
conjugate imaginary according to the determinant of the
surface covariant curvature tensor at that point.
Exercise 87. Why the classification in the previous question
can also be based on the Gaussian curvature of the point?
Exercise 88. The classification in the two previous questions is
related to the number of asymptotic directions at elliptic,
hyperbolic and parabolic points. How?
Exercise 89. Justify the following statement: “A straight line
contained in a surface is an asymptotic line”.
Exercise 90. State a mathematical condition for two directions
at a given point on a surface to be conjugate directions
explaining all the symbols used.
Exercise 91. What “conjugate families of curves on a surface”
means?
Exercise 92. Show that at hyperbolic and elliptic points each
direction has a unique conjugate direction.
Exercise 93. Show that on a surface represented by:
r = r1(u) + r2(v)
the coordinate curves are conjugate families of curves.
Exercise 94. What is the necessary and sufficient condition for
the u1 and u2 coordinate curves on a smooth surface to be
conjugate families?
Chapter 6
Special Surfaces
There are many classifications to surfaces in 3D spaces
depending on their properties and relations. A few of these
classifications are briefly investigated in the following sections
of this chapter.
6.1 Plane Surface
Planes are simple, ruled, connected, elementary surfaces. The
following statements apply to planes:
1. All the coefficients of the surface curvature tensor vanish
identically throughout plane surfaces.
2. The Riemann-Christoffel curvature tensor vanishes
identically over plane surfaces.
3. The Gaussian curvature K and the mean curvature H vanish
identically over planes.
4. Planes are minimal surfaces (see § 6.7↓).
5. All points on planes are flat umbilical.
6. At any point on a plane surface, κ1 = κ2 = 0 and hence all the
directions are principal directions (or there is no principal
direction).
7. At any point on a plane surface, all the directions are
asymptotic.
8. A sufficient and necessary condition for a surface to be
plane is having an identically vanishing surface curvature
tensor.
9. A sufficient and necessary condition for a surface to be
isometric with the plane is having an identically vanishing
Riemann-Christoffel curvature tensor. The same applies for
identically vanishing Gaussian curvature.
6.2 Quadratic Surface
Quadratic surfaces are defined by the following quadratic
equation:
(433) Aij xi xj + Bi xi + C = 0
where i, j = 1, 2, 3, the coefficients Aij and Bi are real-valued
tensors of rank-2 and rank-1 respectively and C is a real scalar.
There are many degenerate and non-degenerate types of
quadratic surface. However, we consider here only six non-
degenerate types which are probably the most commonly
occurring in differential geometry. These six types are:
ellipsoid, hyperboloid of one sheet, hyperboloid of two sheets,
elliptic paraboloid, hyperbolic paraboloid, and quadric cone.
The first five of these quadratic surfaces have been defined in §
1.4.1↑ using parametric forms.
By rigid motion transformations, consisting of translation and
rotation of coordinate system, whose purpose is to put the
center of symmetry or vertex of these surfaces at the origin of
coordinates and orient their axes and planes of symmetry with
the coordinate lines and coordinate planes, these types can be
given in the following canonical forms where we assume a
Euclidean 3D space with a rectangular Cartesian coordinate
system:
1. Ellipsoid (Fig. 51↓ a):
(434) (x2 ⁄ a2) + (y2 ⁄ b2) + (z2 ⁄ c2) = 1
2. Hyperboloid of one sheet (Fig. 51↓ b):
(435) (x2 ⁄ a2) + (y2 ⁄ b2) − (z2 ⁄ c2) = 1
3. Hyperboloid of two sheets (Fig. 51↓ c):
(436) (x2 ⁄ a2) − (y2 ⁄ b2) − (z2 ⁄ c2) = 1
4. Elliptic paraboloid (Fig. 51↓ d):
(437) (x2 ⁄ a2) + (y2 ⁄ b2) − z = 0
5. Hyperbolic paraboloid (Fig. 51↓ e):
(438) (x2 ⁄ a2) − (y2 ⁄ b2) − z = 0
6. Quadric cone (Fig. 51↓ f):
(439) (x2 ⁄ a2) + (y2 ⁄ b2) − (z2 ⁄ c2) = 0
We note that in the above equations a, b, c are real parameters
(see Footnote 33 in § 8↓), and for convenience we use x, y, z
for x1, x2, x3 respectively. Also, for the first three of these
surfaces the origin of coordinates is not a valid surface point,
as seen from their equations.
Figure 51 Quadratic surfaces (from upper row to lower row
left to right): ellipsoid, hyperboloid of one sheet, hyperboloid
of two sheets, elliptic paraboloid, hyperbolic paraboloid, and
quadric cone.
6.3 Ruled Surface
A “ruled surface”, or “scroll”, is a surface generated by a
continuous translational-rotational motion of a straight line in
space. Hence, at each point of the surface there is a straight
line passing through the point and lying entirely in the surface.
Planes, cones, cylinders and Mobius strips (Fig. 1↑) are
common examples of ruled surface. The parabolic cylinder
(Fig. 52↓) is another example of ruled surface. The different
perspectives of the generating line along its movement are
described as the rulings of the surface. A ruled surface that can
be generated by two different families of lines is called doubly-
ruled surface. Examples of doubly-ruled surface are hyperbolic
paraboloids (Fig. 53↓) and hyperboloids of one sheet (Fig.
54↓). At any point of a regular ruled surface, the Gaussian
curvature is non-positive (K ≤ 0).
Figure 52 Parabolic cylinder as a ruled surface.
Figure 53 Hyperbolic paraboloid as a doubly-ruled surface
where the grid demonstrates the two sets of straight line
rulings.
Figure 54 Hyperboloid of one sheet as a doubly-ruled surface
where the two frames demonstrate the two sets of straight line
rulings.

The tangent surface (see § 6.6↓) of a smooth curve is a ruled

surface generated by the tangent line of the curve. The tangent
plane is constant along a branch, represented by the tangent
line at a given point, of the tangent surface of a curve. If P is a
point on a curve C where C has a tangent surface S, then the
tangent plane to S along the ruling that passes through P
coincides with the osculating plane of C at P. Hence, the
tangent surface may be described as the envelope of the
osculating planes of the curve.
6.4 Developable Surface
As defined previously (see § 3.1↑), a surface that can be
flattened into a plane without local distortion is called
developable surface. A developable surface can also be defined
as a surface that is isometric to the Euclidean plane. In 3D
manifolds, all developable surfaces are ruled surfaces but not
all ruled surfaces are developable surfaces. A ruled surface is
developable if the tangent plane is constant along every ruling
of the surface as it is the case with cones and cylinders. The
neighborhood of each point on a sufficiently smooth surface
with no flat points is developable iff the Gaussian curvature
vanishes identically on the surface.
The generators of a developable surface and their orthogonal
trajectories are its lines of curvature. A developable surface,
excluding cylinder and cone, is a tangent surface of a curve
where the osculating planes of the curve form the tangent
planes of the surface. The collection of normal lines to a
surface S along a given curve C on S make a developable
surface iff C is a line of curvature (see § 5.8↑). Intrinsically,
any developable surface is equivalent (i.e. having the same
metric characteristics) to a plane and hence any two
developable surfaces are isometric to each other.
6.5 Isometric Surface
An isometry is an injective mapping from a surface S to a
surface S̃ which preserves distances. If the mapping preserves
distances but it is not injective it is described as local isometry.
As a consequence of preserving the lengths in isometric
mappings, the angles and areas are also preserved. Examples of
isometric surfaces are cylinder and cone which are both
isometric to plane.
Two isometric surfaces, such as a cylinder and a cone or each
one of these and a plane, appear identical to a 2D inhabitant.
Any difference between the two can only be perceived by an
external observer residing in a reference frame in the
enveloping space. Accordingly, two isometric surfaces possess
identical first fundamental forms and hence any difference
between them, as viewed extrinsically from the embedding
space, is based on the difference between their second
fundamental forms.
Isometry is an equivalence relation and hence it is reflective,
symmetric and transitive, that is for three surfaces S1, S2 and S3
we have:
1. S1 ~ S1.
2. S1 ~ S2 ⇔ S2 ~ S1.
3. If S1 ~ S2 and S2 ~ S3 then S1 ~ S3.
where the symbol ~ represents an isometric relation. If two
sufficiently smooth surfaces have constant equal Gaussian
curvature then they are locally isometric. The mapping relation
between the two surfaces then include three constants
corresponding to the three independent coefficients of the first
fundamental form. A surface of revolution is isometric to itself
in infinitely many ways, each of which corresponds to a
rotation of the surface through a given angle around its axis of
symmetry. As indicated before, any surface is isometric to the
plane iff the Gaussian curvature (or the Riemann-Christoffel
curvature tensor) vanishes identically on the surface.
6.6 Tangent Surface
As stated previously, the tangent surface of a space curve is a
surface generated by the assembly of all the tangent lines to the
curve. The tangent lines of the curve are called the generators
or branches of the tangent surface. Accordingly, the equation
of a tangent surface ST to a curve C is given by:
(440) rT = ri + kTi
where rT is an arbitrary point on the tangent surface, ri is a
given point on the curve C, k is a real variable ( − ∞ < k < ∞),
and Ti is the unit vector tangent to C at ri. The tangent surface
is generated by varying i along C and k along the tangent line.
The tangent surface of a curve is made of two parts: one part
corresponds to k > 0 and the other part corresponds to k < 0
where the curve is a border line between these two parts (see
Fig. 55↓). The two parts of the tangent surface are tangent to
each other along the curve which forms a sharp edge between
the two. The curve is, therefore, called the edge of regression
of the surface. The tangent plane is constant along a branch of
the tangent surface of a curve. This tangent plane is the
osculating plane of the curve at the point of contact of the
branch with the curve. According to the definition of involute,
all the involutes of a curve Ce are wholly embedded in the
tangent surface of Ce. The normal to the tangent surface of a
space curve C at a point of a given ruling Z is parallel to the
binormal line of C at the point of contact of C with Z.
Figure 55 Space curve C and the two parts of its tangent
surface, shaded differently at the top and the bottom.
6.7 Minimal Surface
A minimal surface is a surface whose area is minimum
compared to the area of any other surface sharing the same
boundary. Hence, the minimal surface is an extremum with
regard to the integral of area over its domain. A common
physical example of a minimal surface is a soap film formed
between two coaxial circular rings where it takes the minimal
surface shape of a catenoid (Fig. 10↑) due to the surface
tension. This problem, and its alike of investigations related to
the physical realization of minimal surfaces, may be described
as the Plateau problem. Geometric examples of minimal
surface shapes are planes, catenoids, helicoids (Fig. 11↑) and
ennepers (Fig. 13↑).
Since the mean curvature H of a surface at a given point P is a
measure of the rate of change of area of the surface elements in
the neighborhood of P, a minimal surface is characterized by
having an identically vanishing mean curvature and hence the
principal curvatures at each point have the same magnitude and
opposite signs (see Eq. 382↑). A minimal surface is also
characterized by having an orthogonal net of asymptotic lines
and a conjugate net of minimal lines (see Footnote 34 in § 8↓).
In fact, having an orthogonal net of asymptotic lines is a
sufficient and necessary condition for having zero mean
curvature (refer to § 5.9↑). Among surfaces of revolution,
catenoid is the only minimal surface of this type.
6.8 Exercises
Exercise 1. State three features which are specific to plane
surfaces.
Exercise 2. Why all directions are asymptotic at any point on a
plane surface?
Exercise 3. Show that having an identically vanishing surface
curvature tensor is a necessary and sufficient condition for a
surface to be plane.
Exercise 4. Show that plane is the only connected surface of
class C2 whose all points are flat.
Exercise 5. Give the tensor notation form of the equation that
defines quadratic surfaces.
Exercise 6. Name three types of quadratic surface giving their
canonical equation in Cartesian coordinates.
Exercise 7. A surface is represented parametrically by:
r(u, v) = (u + v, u − v, 2uv).
Obtain the surface representation in canonical Cartesian form
and hence determine its type.
Exercise 8. Make a simple 3D plot of a hyperbolic paraboloid
showing the Cartesian coordinate axes and indicating the
parameters of the surface. Use a computer graphic package if
convenient.
Exercise 9. For which types of quadratic surface the origin of
coordinates is not a valid point on the surface according to
their canonical forms and why? Does this also apply to the
non-canonical forms of these surfaces? Assuming a canonical
form, are there any limiting conditions under which the origin
can be included in these surfaces?
Exercise 10. Find the parametric representation of a cylindrical
surface whose intersection with the xy plane is given by:
4x2 + 9y2 = 1
and whose central axis is the z-axis. What is the type of this
surface?
Exercise 11. What is “ruled surface”? What is the other name
given to this type of surface and why?
Exercise 12. Make simple sketches for plane, cone, cylinder
and Mobius strip that demonstrate their nature as ruled
surfaces.
Exercise 13. What “doubly-ruled surface” means? Give an
example of such a surface with a simple sketch.
Exercise 14. Prove that the tangent plane is constant along a
branch of the tangent surface of a space curve.
Exercise 15. Prove that the hyperbolic paraboloid is a doubly-
ruled surface.
Exercise 16. Show that a helix embedded in a circular cylinder
intersects all the generators of the cylinder with constant angle.
Exercise 17. Show that all points on the tangent surface of a
given space curve are parabolic.
Exercise 18. Justify the following statement: “At any point of
a ruled surface the Gaussian curvature is non-positive”. From
this perspective, discuss singly- and doubly-ruled surfaces.
Exercise 19. Prove that the tangent plane to a cylinder or a
cone is constant along their generators.
Exercise 20. Prove that if P is a point on a curve C where C
has a tangent surface S, then the tangent plane to S along the
ruling that passes through P coincides with the osculating
plane of C at P.
Exercise 21. Define “developable surface” giving several
examples of this type of surface with an explanation of why
they are developable surfaces.
Exercise 22. Give an example of a ruled surface which is not
developable.
Exercise 23. State the condition for a ruled surface to be
developable.
Exercise 24. Why any two developable surfaces are equivalent
to each other by having the same metric characteristics?
Exercise 25. Show that the generators of developable surfaces
are lines of curvature.
Exercise 26. Show that a necessary and sufficient condition for
a ruled surface to be developable is that its Gaussian curvature
vanishes identically.
Exercise 27. What is “isometric mapping”? Give an example
of such a mapping between two common types of surface.
Exercise 28. How two isometric surfaces are seen by a 2D
inhabitant? Can he distinguish between the two and why?
Exercise 29. Prove that isometry is a symmetric relation.
Exercise 30. Demonstrate symbolically that isometric mapping
is an equivalence relation.
Exercise 31. How two isometric surfaces are characterized in
terms of their first and second fundamental forms? Provide
detailed explanations.
Exercise 32. Show that catenoid and helicoid are locally
isometric.
Exercise 33. Why a surface of revolution is isometric to itself
in infinitely many ways? Demonstrate your answer by an
example.
Exercise 34. Define “tangent surface” of a space curve
descriptively and mathematically.
Exercise 35. What is the difference between the “tangent
surface” of a curve and the “tangent plane” of a surface? Make
detailed comparisons between the two.
Exercise 36. Derive the equation representing the tangent
surface of a space curve represented by:
r(t) = (t2, t − 2, t3 + 5).
Exercise 37. Why the tangent surface of a space curve is made
of two sections? How these sections meet on the curve?
Exercise 38. Make a simple 3D sketch of an arbitrary twisted
space curve and its tangent surface showing parts of its two
sections.
Exercise 39. Prove that the curve made by the intersection of
the normal plane of a curve C at a given point P with the
tangent surface of C has a cusp at P.
Exercise 40. Write down the equation representing the tangent
surface of a space curve explaining all the symbols used in the
equation.
Exercise 41. In the context of tangent surface of a curve, what
“branch”, “generator”, and “edge of regression” mean? Make
an attempt to justify these names.
Exercise 42. What is the meaning of “minimal surface”? Give
geometric and physical examples of this type of surface.
Exercise 43. Check if the surface represented parametrically
by:
r(θ, φ) = (coshθ cosφ, coshθ sinφ, θ)
is minimal or not.
Exercise 44. Why minimal surfaces are characterized by
having identically vanishing mean curvature?
Exercise 45. What is the implication of having vanishing mean
curvature H on the principal curvatures of the surface at the
points where H vanishes?
Exercise 46. Should a surface with orthogonal families of
asymptotic lines be a minimal surface? If so, why?
Chapter 7
Tensor Differentiation over
Surfaces
The focus of this chapter is the differentiation of tensor fields
over surfaces, where general surface and space coordinates are
generally assumed. In general, tensor differentiation, whether
covariant or absolute, over a 2D surface follows similar rules
to the rules that apply to tensor differentiation over general nD
curved spaces. Some of these rules are:
1. The sum and product rules of differentiation apply to
covariant and absolute differentiation as usual.
2. The covariant and absolute derivatives of tensors are tensors.
3. The covariant and absolute derivatives of scalars and
invariant tensors of higher ranks are the same as the ordinary
derivatives.
4. The covariant and absolute derivative operators commute
with the contraction of indices.
5. The covariant and absolute derivatives of the metric,
Kronecker and permutation tensors (and their associated
tensors) vanish identically in any coordinate system, that is:
(441) aαβ|γ = 0 aαβ|γ = 0
(442) δαβ|γ = 0 δαδβω|γ = 0
(443) εαβ|γ = 0 εαβ|γ = 0
where the sign | represents covariant or absolute differentiation
with respect to the surface coordinate uγ. Hence, these tensors
should be treated like constants in tensor differentiation.
An exception to these rules is the covariant derivative of the
space basis vectors in their covariant and contravariant forms
which is identically zero, that is:
(444) Ei;j = ∂jEi − ΓkijEk = + ΓkijEk − ΓkijEk = 0
(445) Ei;j = ∂jEi + ΓikjEk = − ΓikjEk + ΓikjEk = 0
but this is not the case with the surface basis vectors in their
covariant and contravariant forms, Eα and Eα, whose covariant
derivatives do not vanish identically. The reason is that, due to
curvature, the partial derivatives of the surface basis vectors do
not necessarily lie in the tangent plane and hence the following
relations:
(446) ∂jEi = + ΓkijEk
(447) ∂jEi = − ΓikjEk
which are valid in the enveloping space and are used in Eqs.
444↑ and 445↑, are not valid on the surface anymore.
At a given point P on a sufficiently smooth surface with
geodesic surface coordinates and rectangular Cartesian space
coordinates, the covariant and absolute derivatives reduce
respectively to the partial and total derivatives at P.
The covariant derivative of the surface basis vectors is
symmetric in its two indices, that is:
(448)
Eα;β =
∂βEα − ΓγαβEγ =
∂αEβ − ΓγβαEγ = Eβ;α
The covariant derivative of the surface basis vectors, Eα;β,
represents space vectors which are normal to the surface with
no tangential component.
The covariant derivative of a differentiable rank-1 surface
tensor A in its covariant and contravariant forms with respect
to a surface coordinate uβ is given by:
(449) Aα;β = (∂Aα ⁄ ∂uβ) − ΓγαβAγ (covariant)
(450) Aα;β = (∂Aα ⁄ ∂uβ) + ΓαγβAγ (contravariant)
where the Christoffel symbols are derived from the surface
metric.
The covariant derivative of a differentiable rank-2 surface
tensor A in its covariant, contravariant and mixed forms with
respect to a surface coordinate uγ is given by:
(451) Aαβ;γ = (∂Aαβ ⁄ ∂uγ) − ΓδαγAδβ − ΓδβγAαδ (covariant)
(452) Aαβ;γ = (∂Aαβ ⁄ ∂uγ) + ΓαδγAδβ +
ΓβδγAαδ (contravariant)
(453) Aβα;γ = (∂Aβα ⁄ ∂uγ) − ΓδαγAβδ + ΓβδγAδα (mixed)
More generally, for a differentiable surface tensor A of type
(m, n), the covariant derivative with respect to a surface
coordinate uγ is given by:
(454)
The covariant derivative of a space tensor with respect to a
surface coordinate uα is formed by the inner product of the
covariant derivative of the tensor with respect to the space
coordinates xk by the tensor xkα. This may be considered as a
form of the chain rule of differentiation. For example, the
covariant derivative of Ai with respect to uα is given by:
(455) Ai;α = Ai;k xkα
The covariant derivative with respect to a surface coordinate uβ
of a mixed tensor Aiα, which is contravariant with respect to
transformation in a space coordinate xi and covariant with
respect to transformation in a surface coordinate uα, is given by
(see Footnote 35 in § 8↓):
(456) Aiα;β = (∂Aiα ⁄ ∂uβ) + ΓijkAkα(∂xj ⁄ ∂uβ) − ΓγαβAiγ
where the Christoffel symbols with Latin and Greek indices are
derived respectively from the space and surface metrics. This
pattern can be easily generalized to a mixed tensor

of type (m, n) which is contravariant in transformations of

space coordinates xi and covariant in transformations of
surface coordinates uα. For example, the covariant derivative
of a tensor
with respect to uγ is given by:
(457)
The above rules can be extended further to include tensors with
space and surface contravariant indices and space and surface
covariant indices. For Example, the covariant derivative of a
tensor

with respect to a surface coordinate uγ, where i and j are space

indices and α and β are surface indices, is given by:
(458)
This example can be easily extended to the most general form
of a tensor with any combination of covariant and
contravariant space and surface indices.
The covariant derivative of the surface basis vector Eα, which
in tensor notation is denoted by xiα, is given by:
(459)

From the last equation, we conclude:

(460) xiα;β = xiβ;α
This is because the Christoffel symbols are symmetric in their
paired indices (see Eq. 66↑) and we have ∂αβxi = ∂βαxi and xjα
xkβ = xkβ xjα. This symmetry has also been established earlier
using vector notation for the surface basis vectors (see Eq.
448↑).
The mixed second order covariant derivative of the surface
basis vectors is given by:
(461)
xiα;βγ =
bαβ;γ ni + bαβ ni;γ =
bαβ;γ ni − bαβ aδω bδγ xiω
where the covariant derivative of the surface covariant
curvature tensor is given, as usual, by:
(462) bαβ;γ = (∂bαβ ⁄ ∂uγ) − Γδαγbδβ − Γδβγbαδ
The covariant differentiation operators in mixed derivatives are
not commutative and hence for a contravariant surface vector
Aγ, for instance, we have:
(463) Aγ;αβ − Aγ;βα = Rγδαβ Aδ
where Rγδαβ is the Riemann-Christoffel curvature tensor of the
second kind for the surface. Similarly, the mixed second order
covariant derivatives of the surface basis vectors satisfy the
following relation:
(464) xiα;βγ − xiα;γβ = Rδαβγ xiδ
This, in fact, is an instance of the general relation:
Aj;kl − Aj;lk = Rijkl Ai
which is found in tensor calculus texts.
As defined in tensor calculus books, the absolute or intrinsic
derivative of a tensor field along a t-parameterized curve in an
nD space with respect to the parameter t is the inner product of
the covariant derivative of the tensor and the tangent vector to
the curve. This identically applies to the absolute derivative of
curves contained in 2D surfaces.
Consequently, the absolute derivative of a tensor field along a
t-parameterized curve on a surface with respect to the
parameter t follows similar rules to those of a space curve in a
general nD space, as stated in the literature of tensor calculus.
For example, the absolute derivative of a differentiable surface
vector field A in its covariant and contravariant forms with
respect to the parameter t is given by:
(465) δAα ⁄ δt = (dAα ⁄ dt) − ΓγαβAγ(duβ ⁄ dt)
(466) δAα ⁄ δt = (dAα ⁄ dt) + ΓαγβAγ(duβ ⁄ dt)
where the Christoffel symbols are derived from the surface
metric. It should be remarked that if A is a space vector field
defined along the above surface curve then the above formulae
will take a similar form but with change from surface to space
coordinates, and hence the curve is treated as a space curve,
that is:
(467) δAi ⁄ δt = (dAi ⁄ dt) − ΓjikAj(dxk ⁄ dt)
(468) δAi ⁄ δt = (dAi ⁄ dt) + ΓijkAj(dxk ⁄ dt)
where the Christoffel symbols are derived from the space
metric.
The absolute derivative of the tensor Aiα, which is defined in
the previous paragraphs, along a t-parameterized surface curve
is given by:
(469)
δAiα ⁄ δt =
Aiα;β(duβ ⁄ dt) =
[(∂Aiα ⁄ ∂uβ) + ΓijkAkα(∂xj ⁄ ∂uβ) − ΓγαβAiγ](duβ ⁄ dt)
The pattern of absolute differentiation, as seen in the above
examples, can be easily extended to a more general type of
tensor with covariant and contravariant space and surface
indices, as done for covariant differentiation.
To extend the idea of geodesic coordinates to deal with mixed
tensors of the type Aiα, a rectangular Cartesian coordinate
system over the space and a geodesic system on the surface are
introduced and hence at the poles the absolute and covariant
derivatives become total and partial derivatives respectively.
As indicated above, the covariant and absolute derivatives of
space and surface metric, permutation and Kronecker tensors
in their covariant, contravariant and mixed forms vanish
identically and hence they behave as constants with respect to
tensor differentiation when involved in inner or outer product
operations with other tensors and commute with these
operators. Similarly, the surface covariant and absolute
derivatives of space metric tensor, space Kronecker tensor,
space permutation tensor and space basis vectors vanish
identically, that is:
(470) gij|γ = 0 gij|γ = 0
(471) δij|γ = 0 δijkl|γ = 0
(472) εijk|γ = 0 εijk|γ = 0
(473) Ei|γ = 0 Ei|γ = 0
where the sign | represents covariant or absolute differentiation
with respect to the surface coordinate uγ. Hence, these space
tensors are in lieu of constants with respect to surface tensor
differentiation.
∇
The nabla based differential operations, such as gradient
and divergence, apply to space surface as for any general
curved space and hence the formulae given in the literature of
tensor calculus for a general nD space can be used with the
substitution of the surface coordinates and surface metric
parameters. For example, the divergence of a surface vector
field Aα is given by:
∇
(474) ⋅A = [1 ⁄ √(a)] ∂α[√(a)Aα]
and the Laplacian of a surface scalar field f is given by:
∇
(475) 2f = [1 ⁄ √(a)] ∂α[√(a)aαβ∂βf]
where aαβ is the surface contravariant metric tensor, a is the
determinant of the surface covariant metric tensor and α, β =
1, 2.
7.1 Exercises
Exercise 1. Summarize the main rules that govern the
differentiation of tensor fields over surfaces and compare these
rules to those of nD spaces (n > 2).
Exercise 2. Is there any rule of tensor differentiation that
applies to nD spaces (n > 2) but not to surfaces? If so, which
and why? State your answer with a full formal explanation.
Exercise 3. At the points of a smooth surface with geodesic
surface coordinates and Cartesian spatial coordinates of a flat
embedding 3D space, what happens to the covariant and
absolute derivatives of tensor fields?
Exercise 4. Derive the following identity: Eα;β = Eβ;α.
Exercise 5. Express the identity in the previous exercise in full
tensor notation.
Exercise 6. Is Eα;β a surface vector or a space vector? Discuss
the possible different meanings of these attributes.
Exercise 7. Explain, in detail, the following equation related to
the covariant derivative of space tensors with respect to surface
coordinates: Ai;α = Ai;k xkα.
Exercise 8. Write down the mathematical expression for the
covariant derivative of the tensor Aδjknβ with respect to the
surface coordinate uγ where the Latin and Greek indices
represent space and surface general coordinates.
Exercise 9. Write down the tensor equation for the covariant
derivative of the surface basis vector xmγ with respect to the
index β.
Exercise 10. Complete the following equation which involves
the tensor B where the indices represent surface coordinates:
Bα;γδ − Bα;δγ = ?
Exercise 11. Give a brief descriptive definition of the absolute
differentiation of a tensor field along a curve.
Exercise 12. What is the other name given to the absolute
differentiation?
Exercise 13. Explain the mathematical pattern of absolute
differentiation of tensor fields along surface curves illustrating
this by an example.
Exercise 14. Is the pattern of absolute differentiation of surface
tensor fields along surface curves identical to the pattern of
absolute differentiation of space tensor fields along space
curves? If there is any difference, identify and explain.
Exercise 15. Write, in expanded form, the mathematical
equation of the following intrinsic derivative: (δBkγ ⁄ δt) where
B is a tensor and k and γ are space and surface indices.
Exercise 16. What are the covariant and absolute derivatives of
space and surface metric, permutation and Kronecker tensors
in their covariant, contravariant and mixed forms?
Exercise 17. Do the operators of covariant and absolute
differentiation with respect to space and surface coordinates
commute with the metric tensor involved in an inner or outer
product with another tensor? Explain why.
Exercise 18. Explain the following identity giving detailed
definitions of all the symbols and notations involved: εijk|γ = 0.
Exercise 19. Do the nabla based differential operations apply
to the surface tensor fields as to the tensor fields in curved
spaces of higher dimensionality?
Exercise 20. What is the Laplacian of a differentiable
∇
coordinate-dependent surface scalar field h (i.e 2h)? Write in
your answer the mathematical equation for this operation
defining all the symbols used.
Exercise 21. Compare the equation in the previous question
with the equation of the Laplacian of a scalar field defined over
a general nD space.
References
• L.P. Eisenhart. An Introduction to Differential Geometry.
Princeton University Press, second edition, 1947.
• J. Gallier. Geometric Methods and Applications for Computer
Science and Engineering. Springer, second edition, 2011.
• P. Grinfeld. Introduction to Tensor Analysis and the Calculus
of Moving Surfaces. Springer, first edition, 2013.
• J.H. Heinbockel. Introduction to Tensor Calculus and
Continuum Mechanics. 1996.
• D.C. Kay. Schaum’s Outline of Theory and Problems of
Tensor Calculus. McGraw-Hill, first edition, 1988.
• R. Koch. Mathematics 433/533; Class Notes. University of
Oregon, 2005.
• E. Kreyszig. Differential Geometry. Dover Publications, Inc.,
second edition, 1991.
• M.M. Lipschutz. Schaum’s Outline of Differential Geometry.
McGraw-Hill, first edition, 1969.
• T. Sochi. Tensor Calculus Made Simple. CreateSpace, first
edition, 2016.
• I.S. Sokolnikoff. Tensor Analysis Theory and Applications.
John Wiley & Sons, Inc., first edition, 1951.
• B. Spain. Tensor Calculus: A Concise Course. Dover
Publications, third edition, 2003.
• J.L. Synge; A. Schild. Tensor Calculus. Dover Publications,
1978.
Author Notes
• All copyrights of this book are held by the author.
• This book, like any other academic document, is protected by
the terms and conditions of the universally recognized
intellectual property rights. Hence, any quotation or use of any
part of the book should be acknowledged and cited according
to the scholarly approved traditions.
8 Footnotes
Footnote 1. The surface coordinates may also be called the
Gaussian coordinates.
Footnote 2. The main focus here is the properties of surfaces
although some of these qualifications can be extended to
surface curves.
Footnote 3. We are assuming the curve is embedded in the xz
plane.
Footnote 4. The “length of a straight line segment” may be
taken as an axiomatic concept.
Footnote 5. The “area of a polygonal plane fragment” may be
taken as an axiomatic concept.
Footnote 6. The meaning of “admissible coordinate
transformation” may vary depending on the context. We are
generally assuming linear transformations.
Footnote 7. The points are usually assumed to be totally
connected so that any point on the curve can be reached from
any other point by passing through other curve points or at
least they are piecewise connected. We also consider mostly
open curves with simple connectivity and hence the curve does
not intersect itself.
Footnote 8. In this formula, t is a generic symbol and hence it
stands for s.
Footnote 9. Since we employ Cartesian coordinates in a flat
space, ordinary derivatives (i.e. d ⁄ ds and d ⁄ dt) are used in this
and the following formulae. For general curvilinear
coordinates, these ordinary derivatives should be replaced by
absolute derivatives (i.e. δ ⁄ δs and δ ⁄ δt) along the curves.
Footnote 10. There is a more technical and rigorous definition
of simple surface which the interested readers are advised to
seek in the literature of topology.
Footnote 11. This should be understood in an infinitesimal
sense or, equivalently, as a vector lying in the tangent plane of
the surface at the given point, as will be seen next.
Footnote 12. Using well-known identities from vector algebra
plus what we will see later in this chapter, we obtain:
|E1 × E2| =
√(|E1|2| E2|2 − [ E1⋅ E2]2) =
√(a11a22 − [a12]2) = √(a)
and hence the above equality is fully justified.
Footnote 13. In this type of statements, the meaning is that we
can find a coordinate system on each surface such that these
conditions are satisfied, as indicated early in the book.
Footnote 14. The purpose of this demonstration, which
employs a typical surface with infinitesimal displacement of
the tangent plane, is to provide a visual qualitative impression
for the idea of Dupin indicatrix. We note that the indicatrix is
not necessarily represented by one of these contours on the
surface as the indicatrix is a quadratic equation based on the
second fundamental form at a particular point P on the surface,
and hence it is not necessarily satisfied by the surface in the
extended neighborhood of P.
Footnote 15. It is circle if the point is umbilical (see § 4.10↑).
Footnote 16. We are considering here the part of these curves
in the neighborhood of the point P on the surface.
Footnote 17. This may be demonstrated non-rigorously as:
(ds ⁄ du1)(ds ⁄ du1) =
(Edu1du1) ⁄ (du1du1) = E
since on the u1 coordinate curves we have: IS = (ds)2 = E(du1)2
(see Eq. 233↑). Hence,
du1 ⁄ ds = 1 ⁄ √(E).
Footnote 18. This can be demonstrated non-rigorously as in
the previous footnote by replacing du1 with du2 and E with G.
Footnote 19. The plane containing n is also characterized by
being orthogonal to the tangent plane to the surface at P.
Footnote 20. Alternatively, we can use a single principal
normal vector for both normal sections with the use of the
principal curvatures instead of their absolute values. However,
the sign of the second term on the right hand side of the above
equations should be selected properly as it can be plus or
minus depending on the choice of n direction.
Footnote 21. In brief, “geodesic coordinates” here stands for a
coordinate system on a coordinate patch of a surface whose u
and v coordinate curve families are orthogonal with one of
these families (u or v) being a family of geodesic curves.
Footnote 22. Or the sum depending on the authors although it
will not be a mean anymore.
Footnote 23. These parallels correspond to the two circles
contacting its two tangent planes at the top and bottom which
are perpendicular to its axis of symmetry (see Fig. 38↑).
Footnote 24. This is the common case, however in some
exceptional cases the surface in the neighborhood of a
parabolic point lies on both sides of the tangent plane.
Footnote 25. The latter may also be called spherical umbilic.
Footnote 26. As seen, the spherical indicatrix may be ascribed
to the vector or to the curve; the meaning should be obvious.
Footnote 27. In this type of statement, which is found in
common textbooks on differential geometry, we may need to
add extra restrictions such as excluding closed surfaces or
adding further conditions like “in the immediate neighborhood
of the two points” to make the statement applicable to all types
of surface.
Footnote 28. To put it in a different way, “geodesic curves”
has two common uses: (a) curves with identically vanishing
geodesic curvature and (b) arcs of optimal length connecting
two points, where (b) in a sense is a subset of (a). Here,
“geodesic curves” is used in the first sense.
Footnote 29. There are some details about this issue that could
be elaborated where different conventions should be
considered. The main issue is that: can the number of principal
directions at a given point on a surface exceed two or not.
Hence, some of the future materials may not be based on a
single convention.
Footnote 30. These conditions can be taken together or
separately.
Footnote 31. The discriminant is:
Δ = 4(b12)2 − 4b11b22
while the determinant is:
b = b11b22 − (b12)2
and hence Δ = − 4b.
Footnote 32. The notation (δv ⁄ δu) is not related to the
notation of absolute derivative (see § 7↑).
Footnote 33. The symbols a, b, c here are defined locally and
hence they should not be confused with similar symbols used
previously; in particular a and b which symbolize the
determinants of the metric and curvature tensors.
Footnote 34. Minimal lines are curves of minimal length.
Footnote 35. An example of such a tensor is xiα which was
discussed earlier, e.g. in § 3.3↑.