1

Introduction

1.1 Some history

In the words of S.S. Chern, ”the fundamental objects of study in differential geome-
try are manifolds.” 1 Roughly, an n-dimensional manifold is a mathematical object
that “locally” looks like Rn . The theory of manifolds has a long and complicated his-
tory. For centuries, manifolds have been studied as subsets of Euclidean space, given
for example as level sets of equations. The term ‘manifold’ goes back to the 1851
thesis of Bernhard Riemann, “Grundlagen für eine allgemeine Theorie der Functio-
nen einer veränderlichen complexen Grösse” (“foundations for a general theory of
functions of a complex variable”) and his 1854 habilitation address “Über die Hy-
pothesen, welche der Geometrie zugrunde liegen” (“on the hypotheses underlying
geometry”).

However, in neither reference does Riemann make an attempt to give a precise defini-
tion of the concept. This was done subsequently by many authors, including Riemann
himself. 2 Henri Poincaré in his 1895 work analysis situs, introduces the idea of a
manifold atlas.
2 1 Introduction

The first rigorous axiomatic definition of manifolds was given by Veblen and White-
head only in 1931.
We will see below that the concept of a manifold is really not all that complicated;
and in hindsight it may come as a bit of a surprise that it took so long to evolve. Quite
possibly, one reason is that for quite a while, the concept as such was mainly regarded
as just a change of perspective (away from level sets in Euclidean spaces, towards
the ‘intrinsic’ notion of manifolds). Albert Einstein’s theory of General Relativity
from 1916 gave a major boost to this new point of view; In his theory, space-time
is regarded as a 4-dimensional ‘curved’ manifold with no distinguished coordinates
(not even a distinguished separation into ‘space’ and ‘time’); a local observer may
want to introduce local xyzt coordinates to perform measurements, but all physically
meaningful quantities must admit formulations that are coordinate-free. At the same
time, it would seem unnatural to try to embed the 4-dimensional curved space-time
continuum into some higher-dimensional flat space, in the absence of any physical
significance for the additional dimensions. Some years later, gauge theory once again
emphasized coordinate-free formulations, and provided physics-based motivations
for more elaborate constructions such as fiber bundles and connections.
Since the late 1940s and early 1950s, differential geometry and the theory of
manifolds has developed with breathtaking speed. It has become part of the basic ed-
ucation of any mathematician or theoretical physicist, and with applications in other
areas of science such as engineering and economics. There are many sub-branches,
for example complex geometry, Riemannian geometry, and symplectic geometry,
which further subdivide into sub-sub-branches.

1.2 The concept of manifolds: Informal discussion

To repeat, an n-dimensional manifold is something that “locally” looks like Rn . The

prototype of a manifold is the surface of planet earth:
 1.3 Manifolds in Euclidean space 3

It is (roughly) a 2-dimensional sphere, but we use local charts to depict it as subsets

of 2-dimensional Euclidean spaces. 3

To describe the entire planet, one uses an atlas with a collection of such charts, such
that every point on the planet is depicted in at least one such chart.
This idea will be used to give an ‘intrinsic’ definition of manifolds, as essentially
a collection of charts glued together in a consistent way. One can then try to de-
velop analysis on such manifolds – for example, develop a theory of integration and
differentiation, consider ordinary and partial differential equations on manifolds, by
working in charts; the task is then to understand the ‘change of coordinates’ as one
leaves the domain of one chart and enters the domain of another.

1.3 Manifolds in Euclidean space

In multivariable calculus, you will have encountered manifolds as solution sets of
equations. For example, the solution set of an equation of the form f (x, y, z) = a
in R3 defines a ‘smooth’ hypersurface S ✓ R3 provided the gradient of f is non-
vanishing at all points of S. We call such a value of f a regular value, and hence
S = f 1 (a) a regular level set. Similarly, the joint solution set C of two equations

f (x, y, z) = a, g(x, y, z) = b

defines a smooth curve in R3 , provided (a, b) is a regular value of ( f , g) in the sense

that the gradients of f and g are linearly independent at all points of C. A familiar
4 1 Introduction

example of a manifold is the 2-dimensional sphere S2 , conveniently described as a

level surface inside R3 :

S2 = {(x, y, z) 2 R3 | x2 + y2 + z2 = 1}.

There are many ways of introducing local coordinates on the 2-sphere: For example,
one can use spherical polar coordinates, cylindrical coordinates, stereographic pro-
jections, or orthogonal projections onto the coordinate planes. We will discuss some
of these coordinates below. More generally⇤ , one has the n-dimensional sphere Sn
inside Rn+1 ,

Sn = {(x0 , . . . , xn ) 2 Rn+1 | (x0 )2 + · · · + (xn )2 = 1}.

The 0-sphere S0 consists of two points, the 1-sphere S1 is the unit circle. Another
example is the 2-torus, T 2 . It is often depicted as a surface of revolution: Given real
numbers r, R with 0 < r < R, take a circle of radius r in the x z plane, with center
at (R, 0), and rotate about the z-axis.

The resulting surface is given by an equation,

p 2
T 2 = {(x, y, z)| x2 + y2 R + z2 = r2 }. (1.1)

Not all surfaces can be realized as ‘embedded’ in R3 ; for non-orientable surfaces

one needs to allow for self-intersections. This type of realization is referred to as an
immersion: We don’t allow edges or corners, but we do allow that different parts of
the surface pass through each other. An example is the Klein bottle

⇤ Note that we adopt the superscript notation for indices, so that a point in say R4 is writen
as (x1 , x2 , x3 , x4 ). The justification for this convention has to do with tensor calculus which
is beyond the scope of the current book.
 1.4 Intrinsic descriptions of manifolds 5

The Klein bottle is an example of a non-orientable surface: It has only one side. (In
fact, the Klein bottle contains a Möbius band – see exercises.) It is not possible to
represent it as a regular level set f 1 (0) of a function f : For any such surface one
has one side where f is positive, and another side where f is negative.

1.4 Intrinsic descriptions of manifolds

In this course, we will mostly avoid concrete embeddings of manifolds into any RN .
Here, the term ‘embedding’ is used in an intuitive sense, for example as the real-
ization as the level set of some equations. (Later, we will give a precise definition.)
There are a number of reasons for why we prefer developing an ‘intrinsic’ theory of
manifolds.
1. Embeddings of simple manifolds in Euclidean space can look quite complicated.
The following one-dimensional manifold

is intrinsically, ‘as a manifold’, just a closed curve, that is, a circle. The problem
of distinguishing embeddings of a circle into R3 is one of the goals of knot
theory, a deep and difficult area of mathematics.
2. Such complications disappear if one goes to higher dimensions. For example,
the above knot (and indeed any knot in R3 ) can be disentangled inside R4 (with
R3 viewed as a subspace). Thus, in R4 they become unknots.
3. The intrinsic description is sometimes much simpler to deal with than the extrin-
sic one. For instance, the equation describing the torus T 2 ✓ R3 is not especially
simple or beautiful. But once we introduce the following parametrization of the
torus
x = (R + r cos j) cos q , y = (R + r cos j) sin q , z = r sin j,
where q , j are determined up to multiples of 2p, we recognize that T 2 is simply
a product:
T 2 = S1 ⇥ S1 . (1.2)
That is, T 2 consists of ordered pairs of points on the circle, with the two factors
corresponding to q , j. In contrast to (1.1), there is no distinction between ‘small’
circle (of radius r) and ‘large circle’ (of radius R). The new description suggests
an embedding of T 2 into R4 which is ‘nicer’ then the one in R3 . (But does it
help?)
4. Often, there is no natural choice of an embedding of a given manifold inside RN ,
at least not in terms of concrete equations. For instance, while the triple torus
6 1 Introduction

is easily pictured in 3-space R3 , it is hard to describe it concretely as the level

set of an equation.
5. While many examples of manifolds arise naturally as level sets of equations
in some Euclidean space, there are also many examples for which the initial
construction is different. For example, the set M whose elements are all affine
lines in R2 (that is, straight lines that need not go through the origin) is naturally
a 2-dimensional manifold. But some thought is required to realize it as a surface
in R3 .

1.5 Surfaces
Let us briefly give a very informal discussion of surfaces. A surface is the same
thing as a 2-dimensional manifold. We have already encountered some examples:
The sphere, the torus, the double torus, triple torus, and so on:

All of these are ‘orientable’ surfaces, which essentially means that they have two
sides which you might paint in two different colors. It turns out that these are all
the orientable surfaces, if we consider the surfaces ‘intrinsically’ and only consider
surfaces that are compact in the sense that they don’t go off to infinity and do not
have a boundary (thus excluding a cylinder, for example). For instance, each of the
following drawings depicts a double torus:

We also have one example of a non-orientable surface: The Klein bottle. More ex-
amples are obtained by attaching handles (just like we can think of the torus, double
torus and so on as a sphere with handles attached).
 1.5 Surfaces 7

Are these all the non-orientable surfaces? In fact, the answer is no. We have missed
what is in some sense the simplest non-orientable surface. Ironically, it is the surface
which is hardest to visualize in 3-space. This surface is called the projective plane
or projective space, and is denoted RP2 . One can define RP2 as the set of all lines
(i.e., 1-dimensional subspaces) in R3 . It should be clear that this is a 2-dimensional
manifold, since it takes 2 parameters to specify such a line. We can label such lines
by their points of intersection with S2 , hence we can also think of RP2 as the set of
antipodal (i.e., opposite) points on S2 . In other words, it is obtained from S2 by iden-
tifying antipodal points. To get a better idea of how RP2 looks like, let us subdivide
the sphere S2 into two parts:
(i) points having distance  e from the equator,
(ii) points having distance e from the equator.

If we perform the antipodal identification for (i), we obtain a Möbius strip. If we

perform antipodal identification for (ii), we obtain a 2-dimensional disk (think of it
as the points of (ii) lying in the upper hemisphere). Hence, RP2 can also be regarded
as gluing the boundary of a Möbius strip to the boundary of a disk:

Now, the question arises: Is it possible to realize RP2 smoothly as a surface inside
R3 , possibly with self-intersections (similar to the Klein bottle)? Simple attempts of
joining the boundary circle of the Möbius strip with the boundary of the disk will
always create sharp edges or corners – try it. Around 1900, David Hilbert posed
8 1 Introduction

this problem to his student Werner Boy, who discovered that the answer is yes. The
following picture of Boy’s surface was created by Paul Nylander.

There are some nice videos illustrating the construction of the surface: See in partic-
ular
https://www.youtube.com/watch?v=9gRx66xKXek
and
www.indiana.edu/˜minimal/archive/NonOrientable/NonOrientable/
Bryant-anim/web/
While these pictures are very beautiful, it certainly makes the projective space appear
more complicated than it actually is. If one is only interested in RP2 itself, rather than
its realization as a surface in R3 , it is much simpler to work with the definition (as a
sphere with antipodal identification).
Going back to the classification of surfaces: It turns out that all closed, connected
surfaces are obtained from either the 2-sphere S2 , the Klein bottle, or RP2 , by attach-
ing handles.
Remark 1.1. Another operation for surfaces, generalizing the procedure of ‘attaching
handles’, is the connected sum. Given two surfaces S1 and S2 , remove small disks
around given points p1 2 S1 and p2 2 S2 , to create two surfaces with boundary cir-
cles. Then glue-in a cylinder connecting the two boundary circles, without creating
edges. The resulting surface is denoted
S1 ]S2 .
For example, the connected sum S ]T 2 is S with a handle attached. You may want to
think about the following questions: What is the connected sum of two RP2 ’s? And
what is the connected sum of RP2 with a Klein bottle? Both must be in the list of
2-dimensional surfaces given above.
 NOTES 9

Notes
1 Page 332 of Chern, Chen, Lam: Lectures on Differential Geometry, World Scientific
2 See e.g. the article by Scholz http://www.maths.ed.ac.uk/ aar/papers/scholz.pdf for the
long list of names involved.
3 Note that such a chart will always give a somewhat ‘distorted’ picture of the planet; the

distances on the sphere are never quite correct, and either the areas or the angles (or both)
are wrong. For example, in the standard maps of the world, Canada always appears somewhat
bigger than it really is. (Even more so Greenland, of course.)