Professional Documents
Culture Documents
∗
2018-Summer Study Report
by BHAVNA JOSHI
Mentor: ATHARVA KORDE
∗
Organizer: Maths and Physics Club, IIT Bombay
Preface
I have written this report to conclude my study on the topic, with my best
possible effort. Throughout, I try to interrogate and summarize, while main-
taining a conversational tone. All that I’ve been able to learn and write is due
to the lectures by Professor Claudio Arezzo on differential geometry. Along
with it reference book used was Differential geometry of curves and surfaces by
do carmo.
I want to thank the Maths and Physics Club organizers for the
encouragement.
Introduction
The two aspects of Differential geometry are local and global. Local differ-
ential geometry is the study of properties of a geometric structure that (at each
point) depend only on the neighborhood of a given point. Global properties,
however, deal with the entire manifold (or at least properties of some region
larger than an arbitrarily small neighborhood)–properties that can change when
you alter a part of the manifold
A curvature can be intrinsic and extrinsic. Extrinsic tells how lower dimen-
sional objects are embedded in higher ones, it is something only detectable by
a higher dimensional being. Intrinsic curvature such as Gaussian curvature is
detectable to the inhabitants of a surface and not just outside observers.
CURVES
Starting with building differential geometry for lowest dimensional curved
space, 0D, we find a curved point makes no sense. 1D, a curve which can have
only extrinsic curvature. 2D is a surface so it can have both types of
curvatures and the complexity from this dimension onward increases enough
to extend theory for even higher dimensions.
1.Definition
Curves: An infinitely differentiable (C ∞ ) parametrized curve is a map
α:I → R3 , I -(a,b) ⊂ R
α(t) = (x(t), y(t), z(t)) where x,y,z are (C ∞ ) i.e. smooth functions
α’(t) = (x’(t), y’(t), z’(t)) is the tangent(or velocity) vector at α(t).
Planar curves: α is called planar if ∃ a plane P ⊂ R3 such that the image set
or Trace of map α is a subset of that plane, α(I) ⊂ P.
For planar curves, by rigid motions we can always set P as, say, the plane z=0
and α will be of the form
α(t) = (x(t), y(t), 0)
• Helices
t t t
α(t) = (a cos( √ ), a sin( √ ), b √ )
a2 + b 2 a2 + b2 a2 + b 2
• Twisted cubic (in projective 3-space): The twisted cubic is formed by
taking a parabola in the xy plane and bending the ends in the z direction
so that the shadow of the curve on the xz plane looks like a cubic. Note
that, while the parameterization of the function is extremely simple, its
arclength and Frenet frame are complicated.
α(t) = ~r(t) = t, t2 , t3
• Toroidal spiral
~r(t) = h(2 + sin 3t) cos t, (2 + sin 3t) sin t, cos 3ti
~s(t) = h(2 + sin 6t) cos 2t, (2 + sin 6t) sin 2t, cos 6ti.
Notice that ~s(t) = ~r(2t)
So as observed above, both these vector-valued functions will trace out the
same space curve.
3.Geometric properties of a curve
Length
For defining length of any curve
α:I → R3 , I = [a, b] ⊂ R
We make a choice of partition of [a,b] and consider set P of all partition points.
P = {t0 = a < t1 < t2 < ... < tn = b}
As the norm minimizes, the partitions become denser and denser, length
approaches more to true value of curve length
Z b
s(α) = |α0 (t)|dt
a
By mean value theorem, | α(ti ) − α(ti−1 ) |= f (βi , γi , δi )(ti − ti−1 ) for some
(βi , γi , δi ) ∈ [ti−1 , ti ]
Hence
n
X n
X
b
L a (α, P ) = | α(ti ) − α(ti−1 ) |= f (βi , γi , δi )(ti − ti−1 )
i=1 i=1
and
Z b n Z
X t n Z
X t n
X
0 0 0
|α (t)|dt = |α (t)|dt = | α (pi ) | (ti −ti−1 ) = f (pi , pi , pi )(ti −ti−1 )
a i=1 ti−1 i=1 ti−1 i=1
where pi ∈ [ti−1 , ti ]
Hence proved.
Coments
• Now arises one more complication, if we get equality in the above case, does it
necessarily mean that map α(a) - α(b) is segment? No, curve may look like a
segment but really the particle (just the point on the curve moving
corresponding to t) can be performing motion back and forth between the
points a and b before reaching b based on its parametrization, so the
parametrized path travelled will be more than just, segment-the geometric
path we call length.
Notice, length is not able to reconstruct the way the particle moves. From
geometric point of view it is unimportant what the motion is on the curve as
long as same image is produced.
• Hence we find, there can be infinite number of parametrizations of the same
geometric object. This makes us ask certain questions:
– Given a parametrized equation of a curve, how do we go back to
geometric properties of original* curve?
– what properties are independent of the way the particle moves?
– We very comfortably used the definition of curves as smooth function
but there’s a price that we’re paying: several geometric things can be
described in completely different analytic ways. What is the way to
kill this freedom of possible parametrizations?
– If that’s the case, is there a best possible choice of a
parametrization(s)?
• Given a diffeomorphim
φ : J → I (I and J are open Intervals)
φ is smooth(C∞ ), invertible with smooth (C ∞ ) inverse
and α : I → R3 ,
we can construct a new curve β : J → R3 as α composite φ : β = α ◦ φ
Now β maybe a different map and thus a different curve, but for a geometer it
is the same curve since image of α and β is the same. β is called a
reparametrization of α.
Proposition:
If φ: J → I is a dif f eomorphism
α:I → R3 , [a,b] ⊂ J
φ([a,b])=[c,d]
then Lb a (α◦φ) = Lc d (α)
We consider α: I → R3 and t0 ∈ I
define s: I → R toR be arc length
t
s takes t (∈ I) → t0 |α0 (t)|dt= La b (α)
Curves whose α’(t) never vanishes, have a well defined s(t) which is a
diffeomrophism. Such curves are called regular curves.
The s function takes in parameter t and gives the value of arc length between
the reference point α(t0 ), and α(t).
The φ function is inverse of s (defined because s is regular). Thus its domain
is the set of arc lengths and it outputs that corresponding point t.
Comments
• We have one natural vector of R3 associated with the curve, namely, α0 (s)
at every s, call it T(s) the tangent vector. where | T (s) |= 1 ∀s
• <T(s),T(s)>=1 (scalar product)
<T’(s),T(s)>+ <T(s),T’(s)>= 0 (taking derivative)
<T’(s),T(s)>= 0
Thus derivative of tangent vector is always orthogonal to tangent.
This induces natural definition of a function: let k(s) = | T 0 (s) | and call it
curvature of α at s . Curvature is only continuous in general, it can be smooth
if | T 0 (s) |6= 0∀s
From now on we will assume k(s)>0 ∀s (which means we are eliminating
possibility of even a piece of line)
T 0 (s) T 0 (s)
let N(s) = =
| T 0 (s) | | k(s) |
and call it the unit normal vector to α at s.
The plane determined by T(s) and N(s) is called the osculating plane.
• we have now 2 vectors defined at every point of the curve, those are T(s) and
N(s). to form a complete orthonormal basis of R3 , we want one more vector
perpendicular to both.
Note: T(s) is a set of all tangent vectors t(s) at all points on the curve.
Likewise for N(s) and B(s).
To each value of the parameter s, there are three orthogonal unit vectors t(s),
n(s), b(s). The derivatives, called The Frenet Serret formulas are:
T’(s) = k(s)N(s)
N’(s) = -k(s)T(s) - τ (s)B(s)
B’(s) = τ (s) N(s)
• lines: α(t)=tv+v0
automatically | v | = 1 as parametrized by arc length.
– T(t)=v
– ∴ T 0 (t) = 0 ⇒ k(s) = 0
• Circle:
α(t)= r(cost(t/r), sin(t/r), 0) + c (notice: its planar)
−cos(t/r) −sin(t/r)
T 0 =( , , 0) ∴ curvature k(t) = 1/r (k>0)
r r
T 0 (t)
N = (−cos(t/r), −sin(t/r), 0) =
k(t)
B(t) = (0, 0, 1)
B is constant ⇒ B 0 = 0 ∴ τ =0
−asin(t/c) acos(t/c) b
T(t) = α’(t)= ( , , ) and | T (s) |= 1
c c c
−acos(t/c) −asin(t/c) |a|
T’(t) = ( , , 0) thus k=
c2 c2 c2
−b
τ= (from B’)
c2
Theorem for interpretting k and τ
Theorem: α : I → R3 parametrized with arc length, k>0, is planar iff τ =0
proof: A]
suppose τ =0
∴ B0 = 0
i.e. B=constant vector and | B |= 1
∴ consider<α(s),B>=a
taking derivative <T, v>=0 is satisfying the requirement of T ⊥ B
so, the solution is: <α(s),B>=a which is a plane.
∴ if τ =0 then curve is planar
B]
consider planar curve <α(s),B>=a
<T,v>=0
∴ B = vaconstant ⇒ B 0 = 0
∴τ =0
∴ if curve is planar then τ =0
τ measures the effort you have to put in twisting a planar curve to a non
planar
Now, the way to prove that there is nothing more to discover is to prove that
if given k and τ there will exist a unique (up to rigid motions) curve having
that curvature k and torsion τ . Because if not, then there must be some more
variable(s) about curves that describe it and we are yet to define them.
Why study curves and surfaces? Curves and surfaces are to geometry what
numbers are to algebra. They form the basic ingredients of our visual
perception and inspire the development of far reaching mathematical tools.
Yet despite centuries of pure study, not to mention a wealth of growing
applications in science and technology, there are still numerous open problems
in this area which are strikingly intuitive and elementary to state, pointing to
fundamental gaps in our conceptions of space and shape. According to Ezra
Pound, music begins to atrophy when it departs too far from the dance and
poetry begins to atrophy when it gets too far from music. Perhaps the same
can be said of geometry, and indeed all of mathematics, if it looses sight of its
natural building blocks and connections to the physical world. So, Differential
geometry arose and developed as a result of and in connection to the
mathematical analysis of curves and surfaces. (source linked below)
Historical Outline:
• 250 B.C.; Isoperimetric Problem
• 200 A.D.; Pappus of Alexandria
• 50 A.D.; Dido’s Problem
• 1736; The general idea of natural equations for obtaining curves from
local curvature appears to have been first considered by Leonhard Euler.
• 1795; Monge’s paper.
• 1827; Gauss published his article, titled ’Disquisitiones Generales Circa
Superficies Curvas’
• 1854; The intrinsic geometry of Gauss was generalized to higher
dimensions by B. Riemann.
Q. Among all planar simple closed curves of given length L, which curve
bounds the region of largest area? Indeed intuitive, but not easy to prove,
called the Jordan seperation theorem (one of the early success of modern
topology).
There are various attempts for proving this using basic mathematical tools
from Linear Algebra and Analysis, which can be found
here:(https://www.math.uni-tuebingen.de/ab/GeometrieWerkstatt/
IsoperimetricInequality.pdf)
Thank you for reading. If you find anything wrong or misleading or want to make a comment, please send
your views on: bhavnajoshi45@gmail.com.