Proseminar Curves and Surfaces

Prof. Dr. Franz Pedit

Dr. Allison Tanguay

Handout

The Isoperimetric Inequality

Andreas Hehl
Tuebingen, 9th February 2013

This handout is part of the Proseminar Curves and Surfaces supervised by Prof. Dr.
Franz Pedit and Dr. Allison Tanguay at Eberhard-Karls Universitaet Tuebingen in Wintersemester 2012/2013. It will discuss Topic 4: The Isoperimetric Inequality.
After a short introduction about the History of the Isoperimetric Problem which resulted in
the Isoperimetric Inequality, we will formulate the requirements and the theorem. We will
then give a geometric proof by Jakob Steiner to show an easily comprehensive approach,
followed by a rigorous proof by Erhard Schmidt. This proof may be more technical and
not as elegant as other proofs but it can be given with basic mathematical tools from
Linear Algebra and Analysis.
For further reading and other approaches please refer to [Blaschke].

Contents

Contents
Contents

1 History

2 The
2.1
2.2
2.3
2.4

Isoperimetric Inequality
Formulation . . . . . . .
Proof by Steiner . . . .
Preparation for Proof by
Proof by E. Schmidt . .

3 References

. . . . . . .
. . . . . . .
E. Schmidt
. . . . . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

7
7
8
10
12
15

1 History

1 History
To emphasize the significance of the discussed topic, we will start with the history of the
Isoperimetric Problem/Inequality which mainly has been taken from [Siegel] and [Wiegert].
The so-called isoperimetric problem dates back to antique literature and geometry, giving
physical insight into nature phenomena and answering questions such as why bees build
hives with cells that are hexagonal in shape.
Literary history is dating back the problem to Vergils Aeneid and his tale of the foundation of the city of Carthage. Vergil reports that Queen Dido from Phoenicia was obliged
to flee from her bloodthirsty brother to North Africa. Once there, she made a deal with a
local chieftain: In return for her fortune she would get as much land as she could isolate
with the skin of a single ox. The deal was agreed upon and an ox was sacrificed. Queen
Dido broke the skin of the ox down into very thin strips of leather, tied them together
and constructed a huge semicircle which, together with the natural boundary of the sea,
turned out to be way bigger than anyone would have expected. Upon this land, Carthage
was established. Apparently, the Queen knew the isoperimetric inequality and understood
how to apply this knowledge to gain the best possible solution to her problem.
The history of geometric proofs goes back to the ancient Greeks and was recorded by
Pappus of Alexandria in the fourth century CE. He credited the isoperimetric results to
Zenodorus who lived during the second century BCE and had originally not dealt with
circles but with rectilinear figures. However, according to modern standards their proofs
were incomplete since they apparently did not study the irregular cases. These always
turn out to have less area than the more natural figures they examined and should have
nevertheless been taken into account in order to present any rigorous proof. Archimedes
also studied the problem but his work, like the original writing of Zenodorus, has been lost.
In modern days Steiner realized that the Greek arguments were incomplete and established a better way to proof the inequality by means of showing how any figure that does
not have a circular boundary can be transformed into a new one with the same perimeter but greater area (we see his idea later). However, Weierstrass, by a new formalized
mathematical system, showed that the proof could still not be considered a rigorous proof.
It was Weierstrass himself who did the first rigorous proof as a corollary of his Theory of
Calculus of Several Variables in 1870.
For other approaches (e.g. Hurwitz who used the theory of fourier series) you may refer
to [Blaschke].

2 The Isoperimetric Inequality

2 The Isoperimetric Inequality

2.1 Formulation
Theorem (The Isoperimetric Inequality)
Let c(t) = (x(t), y(t)) be a simple, closed, positively oriented and regular parameterised
C 1 curve with t [a, b]. Denote the area enclosed in the above defined curve c(t) with A.
For a given length l of c(t) = (x(t), y(t)), we then have
l2 4A 0
or equivalently
A

l2
4

with equality iff c(t) is a circle.

2 The Isoperimetric Inequality

2.2 Proof by Steiner

Let c(t) be as described above. First, we will show geometrically that for a given length
l, the biggest area enclosed by c(t) is the area of a circle. This idea is by Jakob Steiner
and is taken from [Froehlich] and [Hopf].
1. The area A must be convex

Figure 1: Reflecting for Non-Convex Areas

If the area is not convex, we can always find two points P1 and P2 on c(t) which
have a connecting line outside of A. Use this line as a mirror axis to reflect c(t)
between the two points to the other side of the connecting line. We see that l has
not changed, however A has become bigger. Thus A has to be convex.
2. A consists of two parts of the same size
Now, choose two points on c(t), call them Q1 and Q2 . While Q1 can be chosen
freely, Q2 has to be the point that bisects c(t) into two parts c1 (t) and c2 (t), with
c1 (t) having exactly the same length as c2 (t). This means that walking along c(t)
starting at Q1 , Q2 will appear exactly after the length 2l . By drawing a line from
Q1 to Q2 , we obtain two new areas A1 and A2 , with A1 + A2 = A (note: A1 and
A2 have to be equal). Again, we can see that if they are not, then WLOG A1 > A2 .
In this case we can again use the connecting line between Q1 and Q2 to reflect A1
to the other side of our connecting line, getting a bigger area A without changing
the length l of c(t).

2 The Isoperimetric Inequality

3. The curves c1 (t) and c2 (t) denote semi-circles

Figure 2: Changing the Angle to Maximise the Area

Again, look at Q1 and Q2 . We showed above that A1 = A2 , therefore it is enough
to show WLOG that c1 (t) is a semi-circle. Suppose c1 (t) is not a semi-circle. Then
there exists a point R on c1 (t) such that the angle at R (which we call ) is not equal
to 90 (note: This is Thales Theorem). Now leave the length of Q1 R and RQ2 fixed
but allow to be changed (and therefore Q1 and Q2 are moved, of course changing
the connecting line between the two points, too) in order to maximise the area of
the triangle Q1 RQ2 . This operation leaves the shaded areas as well as the length of
c1 (t) untouched but for = 90 we have a bigger area for Q1 RQ2 than before. This
however is a contradiction to our assumption. Thus c1 (t) has to be a semi-circle.
Note: While this construction is intuitively accessible, it cannot be seen as a complete
proof of the Isoperimetric Inequality. It merely gives one way to construct the biggest
area enclosed in a curve with fixed length l, namely the circle. Yet, we have not shown
the existence of a (unique) solution, so we have to see this proof as incomplete.

10

2 The Isoperimetric Inequality

2.3 Preparation for Proof by E. Schmidt

We will prepare the proof by Erhard Schmidt using the following lemmata:
Lemma (i) - (Calculation of A)
We can determine A by the following formula:
b

A=
=
=

b
a

1
2

y(t)x (t) dt

y (t)x(t) dt

b
a

(y(t) x(t) y(t)x (t)) dt

Proof
The first equality is Greens Theorem, which is a special case of Stokes Theorem in the
plane. (Stokes Theorem should have been proved in Analysis III.)
The second equality follows from the Fundamental Theorem of Integration and Differentiation. Since c(t) is a closed curve, parameterised by arc-length with t [a,b], we have
that c(a) = c(b):
b
a

y x dt =

b
a

(xy) dt

b
a

x y dt

= [x(b)y(b) x(a)y(a)]
=

b
a

b
a

x y dt

x y dt

Finally, the last equality follows immediately from the second one:
b
a

1 b
1 b
y x dt +
y x dt
2 a
2 a

1 b
1 b
=
y x dt
yx dt
2 a
2 a

1 b
=
(y x yx ) dt
2 a

y x dt =

which concludes this Lemma.

2 The Isoperimetric Inequality

11

Lemma (ii)
We will use the following inequality/equality later in our proof. For better readability and
structure it is shown now and later, we will only quote the result.
Let x, y, z be functions of t and x, y, z C 1 , that is x, y, z exist and are continuous.
Then we have:
(xy zx )2 (x2 + z 2 ) ((x )2 + (y )2 )
with equality (xx + zy )2 = 0 zy = xx
Proof
(x2 + z 2 ) ((x )2 + (y )2 ) (xy zx )2
= x2 (x )2 + x2 (y )2 + z 2 (x )2 + z 2 (y )2 (x2 (y )2 2xy zx + z 2 (x )2 )
= x2 (x )2 + 2xy zx + z 2 (y )2
= (xx + zy )2 0

For equality it has to hold

(xx + zy )2 = 0 xx + zy = 0
zy = xx

Revision (i) - (Inequality of Geometric/Arithmetric Mean)

Recall the Inequality of Geometric/Arithmetic Mean from foundation courses:
Let a, b R+ , then the following inequality holds

ab

a+b
2

with equality a = b
Revision (ii) - (Properties of Special Parameterised Curves)
Remember from previous talks and foundation courses that
c(t)=(x(t),y(t)) is parameterised by arc-length |c(t)| = 1 t

|(x )2 + (y )2 | = 1

c(t)=(x(t),y(t)) is a circle with radius r > 0 x2 + y 2 = r2

12

2 The Isoperimetric Inequality

2.4 Proof by E. Schmidt

This proof was given by E. Schmidt in 1939 and the main concepts have been taken from
[Carmo] and [Hopf]:

Figure 3: Construction of I, I with c(t), k(t) and k(t)

Proof (The Isoperimetric Inequality by E. Schmidt)
Let c(t) = (x(t), y(t)) be an arc-length parameterised and positively oriented curve as
defined above. Then we can find an interval I = [r, +r] such that x(t) I (by setting
two parallel lines touching c(t) and bounding I as seen in Figure 3). Set x(t) in such a
way that it starts on one bound of I WLOG x(0)=+r and x(p1 ) = r. Now define a circle
k(t) = (x(t), z(t)) with radius r and the same x(t) as above in c(t), meaning x(t) of k(t)

2 The Isoperimetric Inequality

13

has the same parameterisation

as in c(t) (You can set z(t) as z(t) = + r2 x(t)2 for

2
0 < t < p1 and z(t) = r x(t)2 for p1 < t < l to ensure this parameterisation of the
circle k(t)). Then k(t) is entirely in I, too. Denote A as the area enclosed in c(t) and B
as the area enclosed by k(t). Then (by Lemma (i)):
A=

l
0

y (t)x(t) dt;

B=

l
0

z(t)x (t) dt = r2

We can add A and B to get:

A + B = A + r2 =

l
0

(y x zx ) dt

l
0

(y x zx )2 dt

(1)

l
2
(x + z 2 ) ((x )2 + (y )2 ) dt

0
=r 2 by R(iv)

l
0

(by L(ii))

(2)

=1 by R(ii)

r dt = lr

Use Revision (i) (setting a=A and b=r2 ) and the calculation above to get the following
inequalities:

A + r2
lr
A r2

2
2

lr
A r2
2
2
2 2
4Ar l r

0 l2 4A

(3)

To get equality in the Isoperimetric Inequality, we have to have equalities instead of

inequalities in all of the above calculations. To get equality in (3) we know by Revision
(i) that since a = b it has to be A = r2 and thus l = 2r without restriction on the
orientation of r. Equality in (1) and (2) implies that
(xy zx )2 = (x2 + z 2 ) ((x )2 + (y )2 )
which, by Lemma (ii), tells us that xx = zy . Substituting this again into the left side
of the above equation gives us

14

2 The Isoperimetric Inequality

(xy zx )2 = (x2 + z 2 ) ((x )2 + (y )2 )

=r 2 by R(ii)

x2 (y )2 2zx xy + z 2 (x )2 = r2

x2 (y )2 + 2(x )2 x2 + z 2 (x )2 = r2
x2 ((x )2 + (y )2 ) +(x )2 (x2 +

=1 by R(ii)

x2 + (x )2 r2 = r2

=1 by R(ii)

) = r2

=r 2 x2 by R(ii)

x2 = r2 ( 1 (x )2 )

=(y )2 by R(ii)

x = ry

To finish this proof, we show that y = rx . To do this we go back to the beginning of our
proof. We now find another Interval I = [
r, r] (by setting the parallel lines perpendicular
to the ones bounding the Intervall I). Again, we can put a circle c(t) in I but this time
c(t) is parameterised as c(t) = (w(t),y(t)). We have to redo the complete proof now for
=
the same A as above but B
r2 . Requiring equalities everywhere as we did before,
2
this leads us to A =
r but since we used the same A as above, we get r = r as well as

xw = yy which finally results (after doing the calculations again) in y = rx .

Eventually, adding x2 and y 2 we get:

x2 + y 2 = r2 ((x )2 + (y )2 ) = r2

=1 by R(ii)

By Revision (ii) then c(t) is a circle.

3 References

15

3 References
[Blaschke] Blaschke, Wilhelm. Die Grundlehren der Mathematischen
Wissenschaften. Berlin: Springer Verlag, 1950. Print.
[Carmo] Carmo, do Manfredo P. Differentialgeometrie von Kurven und
Flchen. Trans. Michael Grter. Wiesbaden: Vieweg Verlag, 1993. Print.
[Froehlich] Froehlich, Steffen. Konvexe Kurven und das Isoperimetrische Problem.
FU Berlin, 2009. Web. 7 Feb 2013.
<http://page.mi.fu-berlin.de/sfroehli/ss2007/vorlesung03.pdf>
[Hopf] Hopf, Heinz. Selected Chapters of Geometry. Trans. H. Samelson
ETH Zuerich, 1940. Web. 7 Feb 2013.
<http://www.math.cornell.edu/~hatcher/Other/hopf-samelson.pdf>
[Siegel] Siegel, Alan. A Historical Review of the Isoperimetric Theorem
in 2-D, and its Place in Elementary Plane Geometry.
Courant Institute of Mathematical Sciences New York University. Web.
7 Feb 2013. <http://www.cs.nyu.edu/faculty/siegel/SCIAM.pdf>
[Wiegert] Wiegert, Jennifer. The Sagacity of Circles: A History of the
Isoperimetric Problem. MathDL. MAA, 2006. Web. 7 Feb 2013.
<http://mathdl.maa.org/mathDL/46/?pa=content&sa=viewDocument
&nodeId=1186&pf=1>