You are on page 1of 16

Brouwer Degree in Two Dimensions

Bob Lutz
Pomona College 13
August 29, 2011
1 Introduction
Having developed out of early 20
th
century work in topology by J. Hadamard
and L. E. J. Brouwer, degree theory is fundamentally concerned with the exis-
tence, number, and nature of solutions to the equation
f(x) = y
in a suitable space. For our purposes, this will be R
2
.
Because its value can guarantee solutions within a domain, the Brouwer de-
gree in particular is a useful tool in the study of nonlinear dierential equations.
The objective of these notes is to motivate a basic comprehension of degree the-
ory to this end by providing an analytical discussion of the Brouwer degree in
two dimensions. We begin by introducing the Kronecker index, an analogous
concept with a tractable integral formulation.
Let F := P(x, y)+Q(x, y) C
2
(U, R
2
) where U R
2
is open, and consider
, the angle of F at (x, y) relative to the positive x-axis, given by
= tan
1
_
Q
P
_
for P = P(x, y) and Q = Q(x, y). Consider also
:= d =
P dQQdP
P
2
+ Q
2
, (1)
which describes the change in as (x, y) varies. When integrated over a closed
curve, measures the total change in the angle of F. We formalize this concept
with the planar notion of the Kronecker index.
Denition 1.1 (Kronecker index). Let F be dened as above, and let D
D U be open such that D is a positively oriented C
1
Jordan curve. Suppose
that F(x, y) ,= (0, 0) for all (x, y) D. The Kronecker index of F around D
is given by
index
K
(F, D)
1
2
_
D
. (2)
1
This expression calculates the number of times the curve F(D) winds
around the origin; ttingly, this is called the winding number of F(D) in the
complex plane. More pertinently, however, the integral expression in (2) also
describes the planar Brouwer degree for the case in which F is C
1
. These
connections (particularly the latter) will be developed throughout the notes as
we expand the above denition by using the Kronecker index to address the
Brouwer degree generally for the case of continuous F and D.
2 Examples
We present several examples and provide computations of the Kronecker index
for each.
2.1 Identity map
Let id be the identity map dened by
id(x, y) = x + y for all (x, y) R
2
, (3)
and let D be the unit disk. From (2) we have that
index
K
(F, D)
1
2
_
S
1
xdy y dx
x
2
+ y
2
.
Parameterizing S
1
by (x, y) = (cos t, sin t) with t [0, 2] gives
1
2
_
S
1
xdy y dx
x
2
+ y
2
=
1
2
_
2
0
cos
2
t + sin
2
t
cos
2
t + sin
2
t
dt
=
1
2
_
2
0
dt
= 1.
Hence, index
K
(F, D) 1.
2.2
Let F(x, y) := y + x for all (x, y) R
2
, and let D be the unit disk. From
(2) we have that
index
K
(F, D)
1
2
_
S
1
xdy y dx
x
2
+ y
2
.
As in 2.1, the parameterization (x, y) = (cos t, sin t) of S
1
gives
index
K
(F, D)
1
2
_
2
0
dt = 1.
2
2.3
Let F(x, y) := (x
2
y
2
) + 2xy for all (x, y) R
2
, and let D be the unit disk.
From (2) we have that
index
K
(F, D)
1
2
_
S
1
(x
2
y
2
) d(2xy) 2xy d(x
2
y
2
)
(x
2
y
2
)
2
+ (2xy)
2
. (4)
Note that
d(2xy) = 2xdy + 2y dx (5)
and
d(x
2
y
2
) = 2xdx 2y dy. (6)
After a bit of algebra, combining (4), (5), and (6) yields
index
K
(F, D)
1
2
_
S
1
(2x
3
+ 2xy
2
) dy (2x
2
y + 2y
3
) dx
(x
2
+ y
2
)
2

_
S
1
xdy y dx
x
2
+ y
2
.
As in 2.1 and 2.2, the parameterization (x, y) = (cos t, sin t) of S
1
gives
index
K
(F, D)
1

_
2
0
dt = 2.
Alternatively, note that F(sin t, cos t) = (sin 2t, cos 2t) on S
1
.
3 Properties of the Kronecker index
It is necessary to establish several properties of the Kronecker index in R
2
. We
denote by | | the Euclidean norm on R
2
, and by | |
C
1
(U)
the C
1
norm on U
given by
|G|
C
1
(U)
= sup
(x,y)U
|G| + sup
(x,y)U
|DG|
for G C
1
(U), where DG denotes the derivative of G. We refer to Stokes
theorem as given in [5, p. 124], i.e.
_
D
=
_
D
d,
where is a dierential 1-form on D.
3.1 Winding number
Given the introductory explanation and examples, it may seem intuitive that
the Kronecker index is integer-valued. To prove this fact, however, we must
examine the index in the complex plane. Expressed as a special case of the
Cauchy integral formula, the complex winding number is readily comparable to
our integral formulation of the Kronecker index.
3
Denition 3.1. Suppose that F C
1
(U, C) where U C is open, and let
D U be open such that D is a positively oriented C
1
Jordan curve. The
winding number of F around D is given by
1
2i
_
D
dF
F
.
Following this denition, suppose that F, U, and D are dened as above,
and write F(x + iy) P(x, y) + iQ(x, y), so that dF = dP + i dQ. We have
that ln F = ln [F[ + i arg F; a brief calculation yields
1
2i
_
D
dF
F
=
1
2i
_
D
d(ln F)
=
1
2i
_
D
d(ln [F[) +
1
2
_
D
d(arg F)
=
1
2
_
D
d
_
tan
1
Q
P
_
= index
K
(F, D),
noting that
_
D
d (ln [F[) = 0, since exact dierential forms are closed.
Using this formulation, it can be shown that the winding number (or, equiv-
alently, the Kronecker index) is an integer. We adapt the argument from [4,
p. 115] and [3, p. 10].
Lemma 3.2. Let F C
1
(U, R
2
), and suppose that F ,= (0, 0) for all (x, y)
D. Then, index
K
(F, D) is an integer.
Proof. Let : [0, 2] D be a parameterization dened by (t) = (x(t), y(t)),
and let (t) := x(t) + iy(t). Consider the function
h(t) :=
_
t
a
d
d
F(())
F(())
d for all t [a, b].
By the fundamental theorem of calculus along curves, we have that
h

(t) =
d
dt
F((t))
F((t))
.
Consider also g(t) := e
h(t)
F((t)). We have that
g

(t) =
d
dt
[F((t))] e
h(t)
h

(t)e
h(t)
F((t))
= e
h(t)
_
d
dt
F((t)) h

(t)F((t))
_
= 0 for all t [a, b].
Thus, g(t) is constant, so g(a) = F((a)) = g(t) for all t [a, b]. It follows that
e
h(t)
=
F((t))
F((a))
.
4
Since (a) = (b), we have that e
h(b)
= 1. Hence, h(b) = 2mi for some m Z,
so that
index
K
(F, D) =
h(b)
2i
= m.
3.2 Basic results
The behavior of index
K
(F, D) depends on the existence and location of equi-
librium points of F: i.e., points (x, y) U where F(x, y) = (0, 0). The following
several theorems describe this relationship explicitly. Initially, we address only
the case of C
2
vector elds and C
1
curves, but the concepts are subsequently
generalized as we address analogous properties of the Brouwer degree.
Unless stated otherwise, U and D are dened as in Denition 1.1. We
begin with the solution property, a result which can guarantee the existence of
equilibrium points of F in D.
Theorem 3.3 (Solution property of the Kronecker index). Let F = P +Q
C
2
(U, R
2
) be such that F ,= (0, 0) for all (x, y) D. If index
K
(F, D) ,= 0,
then there exists (p, q) D such that F(p, q) = (0, 0).
Proof. We prove the contrapositive. Suppose that F(x, y) ,= (0, 0) for all (x, y)
D, and let be dened as in (1). Then is a dierential 1-form on D, since
P
2
+ Q
2
,= 0 on D. We have by Stokes theorem that
index
K
(F, D) =
_
D
d, (7)
where
d = d
_
P
P
2
+ Q
2
_
dQd
_
Q
P
2
+ Q
2
_
dP, (8)
since d
2
Q = d
2
P = 0 for C
2
elds P and Q. Note that
d
_
P
P
2
+ Q
2
_
=
dP
P
2
+ Q
2

2
(P
2
+ Q
2
)
2
(P
2
dP + PQdQ) (9)
and
d
_
Q
P
2
+ Q
2
_
=
dQ
P
2
+ Q
2

2
(P
2
+ Q
2
)
2
(PQdP + Q
2
dQ). (10)
Combining (8), (9), and (10) and applying the anti-commutativity of the wedge
product yields
d
2
=
1
P
2
+ Q
2
dP dQ +
PQdP + Q
2
dQ
(P
2
+ Q
2
)
2
dP
P
2
dP + PQdQ
(P
2
+ Q
2
)
2
dQ.
We use the fact that dQ dQ = dP dP = 0 to obtain
d = 2
_
1
P
2
+ Q
2

1
P
2
+ Q
2
_
dP dQ = 0.
Hence, the statement follows from (7).
5
Much of dierential equation theory investigates conditions which guarantee
solutions within a domain; given this statement of the solution property, then,
its application to the study of dierential equations is more clear. We make use
primarily of the contrapositive in subsequent proofs, most notably in that of the
excision property.
Parameterizing the boundary of a given D as in Section 2 is often dicult,
if not impossible, when computing the Kronecker index. The excision property
allows us to excise open subsets of D not containing equilibrium points of F
while maintaining the value of the index, in order to attain a region with a more
manageable boundary.
Theorem 3.4 (Excision property of the Kronecker index). Let D
1
D be open
such that D
1
is a positively oriented C
1
Jordan curve. Suppose F(x, y) ,= (0, 0)
on D D
1
. Then,
index
K
(F, D) = index
K
(F, D
1
).
Proof. We have by the Jordan curve theorem that D
1
lies in the interior of
D. Since F(x, y) ,= (0, 0) on D D
1
, it follows from the contrapositive of the
solution property that
index
K
(F, D D
1
)
1
2
_
DD1
= 0.
Note that (D D
1
) = DD
1
, since D
1
is traversed as illustrated below.
Figure 1: Positively orienting (D D
1
).
By Stokes theorem, we have that
_
DD1
=
_
D
d
_
D1
d = 0.
Hence,
_
D
=
_
D1
,
from which the statement follows.
6
In some cases, D may be open but disconnected. However, it is not imme-
diately clear from Denition 1.1 how to calculate the index of F around the
corresponding D. This discrepancy is addressed by the additivity property,
which provides a convenient sum formulation for the index in such cases.
Theorem 3.5 (Additivity property). Suppose that D is the union of disjoint
open sets D
1
, . . . , D
m
where D
i
is a C
1
Jordan curve for i = 1, . . . , m. Addi-
tionally, suppose that F(x, y) ,= (0, 0) for all (x, y) D
i
. Then,
index
K
(F, D) =
m

i=1
index
K
(F, D
i
).
Proof. Suppose that
i
: [a, b] D
i
is a parametrization for i = 1, . . . , m.
Then we have for
i
=
i
(t) that
index
K
(F, D) =
1
2
_
D
P(x, y) dQQ(x, y) dP
P(x, y)
2
+ Q(x, y)
2
=
1
2
m

i=1
_
b
a
P(
i
)
d
dt
Q(
i
) Q(
i
)
d
dt
P(
i
)
P(
i
)
2
+ Q(
i
)
2
dt
=
m

i=1
index
K
(F, D
i
).
While it is dicult to generalize about the behavior of the Kronecker index,
the normalization property provides a simple evaluation for a specic case of F.
Theorem 3.6 (Normalization property). Let id : R
2
R
2
be dened as in
Example 2.1. Then,
index
K
(id, D) =
_
1 if (0, 0) D
0 if (0, 0) / D.
Proof. Suppose (0, 0) D. Since D is open, there exists > 0 such that
D

= (x, y) : x
2
+ y
2
< D with F(x, y) ,= (0, 0) on S
1

= D

. Note that
S
1

is C
1
, and that (0, 0) / id(D D

). We have that
index
K
(F, D

)
1
2
_
S
1

xdy y dx
x
2
+ y
2
.
Parametrizing S
1

by (x, y) = ( cos t, sin t) gives


index
K
(F, D)
1
2
_
2
0

2
dt = 1.
Hence, by the excision property,
index
K
(F, D) = index
K
(F, D

) = 1.
Next, suppose (0, 0) / D. Then we have that id(x, y) ,= (0, 0) for all (x, y) D.
It follows from the contrapositive of the solution property that
index
K
(F, D) = 0.
7
4 Dening the Brouwer degree
We begin our approach of the Brouwer degree with a proof of the continuity
of the Kronecker index. This will allow us to properly address the use of close
approximations of F in dening the degree.
Lemma 4.1. Let F, G C
1
(U, R
2
), and suppose that F, G ,= 0 for all (x, y)
D. Then, for every > 0, there exists > 0 such that
|F G|
C
1
(D)
< [index
K
(F, D) index
K
(G, D)[ < .
Proof. Let F := F
1
(x, y) +F
2
(x, y) and G be dened similarly. We have that
[index
K
(F, D) index
K
(G, D)[
=
1
2

_
D
F
1
dF
2
F
2
dF
1
F
2
1
+ F
2
2

_
D
G
1
dG
2
G
2
dG
1
G
2
1
+ G
2
2

Let : [a, b] D be a parameterization dened by (t) = (x(t), y(t)), and let


: R
4
R be dened by
(p, q, r, s) =
ps qr
p
2
+ q
2
. (11)
Writing
F,x
= (F
1
, F
2
, (F
1
)
x
, (F
2
)
x
) and similarly for
F,y
,
G,x
, and
G,y
,
we have after grouping terms that
1
2

_
D
_
F
1
dF
2
F
2
dF
1
F
2
1
+ F
2
2

G
1
dG
2
G
2
dG
1
G
2
1
+ G
2
2
_

=
1
2

_
b
a
[(
F,x

G,x
)x

(t) + (
F,y

G,y
)y

(t)] dt

1
2
_
b
a
[(
F,x

G,x
,
F,y

G,y
)

(t)[ dt.
Let : C
1
(D, R
2
) R
2
be dened by (F) = (
F,x
,
F,y
). We have by the
Cauchy-Schwarz inequality that
1
2
_
b
a
[((F) (G))

(t)[ dt
1
2
_
b
a
|(F) (G)||

(t)| dt

(D)
2
max
(x,y)D
|(F) (G)|,
where (D) is the arclength of D. We claim that for every > 0, there exists
> 0 such that
|F G|
C
1
(D)
< max
(x,y)D
|(F) (G)| <
2
(D)
. (12)
8
Note that = (p, q, r, s) dened in (11) is continuous for p
2
+ q
2
,= 0. Con-
sequently, is uniformly continuous on any compact K R
4
p
2
+ q
2
= 0.
Thus, for every
1
> 0, there exists
1
> 0 such that
|(p
1
, q
1
, r
1
, s
1
) (p
2
, q
2
, r
2
, s
2
)| <
1
[(p
1
, q
1
, r
1
, s
1
) (p
2
, q
2
, r
2
, s
2
)[ <
1
, (13)
for (p
1
, q
1
, r
1
, s
1
), (p
2
, q
2
, r
2
, s
2
) K. In particular, let
1
=

2
(D)
, and let
K =
_
(p, q, r, s) :
o
|(p, q)|
1
+|F|
C
1
(D)
, |(r, s)|
1
+|F|
C
1
(D)
_
.
Note that K is compact, and that if |F G|
C
1
(D)
<
1
, then
(F
1
, F
2
, (F
1
)

, (F
2
)

), (G
1
, G
2
, (G
1
)

, (G
2
)

) K
for = x, y and all (x, y) D, since |G|
1
+sup
(x,y)D
|F|, and |DG|

1
+ sup
(x,y)D
|DF|. Set

= (F
1
, F
2
, (F
1
)

, (F
2
)

) (G
1
, G
2
, (G
1
)

, (G
2
)

).
We have that
|
x
|
2
+|
y
|
2
|F G|
2
C
1
(D)
<
2
1
,
so |

| <
1
. We have by (13) that [
F,

G,
[ <
1
for all (x, y) D; thus,
|(F) (G)|
2
= [
F,x

G,x
[
2
+[
F,y

G,y
[
2
< 2
2
1
.
Hence,
|F G|
C
1
(D)
<
1
|(F) (G)| <

2
1
=
2
(D)
,
so the claim in (12) is proven.
An immediate consequence of Lemma 4.1 is the homotopy invariance of the
Kronecker index. We are particularly concerned with C
1
homotopies.
Denition 4.2. Let G C
1
(U, R
2
), and suppose G(x, y) ,= (0, 0) for all (x, y)
D. A C
1
homotopy between F and G is a function H : D[0, 1] R
2
where
H
t
denotes the function (x, y) H((x, y), t) such that
(i) H
0
= F, H
1
= G, and H
t
C
1
(D) for t [0, 1]
(ii) lim
st
|H
s
H
t
|
C
1
(D)
= 0.
Homotopy invariance allows us to deform F while maintaining the value of
the Kronecker index.
Theorem 4.3 (Homotopy invariance of the Kronecker index). Let F and G
be dened as above, and let H((x, y), t) be a C
1
homotopy between them. If
H((x, y), t) ,= (0, 0) for all ((x, y), t) D [0, 1], then
index
K
(F, D) = index
K
(G, D).
9
Proof. Let h : [0, 1] Z be dened by h(t) = index
K
(H
t
, D). We have by
Lemmata 3.2 and 4.1 that h maps [0, 1] to a connected component in Z; thus,
we have that h(t) = n for some integer n and all t [0, 1]. In particular,
index
K
(H
0
, D) = index
K
(H
1
, D), from which the statement follows.
For F whose Kronecker index is dicult to calculate, it is often simpler to
prove that F is homotopic to a vector eld with a more easily calculable index.
Homotopy invariance allows us this convenience.
In order to extend homotopy invariance and other important properties of
the Kronecker index to the Brouwer degree for continuous F, we require a C
2
approximation. The following result, adapted from [2, p. 4], allows us to choose
such a function in any neighborhood of F.
Proposition 4.4. Suppose that F C(U, R
2
), and let D D U be open
such that D is compact. For any > 0, there exists G C

(U, R
2
) such that
|F G|
C(D)
< .
With this, we have nally established the results necessary to dene the
Brouwer degree explicitly for continuous F and D whose boundary is C
1
.
Denition 4.5 (Brouwer degree). Let F C(U, R
2
) be such that F(x, y) ,=
(0, 0) for all (x, y) D, and let G C
2
(U, R
2
) be such that
|GF|
C(D)
< min
(x,y)D
|F(x, y)|. (14)
The Brouwer degree of F in D over (0, 0), denoted by deg
B
(F, D, (0, 0)), is given
by
deg
B
(F, D, (0, 0)) = index
K
(G, D).
In order to justify the above denition, it suces to show that any two C
2
approximations of F yield the same Kronecker index. We adapt the argument
from [6, p. 18].
Let F C(U, R
2
), and set
= min
(x,y)D
|F|.
Let G
i
C
1
(U, R
2
) for i = 1, 2 be such that |F G
i
|
C(D)
< . Note that
G
i
,= 0 for all (x, y) D, and consider the C
1
homotopy H dened by
H
t
(x, y) = tG
1
(x, y) + (1 t)G
2
(x, y) for all (x, y) D, [0, 1],
where H
t
(x, y) = H((x, y), t). We have that
|H
t
F|
C(D)
= |t(G
1
F) + (1 t)(G
2
F)|
C(D)
< t + (1 t) = ,
so if (x, y) D, then
|H
t
(x, y)| |F(x, y)| |F(x, y) H
t
(x, y)| > 0 for all t [0, 1].
10
Thus, H((x, y), t) ,= 0 for all ((x, y), t) D [0, 1]. We have by homotopy
invariance that
index
K
(G
1
, D) = index
K
(G
2
, D).
Consequently, index
K
(G, D) is constant for all G C
1
(U, R
2
) with G ,= 0
on D such that |F G|
C(D)
< . Our denition of the Brouwer degree is
therefore constant for any C
1
approximation G of F which satises the estimate
in (14).
5 Properties of the Brouwer degree
With a denition of the Brouwer degree now established, we wish to extend sev-
eral key properties of the Kronecker index to the Brouwer degree for continuous
F, beginning with the solution property. In particular, we write deg
B
(F, D) =
deg
B
(F, D, (0, 0)).
Theorem 5.1 (Solution property of the Brouwer degree). If deg
B
(F, D) ,= 0,
then there exists (p, q) D such that F(p, q) = (0, 0).
Proof. We prove the contrapositive. Suppose that F(x, y) ,= (0, 0) for all (x, y)
D, and approximate F by G C
2
(U, R
2
) such that
|GF|
C(D)
< min
(x,y)D
|F(x, y)|.
Observe that G(x, y) ,= (0, 0) for all (x, y) D. We have by Theorem 3.3 that
index
K
(G, D) = 0.
Hence, it follows from Denition 4.5 that deg
B
(F, D) = 0.
The excision property follows naturally.
Theorem 5.2 (Excision property of the Brouwer degree). Let D
1
D be
open such that D is a C
1
Jordan curve. Suppose that F(x, y) ,= 0 for all
(x, y) D D
1
. Then,
deg
B
(F, D) = deg
B
(F, D
1
).
Proof. Approximate F by G C
2
(U, R
2
) such that
|GF|
C(D\D1)
< min
(x,y)D\D1
|F(x, y)|.
We have that
G(x, y) ,= (0, 0) for all (x, y) D D
1
.
Hence, by Theorem 3.4 and Denition 4.5,
deg
B
(F, D) = deg
B
(F, D
1
).
11
As was the case for the Kronecker index, we must rst establish the conti-
nuity of the Brouwer index before discussing homotopy invariance.
Lemma 5.3. Let G C(U, R
2
), and suppose that F, G ,= 0 for all (x, y) D.
Then, for every > 0, there exists > 0 such that
|F G|
C(D)
< [ deg
B
(F, D) deg
B
(G, D)[ < .
Proof. Set > 0, and let
= min
(x,y)D
|F(x, y)|.
Suppose that |G F|
C(D)
< /4. Let

G be a C
1
approximation of G such
that |

G G|
C(D)
< /4. It follows that

G ,= (0, 0) for all (x, y) D.
Furthermore, since
|

GF|
C(D)
|

GG|
C(D)
+|GF|
C(D)
< /2,
we have that
deg
B
(F, D) = index
K
(GD) = deg
B
(G, D).
Hence,
|F G|
C(D)
< /4 [ deg
B
(F, D) deg
B
(G, D)[ = 0 < .
Note that, as a consequence of Lemma 3.2 and Denition 4.5, the Brouwer
degree is integer-valued. This fact allows us to prove the homotopy invariance
of the Brouwer degree.
Theorem 5.4 (Homotopy invariance of the Brouwer degree). Let G C(U, R
2
),
and let H : D [0, 1] R
2
be a continuous homotopy between F and G. If
H((x, y), t) ,= (0, 0) for all ((x, y), t) D [0, 1], then
deg
B
(F, D) = deg
B
(G, D).
Proof. Let H : U[0, 1] R
2
be a continuous map. Suppose that H((x, y), t) ,=
(0, 0) for all ((x, y), t) D [0, 1], and dene the map f : [0, 1] Z by
f(t) = deg
B
(H
t
, D, (0, 0) for all t [0, 1],
where H
t
(x, y) = H((x, y), t) for all ((x, y), t) D [0, 1]. We have by Lemma
5.3 that f maps [0, 1] to a connected component in Z. In particular, f(0) = f(1),
from which the statement follows.
6 Generalizing for continuous D
We wish to generalize our denition of the Brouwer degree to address D with
a continuous boundary. In order to do so, we must construct several curves
12
approximating D; however, we rst require a neighborhood of D on which
F(x, y) ,= (0, 0).
Let D D U be open such that D is a Jordan curve. Additionally, sup-
pose that F ,= (0, 0) for all (x, y) D. Then there exists an open neighborhood
V of the boundary of D such that F ,= (0, 0) for all (x, y) V .
To see why this is the case, note that if (x, y) D, then there exists an
open disk around (x, y) of radius (x, y) > 0 such that |F| > 0 on the closure
of the disk. The collection of disks around all (x, y) D forms an open cover
of D. Since D is compact, there exists a nite subcover of disks whose union
we take as V .
With our neighborhood dened, we require a piecewise smooth approxima-
tion of D which lies in V . The following result, adapted from [1, pp. 588589],
allows guarantees the existence of a polygon in any neighborhood of D.
Lemma 6.1. Suppose that J R
2
is a Jordan curve with parameterization :
[a, b] J. For any > 0, there exists a polygon P R
2
with parameterization
: [a, b] P such that |(t) (t)| < for all t [a, b].
Let (t) : [a, b] D be a parameterization. By Lemma 6.1, there exists a
polygon P with parameterization : [a, b] P such that
|(t) (t)| < dist(D, V ) for all t [a, b].
Note that F ,= (0, 0) for all (x, y) in some neighborhood of P, since P V . By
replacing all p
i
with arcs, it is possible to construct a dierentiable curve in this
neighborhood, thereby smoothing the polygon. Denote the n vertices of P
by p
i
for i = 1, . . . , n, and denote by
i
the smaller of the interior and exterior
angles of each p
i
.
The smoothing procedure is as follows. For each p
i
, we draw a circle in
i
with two points of tangency to P, as illustrated in Figure 2. A curve is then
constructed by removing the vertex as well as the line segments connecting it
to either point of tangency and replacing these with the minor arc of the circle
between the same two points. After this has been done for all n vertices, i.e. P
has been smoothed at all p
i
, the resulting curve

P is C
1
.
Observe that the quantity r
i
tan
i
corresponds to the length of the line
segments removed when smoothing P at p
i
, also as illustrated in Figure 2. We
must choose the radius r
i
of each circle such that
r
i
tan
i
< min
_
|p
i
p
i1
|
2
,
|p
i+1
p
i
|
2
, dist(P, V )
_
,
adjusting the indices appropriately for the cases i = 1 and i = n. The restric-
tions r
i
tan
i
<
1
2
|p
i
p
i1
| and r
i
tan
i
<
1
2
|p
i
p
i+1
| guarantee that the
sum of the lengths of line segments removed from each edge of P does not exceed
the length of the edge itself. The restriction r
i
tan
i
< dist(P, V ) guarantees
that

P V , and furthermore that F ,= (0, 0) for all (x, y)

P.
The curve

P is a C
1
approximation of D with respect to F. By excision,
the Brouwer degree of F over (0, 0) is equal in the interior of any two smooth
13
Figure 2:

P, in blue, replaces red segments with an arc, smoothing P at p
i
.
approximations of a Jordan curve D. With this result, we may state the
following denition.
Denition 6.2. Let F C(U, R
2
), and let D D U be open such that
D is a positively oriented Jordan curve. Suppose that F(x, y) ,= (0, 0) for
all (x, y) D, and let

P be a positively oriented C
1
Jordan curve such that

P V for some neighborhood V of D on which F ,= (0, 0). The Brouwer


degree of F in D over (0, 0) is given by
deg
B
(F, D) = deg
B
(F, int(

P)).
7 Exercises
1. Suppose that F C
2
(U, R
2
), and that F(D) S
1
. Prove that
index
K
(F, D)
1

__
D
det(DF) dxdy.
14
References
[1] C. Jordan, Cours danalyse de l

Ecole polytechnique, vol. 3, Gauthier-Villars,


Paris, 1887.
[2] D. ORegan et al., Topological Degree Theory and Applications, Chapman &
Hall, Boca Raton, 2006.
[3] G. Dinca and J. Mawhin, Brouwer Degree and Applications, Birkhauser,
Basel, 2012 (preprint).
[4] L. Ahlfors, Complex analysis: an introduction to the theory of analytic func-
tions of one complex variable, McGraw- Hill, New York, 1953.
[5] M. Spivak, Calculus on Manifolds: a modern approach to classical theorems
of advanced calculus, Addison- Wesley, Reading, 1965.
[6] N. G. Lloyd, Degree theory, Cambridge University Press, Cambridge, 1978.
15
Appendix
A Solutions to exercises
1. Note that
det(DF) = P
x
Q
y
P
y
Q
x
. (15)
We can write
index
K
(F, D)
1
2
_
D
P dQQdP

1
2
_
D
[(PQ
x
QP
x
) dx + (PQ
y
QP
y
) dy],
since |F|
2
= P
2
+ Q
2
= 1 on D. An application of Greens theorem
yields
index
K
(F, D)
1
2
__
D
(P
x
Q
y
Q
x
P
y
P
y
Q
x
+ Q
y
P
x
) dxdy

1
2
__
D
2(P
x
Q
y
P
y
Q
x
) dxdy.
The statement then follows from (15).
16

You might also like