0% found this document useful (0 votes)
258 views

Differential Geometry Notes

1. The document discusses the osculating circle of a planar curve at a point of nonvanishing curvature. It defines the radius of curvature as the reciprocal of the limiting radius of circles passing through nearby points on the curve. 2. Several exercises are provided to prove properties of curvature and osculating circles. Kneser's Nesting Theorem states that for curves with monotone nonvanishing curvature, osculating circles are pairwise disjoint. 3. Total curvature is discussed. The total curvature of any closed planar curve is at least 2π, with equality if and only if the curve is convex.

Uploaded by

Sanjeev Shukla
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
258 views

Differential Geometry Notes

1. The document discusses the osculating circle of a planar curve at a point of nonvanishing curvature. It defines the radius of curvature as the reciprocal of the limiting radius of circles passing through nearby points on the curve. 2. Several exercises are provided to prove properties of curvature and osculating circles. Kneser's Nesting Theorem states that for curves with monotone nonvanishing curvature, osculating circles are pairwise disjoint. 3. Total curvature is discussed. The total curvature of any closed planar curve is at least 2π, with equality if and only if the curve is convex.

Uploaded by

Sanjeev Shukla
Copyright
© Attribution Non-Commercial (BY-NC)
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Math 4441 Sep 20, 20071

Differential Geometry
Fall 2007, Georgia Tech

Lecture Notes 5

1.13 Osculating Circle and Radius of Curvature


Recall that in a previous section we defined the osculating circle of a planar
curve α : I → R2 at a point a of nonvanishing curvature t ∈ I as the circle
with radius r(t) and center at

α(t) + r(t)N (t)

where
1
r(t) :=
κ(t)
is called the radius of curvature of α. If we had a way to define the osculating
circle independently of curvature, then we could define curvature simply as
the reciprocal of the radius of the osculating circle, and thus obtain a more
geometric definition for curvature.

Exercise 1. Let r(s, t) be the radius of the circle which is tangent to α at


α(t) and is also passing through α(s). Show that

κ(t) = lim r(s, t).


s→t

To do the above exercise first recall that, as we showed in the previous


lecture, curvature is invaraint under rigid motions. Thus, after a rigid motion,
we may assume that α(t) = (0, 0) and α0 (t) is parallel to the x-axis. Then,
we may assume that α(t) = (t, f (t)), for some function f : R → R with
f (0) = 0 and f 0 (0) = 0. Further, recall that

|f 00 (t)|
κ(t) = p .
( 1 + f 0 (t)2 )3
1
Last revised: October 2, 2007

1
Thus
κ(0) = |f 00 (0)|.
Next note that the center of the circle which is tangent to α at (0, 0) must
lie on the y-axis at some point (0, r), and for this circle to also pass through
the point (s, f (s)) we must have:

r2 = s2 + (r − f (s))2 .

Solving the above equation for r and taking the limit as s → 0, via the
L’Hopital’s rule, we have

2|f (s)|
lim = |f 00 (0)| = κ(0),
s→0 f 2 (s) + s2

which is the desired result.

Note 2. The above limit can be used to define a notion of curvature for
curves that are not twice differentiable. In this case, we may define the
upper curvature and lower curvature respectively as the upper and lower
limit of
2|f (s)|
.
f 2 (s) + s2
as s → 0. We may even distinguish between right handed and left handed
upper or lower curvature, by taking the right handed or left handed limits
respectively.

Exercise* 3. Let α : I → R2 be a planar curve and t0 , t1 , t2 ∈ I with


t1 ≤ t0 ≤ t2 . Show that κ(t0 ) is the reciprocal of the limit of the radius of
the circles which pass through α(t0 ), α(t1 ) and α(t2 ) as t1 , t2 → t0 .

1.14 Kneser’s Nesting Theorem


We say that the curvature of a curve is monotone if it is strictly increasing or
decresing. The following result shows that the osculating circels of a curve
with monotone curvature are “nested”, i.e., the lie inside each other:

Theorem 4 (Kneser’s Nesting theorem). Let α : I → R2 be a C 4 curve


with monotone nonvanishing curvature. Then the osculating circles of α are
pairwise disjoint.

2
To prove the above result we need the following Lemma. Note that if
α : I → R2 is a curve with nonvanishing curvature, then the centers of the
osculating circles of α for the curve

β(t) := α(t) + r(t)N (t),

where r(t) := 1/κ(t) is the radius of curvature of α. This curve β is known


as the evolute of α.

Exercise 5. Show that if α : I → R2 is a C 4 curve with monotone non-


vanishing curvature, then its evolute β is a regular curve which also has
nonvanishing curvature. In particular β contains no line segments.

Now we are ready to prove the main result of this section:


Proof of Neser’s Theorem. We may suppose that kα0 k = 1, and its curvature
κ is increasing. We need to show that for every t0 , t1 ∈ I, with t0 < t1 , the
osculating circle at t1 lies inside the osculating circle at t0 . To this end it
suffices to showing that

kβ(t0 ) − β(t1 )k + r(t1 ) < r(t0 ).

To see this end first note that, since β contains no line segments (see the
previous exercise)
Z t1
kβ(t0 ) − β(t1 )k < kβ 0 (t)kdt.
t0

Now a simple computation completes the proof:


Z t1 Z t1
0
kβ (t)k = |r0 (t)| dt
t0 t
Z 0t1
= −r0 (t) dt = r(t0 ) − r(t1 ).
t0

(Here |r0 | = −r0 , becuase, since κ is increasing by assumption, r is decreas-


ing.)
Kneser’s theorem has a number of interesting corollaries:

3
Exercise 6. Show that a curve with monotone curvature cannot have any
self intersections.
Exercise 7. Show that a curve with monotone curvature cannot have any
bitangent lines.
The last two exercises show that a curve with monotone curvature looks
essentially as depicted in the following figure, i.e., it spirals around itself.

1.15 Total Curvature and Convexity


The boundary of X ⊂ Rn is defined as the intersection of the closure of X
with the closure of its complement.
Exercise 8. Is it true that the boundary of any set is equal to its closure
minus its interior? (Hint: Consider a ball with its center removed)
We say that a simple closed curve α : I → R2 is convex provided that
its image lies on one side of every tangent line. A subset of Rn is convex
if it contains the line segment joining each pairs of its points. Clearly the
intersection of convex sets is convex.
Exercise 9. Show that a simple closed planar curve α : I → R2 is convex
only if it lies on the boundary of a convex set. (Hint: By defintion, through
each point p of Γ there passes a line `p with respect to which Γ lies on one
side. Thus each `p defines a closed half plane Hp which contains Γ. Show
that Γ lies on the boundary of the intersection of all these half planes).
The total curvature of a curve α : I → Rn is defined as
Z
κ(t)dt,
I

where t is the arclength parameter.

4
Exercise 10. Show that the total curvature of any convex planar curve is
2π. (Hint: We only need to check that the exterior angles of polygonal
approximations of a convex curve do not change sign. Recall that, as we
showed in a previous section, the sum of these angles is the total signed
curvature. So it follows that the signed curvature of any segment of α is
either zero or has the same sign as any other segment. This in turn implies
that the signed curvature of α does not change sign. So the total signed
curvature of α is equal to its total curvature up to a sign. Since by definition
the curve is simple, however, the total signed curvature is ±2π by Hopf’s
theorem.)
Theorem 11. For any closed planar curve α : I → R2 ,
Z
κ(t)dt ≥ 2π,
I

with equality if and only if α is convex.


First we show that the total curvature of any curve is at least 2π. To this
end recall that when t is the arclength parameter κ(t) = kT 0 (t)k. Thus the
total curvature is simply the total length of the tantrix curve T : I → S2 .
Since T is a closed curve, to show that its total length is bigger than 2π, it
suffices to check that the image of T does not lie in any semicircle.
Exercise 12. Verify the last sentence.
To see the that the image of T does not lie in any semicircle, let u ∈ S1
be a unit vector and note that
Z b Z b
hT (t), uidt = hα0 (t), uidt = hα(b) − α(a), ui = 0.
a a

Since T (t) is not constant (why?), it follows that the function t 7→ hT (t), ui
must change sign. So the image of T must lie on both sides of the line through
the origin and orthogonal to u. Since u was chosen arbitrarily, it follows that
the image of T does not lie in any semicircle, as desired.
Next we show that the total curvature is 2π if and only if α is convex.
The “if” part has been established already in exercise 10. To prove the “only
if” part, suppose that α is not convex, then there exists a tangent line `0 of
α, say at α(t0 ), with respect to which the image of α lies on both sides. Then
α must have two more tangent lines parallel to `0 .

5
Exercise 13. Verify the last sentence (Hint: Let u be a unit vector or-
thogonal to ` and note that the function t 7→ hα(t) − α(t0 ), ui must have a
minimum and a maximum differerent from 0. Thus the derivative at these
two points vanishes.)

Now that we have established that α has three distinct parallel lines, it
follows that it must have at least two parallel tangents. This observation is
worth recording:

Lemma 14. If α : I → R2 is a closed curve which is not convex, then it has


a pair of parallel tangent vectors wich generate disitinct parallel lines.

Next note that

Exercise 15. If α : I → R2 is closed curve whose tantrix T : I → S1 is not


onto, then the total curvature is bigger than 2π. (Hint: This is immediate
consequence of the fact that T is a closed curve and it does not lie in any
semicircle)

So if T is not onto then we are done (recall that we are trying to show
that if α is not convex, then its total curvature is bigger than 2π). We
may assume, therefore, that T is onto. This together with the above lemma
yields that the total curvature is bigger than 2π. To see this note that let
t1 , t2 ∈ I be the two points such that T (t1 ) and T (t2 ) are parallel and the
corresponding tangent lines are distict. Then T restricted to [t1 , t2 ] is a closed
nonconstant. So either T ([t1 , t2 ]) (i) covers some open segment of the circle
twice or (ii) covers the entire circle. Since we have established that T is onto,
the first possibility implies that the legth of T is bigger than 2π. Further,
since, T restricted to I − (t1 , t2 ) is not constant, the second possibility (ii)
would imply the again the first case (i). Hence we conclude that if α is not
convex, then its total curvature is bigger than 2π, which completes the proof
of Theorem 11.

Corollary 16. Any simple closed curve α : I → R2 is convex if and only if


its signed curvature does not change sign.

Proof. Since α is simple, its total signed curature is ±2π by Hopf’s theorem.
After switching the orientation of α, if necessary, we may assume that the
total signed curvature is 2π. Suppose, towards a contradiction, that the
signed curvature does change sign. The integral of the signed curvature over

6
the regions where its is positive must be bigger than 2π, which in turn implies
that the total curvature is bigger than 2π, which contradicts the previous
theorem. So if α is convex, then κ does not change sign.
Next suppose that κ does not change sign. Then the total signed curva-
ture is equal to the total curvature (up to a sign), which, since the curve is
simple, implies, via the Hopf’s theorem, that the total curvature is 2π. So
by the previous theorem the curve is convex.

You might also like