Curvature and torsion

Based on lecture notes by James McKernan

Blackboard 1. Let ~r : I Rn be a C 2 regular curve (i.e., ~r0 (t) 6= ~0 for all t).
The curvature (s) of ~r(s) is the magnitude of the vector
dT~ (s)
T~ 0 (s) =
,
ds
~ is the unit vector pointing in the direction of T~ 0 (s)
and the unit normal vector N
~0
~ (s) = T (s) .
N
kT~ 0 (s)k
One can try to calculate the curvature using the parameter t. By the chain rule,
dT~ (t)
dT~ (s) ds
=
.
dt
ds dt
So
dT~ (s)
=
ds

~ (t)
dT
dt
ds
dt

The denominator is the speed. It follows that ~n and dT~ /dt point in the same
direction.
Note that the normal vector and the unit tangent vector are always orthogonal.
Indeed, more generally
Proposition 2. Let ~v : I Rn . Then
d(~v ~v )
= 2~v 0 ~v ,
dt
and in particular if |~v (t)| is constant then ~v 0 and ~v are orthogonal.
Now,
kT~ (s)k = 1.
~ (s) points in the same direction as T~ 0 (s), it follows that the tangent
and so, as N
vector and the normal vector are orthogonal.
Blackboard 3.
~
~ (s).
B(s)
= T~ (s) N
is called the binormal vector.
~ (s), and B(s)
~
The three vectors T~ (s), N
are unit vectors and pairwise orthogonal,
~ (s), and
that is, these vectors are an orthonormal basis of R3 . Notice that T~ (s), N
~
B(s) are a right handed set.
We call these vectors a moving frame or the Frenet-Serret frame. Now
~
dB
~
(s) B(s)
= 0,
ds
as
~
~
B(s)
B(s)
= 1.
It follows that
~
dB
(s),
ds
1

~ (s).
lies in the plane spanned by T~ (s) and N
~
~)
dB
d(T~ N
(s) T~ (s) =
(s) T~ (s)
ds
ds
!
~
dN
dT~
~
~
(s) N (s) + T (s)
(s) T~ (s)
=
ds
ds
~
~ (s) N
~ (s)) T~ (s) + (T~ (s) dN (s)) T~ (s)
= (s)(N
ds
~
d
N
= 0 + (T~ (s) T~ (s))
(s)
ds
= 0.
It follows that
~
dB
(s)
ds

and

T~ (s),

are orthogonal, and so

~
dB
(s)
ds

is parallel to

~ (s).
N

Blackboard 4. The torsion of the curve ~r(s) is the unique scalar (s) such that
~
dB
~ (s).
(s) = (s)N
ds
If we have a helix, the sign of the torsion distinguishes between a right handed
helix and a left handed helix. The magnitude of the torsion measures how spread
out the helix is (the curvature measures how tight the turns are). Now
~
dN
(s)
ds
~ (s), and so it is a linear combination of T~ (s) and B(s).
~
is orthogonal to N
In fact,
~
~ T~ )
dN
d(B
(s) =
(s)
ds
ds
~
dB
dT~
~
=
(s) T~ (s) + B(s)

(s)
ds
ds
~ (s) T~ (s) + (s)B(s)
~
~ (s)
= (s)N
N
~
= (s)B(s)
(s)T~ (s)
~
= (s)T~ (s) + (s)B(s).
Blackboard 5. We say that ~r(t) is smooth if ~r(t) is C .
Theorem 6 (Frenet Formulae). Let ~r : I R3 be a regular smooth parametrised
curve. Then
~0
~
T (s)
T (s)
0
(s)
0
N

~ 0 (s) = (s)
~ (s) .
0
(s)
N
0
~
~
0
(s)
0
B (s)
B(s)

Of course, s represents the arclength parameter and primes denote derivatives

with respect to s. Notice that the 3 3 matrix A appearing in (6) is skewsymmetric, that is At = A. The way we have written the Frenet formulae, it
appears that we have two 3 1 vectors; strictly speaking these are the rows of two
3 3 matrices.
Theorem 7. Let I R be an open interval and suppose we are given two smooth
functions
: I R
and
: I R,
where (s) > 0 for all s I.
Then there is a regular smooth curve ~r : I R3 parametrised by arclength with
curvature (s) and torsion (s). Further, any two such curves are congruent, that
is, they are the same up to translation and rotation.
Lets consider the example of the helix:
Example 8.
s bs
s
~r(s) = (a cos , a sin , ),
c
c c
where
c2 = a2 + b2 .
Lets assume that a > 0. By convention c > 0. Then
1
s
s
T~ (s) = (a sin , a cos , b).
c
c
c
Hence
dT
a
s
a
s
a ~
s
s
(s) = 2 (cos , sin , 0) = 2 ( cos , sin , 0) = 2 N
(s)
ds
c
c
c
c
c
c
c
It follows that
a
~ (s) = ( cos s , sin s , 0).
(s) = 2
and
N
c
c
c
Finally,




k 

~

B(s)
=  ac sin sc ac cos sc bs
c 
 cos s sin s 0 
c
c
It follows that
b
s b
s a
1
s
s
~
B(s)
= ( sin , cos , ) = (b sin , b cos , a).
c
c c
c c
c
c
c
Finally, note that
~
dB
b
s
s
b ~
(s) = 2 (cos , sin , 0) = 2 N
.
ds
c
c
c
c
Using this we can compute the torsion:
b
(s) = 2 .
c
It is interesting to use the torsion and curvature to characterise various geometric
properties of curves. Lets say that a parametrised differentiable curve ~r : I R3
is planar if there is a plane which contains the image of ~r.
Theorem 9. A regular smooth curve ~r : I R3 is planar if and only if the torsion
is zero.

Proof. We may assume that the curve passes through the origin.
Suppose that ~r is planar. Then the image of ~r is contained in a plane . As
the curve passes through the origin, contains the origin as well. Note that the
~ (s) are contained in . It
unit tangent vector T~ (s) and the unit normal vector N
~
~
follows that B(s)
is a normal vector to the plane; as B(s)
is a unit vector, it must
be constant. But then
~
dB
~ (s),
(s) = ~0 = 0N
ds
so that the torsion is zero.
Now suppose that the torsion is zero. Then
dB
~ = ~0,
(s) = 0N
ds
~
so that B(s)
= B0 , is a constant vector. Consider the function
~
~ 0.
f (s) = ~r(s) B(s)
= ~r(s) B
Then
~ 0)
df
d(~r B
(s) =
(s)
ds
ds
~ 0 = 0.
= T~ (s) B
So f (s) is constant. It is zero when ~r(a) = ~0 (the curve passes through the origin)
so that f (s) = 0. But then ~r(s) is always orthogonal to a fixed vector, so that ~r is
contained in a plane, that is, C is planar.

It is interesting to try to figure out how to characterise curves which are contained
in spheres or cylinders.