The FrenetSerret formulas

Attila Mate
Brooklyn College of the City University of New York
January 19, 2017

Contents

Contents 1

1 The FrenetSerret frame of a space curve 1

2 The FrenetSerret formulas 3

3 The first three derivatives of r 3

4 Examples and discussion 4

4.1 The curvature of a circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4.2 The curvature and the torsion of a helix . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1 The FrenetSerret frame of a space curve

We will consider smooth curves given by a parametric equation in a three-dimensional space. That
is, writing bold-face letters of vectors in three dimension, a curve is described as r = F(t), where F
is continuous in some interval I; here the prime indicates derivative. The length of such a curve
between parameter values t0 I and t1 I can be described as
Z t1 Z t1  

 dr 
(1) (t1 ) = |F (t)| dt =   dt
 dt 
t0 t0

where, for a vector u we denote by |u| its length; here we assume t0 is fixed and t1 is variable, so
we only indicated the dependence of the arc length on t1 . Clearly, is an increasing continuous
function, so it has an inverse 1 ; it is customary to write s = (t). The equation

def
(2) r = F( 1 (s)) s J = {(t) : t I}

is called the re-parametrization of the curve r = F(t) (t I) with respect to arc length. It is clear
that J is an interval. To simplify the description, we will always assume that r = F(t) and s = (t),
Written for the course Mathematics 2201 (Multivariable Calculus) at Brooklyn College of CUNY.

1
so we will just use the variables r, t, and s instead of using function notation. We will use prime
to indicate the differentiation d/dt, while the differentiation d/ds will not be abbreviated. We will
assume that r 6= 0 for any t I; then r is a tangent vector to the curve corresponding to the given
parameter value t.1.1
By the fundamental Theorem of Calculus, equation (1) implies
ds d(t)
(3) s =
= = |F (t)| = |r |.
dt dt
Hence, by the chain rule of differentiation we have
 
d dt d 1 d 1 d
(4) = = = ,
ds ds dt ds/dt dt |r | dt
where the second equation follows from equation (1) by the Fundamental Theorem of calculus. When
using differential operator notation as in d/ds, everything after the differential operator up to the
next + or sign needs to be differentiated; the expression preceding the differential operator is not
to be differentiated. The unit tangent vector T is defined as1.2
def 1 dr
(5) T = r = ;
|r | ds
the second equation follows from equations (4). The curvature is defined as
dT def
(6) =
.
ds
We will assume that 6= 0.1.3 The unit normal vector is defined as
1 dT def
(7) .
N =
ds
Note that |T|2 = T T, so by the product rule of differentiation,
 
dT 1 dT dT
(8) T= T+T = 0;
ds 2 ds ds
hence T is perpendicular to dT/ds, and so N is perpendicular to T. The unit binormal vector is
defined as
def
(9) B = T N.
The vectors T, N, B form the basic unit vectors of a coordinate system especially useful for describing
the the local properties of the curve at the given point. These three vectors form what is called the
FrenetSerret frame. Equation (9) implies that the vectors T, N, B form a right-handed system
of pairwise perpendicular unit vectors. Any cyclic permutation of these vectors also form a right-
handed system of pairwise perpendicular unit vectors; therefore we have
T N = B,
(10) N B = T,
B T = N.
1.1 If the equation r = F(t) describes a moving point, where t is time, then r is the velocity vector of the moving

point at time t. That is, the length of r is its speed, while the direction of r is its direction of its movement. If r = 0,
then the point stopped moving at the given time, and then it may resume its movement in a different direction. This
means that even though the function F is differentiable, the tangent line to the curve described by the function may
not be defined at this point.
1.2 A unit vector is a vector of length 1.
1.3 We will comment later on what happens when 6= 0.

2
2 The FrenetSerret formulas
As |N| = 1, we have |N|2 = N N = 1, and so, similarly to equation (8), we have

dN
N = 0.
ds
That is dN/ds is perpendicular to N, so we have

dN
(11) = T + B
ds
for some numbers and (depending on t).2.1 Here is called the torsion of the curve at the point;
the value of will be determined below. Using this equation, equations (10), and the product rule
of differentiation for vector products, we have

dB d(T N) dT dN
= = N+T
(12) ds ds ds ds
= N N + T (T + B) = T B = N;

the third equation follows by equations (6) and (11). The first term third member2.2 is zero; so is
first term after distributing the cross product in the second term in the third member. Thus the
fourth equation follows; to obtain the last equation we used equation (10). Hence, using equation
(10), we have

dN d(B T) dB dT
= = T+B
(13) ds ds ds ds
= N T + B (N) = B + B N = T + B;

This equation shows that in equation (11) equals ; however, this fact and equation (11) itself is
no longer of any interest, since this equation is subsumed in the last equation. Equations (7), (12),
and (13) are called the FrenetSerret formulas. To summarize these formulas, we have

dT
= N,
ds
dN
(14) = T + B,
ds
dB
= N.
ds

3 The first three derivatives of r

As we mentioned above, we will indicate derivation with respect to t by prime. According to
equations (5) and (3), we have

(15) r = |r |T = s T.
2.1 Every vector can be expressed as a linear combination of the basic unit vectors T, N, B; as dN/ds is perpendicular

to N, the coefficient of N in this linear combination is 0.

2.2 The member between the third and fourth equations

3
We can do further differentiations with respect to t by using equations (14), (4), and (3). We have

dT
(16) r = s T + s T = s T + (s )2 = s T + (s )2 N.
ds
Further, repeatedly using equations (15) and (14), we have

r = s T + s T + 2s s + (s )2 N + (s )2 N

dT dN
+ 2s s + (s )2 N + (s )3


= s T + s s
(17) ds ds
= s T + s s N + 2s s + (s )2 N + (s )3 (T + B)

= s (s )3 T + 3s s + (s )2 N + (s )3 B.

It is now easy to express and in terms of derivatives with respect to t. Equations (15), (16),
and (10) give

(18) r r = (s )3 B,

so using equation (3), we obtain that

|r r |
(19) =
|r |3

by noting that |B| = 1. By equations (18) and (17) we have

(r r ) r = (s )6 2 .

Hence, using equation 18) and noting that |B| = 1, we obtain that

(r r ) r
(20) = .
|r r |2

4 Examples and discussion

Since the curvature and the torsion are defined in terms of the local coordinate frame T, N, B
in arc-length parametrization, they only depend on the shape of the curve and not on the choice of
the coordinate system x, y, z and the choice of the parameter t. Hence, for a curve that we want
to calculate the curvature or the torsion, we may set up the coordinate system x, y, z and choose a
parametrization that make these calculations especially simple.

4.1 The curvature of a circle

We consider the circle of radius R > 0 lying in the x, y plane, and centered at the origin. This circle
can be parametrized by the equation

r = R(i cos t + j sin t),

where i, j, and k are the unit coordinate vectors in the directions of the positive x, y, and z axes,
respectively. We have
r = R(i sin t + j cos t)

4
and
r = R(i cos t + j sin t).
Hence  
 i j k
R cos t 0  = R2 (sin2 t + cos2 t) k = R2 k.


r r =  R sin t
R cos t R sin t 0 
We also have p
|r | = R cos2 t + sin2 t = R.
Hence equation (19) gives that
R2 1
== .
R3 R
Thus, the curvature of a circle is the reciprocal of the radius. For this reason, given any curve,
1/ is called the radius of curvature; it is the radius of the osculating circle: given an curve by the
equation r = F(t), the osculating circle at a point corresponding to the parameter value t = t0 (i.e.,
at the point with position vector r0 = F(t0 )) is a circle with equation r = G(t) at the for which
(21) G(t0 ) = F(t0 ), G (t0 ) = F (t0 ), and G (t0 ) = F (t0 ).
For the existence of such a circle, one needs to assume that F (t0 ) 6= 0 and F (t0 ) 6= 0. The case
that F (t0 ) = 0 is a case of bad parametrization, when the curve may or may not have a tangent line
and a curvature, but the equation is not suitable for determining the tangent line or the curvature.
In this case, one needs to re-parametrize the curve in such a way that the derivative at the point
with position vector F(t0 ) is not zero.4.1 If F (t0 ) 6= 0 but F (t0 ) = 0, the curvature is 0, and
the osculating circle degenerates into a straight line; in fact, the tangent line can be considered the
osculating circle. In this case one sometimes says that the radius of curvature is infinite.

4.2 The curvature and the torsion of a helix

A helix in the standard position can be described by the equation
r = i R cos t + j R sin t + ct k (R > 0).
We have
r = i R sin t + j R cos t + c k,
r = i R cos t j R sin t,
r = i R sin t j R cos t.
Therefore q p

|r | = R2 (sin2 t + cos2 t) + c2 = R 2 + c2 .
Furthermore,
 
 i j k
c  = i cR sin t j cR cos t + R2 (sin2 t + cos2 t) k

r r =  R sin t R cos t
R cos t R sin t 0
= i cR sin t j cR cos t + R2 k.
4.1 Such a parametrization may not exist. If there is a suitable re-parametrization, the re-parametrization with

respect to arc length will work. It is, however, possible, that the derivative of the arc-length parametrization at the
given point does not exist in which case there is no smooth re-parametrization at the point.

5
Thus q p
|r r | = c2 R2 (cos2 t + sin2 t) + R4 = R c2 + R 2
and p
(r r ) r = R c2 + R2 = cR2 sin2 t + cR2 cos2 t = cR2 .
Hence, using formula (19) we obtain

R c2 + R 2 R
= = 2 .
2 2
(R + c ) 3/2 R + c2

Similarly, using formula (20)

cR2 c
= = 2 .
R2 (c2 + R2 ) R + c2
An osculating curve to a given curve r = F(t) is a curve r = G(t) satisfying equations (21); these
equations can also be described by saying that the curves r = F(t) and r = G(t) have a second
order contact at the given point. One can generalize this to an arbitrary integer n by saying that the
curve r = F(t) and r = F(t) have an order n contact for a given parameter value t = t0 if F (t0 ) 6= 0
and4.2

(22) G(k) (t0 ) = F(k) (t0 ) for all k with 0 k n.

The osculating plane is the plane spanned by the vectors T and N. The osculating circle lies in
this place. The osculating plane has a second order contact with the curve at the point given by a
parameter value t = t0 . More generally, if F (t0 ) 6= 0 and F (t0 ) 6= 0, then any plane curve that
has a second order contact with the curve r = F(t) at the parameter value t = t0 lies entirely in the
osculating plane. If F (t0 ) = 0 then the osculating plane is not determined since = 0 in this case
so the vector N is not determined (cf. equations (19) and (7) to see this). The torsion expresses
the speed with which the osculating place turns as the arc-length parameter changes (indeed, this
follows from the third equation in (14), since B is normal to the osculating plane).
The derivation of the FrenetSerret formulas (14) shows the theoretical usefulness of arc-length
parametrization. Re-parametrizing a curve with respect to arc-length is rarely done in practice,
since the integrals involved cannot usually be evaluated, and a more useful procedure is to rewrite
the formulas derived with arc-length parametrization in terms of the original parameter, as was done
in formulas (19) and (20).

4.2 This condition just mentioned may be dropped if one is only interested in the curve as a moving points. If

F (t0 ) = 0 then this condition expresses only that the points near the time t = t0 have a higher order closeness,
but since both points stop moving at this time, this has no implication for the geometry of the two curves.