Chaptrer 2

Curvilinear Abscissa

I Definition and Properties of Curvilinear Abscissa

In this chapter we will consider the arc length as a parameter of a parametric curve Γ, 𝑀 .

Definition 1.1: Let Γ, 𝑀 be a parametric curve of class 𝐶1 . For 𝑎 𝑓𝑖𝑥𝑒𝑑 𝑡0 belonging to

the interval I (I is the domain of definition of the parametric representation M ), we call the
curvilinear abscissa of the parametric curve Γ, 𝑀 , the following function :

𝑡
𝑠=𝑆 𝑡 = 𝑀′(𝜏) 𝑑𝜏.
𝑡0

The fonction 𝜏 → 𝑀′(𝜏) is continuous on I, so S is primitive of this fonction.

If 𝑡 ≥ 𝑡0 , 𝑆 𝑡 is equal to the length of the arc of Γ taken between 𝑀(𝑡) and 𝑀(𝑡0 ).

If 𝑡 ≤ 𝑡0 , 𝑆 𝑡 is equal to minus the length of the arc of Γ taken between 𝑀(𝑡) and 𝑀(𝑡0 ).
From its definition, curvilinear abscissa S is a continuous function on I. Its image 𝐽 =
𝑆(𝐼) ⊂ ℝ is therefore an interval of ℝ.

𝑡0
The point 𝑀 𝑡0 ∈ 𝑀 𝐼 = Γ, corresponding to 𝑠 = 𝑆 𝑡0 = 𝑡0
𝑀′(𝜏) 𝑑𝜏 = 0, is the

origin of the curvilinear abscissa S.

Proposition 1.1: If the set of values of parameter 𝑡, for which 𝑀(𝑡) is a regular point of
Γ, 𝑀 , is dense in 𝐼 = 𝐷𝑜𝑚𝑀, i.e. any interval (∅ ≠) 𝛼, 𝛽 ⊂ 𝐼 contains a 𝑡 such that
𝑀 𝑡 is regular, then the function S (i.e. curvilinear abscissa) is then an homeomorphism
of 𝑡 ∈ 𝐼 to the interval 𝐽 = 𝑆(𝐼).

Therefore, with the assumption of the above proposition, we can define a

reparametrization 𝑀 = 𝑀 ∘ 𝑆 −1 of Γ, 𝑀 , called reparametrization of Γ, 𝑀 by a
curvilinear abscissa; in other words Γ, 𝑀 , 𝑀: 𝐽 ⊂ ℝ → ℝ𝑛 is a curve parametrized by a
curvilinear abscissa.

16
 In general a reparametrization of Γ, 𝑀 by a curvilinear abscissa is not of class 𝐶 1 ;
however, in the special case where all points 𝑀 𝑡 , 𝑡 ∈ 𝐼, are regular points , we have the
following theorem :

Theorem 1.1: Let Γ, 𝑀 a parametric curve of class 𝐶 𝑟 , 𝑟 ≥ 1, and all points 𝑀 𝑡 are
regular. A reparametrization 𝑀 = 𝑀 ∘ 𝑆 −1 by a curvilinear abscissa is then of class 𝐶 𝑟 and
moreover we have :

𝑑𝑀
(𝑠) = 1.
𝑑𝑠

Proof :
1) Let us show 𝑀 is of class 𝐶 𝑟 .

- If 𝑟 = 1, i.e. 𝑀 is of class 𝐶 1 , then the function 𝑡 → 𝑀′(𝑡) is continuoue and therefore

𝑡
curvilinear abscissa 𝑡 → 𝑆 𝑡 = 𝑡0
𝑀′(𝜏) 𝑑𝜏 is a differentiable function dérivable and we

have 𝑡 → 𝑆 ′ 𝑡 = 𝑀′(𝑡) which is continuous ; thus 𝑆 is of class 𝐶 1 . Moreover, since by

hypothesis all the points 𝑀 𝑡 are regular, then ∀𝑡 ∈ 𝐼, 𝑀′(𝑡) ≠ 0 and therefore 𝑆 is a

𝐶 1 -diffeomorphism (consequence of proposition 4.1 of the previous chapter ).

- If 𝑟 ≥ 2, i.e. 𝑀 is of classe 𝐶 𝑟 , so the application 𝑡 → 𝑀′(𝑡) is of class 𝐶 𝑟−1 and the

same is true of the fonction 𝑡 → 𝑀′(𝑡). 𝑀′(𝑡) as well as of the function 𝑡 → 𝑆 ′ 𝑡 =

1/2
𝑀′(𝑡) = 𝑀′(𝑡). 𝑀′(𝑡) (because 𝑆 ′ 𝑡 ≠ 0, ∀𝑡 ∈ 𝐼). This implies that 𝑆 is of class

𝐶 𝑟 and by the proposition 4.1. of the previous chapiter 𝑆 is a 𝐶 𝑟 -diffeomorphism.

𝑑𝑀
2) Let us show that (𝑠) = 1.
𝑑𝑠

𝑡
Since the function 𝑡 → 𝑀′(𝑡) is continuous, then the function 𝑡 → 𝑆 𝑡 = 𝑡0
𝑀′(𝜏) 𝑑𝜏

is derivable and we have : ∀𝑡 ∈ 𝐼, 𝑆 ′ 𝑡 = 𝑀′(𝑡) ≠ 0 (since by hypothesis all the points

𝑀 𝑡 are regular).
By differentiating the composite application 𝑀 = 𝑀 ∘ 𝑆 we obtain :

𝑑𝑀 𝑑𝑀 𝑑𝑆 𝑑𝑀
𝑡 = 𝑆 𝑡 . 𝑡 = 𝑆 𝑡 . 𝑀′(𝑡)
𝑑𝑡 𝑑𝑠 𝑑𝑡 𝑑𝑠
From which, by taking the norms of the two members :

17
 𝑑𝑀
𝑀′(𝑡) = 𝑆 𝑡 . 𝑀′(𝑡)
𝑑𝑠

and, since 𝑀′(𝑡) ≠ 0 :

𝑑𝑀
𝑆 𝑡 = 1, ∀𝑡 ∈ 𝐼
𝑑𝑠

Or also :

𝑑𝑀
𝑠 = 1, ∀𝑠 ∈ 𝐽. ∎
𝑑𝑠

In the following we will pose:

𝑑𝑀
𝑇(𝑠) = 𝑠 .
𝑑𝑠
𝑑𝑀 𝑑𝑀
as 𝑠 ≠ 0 Since 𝑠 = 1, 𝑇(𝑠) is therefore a unit vector tangent to Γ = 𝑀 𝐽 =
𝑑𝑠 𝑑𝑠

𝑀 𝐼 to the point 𝑀 𝑠 = 𝑀 ∘ 𝑆 −1 (𝑠) .

II Local Geometric study of a Parametric curve

II.1 Curvature
Throughout this section, Γ, 𝑀 is a parametric curve of class 𝐶 2 whose points are all regular.
Let 𝑀 𝑡0 be a fixet point fixé of Γ = 𝑀 𝐼 , and let the 𝐶 2 -reparametrization
𝑀 = 𝑀 ∘ 𝑆 −1 : 𝐽 → ℝ𝑛 where 𝑆 ∶ 𝑡 ∈ 𝐼 → 𝑆 𝑡 = 𝑠 ∈ 𝐽 (= 𝑆 𝐼 ) is a curvilinear abscissa
of the curve Γ, 𝑀 (the 𝐶 2 -reparametrization 𝑀 exists by virtue of proposition 4.1 of the
previous chapter and theorem 1.1), and let 𝑠0 = 𝑆 𝑡0 .

Definition 2.1: The curvature of Γ = 𝑀 𝐽 = 𝑀 𝐼 at the point 𝑀 𝑠0 = 𝑀(𝑡0 ) is

the real number

𝑑𝑇
𝜌 𝑠0 = 𝑠 ≥0 .
𝑑𝑠 0

𝑑𝑇 𝑑𝑀
This number exists because the derivative exists ( since 𝑇(𝑠) = 𝑠 and 𝑀 is of
𝑑𝑠 𝑑𝑠

class 𝐶 2 ).

Example 1 : Let the parametric curve Γ, 𝑀 of ℝ𝑛 defined by :

18
𝑂𝑀(𝑡) = 𝑡. 𝑢, 𝑡 ∈ ℝ and 𝑢 = 𝑢1 , ⋯ , 𝑢𝑛 ≠ 0 (𝑢 direction vector of the straight line). The
equation of this straight line can also be written : 𝑀 𝑡 = 𝑡 𝑢1 , ⋯ , 𝑢𝑛 = 𝑢1 . 𝑡, ⋯ , 𝑢𝑛 . 𝑡 .𝑀
is of class 𝐶 ∞ and ∀𝑡 ∈ ℝ, 𝑀′ 𝑡 = 𝑢1 , ⋯ , 𝑢𝑛 ≠ 0, ⋯ ,0 , so ∀𝑡 ∈ ℝ, 𝑀 𝑡 is regular. We
can therefore reparametrize the straight line by a curvilinear abscissa (proposition 1.1) and
calculate the curvature at each point of Γ since (theoreme 1.1) 𝑀 is the same class as 𝑀 (it
is enough that 𝑀 is the class 𝐶 2 ).
If we take for example 𝑡0 = 0, we have :
𝑡 𝑡
𝑠=𝑆 𝑡 = 𝑀′(𝜏) 𝑑𝜏 = 𝑢 𝑑𝜏 = 𝑢 . 𝑡 ,
𝑡 0 =0 0

Which is a curvilinear abscissa of Γ, 𝑀 ( for each 𝑡0 we can define a curvilinear abscissa).

Let 𝑀 = 𝑀 ∘ 𝑆 −1 the reparametrization associated with this curvilinear abscissa.
𝑠
𝑠 = 𝑆 𝑡 = 𝑢 . 𝑡 ⟺ 𝑡 = 𝑆 −1 𝑠 = , 𝑠 ∈ ℝ. This implies
𝑢
𝑠 𝑠
𝑀 𝑠 = 𝑀 𝑆 −1 (𝑠) = 𝑀 = ∙ 𝑢1 , ⋯ , 𝑢𝑛 .
𝑢 𝑢
By differentiating 𝑀 we obtain 𝑇(𝑠) the unit vector tangent to Γ
𝑠
𝑑𝑀 𝑑 𝑢
∙𝑢 𝑢
𝑇(𝑠) = 𝑠 = = ∙
𝑑𝑠 𝑑𝑠 𝑢
So
𝑑𝑇 𝑑𝑇
∀𝑠 ∈ ℝ, 𝑠 =0 and consequently ∀𝑠 ∈ ℝ, 𝜌 𝑠 = 𝑠 = 0. That is the
𝑑𝑠 𝑑𝑠

curvature of straight line is zero at each of its points (logical).

Example 2 : Let the parametrized circle Γ, 𝑀 be defined by :

𝑀 𝑡 = 𝑅 cos 𝑡 , 𝑅 sin 𝑡 , 𝑡 ∈ 0, 2𝜋
𝑀 is of class 𝐶 ∞ and 𝑀′ 𝑡 is equal to :

𝑀′ 𝑡 = −𝑅 sin 𝑡 , 𝑅 cos 𝑡 ≠ 0,0 , ∀𝑡 ∈ 0, 2𝜋 .

This implies that 𝑀 𝑡 is regular, ∀𝑡 ∈ 0, 2𝜋 .
We can therefore reparametrize the circle by a curvilinear abscissa.𝑠 = 𝑆(𝑡) (proposition
1.1) and 𝑀 = 𝑀 ∘ 𝑆 −1 is of class 𝐶 ∞ (theorem 1.1).
Let curvilinear abscissa of Γ, 𝑀
𝑡 𝑡
𝑠=𝑆 𝑡 = 𝑀′(𝜏) 𝑑𝜏 = 𝑅𝑑𝜏 = 𝑅𝑡.
0 0

19
 𝑠
𝑠 = 𝑆 𝑡 = 𝑅𝑡 ⟺ 𝑡 = 𝑆 −1 𝑠 = , où 𝑠 = 𝑅𝑡 ∈ 0, 2𝜋𝑅 . This implies
𝑅
𝑠 𝑠 𝑠
𝑀 𝑠 = 𝑀 𝑆 −1 (𝑠) = 𝑀 = 𝑅 cos , 𝑅 sin .
𝑅 𝑅 𝑅
So

𝑑𝑀 𝑠 𝑠
𝑇(𝑠) = 𝑠 = − sin , cos
𝑑𝑠 𝑅 𝑅
And therefore

𝑑𝑇 1 𝑠 1 𝑠 1
𝜌 𝑠 = 𝑠 = − cos , − sin = , ∀𝑠 ∈ 0, 2𝜋𝑅 .
𝑑𝑠 𝑅 𝑅 𝑅 𝑅 𝑅
The curvature of a circle is therefore constant (logical) and equal to the inverse of its radius.

Let us show,more the curvature at point 𝑀 𝑡0 is larger and more 𝑀 𝑡 moves away from
its tangent at 𝑀 𝑡0 .

Let 𝑄 the parametric representation of the tangent of Γ=𝑀 𝐼 =𝑀 𝐽 at 𝑀 𝑡0 = 𝑀 𝑠0

defined by :

𝑑𝑀
𝑂𝑄 𝑠 = 𝑂𝑀 𝑠0 + 𝑠 − 𝑠0 𝑠 = 𝑂𝑀 𝑠0 + 𝑠 − 𝑠0 𝑇(𝑠0 )
𝑑𝑠 0
⟺ 𝑀 𝑠0 𝑄 𝑠 = 𝑠 − 𝑠0 𝑇(𝑠0 ) (1).
Let

𝐹 𝑠 = 𝛿 𝑄 𝑠 , 𝑀(𝑠) = 𝑄 𝑠 𝑀 𝑠 (2).

𝐹 𝑠 be the distance between a point 𝑀(𝑠) on the curve and a point 𝑄 𝑠 on its tangent at
the point 𝑀 𝑠0 in the vicinity of 𝑠 = 𝑠0 .
Let us write the Taylor formula of order 2 at 𝑠 = 𝑠0 for the application 𝑀 :
𝑠−𝑠0 2 𝑑𝑇
𝑀 𝑠0 𝑀 𝑠 = 𝑂𝑀 𝑠 − 𝑂𝑀 𝑠0 = 𝑠 − 𝑠0 𝑇 𝑠0 + 2 𝑑𝑠
𝑠0 + 𝜀(𝑠) (3),

where lim𝑠→𝑠0 𝜀(𝑠) = 0.

From:

𝑄 𝑠 𝑀 𝑠 =
1 3 𝑠−𝑠0 2 𝑑𝑇
𝑀 𝑠0 𝑀 𝑠 − 𝑀 𝑠0 𝑄 𝑠 = 𝑀 𝑠0 𝑀 𝑠 − 𝑠 − 𝑠0 𝑇 𝑠0 = 𝑠0 + 𝜀(𝑠) .
2 𝑑𝑠

If 𝜌 𝑠0 ≠ 0, then we have

20
 𝑠−𝑠0 2 𝑑𝑇
𝐹(𝑠) 2 𝑄 𝑠 𝑀 𝑠 𝑠0 + 𝜀(𝑠)
2 𝑑𝑠
lim 2 = lim 𝑠−𝑠0 2
= lim 𝑠−𝑠0 2
= 1.
𝑠→𝑠0 𝑠−𝑠0 𝑠→𝑠0 𝑠→𝑠0
𝜌 𝑠0 𝜌 𝑠0 𝜌 𝑠0
2 2 2
𝑠−𝑠0 2
We deduse that 𝐹 𝑠 ~ 𝜌 𝑠0 when 𝑠 → 𝑠0 . Consequently, in the neighborhood of
2

𝑠0 , whenever 𝜌 𝑠0 increases, the gap between the points 𝑀 𝑠 and 𝑄 𝑠 increases.

Definition 2.2: Let Γ, 𝑀 a parametric curve of class 𝐶 𝑟 , 𝑟 ≥ 2. A regular point 𝑀 𝑡0

of Γ is said to be biregular if 𝑀′ 𝑡0 and 𝑀′′ 𝑡0 are linearly independent. We say that
Γ, 𝑀 is biregular, if all the points of Γ are biregular.

Proposition 2.1: A point 𝑀 𝑡0 ∈ Γ = 𝑀 𝐼 is biregular if and only if the curvature of Γ

at 𝑀 𝑡0 is nonzero.

Proof: Let us differntiate twice the map 𝑀 = 𝑀 ∘ 𝑆, where 𝑀 is the reparametrization of Γ

by curvilinear abscissa 𝑆.

𝑑𝑀 𝑑𝑀 𝑑𝑆 𝑑𝑆
𝑡0 = 𝑆 𝑡0 . 𝑡0 = 𝑇 𝑆 𝑡0 . 𝑡 , (1)
𝑑𝑡 𝑑𝑠 𝑑𝑡 𝑑𝑡 0

𝑑𝑀 𝑑𝑀
𝑡0 and 𝑠0 = 𝑇 𝑠0 𝑠0 = 𝑆 𝑡0 are therefore collinear .
𝑑𝑡 𝑑𝑠

𝑑2 𝑀 𝑑 𝑑𝑀 𝑑𝑆
𝑡0 = 𝑆 . . . 𝑡0
𝑑𝑡 2 𝑑𝑡 𝑑𝑠 𝑑𝑡

𝑑 𝑑𝑀 𝑑𝑆 𝑑𝑀 𝑑2 𝑆
= 𝑆 . 𝑡0 . 𝑡0 + 𝑆 𝑡0 . 2 𝑡0
𝑑𝑡 𝑑𝑠 𝑑𝑡 𝑑𝑠 𝑑𝑡

2
𝑑2 𝑀 𝑑𝑆 𝑑2 𝑆
= 𝑆 𝑡0 . 𝑡 + 𝑇 𝑠0 . 𝑡
𝑑𝑠 2 𝑑𝑡 0 𝑑𝑡 2 0
2
𝑑𝑇 𝑑𝑆 𝑑2 𝑆
= 𝑠0 . 𝑡 + 𝑇 𝑠0 . 2 𝑡0 (2)
𝑑𝑠 𝑑𝑡 0 𝑑𝑡

i) Necessary condition :

21
 2
𝑑𝑇 𝑑𝑆
If 𝑀 𝑡0 is biregular, it follows from (2) that 𝑠0 . 𝑡0 ≠ 0 since 𝑀′′ 𝑡0 would
𝑑𝑠 𝑑𝑡

be collinear with 𝑀′ 𝑡0 since from (1) the latter vector is collinear with 𝑇 𝑠0 .
𝑑𝑇 𝑑𝑇
Consequently 𝑠0 ≠ 0 et donc 𝜌 𝑠0 = 𝑠0 ≠ 0.
𝑑𝑠 𝑑𝑠

ii) Sufficient Condition :

If 𝜌 𝑠0 ≠ 0, the linear combination :

𝑑𝑀 𝑑2 𝑀
𝛼 𝑡0 + 𝛽 2 𝑡0 = 0, (𝛼, 𝛽) ∈ ℝ2
𝑑𝑡 𝑑𝑡
Is written, using (1) and (2), as :
2
𝑑𝑆 𝑑2 𝑆 𝑑𝑆 𝑑𝑇
𝛼 𝑡 + 𝛽 2 𝑡0 𝑇 𝑠0 +𝛽 𝑡 𝑠 = 0. (3)
𝑑𝑡 0 𝑑𝑡 𝑑𝑡 0 𝑑𝑠 0
2
Since 𝑇 𝑠 = 𝑀′ 𝑠 = 1, then 𝑇 𝑠 . 𝑇 𝑠 = 𝑀′ 𝑠 = 1. this implies that

𝑑 𝑇 𝑠 .𝑇 𝑠 𝑑𝑇 𝑑𝑇
= 2𝑇 𝑠 . 𝑠 =0 ⟹ 𝑇 𝑠 . 𝑠 = 0.
𝑑𝑠 𝑑𝑠 𝑑𝑠
𝑑𝑇
We deduce that by making the scalar product of (3) by 𝑠0 , we obtain :
𝑑𝑠
2 2
𝑑𝑆 𝑑𝑇 𝑑𝑇 𝑑𝑆 2
2
2
𝛽 𝑡 𝑠0 . 𝑠0 = 𝛽 𝑡 𝜌 𝑠0 = 𝛽 𝑀′ 𝑡0 𝜌 𝑠0 = 0.
𝑑𝑡 0 𝑑𝑠 𝑑𝑠 𝑑𝑡 0

As by hypothesis, in this paragraph, all the points 𝑀(𝑡) are regular, So 𝑀′ 𝑡0 ≠ 0, and

as sufficient condition 𝜌 𝑠0 ≠ 0, then 𝛽 = 0 and by substituting in (3), we will also have

𝛼 = 0. Therefore 𝑀′ 𝑡0 et 𝑀′′ 𝑡0 are linearly independent, so 𝑀 𝑡0 is biregular point

of Γ, 𝑀 .∎

Let Γ, 𝑀 a parametric curve of classe 𝐶 2 and biregular. The expression of the

curvature of Γ at a point 𝑀(𝑡0 ), where 𝑀 is not necessarily a parametrization by a
curvilinear abscissa, is giyen by :
- If Γ ⊂ ℝ𝑛
1 2 2 2 1/2
𝜌 𝑡0 = 3 𝑀′′ 𝑡0 𝑀′ 𝑡0 − 𝑀′′ 𝑡0 . 𝑀′ 𝑡0 .
𝑀′ 𝑡0

- If Γ is a plane curve

22
and 𝑀′ 𝑡0 and 𝑀′′ 𝑡0 are expressed with respect to an orthonormal basis of ℝ2 , then we
have :

𝑑𝑒𝑡 𝑀′ 𝑡0 , 𝑀′′ 𝑡0
𝜌 𝑡0 = 3 .
𝑀′ 𝑡0

- If Γ is a curve in space, we have :

𝑀′ 𝑡0 ⋀ 𝑀′′ 𝑡0
𝜌 𝑡0 = 3 .
𝑀′ 𝑡0

Reminder : Let 𝑢1 et 𝑢2 be two vectors in ℝ3 written in the canonical basis 𝑖, 𝑗, 𝑘 ,

or in a direct orthonormal basis, as follows :
𝑢1 = 𝑥1 𝑖 + 𝑦1 𝑗 + 𝑧1 𝑘et𝑢2 = 𝑥2 𝑖 + 𝑦2 𝑗 + 𝑧2 𝑘.
𝑢1 ∧ 𝑢2 is a vector orthonormal to 𝑢1 and 𝑢2 equal to :

𝑖 𝑗 𝑘
𝑢1 ∧ 𝑢2 = 𝑦1 𝑧2 − 𝑧1 𝑦2 𝑖 + 𝑧1 𝑥2 − 𝑥1 𝑧2 𝑗 + 𝑥1 𝑦2 − 𝑦1 𝑥2 𝑘 = 𝑥1 𝑦1 𝑧1 .
𝑥2 𝑦2 𝑧2

Definition 2.3 : Let Γ, 𝑀 be a parametric curve in ℝ𝑛 of class 𝐶 2 and regular, and

let 𝑀 𝑡0 be biregular point . Let 𝑀 = 𝑀 ∘ 𝑆 −1 : 𝐽 ⟶ ℝ𝑛 be a reparametrization by a
curvilinear abscissa such that 𝑡 ∈ 𝐼 ⟶ 𝑠 = 𝑆 𝑡 ∈ 𝐽 = 𝑆(𝐼), 𝑠0 = 𝑆(𝑡0 ).
1) We call the radius of curvature of Γ at 𝑀 𝑡0 = 𝑀 𝑠0 the number
1
𝑅 𝑠0 = >0 .
𝜌 𝑠0
2) We call the principal normal vector to Γ at 𝑀 𝑡0 = 𝑀 𝑠0 the vector
1 𝑑𝑇 𝑑𝑇
𝑁 𝑠0 = 𝑠0 = 𝑅 𝑠0 𝑠 .
𝜌 𝑠0 𝑑𝑠 𝑑𝑠 0
3) We call the center of curvature of Γ at 𝑀 𝑡0 = 𝑀 𝑠0 , the point 𝐶 𝑠0 such that

𝑂𝐶 𝑠0 − 𝑂𝑀 𝑠0 = 𝑀 𝑠0 𝐶 𝑠0 = 𝑅 𝑠0 𝑁 𝑠0 .
the center of curvature of e Γ at 𝑀 𝑡0 is the center of the tangent circle to Γ at 𝑀 𝑡0
and the radius equal to radius of curvature of Γ at 𝑀 𝑡0 .

From the definition of the principal normal vector 𝑁 𝑠0 we deduce that

23
 1 𝑑𝑇 1
𝑁 𝑠0 = 𝑠 = ∙ 𝜌 𝑠0 = 1,
𝜌 𝑠0 𝑑𝑠 0 𝜌 𝑠0
2
i.e. 𝑁 𝑠0 is a unit vector . Moreover, since 𝑇 𝑠 . 𝑇 𝑠 = 𝑇 𝑠 = 12 = 1, then

𝑑 𝑇 𝑠 .𝑇 𝑠 𝑑𝑇
= 2𝑇 𝑠 . 𝑠 = 0.
𝑑𝑠 𝑑𝑠
𝑑𝑇
This implies that 𝑠 is orthogonal to 𝑇 𝑠 and therefore 𝑁 𝑠 , which is collinear to
𝑑𝑠

𝑑𝑇
𝑠 , it is orthogonal to 𝑇 𝑠 . Therefore 𝑇 𝑠 , 𝑁 𝑠 is an orthonormal basis of ℝ2 .
𝑑𝑠

II.2 Plane Curves

Let Γ, 𝑀 be a parametric plane curve of class 𝐶 3 , the parameter being a curvilinear

abscissa 𝑠 (𝑠 ∈ 𝐼 ⟶ 𝑀 𝑠 ∈ Γ). Suppose that all the points of Γ are biregular. The radius
1
of curvature 𝑅 𝑠 = 𝜌 𝑠 of Γ at any point 𝑀 𝑠 is then defined (since by hypothesis Γ is

biregular and therefore from proposition 2.1 ∀𝑠 ∈ 𝐼, 𝜌 𝑠 ≠ 0) and the same is true of the

center of curvature 𝐶(𝑠) of Γ at 𝑀 𝑠 𝑀 𝑠 𝐶 𝑠 =𝑅 𝑠 𝑁 𝑠 .

𝑑𝑇
Since by hypothesis Γ, 𝑀 is of class 𝐶 3 , therefore is of class 𝐶 1 since by definition
𝑑𝑠

𝑑𝑀
𝑇= . The radius of curvature 𝑅 𝑠 therefore has continuous derivative expressed by :
𝑑𝑠

𝑑𝑇 𝑑2𝑇
𝑑𝑅 𝑠 . 2 𝑠
𝑠 = − 𝑑𝑠 3 𝑑𝑠 .
𝑑𝑠 𝜌 (𝑠)
Indeed, we have :
𝑑𝑅 𝑑 1 𝜌′ 𝑠
𝑠 = 𝑠 =− 2 . (1)
𝑑𝑠 𝑑𝑠 𝜌 𝑠 𝜌 (𝑠)
Moreover, as
2
𝑑𝑇 𝑑𝑇 𝑑𝑇
𝜌2 (𝑠) = 𝑠 = 𝑠 . 𝑠 , (2)
𝑑𝑠 𝑑𝑠 𝑑𝑠

Then,
′

2 ′
𝑑𝑇 𝑑𝑇 ′
𝑑𝑇 2𝑑2 𝑇
𝜌 (𝑠) 𝑠 = 𝑠 . 𝑠 ⟺ 2𝜌 𝑠 𝜌 𝑠 = 2 𝑠 . 2 𝑠 ⟹
𝑑𝑠 𝑑𝑠 𝑑𝑠 𝑑𝑠
𝑠

24
 𝑑𝑇 𝑑2𝑇
𝑠 . 𝑑𝑠 2 𝑠
𝑑𝑠
⟹ 𝜌′ 𝑠 = .
𝜌 𝑠
We deduce that
𝑑𝑇 𝑑2𝑇
𝑑𝑅 1 𝑠 . 2 𝑠
𝑠 = − 𝑑𝑠 3 𝑑𝑠
𝑑𝑠 𝜌 (𝑠)
𝑑𝑅
Since 𝑀 is of class 𝐶 3 , then is continuous.
𝑑𝑠

𝑑𝑇
Since is of class 𝐶 1 and we have just shown that the function 𝑠 ⟶ 𝑅(𝑠) is also of class
𝑑𝑠

𝑑𝑇
𝐶 1 , then the principal normal vector 𝑁 = 𝑅 𝑑𝑠 is of class 𝐶 1 and thus, is the same for the

application 𝑠 ∈ 𝐼 ⟶ 𝐶(𝑠) ∈ ℝ2 which gives the centers of curvature of Γ 𝑂𝐶 𝑠 =

𝑂𝑀 𝑠 + 𝑅 𝑠 𝑁 𝑠 . This application défines a parametric curve and its image Γ𝐷 = 𝐶(𝐼) is

called the the evolute of the curve Γ.
𝑑𝑁 𝑑𝑂𝐶
Let us express 𝑠 and 𝑠 in the orthonormal basis 𝑇 𝑠 , 𝑁 𝑠 , let us first show
𝑑𝑠 𝑑𝑠
the following result :

Proposition 2.2. Let us 𝑢1 and 𝑢2 be applications of 𝐼 ⊂ ℝ ⟶ ℝ2 such us

∀𝑡 ∈ 𝐼, 𝑢1 (𝑡) , 𝑢2 (𝑡) is an orthonormal basis of ℝ2 . Suppose that 𝑢1 and 𝑢2 are

differentiable on I. Then the matrix of coordinates of 𝑢′ 1 𝑡 and 𝑢′2 (𝑡), whith respect to
𝑢1 (𝑡) , 𝑢2 (𝑡) , is antisymmetric (i.e. ∀𝑖, 𝑗𝑎𝑖𝑗 = 𝑎𝑗𝑖 ).

2
Proof : For 𝑖 = 1, 2, 𝑢𝑖 (𝑡) = 1 ⟹ 𝑢𝑖 𝑡 = 1 and we have :

𝑑(1) 𝑑 2 𝑑 𝑑𝑢𝑖 𝑡
0= = 𝑢𝑖 𝑡 = 𝑢 𝑡 . 𝑢𝑖 𝑡 = 2𝑢𝑖 𝑡 . (∗)
𝑑𝑡 𝑑𝑡 𝑑𝑡 𝑖 𝑑𝑡
On the other hand , since 𝑢1 (𝑡) and 𝑢2 (𝑡) are orthogonal 𝑢1 (𝑡) . 𝑢2 (𝑡) = 0 , then

𝑑(0) 𝑑 𝑑𝑢1 𝑡 𝑑𝑢2 𝑡

0= = 𝑢 𝑡 . 𝑢2 𝑡 = ∙ 𝑢2 𝑡 + 𝑢1 𝑡 . (∗∗)
𝑑𝑡 𝑑𝑡 1 𝑑𝑡 𝑑𝑡
Let us write 𝑢′1 (𝑡) et 𝑢′2 (𝑡) in the basis 𝑢1 (𝑡) , 𝑢2 (𝑡) :

𝑢′1 (𝑡) = 𝛼𝑢1 (𝑡) + 𝛽𝑢2 (𝑡)et𝑢′2 (𝑡) = 𝛾𝑢1 (𝑡) + 𝛿𝑢2 (𝑡),
This implies that

25
 (∗)
𝑢′1 (𝑡). 𝑢1 (𝑡) = 𝛼𝑢1 (𝑡) + 𝛽𝑢2 (𝑡) . 𝑢1 (𝑡) = 𝛼 = 0,

(∗)
𝑢′2 (𝑡). 𝑢2 (𝑡) = 𝛾𝑢1 (𝑡) + 𝛿𝑢2 (𝑡) . 𝑢2 (𝑡) = 𝛿 = 0
and
(∗∗)
𝛽 = 𝛼𝑢1 (𝑡) + 𝛽𝑢2 (𝑡) . 𝑢2 (𝑡) = 𝑢′1 (𝑡). 𝑢2 (𝑡) = − 𝑢1 𝑡 . 𝑢′ 2 𝑡

= −𝑢1 𝑡 𝛾𝑢1 𝑡 + 𝛿𝑢2 𝑡 = −𝛾.

Therefore
𝛼 𝛽 0 𝛽
= .∎
𝛾 𝛿 −𝛽 0

𝑑𝑇 𝑑𝑁
This proposition implies that the matrix of coordinates of 𝑠 and 𝑠 with respect to
𝑑𝑠 𝑑𝑠

the basis 𝑇 𝑠 , 𝑁 𝑠 is antisymmetric, i.e.

𝑎11 𝑎12 0 𝑎12
𝑎21 𝑎22 = −𝑎12 0
. (∗)

By definition of the principal normal vector , we have :

𝑑𝑇
𝑠 = 𝜌 𝑠 𝑁 𝑠 = 0. 𝑇 𝑠 + 𝜌 𝑠 𝑁 𝑠 = 𝑎11 s 𝑇 𝑠 + 𝑎12 𝑠 𝑁 𝑠 . (∗∗)
𝑑𝑠
From (∗) et (∗∗) we have therefore
𝑑𝑁
𝑠 = 𝑎21 s 𝑇 𝑠 + 𝑎22 𝑠 𝑁 𝑠 = −𝜌 𝑠 𝑇 𝑠 + 0. 𝑁 𝑠 = −𝜌 𝑠 𝑇 𝑠 .
𝑑𝑠
𝑑𝑂𝐶
Let us calculate 𝑠 :
𝑑𝑠

𝑑𝑂𝐶 𝑑𝑂𝑀 𝑑𝑅 𝑑𝑁 𝑑𝑅
𝑠 = 𝑠 + 𝑠 𝑁 𝑠 +𝑅 𝑠 𝑠 =𝑇 𝑠 + 𝑠 𝑁 𝑠 −𝑅 𝑠 𝜌 𝑠 𝑇 𝑠
𝑑𝑠 𝑑𝑠 𝑑𝑠 𝑑𝑠 𝑑𝑠
𝑑𝑅 𝑑𝑅
=𝑇 𝑠 + 𝑠 𝑁 𝑠 − 1. 𝑇 𝑠 = 𝑠 𝑁 𝑠 .
𝑑𝑠 𝑑𝑠

We deduce from this last formula that , if the point 𝐶(𝑠) of Γ𝐷 = 𝐶(𝐼) ( the evolute of Γ )
𝑑𝑂𝐶
is regular (i.e. 𝑠 ≠ 0), the tangent to Γ𝐷 at point 𝐶(𝑠) coincides with the normal to
𝑑𝑠

Γ = 𝑀(𝐼) at point 𝑀(𝑠).

II.3 Curves of space –Frenet Frame

26
Let Γ, 𝑀 , 𝑀: 𝐼 ⟶ ℝ3 , be a parametric curve of the space (also called a left curve), of class
𝐶 3 , all of whose points are biregular. The parameter being curvilinear abscissa 𝑠.

Definition 2.4:
1) The vector 𝐵 𝑠 = 𝑇 𝑠 ∧ 𝑁 𝑠 is called a binormal vector to Γ at the point 𝑀(𝑠).
2 2 2
𝐵 𝑠 = 𝑇 𝑠 ∧ 𝑁 𝑠 = 𝑇 𝑠 𝑁 𝑠 − 𝑇 𝑠 .𝑁 𝑠 = 12 . 12 − 02 = 1.

2 The direct orthonormé frame 𝑀 𝑠 , 𝑇 𝑠 , 𝑁 𝑠 , 𝐵 𝑠 is called the Frénet frame of Γ

at the point 𝑀(𝑠).

3) - The plane defined by 𝑀 𝑠 , 𝑇 𝑠 , 𝑁 𝑠 is called the osculating plane to Γ at the

point 𝑀(𝑠) (is the plane that best contains Γ).

- The plane defined by 𝑀 𝑠 , 𝑁 𝑠 , 𝐵 𝑠 is called the normal plane to Γ at the point

𝑀(𝑠).

The plane defined by 𝑀 𝑠 , 𝑇 𝑠 , 𝐵 𝑠 is called the rectifying to Γ at the point 𝑀(𝑠).

1 𝑑𝑇
Since Γ, 𝑀 is of class 𝐶 3 , then 𝑁 𝑠 = 𝜌 𝑠 𝑠 is differentiable. Therefore
𝑑𝑠

𝐵 𝑠 = 𝑇 𝑠 ∧ 𝑁 𝑠 is also differentiable.

𝑑𝑁
Definition 2.5: The real 𝜏 𝑠 = 𝐵 𝑠 ∙ 𝑑𝑠 (𝑠) is called torsion of the curve Γ at the point
𝑀(𝑠).

Théorème 2.1: The derivatives of the vectors of the Frénet frame of Γ at a point 𝑀(𝑠)
are expressed, with respect to this same frame, by :

27
 𝑑𝑇
𝑠 = 𝜌 𝑠 𝑁 𝑠
𝑑𝑠
𝑑𝑁
𝑠 = −𝜌 𝑠 𝑇 𝑠 + 𝜏 𝑠 𝐵 𝑠
𝑑𝑠
𝑑𝐵
𝑠 = −𝜏 𝑠 𝑁 𝑠
𝑑𝑠
These formulas are called Frénet formulas.

Proof : By definition of the principal normal vector, we have :

𝑑𝑇
𝑠 =𝜌 𝑠 𝑁 𝑠 ,
𝑑𝑠
That is the first Frenet formula.
𝑑𝑇 𝑑𝑁 𝑑𝐵
We know that matrix of coordinates of 𝑠 , 𝑠 and 𝑠 in the basis 𝑇 𝑠 , 𝑁 𝑠 ,
𝑑𝑠 𝑑𝑠 𝑑𝑠

𝐵 𝑠 is antisymmetric, so we have:
𝑑𝑇
𝑠 0 𝜌 𝑠 0 𝑇 𝑠 𝜌 𝑠 𝑁 𝑠
𝑑𝑠
𝑑𝑁
𝑠 = −𝜌 𝑠 0 𝛼 𝑁 𝑠 = −𝜌 𝑠 𝑇 𝑠 + 𝛼𝐵 𝑠 (∗)
𝑑𝑠
𝑑𝐵 0 −𝛼 0
𝑠 𝐵 𝑠 −𝛼𝑁 𝑠
𝑑𝑠
It remains to determine 𝑡𝑜 𝑓𝑖𝑛𝑑 𝑡𝑕𝑒 𝑙𝑎𝑠𝑡 𝑡𝑤𝑜 Frénet.formulas.

The vector 𝐵 𝑠 = 𝑇 𝑠 ∧ 𝑁 𝑠 is orthogonal to the vectors 𝑇 𝑠 and 𝑁 𝑠 , so

𝑑 0 𝑑 𝑑𝐵 𝑑𝑁
𝐵 𝑠 ∙ 𝑁 𝑠 =0 ⟹ 0= = 𝐵 𝑠 ∙ 𝑁 𝑠 = 𝑠 .𝑁 𝑠 + 𝐵 𝑠 . 𝑠
𝑑𝑠 𝑑𝑠 𝑑𝑠 𝑑𝑠
𝑑𝐵
= 𝑠 .𝑁 𝑠 + 𝜏 𝑠
𝑑𝑠
𝑑𝐵
⟹ 𝑠 . 𝑁 𝑠 = −𝜏 𝑠 . (∗∗)
𝑑𝑠
From (∗) and (∗∗) we have :
∗∗ 𝑑𝐵 ∗ 2
−𝜏 𝑠 = 𝑠 . 𝑁 𝑠 = − 𝛼𝑁 𝑠 . 𝑁 𝑠 = −𝛼 𝑁 𝑠 = −𝛼. 1 ⟹ 𝛼 = 𝜏 𝑠 . ∎
𝑑𝑠

Projection of 𝚪, 𝑴 onto the osculating, normal and rectifying planes :

Let Γ, 𝑀 a curve in space, parametrized by curvilinear abscissa 𝑠, of class 𝐶 3 , *

28
biregular and whose torsion 𝜏 𝑠0 at a point 𝑀(𝑠0 ) is non-zero. The Taylor formula at
order 3 in 𝑠0 of the application 𝑀 is written :
2 3
𝑑𝑂𝑀 𝑠 − 𝑠0 𝑑 2 𝑂𝑀 𝑠 − 𝑠0 𝑑 3 𝑂𝑀
𝑂𝑀 𝑠 = 𝑂𝑀 𝑠0 + 𝑠 − 𝑠0 𝑠0 + 2
𝑠0 + 𝑠0
𝑑𝑠 2! 𝑑𝑠 3! 𝑑𝑠 3
3
𝑠 − 𝑠0
+ 𝑢 𝑠 , (1)
3!
with lim 𝑢 𝑠 = 0.
𝑠→𝑠0

𝑑𝑂𝑀 𝑑 2 𝑂𝑀 𝑑𝑇
As 𝑠0 = 𝑇 𝑠0 et 𝑠0 = 𝑠0 = 𝜌 𝑠0 𝑁 𝑠0 , and using Frénet’s formulas ,
𝑑𝑠 𝑑𝑠 2 𝑑𝑠

the Taylor formula above, with respect to the Frénet frame 𝑀 𝑠0 , 𝑇 𝑠0 , 𝑁 𝑠0 , 𝐵 𝑠0 ,

is written :

𝑂𝑀 𝑠 − 𝑂𝑀 𝑠0 = 𝑀 𝑠0 𝑀 𝑠
2
𝑠 − 𝑠0
= 𝑠 − 𝑠0 1 + 𝛼 𝑠 𝑇 𝑠0 + 𝜌 𝑠0 + 𝛽 𝑠 𝑁 𝑠0
2!
3
𝑠 − 𝑠0
+ 𝜌 𝑠0 𝜏 𝑠0 + 𝑕 𝑠 𝐵 𝑠0 ,
3!
or

𝑢 𝑠 = 𝑓 𝑠 𝑇 𝑠0 + 𝑔 𝑠 𝑁 𝑠0 + 𝑕 𝑠 𝐵 𝑠0 ,
𝑠 − 𝑠0 2
𝛼 𝑠 = −𝜌2 𝑠0 + 𝑓(𝑠) ,
3!
𝑠 − 𝑠0 𝑑𝜌
𝛽 𝑠 = 𝑠 + 𝑔(𝑠)
3 𝑑𝑠 0
and
lim 𝛼 𝑠 = lim 𝛽 𝑠 = lim 𝑓 𝑠 = lim 𝑔 𝑠 = lim 𝑕 𝑠 = 0.
𝑠→𝑠0 𝑠→𝑠0 𝑠→𝑠0 𝑠→𝑠0 𝑠→𝑠0

We deduce the shape , in the vicinity of 𝑀(𝑠0 ), of the orthogonal projections of the curve Γ
onto its osculating, normal and rectifying planes at point 𝑀(𝑠0 ).

1 - Orthogonal Projection of onto its osculating plane at point 𝑀(𝑠0 )

29
 2 - Projection orthogonale de Γ sur son plan normal au point 𝑀(𝑠0 )

3 - Projection orthogonale de Γ sur son plan rectifiant au point 𝑀(𝑠0 )

Expression of torsion with respect to any parameter t

Let Γ, 𝑀 be a parametrized curve in espace, of class 𝐶 3 and all of whose points 𝑀(𝑡) are
biregular. The torsion 𝜏 𝑡 of Γ at a point 𝑀(𝑡) is equal to:

30
 𝑀′ 𝑡 , 𝑀′′ 𝑡 , 𝑀′′′ 𝑡
𝜏 𝑡 = 2 ,
𝑀′ 𝑡 ⋀ 𝑀′′ 𝑡

where

𝑀′ 𝑡 , 𝑀′′ 𝑡 , 𝑀′′′ 𝑡 = 𝑀′ 𝑡 . 𝑀′ 𝑡 ⋀ 𝑀′′ 𝑡

In the mixed product of the vectors 𝑀′ 𝑡 , 𝑀′′ 𝑡 and 𝑀′′′ 𝑡 .

The mixed product is an alternating multilinear form; this implies that

𝑀′ 𝑡 , 𝑀′′ 𝑡 , 𝑀′′′ 𝑡 = − 𝑀′′ 𝑡 , 𝑀′ 𝑡 , 𝑀′′′ 𝑡 .

In the canonical basis 𝑖, 𝑗, 𝑘 or in a direct orthonormal basis, the mixed product is equal
to :

𝑀′ 𝑡 , 𝑀′′ 𝑡 , 𝑀′′′ 𝑡 = 𝑑𝑒𝑡 𝑀′ 𝑡 , 𝑀′′ 𝑡 , 𝑀′′′ 𝑡 .

Theorem 2.2: Let Γ, 𝑀 be a parametric curve in space, of class 𝐶 3 and whose points
𝑀(𝑡) are all biregular. The curve Γ is planar if and only if its torsion is identically zero (i.e.
∀𝑡 ∈ 𝑑𝑜𝑚𝑀, 𝜏 𝑡 = 0).

31