Parametrized Curves and Arclength: Appendix E
Parametrized Curves and Arclength: Appendix E
Parametrized Curves
and Arclength
For many purposes, the precise parametrization of a curve 𝜎 is not
important, in the sense that some property of the curve that we are
interested in is unchanged if we “reparametrize” the curve. Let us
look at just what reparametrization means. Suppose that 𝑡 is a 𝐶 1
function with a strictly positive derivative on a closed interval [𝛼, 𝛽].
Then 𝑡 is strictly monotonic, and hence it maps [𝛼, 𝛽] one-to-one
onto some other closed interval [𝑎, 𝑏]. Thus if 𝜎 : [𝑎, 𝑏] → R𝑛 is a
𝐶 1 parametrized curve, then 𝜎 ˜ = 𝜎 ∘ 𝑡 : [𝛼, 𝛽] → R𝑛 is another 𝐶 1
parametrized curve which clearly has the same image as 𝜎 and is called
the reparametrization of 𝜎 defined by the parameter transformation
𝑡. (If you like, you can think of 𝑡 as a “variable that parameterizes
the points of the interval [𝛼, 𝛽] by points of the interval [𝑎, 𝑏]” and
with this interpretation 𝜎 and 𝜎 ˜ become “the same”.) In particular,
given any interval [𝛼, 𝛽], we always find an affine map 𝑡(𝜏 ) = 𝑐𝜏 + 𝑘
that maps it onto [𝑎, 𝑏], so reparametrization allows us to adjust a
parameter interval as convenient in situations where parametrization
is not relevant.
A reparametrization of 𝜎 : [𝑎, 𝑏] → R can always be thought of
as arising by starting from a positive, continuous function 𝜌 : [𝑎, 𝑏] →
R and letting 𝑡 ∫be the inverse function of its indefinite integral, 𝜏 .
𝑡
In fact 𝜏 (𝑡) := 𝑎 𝜌(𝜉) 𝑑𝜉 is a smooth 𝐶 1 function with a positive
derivative, so it does indeed map [𝑎, 𝑏] one-to-one onto some interval
[𝛼, 𝛽], and by the inverse function theorem 𝑡 := 𝜏 −1 : [𝛼, 𝛽] → R is
𝐶 1 with a positive derivative. A very important special case of this is
reparametrization by arclength. Suppose that 𝜎 is nonsingular, i.e.,
255
256 E. Parametrized Curves and Arclength
∫𝑡
𝜎 ′ never vanishes. Define 𝑠 : [𝑎, 𝑏] → R by 𝑠(𝑡) := 𝑎 ∥𝜎 ′ (𝜉)∥ 𝑑𝜉, and
recall that by definition it gives the arclength along 𝜎 from 𝑎 to 𝑡.
This is a smooth map with positive derivative ∥𝜎 ′ (𝑡)∥ mapping [𝑎, 𝑏]
onto [0, 𝐿], where 𝐿 is the length of 𝜎. The inverse function, 𝑡(𝑠),
mapping [0, 𝐿] to [𝑎, 𝑏], gives the point of [𝑎, 𝑏] where the arclength
of 𝜎 measured from its left endpoint is 𝑠, and the curve 𝑠 → 𝜎(𝑡(𝑠))
is a reparametrization of 𝜎 called its reparametrization by arclength.
More generally, we say that a curve 𝜎 : [𝑎, 𝑏] → R𝑛 is parameterized
by arclength if the length of 𝜎 between 𝜎(𝑎) and 𝜎(𝑡) is equal to 𝑡 − 𝑎,
and we say that 𝜎 is parametrized proportionally to arclength if that
length is proportional to 𝑡 − 𝑎.
⊳ Exercise E–3. Prove the old saying, “A straight line is the short-
est distance between two points.” That is, if 𝜎 : [𝑎, 𝑏] → R𝑛 is
a 𝐶 1 path of length 𝐿, then ∥𝜎(𝑏) − 𝜎(𝑎)∥ ≤ 𝐿, with equality if
and only if 𝜎 is a straight line from 𝜎(𝑎) to 𝜎(𝑏). (Hint: As we
have just seen, we can assume without loss of generality that 𝜎 is
parametrized proportionally to arclength, i.e., that ∥𝜎 ′ (𝑡)∥ is a con-
stant. Let 𝑣 := 𝜎(𝑏) − 𝜎(𝑎), so that what we must show is that
∥𝑣∥ ≤ 𝐿 with equality if and only if 𝜎 ′ is a constant. If 𝑣 = 0, i.e., if
𝜎(𝑏) = 𝜎(𝑎), the result is trivial so we can assume 𝑣 ∕= 0 and define
𝑣
∫𝑏
a unit vector 𝑒 = ∥𝑣∥ , so that ∥𝑣∥ = ⟨𝑣, 𝑒⟩. Now 𝑣 = 𝑎 𝜎 ′ (𝑡) 𝑑𝑡, and
∫𝑏
since 𝑒 is a constant vector, ∥𝑣∥ = ⟨𝑣, 𝑒⟩ = 𝑎 ⟨𝜎 ′ (𝑡), 𝑒⟩ 𝑑𝑡. Finally
note that by the Schwarz Inequality, ⟨𝜎 ′ (𝑡), 𝑒⟩ ≤ ∥𝜎 ′ (𝑡)∥ and equality
holds for all 𝑡 if and only if 𝜎 ′ (𝑡) is a multiple of 𝑒 for each 𝑡, and this
multiple must be a constant since ∥𝜎 ′ (𝑡)∥ is a constant.)