Polygonal lengths record the path in order
Let be continuously differentiable on , where . For a partition , add the lengths of consecutive chords:
Definition 5.9.1
(Traced length and speed)
The traced length of the path is the supremum of over all finite partitions of . The speed is .
This definition follows the parameter order: if a path traverses a segment twice, the two traversals contribute separately. The speed is defined, and may be zero, at times when both derivatives vanish.
Why the speed integral equals the traced length
Theorem 5.9.2
(Arc length of a parametric path)
If is continuously differentiable on , its traced length is
No assumption that the speed is nonzero is needed.
Proof
(Chords below the speed integral, then close to it)
For each subinterval, the fundamental theorem gives componentwise. Its norm is at most the integral of the speed. To see this, if the chord is nonzero, take its unit direction and write its norm as ; the zero chord is immediate. Summing gives for every partition.
For the reverse bound, uniform continuity of on supplies a modulus such that whenever . If the mesh of is at most , the chord on differs from by a vector of norm at most . The reverse triangle inequality therefore bounds the chord below by the left-endpoint speed times the interval width minus that error.
The speed itself varies by at most on the same cell, so its integral differs from the corresponding left-endpoint sum by at most . Combining the two estimates gives
Arbitrarily fine finite partitions exist. Letting and using the upper bound shows that the supremum is exactly
No nonzero-speed assumption was needed. The continuity hypothesis supplies both the integral and the uniform error control; a different regularity class needs its own justification.
Proof X-ray
Two inequalities, and where continuity enters
The proof squeezes the supremum between two bounds, and only one of them needs the hypothesis. The upper bound holds for every partition and uses only the fundamental theorem and the unit-vector trick: pairing the chord with its own direction turns a norm into an ordinary integral that can be bounded by the integral of . The non-obvious move is the lower bound. A chord is not the integral of , but on a fine cell is almost constant, so the chord is almost , and the reverse triangle inequality turns that closeness into an inequality. Uniform continuity of on the closed interval is what makes one work on every cell at once; without it the error could not be summed over cells. The argument closes because both bounds have the same limit as the mesh tends to : the lower bound loses at most . Nothing divides by the speed, which is why zero speed does no harm.
Worked example: The length of a parabolic path
Worked example 5.9.3
(Length of a parabolic path)
Problem. Find the traced length of for .
Solution. The velocity is , so the length is . To derive a primitive, integrate by parts:
Since , rearrangement gives . Differentiating gives ; its argument is positive here. Therefore
The coordinate is strictly increasing, so this path does not retrace its image. Substituting into the graph-length formula gives the same integral as an independent check.
Worked example: Zero speed at a cusp
Worked example 5.9.4
(A cusp where the velocity vanishes)
Let for . Then , which is zero at , and the speed is . The speed is even, so
The substitution , was used for . The path is not smooth as a curve at the origin, since the tangent direction reverses there, yet the formula applies because the parametrization is continuously differentiable. The map is injective, so nothing is retraced. A numerical check: , so , and this exceeds the straight-line distance between the endpoints and , as it must.
Worked example: Reparametrizing does not change the length
Worked example 5.9.5
(Same parabola, different clock)
Let for . This follows the same parabola from to . Now and the speed is . With , ,
the integral of the first example. The two parametrizations traverse the same curve once in the same direction, and they give the same length. The speed itself is different at corresponding points, but the integral of speed against time is not.
The image can be shorter than the journey
The length of a path and the length of its image can be different numbers.
Counterexample mode
The speed integral is not the length of the image
Claim: is the length of the geometric curve . Let for . The velocity is , so the speed is . Split at the reversal:
The image is only the horizontal segment from to , whose geometric length is . The path travels from to and then back. Its velocity vanishes at the turning time, but the length formula remains valid. The formula measures the journey; it equals the length of the image only when the path covers its image once, for example when is injective.
Common mistakes
Common mistake
(Adding the component speeds instead of taking the norm)
For the parabola the sum is not the length: it exceeds . The speed is the Euclidean norm of the velocity, not the sum of the component sizes.
Common mistake
(Integrating the signed velocity)
For on , the integral is the net displacement, not the distance traveled. Length needs the absolute value of the velocity, which is where the reversal at is detected.
Exercises
Quick check
(Find the traced length of for .)
Compute the speed first.
Solution · Answer
has speed , so . The image is a half circle of radius , traversed once.
Quick check
(The same circle is run for . What are the traced length and the length of the image?)
The speed does not change.
Solution · Answer
The traced length is . The image is the unit circle, whose length is , since the path goes around twice.
Exercise 5.9.6
(Find the traced length of for .)
Core
Solution(3)
- 1
Differentiate each component
.
- 2
Compute the speed
Key step
, so the speed is .
Because: The cross terms cancel when the squares are added.
- 3
Integrate
.