Calculus › Section P · Vector analysis › Topic P.2
Line integrals
An integral along a curve instead of an axis: one form weighs a wire, the other measures a force field's work along a path. The difference between them — orientation — is the thread the whole chapter pulls.
Weighing a wire
An ordinary integral adds \(f(x)\,dx\) along a straight axis; a line integral adds \(f(x, y)\) along a curve. Picture a thin wire bent along a curve \(C\) with density \(f(x, y)\) mass per unit length: slice \(C\) into short pieces of length \(ds\), let each contribute mass \(f \cdot ds\), and sum:
The same integral prices a curved pipe by cost per metre or totals exposure along a road. The \(ds\) is exactly the arc-length element from the arc-length topic: on a graph \(y = g(x)\), \(ds = \sqrt{1 + (g'(x))^2}\,dx\). Only the values of \(f\) on \(C\) itself matter — the field of values off the curve is irrelevant.
Computing with a parametrisation
To compute \(\int_C f\,ds\), parametrise \(C\) as \(\mathbf{r}(t) = \langle x(t), y(t)\rangle\), \(a \le t \le b\), the way the vector-valued functions topic taught. Since \(|\mathbf{r}'(t)|\) is speed, the arc length travelled in time \(dt\) is \(ds = |\mathbf{r}'(t)|\,dt\), and the line integral becomes a plain one:
For a graph, \(\mathbf{r}(t) = \langle t, g(t)\rangle\) gives \(|\mathbf{r}'| = \sqrt{1 + (g')^2}\), reproducing the graph formula. A piecewise-smooth curve — a rectangle, say — is handled piecewise: one integral per smooth side, then add the results.
Worked example: a wire's mass
A wire lies along the parabola \(y = x^2\) from \((0, 0)\) to \((1, 1)\) with density \(f(x, y) = 3x\) mass per unit length. Parametrise \(\mathbf{r}(t) = \langle t, t^2\rangle\) for \(0 \le t \le 1\): \(\mathbf{r}'(t) = \langle 1, 2t\rangle\), \(|\mathbf{r}'| = \sqrt{1 + 4t^2}\), and the density on the curve is \(3t\), so
Substitute \(w = 1 + 4t^2\), \(dw = 8t\,dt\): the integral is \(\tfrac{3}{8}\int_1^5 \sqrt{w}\,dw = \tfrac{3}{8} \cdot \tfrac{2}{3}\big[w^{3/2}\big]_1^5 = \tfrac{1}{4}\big(5\sqrt{5} - 1\big) \approx 2.55\). A numerical check: Simpson's rule with nodes \(0, \tfrac{1}{2}, 1\) estimates the integral at \(\tfrac{1}{6}(0 + 4 \cdot 1.5\sqrt{2} + 3\sqrt{5}) \approx 2.53\), within 0.02 of the substitution's answer.
Line integrals of vector fields
A force field \(\mathbf{F}\) does work along a path: with each small step \(d\mathbf{r}\), the field contributes \(\mathbf{F} \cdot d\mathbf{r}\). The total work is
Written in components this is \(\int_C M\,dx + N\,dy\), plus \(P\,dz\) in space — the same integral, its notation borrowed from differentials. The dot product carries the physics: \(\mathbf{F} \cdot \mathbf{r}'\) is the forward push per unit time, positive when the field helps the motion, negative when it resists. Unlike the scalar line integral, this one is oriented — walk the path backwards and every \(d\mathbf{r}\) flips sign, negating the integral.
Worked example: a swirl's work
Compute the work done by \(\mathbf{F} = \langle -y, x\rangle\) on a particle making one counter-clockwise lap of the unit circle. Parametrise \(\mathbf{r}(t) = \langle\cos t, \sin t\rangle\), \(0 \le t \le 2\pi\): then \(\mathbf{r}'(t) = \langle-\sin t, \cos t\rangle\) and \(\mathbf{F}(\mathbf{r}(t)) = \langle-\sin t, \cos t\rangle\), so the dot product is \(\sin^2 t + \cos^2 t = 1\) and \(W = \int_0^{2\pi} 1\,dt = 2\pi\). Second route, geometry: on the unit circle \(\mathbf{F}\) is exactly the unit tangent vector to the circle — the field pushes forward with magnitude 1 at every point — so the work is magnitude 1 times path length \(2\pi\). No components were ever dotted, and the answer matches.
Two kinds of line integral
The two forms must not be confused. \(\int_C f\,ds\) is scalar: \(f\) is a number per point and \(ds\) is length, so the answer does not care which way \(C\) is walked — a wire weighs the same in either direction. \(\int_C M\,dx + N\,dy\) is vector: \(dx\) and \(dy\) are signed displacements that flip with orientation, and reversing \(C\) negates the whole integral. The hybrid \(\int_C M\,dx\) belongs to the vector family — it is \(\int_C \langle M, 0\rangle \cdot d\mathbf{r}\). Two anchors keep them straight: \(\int_C 1\,ds\) is the length of \(C\), while \(\int_C 1\,dx\) is \(x\) at the end minus \(x\) at the start, which is zero on any closed loop.
A preview of path independence
Compute the work of \(\mathbf{F} = \langle y, x\rangle\) from \((0, 0)\) to \((1, 1)\) along two different paths. Straight diagonal, \(\mathbf{r}(t) = \langle t, t\rangle\): \(\mathbf{F} = \langle t, t\rangle\), \(\mathbf{r}' = \langle 1, 1\rangle\), and \(W = \int_0^1 2t\,dt = 1\). L-shaped path: along the x-axis \(\mathbf{r}(t) = \langle t, 0\rangle\) where \(\mathbf{F} \cdot \mathbf{r}' = 0\), then up the line \(x = 1\), \(\mathbf{r}(t) = \langle 1, t\rangle\), where \(\mathbf{F} \cdot \mathbf{r}' = 1\), giving \(0 + 1 = 1\). Both paths give 1, and 1 is no coincidence: \(\mathbf{F} = \nabla(xy)\), so each route is secretly computing \(f(1, 1) - f(0, 0) = 1\). The conservative-fields topic proves this shortcut; the swirl field above, by contrast, delivers \(2\pi\) per lap of a closed circle — the two kinds of fields behave differently, and the difference is the next topic's whole subject.
Common mistakes to avoid
First, forgetting \(|\mathbf{r}'(t)|\) in a \(ds\) integral — the most common error in the chapter; without it you measured \(f\) along a phantom straight path. Second, wrong limits: the limits on \(t\) are the parameter's range, not the \(x\)-range — for \(\mathbf{r}(t) = \langle\cos t, \sin t\rangle\) they are 0 to \(2\pi\), not 0 to 1. Third, \(ds\) versus \(dx\): \(\int_C f\,dx\) uses \(dx = x'(t)\,dt\), which is signed and vanishes on vertical segments; only \(\int_C f\,ds\) carries the square root. Fourth, reversing orientation: recompute with a genuinely reversed parametrisation rather than assuming a minus sign lands somewhere convenient. Fifth, units: density times length is mass, force times distance is work — the answer's units should match the question's.
Key takeaways
The scalar line integral \(\int_C f\,ds\) adds \(f\) per unit length along a curve; compute it as \(\int_a^b f(\mathbf{r}(t))\,|\mathbf{r}'(t)|\,dt\) with any parametrisation, summing piecewise across corners. The work integral \(\int_C \mathbf{F} \cdot d\mathbf{r} = \int_a^b \mathbf{F}(\mathbf{r}(t)) \cdot \mathbf{r}'(t)\,dt = \int_C M\,dx + N\,dy\) weighs the field's forward push along the motion; reversing \(C\) negates it, while scalar integrals never notice orientation. The swirl \(\langle -y, x\rangle\) delivers \(2\pi\) per lap of the unit circle, whereas the gradient field \(\langle y, x\rangle = \nabla(xy)\) delivers endpoint-only work — the fork that the conservative-fields topic formalises.
Flashcards
Flashcard
Add f times arc length along C
Test your understanding
Question 1 of 5
A wire lies along y = x² from (0,0) to (1,1) with density f(x,y) = 3x per unit length. Which integral gives its mass?
Question 2 of 5
What is the work done by F = ⟨−y, x⟩ on a particle making one counter-clockwise lap of the unit circle?
Question 3 of 5
If the orientation of C is reversed, what happens to ∫_C F · dr?
Question 4 of 5
For the parametrisation r(t) = ⟨t, t²⟩ of the parabola y = x², what is |r′(t)|?
Question 5 of 5
Which line integral does not depend on the orientation of C?
Standards: Beyond AP
