Calculus › Section P · Vector analysis › Topic P.3

Conservative fields and path independence

Some force fields keep a promise: the work between two points is the same on every route. A potential function is the promise in writing, and a one-line test tells you when to expect it.

The fundamental theorem for line integrals

The single-variable fundamental theorem says \(\int_a^b f'(x)\,dx = f(b) - f(a)\): integrating a derivative returns endpoint values. Its line-integral sibling does the same along curves: if \(\mathbf{F} = \nabla f\) and \(C\) runs from \(A\) to \(B\), then

\int_C \mathbf{F} \cdot d\mathbf{r} = f(B) - f(A)

The proof is one line of the several-variable chain rule: \(\frac{d}{dt} f(\mathbf{r}(t)) = \nabla f(\mathbf{r}(t)) \cdot \mathbf{r}'(t) = \mathbf{F}(\mathbf{r}(t)) \cdot \mathbf{r}'(t)\), so the integral of the right side over \(a \le t \le b\) telescopes to \(f(\mathbf{r}(b)) - f(\mathbf{r}(a))\). The parametrisation of \(C\) never mattered — only its endpoints did. Work computed, curve discarded.

Path independence and closed loops

The theorem makes conservative fields predictable: work depends only on where you start and end, never on the route — path independence. On an open connected region, three statements become equivalent. \(\mathbf{F}\) is conservative (has a potential); work is path-independent; and the work around every closed loop is zero, since returning to the start makes \(f(B) - f(A) = 0\). Each implies the others: zero loop-work gives path independence, because any two routes from \(A\) to \(B\) form a closed loop that cancels their difference; and path independence builds a potential, by defining \(f(P)\) as the work done from a fixed base point to \(P\) — a definition that only makes sense if every route agrees on the answer.

Testing for a potential

Before hunting a potential, test whether one exists. In two dimensions, \(\mathbf{F} = \langle M, N\rangle\) has a potential only if the potential's mixed partials agree: since \(f_{xy} = f_{yx}\), a conservative field must satisfy \(\frac{\partial M}{\partial y} = \frac{\partial N}{\partial x}\). In three dimensions the same demand reads \(\operatorname{curl}\mathbf{F} = \mathbf{0}\), the identity from the vector-fields topic. Both tests carry fine print about the region: it must be simply connected — every loop in it can shrink to a point without leaving the region, so no punctures or slits. On a plane with the origin removed, a field can pass the test and still fail to be conservative; the hole card below shows exactly how.

Worked example: hunting a potential

Is \(\mathbf{F} = \langle 2xy, x^2 + 2y\rangle\) conservative, and if so, what is its potential? Test first: \(\frac{\partial(2xy)}{\partial y} = 2x\) and \(\frac{\partial(x^2 + 2y)}{\partial x} = 2x\) — equal, and the whole plane is hole-free, so a potential exists. Hunt it: \(f_x = 2xy\) forces \(f = x^2y + g(y)\) for some leftover function \(g\). Then \(f_y = x^2 + g'(y)\) must equal \(N = x^2 + 2y\), so \(g'(y) = 2y\) and \(g(y) = y^2\). The potential is \(f = x^2y + y^2 + C\). Audit by differentiating: \(\nabla f = \langle 2xy, x^2 + 2y\rangle\), exactly \(\mathbf{F}\) — the constant \(C\) vanishes, as constants do under differentiation.

Worked example: the payoff

Compute the work of \(\mathbf{F} = \langle 2xy, x^2 + 2y\rangle\) from \((0, 0)\) to \((2, 1)\). With the potential \(f = x^2y + y^2\) in hand, the fundamental theorem finishes it:

W = f(2, 1) - f(0, 0) = (4 + 1) - 0 = 5

Second route, to keep the theorem honest: parametrise the straight line \(\mathbf{r}(t) = \langle 2t, t\rangle\). There \(\mathbf{F} = \langle 4t^2, 4t^2 + 2t\rangle\) and \(\mathbf{r}' = \langle 2, 1\rangle\), so \(\mathbf{F} \cdot \mathbf{r}' = 8t^2 + 4t^2 + 2t = 12t^2 + 2t\) and \(W = \int_0^1 (12t^2 + 2t)\,dt = 4 + 1 = 5\). Same answer — but note which route needed a parametrisation, a dot product and an integral, and which needed one subtraction.

Worked example: three variables

In three dimensions the test runs through curl. Take \(\mathbf{F} = \langle y, x, z\rangle\): the curl is \(\langle \frac{\partial z}{\partial y} - \frac{\partial x}{\partial z}, \frac{\partial y}{\partial z} - \frac{\partial z}{\partial x}, \frac{\partial x}{\partial x} - \frac{\partial y}{\partial y}\rangle = \langle 0, 0, 0\rangle\), so a potential exists. Hunt it one variable at a time: \(f_x = y\) gives \(f = xy + g(y, z)\); then \(f_y = x + g_y\) must equal \(x\), forcing \(g_y = 0\), so \(g\) is really \(h(z)\); finally \(f_z = h'(z) = z\) gives \(h = \tfrac{z^2}{2}\). The potential is \(f = xy + \tfrac{z^2}{2} + C\). Each variable knocks the remaining freedom down a dimension until only the constant survives.

The hole in the fine print

Here is why 'simply connected' earns its place in the theorem. On the plane punctured at the origin, take \(\mathbf{F} = \langle\frac{-y}{x^2 + y^2}, \frac{x}{x^2 + y^2}\rangle\). It passes the cross-partial test: a direct computation gives \(\frac{\partial M}{\partial y} = \frac{\partial N}{\partial x} = \frac{y^2 - x^2}{(x^2 + y^2)^2}\) everywhere except the origin. Yet its work around the unit circle is \(2\pi\): on that circle \(x^2 + y^2 = 1\), so the field is exactly the swirl from the line-integrals topic, worth \(2\pi\) per lap. The resolution: the would-be potential is the polar angle \(\theta\), since \(\nabla\theta = \mathbf{F}\) — but \(\theta\) is multivalued, gaining \(2\pi\) with every lap, and a multivalued function cannot serve as a potential. Locally conservative everywhere, globally conservative nowhere: remove the hole and the paradox dissolves.

Common mistakes to avoid

First, skipping the test: integrating \(M\) and matching is wasted labour if no potential exists — check \(\frac{\partial M}{\partial y} = \frac{\partial N}{\partial x}\) before hunting. Second, forgetting the region: the test certifies a potential only on hole-free regions; the punctured-plane field passes the test and fails the conclusion. Third, stopping mid-hunt: after \(f = x^2y + g(y)\), the leftover \(g\) must be pinned down by \(f_y = N\); silently setting \(g = 0\) invents a constraint the field never made. Fourth, sign errors in the payoff: the theorem gives \(f(\text{end}) - f(\text{start})\), so reversing the direction reverses that difference, consistent with the orientation rule from the line-integrals topic. Fifth, hunting 'the' constant: any \(C\) works, and two correct answers that differ by a constant are equally correct.

Key takeaways

For a conservative field \(\mathbf{F} = \nabla f\), the fundamental theorem for line integrals gives \(\int_C \mathbf{F} \cdot d\mathbf{r} = f(B) - f(A)\): work depends only on endpoints, closed loops cost nothing, and no parametrisation is ever needed. Test before trusting: in two dimensions require \(\frac{\partial M}{\partial y} = \frac{\partial N}{\partial x}\), in three require \(\operatorname{curl}\mathbf{F} = \mathbf{0}\), both on a simply connected region — the punctured-plane field \(\langle\frac{-y}{x^2+y^2}, \frac{x}{x^2+y^2}\rangle\) shows the fine print is real, its 'potential' being the multivalued polar angle. Hunt potentials by integrating one component and matching the rest, as with \(f = x^2y + y^2\) for \(\langle 2xy, x^2 + 2y\rangle\). The first move on any work integral is always the test — one subtraction can replace a page of parametrisation.

Flashcards

Flashcard

Fundamental theorem for line integrals

∫_C ∇f · dr = f(end) − f(start)

The single-variable FTC's younger sibling.

Test your understanding

Question 1 of 5

For the conservative field F = ⟨2xy, x² + 2y⟩ with potential f = x²y + y², what is ∫_C F · dr from (0,0) to (2,1)?

Question 2 of 5

Which of these fields is conservative on the whole plane?

Question 3 of 5

A potential function for F = ⟨2xy, x² + 2y⟩ is...

Question 4 of 5

Which potential works for the 3D field F = ⟨y, x, z⟩?

Question 5 of 5

For a conservative field on a connected region, the work around any closed loop is always...

Standards: Beyond AP

← Previous topicNext topic →