Conservative Fields and Potentials

Suppose you push a particle from point AA to point BB through a force field. You slide it along two different paths. Does the work done depend on which path you chose? For some fields the answer is a resounding no. The work depends only on the endpoints. Such fields are called conservative. This single question can be asked of gravitational fields, electric fields, and countless other physical systems. It turns out to have a remarkably clean analytic answer. That answer is phrased entirely in terms of the Line Integrals machinery already at hand. The idea matters because it converts a hard problem into an easy one. Integrating a vector field along an arbitrary curve becomes evaluating a scalar function at two points. When a field is conservative, all of its integral information collapses into a single scalar "height function." This is called the potential. Computing work becomes subtraction, exactly as in single-variable calculus. There, ∫abf′(x) dx=f(b)−f(a)\int_a^b f'(x)\,dx = f(b)-f(a). This article builds the precise bridge between three notions. These are path independence of line integrals, the existence of a potential function (gradient fields), and a local differential test (the curl-free, or "mixed partials," condition). The big picture is a chain of equivalences. These are tightened by increasingly strong hypotheses on the underlying domain. On any open connected set, path independence and the existence of a potential are equivalent. This is the multivariable Fundamental Theorem for Line Integrals. The curl-free condition is always a necessary consequence of having a potential. This follows by equality of mixed partial derivatives. But it is only sufficient to guarantee a potential when the domain has no "holes." This condition is called simple connectivity. Getting this hierarchy exactly right is the heart of the subject. So is knowing precisely where each implication can fail. This article delivers the full logical structure. It gives a rigorous definition of conservative fields and potentials. It provides a complete proof of the fundamental theorem for line integrals. It offers a careful statement and proof of the curl-free necessary condition. It also gives an explicit counterexample. This is the classic "angle field" on the punctured plane. This example shows why simple connectivity cannot be dropped. It closes by pointing toward Green's Theorem and Stokes' Theorem. These supply the topological machinery that finally makes the curl-free condition sufficient.

Key takeaways
  • A vector field F\mathbf{F} is conservative on an open set if ∫CF⋅dr\int_C \mathbf{F}\cdot d\mathbf{r} depends only on the endpoints of CC, for every path CC in that set.
  • Equivalently (on open connected domains), F=∇f\mathbf{F}=\nabla f for some scalar potential function ff.
  • The Fundamental Theorem for Line Integrals says ∫C∇f⋅dr=f(B)−f(A)\int_C \nabla f \cdot d\mathbf{r} = f(B)-f(A), the multivariable analogue of the single-variable FTC.
  • If F=∇f\mathbf{F}=\nabla f with ff twice continuously differentiable, then F\mathbf{F} is curl-free (∇×F=0\nabla\times\mathbf{F}=\mathbf{0}, or ∂Q/∂x=∂P/∂y\partial Q/\partial x=\partial P/\partial y in the plane) — this follows from equality of mixed partials.
  • The converse of curl-free ⇒\Rightarrow conservative requires the domain to be simply connected; the punctured plane is the standard counterexample.
  • Key idea: conservative fields are exactly gradient fields, and testing "curl = 0" is a cheap local diagnostic whose global validity depends entirely on the shape of the domain.

Where this lives today

Conservative vector fields and potentials are the mathematical backbone of classical mechanics: gravitational and electrostatic force fields are conservative (away from point masses/charges), which is precisely why physicists can define potential energy and invoke conservation of total mechanical energy, a direct restatement of the fundamental theorem for line integrals. In electromagnetism, the electric field is conservative only in the static case (∇×E=−∂B/∂t=0\nabla \times \mathbf{E} = -\partial \mathbf{B}/\partial t = 0), and the scalar electric potential exists exactly because of this; when magnetic fields change in time, the curl of E\mathbf{E} becomes nonzero and no scalar potential exists, forcing the introduction of vector potentials, a direct physical instance of the curl-free/conservative distinction. In thermodynamics, state functions like entropy and internal energy are potentials for exact differentials, and the question of whether an infinitesimal quantity like heat is "exact" (a gradient) versus merely a path-dependent differential is exactly this article's dichotomy in disguise. In topology and geometry, the punctured-plane counterexample is the seed of de Rham cohomology: the failure of closed forms to be exact measures the "holes" in a manifold, with the angle field representing a generator of the first cohomology group of the punctured plane. In numerical optimization and machine learning, recognizing that a force/gradient field is conservative justifies working with a single scalar loss function rather than tracking a vector field of updates, and non-conservative "curl" components in learned dynamics are used diagnostically in physics-informed and Hamiltonian neural networks. In complex analysis, the sufficiency direction (curl-free plus simply connected implies conservative) is the real-variable shadow of Cauchy's theorem for holomorphic functions on simply connected domains.

Ask the tutor: Can you show why Green's Theorem is exactly the missing ingredient that upgrades "curl-free" to "conservative" on simply connected domains?

DefinitionConservative field, potential function, path independence

Let D⊆RnD \subseteq \mathbb{R}^n (n=2n = 2 or 33 for concreteness) be an open connected set, and let F:D→Rn\mathbf{F}: D \to \mathbb{R}^n be a continuous vector field. 1. Path independence. We say the line integral of F\mathbf{F} is path-independent in DD if for every pair of points A,B∈DA, B \in D and every two piecewise-C1C^1 paths C1,C2C_1, C_2 lying in DD from AA to BB,

∫C1F⋅dr=∫C2F⋅dr.\int_{C_1} \mathbf{F} \cdot d\mathbf{r} = \int_{C_2} \mathbf{F} \cdot d\mathbf{r}.

2. Conservative field. F\mathbf{F} is called conservative on DD if its line integral is path-independent in DD, equivalently (as shown below) if ∫CF⋅dr=0\int_C \mathbf{F}\cdot d\mathbf{r} = 0 for every piecewise-C1C^1 closed curve CC in DD. 3. Potential function. A scalar function f:D→Rf: D \to \mathbb{R} is a potential for F\mathbf{F} if ff is differentiable on DD and

∇f=Fon D.\nabla f = \mathbf{F} \quad \text{on } D.

4. Curl-free condition. If F=(P,Q)\mathbf{F} = (P, Q) on a planar domain with P,Q∈C1P, Q \in C^1, F\mathbf{F} is curl-free if

∂Q∂x=∂P∂yon D.\frac{\partial Q}{\partial x} = \frac{\partial P}{\partial y} \quad \text{on } D.

In space, with F=(P,Q,R)∈C1\mathbf{F} = (P,Q,R) \in C^1, curl-free means ∇×F=0\nabla \times \mathbf{F} = \mathbf{0}, i.e.

∂R∂y=∂Q∂z,∂P∂z=∂R∂x,∂Q∂x=∂P∂y.\frac{\partial R}{\partial y} = \frac{\partial Q}{\partial z}, \qquad \frac{\partial P}{\partial z} = \frac{\partial R}{\partial x}, \qquad \frac{\partial Q}{\partial x} = \frac{\partial P}{\partial y}.

Intuition

Think of a scalar function ff as an elevation map, its graph a mountainous terrain, and F=∇f\mathbf{F} = \nabla f as the field of steepest-ascent vectors sitting over the map. Walking along any trail from a trailhead AA to a summit BB, the total "uphill work" you do, integrated infinitesimally, is exactly the net change in elevation f(B)−f(A)f(B) - f(A) — utterly indifferent to how winding the trail was, how many switchbacks it took, or whether you detoured through a valley. That is path independence made visual: the integral only sees the start and end altitudes. Conversely, if a field is not a gradient — if it has some rotational, swirling component — then walking around a loop can net nonzero work, because the field keeps "pushing along" the direction of travel in a way no elevation function could encode (elevation always returns to its starting value after a loop). The curl-free condition is the local, calculus-based fingerprint of "no swirling": it says the field's rotation vanishes at every point. But a field can be locally swirl-free and still fail to be a gradient, if the domain itself has a hole that the loop wraps around — the field's rotation, integrated around that hole, need not be zero even though no single point carries any curl. This is exactly why the full converse needs simple connectivity, previewing the role of Green's Theorem in relating a boundary integral to circulation trapped inside a region.

Figure
Elevation surface with gradient vectors: work along a path equals the net rise f(B)−f(A)f(B)-f(A), independent of the route
f = 3 (summit B) f = 0 (base A) A B ∫ F·dr along either path = f(B) − f(A) = 3
TheoremFundamental Theorem for Line Integrals

Let D⊆RnD \subseteq \mathbb{R}^n be open and connected, and let f:D→Rf: D \to \mathbb{R} be differentiable with ∇f\nabla f continuous on DD. Let CC be a piecewise-C1C^1 curve in DD parametrized by r(t)\mathbf{r}(t) for t∈[a,b]t \in [a,b], with r(a)=A\mathbf{r}(a) = A and r(b)=B\mathbf{r}(b) = B. Then

∫C∇f⋅dr=f(B)−f(A).\int_C \nabla f \cdot d\mathbf{r} = f(B) - f(A).

Consequently: (i) The line integral of F=∇f\mathbf{F} = \nabla f is path-independent in DD: it depends only on the endpoints A,BA, B, not on the particular curve joining them. (ii) ∫C∇f⋅dr=0\int_C \nabla f \cdot d\mathbf{r} = 0 for every piecewise-C1C^1 closed curve CC in DD.

Converse. Conversely, if F\mathbf{F} is continuous on open connected DD and its line integral is path-independent in DD (equivalently, vanishes on every closed loop in DD), then there exists a differentiable function f:D→Rf: D \to \mathbb{R} with ∇f=F\nabla f = \mathbf{F} on DD; that is, F\mathbf{F} has a potential.

Proof

Forward direction (fundamental theorem). We first prove the identity for a C1C^1 curve, then extend to piecewise-C1C^1 by additivity. Suppose r:[a,b]→D\mathbf{r}: [a,b] \to D is C1C^1. Define g:[a,b]→Rg: [a,b] \to \mathbb{R} by

g(t)=f(r(t)).g(t) = f(\mathbf{r}(t)).

Since ff is differentiable on DD with continuous gradient, and r\mathbf{r} is C1C^1, the The Chain Rule in Several Variables gives that gg is differentiable on (a,b)(a,b), continuous on [a,b][a,b], with

g′(t)=∇f(r(t))⋅r′(t).g'(t) = \nabla f(\mathbf{r}(t)) \cdot \mathbf{r}'(t).

The right-hand side is continuous in tt because ∇f\nabla f is continuous, r\mathbf{r} is C1C^1, and dot product and composition of continuous maps are continuous. Hence gg is continuously differentiable on [a,b][a,b] (one-sided derivatives at the endpoints agree with the one-sided limits of g′g', by the same chain rule argument applied on half-open intervals), and the single-variable Fundamental Theorem of Calculus applies:

g(b)−g(a)=∫abg′(t) dt.g(b) - g(a) = \int_a^b g'(t)\, dt.

By definition of the line integral of a vector field along a parametrized curve,

∫C∇f⋅dr:=∫ab∇f(r(t))⋅r′(t) dt=∫abg′(t) dt=g(b)−g(a)=f(r(b))−f(r(a))=f(B)−f(A).\int_C \nabla f \cdot d\mathbf{r} := \int_a^b \nabla f(\mathbf{r}(t)) \cdot \mathbf{r}'(t)\, dt = \int_a^b g'(t)\,dt = g(b) - g(a) = f(\mathbf{r}(b)) - f(\mathbf{r}(a)) = f(B) - f(A).

Now suppose CC is only piecewise-C1C^1, i.e. there is a partition a=t0<t1<⋯<tk=ba = t_0 < t_1 < \cdots < t_k = b such that r\mathbf{r} restricted to each [ti−1,ti][t_{i-1}, t_i] is C1C^1. By definition, the line integral over CC is the sum of the line integrals over each smooth piece CiC_i. Applying the C1C^1 case to each piece,

∫C∇f⋅dr=∑i=1k∫Ci∇f⋅dr=∑i=1k[f(r(ti))−f(r(ti−1))].\int_C \nabla f \cdot d\mathbf{r} = \sum_{i=1}^k \int_{C_i} \nabla f \cdot d\mathbf{r} = \sum_{i=1}^k \big[f(\mathbf{r}(t_i)) - f(\mathbf{r}(t_{i-1}))\big].

This is a telescoping sum: every intermediate value f(r(ti))f(\mathbf{r}(t_i)) for 0<i<k0 < i < k appears once with a plus sign (from piece ii) and once with a minus sign (from piece i+1i+1), so it cancels. What remains is

∫C∇f⋅dr=f(r(tk))−f(r(t0))=f(B)−f(A).\int_C \nabla f \cdot d\mathbf{r} = f(\mathbf{r}(t_k)) - f(\mathbf{r}(t_0)) = f(B) - f(A).

This proves the main identity.

Part (i) is immediate: the right-hand side f(B)−f(A)f(B) - f(A) depends only on the endpoints AA and BB, not on the path CC chosen between them, since it makes no reference to CC at all.

Part (ii) follows by taking A=BA = B (a closed curve has r(a)=r(b)\mathbf{r}(a) = \mathbf{r}(b)), giving f(B)−f(A)=f(A)−f(A)=0f(B) - f(A) = f(A) - f(A) = 0.

Converse direction. Assume F\mathbf{F} is continuous on open connected DD and path-independent there. We construct an explicit potential. Fix any base point A0∈DA_0 \in D. Since DD is open and connected in Rn\mathbb{R}^n, it is polygonally connected: any two points of DD can be joined by a finite chain of straight segments lying entirely in DD. (This is a standard fact: the set of points reachable from A0A_0 by a polygonal path in DD is open, since DD is open, one can always extend a polygonal path a little further inside a ball; its complement in DD is also open by the same argument; since DD is connected and this reachable set is nonempty, it must be all of DD.) For each X∈DX \in D, choose any piecewise-linear path CXC_X in DD from A0A_0 to XX, and define

f(X):=∫CXF⋅dr.f(X) := \int_{C_X} \mathbf{F} \cdot d\mathbf{r}.

This is well-defined: by path independence, the value does not depend on the choice of CXC_X, since any two such paths from A0A_0 to XX inside DD give equal integrals by hypothesis. We now show ∇f=F\nabla f = \mathbf{F} on DD, i.e. that all partial derivatives of ff exist and match the components of F\mathbf{F}. Fix X∈DX \in D and write F=(F1,…,Fn)\mathbf{F} = (F_1, \dots, F_n). Since DD is open, there is r>0r > 0 with the open ball B(X,r)⊆DB(X, r) \subseteq D. For ∣h∣<r|h| < r, let ej\mathbf{e}_j be the jj-th standard basis vector, and build a path from A0A_0 to X+hejX + h\mathbf{e}_j by first following CXC_X from A0A_0 to XX, then appending the straight segment ShS_h from XX to X+hejX + h\mathbf{e}_j (this segment lies in DD because it lies in B(X,r)⊆DB(X,r) \subseteq D). By path independence (using this concatenated path as our chosen path for the point X+hejX + h\mathbf{e}_j),

f(X+hej)=∫CXF⋅dr+∫ShF⋅dr=f(X)+∫ShF⋅dr.f(X + h\mathbf{e}_j) = \int_{C_X} \mathbf{F}\cdot d\mathbf{r} + \int_{S_h} \mathbf{F} \cdot d\mathbf{r} = f(X) + \int_{S_h} \mathbf{F} \cdot d\mathbf{r}.

Parametrize ShS_h by r(t)=X+th ej\mathbf{r}(t) = X + t h\, \mathbf{e}_j, t∈[0,1]t \in [0,1], so r′(t)=hej\mathbf{r}'(t) = h\mathbf{e}_j and

∫ShF⋅dr=∫01F(X+thej)⋅hej dt=h∫01Fj(X+thej) dt.\int_{S_h} \mathbf{F}\cdot d\mathbf{r} = \int_0^1 \mathbf{F}(X + th\mathbf{e}_j) \cdot h\mathbf{e}_j \, dt = h \int_0^1 F_j(X + th\mathbf{e}_j)\, dt.

Therefore

f(X+hej)−f(X)h=∫01Fj(X+thej) dt.\frac{f(X + h\mathbf{e}_j) - f(X)}{h} = \int_0^1 F_j(X + th\mathbf{e}_j)\, dt.

As h→0h \to 0, the integrand Fj(X+thej)→Fj(X)F_j(X + th\mathbf{e}_j) \to F_j(X) uniformly for t∈[0,1]t \in [0,1], because FjF_j is continuous at XX: given ε>0\varepsilon > 0, continuity gives δ>0\delta > 0 such that ∣Fj(Y)−Fj(X)∣<ε|F_j(Y) - F_j(X)| < \varepsilon whenever ∣Y−X∣<δ|Y - X| < \delta, and for ∣h∣<δ|h| < \delta we have ∣thej∣=t∣h∣≤∣h∣<δ|th\mathbf{e}_j| = t|h| \le |h| < \delta for all t∈[0,1]t \in [0,1], so

∣∫01Fj(X+thej) dt−Fj(X)∣=∣∫01(Fj(X+thej)−Fj(X)) dt∣≤ε.\left| \int_0^1 F_j(X+th\mathbf{e}_j)\,dt - F_j(X) \right| = \left| \int_0^1 \big(F_j(X + th\mathbf{e}_j) - F_j(X)\big)\, dt \right| \le \varepsilon.

Hence

∂f∂xj(X)=lim⁡h→0f(X+hej)−f(X)h=Fj(X).\frac{\partial f}{\partial x_j}(X) = \lim_{h\to 0} \frac{f(X+h\mathbf{e}_j) - f(X)}{h} = F_j(X).

Since jj was arbitrary, all partial derivatives of ff exist at every X∈DX \in D and equal the corresponding components of F\mathbf{F}; since F\mathbf{F} is continuous, these partials are continuous, so f∈C1(D)f \in C^1(D) and in particular ff is differentiable with ∇f=F\nabla f = \mathbf{F} on DD. This exhibits the required potential and completes the proof. ■\blacksquare

∎
LemmaCurl-free is necessary for a potential

Let D⊆R2D \subseteq \mathbb{R}^2 (respectively R3\mathbb{R}^3) be open, and suppose f:D→Rf: D \to \mathbb{R} is C2C^2 (twice continuously differentiable) with ∇f=F=(P,Q)\nabla f = \mathbf{F} = (P,Q) (respectively (P,Q,R)(P,Q,R)). Then F\mathbf{F} is curl-free on DD.

Proof

This uses Clairaut's theorem / equality of mixed partial derivatives: if ff is C2C^2 on an open set, then ∂2f/∂x ∂y=∂2f/∂y ∂x\partial^2 f/\partial x\,\partial y = \partial^2 f/\partial y\,\partial x at every point of that set (a standard consequence of the mean value theorem applied twice to a difference quotient, given continuity of the second partials). In the plane, P=∂f/∂xP = \partial f/\partial x and Q=∂f/∂yQ = \partial f/\partial y. Then

∂Q∂x=∂∂x(∂f∂y)=∂2f∂x ∂y=∂2f∂y ∂x=∂∂y(∂f∂x)=∂P∂y,\frac{\partial Q}{\partial x} = \frac{\partial}{\partial x}\left(\frac{\partial f}{\partial y}\right) = \frac{\partial^2 f}{\partial x \, \partial y} = \frac{\partial^2 f}{\partial y\,\partial x} = \frac{\partial}{\partial y}\left(\frac{\partial f}{\partial x}\right) = \frac{\partial P}{\partial y},

which is exactly the curl-free condition ∂Q/∂x=∂P/∂y\partial Q/\partial x = \partial P/\partial y. In space, P=fx,Q=fy,R=fzP = f_x, Q = f_y, R = f_z, and applying Clairaut's theorem to each of the three pairs of variables gives Ry=fzy=fyz=QzR_y = f_{zy} = f_{yz} = Q_z, and similarly Pz=RxP_z = R_x and Qx=PyQ_x = P_y, which together say precisely ∇×F=0\nabla \times \mathbf{F} = \mathbf{0}. ■\blacksquare

∎

Worked examples

Example 1 (a genuine, asymmetric conservative field in the plane). Let D=R2D = \mathbb{R}^2 and consider

F(x,y)=(P,Q)=(2xy3+cos⁡x, 3x2y2−1).\mathbf{F}(x,y) = \big(P, Q\big) = \big(2xy^3 + \cos x,\ 3x^2y^2 - 1\big).

Step 1: check curl-free. We compute

∂Q∂x=6xy2,∂P∂y=6xy2.\frac{\partial Q}{\partial x} = 6xy^2, \qquad \frac{\partial P}{\partial y} = 6xy^2.

These agree, so F\mathbf{F} passes the necessary test — but since D=R2D = \mathbb{R}^2 has no holes (it is simply connected, as will be discussed), this is also sufficient, foreshadowing the role of Green's Theorem in proving such sufficiency rigorously. Step 2: construct a potential. We seek ff with fx=2xy3+cos⁡xf_x = 2xy^3 + \cos x. Integrating in xx (treating yy as constant):

f(x,y)=x2y3+sin⁡x+h(y)f(x,y) = x^2y^3 + \sin x + h(y)

for some function hh depending only on yy. Differentiate in yy:

fy=3x2y2+h′(y).f_y = 3x^2y^2 + h'(y).

We need this to equal Q=3x2y2−1Q = 3x^2y^2 - 1, so h′(y)=−1h'(y) = -1, giving h(y)=−y+Ch(y) = -y + C. Take C=0C=0:

f(x,y)=x2y3+sin⁡x−y.f(x,y) = x^2y^3 + \sin x - y.

Check: fx=2xy3+cos⁡x=Pf_x = 2xy^3 + \cos x = P ✓, and fy=3x2y2−1=Qf_y = 3x^2y^2 - 1 = Q ✓. Step 3: use the fundamental theorem to compute work. Suppose CC is any piecewise-C1C^1 path from A=(0,1)A = (0,1) to B=(1,2)B = (1, 2) — say a wild spiral, it doesn't matter. By the fundamental theorem for line integrals,

∫CF⋅dr=f(1,2)−f(0,1)=(12⋅23+sin⁡1−2)−(0+sin⁡0−1)=(8+sin⁡1−2)−(−1)=7+sin⁡1.\int_C \mathbf{F}\cdot d\mathbf{r} = f(1,2) - f(0,1) = \big(1^2\cdot 2^3 + \sin 1 - 2\big) - \big(0 + \sin 0 - 1\big) = (8 + \sin 1 - 2) - (-1) = 7 + \sin 1.

Numerically, sin⁡1≈0.8415\sin 1 \approx 0.8415, so the integral is approximately 7.84157.8415 — obtained without ever parametrizing a path.

Example 2 (using path independence to detect non-conservativeness, and to compute an integral by a clever detour). Let D=R2D = \mathbb{R}^2 and

F(x,y)=(P,Q)=(−y, x).\mathbf{F}(x,y) = (P,Q) = (-y,\ x).

Compute ∂Q/∂x=1\partial Q/\partial x = 1 and ∂P/∂y=−1\partial P/\partial y = -1. These are unequal, so F\mathbf{F} fails the curl-free test everywhere, hence by the lemma above (contrapositive) F\mathbf{F} admits no potential on any open subset of R2\mathbb{R}^2, and its line integral is not path-independent. We verify this directly and see the path-dependence explicitly. Let A=(1,0)A = (1,0), B=(−1,0)B = (-1, 0). Path C1C_1: the straight segment from AA to BB, parametrized r(t)=(1−2t,0)\mathbf{r}(t) = (1-2t, 0), t∈[0,1]t\in[0,1], so r′(t)=(−2,0)\mathbf{r}'(t) = (-2, 0). Then F(r(t))=(0,1−2t)\mathbf{F}(\mathbf{r}(t)) = (0, 1-2t), and

∫C1F⋅dr=∫01(0, 1−2t)⋅(−2,0) dt=∫010 dt=0.\int_{C_1} \mathbf{F}\cdot d\mathbf{r} = \int_0^1 (0,\,1-2t)\cdot(-2,0)\, dt = \int_0^1 0 \, dt = 0.

Path C2C_2: the upper semicircle from AA to BB, r(t)=(cos⁡t,sin⁡t)\mathbf{r}(t) = (\cos t, \sin t), t∈[0,π]t \in [0, \pi], so r′(t)=(−sin⁡t,cos⁡t)\mathbf{r}'(t) = (-\sin t, \cos t). Then F(r(t))=(−sin⁡t,cos⁡t)\mathbf{F}(\mathbf{r}(t)) = (-\sin t, \cos t), and

∫C2F⋅dr=∫0π[(−sin⁡t)(−sin⁡t)+(cos⁡t)(cos⁡t)] dt=∫0π(sin⁡2t+cos⁡2t) dt=∫0π1 dt=π.\int_{C_2} \mathbf{F}\cdot d\mathbf{r} = \int_0^\pi \big[(-\sin t)(-\sin t) + (\cos t)(\cos t)\big]\, dt = \int_0^\pi (\sin^2 t + \cos^2 t)\, dt = \int_0^\pi 1 \, dt = \pi.

Indeed 0≠π0 \ne \pi: two paths between the same endpoints give different integrals, confirming directly that F=(−y,x)\mathbf{F} = (-y,x) is not conservative, exactly as the curl test predicted. (This same field, restricted to going all the way around the full unit circle once counterclockwise, gives ∫F⋅dr=2π≠0\int \mathbf{F}\cdot d\mathbf{r} = 2\pi \ne 0, the signature of nonzero circulation — the topic taken up by Green's Theorem.)

CounterexampleWhy simple connectivity cannot be dropped from the converse of the curl-free lemma

The curl-free condition is necessary for a potential (proved above) whenever f∈C2f \in C^2. One might hope for the converse in general: curl-free ⇒\Rightarrow conservative. This is false if the domain DD is not simply connected (informally: has a hole that a loop can wrap around and not contract to a point inside DD). Let D=R2∖{(0,0)}D = \mathbb{R}^2 \setminus \{(0,0)\}, the punctured plane, and define the classic angle field

F(x,y)=(−yx2+y2, xx2+y2)=:(P,Q).\mathbf{F}(x,y) = \left(\frac{-y}{x^2+y^2},\ \frac{x}{x^2+y^2}\right) =: (P, Q).

DD is open (complement of a single closed point) but not simply connected: the unit circle centered at the origin cannot be continuously shrunk to a point without leaving DD, since the origin is missing.

Curl-free check. Direct computation (quotient rule) gives, on all of DD,

∂P∂y=−(x2+y2)+y(2y)(x2+y2)2=y2−x2(x2+y2)2,\frac{\partial P}{\partial y} = \frac{-(x^2+y^2) + y(2y)}{(x^2+y^2)^2} = \frac{y^2 - x^2}{(x^2+y^2)^2},
∂Q∂x=(x2+y2)−x(2x)(x2+y2)2=y2−x2(x2+y2)2.\frac{\partial Q}{\partial x} = \frac{(x^2+y^2) - x(2x)}{(x^2+y^2)^2} = \frac{y^2 - x^2}{(x^2+y^2)^2}.

These are equal at every point of DD, so F\mathbf{F} is curl-free on DD.

Yet F\mathbf{F} has no potential on DD, because its integral around the unit circle is nonzero. Parametrize the unit circle counterclockwise, r(t)=(cos⁡t,sin⁡t)\mathbf{r}(t) = (\cos t, \sin t), t∈[0,2π]t \in [0, 2\pi], entirely inside DD since it avoids the origin. Then x2+y2=1x^2+y^2 = 1 along this curve, r′(t)=(−sin⁡t,cos⁡t)\mathbf{r}'(t) = (-\sin t, \cos t), and F(r(t))=(−sin⁡t,cos⁡t)\mathbf{F}(\mathbf{r}(t)) = (-\sin t, \cos t), so

∮CF⋅dr=∫02π[(−sin⁡t)(−sin⁡t)+(cos⁡t)(cos⁡t)] dt=∫02π1 dt=2π≠0.\oint_{C} \mathbf{F} \cdot d\mathbf{r} = \int_0^{2\pi} \big[(-\sin t)(-\sin t) + (\cos t)(\cos t)\big]\, dt = \int_0^{2\pi} 1\, dt = 2\pi \ne 0.

If a potential ff existed on all of DD, part (ii) of the fundamental theorem would force this closed-loop integral to vanish. Since it does not, no potential for F\mathbf{F} exists on the full punctured plane DD, despite F\mathbf{F} being curl-free everywhere on DD. (A potential does exist locally: on the slit plane D′=R2∖{(x,0):x≤0}D' = \mathbb{R}^2 \setminus \{(x,0): x \le 0\}, which is simply connected, f(x,y)=θ(x,y)=f(x,y) = \theta(x,y) = the continuous branch of arg⁡(x+iy)∈(−π,π)\arg(x+iy) \in (-\pi, \pi) satisfies ∇f=F\nabla f = \mathbf{F}. This local potential just cannot be extended continuously all the way around the origin, since θ\theta must jump by 2π2\pi somewhere.) This shows precisely which hypothesis is essential: the implication "curl-free ⇒\Rightarrow conservative" needs the domain to be simply connected, and the failure is entirely topological, not a failure of the differential computation.

Scope, edge cases & common traps

SituationWhat holdsWhat can fail
DD open, connected, F=∇f\mathbf{F}=\nabla f, ff differentiable, ∇f\nabla f continuousFundamental theorem: ∫CF⋅dr=f(B)−f(A)\int_C \mathbf{F}\cdot d\mathbf{r}=f(B)-f(A), path independence, zero on loopsNothing — this direction always holds
F\mathbf{F} path-independent (or zero on all loops) on open connected DDA potential ff exists on DD (converse direction, proved above)The potential is unique only up to an additive constant on each connected component
f∈C2f \in C^2F=∇f\mathbf{F}=\nabla f is curl-free (Clairaut)If ff is only C1C^1 (mixed partials not known continuous), curl-free may fail to even make sense / Clairaut may not apply
F\mathbf{F} curl-free, DD simply connectedF\mathbf{F} is conservative on DD (sufficiency; proved via Green's Theorem / Stokes' Theorem, beyond this article's scope)—
F\mathbf{F} curl-free, DD not simply connectedNothing guaranteed globallyCounterexample above: angle field on punctured plane is curl-free but not conservative

The mathematician behind it

Origins · 1 minGeorge Green and how this was discovered
GG
George Green
1793–1841 · self-taught English miller and mathematician, working largely in isolation in Nottingham before a brief, late academic career at Cambridge

Green developed the mathematical theory that would eventually make precise exactly when a curl-free field is conservative, in his 1828 Essay on the Application of Mathematical Analysis to the Theories of Electricity and Magnetism. Working almost entirely self-taught, with only a single year of formal schooling, he introduced what we now call Green's functions and the theorem relating a line integral around a boundary to a double integral of a "curl-like" quantity over the enclosed region — the very tool needed to convert the local curl-free condition into a global statement about circulation. His work was largely unnoticed until after his death, when William Thomson (Lord Kelvin) rediscovered the essay and helped bring it to prominence.

History
- 1793 — George Green born near Nottingham, England, son of a baker/miller; almost entirely self-educated in mathematics.
- 1828 — Green privately publishes his Essay, introducing what becomes Green's theorem and the potential-theoretic framework linking curl and circulation.
- 1833 — At age 40, Green enrolls as an undergraduate at Cambridge, remarkably late for a career mathematician.
- 1846 — William Thomson (Lord Kelvin) rediscovers Green's essay, republishes and popularizes it, cementing its role in vector calculus and physics.
- Late 1800s — Poincaré, Stokes, and others generalize the curl/circulation relationship into what becomes Stokes' Theorem, completing the modern chain of equivalences discussed here.
What exactly is the difference between "conservative" and "curl-free"?

"Conservative" is a global statement: the line integral of F\mathbf{F} between any two points is path-independent, equivalently F\mathbf{F} has a potential function defined on the whole domain. "Curl-free" is a local, pointwise differential condition — a statement purely about partial derivatives at each point, checkable without reference to any path or loop. Every conservative C2C^2 field is curl-free (via equality of mixed partials), but the converse needs the domain to have no holes (simple connectivity). On simply connected domains the two notions coincide; on domains with holes, like the punctured plane, they can diverge, as the angle field example demonstrates concretely.

If a field is conservative, is its potential function unique?

No — only unique up to an additive constant on each connected component of the domain. If ∇f1=∇f2=F\nabla f_1 = \nabla f_2 = \mathbf{F} on a connected open set DD, then ∇(f1−f2)=0\nabla (f_1 - f_2) = \mathbf{0} on DD, so all partial derivatives of f1−f2f_1-f_2 vanish identically. A standard connectedness argument (using that DD is polygonally connected, and that a function with zero gradient is locally constant, hence constant along any path) shows f1−f2f_1 - f_2 is constant on all of DD. If DD has several connected components, the constant can differ from component to component.

How do you actually test whether a given field is conservative in practice?

First check the curl-free condition componentwise; if it fails at even one point, the field is definitely not conservative on any domain containing that point (by the necessity lemma). If curl-free holds everywhere on the domain, check whether the domain is simply connected (no holes, and in the plane, "no missing points/curves that a loop could encircle"). If both hold, the field is conservative (a fact proved via Green's Theorem), and you can construct a potential explicitly by successive partial integration, exactly as in Example 1. If the domain has a hole, you must additionally check directly whether the integral around a loop encircling the hole vanishes.

Why does the fundamental theorem for line integrals require ∇f\nabla f to be continuous, not just ff differentiable?

Continuity of ∇f\nabla f is what makes the composite function g(t)=f(r(t))g(t) = f(\mathbf{r}(t)) continuously differentiable, which is exactly the hypothesis needed to invoke the single-variable Fundamental Theorem of Calculus on gg. Without continuity of the derivative, g′g' could fail to be Riemann integrable, or the equality ∫abg′(t) dt=g(b)−g(a)\int_a^b g'(t)\,dt = g(b)-g(a) could fail even if gg is differentiable everywhere (there exist differentiable functions with unbounded, hence non-Riemann-integrable, derivatives). Requiring ∇f\nabla f continuous sidesteps these pathologies cleanly.

Is every gradient field automatically curl-free, with no extra hypothesis at all?

You need ff to be C2C^2, meaning its second partial derivatives exist and are continuous, for the curl-free identity to follow from Clairaut's theorem on equality of mixed partials. If ff is merely C1C^1 (first partials continuous but second partials not assumed to exist or be continuous), the argument for equality of mixed partials can break down, and pathological examples exist where mixed partials fail to agree at isolated points. In virtually all applied settings the potential is smooth enough that this subtlety never arises, but it is a genuine hypothesis worth stating.

How does this topic connect to Green's Theorem and Stokes' Theorem?

Green's Theorem states that for a curl-free field on a simply connected planar domain, circulation around any closed curve equals a double integral of (∂Q/∂x−∂P/∂y)(\partial Q/\partial x - \partial P/\partial y) over the enclosed region — which is identically zero precisely because the field is curl-free. That is exactly the missing sufficiency step this article flags as "beyond scope": it shows why, on a hole-free domain, curl-free forces every closed-loop integral to vanish, hence forces conservativeness. Stokes' Theorem generalizes this same idea to curves and surfaces in three-dimensional space, giving the three-dimensional criterion for when ∇×F=0\nabla\times\mathbf{F}=\mathbf{0} guarantees a potential.

Does the theory change if the domain is disconnected?

The definitions and the forward direction of the fundamental theorem still make sense on each connected component separately, but "path independence throughout DD" only meaningfully compares points within the same component, since there is no path in DD joining points in different components. The converse construction (building ff from a base point) must be carried out separately on each connected component, using a possibly different base point and additive constant on each; there is no way to compare potentials across components without extra information.

Before this

Where this leads