The Curl of a Vector Field

The curl is one of the two first-order differential operators (alongside the divergence) that act on a vector field to produce another field. Where the divergence extracts the part of the field's variation associated with sources and sinks (outflow), the curl extracts the part associated with rotation. Physically, if you imagine a vector field as the velocity field of a fluid, the curl at a point tells you how a tiny paddle wheel placed there would spin: the axis of the wheel aligns with $\nabla\times\mathbf{F}$ and the spin rate is proportional to its magnitude. Formally the curl is built from the del operator $\nabla$ via the cross product, $\nabla\times\mathbf{F}$ . Because the cross product yields a vector orthogonal to its factors, the curl is itself a vector field — but only because we live in three dimensions, where the cross product is defined. More fundamentally, $\nabla\times\mathbf{F}$ packages the antisymmetric combinations of partial derivatives $\partial_i F_j-\partial_j F_i$ into a vector through the Levi-Civita symbol. The operational meaning of curl is fixed by the curl (Stokes) theorem, which equates the flux of $\nabla\times\mathbf{F}$ through a surface with the circulation of $\mathbf{F}$ around the bounding loop; shrinking the loop to a point reveals the curl as a circulation density. This makes the curl the natural object on the right-hand side of Faraday's and Ampère's laws, where changing magnetic flux and electric current drive circulating fields. Two identities recur throughout: the curl of a gradient vanishes — underlying the electrostatic potential — and the divergence of a curl vanishes, guaranteeing that $\mathbf{B}=\nabla\times\mathbf{A}$ is automatically divergence-free.

Key takeaways

The curl $\nabla\times\mathbf{F}$ is a vector field that measures the local circulation density of $\mathbf{F}$ : its magnitude is the rate of swirl per unit area and its direction is the axis of rotation by the right-hand rule.
In Cartesian coordinates $\nabla\times\mathbf{F}=(\partial_y F_z-\partial_z F_y,\,\partial_z F_x-\partial_x F_z,\,\partial_x F_y-\partial_y F_x)$ , conveniently written as a symbolic determinant with $\nabla$ and the unit vectors.
Stokes' theorem ties curl to circulation: $\oint_{\partial S}\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell}=\int_S(\nabla\times\mathbf{F})\cdot\mathrm{d}\mathbf{a}$ , making the curl the circulation per unit area in the limit of a shrinking loop.
A field with $\nabla\times\mathbf{F}=\mathbf{0}$ everywhere (in a simply connected region) is irrotational and equals the gradient of a scalar potential; the curl of any gradient vanishes identically.
The curl is central to Maxwell's equations: $\nabla\times\mathbf{E}=-\partial_t\mathbf{B}$ (Faraday) and $\nabla\times\mathbf{B}=\mu_0\mathbf{J}+\mu_0\varepsilon_0\partial_t\mathbf{E}$ (Ampère–Maxwell).
Curl is a genuine vector only in 3D; it is the antisymmetric part of the derivative tensor $\partial_i F_j$ , and its divergence is identically zero.
Key idea: The curl is circulation per unit area: $(\nabla\times\mathbf{F})\cdot\hat{\mathbf{n}}=\lim_{S\to 0}\frac{1}{|S|}\oint_{\partial S}\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell}$ , a vector whose component along any direction $\hat{\mathbf{n}}$ is the swirl of $\mathbf{F}$ in the plane perpendicular to $\hat{\mathbf{n}}$ .

Where this lives today

Meteorology runs on curl. The vorticity of the wind field — $\nabla\times\mathbf{v}$ — is a primary forecast variable: hurricanes are vorticity maxima, and the spin-up of a storm is tracked as vorticity stretching and advection in every numerical weather model. Aerodynamics is the same mathematics at smaller scale: lift on a wing is bound circulation (Kutta–Joukowski), and the wingtip vortices that force aircraft spacing at airports are shed curl, conserved and long-lived precisely because vorticity obeys its own transport theorem. Space physics measures the operator directly. NASA's four MMS spacecraft fly in a tetrahedron through Earth's magnetosphere and compute the "curlometer" estimate: differences of $\mathbf{B}$ between the four points give $\nabla\times\mathbf{B}$ , hence the local current density via Ampère's law — a literal implementation of the shrinking-loop definition (4), with the loop replaced by a 10-km spacecraft formation. Fusion devices do the inverse daily: tokamak control systems reconstruct internal current profiles from magnetic measurements, solving curl equations in real time to keep the plasma stable.

The physical picture

Picture the vector field as a flowing fluid. Drop a tiny rigid paddle wheel into the flow and let it freely pivot. The curl at that point is twice the angular velocity vector the wheel settles into: it spins about the curl's axis at a rate set by the curl's magnitude. No spin means zero curl. The crucial — and counterintuitive — point is that the wheel responds to differences in the flow across its own small width, not to whether the streamlines look curved.

Figure

The paddle-wheel test — visual swirl is not curl and straight lines are not innocence

The curl measures the difference in the field across a loop, not its absolute swirl. Walk a small rectangle: the line integral $\oint\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell}$ accumulates net circulation only if the field is stronger on one side of the loop than the opposite side, or points differently along the two opposite edges. That net imbalance, divided by the area, is the curl component out of the loop. There is also a tidy accounting behind the divergence/curl division of labour. Decompose the derivative tensor $\partial_i F_j$ into symmetric and antisymmetric parts. The trace ( $\partial_i F_i$ ) is the divergence; the antisymmetric part $\tfrac12(\partial_i F_j-\partial_j F_i)$ has three independent components — exactly the curl. So divergence and curl together (plus a symmetric traceless shear) exhaust all first derivatives of a vector field. Curl points along the axis of swirl by the right-hand rule: curl your right fingers along the direction the field circulates and your thumb points along $\nabla\times\mathbf{F}$ . The E&M payoff arrives in two instalments: Faraday's law will say that a changing magnetic field drives a circulating electric field ( $\nabla\times\mathbf{E}=-\partial_t\mathbf{B}$ — the principle behind every generator), and Ampère's law will say that currents drive circulating magnetic fields ( $\nabla\times\mathbf{B}=\mu_0\mathbf{J}$ in magnetostatics). Whenever a field wraps around something, a curl equation is talking.

The formulas

In Cartesian coordinates the curl collects the three antisymmetric derivative differences. Each component measures swirl in the plane perpendicular to it — the $z$ -component, for instance, compares how $F_y$ grows along $x$ against how $F_x$ grows along $y$ :

\nabla\times\mathbf{F}=\left(\frac{\partial F_z}{\partial y}-\frac{\partial F_y}{\partial z}\right)\hat{\mathbf{x}}+\left(\frac{\partial F_x}{\partial z}-\frac{\partial F_z}{\partial x}\right)\hat{\mathbf{y}}+\left(\frac{\partial F_y}{\partial x}-\frac{\partial F_x}{\partial y}\right)\hat{\mathbf{z}} \qquad (1)

where: $\mathbf{F}=(F_x,F_y,F_z)$ — vector field whose curl is taken; $\partial_i$ — partial derivative with respect to coordinate $i$ ; $\hat{\mathbf{x}},\hat{\mathbf{y}},\hat{\mathbf{z}}$ — Cartesian unit vectors. Notice that only off-diagonal derivatives appear — the diagonal ones ( $\partial_x F_x$ etc.) all went to the divergence. The two operators split the field's first-derivative information without overlap. Nobody memorizes equation (1) directly; the working mnemonic is the symbolic determinant, exactly as for an ordinary cross product but with derivative operators in the second row:

\nabla\times\mathbf{F}=\begin{vmatrix}\hat{\mathbf{x}}&\hat{\mathbf{y}}&\hat{\mathbf{z}}\\[2pt]\dfrac{\partial}{\partial x}&\dfrac{\partial}{\partial y}&\dfrac{\partial}{\partial z}\\[4pt]F_x&F_y&F_z\end{vmatrix} \qquad (2)

where: the determinant is expanded along the top row, and each operator must sit to the left of the component it differentiates. Like all del-determinant tricks it is Cartesian-only: in cylindrical or spherical coordinates the moving basis vectors add scale-factor terms, and the tabulated formulas must be used. For proving identities the sleekest packaging uses the Levi-Civita symbol:

(\nabla\times\mathbf{F})_i=\varepsilon_{ijk}\,\partial_j F_k \qquad (3)

where: $\varepsilon_{ijk}$ — Levi-Civita symbol: $+1$ for even permutations of $(1,2,3)$ , $-1$ for odd, $0$ if any index repeats; repeated indices $j,k$ are summed (Einstein convention). One line of epsilon algebra replaces a page of component bookkeeping — the curl-of-curl identity below is the standard demonstration. The definition that carries the meaning, though, is geometric — circulation per unit area:

(\nabla\times\mathbf{F})\cdot\hat{\mathbf{n}}=\lim_{|S|\to0}\frac{1}{|S|}\oint_{\partial S}\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell} \qquad (4)

where: $S$ — small planar surface with unit normal $\hat{\mathbf{n}}$ ; $\partial S$ — its boundary loop, oriented by the right-hand rule relative to $\hat{\mathbf{n}}$ ; $|S|$ — area; $\mathrm{d}\boldsymbol{\ell}$ — line element along the loop. Like the flux definition of divergence, equation (4) mentions no axes, which is what makes the curl a genuine geometric object. Summed over a mosaic of small loops tiling a finite surface — interior edges cancelling — it becomes the curl (Stokes) theorem:

\oint_{\partial S}\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell}=\int_S(\nabla\times\mathbf{F})\cdot\mathrm{d}\mathbf{a} \qquad (5)

where: $\mathrm{d}\mathbf{a}$ — vector area element $\hat{\mathbf{n}}\,\mathrm{d}a$ . Circulation around the rim equals curl-flux through the membrane; the full anatomy of this theorem gets its own article.

Figure

Circulation around a small rectangle — opposite edges disagree, and the disagreement is the curl

Derivation AThe Cartesian curl from circulation around a small rectangle

We are deriving the component formula (1) directly from the circulation definition (4) — the calculation that explains why the curl is those particular derivative differences. The strategy: walk counterclockwise around a small rectangle of sides $\Delta x, \Delta y$ , centered at ( $x_0, y_0$ ), in a plane $z=z_0$ , Taylor-expand the field along each edge, and watch the constant parts cancel between opposite edges while the cross-derivatives survive. The sign discipline is the whole game — opposite edges are traversed in opposite directions, so each pair reports a difference of the field, which becomes a derivative in the limit.

Step 1. Bottom edge, traversed in $+x$ at height $y_0-\tfrac{\Delta y}{2}$ :

\int_{\text{bot}}\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell}\approx F_x\big(x_0,\,y_0-\tfrac{\Delta y}{2}\big)\,\Delta x

Why: along a horizontal edge only $F_x$ projects onto the path; over a short edge the field is well approximated by its midpoint value.

Step 2. Top edge, traversed in $-x$ at height $y_0+\tfrac{\Delta y}{2}$ , picks up the opposite sign:

\int_{\text{top}}\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell}\approx -F_x\big(x_0,\,y_0+\tfrac{\Delta y}{2}\big)\,\Delta x

Why: counterclockwise travel means the return leg runs against $+\hat{\mathbf{x}}$ ; the two horizontal edges together therefore measure how much $F_x$ differs across the rectangle's height.

Step 3. Combine the horizontal pair by Taylor expansion:

\int_{\text{bot}}+\int_{\text{top}}\approx-\frac{\partial F_x}{\partial y}\,\Delta y\,\Delta x

Why: the difference of $F_x$ across a separation $\Delta y$ is $(\partial F_x/\partial y)\Delta y$ to first order — a field growing with $y$ fights the counterclockwise walk, hence the minus sign.

Step 4. The vertical pair works identically with roles swapped and signs reversed (right edge up, left edge down):

\int_{\text{right}}+\int_{\text{left}}\approx+\frac{\partial F_y}{\partial x}\,\Delta x\,\Delta y

Why: $F_y$ growing along $x$ aids the counterclockwise walk on the right edge more than it hinders on the left.

Step 5. Add all four edges and divide by the area $\Delta x\,\Delta y$ :

\frac{1}{\Delta x\,\Delta y}\oint\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell}\;\longrightarrow\;\frac{\partial F_y}{\partial x}-\frac{\partial F_x}{\partial y} = (\nabla\times\mathbf{F})_z

Cyclic permutation of the axes delivers the $x$ - and $y$ -components, completing formula (1). For the linear shear $\mathbf{F}=y\hat{\mathbf{x}}$ the recipe gives exactly $-1$ with no limit needed, as the worked example below confirms numerically.

Beyond the first derivatives, three curl identities carry most of electromagnetism. The curl of a gradient vanishes,

\nabla\times(\nabla\phi)=\mathbf{0} \qquad (6)

where: $\phi$ — any twice-differentiable scalar field; each component is a difference of mixed partials, e.g. $\partial_x\partial_y\phi - \partial_y\partial_x\phi = 0$ by Clairaut's theorem. This is why the electrostatic field, which we will show is curl-free, can be written as $-\nabla V$ . Its mirror image, proved in the divergence article, is

\nabla\cdot(\nabla\times\mathbf{A})=0 \qquad (7)

where: $\mathbf{A}$ — any twice-differentiable vector field. Curls never diverge — so defining $\mathbf{B}=\nabla\times\mathbf{A}$ builds "no magnetic monopoles" into the formalism for free. And applying curl twice folds back onto the other operators:

\nabla\times(\nabla\times\mathbf{F})=\nabla(\nabla\cdot\mathbf{F})-\nabla^2\mathbf{F} \qquad (8)

where: $\nabla^2\mathbf{F}$ — the vector Laplacian, componentwise in Cartesian coordinates. This is the crank that turns Maxwell's equations into the electromagnetic wave equation. Finally, the Leibniz rule for a scalar times a field:

\nabla\times(f\mathbf{F})=(\nabla f)\times\mathbf{F}+f(\nabla\times\mathbf{F}) \qquad (9)

where: $f$ — scalar field; note the first term is a cross product of the gradient with $\mathbf{F}$ — the sign and order matter.

Go deeper · Derivation B · 1 minThe field around a wire is curl-free off the axis

Derivation BThe field around a wire is curl-free off the axis

We are computing the curl of $\mathbf{F}=(k/s)\hat{\boldsymbol{\phi}}$ — the shape of the magnetic field around a straight current-carrying wire — and showing it vanishes everywhere except on the axis, where all the curl hides as a delta function. This is the curl-side twin of the Coulomb-field divergence calculation, and it teaches the same lesson: a field can look maximally swirly while being locally rotation-free, its entire "vorticity" concentrated on its source. We work in cylindrical coordinates, where the relevant curl component is $(\nabla\times\mathbf{F})_z=\frac{1}{s}\left[\partial_s(sF_\phi)-\partial_\phi F_s\right]$ ; the factor of $s$ inside the derivative is the step to watch.

Step 1. Identify the components: $F_s=0$ , $F_\phi=k/s$ , $F_z=0$ , with dependence on $s$ only. Why: the field circulates azimuthally with magnitude falling as $1/s$ — the ansatz Ampère's law will later derive for a line current with $k=\mu_0 I/2\pi$ .

Step 2. Apply the cylindrical curl's $z$ -component:

(\nabla\times\mathbf{F})_z=\frac{1}{s}\,\partial_s\!\left(s\cdot\frac{k}{s}\right)-\frac{1}{s}\,\partial_\phi(0) = \frac{1}{s}\,\partial_s(k)=0 \quad (s\neq 0)

Why: the combination $sF_\phi=k$ is constant — the $1/s$ falloff is precisely the rate at which a paddle wheel's outer edge is pushed less hard, cancelling its inner edge's advantage. The $s$ - and $\phi$ -components of the curl vanish too, since $F_z=F_s=0$ and nothing depends on $z$ or $\phi$ .

Step 3. Interrogate the axis with Stokes' theorem. Around a circle of radius $s$ centred on the axis,

\oint\mathbf{F}\cdot\mathrm{d}\boldsymbol{\ell}=\frac{k}{s}\cdot 2\pi s = 2\pi k \neq 0

Why: nonzero circulation with zero pointwise curl everywhere on the disc except one point forces the curl to be a delta function on the axis — the exact analogue of the Coulomb field's $4\pi\delta^3(\mathbf{r})$ .

Step 4. Read it as magnetostatics: $\nabla\times\mathbf{B} = \mu_0\mathbf{J}$ with $\mathbf{B}=\frac{\mu_0 I}{2\pi s}\hat{\boldsymbol{\phi}}$ gives a current density concentrated on the axis carrying total current $I$ — Ampère's law for a line current, with the wire itself as the only place the field truly rotates. The circulation $2\pi k$ is radius-independent, as a delta-source demands.

Go deeper · Derivation C · 1 minThe curl of a rigid rotation field is twice the angular velocity

Derivation CThe curl of a rigid rotation field is twice the angular velocity

We are proving the calibration fact of the whole subject: for a velocity field that rotates like a rigid body, $\mathbf{F}=\boldsymbol{\omega}\times\mathbf{r}$ , the curl is exactly $2\boldsymbol{\omega}$ — uniform everywhere, twice the angular velocity. This anchors the paddle-wheel picture quantitatively (the wheel spins at $\omega$ , half the curl) and explains why fluid dynamicists call the curl of velocity the vorticity. The route is direct Cartesian computation with $\boldsymbol{\omega}=\omega\hat{\mathbf{z}}$ ; the factor of 2 emerges from the two off-diagonal derivatives contributing equally — a signature worth remembering when factor-of-2 doubts strike.

Step 1. Write out the field for constant $\boldsymbol{\omega}=\omega\hat{\mathbf{z}}$ :

\mathbf{F}=\omega\hat{\mathbf{z}}\times(x\hat{\mathbf{x}}+y\hat{\mathbf{y}}+z\hat{\mathbf{z}})=\omega(-y\,\hat{\mathbf{x}}+x\,\hat{\mathbf{y}})

Why: $\hat{\mathbf{z}}\times\hat{\mathbf{x}}=\hat{\mathbf{y}}$ and $\hat{\mathbf{z}}\times\hat{\mathbf{y}}=-\hat{\mathbf{x}}$ ; each point moves azimuthally with speed $\omega s$ , the rigid-carousel profile.

Step 2. Only the $z$ -component of the curl can survive (the field is planar and $z$ -independent):

(\nabla\times\mathbf{F})_z=\partial_x F_y-\partial_y F_x=\partial_x(\omega x)-\partial_y(-\omega y) = \omega+\omega = 2\omega

Why: both antisymmetric derivatives contribute $+\omega$ — one from the azimuthal speed growing with $x$ , one from it growing with $y$ . Neither alone is the rotation rate; their sum is double it.

Step 3. Restore vector form:

\nabla\times(\boldsymbol{\omega}\times\mathbf{r})=2\boldsymbol{\omega}

A paddle wheel anywhere in the flow — centre or rim — spins at $\omega$ , confirming that curl is a local, position-independent property of rigid rotation. Contrast with the wire field $k/s\,\hat{\boldsymbol\phi}$ : same circular streamlines, radically different curl. The vorticity of solid-body rotation is $2\boldsymbol{\omega}$ ; the same result also follows in one line from the product-rule identity for $\nabla\times(\mathbf{A}\times\mathbf{r})$ with constant $\mathbf{A}$ .

Limiting cases

Reading the curl in simple regimes:

Case	Result	Meaning
Uniform field	$\mathbf{F}=\mathbf{F}_0=\text{const}\;\Rightarrow\;\nabla\times\mathbf{F}=\mathbf{0}$	A constant field has no spatial variation, hence no shear and no circulation density.
Pure gradient (irrotational)	$\mathbf{F}=\nabla\phi\;\Rightarrow\;\nabla\times\mathbf{F}=\mathbf{0}$	Any field derived from a potential has identically zero curl; the converse holds in simply connected regions. The electrostatic field is the prototype.
Rigid rotation	$\mathbf{F}=\boldsymbol{\omega}\times\mathbf{r}\;\Rightarrow\;\nabla\times\mathbf{F}=2\boldsymbol{\omega}$	A field rotating like a rigid body has uniform curl equal to twice the angular velocity — the cleanest illustration of curl as rotation.
Straight-line shear	$\mathbf{F}=y\,\hat{\mathbf{x}}\;\Rightarrow\;\nabla\times\mathbf{F}=-\hat{\mathbf{z}}$	Perfectly straight field lines, nonzero curl: sideways variation of magnitude is rotation in disguise.
Field around a wire	$\mathbf{F}=\frac{k}{s}\,\hat{\boldsymbol{\phi}}\;\Rightarrow\;\nabla\times\mathbf{F}=\mathbf{0}\ (s\neq0)$	Visually swirling yet curl-free away from the axis: the field weakens as $1/s$ exactly fast enough to cancel circulation density. All the curl is concentrated on the axis (a delta function), modelling a line current.

Worked examples

Example 1 — curl of a concrete field, with a free identity check (confidence builder). Let $\mathbf{F} = xy\,\hat{\mathbf{x}} + 2yz\,\hat{\mathbf{y}} + 3zx\,\hat{\mathbf{z}}$ . Find $\nabla\times\mathbf{F}$ at $P=(1,2,3)$ .

Step 1. Expand the determinant (2) component by component:

\nabla\times\mathbf{F} = \big(\partial_y(3zx)-\partial_z(2yz)\big)\hat{\mathbf{x}} + \big(\partial_z(xy)-\partial_x(3zx)\big)\hat{\mathbf{y}} + \big(\partial_x(2yz)-\partial_y(xy)\big)\hat{\mathbf{z}}

Step 2. Evaluate each derivative: $\partial_y(3zx)=0$ , $\partial_z(2yz)=2y$ ; $\partial_z(xy)=0$ , $\partial_x(3zx)=3z$ ; $\partial_x(2yz)=0$ , $\partial_y(xy)=x$ . Hence

\nabla\times\mathbf{F} = -2y\,\hat{\mathbf{x}} - 3z\,\hat{\mathbf{y}} - x\,\hat{\mathbf{z}}, \qquad \nabla\times\mathbf{F}\big|_P = -4\,\hat{\mathbf{x}} - 9\,\hat{\mathbf{y}} - 1\,\hat{\mathbf{z}}

Step 3. Free check via identity (7): $\nabla\cdot(\nabla\times\mathbf{F}) = \partial_x(-2y) + \partial_y(-3z) + \partial_z(-x) = 0 + 0 + 0 = \mathbf{0}$ — the curl field is divergence-free everywhere, exactly as it must be. Any nonzero answer here means an arithmetic slip upstream; this one-line audit catches a large fraction of exam errors.

Example 2 — circulation of the wire field and Ampère's law (exam level). A long straight wire carries $I = 10\ \mathrm{A}$ , producing $\mathbf{B} = \dfrac{\mu_0 I}{2\pi s}\hat{\boldsymbol{\phi}}$ with $\mu_0 = 4\pi\times 10^{-7}\ \mathrm{T\,m/A}$ .

Step 1. Field magnitudes: at $s_1 = 2\ \mathrm{cm}$ ,

B_1 = \frac{(4\pi\times 10^{-7})(10)}{2\pi(0.02)} = 1.0\times 10^{-4}\ \mathrm{T}

and at $s_2 = 5\ \mathrm{cm}$ , $B_2 = 4.0\times 10^{-5}\ \mathrm{T}$ .

Step 2. Circulation around the circle of radius $s_1$ (field parallel to path everywhere):

\oint\mathbf{B}\cdot\mathrm{d}\boldsymbol{\ell} = B_1\cdot 2\pi s_1 = (1.0\times 10^{-4})(2\pi)(0.02) = 1.257\times 10^{-5}\ \mathrm{T\,m}

Step 3. Repeat at $s_2$ : $B_2\cdot 2\pi s_2 = (4.0\times 10^{-5})(2\pi)(0.05) = 1.257\times 10^{-5}\ \mathrm{T\,m}$ — identical. The circulation is radius-independent and equals $\mu_0 I = 4\pi\times 10^{-7}\times 10 = 1.257\times 10^{-5}\ \mathrm{T\,m}$ exactly.

Step 4. Interpret with Stokes' theorem (5): the flux of $\nabla\times\mathbf{B}$ through any disc bounded by these circles is the same, yet the derivation above showed $\nabla\times\mathbf{B}=0$ at every off-axis point. All the curl is a delta function on the axis, of strength $\mu_0 I$ — the current itself. This is Ampère's law seen from the vector-calculus side, two chapters early.

Applications

Faraday's law of induction. The differential form $\nabla\times\mathbf{E}=-\partial_t\mathbf{B}$ states that a time-varying magnetic field generates a circulating electric field. Integrating with Stokes' theorem yields the EMF $\oint\mathbf{E}\cdot\mathrm{d}\boldsymbol{\ell}=-\mathrm{d}\Phi_B/\mathrm{d}t$ around any loop, the principle behind generators and transformers. The curl makes precise that induced fields swirl around changing flux.

Ampère–Maxwell law and magnetic fields. $\nabla\times\mathbf{B}=\mu_0\mathbf{J}+\mu_0\varepsilon_0\,\partial_t\mathbf{E}$ identifies current density and displacement current as sources of circulating magnetic field. For a steady line current this reproduces $\mathbf{B}=\mu_0 I/(2\pi s)\,\hat{\boldsymbol{\phi}}$ , whose curl is concentrated on the wire — exactly the off-axis-curl-free field analysed above.

Vector potential and gauge freedom. Because $\nabla\cdot(\nabla\times\mathbf{A})=0$ identically, defining $\mathbf{B}=\nabla\times\mathbf{A}$ automatically enforces $\nabla\cdot\mathbf{B}=0$ (no magnetic monopoles). The freedom to add any gradient $\nabla\chi$ to $\mathbf{A}$ without changing $\mathbf{B}$ (since $\nabla\times\nabla\chi=0$ ) is the gauge freedom of electromagnetism.

Conservative fields and potential energy. A static electric field satisfies $\nabla\times\mathbf{E}=\mathbf{0}$ , so by the converse to the gradient identity $\mathbf{E}=-\nabla V$ for a scalar potential $V$ . This is why electrostatic work is path-independent and why a potential energy can be defined at all — the vanishing curl is the field-level statement of conservativeness.

Origins · 1 minJames Clerk Maxwell and how this was discovered

James Clerk Maxwell

1831–1879 · Cambridge, Aberdeen and King's College London

Maxwell not only wrote the equations in which the curl stars — he named the operator. In an 1871 note to the London Mathematical Society he auditioned names for the antisymmetric derivative of a vector field: twist he rejected as too mechanical, twirl as "too dynamical for mathematicians", before settling on curl — with the alternative version held in reserve. The physics came first: his 1861 paper "On Physical Lines of Force" modelled magnetic fields as arrays of molecular vortices, with electric current as the motion of idle wheels between them — a gloriously literal machine of curls from which he extracted, correctly, that circulating $\mathbf{B}$ wraps around currents and circulating $\mathbf{E}$ wraps around changing $\mathbf{B}$ . By the 1873 Treatise the scaffolding was gone and the mathematics stood alone. When Heaviside later compressed the theory to four equations, two of the four were curls — Maxwell's paddle wheels, immortalized in notation.

History

1858

— Helmholtz's vortex theorems make the rotation of a velocity field (vorticity) a central object of fluid mechanics

1861

— Maxwell's molecular-vortex model derives the laws of circulating electric and magnetic fields mechanically

1871

— Maxwell coins the name curl for the operator, rejecting twist and twirl

1873

— The Treatise on Electricity and Magnetism embeds the curl in the field equations of electromagnetism

1883

— Heaviside and Gibbs's vector analysis gives the curl its modern cross-product notation

1892

— Heaviside's four vector Maxwell equations put the two curl laws — Faraday and Ampère–Maxwell — at the theory's core

Going further

References. Griffiths, Introduction to Electrodynamics, 4th ed., §1.2.5 and §1.3.5 · Jackson, Classical Electrodynamics, 3rd ed., front-cover vector identities and §1 · Purcell & Morin, Electricity and Magnetism, 3rd ed., Chapter 2 · Zangwill, Modern Electrodynamics, §1.2 The key symbolic results developed in this article are: (1) the rectangular-loop derivation giving $(\nabla\times\mathbf{F})_z=\partial_x F_y-\partial_y F_x$ ; (2) $\nabla\times((k/s)\hat{\boldsymbol{\phi}})=0$ for $s\neq0$ via the cylindrical curl; (3) $\nabla\times(\boldsymbol{\omega}\times\mathbf{r})=2\boldsymbol{\omega}$ ; and (4) the line-integral value $2\pi k$ yielding Ampère's law with $k=\mu_0 I/2\pi$ . Numerical constants $\mu_0,\varepsilon_0$ are not quoted; only standard symbolic relations appear.

Builds on: Scalar and Vector Fields · The Del Operator (Nabla) · The Divergence of a Vector Field · Line Integrals Leads to: The Curl (Stokes) Theorem · Second Derivatives and the Laplacian · Surface Integrals and Flux

Why does the paddle wheel spin at half the curl rather than at the curl itself?

The factor of 2 is baked into rigid rotation: a carousel turning at $\omega$ has velocity curl $2\omega$ , because two independent derivative terms ( $\partial_x F_y$ and $-\partial_y F_x$ ) each contribute $\omega$ . A free paddle wheel matches the local rigid-rotation rate of the fluid, i.e. $\omega = \frac{1}{2}|\nabla\times\mathbf{v}|$ . Fluid dynamicists live with the 2 (vorticity is defined as the full curl); some older texts define "rotation" as half the curl to absorb it. Nothing deep — just bookkeeping fixed by the $2\boldsymbol{\omega}$ calibration derivation.

How can the field around a wire circulate globally but have zero curl locally?

Because circulation and curl live at different scales. Take a paddle wheel orbiting the wire at radius $s$ : its outer half sits in weaker field ( $\propto 1/s$ ), its inner half in stronger field — and for exactly the $1/s$ profile, the "carried around" tendency and the "sheared backwards" tendency cancel. The wheel revolves around the wire without rotating about its own axle, like a gondola on a Ferris wheel. Any loop not enclosing the wire gets zero circulation; loops enclosing it get $2\pi k$ from the singular axis alone. Global swirl, local innocence.

Is the curl I compute from the determinant guaranteed to be independent of my choice of axes?

Yes — though the determinant makes it look coordinate-bound, the circulation definition (4) rescues it: circulation per unit area around a shrinking loop mentions no axes, and the derivation shows the Cartesian formula reproduces that geometric limit. Rotate your frame and the components remix exactly as a (pseudo)vector's should. The one caveat is reflections: under a mirror flip the curl gains an extra sign relative to a polar vector, because the right-hand rule is handed. That is why $\mathbf{B}$ , defined through curls, is an axial vector.

What happens to the curl in two dimensions?

It collapses to a single number: for $\mathbf{F}=(F_x(x,y),\,F_y(x,y))$ the only surviving component is the scalar $\partial_x F_y - \partial_y F_x$ — what Green's theorem integrates over a plane region. The counting explains it: antisymmetric derivative combinations number $n(n-1)/2$ , giving 1 in 2D, 3 in 3D (luckily assemblable into a vector), and 6 in 4D, where the object is irreducibly a tensor — which is exactly what the electromagnetic field strength $F_{\mu\nu}$ turns out to be in the relativistic formulation at the end of this track.

If a field has zero curl everywhere, is it automatically a gradient field?

Almost — the fine print is topology. In a simply connected region (no holes), yes: $\nabla\times\mathbf{F}=0$ guarantees $\mathbf{F}=\nabla\phi$ for some scalar $\phi$ , by building $\phi$ from path-independent line integrals. Puncture the region — remove the wire's axis, say — and the guarantee fails: the field $k\hat{\boldsymbol\phi}/s$ is curl-free in the punctured plane but has nonzero circulation around the hole, so no single-valued potential exists (the candidate, $k\phi$ , is multivalued). This subtlety becomes real physics in the Aharonov–Bohm effect and in magnetic scalar potentials.

Why does the curl appear on the left of Faraday's and Ampère's laws rather than the divergence?

Because the physics of both laws is circulation. Faraday's discovery was that changing magnetic flux drives current around a loop — an EMF, which is a line integral of $\mathbf{E}$ around a closed path. Ørsted's compass needles circled a wire — $\mathbf{B}$ wraps around currents. Localizing "around a loop" statements via the shrinking-loop definition (4) hands you curl equations, just as localizing "through a closed surface" statements (Gauss) hands you divergence equations. The grammar of the four Maxwell equations mirrors the geometry of the four founding experiments.

Do I ever need the middle part of the derivative tensor — the symmetric non-trace piece?

Not in electromagnetism proper, but knowing it exists prevents a misconception. The nine derivatives $\partial_i F_j$ split into divergence (1 number: isotropic expansion), curl (3 numbers: rigid rotation), and a symmetric traceless shear (5 numbers: stretching without volume change or net rotation). E&M's field equations constrain only the first two — which is precisely the content of the Helmholtz theorem, coming later in this chapter: a field vanishing at infinity is fully reconstructible from its divergence and curl. The shear piece matters in elasticity and viscous flow, where it encodes deformation.