Divergence and Curl
A vector field \mathbf{F}(x,y,z) = (P, Q, R)
attaches an arrow to every point of space — a velocity of flowing water, an electric field,
a wind map. The operator \nabla (called "del" or "nabla")
\nabla = \left(\frac{\partial}{\partial x},\ \frac{\partial}{\partial y},\ \frac{\partial}{\partial z}\right)
is a vector of
partial derivatives.
Combine it with a field in the two natural ways a vector combines with another vector — the
dot product and the cross product — and you get the two
first-order ways a field can vary: how much it spreads, and how much it
spins.
Divergence: net outflow per unit volume
The divergence is the dot product \nabla \cdot \mathbf{F}
— a scalar at each point measuring the net rate at which the field flows
out of a tiny region around it. Treating \nabla as a
vector and dotting it with (P, Q, R):
\operatorname{div}\mathbf{F} = \nabla \cdot \mathbf{F} = \frac{\partial P}{\partial x} + \frac{\partial Q}{\partial y} + \frac{\partial R}{\partial z}.
Positive divergence at a point marks a source (more flows out than in);
negative marks a sink; zero means whatever flows in flows back out
(an incompressible flow).
Worked example — divergence
Take the outward "explosion" field \mathbf{F} = (x,\ y,\ z),
every arrow pointing radially away from the origin.
Step 1 — read off the components.
P = x, Q = y,
R = z.
Step 2 — differentiate each by its matching variable.
\frac{\partial P}{\partial x} = \frac{\partial x}{\partial x} = 1, \qquad \frac{\partial Q}{\partial y} = 1, \qquad \frac{\partial R}{\partial z} = 1.
Step 3 — add.
\nabla \cdot \mathbf{F} = 1 + 1 + 1 = 3.
A constant +3 everywhere: this field is a source at every point,
spreading uniformly. (In two dimensions the same (x,y) would give
2 — one per spatial dimension.)
Curl: local rotation
The curl is the cross product \nabla \times \mathbf{F}
— a vector at each point whose direction is the axis of local spin (by the
right-hand rule) and whose length is twice the local angular speed. Expanding the formal
determinant,
\operatorname{curl}\mathbf{F} = \nabla \times \mathbf{F} = \begin{vmatrix} \mathbf{i} & \mathbf{j} & \mathbf{k} \\[2pt] \dfrac{\partial}{\partial x} & \dfrac{\partial}{\partial y} & \dfrac{\partial}{\partial z} \\[4pt] P & Q & R \end{vmatrix} = \left(\frac{\partial R}{\partial y} - \frac{\partial Q}{\partial z},\ \ \frac{\partial P}{\partial z} - \frac{\partial R}{\partial x},\ \ \frac{\partial Q}{\partial x} - \frac{\partial P}{\partial y}\right).
If you placed a tiny paddle wheel in the flow, the curl tells you which way and how fast it
would turn. A field with zero curl everywhere is called irrotational.
Worked example — curl
Take the "merry-go-round" field \mathbf{F} = (-y,\ x,\ 0), which
circulates counter-clockwise about the z-axis.
Step 1 — components.
P = -y, Q = x,
R = 0.
Step 2 — the \mathbf{i} component,
\partial R/\partial y - \partial Q/\partial z. Both
R = 0 and Q = x are free of
z and y here:
\frac{\partial R}{\partial y} - \frac{\partial Q}{\partial z} = 0 - 0 = 0.
Step 3 — the \mathbf{j} component,
\partial P/\partial z - \partial R/\partial x:
\frac{\partial P}{\partial z} - \frac{\partial R}{\partial x} = 0 - 0 = 0.
Step 4 — the \mathbf{k} component,
\partial Q/\partial x - \partial P/\partial y:
\frac{\partial Q}{\partial x} - \frac{\partial P}{\partial y} = \frac{\partial x}{\partial x} - \frac{\partial(-y)}{\partial y} = 1 - (-1) = 2.
Step 5 — assemble.
\nabla \times \mathbf{F} = (0,\ 0,\ 2) = 2\,\mathbf{k}.
The curl points along +\mathbf{k} — the spin axis of a
counter-clockwise swirl in the xy-plane — and its magnitude
2 is twice the angular speed 1. Notice
this field has \nabla \cdot \mathbf{F} = -0 + 0 + 0 = 0: pure
rotation, no spreading.
Two identities: "div of curl" and "curl of grad"
Mixed second derivatives commute (Clairaut's theorem),
\partial^2/\partial x\,\partial y = \partial^2/\partial y\,\partial x.
That single fact forces two clean cancellations.
(a) The divergence of a curl is zero
Step 1 — take the divergence of the curl. Dot
\nabla into the curl vector from above:
\nabla \cdot (\nabla \times \mathbf{F}) = \frac{\partial}{\partial x}\!\left(\frac{\partial R}{\partial y} - \frac{\partial Q}{\partial z}\right) + \frac{\partial}{\partial y}\!\left(\frac{\partial P}{\partial z} - \frac{\partial R}{\partial x}\right) + \frac{\partial}{\partial z}\!\left(\frac{\partial Q}{\partial x} - \frac{\partial P}{\partial y}\right).
Step 2 — expand into six terms.
= R_{yx} - Q_{zx} + P_{zy} - R_{xy} + Q_{xz} - P_{yz}.
Step 3 — pair them off. By Clairaut, R_{yx} = R_{xy},
Q_{zx} = Q_{xz}, and P_{zy} = P_{yz},
so the six terms cancel in three pairs:
\nabla \cdot (\nabla \times \mathbf{F}) = 0.
Curls never have sources — a swirl neither creates nor destroys flow.
(b) The curl of a gradient is zero
Step 1 — start from a scalar field's
gradient
\nabla f = (f_x,\ f_y,\ f_z), and take its curl. The
\mathbf{k} component is
(\nabla \times \nabla f)_{\mathbf{k}} = \frac{\partial}{\partial x}(f_y) - \frac{\partial}{\partial y}(f_x) = f_{yx} - f_{xy}.
Step 2 — apply Clairaut. f_{yx} = f_{xy}, so this
component is 0; the same cancellation kills the
\mathbf{i} and \mathbf{j} components.
\nabla \times (\nabla f) = \mathbf{0}.
Gradient fields never swirl — they are exactly the irrotational (conservative) fields. These
two identities are the algebraic skeleton behind
Stokes' theorem
and
the divergence theorem.
For a vector field \mathbf{F} = (P, Q, R) with continuous
second partials and a scalar field f:
-
Divergence is the scalar
\nabla \cdot \mathbf{F} = P_x + Q_y + R_z — the net outflow
per unit volume (source density). {>}\,0 a source,
{<}\,0 a sink, 0 incompressible.
-
Curl is the vector
\nabla \times \mathbf{F} = (R_y - Q_z,\ P_z - R_x,\ Q_x - P_y)
— the axis and twice the rate of local rotation.
-
Divergence of a curl vanishes:
\nabla \cdot (\nabla \times \mathbf{F}) = 0.
-
Curl of a gradient vanishes:
\nabla \times (\nabla f) = \mathbf{0} — gradient fields are
irrotational.
Two imaginary probes make the operators physical. Drop a tiny paddle wheel
into the flow: it spins about the curl vector at half its length, so it stays still exactly
where the field is irrotational and whirls fastest where the curl is largest. The
(-y, x, 0) merry-go-round above spins it about
+\mathbf{k}; the radial (x,y,z) leaves
it dead still.
Now imagine a tiny balloon drifting with the flow. Where the divergence is
positive the surrounding field carries its skin outward and the balloon inflates; where it
is negative the balloon shrinks; where it is zero the volume is preserved even as the shape
distorts. Divergence measures change of size, curl measures change of
orientation — together they capture the entire first-order behaviour of a flow.
See it: spread vs spin
Switch between the two archetypes. The radial field
\mathbf{F} = (x, y) is all divergence and no curl; the
circulating field \mathbf{F} = (-y, x) is all curl and no
divergence. The computed values appear beneath each.