Divergence and curl
A moving fluid can spread out from a point, and it can swirl around a point. How do the partial derivatives of its velocity field measure each of these, and why do the measurements not depend on the choice of axes?
Where you are 8 in Vector calculus
Why this exists
Green's theorem produced two integrands from a plane field \langle P, Q \rangle: the scalar curl Q_x - P_y, circulation per unit area, and P_x + Q_y, whose integral is outward flux. Both are written in fixed axes, though circulation and flux are not. This lesson names the two in space, the divergence and the curl, and proves what they measure at a point; that is why they do not depend on the axes, and it is what Stokes' theorem and The divergence theorem integrate.
Definition
Let \mathbf{F} = \langle P, Q, R \rangle be a vector field (Vector fields) on an open set in \mathbb{R}^3 whose components have first partial derivatives (Partial derivatives). Write \nabla = \left\langle \dfrac{\partial}{\partial x}, \dfrac{\partial}{\partial y}, \dfrac{\partial}{\partial z} \right\rangle as a formal vector, so that \nabla f is the gradient.
- The divergence of \mathbf{F} is the function
- The curl of \mathbf{F} is the vector field
the formal determinant expanded along its first row as for the The cross product and the triple product, a “product” of a derivative with a function read as taking that derivative. - A plane field \langle P, Q \rangle has divergence P_x + Q_y. As the field \langle P, Q, 0 \rangle in space, with P and Q free of z, its curl is (Q_x - P_y)\,\mathbf{k}: the scalar curl is the \mathbf{k}-component of the curl. - The Laplacian of f is \nabla^2 f = \operatorname{div} \nabla f = f_{xx} + f_{yy} + f_{zz}; f is harmonic if \nabla^2 f = 0. - \mathbf{F} is source-free (for a velocity field, incompressible) if \operatorname{div} \mathbf{F} = 0, and irrotational if \operatorname{curl} \mathbf{F} = \mathbf{0}.
A field or function is C^1 when its components have continuous first partial derivatives, and C^2 when they have continuous second partials.
Intuition
Think of \mathbf{F} as the velocity of a fluid. For a small box with sides \Delta x, \Delta y, \Delta z, the two faces perpendicular to the x-axis pass a net outflow of about \big(P(x + \Delta x, y, z) - P(x, y, z)\big)\Delta y\,\Delta z \approx P_x\,\Delta V; the other pairs give Q_y\,\Delta V and R_z\,\Delta V. So \operatorname{div} \mathbf{F} is net outflow per unit volume: positive where fluid is produced or expands, negative where it drains or compresses.
Curl measures turning. Take a small circle of radius \varepsilon about \mathbf{p}, perpendicular to a unit vector \mathbf{n}, with unit tangent \mathbf{T}. A particle on it moves around \mathbf{p} with angular velocity \mathbf{F} \cdot \mathbf{T}/\varepsilon, whose average over the circle is
By Result 3 this tends to \displaystyle \frac12 \operatorname{curl} \mathbf{F}(\mathbf{p}) \cdot \mathbf{n}: a tiny paddle wheel with axis \mathbf{n} turns at that rate, fastest with its axis along the curl. Curl is local spin, not curved paths: the shear field \langle y, 0 \rangle of Figure 1 has straight flow lines, but faster flow above than below turns a wheel clockwise, and \operatorname{curl} \langle y, 0, 0 \rangle = \langle 0, 0, -1 \rangle.
Results
Result 1 (Green’s theorem in vector form). For D, P and Q as in Result 2 of Green's theorem, with \partial D positively oriented and outward unit normal \mathbf{n},
Result 2 (divergence is flux per unit size). Let \mathbf{F} be C^1 on an open set containing \mathbf{p}.
(a) In the plane, with C_\varepsilon the circle of radius \varepsilon about \mathbf{p} and \mathbf{n} its outward unit normal,
(b) In space, the outward flux of \mathbf{F} through the boundary of a box B = [a_1, b_1] \times [a_2, b_2] \times [a_3, b_3] is the sum over its six faces of \iint \mathbf{F} \cdot \mathbf{n}\,dA, with \mathbf{n} the face’s outward unit normal. If \mathbf{F} is C^1 on an open set containing B, this flux is \iiint_B \operatorname{div} \mathbf{F}\,dV. For the cube B_h of side h centered at \mathbf{p}, the flux divided by h^3 tends to \operatorname{div} \mathbf{F}(\mathbf{p}) as h \to 0^+.
Result 3 (curl is circulation per unit area). Let \mathbf{F} be C^1 on an open set in \mathbb{R}^3 containing \mathbf{p}, and let \mathbf{n} be a unit vector. Choose unit vectors \mathbf{u}, \mathbf{v} with \mathbf{u} \cdot \mathbf{v} = 0 and \mathbf{u} \times \mathbf{v} = \mathbf{n}, and let C_\varepsilon be the circle \mathbf{r}(\theta) = \mathbf{p} + \varepsilon\cos\theta\,\mathbf{u} + \varepsilon\sin\theta\,\mathbf{v}, 0 \le \theta \le 2\pi, counterclockwise as seen from the tip of \mathbf{n}. Then
If \operatorname{curl} \mathbf{F}(\mathbf{p}) \ne \mathbf{0}, this circulation density is largest, with value \lVert \operatorname{curl} \mathbf{F}(\mathbf{p}) \rVert, exactly when \mathbf{n} points along \operatorname{curl} \mathbf{F}(\mathbf{p}).
Results 2 and 3 express the divergence and every component of the curl through fluxes and circulations alone, so for C^1 fields both are the same in every right-handed system of perpendicular axes.
Result 4 (rules). Divergence and curl are linear, and for a C^1 function f and C^1 field \mathbf{F},
Result 5 (two zeros). If f is C^2, then \operatorname{curl} \nabla f = \mathbf{0}. If \mathbf{F} is C^2, then \operatorname{div} \operatorname{curl} \mathbf{F} = 0.
Result 6 (potentials). (a) A C^1 field conservative on an open set has curl \mathbf{0} there (Result 4 of Conservative fields and potential functions). Conversely, a C^1 field with curl \mathbf{0} on an open simply connected set is conservative there, proved in Stokes' theorem.
(b) If \mathbf{F} = \langle P, Q, R \rangle is C^1 on \mathbb{R}^3 with \operatorname{div} \mathbf{F} = 0, then \mathbf{F} = \operatorname{curl} \mathbf{G} for the vector potential \mathbf{G} = \langle G_1, G_2, 0 \rangle,
By Result 5, \mathbf{G} + \nabla h is another vector potential for every C^2 function h.
Why it is true
Result 1. For \langle P, Q, 0 \rangle, (\operatorname{curl} \mathbf{F}) \cdot \mathbf{k} = Q_x - P_y and \operatorname{div} \mathbf{F} = P_x + Q_y, so these are Results 2 and 4 of Green's theorem.
Result 2. (a) Repeat the proof of Result 5 of Green's theorem with the flux form in place of Result 1.
(b) The faces x = b_1 and x = a_1, with normals \mathbf{i} and -\mathbf{i}, contribute \iint \big( P(b_1, y, z) - P(a_1, y, z) \big)\,dA over [a_2, b_2] \times [a_3, b_3]. The integrand is \int_{a_1}^{b_1} P_x\,dx by the fundamental theorem of calculus, so by Fubini’s theorem on the box (Triple integrals) these faces give \iiint_B P_x\,dV. The other pairs give \iiint_B Q_y\,dV and \iiint_B R_z\,dV. This is the box case of The divergence theorem. For the limit, let g = \operatorname{div} \mathbf{F} and \eta > 0. For small h, \lvert g - g(\mathbf{p}) \rvert < \eta on B_h, so the flux over h^3 differs from g(\mathbf{p}) by \displaystyle \frac{1}{h^3} \left\lvert \iiint_{B_h} \big( g - g(\mathbf{p}) \big)\,dV \right\rvert \le \eta.
Result 3. For \mathbf{w} = \langle w_1, w_2, w_3 \rangle write \mathbf{F}_{\mathbf{w}} = w_1\mathbf{F}_x + w_2\mathbf{F}_y + w_3\mathbf{F}_z, where \mathbf{F}_x = \langle P_x, Q_x, R_x \rangle and so on. By the chain rule (The multivariable chain rule), \mathbf{F}_{\mathbf{w}}(\mathbf{q}) is the derivative of h \mapsto \mathbf{F}(\mathbf{q} + h\mathbf{w}) at h = 0.
Step 1: \mathbf{F}_{\mathbf{u}} \cdot \mathbf{v} - \mathbf{F}_{\mathbf{v}} \cdot \mathbf{u} = \operatorname{curl} \mathbf{F} \cdot (\mathbf{u} \times \mathbf{v}) for all \mathbf{u}, \mathbf{v}. Both sides are linear in each of \mathbf{u} and \mathbf{v} and change sign when they are swapped, so both vanish when \mathbf{u} = \mathbf{v}. Expanding in \mathbf{i}, \mathbf{j}, \mathbf{k} reduces the claim to the pairs (\mathbf{i}, \mathbf{j}), (\mathbf{j}, \mathbf{k}), (\mathbf{k}, \mathbf{i}), the reversed pairs following by the sign change. For (\mathbf{i}, \mathbf{j}) the left side is \mathbf{F}_x \cdot \mathbf{j} - \mathbf{F}_y \cdot \mathbf{i} = Q_x - P_y, and \mathbf{i} \times \mathbf{j} = \mathbf{k}. Likewise (\mathbf{j}, \mathbf{k}) gives R_y - Q_z and (\mathbf{k}, \mathbf{i}) gives P_z - R_x.
Step 2: reduce to the plane. For (s, t) near (0, 0) let G_1(s, t) = \mathbf{F}(\mathbf{p} + s\mathbf{u} + t\mathbf{v}) \cdot \mathbf{u} and G_2(s, t) = \mathbf{F}(\mathbf{p} + s\mathbf{u} + t\mathbf{v}) \cdot \mathbf{v}. By the chain rule, \partial_s G_2 = \mathbf{F}_{\mathbf{u}} \cdot \mathbf{v} and \partial_t G_1 = \mathbf{F}_{\mathbf{v}} \cdot \mathbf{u} at \mathbf{p} + s\mathbf{u} + t\mathbf{v}, both continuous. On C_\varepsilon, (s, t) = (\varepsilon\cos\theta, \varepsilon\sin\theta) and \mathbf{r}' = s'\mathbf{u} + t'\mathbf{v}, so \mathbf{F} \cdot \mathbf{r}' = G_1 s' + G_2 t': the circulation of \mathbf{F} around C_\varepsilon is that of \langle G_1, G_2 \rangle around the circle of radius \varepsilon about the origin of the st-plane, counterclockwise. By Result 5 of Green's theorem, over \pi\varepsilon^2 it tends to \partial_s G_2 - \partial_t G_1 at (0, 0), which is \operatorname{curl} \mathbf{F}(\mathbf{p}) \cdot \mathbf{n} by Step 1.
Finally \operatorname{curl} \mathbf{F}(\mathbf{p}) \cdot \mathbf{n} = \lVert \operatorname{curl} \mathbf{F}(\mathbf{p}) \rVert \cos\varphi (The dot product and the angle between vectors), largest exactly when \varphi = 0.
Result 4. Differentiation is linear. By the product rule,
The curl rule is Exercise 14.
Result 5. With P = f_x, Q = f_y, R = f_z, the components of \operatorname{curl} \nabla f are f_{zy} - f_{yz}, f_{xz} - f_{zx} and f_{yx} - f_{xy}, each 0 by Clairaut’s theorem (Partial derivatives). For C^2 fields,
and Clairaut’s theorem makes each bracket 0.
Result 6. (a) is quoted. For (b): since P_x and Q_y are continuous, they may be differentiated under the integral sign in \int_0^z \cdots\,dt (proved in a course on real analysis), and the fundamental theorem of calculus handles the upper limits. With G_3 = 0, \operatorname{curl} \mathbf{G} = \langle -\partial_z G_2,\ \partial_z G_1,\ \partial_x G_2 - \partial_y G_1 \rangle. Here -\partial_z G_2 = P, and \partial_z G_1 = Q since the second integral in G_1 is free of z. Using P_x + Q_y = -R_z,
Worked examples
Reading a field at a point
Let \mathbf{F} = \langle xyz, x^2 - z, y^2 + xz \rangle and \mathbf{p} = (1, 2, -1). Find \operatorname{div} \mathbf{F}(\mathbf{p}) and \operatorname{curl} \mathbf{F}(\mathbf{p}), the axis of fastest turning for a small paddle wheel at \mathbf{p}, and an estimate of the circulation around the circle of radius 0.01 about \mathbf{p} perpendicular to \displaystyle \mathbf{n} = \frac13\langle 2, -1, 2 \rangle, counterclockwise as seen from the tip of \mathbf{n}.
Solution\operatorname{div} \mathbf{F} = yz + 0 + x, which is -1 at \mathbf{p}: near \mathbf{p} more fluid flows in than out, at net rate 1 per unit volume (Result 2). The curl is
By Result 3 the wheel turns fastest with axis \displaystyle \frac{1}{\sqrt{43}}\langle 5, 3, 3 \rangle, circulation density \displaystyle \sqrt{43} \approx 6.56. About \mathbf{n} it is \displaystyle \langle 5, 3, 3 \rangle \cdot \frac13\langle 2, -1, 2 \rangle = \frac{13}{3}, so the circulation is about
A rigidly rotating body
A body turns about an axis through the origin with constant angular velocity \boldsymbol{\omega} = \langle a, b, c \rangle, so its point at \mathbf{r} = \langle x, y, z \rangle has velocity \mathbf{F} = \boldsymbol{\omega} \times \mathbf{r}. Find \operatorname{div} \mathbf{F} and \operatorname{curl} \mathbf{F}.
SolutionExpanding, \mathbf{F} = \langle bz - cy,\ cx - az,\ ay - bx \rangle. Each component is free of its own variable, so \operatorname{div} \mathbf{F} = 0: a rigid motion neither expands nor compresses. The curl is
So the paddle-wheel rate \displaystyle \frac12 \operatorname{curl} \mathbf{F} of the Intuition is exactly the body’s angular velocity, at every point.
Layla’s waterspout
In a meteorology lab, Layla models the horizontal wind of a waterspout whose core of radius 50 meters turns like a rigid body at 0.8 radians per second, the wind speed falling off as 1/r outside:
with \displaystyle r = \sqrt{x^2 + y^2}, in meters per second (Figure 2). Find the divergence and scalar curl inside and outside the core, and the circulation around the circle C of radius 100 meters, counterclockwise, directly and from the curl.
SolutionThe formulas agree on r = 50, each giving speed 40 meters per second. Inside, as in Example 2 with \boldsymbol{\omega} = 0.8\,\mathbf{k}, the divergence is 0 and Q_x - P_y = 0.8 + 0.8 = 1.6 per second. Outside, \mathbf{F} is 2000 times the field \langle -y, x \rangle/(x^2 + y^2) of Conservative fields and potential functions, with scalar curl 0, and
Directly: on C the wind is tangent with speed 2000/100 = 20 meters per second, so the circulation is 20 \cdot 2\pi \cdot 100 = 4000\pi \approx 12{,}566 square meters per second.
From the curl: \mathbf{F} is not C^1 across r = 50, so cut there. On the core, \mathbf{F} is the C^1 field 0.8\langle -y, x \rangle; on the ring 50 \le r \le 100 it is the second formula, C^1 off the origin. Apply Result 1 to each piece (the ring cut by the axes, as in Result 2 of Green's theorem) and add. The circle r = 50 is traced once each way with the same values of \mathbf{F} and cancels, so
the same for every circle outside the core. There the air circles the center, yet a paddle wheel does not turn.
Proof: which radial fields are source-free
Let \mathbf{r} = \langle x, y, z \rangle, \rho = \lVert \mathbf{r} \rVert and k real. Prove that off the origin \operatorname{div}(\rho^k \mathbf{r}) = (k + 3)\rho^k and \operatorname{curl}(\rho^k \mathbf{r}) = \mathbf{0}, and deduce \nabla^2 \rho^m = m(m + 1)\rho^{m - 2}.
SolutionFrom \rho^2 = x^2 + y^2 + z^2, \rho_x = x/\rho, so by the chain rule \nabla \rho^k = k\rho^{k - 2}\,\mathbf{r}. Also \operatorname{div} \mathbf{r} = 3 and \operatorname{curl} \mathbf{r} = \mathbf{0}. By Result 4,
So the only source-free field of this form is the inverse-square field \mathbf{r}/\rho^3 of Vector fields (k = -3). Then \nabla^2 \rho^m = \operatorname{div}(m\rho^{m - 2}\mathbf{r}) = m(m - 2 + 3)\rho^{m - 2}; in particular 1/\rho is harmonic off the origin.
A curl that is not a gradient
Let \mathbf{F} = \langle x^2, -2xy + z, y \rangle. Show that \mathbf{F} is conservative on no open set, and find a vector potential for it.
Solution\operatorname{curl} \mathbf{F} = \langle 1 - 1,\ 0 - 0,\ -2y - 0 \rangle = \langle 0, 0, -2y \rangle. Every open set contains points with y \ne 0, where the curl is not \mathbf{0}, so Result 6(a) rules out a potential. The divergence is 2x - 2x + 0 = 0, so by Result 6(b)
Check: \mathbf{G} = \langle G_1, G_2, 0 \rangle has curl \langle x^2,\ -2xy + z,\ -2xz - (-2xz - y) \rangle, which is \mathbf{F}.
Boundaries
A zero at every point is not a global statement. \langle -y, x, 0 \rangle/(x^2 + y^2) has curl \mathbf{0} off the z-axis but circulation 2\pi around the unit circle, so it is not conservative on that set, which is not simply connected (Exercise 16 of Conservative fields and potential functions). Likewise \mathbf{r}/\rho^3 has divergence 0 off the origin, yet its outward flux through every sphere centered at the origin is 4\pi, a computation for Parametric surfaces and surface integrals; a curl has zero flux out of every sphere in its domain (Stokes' theorem), so \mathbf{r}/\rho^3 has no vector potential there.
Results 2 and 3 need C^1. The scalar curl in Example 3 is 1.6 on one side of r = 50 and 0 on the other. At a point of that circle, the argument of Example 3 shows that the circulation per unit area around small circles tends to 0.8, neither one-sided value.
Where this goes next
Stokes' theorem integrates Result 3 over a surface (Parametric surfaces and surface integrals), and The divergence theorem integrates Result 2(b) over a solid. Harmonic functions continues Example 4.
Exercises
Divergence and curl at a point
- Find the divergence and curl of \mathbf{F} = \langle xy^2, yz^2, zx^2 \rangle at (1, -1, 2).
- Find the divergence and curl of \mathbf{F} = \langle e^x \sin y, e^x \cos y, z \rangle, and a potential function for it.
- For the plane field \mathbf{F} = \langle xe^y, -ye^x \rangle, find the divergence and the scalar curl at (0, 1), and say what each tells you about the flow there.
- Show that \mathbf{F} = \langle -3y + 2z, 3x - z, -2x + y \rangle is \boldsymbol{\omega} \times \mathbf{r} for a constant \boldsymbol{\omega}, and find its curl, its divergence and the angular speed.
- Let \mathbf{F} = \langle z - y, x + z^2, xy \rangle and \mathbf{p} = (1, 0, 1). (a) Find the largest circulation density at \mathbf{p} and the unit axis giving it. (b) Estimate the circulation around the circle of radius 0.02 about \mathbf{p} perpendicular to \displaystyle \mathbf{n} = \frac13\langle 2, 1, 2 \rangle, counterclockwise as seen from the tip of \mathbf{n}.
- For \mathbf{F} = \langle x^3 - y, xy \rangle, estimate the outward flux across the circle of radius 0.05 about (1, 2) with Result 2, then find it exactly with Result 1.
Flows in the lab
- Samir models a smaller waterspout as in Example 3, with a core of radius 30 meters and wind speed 45 meters per second at its edge. Find the core’s angular velocity, the curl in the core, the wind speed 90 meters from the center, and the circulation around the circle of radius 90 meters, counterclockwise.
- Near the ground, Samir measures a wind \mathbf{F} = \langle 0.02z, 0, 0 \rangle meters per second at height z meters. Find its divergence and curl, and the axis and rate at which a small paddle wheel turns fastest.
- Layla models a downburst spreading over the ground by \mathbf{F} = \langle x, y, -2z \rangle. Show that it is incompressible and irrotational, find a potential f, and show that f is harmonic.
Potentials and the Laplacian
- Decide whether each field is a gradient on \mathbb{R}^3 and whether it is a curl on \mathbb{R}^3: (a) \langle x, y, z \rangle; (b) \langle y, z, x \rangle; (c) \langle yz, xz, xy \rangle.
- Use Result 6(b) to find a vector potential for \langle y, z, x \rangle, and give a second one.
- Find the Laplacian and say which are harmonic: (a) e^{3x}\cos 3y; (b) \ln(x^2 + y^2), off the origin of the plane; (c) \dfrac{1}{x^2 + y^2 + z^2}; (d) x^2 + y^2 - 2z^2 + xyz.
- In the plane, let \displaystyle r = \sqrt{x^2 + y^2}. Show that \operatorname{div}(r^k\langle x, y \rangle) = (k + 2)r^k off the origin. For k = -2, find the outward flux across the circle of radius a, and explain why it does not contradict Result 1.
Proofs
- Prove that \operatorname{curl}(f\mathbf{F}) = \nabla f \times \mathbf{F} + f \operatorname{curl} \mathbf{F} for a C^1 function f and a C^1 field \mathbf{F}.
- Prove that \operatorname{div}(\mathbf{F} \times \mathbf{G}) = \mathbf{G} \cdot \operatorname{curl} \mathbf{F} - \mathbf{F} \cdot \operatorname{curl} \mathbf{G} for C^1 fields \mathbf{F} and \mathbf{G}.
- Prove that \nabla f \times \nabla g is source-free when f and g are C^2.
- Prove that \operatorname{curl} \operatorname{curl} \mathbf{F} = \nabla(\operatorname{div} \mathbf{F}) - \nabla^2 \mathbf{F} for a C^2 field, where \nabla^2 \mathbf{F} = \langle \nabla^2 P, \nabla^2 Q, \nabla^2 R \rangle.
- Let \mathbf{F} be C^1 on an open set U \subseteq \mathbb{R}^3. Prove: (a) if the circulation of \mathbf{F} around every circle in U is 0, then \operatorname{curl} \mathbf{F} = \mathbf{0} on U; (b) if the outward flux of \mathbf{F} out of every cube in U with edges parallel to the axes is 0, then \operatorname{div} \mathbf{F} = 0 on U.
Answers
- Divergence 6, curl \langle 4, -4, 2 \rangle, from \operatorname{div} \mathbf{F} = y^2 + z^2 + x^2 and \operatorname{curl} \mathbf{F} = \langle -2yz, -2zx, -2xy \rangle.
- Divergence e^x\sin y - e^x\sin y + 1 = 1; curl \langle 0, 0, e^x\cos y - e^x\cos y \rangle = \mathbf{0}. A potential is \displaystyle f = e^x\sin y + \frac12 z^2.
- Divergence e^y - e^x = e - 1 > 0: the flow spreads out there. Scalar curl -ye^x - xe^y = -1 < 0: a small paddle wheel turns clockwise.
- \boldsymbol{\omega} = \langle 1, 2, 3 \rangle, matching Example 2’s \langle bz - cy, cx - az, ay - bx \rangle with c = 3, b = 2, a = 1. Curl \langle 2, 4, 6 \rangle, divergence 0, angular speed \displaystyle \sqrt{14} \approx 3.74.
- \operatorname{curl} \mathbf{F} = \langle x - 2z, 1 - y, 2 \rangle = \langle -1, 1, 2 \rangle at \mathbf{p}. (a) \displaystyle \sqrt6 \approx 2.45, about \displaystyle \frac{1}{\sqrt6}\langle -1, 1, 2 \rangle. (b) The density about \mathbf{n} is \displaystyle \langle -1, 1, 2 \rangle \cdot \frac13\langle 2, 1, 2 \rangle = 1, so about \pi(0.02)^2 = 0.0004\pi \approx 1.26 \times 10^{-3}.
- \operatorname{div} \mathbf{F} = 3x^2 + x, which is 4 at (1, 2): about 4\pi(0.05)^2 = 0.01\pi \approx 0.0314. Exactly: with x = 1 + u the divergence is 4 + 7u + 3u^2; over the disk of radius \varepsilon, u integrates to 0 and u^2 to \displaystyle \frac{\pi\varepsilon^4}{4}, so Result 1 gives \displaystyle \pi\varepsilon^2\left(4 + \frac{3\varepsilon^2}{4}\right) = 0.0100046875\,\pi \approx 0.031431.
- Angular velocity 45/30 = 1.5 radians per second; curl 3\,\mathbf{k} per second. Outside, \mathbf{F} = 1350\,\langle -y, x \rangle/(x^2 + y^2), so the speed at 90 meters is 1350/90 = 15 meters per second, and the circulation is 15 \cdot 2\pi \cdot 90 = 2700\pi \approx 8482 square meters per second, also 3 \cdot \pi(30)^2.
- Divergence 0; curl \langle 0, 0.02, 0 \rangle per second. The axis is horizontal along the y-axis, across the wind, and the rate \displaystyle \frac12 \cdot 0.02 = 0.01 radians per second, counterclockwise as seen from the positive y-axis, since the wind is faster above.
- \operatorname{div} \mathbf{F} = 1 + 1 - 2 = 0, and each component depends only on its own variable, so \operatorname{curl} \mathbf{F} = \mathbf{0}. A potential is \displaystyle f = \frac12(x^2 + y^2 - 2z^2), and \nabla^2 f = \operatorname{div} \mathbf{F} = 0.
- (a) A gradient, of \displaystyle \frac12(x^2 + y^2 + z^2); not a curl, since its flux out of the unit sphere is 4\pi, a curl’s is 0 (Boundaries). (b) Not a gradient, since its curl is \langle -1, -1, -1 \rangle; a curl, since its divergence is 0 (Result 6(b)). (c) Both: it is \nabla(xyz), with divergence 0.
- \displaystyle \mathbf{G} = \langle \frac12 z^2 - xy,\ -yz,\ 0 \rangle, from G_1 = \int_0^z t\,dt - \int_0^y x\,ds and G_2 = -\int_0^z y\,dt; its curl is \langle y, z, 0 - (-x) \rangle. A second is \displaystyle \mathbf{G} + \nabla(xyz) = \langle \frac12 z^2 - xy + yz,\ -yz + xz,\ xy \rangle, by Result 5.
- (a) 9e^{3x}\cos 3y - 9e^{3x}\cos 3y = 0, harmonic. (b) \displaystyle \frac{2(y^2 - x^2) + 2(x^2 - y^2)}{(x^2 + y^2)^2} = 0, harmonic. (c) Example 4 with m = -2 gives \displaystyle \frac{2}{(x^2 + y^2 + z^2)^2}, not harmonic. (d) 2 + 2 - 4 + 0 = 0, harmonic.
- As in Example 4, \nabla r^k = kr^{k - 2}\langle x, y \rangle and \operatorname{div}\langle x, y \rangle = 2, so Result 4 gives kr^k + 2r^k. For k = -2, \mathbf{F} \cdot \mathbf{n} = 1/a on the circle, so the flux is 2\pi for every a, though the divergence is 0. Result 1 needs \mathbf{F} to be C^1 on the whole disk, and it is undefined at the center.
- The \mathbf{i}-component of \operatorname{curl}(f\mathbf{F}) is (fR)_y - (fQ)_z = (f_yR - f_zQ) + f(R_y - Q_z) by the product rule: the \mathbf{i}-component of \nabla f \times \mathbf{F} plus f times that of \operatorname{curl} \mathbf{F}. The cyclic change x \to y \to z \to x, P \to Q \to R \to P carries each component formula of the curl and of the cross product to the next, giving the other two.
- \mathbf{F} \times \mathbf{G} has components F_2G_3 - F_3G_2, F_3G_1 - F_1G_3 and F_1G_2 - F_2G_1. Apply the product rule to \partial_x of the first, \partial_y of the second and \partial_z of the third. The terms with derivatives of \mathbf{F} are G_1(\partial_yF_3 - \partial_zF_2), G_2(\partial_zF_1 - \partial_xF_3) and G_3(\partial_xF_2 - \partial_yF_1), adding to \mathbf{G} \cdot \operatorname{curl} \mathbf{F}. Those with derivatives of \mathbf{G} are -F_1(\partial_yG_3 - \partial_zG_2), -F_2(\partial_zG_1 - \partial_xG_3) and -F_3(\partial_xG_2 - \partial_yG_1), adding to -\mathbf{F} \cdot \operatorname{curl} \mathbf{G}.
- \nabla f and \nabla g are C^1, so by Exercise 15, \operatorname{div}(\nabla f \times \nabla g) = \nabla g \cdot \operatorname{curl} \nabla f - \nabla f \cdot \operatorname{curl} \nabla g, and both curls are \mathbf{0} by Result 5.
- With B = P_z - R_x and C = Q_x - P_y, the \mathbf{i}-component of \operatorname{curl} \operatorname{curl} \mathbf{F} is C_y - B_z = Q_{xy} - P_{yy} - P_{zz} + R_{xz}. That of the right side is (P_x + Q_y + R_z)_x - \nabla^2 P, which is Q_{yx} + R_{zx} - P_{yy} - P_{zz}. Clairaut’s theorem gives Q_{xy} = Q_{yx} and R_{xz} = R_{zx}. The cyclic change of Exercise 14 gives the other components.
- (a) Let \mathbf{p} \in U and \mathbf{n} a unit vector. For small \varepsilon the circle C_\varepsilon of Result 3 lies in U, so its circulation is 0, and Result 3 gives \operatorname{curl} \mathbf{F}(\mathbf{p}) \cdot \mathbf{n} = 0. Taking \mathbf{n} = \mathbf{i}, \mathbf{j}, \mathbf{k} gives \operatorname{curl} \mathbf{F}(\mathbf{p}) = \mathbf{0}. (b) For small h the cube B_h of Result 2(b) lies in U, so the flux over h^3 is 0, and its limit \operatorname{div} \mathbf{F}(\mathbf{p}) is 0.