The divergence theorem
Green’s theorem and Stokes’ theorem turn an integral around an edge into an integral over what the edge encloses. A closed surface has no edge. What does the flux of a field out through it equal?
Where you are 8 in Vector calculus
Why this exists
Result 2(b) of Divergence and curl showed that the outward flux of a C^1 field through the six faces of a box equals the integral of its divergence over the box. Boxes never fit a ball or a cone exactly. The divergence theorem extends the box result to solids bounded by closed piecewise smooth surfaces: the net rate at which a field flows out through the surface equals the total of its sources inside. It joins Green's theorem and Stokes' theorem, one dimension up.
Definition
Write \mathbf{F} = \langle P, Q, R \rangle, with C^1 as in Divergence and curl. As in Triple integrals, a solid is z-simple if
where D is a closed region of the xy-plane that is type I, type II or radially simple, and u_1 \le u_2 are continuous functions of (x, y) on D.
- E is elementary in the z-direction if, in addition, its top S_2 (the graph of u_2) and its bottom S_1 (the graph of u_1) are each made of finitely many smooth parametric surfaces (Parametric surfaces and surface integrals) \mathbf{r}(s, t) = \langle x, y, u_i(x, y) \rangle with (x, y) = T(s, t), (s, t) in a region \Omega, where the shadow map T is C^1 on an open set containing \Omega, one-to-one on the interior of \Omega with Jacobian \dfrac{\partial(x, y)}{\partial(s, t)} > 0 there, and the shadows T(\Omega) of the pieces cut D into regions. If u_i has continuous first partial derivatives on an open set containing D, the graph \langle x, y, u_i(x, y) \rangle itself qualifies, its shadow map the identity. The rest of \partial E, the points over \partial D with u_1 \le z \le u_2, is the wall S_3 (Figure 1).
- Elementary in the x- or y-direction is defined the same way with the letters exchanged, and E is elementary if it is elementary in all three directions. Balls, ellipsoids, boxes, tetrahedra, solid cylinders and solid cones are elementary; Example 2 checks a half-ball, and the others are alike.
- \partial E is oriented by the outward unit normal \mathbf{n}, which points up on the top and down on the bottom. The outward flux \iint_{\partial E} \mathbf{F} \cdot d\mathbf{S} is the sum of the fluxes through the smooth pieces of \partial E, each computed by Result 2 of Parametric surfaces and surface integrals with a parametrization compatible with \mathbf{n}. It does not depend on how \partial E is cut into pieces (Result 5 there and additivity; used here without further proof).
- A solid is cut into elementary solids E_1, \dots, E_k if it is their union and any two meet at most along faces. Its boundary \partial E consists of the faces that belong to only one E_i. A face between the solid and a cavity inside it is part of \partial E, and there the outward normal points into the cavity.
Intuition
Think of \mathbf{F} as the velocity of a fluid, and cut E into small boxes. By Result 2(b) of Divergence and curl, the fluid leaving a box through its faces is about \operatorname{div} \mathbf{F}\,\Delta V, the fluid produced inside it. Add over the boxes. A face shared by two boxes has opposite outward normals in them, so what one box loses through it the other gains. Only the outside faces remain: the total outflow through the surface is the total production inside.
Boxes do not fit a curved surface, so the proof takes one direction at a time. The part R\,\mathbf{k} of the field crosses only the top and the bottom, since the wall’s normal is horizontal, and top minus bottom is an integral of R_z along each vertical segment (Figure 1).
Results
Result 1 (the divergence theorem). Let E be an elementary solid, or a solid cut into elementary solids, with \partial E oriented outward, and let \mathbf{F} be C^1 on an open set containing E. Then
It is also called Gauss’s theorem. It holds for every bounded solid whose boundary is a piecewise smooth closed surface; that general case is part of the Stokes theorem on manifolds, proved in a course on analysis on manifolds.
Result 2 (one direction at a time). Let E be elementary in the z-direction and let R be C^1 on an open set containing E. Then
The same holds with P\,\mathbf{i} and P_x in the x-direction, and with Q\,\mathbf{j} and Q_y in the y-direction.
Result 3 (divergence is outflow per unit volume). Let \mathbf{F} be C^1 on an open set containing \mathbf{p}, and let B_\varepsilon be the closed ball of radius \varepsilon about \mathbf{p}. Then
Result 4 (cavities and point sources). Let E be as in Result 1.
(a) Let \partial E consist of an outer surface S and the surfaces S_1, \dots, S_m of cavities inside E, and let \mathbf{F} be C^1 on an open set containing E with \operatorname{div} \mathbf{F} = 0 on E. Then the flux out through S equals the sum of the fluxes through the S_j, each oriented by the normal pointing out of its cavity.
(b) (Gauss’s law) Let \mathbf{r} = \langle x, y, z \rangle, \rho = \lVert \mathbf{r} \rVert and \mathbf{F} = \mathbf{r}/\rho^3, and let the origin not lie on \partial E. The flux of \mathbf{F} out of \partial E is 0 if the origin is outside E. It is 4\pi if the origin is inside E and, for small \varepsilon, E with the open ball of radius \varepsilon about the origin removed is also cut into elementary solids.
Why it is true
Result 2. Take a piece of the top, \mathbf{r}(s, t) = \langle x, y, u_2(x, y) \rangle with (x, y) = T(s, t) on \Omega. The \mathbf{k}-component of \mathbf{r}_s \times \mathbf{r}_t is x_s y_t - x_t y_s = \dfrac{\partial(x, y)}{\partial(s, t)} > 0, so \mathbf{r} is compatible with the upward, outward normal. By Result 2 of Parametric surfaces and surface integrals, then Theorem 2 of Change of variables and the Jacobian applied to T,
Adding the pieces, the top contributes \iint_D R(x, y, u_2(x, y))\,dA. On the bottom the outward normal points down, opposite to the compatible one, so the bottom contributes -\iint_D R(x, y, u_1(x, y))\,dA. The wall contributes 0: over a piece of \partial D traced by (x(t), y(t)) it is parametrized by (t, z) \mapsto \langle x(t), y(t), z \rangle, with normal \langle x', y', 0 \rangle \times \mathbf{k} = \langle y', -x', 0 \rangle, so \mathbf{k} \cdot \mathbf{n} = 0. By the fundamental theorem of calculus and Theorem 4 of Triple integrals,
The other two directions are alike, with the letters exchanged.
Result 1. For an elementary solid, \mathbf{F} \cdot \mathbf{n} is the sum of P\,\mathbf{i} \cdot \mathbf{n}, Q\,\mathbf{j} \cdot \mathbf{n} and R\,\mathbf{k} \cdot \mathbf{n}, and Result 2 in the three directions gives \iiint_E (P_x + Q_y + R_z)\,dV. For a solid cut into E_1, \dots, E_k, apply this to each piece and add. A face shared by E_i and E_j carries opposite outward normals, so its two fluxes cancel and the flux out of \partial E remains. The triple integrals add to the one over E, since the pieces share only boundary points (Result 2 of Triple integrals).
Result 3. The plane z = p_3 cuts B_\varepsilon into two half-balls, elementary as in Example 2, so the flux equals \iiint_{B_\varepsilon} g\,dV with g = \operatorname{div} \mathbf{F}, continuous at \mathbf{p}. Given \eta > 0, \lvert g - g(\mathbf{p}) \rvert < \eta on B_\varepsilon for small \varepsilon, so the flux divided by the volume \displaystyle \frac43\pi\varepsilon^3 differs from g(\mathbf{p}) by at most \eta.
Result 4. (a) By Result 1 the flux out of \partial E is \iiint_E 0\,dV = 0. On each S_j the outward normal of E is the opposite of the normal pointing out of the cavity, so 0 is the flux through S minus the sum of the fluxes through the S_j.
(b) \mathbf{F} is C^1 on the open set \mathbb{R}^3 \setminus \{\mathbf{0}\} with divergence 0 there (Example 4 of Divergence and curl). If the origin is outside E, that set contains E, and Result 1 gives 0. If it is inside, then for small \varepsilon removing the open ball of radius \varepsilon leaves a solid with a cavity. By (a) the flux out of \partial E equals the flux out of the sphere \rho = \varepsilon, where \mathbf{n} = \mathbf{r}/\varepsilon and \mathbf{F} \cdot \mathbf{n} = 1/\varepsilon^2, which is 4\pi\varepsilon^2/\varepsilon^2 = 4\pi.
Worked examples
Both sides on a bowl
Verify the divergence theorem for \mathbf{F} = \langle 2x, -y, z^2 \rangle on the solid E between the paraboloid z = x^2 + y^2 and the plane z = 4.
SolutionE is z-simple over the disk x^2 + y^2 \le 4, with u_1 = x^2 + y^2 and u_2 = 4, and is elementary as in the Definition. The divergence is 2 - 1 + 2z = 1 + 2z, and in cylindrical coordinates (Cylindrical and spherical coordinates)
Now the flux. The lid z = 4 has \mathbf{n} = \mathbf{k} and \mathbf{F} \cdot \mathbf{k} = 16 over area 4\pi, giving 64\pi. The bowl is the graph of g = x^2 + y^2, whose upward normal is \langle -2x, -2y, 1 \rangle (Result 3 of Parametric surfaces and surface integrals); the outward normal points down, so use \langle 2x, 2y, -1 \rangle. Then \mathbf{F} \cdot \langle 2x, 2y, -1 \rangle = 4x^2 - 2y^2 - z^2 with z = r^2, and
The total is \displaystyle 64\pi - \frac{40\pi}{3} = \frac{152\pi}{3}, as Result 1 requires.
Closing a hemisphere
Find the upward flux of
through the hemisphere S: x^2 + y^2 + z^2 = 9, z \ge 0.
SolutionThe integrand on S is full of e^{xz} and \ln(1 + y^2), so close S with the disk B: x^2 + y^2 \le 9, z = 0; together they bound the half-ball E.
E is elementary. In the z-direction its bottom is flat, and its top has the spherical parametrization of Result 4 of Parametric surfaces and surface integrals on \displaystyle [0, \frac{\pi}{2}] \times [0, 2\pi], whose shadow map (3\sin\varphi\cos\theta, 3\sin\varphi\sin\theta) is one-to-one on the interior with Jacobian 9\sin\varphi\cos\varphi > 0 there. In the x-direction the front face \displaystyle x = \sqrt{9 - y^2 - z^2} is the part with \displaystyle -\frac{\pi}{2} \le \theta \le \frac{\pi}{2}, with shadow (y, z) = (3\sin\varphi\sin\theta, 3\cos\varphi) and Jacobian 9\sin^2\varphi\cos\theta > 0 inside; the back face (with the parameters swapped) and the y-direction are alike.
The divergence is z^2 + x^2 + y^2 = \rho^2, so by Result 1
On B the outward normal is -\mathbf{k} and \mathbf{F} \cdot (-\mathbf{k}) = -3, so B carries -27\pi. Hence the flux through S is \displaystyle \frac{486\pi}{5} + 27\pi = \frac{621\pi}{5} \approx 390.2.
Naomi’s haystack
Naomi and Hiro build a conical haystack of damp grass, with base radius 1.5 meters and height 1 meter; microbes breaking the grass down heat it. Hiro models the heat produced as h = 60 - 40z watts per cubic meter at height z meters, and measures that the ground under the stack absorbs heat at 5 watts per square meter. Once the temperatures in the stack stop changing, at what rate does heat leave through its sloping surface?
SolutionLet \mathbf{q} be the heat flux, in watts per square meter, assumed C^1. With temperatures steady, the heat leaving any small ball B_\varepsilon in the stack equals the heat produced in it. Divide both by the volume and let \varepsilon \to 0^+: Result 3 and the continuity argument in its proof give \operatorname{div} \mathbf{q} = h.
By Result 1, the heat leaving the cone E is \iiint_E h\,dV. The column at distance r from the axis has height \displaystyle H = 1 - \frac{2r}{3} and \int_0^H (60 - 40z)\,dz = 60H - 20H^2. With r = 1.5(1 - H),
The ground takes 5 watts per square meter over the base of area 2.25\pi, that is 11.25\pi watts. The sloping surface releases the rest, 37.5\pi - 11.25\pi = 26.25\pi \approx 82.47 watts.
A point source in a cube
Let \mathbf{F} = \mathbf{r}/\rho^3. Find its outward flux through the boundary of the cube [-1, 1]^3 and of the box [1, 3] \times [-1, 1] \times [-1, 1], and deduce \displaystyle \displaystyle\int_{-1}^{1} \int_{-1}^{1} \frac{dy\,dz}{(1 + y^2 + z^2)^{3/2}}.
SolutionThe origin is inside the cube. For 0 < \varepsilon < 1, the coordinate planes cut the cube minus the open ball of radius \varepsilon into eight pieces like [0, 1]^3 minus an eighth of the ball, each elementary (the curved face is parametrized as in Example 2). Result 4(b) gives 4\pi.
On the face x = 1, \mathbf{n} = \mathbf{i} and \mathbf{F} \cdot \mathbf{n} = (1 + y^2 + z^2)^{-3/2}. On x = -1, \mathbf{n} = -\mathbf{i} and -x/\rho^3 is the same function of (y, z), and each other face gives the same function of its two free coordinates. The six fluxes are equal, each \displaystyle \frac{4\pi}{6}:
The box does not contain the origin, so its outward flux is 0: the flux \displaystyle \frac{2\pi}{3} entering through its face x = 1 leaves through its other five faces.
Proof: Green’s first identity
Let E be as in Result 1, f be C^1 and g be C^2 on an open set containing E. Prove that
and in particular that the outward flux of \nabla g is \iiint_E \nabla^2 g\,dV.
Solution\nabla g is C^1, so f\,\nabla g is C^1. By Result 4 of Divergence and curl, \operatorname{div}(f\,\nabla g) = \nabla f \cdot \nabla g + f \operatorname{div} \nabla g, and \operatorname{div} \nabla g = \nabla^2 g. Result 1 applied to f\,\nabla g gives the identity, and f = 1, with \nabla f = \mathbf{0}, gives the particular case.
Boundaries
\mathbf{F} must be C^1 on all of E. On the unit ball, \mathbf{r}/\rho^3 has divergence 0 at every point but the origin, where it is undefined, yet its outward flux is 4\pi. Result 4(b) accounts for the missing point.
The surface must be closed and oriented outward. For a surface with a rim, Result 1 says nothing by itself, so Example 2 and Exercise 6 close it first. With the inward normal the left side changes sign and the right side does not.
Elementary is a hypothesis of the proof, not of the theorem. A solid torus around the z-axis is not elementary, since a line parallel to the x-axis can cross it twice, but the planes x = 0 and y = 0 cut it into four pieces that are.
Where this goes next
The plane version is the flux form of Green’s theorem (Result 4 of Green's theorem). Green’s theorem, Stokes’ theorem, the divergence theorem and the fundamental theorem for line integrals (Conservative fields and potential functions) are cases of one theorem, \int_{\partial M} \omega = \int_M d\omega, for a manifold M with boundary. In physics, Result 4(b) times a charge is Gauss’s law, and the argument of Example 3 turns a conservation law into an equation for a divergence. Harmonic functions builds on Green’s identities (Example 5, Exercises 14 and 15).
Exercises
Both sides of the theorem
- Verify the divergence theorem for \mathbf{F} = \langle x, 2y, 3z \rangle on the solid cylinder x^2 + y^2 \le 9, 0 \le z \le 2, computing the flux through the wall, the top and the bottom.
- Verify the divergence theorem for \mathbf{F} = \langle xz, yz, z \rangle on the half-ball x^2 + y^2 + z^2 \le 1, z \ge 0.
Choosing the easier side
- Find the outward flux of \mathbf{F} = \langle xy^2 + \cos z,\ x^2y + e^{-x},\ \sin(xy) \rangle through the sphere x^2 + y^2 + z^2 = 4.
- Find the outward flux of \mathbf{F} = \langle 2x + yz,\ 3y - xz^2,\ z + x^2y \rangle through the ellipsoid \displaystyle \frac{x^2}{4} + \frac{y^2}{9} + z^2 = 1.
- A can without a bottom is the wall x^2 + y^2 = 1, 0 \le z \le 2, together with the top disk at z = 2. Find the outward flux of \mathbf{F} = \langle x^3 + \sin(yz),\ y^3 + \cos(xz),\ z + 1 \rangle through it.
- Show that \mathbf{F} = \langle y^2 + z,\ xz,\ x^2 + 1 \rangle is source-free, and find its upward flux through the hemisphere x^2 + y^2 + z^2 = 4, z \ge 0.
More haystacks
- Naomi builds a second stack in the shape of the half-ball x^2 + y^2 + z^2 \le 1, z \ge 0, in meters, which produces 50(1 - \rho) watts per cubic meter at distance \rho from the center of its base. In the steady state, find the total rate at which heat leaves it, through the air and the ground together.
- Hiro stands a cylindrical round bale of radius 0.6 meters and height 1 meter on end. It produces 80 watts per cubic meter throughout, and the ground absorbs 10 watts per square meter through its base. In the steady state, find the rate at which heat leaves through its top and its curved side.
Point sources and cavities
- For \mathbf{F} = \mathbf{r}/\rho^3, find the outward flux through (a) the ellipsoid \displaystyle \frac{x^2}{4} + \frac{y^2}{9} + \frac{z^2}{16} = 1; (b) the sphere (x - 3)^2 + y^2 + z^2 = 1.
- Let \mathbf{p} = (0, 0, 2) and \mathbf{F} = \dfrac{\mathbf{r}}{\rho^3} - \dfrac{2(\mathbf{r} - \mathbf{p})}{\lVert \mathbf{r} - \mathbf{p} \rVert^3}, two point sources of opposite sign. Find the outward flux through the spheres \rho = 1 and \rho = 3.
- Verify the divergence theorem for \mathbf{F} = \mathbf{r}/\rho on the shell 1 \le \rho \le 2, computing the flux through both of its spheres.
- For \mathbf{F} = \langle x^2y, yz, xz^2 \rangle, estimate the outward flux through the sphere of radius 0.1 about (1, 2, -1) with Result 3, then find it exactly.
Proofs
- Prove that a solid E as in Result 1 has volume \displaystyle \frac13 \iint_{\partial E} \mathbf{r} \cdot d\mathbf{S}, and use it to find the volume of the solid cone with apex at the origin, axis along the positive z-axis, height h and base radius a.
- (Green’s second identity.) For E as in Result 1 and f, g C^2 on an open set containing E, prove that
- Let E be elementary, and let u and v be C^2 on an open set containing E, harmonic on E, with u = v on \partial E. Prove that u = v on all of E.
- For E as in Result 1, prove: (a) if \mathbf{G} is C^2 on an open set containing E, the outward flux of \operatorname{curl} \mathbf{G} through \partial E is 0; (b) if u is C^2 there and harmonic, the outward flux of \nabla u is 0.
- (a) For E as in Result 1 and f C^1 on an open set containing E, prove that \iint_{\partial E} f\,\mathbf{n}\,dS = \iiint_E \nabla f\,dV, both vector integrals taken componentwise. (b) A body fills E under water whose surface is the plane z = 0. The pressure is p = -\delta g z, with \delta the density of water and g the acceleration of gravity, and pushes on each patch of \partial E with force -p\,\mathbf{n}\,dS. Prove that the total force is \delta g\,V(E)\,\mathbf{k} (Archimedes’ principle).
Answers
- 108\pi both ways. The divergence is 1 + 2 + 3 = 6 and the volume 18\pi. On the wall \langle 3\cos\theta, 3\sin\theta, z \rangle the outward normal is \langle 3\cos\theta, 3\sin\theta, 0 \rangle and \mathbf{F} \cdot \mathbf{N} = 9\cos^2\theta + 18\sin^2\theta, giving 27\pi \cdot 2 = 54\pi. The top has \mathbf{F} \cdot \mathbf{k} = 6 over area 9\pi, giving 54\pi; on the bottom z = 0, so 0.
- \displaystyle \frac{7\pi}{6} both ways. The divergence is 2z + 1; \displaystyle \iiint z\,dV = 2\pi \cdot \frac14 \cdot \frac12 = \frac{\pi}{4} and the volume is \displaystyle \frac{2\pi}{3}, so \displaystyle \frac{\pi}{2} + \frac{2\pi}{3}. On the dome \mathbf{n} = \mathbf{r}, so \mathbf{F} \cdot \mathbf{n} = z(x^2 + y^2) + z^2, which is \cos\varphi\sin^2\varphi + \cos^2\varphi; with dS = \sin\varphi\,d\varphi\,d\theta the flux is \displaystyle 2\pi\left( \frac14 + \frac13 \right). On the base \mathbf{F} = \mathbf{0}.
- \displaystyle \frac{256\pi}{15}. The divergence is y^2 + x^2 = \rho^2\sin^2\varphi, and over the ball of radius 2 the integral is \displaystyle 2\pi \cdot \frac{32}{5} \cdot \frac43.
- 48\pi. The divergence is 2 + 3 + 1 = 6, and the ellipsoid has volume \displaystyle \frac43\pi \cdot 2 \cdot 3 \cdot 1 = 8\pi.
- 6\pi. Close the can with the bottom disk to get the solid cylinder. The divergence is 3x^2 + 3y^2 + 1, and \int_0^{2\pi} \int_0^1 (3r^2 + 1)\,r\,dr\,d\theta \cdot 2 = 5\pi. On the bottom, \mathbf{n} = -\mathbf{k} and \mathbf{F} \cdot \mathbf{n} = -1, giving -\pi, so the can carries 5\pi - (-\pi).
- 8\pi. Each component is free of its own variable, so the divergence is 0. By Result 1 on the half-ball, the flux up through the hemisphere equals the flux up through the disk x^2 + y^2 \le 4, z = 0, where \mathbf{F} \cdot \mathbf{k} = x^2 + 1. The term x^2 integrates to \int_0^{2\pi} \int_0^2 r^3\cos^2\theta\,dr\,d\theta = 4\pi and the term 1 to the area 4\pi.
- \displaystyle \frac{25\pi}{3} \approx 26.18 watts. As in Example 3, the heat flux has divergence 50(1 - \rho). In spherical coordinates the angles give 2\pi, so the total is \displaystyle 2\pi \int_0^1 50(1 - \rho)\rho^2\,d\rho = 100\pi\left( \frac13 - \frac14 \right).
- 25.2\pi \approx 79.17 watts. The bale produces 80 \cdot 0.36\pi \cdot 1 = 28.8\pi watts, and the ground takes 10 \cdot 0.36\pi = 3.6\pi.
- (a) 4\pi. The origin is inside, and the coordinate planes cut the ellipsoid minus a small ball into eight elementary pieces, so Result 4(b) applies. (b) 0: the origin is outside the ball.
- 4\pi and -4\pi. Apply Result 4(b) to each term, to the second after moving the origin to \mathbf{p}. The ball \rho \le 1 contains the origin but not \mathbf{p}, so only the first term counts. The ball \rho \le 3 contains both: 4\pi - 2 \cdot 4\pi.
- 12\pi both ways. By Result 4 of Divergence and curl, the divergence is \nabla(1/\rho) \cdot \mathbf{r} + 3/\rho = 2/\rho, whose integral over the shell is 2 \cdot 4\pi \int_1^2 \rho\,d\rho = 12\pi; the coordinate planes cut the shell into eight elementary pieces. On a sphere of radius a, \displaystyle \mathbf{F} \cdot \frac{\mathbf{r}}{a} = 1. The outer sphere gives 16\pi; on the inner one the outward normal of the shell points to the origin, giving -4\pi.
- Both \displaystyle \frac{\pi}{750} \approx 0.004189. The divergence is 2xy + z + 2xz, which is 4 - 1 - 2 = 1 at the center, so the estimate is \displaystyle 1 \cdot \frac43\pi(0.1)^3. With x = 1 + u, y = 2 + v, z = -1 + w the divergence is 1 + 2u + 2v + 3w + 2uv + 2uw, and every term but 1 is odd in u, v or w, so its integral over the ball about the center is 0. The estimate is exact.
- \operatorname{div} \mathbf{r} = 3, so Result 1 gives \iint_{\partial E} \mathbf{r} \cdot d\mathbf{S} = 3V(E). The cone is elementary. Its side is t\langle a\cos\theta, a\sin\theta, h \rangle, 0 \le t \le 1, where \mathbf{r} = t\,\mathbf{r}_t is tangent to the surface, so \mathbf{r} \cdot \mathbf{n} = 0 there. On the base z = h, \mathbf{n} = \mathbf{k} and \mathbf{r} \cdot \mathbf{n} = h over area \pi a^2. So \displaystyle V = \frac13 \pi a^2 h.
- Example 5 applies to the pair f, g and, since f is also C^2, to the pair g, f. Subtract the second identity from the first: the terms \nabla f \cdot \nabla g and \nabla g \cdot \nabla f cancel, leaving the claim.
- Let w = u - v, which is C^2, harmonic on E and 0 on \partial E. In Example 5 with f = g = w the left side is 0, since w = 0 on \partial E, and \nabla^2 w = 0, so \iiint_E \lVert \nabla w \rVert^2\,dV = 0. If \lVert \nabla w(\mathbf{q}) \rVert^2 = c > 0 at an interior point \mathbf{q}, then by continuity it exceeds \displaystyle \frac{c}{2} on a small ball about \mathbf{q} inside E, and the integral is positive, a contradiction. So \nabla w = \mathbf{0} at interior points of E. E is z-simple: each point (x, y, z) of E lies on the segment from (x, y, u_1(x, y)), in \partial E, to (x, y, u_2(x, y)). Over \partial D the segment lies in the wall, where w = 0. Over a point inside D, the points with u_1 < z < u_2 are interior to E (u_1, u_2 continuous), so w_z = 0 there, and w is continuous on the closed segment; by the mean value theorem (The mean value theorem) w(x, y, z) = w(x, y, u_1(x, y)) = 0.
- (a) \operatorname{curl} \mathbf{G} is C^1 with divergence 0 (Result 5 of Divergence and curl), so Result 1 gives \iiint_E 0\,dV = 0. This agrees with Result 3(b) of Stokes' theorem. (b) By Example 5 with f = 1, the flux is \iiint_E \nabla^2 u\,dV = 0.
- (a) For a constant vector \mathbf{c}, f\mathbf{c} is C^1 with \operatorname{div}(f\mathbf{c}) = \nabla f \cdot \mathbf{c} (Result 4 of Divergence and curl), so by Result 1 the flux of f\mathbf{c} equals \iiint_E \nabla f \cdot \mathbf{c}\,dV. Taking \mathbf{c} = \mathbf{i}, \mathbf{j}, \mathbf{k} gives the three components. (b) p is a polynomial, C^1 everywhere, with \nabla p = \langle 0, 0, -\delta g \rangle. By (a) the force is -\iiint_E \nabla p\,dV = \delta g\,V(E)\,\mathbf{k}: upward, equal to the weight of the water the body displaces.