Calculus › Section P · Vector analysis › Topic P.5

Parametric surfaces

One parameter swept a curve; two sweep a surface. The cross product of the two sweep directions then hands you the surface's normal and its area element — the tools the final theorems need.

Curves have a big sibling

A parametrised curve \(\mathbf{r}(t)\) swept a line's worth of points with one parameter; a parametrised surface sweeps a whole patch with two:

\mathbf{r}(u, v) = \langle x(u, v), y(u, v), z(u, v)\rangle

for \((u, v)\) in a flat domain \(D\). Fix \(v\) and let \(u\) run: a grid curve. Fix \(u\): another. Together they draw the wire mesh familiar from every 3D globe. The sphere of radius \(a\) is \(\mathbf{r}(\phi, \theta) = \langle a\sin\phi\cos\theta, a\sin\phi\sin\theta, a\cos\phi\rangle\) — the spherical coordinates of the triple-integrals topic wearing parametrisation clothes. A cylinder is \(\langle a\cos u, a\sin u, v\rangle\); a graph \(z = g(x, y)\) needs no more than \(\mathbf{r} = \langle x, y, g\rangle\). The rest of the chapter runs on parametrisations like these.

Tangent vectors and normals

Hold \(v\) fixed and the \(u\)-curve \(\mathbf{r}(u, v_0)\) moves with velocity \(\mathbf{r}_u = \langle x_u, y_u, z_u\rangle\); the \(v\)-curve likewise moves with velocity \(\mathbf{r}_v\). Both vectors lie in the tangent plane at the point, and when they are not parallel they span it. Their cross product is therefore a normal vector:

\mathbf{r}_u \times \mathbf{r}_v

the surface's own normal, computed straight from the parametrisation. A point where \(\mathbf{r}_u \times \mathbf{r}_v \ne \mathbf{0}\) is smooth. Where the cross product vanishes — a cone's tip, or the sphere's poles under the \((\phi, \theta)\) mesh — the parametrisation degenerates even though the surface itself may be perfectly fine there; the sphere's poles are smooth points where the latitude-longitude grid simply pinches shut.

The area element

The image of a tiny \(du\)-by-\(dv\) patch of the domain is a tiny near-parallelogram on the surface with sides \(\mathbf{r}_u\,du\) and \(\mathbf{r}_v\,dv\) — the same geometry the Jacobian used, promoted to three dimensions. Its area is the cross product's magnitude:

dS = |\mathbf{r}_u \times \mathbf{r}_v|\,du\,dv

Summing over the domain gives the surface area formula \(\iint_D |\mathbf{r}_u \times \mathbf{r}_v|\,du\,dv\). The factor \(|\mathbf{r}_u \times \mathbf{r}_v|\) is the surface's own Jacobian: for the sphere it is \(a^2\sin\phi\), shrinking the mesh near the poles exactly as the coordinate slices shrink there. Area on a curved surface is measured in these tilted parallelograms, never in raw \(du\,dv\).

Worked example: the sphere's area

Compute the surface area of the sphere of radius \(a\). The partials of \(\mathbf{r}(\phi, \theta) = \langle a\sin\phi\cos\theta, a\sin\phi\sin\theta, a\cos\phi\rangle\) are \(\mathbf{r}_\phi = \langle a\cos\phi\cos\theta, a\cos\phi\sin\theta, -a\sin\phi\rangle\) and \(\mathbf{r}_\theta = \langle -a\sin\phi\sin\theta, a\sin\phi\cos\theta, 0\rangle\), and the cross product works out to \(a^2\sin\phi\,\langle\sin\phi\cos\theta, \sin\phi\sin\theta, \cos\phi\rangle\) — that is \(a^2\sin\phi\) times the outward unit radial vector, so its magnitude is \(a^2\sin\phi\).

\text{area} = \int_0^{2\pi}\int_0^{\pi} a^2\sin\phi\,d\phi\,d\theta = a^2(2)(2\pi) = 4\pi a^2

the famous formula, sibling to the volume \(\tfrac{4}{3}\pi a^3\). The normal direction came free: it points straight outward, as a sphere's should.

Worked example: a cone

Parametrise the cone \(z = \sqrt{x^2 + y^2}\) as \(\mathbf{r}(u, v) = \langle v\cos u, v\sin u, v\rangle\) — \(v\) is the distance from the axis, \(u\) the angle, and the height equals the radius. The partials are \(\mathbf{r}_u = \langle -v\sin u, v\cos u, 0\rangle\) and \(\mathbf{r}_v = \langle\cos u, \sin u, 1\rangle\), and the cross product is \(\langle v\cos u, v\sin u, -v\rangle\) with magnitude \(v\sqrt{2}\). The lateral surface from the tip up to height \(a\):

\int_0^{2\pi}\int_0^a v\sqrt{2}\,dv\,du = 2\pi\sqrt{2}\,\frac{a^2}{2} = \sqrt{2}\pi a^2

Second route, geometry: a cone's lateral area is \(\pi\) times base radius times slant height; here radius \(a\) meets slant \(a\sqrt{2}\), giving \(\pi a \cdot a\sqrt{2} = \sqrt{2}\pi a^2\). Both routes agree.

The graph case, recovered

For a graph \(z = g(x, y)\), parametrised \(\mathbf{r} = \langle x, y, g\rangle\), the cross product is \(\mathbf{r}_x \times \mathbf{r}_y = \langle -g_x, -g_y, 1\rangle\), so

|\mathbf{r}_x \times \mathbf{r}_y| = \sqrt{1 + g_x^2 + g_y^2}

— the surface-area formula from the surface-area topic, here unmasked as a Jacobian like every other stretching factor in the book. And the normal \(\langle -g_x, -g_y, 1\rangle\) is an old acquaintance: it lies along the same line as the gradient-based normal of the tangent-planes topic (the gradient of \(g(x, y) - z\) is \(\langle g_x, g_y, -1\rangle\), the same direction scaled by \(-1\)). One parametrisation, both formulas.

Choosing an orientation

A parametrised surface comes with a built-in side choice: \(\mathbf{r}_u \times \mathbf{r}_v\) picks one of the two normal directions, and swapping the order of \(u\) and \(v\) flips it. Choosing matters for the flux integrals ahead — 'outward' for closed surfaces like spheres and cylinders, 'upward' for graphs, where \(\langle -g_x, -g_y, 1\rangle\) is the friendly choice. Not every surface has a consistent side: a Möbius strip, given a half-twist before its ends join, has no way to paint an outward face continuously, so it cannot be oriented. The theorems of this chapter politely insist on surfaces that can.

Common mistakes to avoid

First, order confusion: \(|\mathbf{r}_u \times \mathbf{r}_v| = |\mathbf{r}_v \times \mathbf{r}_u|\) for area, but the normal flips with the order — settle the orientation before quoting a direction. Second, dropping the factor: \(dS\) is \(du\,dv\) only on a flat surface; every curvature costs a cross product. Third, integrating over the wrong region: the area integral runs over the parameter domain \(D\), whose bounds come from the parametrisation, not from the \(xy\)-shadow. Fourth, trusting the mesh at its degenerate points: \(|\mathbf{r}_\phi \times \mathbf{r}_\theta| = a^2\sin\phi\) vanishes at the poles, so the normal needs a re-parametrisation there even though the sphere is smooth. Fifth, half-reading the surface: \(z = \sqrt{x^2 + y^2}\) is only the upper half of the double cone; the mirror \(z = -\sqrt{x^2 + y^2}\) is a second surface.

Key takeaways

A parametrised surface \(\mathbf{r}(u, v) = \langle x(u, v), y(u, v), z(u, v)\rangle\) sweeps a patch from a flat domain \(D\), with grid curves at constant \(u\) or \(v\). The partial velocities \(\mathbf{r}_u\) and \(\mathbf{r}_v\) span the tangent plane where they are independent, and \(\mathbf{r}_u \times \mathbf{r}_v\) is the surface's normal — the choice of order choosing the side. The area element \(dS = |\mathbf{r}_u \times \mathbf{r}_v|\,du\,dv\) turns surface area into a double integral over \(D\): \(4\pi a^2\) for the sphere, \(\sqrt{2}\pi a^2\) for the height-\(a\) cone, and \(\sqrt{1 + g_x^2 + g_y^2}\) for a graph — the surface-area topic's formula revealed as the surface's own Jacobian.

Flashcards

Flashcard

Parametrised surface

r(u, v) = ⟨x(u,v), y(u,v), z(u,v)⟩

Two parameters sweep a patch of surface.

Test your understanding

Question 1 of 5

For the parametrised surface r(u, v) = ⟨u, v, u² + v²⟩, what is r_u × r_v?

Question 2 of 5

For the sphere r(φ,θ) = ⟨a sinφ cosθ, a sinφ sinθ, a cosφ⟩, what is |r_φ × r_θ|?

Question 3 of 5

By integrating |r_φ × r_θ| over the full parameter domain, what is the surface area of a sphere of radius 3?

Question 4 of 5

For the graph surface r = ⟨x, y, g(x,y)⟩, what is |r_x × r_y|?

Question 5 of 5

At a smooth point of a parametrised surface, the tangent plane is spanned by which pair of vectors?

Standards: Beyond AP

← Previous topicNext topic →