Calculus › Section O · Multiple integrals › Topic O.5
Surface area
How much fabric covers a curved roof? One double integral answers, and its integrand measures how the slopes stretch a flat patch onto the curve.
Measuring a curved skin
Flat regions have area; solids have volume; but a curved surface, like a dome's cladding or a sail catching wind, has a surface area of its own. The shadow of a hemisphere on the floor has area \(\pi\), yet the hemisphere itself covers \(2\pi\): tilting and stretching inflate area above its projection. To compute the true figure for a surface \(z = f(x, y)\), you add up the areas of thousands of tiny patches that follow the surface, and each patch's area is a stretched version of its flat footprint. One factor captures the whole stretching story, and it is built from the two slopes.
Why the patch stretches: a cross product
Slice the surface above a small rectangle of width \(\Delta x\) and depth \(\Delta y\) with the vertical planes through its edges. The patch of surface trapped above is roughly a parallelogram. Walking along its edges: a step of \(\Delta x\) in the x-direction moves along the tangent vector \(\langle 1, 0, f_x \rangle\), and a step of \(\Delta y\) in the y-direction moves along \(\langle 0, 1, f_y \rangle\). The parallelogram they span has area \(\Delta x \Delta y\) times the magnitude of their cross product:
whose length is \(\sqrt{1 + f_x^2 + f_y^2}\). That is the stretching factor: the local slope of the surface in both directions.
The surface area formula
Add the stretched patches over the region \(R\) where the surface stands, and let the patches shrink:
The formula is honest about its limits: it applies to graphs \(z = f(x, y)\) whose partials are continuous on \(R\). Fully general curved surfaces, described parametrically the way a torus or a Möbius band must be, need the parametric machinery of vector analysis, later in the course. Note what does not matter: the height of \(f\). A raised ceiling covers no more area than the original; only the slopes enter.
Worked example: a tilted plane
Find the area of the part of the plane \(z = 2x + y\) above the unit square \(0 \le x \le 1\), \(0 \le y \le 1\). The slopes are \(f_x = 2\) and \(f_y = 1\), so the stretching factor \(\sqrt{1 + 4 + 1} = \sqrt{6}\) is constant over the square. The area is
Check it without any calculus: the patch is a genuine parallelogram whose edges are the tangent vectors \(\langle 1, 0, 2 \rangle\) and \(\langle 0, 1, 1 \rangle\), and their cross product \(\langle -2, -1, 1 \rangle\) has length \(\sqrt{4 + 1 + 1} = \sqrt{6}\). Both methods agree because for a plane the tangent approximation is exact.
Worked example: a paraboloid cap
Find the area of \(z = x^2 + y^2\) above the unit disk. The slopes are \(2x\) and \(2y\), so the factor is \(\sqrt{1 + 4x^2 + 4y^2} = \sqrt{1 + 4r^2}\) in polar form, and the disk begs for polar coordinates anyway:
The substitution \(u = 1 + 4r^2\), \(du = 8r\,dr\), handles the inner integral. Sanity check: the answer exceeds the disk's area \(\pi \approx 3.14\), because the bowl tilts every patch away from flat, and the rim, where the slopes are largest, stretches the most.
Worked example: a hemisphere, checked against history
The formula should reproduce known results, so test it on the upper hemisphere \(z = \sqrt{1 - x^2 - y^2}\) above the unit disk. Write \(z\) for the height itself: \(f_x = -x/z\), \(f_y = -y/z\), and
using the sphere identity \(x^2 + y^2 + z^2 = 1\). So \(S = \iint \frac{1}{z}\,dA\), or in polar coordinates \(\int_0^{2\pi}\int_0^1 \frac{r}{\sqrt{1 - r^2}}\,dr\,d\theta\). The inner integral is \(\big[-\sqrt{1 - r^2}\big]_0^1 = 1\), giving \(S = 2\pi\). That is exactly the classical hemisphere area: a unit sphere covers \(4\pi\), so its top half covers \(2\pi\). When a formula survives contact with a result known since Archimedes, it earns trust.
Strategy for setting up areas
Three habits make these integrals tractable. First, compute \(f_x\) and \(f_y\) before touching the region; the factor under the root often simplifies before any integration happens. Second, match coordinates to symmetry: if \(f_x^2 + f_y^2\) depends on \(x^2 + y^2\), the integrand is a function of \(r\) and polar coordinates are usually the only sane route, as both worked examples showed. Third, remember what the integral needs from you: the region \(R\) is the shadow of the patch you want, so a request for "the part of the surface above this disk" means the limits describe the disk, not the surface.
Common mistakes to avoid
The most damaging error is dropping the 1 under the root: \(\sqrt{f_x^2 + f_y^2}\) is the stretching only for near-vertical walls and badly understates gently sloped surfaces, where the true factor is near 1. Second, do not confuse this with the volume integral \(\iint f\,dA\) or the region's own area \(\iint 1\,dA\); three different questions, three different integrands. Third, if you convert to polar coordinates, the factor \(r\) from \(dA = r\,dr\,d\theta\) joins the square root; leaving it out silently rescales everything. Fourth, \(f\) may be negative: the formula uses slopes only, so no \(f \ge 0\) condition is ever required.
Key takeaways
The area of \(z = f(x, y)\) above a region \(R\) is \(S = \iint_R \sqrt{1 + f_x^2 + f_y^2}\,dA\), where the stretching factor is the magnitude of the normal vector \(\langle -f_x, -f_y, 1 \rangle\) built from the tangent vectors \(\langle 1, 0, f_x \rangle\) and \(\langle 0, 1, f_y \rangle\). For planes the factor is constant; for spheres and paraboloids it simplifies to a function of \(x^2 + y^2\) and polar coordinates finish the job. The hemisphere computation returning \(2\pi\) confirms the formula against classical geometry, and the 1 under the root is what keeps every answer honest.
Flashcards
Flashcard
S = ∬ √(1 + fx² + fy²) dA
Test your understanding
Question 1 of 5
What is the surface area of the graph of z = 3x above the unit square 0 ≤ x ≤ 1, 0 ≤ y ≤ 1?
Question 2 of 5
Which integral computes the area of z = x² + y² above the unit disk?
Question 3 of 5
For the hemisphere z = √(1 - x² - y²), what does the factor √(1 + fx² + fy²) simplify to?
Question 4 of 5
What happens to the surface area integral when f is a constant function over R?
Question 5 of 5
In the derivation, the tangent vectors along the patch edges are ⟨1, 0, fx⟩ and ⟨0, 1, fy⟩. What is the magnitude of their cross product?
Standards: Beyond AP
