Surface Integrals

From line integrals to surface integrals

A line integral lets you add up the values of a function along a curve — say the work done by a force as a particle traces a path. A surface integral does the same thing one dimension up: you spread the function across a curved sheet and add up its values, weighted by the local area of each patch.

The cleanest way to set this up is with a parametrization. Pick two parameters u and v that label points on the surface, and write the surface as a vector function:

r(u, v) = (x(u,v), y(u,v), z(u,v)), with (u,v) ranging over a flat parameter region D in the plane.

As (u,v) varies through D, r(u,v) traces out the curved sheet S. A small rectangle in D of width du and height dv maps to a small parallelogram on S whose sides are the tangent vectors r_u du and r_v dv. The area of that parallelogram is the magnitude of the cross product:

dS = |r_u × r_v| du dv

This is the surface area element. It tells you how much surface area corresponds to one unit of parameter area. The cross product r_u × r_v also points perpendicular to the surface — that direction is the unit normal n̂ once we divide by its magnitude.

The scalar surface integral

The scalar surface integral of a function f(x,y,z) over a surface S is

∬_S f dS = ∬_D f(r(u,v)) · |r_u × r_v| du dv.

Imagine the surface as a thin metal sheet of variable density σ(x,y,z) kg/m². The scalar integral ∬_S σ dS is the total mass of the sheet. If f ≡ 1, the integral is the surface area itself.

The vector surface integral (flux)

The vector form measures how much of a vector field F(x,y,z) crosses the surface. Think of F as the velocity of a fluid: the flux through S is the rate at which fluid passes through, in m³/s.

∬_S F·dS = ∬_S F·n̂ dS = ∬_D F(r(u,v)) · (r_u × r_v) du dv.

The dot product F·n̂ extracts the component of F perpendicular to the surface — that's the part actually crossing through. Components tangent to the surface contribute nothing; they slide along it.

Notice that the vector form depends on orientation: flipping n̂ to point the other way negates the integral. The scalar form, which uses |r_u × r_v|, doesn't care about orientation.

Scalar vs vector surface integrals

	Scalar surface integral	Vector surface integral (flux)
Notation	∬_S f dS	∬_S F·dS or ∬_S F·n̂ dS
Integrand type	Scalar function f(x,y,z)	Vector field F(x,y,z)
Surface element	dS = |r_u × r_v| du dv	dS = (r_u × r_v) du dv (vector)
Depends on orientation?	No	Yes — sign flips with n̂
Geometric meaning	Mass, charge, total surface temperature	Flux: fluid, electric, magnetic flow rate
Works on Möbius strip?	Yes	No — non-orientable
Connects to	Surface area when f = 1	Divergence theorem, Stokes' theorem

Worked example: surface area of a sphere

Let S be the sphere of radius R centered at the origin. Parametrize with spherical angles φ (polar, 0 to π) and θ (azimuthal, 0 to 2π):

r(φ,θ) = (R sinφ cosθ, R sinφ sinθ, R cosφ).

Compute the partial derivatives:

• r_φ = (R cosφ cosθ, R cosφ sinθ, −R sinφ)
• r_θ = (−R sinφ sinθ, R sinφ cosθ, 0)

The cross product works out to r_φ × r_θ = (R² sin²φ cosθ, R² sin²φ sinθ, R² sinφ cosφ), with magnitude

|r_φ × r_θ| = R² sinφ.

The surface area is then

∬_S 1 dS = ∫₀^{2π} ∫₀^π R² sinφ dφ dθ = 2π · R² · [−cosφ]₀^π = 2π · R² · 2 = 4πR².

That single integral recovers the famous formula. Replace the integrand with a charge density σ(φ,θ) and you immediately get total surface charge — same machinery.

Worked example: flux through a sphere

Let F(x,y,z) = (x, y, z) — the radial position field. The outward unit normal on a sphere of radius R is n̂ = (x,y,z)/R, so on the surface F·n̂ = (x² + y² + z²)/R = R²/R = R, a constant. Then

∯_S F·dS = R · ∬_S 1 dS = R · 4πR² = 4πR³.

You can also verify this with the divergence theorem: ∇·F = 3, and ∭_V 3 dV = 3 · (4/3)πR³ = 4πR³. Same answer, two paths.

Common parametrizations

Surface	Parametrization r(u,v)	|r_u × r_v|
Graph z = g(x,y)	(x, y, g(x,y))	√(1 + g_x² + g_y²)
Sphere (radius R)	(R sinφ cosθ, R sinφ sinθ, R cosφ)	R² sinφ
Cylinder (radius R)	(R cosθ, R sinθ, z)	R
Cone z = √(x²+y²)	(r cosθ, r sinθ, r)	r√2
Plane through point with vectors a, b	p₀ + u a + v b	|a × b|
Torus (R, r)	((R+r cosφ)cosθ, (R+r cosφ)sinθ, r sinφ)	r(R + r cosφ)

Memorising these saves real time. The graph form |√(1 + g_x² + g_y²)| in particular shows up whenever a surface is given as z = g(x,y) — a common case in physics and engineering.

Where surface integrals show up

• Electromagnetism. Gauss's law: ∯_S E·dS = Q_enc/ε₀. The flux of the electric field out of a closed surface equals the enclosed charge over ε₀.
• Fluid dynamics. Mass flux through a control surface: ∯_S (ρv)·dS gives the rate of mass leaving a volume.
• Heat transfer. The total heat radiated by a body of variable temperature is ∬_S σ T⁴ dS (Stefan–Boltzmann, integrated).
• Differential geometry. The Gauss–Bonnet theorem relates ∬_S K dS (Gaussian curvature integrated) to topology — for a sphere, the answer is 4π regardless of how it's deformed.
• Computer graphics. Image-based lighting samples an environment map by integrating radiance over the unit sphere of incoming directions.

Stokes' theorem and the divergence theorem

Two big theorems convert surface integrals to other kinds of integrals.

Stokes' theorem: for an oriented surface S with boundary curve C,

∬_S (∇ × F)·dS = ∮_C F·dr.

The flux of the curl through S equals the circulation of F around the boundary. This is the surface-integral cousin of Green's theorem and the basis of Faraday's law.

Divergence theorem: for a closed surface S enclosing volume V (with outward normal),

∯_S F·dS = ∭_V (∇·F) dV.

Both theorems let you trade a surface integral for something easier — a line integral or a triple integral — depending on which side has more structure.

Common mistakes

• Wrong orientation. A vector surface integral changes sign if you flip the normal. The phrase "outward unit normal" or "upward unit normal" must be honored — check that r_u × r_v points the way the problem requires, and swap u and v if not.
• Missing the cross-product magnitude. dS is not du dv. Always multiply by |r_u × r_v|; forgetting it computes integrals over the parameter domain, not the surface.
• Confusing scalar and vector forms. ∬_S f dS uses |r_u × r_v|; ∬_S F·dS uses (r_u × r_v) without the magnitude. The first is a number; the second is a signed flux.
• Using dx dy on a graph z = g(x,y) without the √(1 + g_x² + g_y²) factor. Tilted patches have larger area than their shadow on the xy-plane.
• Trying to flux-integrate a Möbius strip. Non-orientable surfaces have no consistent n̂; the vector surface integral is simply undefined there.