JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

Surface Integrals & Flux

A line integral adds up a field along a curve; a surface integral adds it up over a sheet — and when the field is a flow, the right sum measures flux, the net amount streaming through. The whole trick is choosing a side of the surface and never letting go of it.

From a curve to a sheet

The previous guide added a vector field up along a one-dimensional path: a line integral, the total work done as you drag something along a curve. Now we climb one dimension. Instead of a wire we have a sheet — a piece of a sphere, a square of plane, the curved wall of a cylinder — and we want to add something up over that two-dimensional surface. This is a surface integral, and the move from curve to sheet is exactly the move from the ordinary definite integral of Volume I to the double integral: one integration sign becomes two, because a sheet has two directions to sweep across.

There are really two flavors, and it pays to keep them apart from the start. The first is the surface integral of a scalar — integrate a scalar field f over the sheet, written as a double integral of f times dS. Picture spreading paint of varying thickness over a curved roof and asking for the total mass of paint: you do not care which way the roof faces, you only weigh what sits on each little patch. The second flavor is the flux integral of a vector field, and there the facing of the surface is everything. This guide builds the scalar kind first because it is gentler, then uses it to assemble flux.

Parametrize the sheet, then weigh each patch

To integrate over a curved surface you first describe it. The standard device is a parametrization: a map r(u, v) that takes a flat rectangle of (u, v) values and bends it into the surface sitting in space, r(u, v) = (x(u, v), y(u, v), z(u, v)). The unit sphere, for instance, is swept out by two angles; a graph z = g(x, y) is the easy case where x and y are themselves the parameters. The parameters u and v are like latitude and longitude lines printed on the sheet: every point of the surface gets an address (u, v), and a tiny rectangle du by dv in parameter space maps to a tiny curved patch on the surface.

How big is that patch? Hold v fixed and nudge u: the point slides along the surface with velocity r_u = partial r / partial u, the partial derivative of the position with respect to u. Nudge v instead and you get r_v = partial r / partial v. These two tangent vectors span the little patch, which is very nearly a parallelogram with sides r_u du and r_v dv. The area of a parallelogram built on two vectors is the length of their cross product, so the patch has area |r_u cross r_v| du dv. That magnitude is the local stretch factor — how much the flat parameter rectangle gets enlarged when it is bent onto the surface — and it is the surface analogue of the Jacobian determinant that rescales area under any change of variables.

Now the scalar surface integral assembles itself. The element of surface area is dS = |r_u cross r_v| du dv, and the surface integral of f over the sheet S is the double integral over the parameter rectangle of f(r(u, v)) times that dS. Reading it aloud: at each patch, take the value of the field there, multiply by the true (stretched) area of the patch, and add up over the whole rectangle. If f is constant 1 you simply recover the surface area itself — a reassuring check that the machinery is measuring what it should.

Orientation: choosing a side and sticking to it

For flux we need more than area — we need to know which way the sheet faces. At every point a surface has two unit normal vectors, pointing out of its two sides; the cross product gives a natural one, n = (r_u cross r_v) / |r_u cross r_v|, a unit-length arrow perpendicular to both tangent directions. To orient the surface is to make a global, continuous choice between the two: outward versus inward for a sphere, upward versus downward for a graph. Once chosen, that choice of n must vary smoothly as you slide across the sheet — no sudden flips.

Reversing the orientation — choosing the other normal, -n — flips the sign of every flux you compute, exactly as reversing the direction you traverse a curve flips the sign of a line integral. So an oriented surface integral is only defined once you have committed to a side, and you must report which. There is also a boundary convention that will matter enormously next guide: if the surface has an edge, the orientation of the sheet picks a direction to walk its boundary curve, fixed by the right-hand rule — point your right thumb along n and your fingers curl in the positive direction around the rim. Hold that picture; it is the hinge on which Stokes' theorem turns.

Flux: how much of the field pours through

Now picture the vector field F as a flow — the velocity of a fluid, say, an arrow at every point telling you which way and how fast the water moves. The flux of F through an oriented surface is the net volume of fluid crossing it per unit time, counted positive in the direction of n and negative against it. The key insight is that only the part of the flow that pierces the sheet contributes; flow that slides along the surface, parallel to it, carries nothing through. The instrument that extracts the piercing part is the dot product: F . n is the component of the flow along the normal, the speed at which fluid actually crosses.

So the flux integral is the surface integral of that normal component: the double integral over S of F . n dS. Patch by patch, you take how fast fluid crosses there (F . n), multiply by the patch's area (dS), and add. There is a beautifully compact shortcut once you recall n dS = (r_u cross r_v / |r_u cross r_v|) times |r_u cross r_v| du dv — the magnitudes cancel and you are left with the vector surface element dS = (r_u cross r_v) du dv, an arrow whose direction is the normal and whose length is the patch area. Then flux is just the double integral of F . (r_u cross r_v) du dv, with no square roots to compute. The orientation lives entirely in the order r_u cross r_v versus r_v cross r_u.

A flux, computed end to end

Let us measure the flux of the radial field F = (x, y, z) outward through the unit sphere — the cleanest example there is, and a number we can later check against the divergence theorem. Parametrize by the two angles: r(theta, phi) = (sin phi cos theta, sin phi sin theta, cos phi), where phi runs from 0 (north pole) to pi (south pole) and theta from 0 to 2 pi around. The steps below carry it through; the only craft is computing one cross product and one dot product, then a routine double integral.

  1. Tangent vectors: r_theta = (-sin phi sin theta, sin phi cos theta, 0) and r_phi = (cos phi cos theta, cos phi sin theta, -sin phi).
  2. Vector surface element: r_phi cross r_theta works out to sin phi times (sin phi cos theta, sin phi sin theta, cos phi) = sin phi . r. It points radially outward (the order phi-then-theta was chosen so it does) — the sanity check passes.
  3. Dot with the field: on the sphere F = r, so F . (r_phi cross r_theta) = sin phi (r . r) = sin phi, because r . r = 1 on the unit sphere.
  4. Integrate: flux = integral over theta from 0 to 2 pi, integral over phi from 0 to pi, of sin phi dphi dtheta = 2 pi times (integral of sin phi from 0 to pi) = 2 pi times 2 = 4 pi.

The answer, 4 pi, is exactly the surface area of the unit sphere — sensible, since the radial field has unit length pointing straight out everywhere on the sphere, so F . n = 1 at every patch and the flux is just the total area. The result is also worth filing away: it is the same 4 pi that appears in Gauss's law for a point charge, and in the next-but-one guide the divergence theorem will reproduce it instantly by turning this surface integral into a volume integral of the divergence of F, which for F = (x, y, z) is the constant 3 — and 3 times the volume of the unit ball, 3 . (4/3) pi, is again 4 pi. Two roads, one number; that agreement is the whole point of the theorems ahead.

Why the bookkeeping matters

It is tempting to treat orientation as a fussy detail, but it is the load-bearing idea of everything that follows. The great integral theorems all have the same architecture: an integral over a region equals an integral over its boundary — provided the boundary is oriented compatibly with the region. Stokes' theorem equates the circulation of F around a closed curve to the flux of its curl through any surface the curve bounds; the right-hand rule we met above is precisely what makes the two sides agree in sign rather than cancel. The divergence theorem equates the outward flux through a closed surface to the integral of the divergence inside; here outward is the non-negotiable orientation. Get the side wrong and a true theorem becomes a sign error.

There is a smoothness caveat too, in the same honest spirit as the orientation one. The surface element dS came from a cross product of tangent vectors, which assumes the surface has well-defined tangents — it must be smooth, or at worst made of finitely many smooth pieces. At a sharp cone tip or the seam of a cube the normal is undefined, and you handle such surfaces by splitting them into smooth faces, orienting each consistently, and summing. The theorems tolerate piecewise-smooth surfaces precisely because the flux through a corner edge, a set of zero area, contributes nothing. Smoothness and orientability are the two standing hypotheses; name them, and the grammar of Green, Stokes, and Gauss is ready to be spoken.