Parametric surfaces — tangent planes, surface area: Visual walkthrough

Step 1 — Two knobs paint a surface

WHAT. A flat sheet of "control paper" has two dials on it. We call the horizontal dial $u$ and the vertical dial $v$ . Turning them picks a point $(u, v)$ inside a flat region we name $D$ (just "the set of dial settings we allow"). A machine called $r$ reads $(u, v)$ and spits out a point in 3D space: $r (u, v) = ⟨ x (u, v), y (u, v), z (u, v) ⟩$

Term by term: $x, y, z$ are three ordinary height/width/depth numbers, and each is its own recipe that depends on both dials. The angle brackets $⟨ ⟩$ just bundle the three numbers into one arrow from the origin — its tip is the surface point.

WHY two dials and not one? A curve is a wire — you only ever go forward or back, so one dial ( $t$ ) is enough. A surface is a bedsheet — you can slide two independent ways. Two freedoms → two dials. This is the whole reason we later differentiate twice, once per dial.

PICTURE. Left: flat control paper $D$ with a grid of dial settings. Right: the machine $r$ warps that grid onto a curvy sheet in space.

Figure — Parametric surfaces — tangent planes, surface area

Step 2 — Freeze one dial, get a curve you can ride

WHAT. Hold $v$ still at some value $v_{0}$ and turn only the $u$ dial. Now only one thing changes, so the tip of $r$ draws a wire on the surface. We call it the $u$ -curve. Symmetrically, freezing $u = u_{0}$ and turning $v$ gives the $v$ -curve.

WHY do this? Area is a 2D thing, but we only know how to measure motion along one direction at a time (a velocity). So we break the 2D freedom into two 1D rides. Each ride is a curve, and a curve has a velocity vector — the same object from Tangent Lines and Velocity Vectors.

The velocity of the $u$ -ride is written $r_{u}$ and means "how fast and which way the surface point moves when I nudge only $u$ ": $r_{u} = \frac{\partial r}{\partial u} = ⟨ \frac{\partial x}{\partial u}, \frac{\partial y}{\partial u}, \frac{\partial z}{\partial u} ⟩$

The curly $\partial$ (say "partial") is an ordinary derivative that treats $v$ as a frozen constant. Each slot answers: "as $u$ changes by a hair, how much does this coordinate change?" Likewise $r_{v} = \partial r / \partial v$ freezes $u$ .

PICTURE. The two grid curves through one point, with their tangent arrows $r_{u}$ (yellow) and $r_{v}$ (blue) glued to the surface.

Step 3 — A tiny dial-rectangle becomes a curved tile

WHAT. On the flat control paper, take a tiny rectangle: width $Δ u$ (a small step of the $u$ dial) and height $Δ v$ (a small step of the $v$ dial). Here $Δ$ (say "delta") just means "a little bit of." That rectangle has flat area $Δ u Δ v$ . Feed its four corners through $r$ and you get a small curved tile on the surface.

WHY start this small? A whole curved surface has no simple area formula — but a tiny enough piece is nearly flat, and flat pieces we can measure. This is the master trick of all of Double Integrals: chop, measure the crumb, add up.

PICTURE. One flat $Δ u \times Δ v$ rectangle on $D$ mapping to one warped tile on the surface.

Step 4 — Straighten the tile into a parallelogram

WHAT. Two edges of the curved tile start at the corner point $P_{0}$ . One edge runs along the $u$ -curve for a step $Δ u$ ; the other runs along the $v$ -curve for a step $Δ v$ . Because the steps are tiny, "moving along the curve" is almost the same as "moving straight along the tangent velocity." So the edges are approximately the two straight arrows $edge 1 r_{u} Δ u and edge 2 r_{v} Δ v$

Reading each: $r_{u}$ is the direction & speed of the $u$ -ride, and multiplying by the small number $Δ u$ turns speed into an actual small displacement. Same for $r_{v} Δ v$ .

WHY is this legal? This is the linear (first-order) approximation: near a point, a smooth curve is indistinguishable from its tangent line — exactly the idea behind Tangent Lines and Velocity Vectors and behind Arc Length's little straight segments. The two straight edges span a parallelogram that hugs the curved tile.

PICTURE. The curved tile with the two tangent edges drawn straight, closing up into a flat parallelogram (chalk-pink).

Step 5 — The area of that parallelogram is a cross product

WHAT. We need the area of the parallelogram built from two arrows $a = r_{u} Δ u$ and $b = r_{v} Δ v$ . The rule (from Cross Product) is $area = a \times b = ∣ a ∣ ∣ b ∣ sin θ$ where $θ$ is the angle between the two edges, and $∣ \cdot ∣$ means "length of the arrow."

Why does $sin θ$ appear and not $cos θ$ ? Slide the tail of $b$ onto $a$ . The height of the parallelogram is the part of $b$ that sticks out sideways from $a$ — that is $∣ b ∣ sin θ$ . Area = base $\times$ height = $∣ a ∣ \cdot ∣ b ∣ sin θ$ . The sideways part is a $sin$ , not the along-part $cos$ .

WHY cross and not dot? A dot product uses $cos θ$ : it is biggest when the arrows line up and zero when they are perpendicular. Area is the opposite mood — perpendicular edges fence off the most area, parallel edges fence off none. Only $sin θ$ (the cross product) behaves that way. This is the "dot vs cross" mistake the parent note warns about, seen geometrically.

Pull the small numbers out. Since $Δ u, Δ v$ are just positive numbers: $r_{u} Δ u \times r_{v} Δ v = r_{u} \times r_{v} Δ u Δ v$ So the tiny tile's area is $Δ S \approx stretch factor r_{u} \times r_{v} flat crumb area Δ u Δ v$

Read it: the flat control-paper crumb had area $Δ u Δ v$ ; the machine $r$ stretched it by the number $∣ r_{u} \times r_{v} ∣$ into the real tile on the surface.

PICTURE. The base–height decomposition of the parallelogram, showing $∣ b ∣ sin θ$ as the true height.

Step 6 — Degenerate cases: when the tile collapses

WHAT. Watch the stretch factor $∣ r_{u} ∣ ∣ r_{v} ∣ sin θ$ in the broken situations. There are exactly three ways for it to vanish, and each is a real geometric event:

What goes to zero	Meaning	Picture of the tile
$sin θ = 0$ ( $θ = 0$ or $18 0^{\circ}$ )	the two edges are parallel	tile squashes to a line — no area
$∣ r_{u} ∣ = 0$	the $u$ -ride stalls (surface pinched)	tile has no width
$∣ r_{v} ∣ = 0$	the $v$ -ride stalls	tile has no height

WHY it matters. At any point where $r_{u} \times r_{v} = 0$ the surface is not smooth — there is no honest tangent plane and the parallelogram picture breaks. A famous example: the north pole of the sphere, where all the $θ$ -circles shrink to a single point ( $sin ϕ = 0$ ), so $∣ r_{ϕ} \times r_{θ} ∣ = R^{2} sin ϕ = 0$ . The integral still works because that stretch factor is exactly zero there — a single pinched line contributes no area — so nothing goes wrong in the sum. Degenerate points are allowed to sit inside the integral as long as they form a thin set.

PICTURE. Three collapsing tiles side by side: parallel edges, stalled $u$ , stalled $v$ .

Step 7 — Add up every tile: the sum becomes an integral

WHAT. Cover all of $D$ with these tiny rectangles, one stretched tile each. Total surface area is the sum of all the crumb-areas: $A (S) \approx \sum_{tiles} r_{u} \times r_{v} Δ u Δ v$ Now shrink every rectangle toward zero size. Two things happen at once: the parallelogram approximation becomes exact, and the sum becomes a double integral: $A (S) = \iint_{D} r_{u} \times r_{v} d A$

Term by term one last time:

$\iint_{D}$ — "sweep over every dial setting in the region $D$ ."
$∣ r_{u} \times r_{v} ∣$ — the stretch number for the tile at that setting.
$d A = d u d v$ — the flat crumb of control-paper area being stretched.

WHY the stretch factor is not optional. If you wrote $\iint_{D} d A$ you would only get the flat area of the dial paper, not the surface. The factor $∣ r_{u} \times r_{v} ∣$ is precisely the Jacobian of the map $r$ — it converts flat-paper area into real curved-surface area. Drop it and you are measuring the wrong sheet.

PICTURE. Many small tiles tiling the surface, each shaded by how much it stretched, with the sum-arrow becoming the integral sign.

Reality check — the sphere in one breath

The one-picture summary

One flat crumb $d u d v$ on the dial paper → machine $r$ → a tilted parallelogram spanned by $r_{u} d u$ and $r_{v} d v$ → its area is $∣ r_{u} \times r_{v} ∣ d u d v$ → add them all with $\iint_{D}$ .

Recall Feynman retelling — the whole walk in plain words

Picture a flat sheet of graph paper with two dials, $u$ and $v$ . A machine reads each dot on the paper and glues it somewhere in the air, so the flat sheet becomes a bumpy bedsheet. To find how much bedsheet there is, cut the graph paper into tiny squares. Each square lands on the bedsheet as a slightly tilted little diamond. To get one diamond's area, draw the two arrows along its sides — one points the way you go when you turn the $u$ dial ( $r_{u}$ ), one for the $v$ dial ( $r_{v}$ ) — and ask how much space they fence off. That "fenced area" is the cross product's length, $∣ r_{u} \times r_{v} ∣$ , times the little step sizes. It is big when the arrows stand apart and shrinks to nothing when they line up (or when one arrow dies, like at the poles of a globe). Add up every diamond and, letting the squares get infinitely small, the pile of sums turns into the double integral $\iint_{D} ∣ r_{u} \times r_{v} ∣ d A$ . That's the whole story: chop, measure one tilted crumb with a cross product, add them all.

Recall

Why do we differentiate

r

twice (once in

u

, once in

v

A surface has two independent directions to move; each dial gives one velocity,

r_{u}

and

r_{v}

Why does the tiny tile's area use

sin θ

(cross) not

cos θ

(dot)?

Area is the sideways spread of the edges — height

= ∣ r_{v} ∣ sin θ

; it is maximal when edges are perpendicular and zero when parallel.

What does

∣ r_{u} \times r_{v} ∣

physically represent in the integral?

The local area-stretch (Jacobian) that converts flat crumb area

d u d v

into real surface area.

At a sphere's pole the stretch factor is

R^{2} sin ϕ = 0

— is that a problem?

No; the tile genuinely collapses to a point there, contributing no area, so the integral is unaffected.