Before we start, one reminder of the two objects everything is built from — the two tangent arrows on the surface's grid.
The red arrow ru is the velocity if you walk in the u-direction; the violet arrow rv if you walk in the v-direction (this is exactly the idea of a velocity vector). Their cross productru×rv points straight out of the surface (the normal) and its length is the area of the little parallelogram they fence off.
For that same surface, is it smooth? (Recall: smooth means ru×rv=0.)
Recall Solution
Compute the cross product using the component rule
a×b=⟨a2b3−a3b2,a3b1−a1b3,a1b2−a2b1⟩.
With ru=⟨1,0,3⟩, rv=⟨0,1,−2⟩:
ru×rv=⟨(0)(−2)−(3)(1),(3)(0)−(1)(−2),(1)(1)−(0)(0)⟩=⟨−3,2,1⟩.
This is not the zero vector, so yes, the surface is smooth everywhere. (It is in fact a flat plane, which is as smooth as it gets.)
Compute ∣ru×rv∣ for the coner(u,v)=⟨ucosv,usinv,u⟩ (with u>0).
Recall Solution
ru=⟨cosv,sinv,1⟩,rv=⟨−usinv,ucosv,0⟩.
Cross product:
ru×rv=⟨(sinv)(0)−(1)(ucosv),(1)(−usinv)−(cosv)(0),(cosv)(ucosv)−(sinv)(−usinv)⟩=⟨−ucosv,−usinv,ucos2v+usin2v⟩=⟨−ucosv,−usinv,u⟩.
Length:
∣ru×rv∣=u2cos2v+u2sin2v+u2=u2+u2=2u.
(We used cos2v+sin2v=1 twice — the workhorse identity for every trig-parametrized surface.)
Consider r(u,v)=⟨u2,uv,v2⟩. Find every point where the surface fails to be smooth, and explain geometrically what happens there.
Recall Solution
ru=⟨2u,v,0⟩,rv=⟨0,u,2v⟩.ru×rv=⟨(v)(2v)−(0)(u),(0)(0)−(2u)(2v),(2u)(u)−(v)(0)⟩=⟨2v2,−4uv,2u2⟩.
This is 0 exactly when all three components vanish:
2v2=0,−4uv=0,2u2=0⟹u=0andv=0.Only the single parameter point (u,v)=(0,0) fails, mapping to the surface point (0,0,0).
Geometry: at (0,0) both tangent arrows collapse to the zero vector (ru=rv=0). There is no parallelogram to speak of, so no plane is spanned — the parametrization pinches to a point (a degenerate patch). The surface may still look fine there; it's the map that is singular.
Two students parametrize the same flat unit square in the xy-plane:
Student A uses r(u,v)=⟨u,v,0⟩, (u,v)∈[0,1]2.
Student B uses s(u,v)=⟨u+v,u−v,0⟩, (u,v)∈DB.
Both should get area... but B's stretch factor is not 1. Compute B's ∣su×sv∣ and explain how the area still comes out right.
Recall Solution
su=⟨1,1,0⟩,sv=⟨1,−1,0⟩.su×sv=⟨(1)(0)−(0)(−1),(0)(1)−(1)(0),(1)(−1)−(1)(1)⟩=⟨0,0,−2⟩,
so ∣su×sv∣=2.
Why area still matches: this factor of 2 is the local area-stretch of B's map — it is precisely the Jacobian of the linear substitution (x,y)=(u+v,u−v). B's parameter region DB is smaller by a factor of 2 than A's unit square (because B's map squashes/rotates the (u,v) grid onto the square). Integrating the stretch factor 2 over that half-sized region gives 2×21=1: the same physical area A gets from ∬[0,1]21dA=1. The stretch factor and the region size always compensate.
Find the surface area of the part of the paraboloidz=x2+y2 lying below z=4 (i.e. over the disk x2+y2≤4).
Recall Solution
Use the graph parametrization r(x,y)=⟨x,y,x2+y2⟩. With f=x2+y2:
fx=2x,fy=2y,∣rx×ry∣=fx2+fy2+1=4x2+4y2+1.
So
A=∬x2+y2≤44x2+4y2+1dA.
The integrand depends only on r2=x2+y2, so switch to polar coordinates (x=rcosθ,y=rsinθ, area element dA=rdrdθ — this factor r is itself a Jacobian, see Double Integrals):
A=∫02π∫024r2+1rdrdθ.
Inner integral: let w=4r2+1, so dw=8rdr⇒rdr=8dw. As r:0→2, w:1→17.
∫024r2+1rdr=81∫117wdw=81⋅32w3/2117=121(173/2−1).
Multiply by ∫02πdθ=2π:
A=122π(173/2−1)=6π(1717−1)≈36.18
The helicoidr(u,v)=⟨ucosv,usinv,v⟩ over 0≤u≤1, 0≤v≤π. Find its area.
Recall Solution
ru=⟨cosv,sinv,0⟩,rv=⟨−usinv,ucosv,1⟩.ru×rv=⟨(sinv)(1)−(0)(ucosv),(0)(−usinv)−(cosv)(1),ucos2v+usin2v⟩=⟨sinv,−cosv,u⟩.∣ru×rv∣=sin2v+cos2v+u2=1+u2.A=∫0π∫011+u2dudv.
The standard antiderivative ∫1+u2du=21(u1+u2+arcsinhu) (this is the Arc Length integrand's cousin). Evaluate 0→1:
∫011+u2du=21(2+ln(1+2)).
Then multiply by ∫0πdv=π:
A=2π(2+ln(1+2))≈3.606
Prove the graph formula from scratch: for r(x,y)=⟨x,y,f(x,y)⟩, show ∣rx×ry∣=fx2+fy2+1, and use it to argue that a non-flat graph over a region D always has more area than D itself.
Recall Solution
Compute the cross product.rx=⟨1,0,fx⟩,ry=⟨0,1,fy⟩.rx×ry=⟨(0)(fy)−(fx)(1),(fx)(0)−(1)(fy),(1)(1)−(0)(0)⟩=⟨−fx,−fy,1⟩.Length.∣rx×ry∣=(−fx)2+(−fy)2+12=fx2+fy2+1.The inequality. Since fx2+fy2≥0 always, the integrand satisfies
fx2+fy2+1≥0+0+1=1,
with equality only where fx=fy=0 (a locally flat, horizontal spot). Integrating over D:
A(S)=∬Dfx2+fy2+1dA≥∬D1dA=area(D).
If f is non-flat, it has a slope on a region of positive area, so the inequality is strict there and A(S)>area(D): tilting a surface can only add area — the flat shadow is always the smallest. (The z-component of rx×ry being 1 is exactly why: it says the normal always "leans back" toward vertical, and length ≥ any single component.)
Mastery / connection. For a general surface, the scalar surface area element is dS=∣ru×rv∣dA, while the vector element used for flux is dS=(ru×rv)dA. Show these agree in magnitude and explain, in one sentence each, why area uses one and flux uses the other.
Recall Solution
Magnitudes agree because ∣dS∣=∣(ru×rv)dA∣=∣ru×rv∣dA=dS (the scalar dA>0 pulls straight out of the magnitude).
Area measures how much surface there is — a size, no direction — so it uses the length∣ru×rv∣, the size of each little parallelogram.
Flux measures how much of a vector field passes through the surface, which needs a direction to say "through which way." So it keeps the full vector ru×rv, whose direction is the normal and whose length is still the patch area. One quantity, two uses: length for how-much, full vector for through-which-way.
The normal and the area factor come from the sameru×rv: the vector is the normal (for planes/flux), its length is the area stretch (for surface area).
When does ∣ru×rv∣=∣ru∣∣rv∣ hold?
Only when the grid curves cross at 90∘ (so sinθ=1); otherwise the cross product is strictly smaller.
Where can a parametrization fail to be smooth?
Wherever ru×rv=0 — the tangent arrows are parallel or collapse, spanning no plane (a pinch or fold in the map).