A M
Posted on Tuesday, 18 February, 2003 - 07:03 pm:

is the nabla(grad, or upside down delta of a function) vector orthogonal to the suface of the plane? If so, how come? Is there a proof?
Dan Goodman
Posted on Tuesday, 18 February, 2003 - 07:28 pm:

Yes it is, assuming you have a function f: R3 R with the surface defined to be the set of points with f(x)=0. The proof is not completely easy and uses the higher dimensional chain rule for differentiation, but here goes.

Suppose you have a vector v in the tangent plane at a point x on the surface. Draw a curve on the surface φ(t) going through x at t=0 with tangent vector v at x, i.e. φ<quote/>(0)=v. Now, because φ(t) is a curve in the surface, you have that f(φ(t))=0. So if we let g(t)=f(φ(t)) then g<quote/>(t)=0. Using the chain rule we have that 0=g<quote/>(0)=(f)(x).φ<quote/>(0)=(f)(x).v, which is what we wanted.

So, I've used the higher dimensional chain rule which says that if f: R3 R and g:R R3 and h:RR defined by h(t)=f(g(t)) then h<quote/>(t)=f.g<quote/>(t).

David Loeffler
Posted on Tuesday, 18 February, 2003 - 07:36 pm:

Yes. When you have a 2-D surface in 3-D space defined by f(x,y,z)=0 where f is some (sufficiently smooth) function, then the vector f(x,y,z) is orthogonal to the surface at the point (x,y,z).

There are various ways of looking at this. One is to use Taylor's theorem. Let ( x0 , y0 , z0 ) be a point on the surface, so f( x0 , y0 , z0 )=0.

Then the higher-dimensional form of Taylor's theorem states that


f( x0 +r, y0 +s, z0 +t)=f( x0 , y0 , z0 )+r f x ( x0 , y0 , z0 )+s f y ( x0 , y0 , z0 )+t f z ( x0 , y0 , z0 )

plus some terms of the order of r, s and t squared.

You can see that if ( x0 , y0 , z0 ) is on the surface f(x,y,z)=0, then the first term is zero. The second term you can recognise as


(r,s,t).f( x0 , y0 , z0 )

What does this mean? It means that locally the surface 'looks like' the surface (x,y,z).f= constant. This is just a plane with normal f. So the normal at that one point ( x0 , y0 , z0 ) is parallel to f at that point.

(This stuff is fairly hard, so if there's anything you don't understand, please let me know and I'll explain it more carefully.)

David

David Loeffler
Posted on Tuesday, 18 February, 2003 - 07:39 pm:

Oops. Take your pick of the available proofs

David
A M
Posted on Wednesday, 26 February, 2003 - 08:13 pm:

That took some time to grasp, and Thanks!