Carlos C.

Rodríguez

Checking Stokes’ Theorem for a general triangle in 3D

Given the positions of three poins ~p0 ,~p1 ,~p2 where ~p j = x j x̂ + y j ŷ + z j ẑ

for j = 0, 1, 2, they define an oriented plane segment S. This plane segment is the triangle with
the points as its vertices and oriented by the order [~p0 ,~p1 ,~p2 ] which means that we go around the
boundary visiting the vertices in that order. This order corresponds to the choice of normal vector
field ~N = (~p1 −~p0 ) × (~p2 −~p0 )
The Fundamental Theorem of (all) Calculus says that:
Z Z
ω= dω
∂S S
where
ω = f1 (x, y, z)dx + f2 (x, y, z)dy + f3 (x, y, z)dz
is any 1-form with the scalar functions f1 , f2 , f3 smooth functions defined on a region that
contains S and where ∂S is the oriented boundary of S and the differential is simply,

dω = d f1 ∧ dx + d f2 ∧ dy + d f3 ∧ dz

1
A 1-form, like ω is just a linear function of (tangent) vectors or if we define the concatenation of
two curves as their sum we can think of ω as a linear function on curves. Give a one dimensional
oriented smooth thingee (a parameterized curve) to ω and it will give you back a number
independent of the way in which you choose to parameterize the curve. This number is the line
integral of the 1-form along the curve and the maping from curves to numbers is linear. From this
point of view the purely algebraic fact that d 2 ω = 0 (i.e. taking the differential twice) is just the
purely geometrical fact that ∂2 S = φ (i.e. the boundary of a boundary is empty; In math lingo
taking the boundary is the adjoint to taking the differential).
The algebra of differential forms unifies all the scattered around fundamental theory like theorems
including: The standard fundamental theorem of calculus (I), i.e.,
Z
d f (x) = − f (a) + f (b)
[a,b]
The fundamental theorem for line integrals, i.e.,
Z
dV (x, y, z) = −V (~A) +V (~B)
C

where C is any smooth curve from ~A to ~B and V (x, y, z) is a scalar field (the scalar potential).
Green’s theorem on the plane, i.e.

∂Q ∂P
Z Z  
P(x, y)dx + Q(x, y)dy = − dA
∂D D ∂x ∂y
Stokes’ theorem,
Z Z
~F · d~r = (∇ × ~F) · d~S
∂S S
and Gauss’ divergence theorem,
Z Z
~F · d~S = ∇ · ~F dV
∂V V

In this notebook we show that Stokes’ theorem is in fact true for a general triangle in 3D. Now, if
you think about it a little, by noticing that any smooth surface can be well approximated by
splitting it into little triangles (a.k.a. triangulation duh!) we are (allmost) proving that Stokes
Theorem is true for any (at least triangulatable (is that a word?)) surface.

So here is the surface S.

2
We are going to need line and surface integration so we "attach" the functions SurfInt.sage
defined on our web site.
Sage code
attach(DATA+"SurfInt.sage")

Start by defining the points (p0 , p1 , p2 ) by their general coordinates and the boundary segments
C01 ,C02 ,C12 connecting the points (e.g. C12 is the straight line segment from p1 to p2 ).
Sage code
var(’x y z x0 y0 z0 x1 y1 z1 x2 y2 z2 t’)
p0 = vector(x0,y0,z0)
p1 = vector(x1,y1,z1)
p2 = vector(x2,y2,z2)
C01(t) = tuple((1-t)*p0+t*p1)
C02(t) = tuple((1-t)*p0+t*p2)
C12(t) = tuple((1-t)*p1+t*p2)

Since we have already defined all these goodies, let’s not miss the opportunity to check that the
area of the triangle computed with the aid of the geometric definition of the cross product (i.e.,
A0), coincides with the value given by integration over the surface S (what we call "Area").
Sage code
f(x,y,z)=1
S(u,v) = tuple((1-u-v)*p0+u*p1+v*p2)
Area = surf_intds(f,S,(u,0,1-v),(v,0,1))
A0 = (p1-p0).cross_product(p2-p0).norm()/2
(Area-A0).full_simplify()

0
nice!

3
 Sage code
F(x,y,z)=(0,0,2*x+y)
cF(x,y,z)=curl(F)
Flux1 = surf_int(cF,S,(u,0,1-v),(v,0,1))
Flux1

1 1
2 (z0 − z2 )(y0 − y1 ) + (z0 − z2 )(x0 − x1 ) − 2 (z0 − z1 )(y0 − y2 ) − (z0 − z1 )(x0 − x2 )
Sage code
L1 = line_int(F,C01,(t,0,1))
L2 = line_int(F,C12,(t,0,1))
L3 = line_int(F,C02,(t,1,0))
L1;L2;L3

− 12 (z0 − z1 )(2 x0 + 2 x1 + y0 + y1 )
− 12 (z1 − z2 )(2 x1 + 2 x2 + y1 + y2 )
1
2 (z0 − z2 )(2 x0 + 2 x2 + y0 + y2 )
Sage code
(Flux1 - (L1+L2+L3)).full_simplify()

0
In the next cell we try something a bit more general for the field. Notice that every vector field in
a region of R3 can be written as, F = f x̂ + gŷ + hẑ where f , g, h are smooth functions of (x, y, z) in
the region where F is defined. We have that

∇ × F = ∇ × ( f x̂) + ∇ × (gŷ) + ∇ × (hẑ)

Therefore, if we can show that Stokes’ theorem is true for any smooth vector field parallel to
one of the axis then by linearity it will be true for every vector field. Moreover, every smooth
enough scalar function can be well approximated by a (multivariate) power series. With these
considerations, the fact that Stokes’ theorem holds for the vector fields of the form

F = xi y j zk ẑ,

as we check in the cell below, makes it very likely that it will hold for all vector fields.

4
 Sage code
n = floor(6*random()) for i in range(3)
F(x,y,z) = (0,0,x^n0*y^n1*z^n2)
cF(x,y,z) = curl(F)
Flux2 = surf_int(cF,S,(u,0,1-v),(v,0,1),evaluate=True)
L1 = line_int(F,C01,(t,0,1))
L2 = line_int(F,C12,(t,0,1))
L3 = line_int(F,C02,(t,1,0))
(Flux2 - (L1+L2+L3)).full_simplify(), n

(0, [3, 3, 3])

Now Let’s Interact!

Sage interact code
%hide
x,y,z = var(’x,y,z’)
interact
def _(f = input_box(default=y*z), g=input_box(default=-x*y+x^3-z),
h=input_box(default=0),
p0=input_box(default=(-1,-1,-1)),
p1=input_box(default=(0,1,0)),
p2=input_box(default=(0,0,1))):
vector_field =plot_vector_field3d((f,g,h),
(x,-1,1),(y,-1,1),(z,-1,1))
p01 = arrow3d(p0,p1)
p12 = arrow3d(p1,p2)
p20 = arrow3d(p2,p0)
tri = p01+p12+p20
result = vector_field+tri
F(x,y,z)=(f,g,h)
v0 = vector(p0)
v1 = vector(p1)
v2 = vector(p2)
C01(t) = tuple((1-t)*v0+t*v1)
C02(t) = tuple((1-t)*v0+t*v2)
C12(t) = tuple((1-t)*v1+t*v2)
S(u,v) = tuple((1-u-v)*v0+u*v1+v*v2)
cF(x,y,z) = curl(F)

5
 Flux = surf_int(cF,S,(u,0,1-v),(v,0,1),evaluate=True)
L1 = line_int(F,C01,(t,0,1))
L2 = line_int(F,C12,(t,0,1))
L3 = line_int(F,C02,(t,1,0))
LI = (L1+L2+L3)
html(r"Make sure the points are in the $ (-1,1)^{3} $
cube.")
html(r" ") # new line
html(r"$\int_{S}(\nabla\times F)\cdot\,dS = %s $ and
$\int_{\partial S} F\cdot\,
dr = %s + %s + %s = %s"
%(latex(Flux),latex(L1),latex(L2),latex(L3),
latex(LI)))
result.show(xmin=-1,xmax=1,ymin=-1,ymax=1,zmin=-1,
zmax=1)

Carlos Rodriguez <[email protected]>