18.8. EXAMPLES OF STOKE’S THEOREM 491
As before, this is a line integral of the form∫
C Pdx+Qdy+Rdz. where C is an orientedcurve bounding the surface S. This orientation will determine the direction of the vector pabove.
18.8.5 The Divergence TheoremIn this case, we have a parameter domain [a,b]⊆ Rp and the differential form is
ω =p
∑r=1
αr (x)(−1)r−1 dx1∧·· ·∧dx̂r ∧·· ·∧dxp
where dx̂r with the hat means that dxr is omitted. The reason for the (−1)r−1 is to makeminus signs disappear in dω . This led to the divergence theorem
±∫r([a,b])
p
∑j=1
∂α j
∂x j(x)dH k =
∫r(∂ [a,b])
p
∑j=1
α j (x)N j (x)dH k−1 (x)
where the + sign holds if and only if∂(x1,··· ,xp)∂(u1,··· ,up)
≥ 0. We needed to have r is one to one off
a set Ŝ where r(Ŝ)
has H p measure zero and rbl ,ral are also one to one or at least one toone off a set S where rbl (S) ,ral (S) have H p−1 measure zero. It remains to consider thevector N which has jth component N j. I want to argue that this vector is a.e. normal tor (∂ [a,b]) and that sometimes it is an outer normal. Recall that for x= rbl (ul) , N j (x) =
(−1) j+l ∂(x1,··· ,x̂ j ,··· ,xp)∂(u1,··· ,ûl ··· ,up)
1J∗(x)
. Thus for a point on the boundary, for i ̸= l,
xui ·N =p
∑j=1
x j,uiNj (x) =
p
∑j=1
x j,ui (−1)l+ j ∂ (x1, · · · , x̂ j, · · · ,xp)
∂ (u1, · · · , ûl · · · ,up)
1J∗ (x)
=1
J∗ (x)det(xu1 · · · xui · · · xui · · · xup
)= 0
since it is a determinant with two equal columns. This involved expanding along the lth
column which was filled by xui . Also, expanding along this column,
xul ·N =1
J∗ (x)det(Dr) = sgn(det(Dr))
It follows that N is perpendicular to r (∂ [a,b]) and that the angle between xul and N is
no more than 90 degrees if∂(x1,··· ,xp)∂(u1,··· ,up)
≥ 0. Now on the face where ul = bl , xul points away
from this face and so N points in roughly the same direction and is an “outer” normal.Similar considerations apply when ul = al but here the −xul points away and we use −Nbecause of the subtraction in the boundary integrals. This yields the following.
Theorem 18.8.1 Let r be Lipschitz and one to one off S where r (S) has H p mea-sure zero and suppose a similar condition holds for ral and rbl . Let α j be C1
ω =p
∑j=1
α j (x)(−1) j−1 dx1∧·· ·∧dx̂ j ∧·· ·∧dxp, α(x) = (α1 (x) , ...,α p (x))