30.4. STOKE’S THEOREM AND THE ORIENTATION OF ∂Ω 1073

define an atlas for ∂Ω,{

Vj,S j}

where Vj ≡ ∂Ω∩U j and S j is just the restriction of R j toVj. Then this is an oriented atlas for ∂Ω and∫

∂Ω

ω =∫

where the two integrals are taken with respect to the given oriented atlass.

What if aI is only the restriction to Ω of a function in W 1,p (Rm) , p > 1? Would thesame formula still hold? Let φ ε be a mollifier and let aIε ≡ aI ∗φ ε . Then Stoke’s theoremapplies to the mollified situation and it follows∫

dωε

= ∑I

m

∑k=1

p

∑j=1

∫R j(U j)

(ψ jaIε

)∂xk

(R−1

j (u))

∂(xk,xi1 · · ·xin−1

)∂ (u1, · · · ,un)

du

= ∑I

p

∑j=1

∫S jV j

ψ jaIε ◦S−1j (u2, · · · ,un)

∂(xi1 · · ·xin−1

)∂ (u2, · · · ,un)

(0,u2, · · · ,un)du1

≡∫

∂Ω

ωε

Now if you let ε → 0, it follows from the definition of convolution that

(ψ jaIε

)∂xk

→∂

(ψ jaI

)∂xk

in Lp (Rm)

and so there is a subsequence such that for each k,

(ψ jaIε

)∂xk

(x)→∂

(ψ jaI

)∂xk

(x)

pointwise a.e. Since R−1j ,R j are Lipschitz, they take sets of measure zero to sets of measure

zero. Hence∂

(ψ jaIε

)∂xk

◦R−1j →

(ψ jaI

)∂xk

◦R−1j

pointwise a.e. on R j (U j) . Similar considerations apply to aIε . Using the Vitali conver-gence theorem in

∫Ω

dωε ,∫

Ωωε , it is possible to pass to the limit. This is because the

integrands are bounded in Lp and so they are uniformly integrable. This proves the follow-ing corollary.

Corollary 30.4.2 Let Ω be an oriented Lipschitz manifold and let

ω = ∑I

aI (x)dxi1 ∧·· ·∧dxin−1 .

30.4. STOKE’S THEOREM AND THE ORIENTATION OF 0Q 1073define an atlas for 0Q, {V;,S;} where Vj = QU; and §; is just the restriction of Rj toV;. Then this is an oriented atlas for OQ. and[ o= | doaQ Qwhere the two integrals are taken with respect to the given oriented atlass.What if a; is only the restriction to Q of a function in W!? (IR), p > 1? Would thesame formula still hold? Let @, be a mollifier and let aye = ay * @,. Then Stoke’s theoremapplies to the mollified situation and it followsdQ@¢Qm Pp 9 (ware) ; O (Xt Xi)— i in— drd haw OX ( J (u)) 0 (u1,-** Un)P O (Xi, +++ Xi,_;)= YL hy, VW jae 0S; (ur, Un) Dun, “tad (0,u2,-°- ,Un) duy= WedQNow if you let € — 0, it follows from the definition of convolution that° (vier) > ° (vi) in L? (R")OX, Xkand so there is a subsequence such that for each k,0 (vate) tz) 0 (var)x) > —.—— (xOX, OX, ( )pointwise a.e. Since Rj! ,R; are Lipschitz, they take sets of measure zero to sets of measurezero. Hence0 (vate) 0 (var)-1 -1—-_—— oR, > ——— oR,OX, J OX, Jpointwise a.e. on Rj (Uj). Similar considerations apply to aye. Using the Vitali conver-gence theorem in fo d@e¢, Jo We, it is possible to pass to the limit. This is because theintegrands are bounded in L? and so they are uniformly integrable. This proves the follow-ing corollary.Corollary 30.4.2 Let Q be an oriented Lipschitz manifold and letoO= Ya (x) dx;, Aves Adxj,_, .I