17.3. THE AREA FORMULA 459

In case G is not bounded, apply the above to Gn ≡ G∩B(0,n) for n ∈ N. Then pass to alimit.

Now consider h is only Lipschitz, maybe not one to one. Adding over k in 17.11,∫h(B)

∑k

Xh(Ek∩B) (y)dH n (y) =∫h(B)

#(y)dH n (y) =∫

B|det(U (x))|dmn (x)

This is because if x ∈ B and h(x) = y then x ∈ B∩Ek for some values of k but h is oneto one on B∩Ek and so this happens for at most one x ∈ Ek ∩B.

Now suppose F is a Borel set in h(G) so h−1 (F) is a Borel set in Rn. In the above letB = h−1 (F)∩A+. Then∫

h(h−1(F)∩A+)#(y)dH n (y) =

∫h−1(F)∩A+

|det(U (x))|dmn (x)

Then this is∫h(A)

XF (y)#(y)dH n (y) =∫h(A+)

XF (y)#(y)dH n (y)

=∫

A+XF (h(x)) |det(U (x))|dmn (x)

=∫

AXF (h(x)) |det(U (x))|dmn (x)

because h(A\A+) has H n measure zero and on A\A+, |det(U (x))|= 0. Since h is Lip-schitz, Rademacher’s theorem implies that G\A has mn measure zero and so also h(G\A)has H n measure zero. Thus for F a Borel set,∫

h(G)XF (y)#(y)dH n (y) =

∫G

XF (h(x)) |det(U (x))|dmn (x)

For {Ek} disjoint bounded Borel sets whose union is A+ which are described above,consider λ (E) ≡H n (E ∩h(Ek)) for E an H n measurable set. This makes sense and isa measure because h is one to one on Ek, h(Ek) is H n measurable because h is Lipschitzon Ek and Ek is a Borel set, hence by Lemma 17.1.4 h(Ek) is H n measurable. λ is a finitemeasure because these Ek are all bounded and from what was shown above,

λ (Rm) = H n (h(Ek)) =∫

Ek

det(U (x))dx < ∞

By Lemma 9.8.4, λ is regular on Borel sets. However, by Theorem 16.6.1, whenever Eis a H n measurable set, there exists a Borel set F such that λ (E) = λ (F) and F ⊇ E.Therefore, by Lemma 9.8.4, λ is a regular measure and if E is H n measurable, there existF,H with these being Borel sets and such that F ⊆ E ⊆ H and λ (H \F) = 0. Therefore,

XF (h(x))det(U (x))≤XE (h(x))det(U (x))≤XH (h(x))det(U (x))

and ∫Ek

(XH (h(x))det(U (x))−XF (h(x))det(U (x)))dx = 0

17.3. THE AREA FORMULA 459In case G is not bounded, apply the above to G, = GNB(0,n) for n € N. Then pass to alimit.Now consider h is only Lipschitz, maybe not one to one. Adding over k in 17.11,J Piero) (y)a7e"(y) = | lyase" (y) = f |aet(U (@))| army (x)A(B)mB)This is because if « € B and h(a) = y then a € BOE, for some values of k but h is oneto one on BM E, and so this happens for at most one xz € Ex, NB.Now suppose F is a Borel set in h(G) so h7! (F) is a Borel set in R”. In the above letB=h'!(F)NA*. Thenw(y)ae(y)= fo (det(U (a))| dim (2hl (F)nAthee (FynA*)Then this isXr (yt (y) dO" (y)h(At)= I. De (h(w)) \det (U (w)) | dimy (x)[2 2X qe (h(a) |det (U (w))| dy (x)in Be (y)#(y) dO" (y)because h (A \At) has #” measure zero and on A \ A*, |det (U (a))| = 0. Since h is Lip-schitz, Rademacher’s theorem implies that G\\A has m, measure zero and so also h(G\A)has .#”" measure zero. Thus for F a Borel set,hig Xe (y)# (yd (y = Xq (h(a) |det (U (w))| dm (x)For {E,} disjoint bounded Borel sets whose union is At which are described above,consider A (FE) = #" (EN h(E,)) for E an #” measurable set. This makes sense and isa measure because h is one to one on Ex, h(E) is 4" measurable because h is Lipschitzon E; and E; is a Borel set, hence by Lemma 17.1.4 h(E,) is #” measurable. A is a finitemeasure because these E; are all bounded and from what was shown above,A(R”) = 92" (h(Ex)) = | det(U (x)) dx <0ExBy Lemma 9.8.4, A is regular on Borel sets. However, by Theorem 16.6.1, whenever Eis a .#" measurable set, there exists a Borel set F such that A(E) =A(F) and F DE.Therefore, by Lemma 9.8.4, A is a regular measure and if E is #” measurable, there existF,H with these being Borel sets and such that F C E CH and A (H \ F) = 0. Therefore,Xp (h(ax)) det (U (a)) < Ze (h(a)) det (U (a)) < 2x (h(ax)) det (U (x))and[ (%y (h(a) det (U (w)) — Zp (h(a) det (U (a))) dx =0