lebesgue area and hausdorff measure

14
205 LEBESGUE AREA AND HAUSDORFF MEASURE by Earl J. Mickle (Columbus, U.S.A.) 1. - INTRODUCTION. I.I. Let (1) T: x, = x,(u, v) x2 = x2(u, v), x3 = x3(u, v), (u, v)~Q, be a continuous mapping from the unit square Q: 0~u-~l, 0~v~l, in the uv-plane into Euclidean xlx2x3-space $3 and let A(T) denote the Lebesgue area of the surface of the type of the 2-cell represented by T. The reader is referred to Rad6 [14] for background material and definitions (square brackets refer to the bibliography at the end of this paper). One of the basic problems in the study of surface area has been to introduce a suitable multiplicity func- tion and a suitable measure so that A(T) is given by integrating the given multiplicity function with respect to the given measure. The reader is referred to Federer [8] for an exellent discussion of this problem. Let H~(E) designate the Hausdorff 2-dimensional measure of a set EC S~ (see w 2.2 for the precise definition), let x designate a point in $3 and for a set EC Q let N(x, T, E) designate the number (possibly --[- co) of points in E A T -l(x). Then a suitable choice of a multiplicity function and measure might be given by =fN(x, T, Q)dH ~. AH(T) However A(T)~ A~(T) and simple examples show that in general the equality sign does not hold. Special cases where the equality sign holds have been given by Cesari [3], Besicovitch [l], Federer [7], Mickle and Rad6 Ill] and Rad6 and Reichelderfer [16].

Upload: earl-j-mickle

Post on 23-Aug-2016

231 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Lebesgue area and Hausdorff measure

205

LEBESGUE AREA AND HAUSDORFF MEASURE

by Earl J. Mickle (Columbus, U.S.A.)

1. - I N T R O D U C T I O N .

I.I . Let

(1) T: x, = x,(u, v) x2 = x2(u, v), x3 = x3(u, v), (u, v )~Q,

be a continuous mapping from the unit square Q: 0 ~ u - ~ l , 0 ~ v ~ l , in

the uv-plane into Euclidean xlx2x3-space $3 and let A(T) denote the Lebesgue

area of the surface of the type of the 2-cell represented by T. The reader is

referred to Rad6 [14] for background material and definitions (square brackets

refer to the bibliography at the end of this paper). One of the basic problems

in the study of surface area has been to introduce a suitable multiplicity func-

tion and a suitable measure so that A(T) is given by integrating the given

multiplicity function with respect to the given measure. The reader is referred

to Federer [8] for an exellent discussion of this problem. Let H~(E) designate

the Hausdorff 2-dimensional measure of a set E C S~ (see w 2.2 for the precise

definition), let x designate a point in $3 and for a set E C Q let N(x, T, E) designate the number (possibly --[- co) of points in E A T -l(x). Then a suitable

choice of a multiplicity function and measure might be given by

=fN(x, T, Q)dH ~. AH(T)

However A ( T ) ~ A~(T) and simple examples show that in general the equality

sign does not hold. Special cases where the equality sign holds have been

given by Cesari [3], Besicovitch [l], Federer [7], Mickle and Rad6 Ill] and Rad6

and Reichelderfer [16].

Page 2: Lebesgue area and Hausdorff measure

2 0 6 E A R L J. M I C K L E

For xES 3, T - ' ( x ) ~ ~1, a component of T-'(x) is called a maximal model continuum (m.m.c.) for 7". For E C Q let N*(x, T, E) designate the number of

m.m.c. "s under T that intersect E ~ T -1(x). Then a suitable multiplicity func-

tion and a suitable measure might be given by

a~(r) = fN*(x , T, Q)dH ~.

However, as in the case of AH(T), A(T)~< A'~(T) and simple examples show

that in general the equality sign does not hold.

1.2. At a point (u, v) E Q where the partial derivatives of the functions

defining E in (1) that are involved exist we define the Jacobians

J, (u, v) = x~o x ~ - x ~ x ~ , ]~ (u, v) = x~o x , , - x,~ x ~ , . /3(u, v) = x,~ x~, - x ~ x , , ,

I

j (u, v) = [Z + ]~ + J~]~.

For a plane set E let L~(E) designate the Lebesgue exterior planar measure

of E. If B0 is a Borel set in Q such that J(u, v) exists everywhere on B0,

Federer [5] has shown that

(2) /J(u, v)dL, = f N(x, T, Bo)dH'. Bo

If J(u, v) exists everywhere on Q except on a set of L,-measure zero, is sum-

mable on Q and

(3) A(T) = fJ(u, v)dL~, Q

then a Borel set B oC Q can be found so that J(u, v) exists everywhere on Bo,

L 2 ( Q - B o ) = 0 and for which by (2) and (3),

A(T) = fN(x , T, (4 ) Bo) d H ~ .

Cesari [4] has shown that every surface of the type of the 2-cell of finite Le-

besgue area has a representative T for which J(u, v)exists everywhere on Q

except on a set of L2-measure zero, (3) holds and hence for which there is a

Borel set BoC Q such that J(u, v) exists everywhere on Bo, L 2 ( Q - Bo)= 0 and (4) holds.

1.3. Federer [6], Heisel and Rad6 [9], Mickle and Rad6 [10] have shown

that a suitable measure and (by t-deformations of a mapping) a suitable multi-

plicity function can be obtained so that A(T) is given by integrating the given

Page 3: Lebesgue area and Hausdorff measure

LEBESGUE AREA AND HAUSDORFF MEASURE 207

multiplicity function with respect to the given measure. However, in each of

these cases, the measures are not over sets in $3. In this direction Feder [8]

introduces a multiplicity function L~(x, T, Q) by an averaging process and

shows that A(T) is given by integrating L~(x, T, Q) with respect to the

measure H 2. It is the purpose of this paper to contribute to the lines of inve-

stigation given above by establishing the following result. Among the m.m.c.'s

under T certain ones will be defined to be essential maximal model continua

(e.m.m.c.) under T. We show that if K(x, T, Q) designates the number

(possibly ~ ) of e.m.m.c.'s under T in T-~(x) then

A(T) T, Q)aH ~.

2. HAUSDORFF AND SOME ALLIED MEASURES.

2.1. For a set E CS3 we denote by d(E) the diameter of the set E and

for xES3, r > 0 we denote by s(x, r) the open sphere with center at x and

radius r.

2.2. For a set E C Ss. Let ~ be a generic notation for a finite or denume-

table number of sets e~ e2, .. which satisfy the conditions (i) E C 1.3 ej and (if) for

each j, d(ej) < E, ej ~ ~ and let x~ be a generic notation for a finite or denu-

merable number of open spheres sl, s2 . . . . which satisfy the conditions (i)

E C U sj and (if) for each j, d(sj)~ ~. We define

H~(E) = gr.l.b.~7~/4d(ej) 2 for all ~ ,

S~(E):gr.l.b.Z7~/4d(s) 2 for all -c~ j J �9

Since the set function H~(E) increases when e decreases, it has a limit (finite

or infinite) for r ~ 0. The Hausdorff 2-dimensional measure of E is defined to be

H2(E) = lim H~(E). ~.~o

H2(E) is a measure over a completely additive classe of sets I' in $3 which

contains all open sets, all closed sets, all Borel sets and all analytic sets.

2.3. LEMMA. For any set E C $3, ~ > O, S~,(E)~< 4H2(E).

Proof. For E C S 3 , e > 0 , let e > ~ > 0 be given. Let el, e 2 . . . . . be a ~

and let ~ be such that 0 < ~ < 8, (1 + ~ ) d ( e j ) < ~ , j = 1, 2 . . . . . Let x" be

a point in ej and let sj be the open sphere with center at x i and radius

Page 4: Lebesgue area and Hausdorff measure

208 E A R L J. M I C K L E

r j = max [(1 + 5i)d(ej), a/2J+~]. Then E C U e j C j U s j and d(si)<.2s. Thus

(1) S~(E) ~< Ejrc/4d(sj) 2 ~< 4(1 + ~)2~r~/4d(ej) 2 + 82~/4 for all ~,.

The relation (1) implies that S~ (E)~4 (1 + ~)2 H2 (E)_[_~ ~ / 4 ~ 4 (1 + ~)2H2 (E) + ~2 n/4.

Since ~ > 0 is arbitrary, it follows that S]~(E)-~< 4H2(E).

2.4. For a set EEl" and xES3 we define

h (x, E) = lim sup H 2 [E C) s (x, r)]/n P. r - ]~o

2.5. LEMMA. I f E ~ l", Eo = [x ] x E E, h (x, E) = 0], then H 2 (Eo) = 0.

Proof. Case 1. H2(E) < co. For each pair of positive integers n, m we set

E , , , ~ = [ x i x ~ E , H 2 l E n s ( x , r ) ] ~ n d / n for 0 < r < l/m]. ~-

Then Eo = ~ ~A E .... C ~ E3,,,~ and it is sufficient to show that H*(E~2,,~)= 0 n ~ l m = l m = l

for each integer m. For a given m set 2E ----- l/m. Let s;, s~, . . . be open spheres

such that E~,m C U s;, for each j, cl(s~ < 2~ and there is a xJEE3~,,~ A s.;. For

rj = d(s~,

(2) H~(E32,~) ~< EH2[E32,., N s(x j, rj)] ~ 1/32Y, nry = 1/8~(~/4)d(s~ ~. J j 1

From (2) and the lemma in w 2.3, 2 !

(3) H~(E32,~,) ~< 1/8S2,(Es2,m) ~< ~-H"(Es2,,,).

2 Since H (s2,,,) < co, (3) implies that H2(E32,m) = O. Case 2. H~(E)= co. Set

E t = [ x l xEE, H ~[ENs(x , I ) ] = co for r > 0 ] .

Since h(x, E ) = co for x EEt, E o C E - E 1. Since E - E~ is the union of a

denumerable number of sets of finite HLmeasure , the fact that H*(Eo)= 0 follows from Case 1.

2.6. LEMMA. I f E ~ l", H ~ (E) < ~ , Eo = [x [ x ~ E, h (x, E) > 0], then H ~ (Eo) ----- 0.

Proof. For each positive integer n let

(4) E, = (x ix ~ E, h (x, E) > 1/n].

Then E o = ~ E, and hence it is sufficient to show that H ~(EJ = 0 for each n.

Let ~ > 0 be given. Then E~l", H~(E) < co implies that there is a closed set

F C E such that H~ F) < ~. Then for the sets E, defined in (4)

E, = [x lx$E, h(x, E - - F) > l/n].

Page 5: Lebesgue area and Hausdorff measure

L E B E S C . U E A R E A A N D H A U S D O R F F M E A S U R E 209

For s > 0 given and x E E , there is a sequence of open spheres s(x, r'~) such

that r~, < , / 1 0 , r ~ . ~ 0 for m . ~ o o and H ~ [ ( E - - F ) O s ( x , r'.,)]>~r~/n. These

open spheres cover E. in the sense of Vitali and hence there is a disjoint se-

quence s(x i, q), s(x "z, r ~ ) . . , of them such that E . C U s(x i, 5 0 (see Morse J

[12], w 3.10). Thus

= 25X, r ..< 25nXH [(E - - n s(x' 0 ] (5 ) J i '

~< 25nH2(E - - F) < 25n~

(5) implies that H2(s ~< 25n~ and since ~ > 0 is arbitrary, H2(E,) = O.

2.7. Let �9 be a measure over the classe of H2-measurable sets ~ such that

for EEF there are sets E~, E~EF satisfying the following conditions (i) E ~ E ~ U E ~

E~ f~ E~ = ~. (ii) �9 (E) = @ (E~) = H ~ (E~). (iii) �9 (E~) = 0. Set

~o(x, E) : lira sup O[E N s(x, r)]/~f z. r - .~ - o

LEMMA. For EEF, E o : [x lxEE, ~r E ) > 0] we have that (a) EoEI',

(b) �9 (E) = �9 (Eo) = H e (Eo) and (c) �9 (E - - Eo) = 0.

Proof. It is easily verified that for each pair of positive integers n, m the set

E,,m = [x I x E E, �9 [E N s (x, r)] >1- 7c r2/n for 0 < r < 1/m] is closed with respect

to E and hence E,,,~ E F. Since E - Eo : ~ ~ E,,,~ it follows that ( E - Eo)6I' n ~ l m = l

and hence EoEF. Thus (a) holds. To prove (b) and (c)we shall have to consider

several cases.

Case 1. O(E) < co. It follows easily that for the sets E~, E~ given above

that H ~ (E') = �9 (E') for E ' C E~, E' E 11. Hence ~ (x, E) ~ ~o (x, E~) = h (x, E~).

By the lemma in w 2.5, H ~ ( E ~ - - E o ) : O and by lemma in w 2.6, H~(Eo--E~)=O.

Thus

(6) H2(Eo) = H2(Eo-- E~) + H2(Eo n E~) = H~(eo N E~) + H~(E~-- Co) = H~(E,) = O(E),

(7) O(E - - Eo) = O[(E - - Eo) A E,] : O(E~ - - Eo) --< H2(E~ - - Eo) = O.

From (6) and (7)i t follows that O ( E - E o ) = 0 and * ( E ) = * ( E o ) = H2(Eo).

Case 2. * ( E ) = c ~ , E = ~ E , , E,,EF, * ( E ~ ) < o % n = l , 2 . . . . . By case n = l

1, for the sets Eo : [ x l x E E , , ~(x, E , ) > 0], O ( E , - E o ) ~ 0 and since oo

C E - - E o U(E,--Eo), O(E--Eo):O. Thus (I)(E)=*(Eo) and since He(Eo)>~O(Eo)�88 n = l

H (Eo) = r (Eo). Case 3. O ( E ) = co. Set

E~ = [x lx r [E f-) s (x, r)] = oo for r > 0], E, = E - - E,. 1~ _ R e n d . C i r c . M a t e m . P a l e r m o - - S e r i e 11 - T o m o I V - 1 9 5 5

Page 6: Lebesgue area and Hausdorff measure

2 1 0 ~ a ~ J. M | C K L I g

It is easily verified that Ei is a closed set. Hence, since Et N E~=I~, q~(x, E ) =

q~(x, E=) for x EE2. Since E2 is the union of a finite or denumerable number of

sets in r of finite O-measure, for E ~ = [xlxEE~, ~(x, E~)> 0]-----E2NEo it

follows from case 2 that �9 (E~ - - Eo) = 0. Since ~ (x, E) = co for x E E f3 Et it

follows that E - - E o - - - E 2 - E ~ and O (E - - Eo) ----- 0. By the same reasoning

as used in case 2 �9 (E) = �9 (Eo) = H2(Eo) ~- co.

2.8. Let r o be a subclass of the class of H2-measurable sets P that sa-

tisfies the following conditions: (it E E r , EoEFo, E C E o implies that E E F o.

(if) E E P implies that there is an E o E r o such that Eo C E . (iii) E, EPo, n = 1, 2 . . . .

implies that (U E,)EP. For E E r we set

(8) O(E) ----- g r . l . b .H~(E- - Eo) for sets E o E r o.

LEMMA. The set function �9 (E) defined in (8) is a measure on the class of

sets r and satisfies the following conditions: (at E E r o implies that ~ ( E ) = O.

(b) E E P implies that there is a set E, EP o. E, C E such that for g , ~ E - - E,,

,I, ( ~ = ,I, (Eo) = H ~ (G).

Proof. (if) implies that O(E) is defined for E E P and O(E) is obviously

non-negative. It follows from (8) that O(E) ~ 0 for E E P o. From (iii) it follows

that there is an E , ( r o , E, C E such that for E~----E--E, , t~(E)=O(E~)=H~(E~).

Thus (at and (b) are satisfied.

Let E,, e r, n = l, 2 , . . . , E, n Ej = fJ for i ~ j, E = U E,,. For Eo C E,

Eo E Po, n~ (E - - Eo) = H e (E, - - E, N eo) -]- H 2 (e 2 - - E, N Eo) q-- �9 �9 �9 ~ (I) (E~) -+-

~(E2) q- ' ' ' . Hence

n

For eoC G , G~ro , n = l , 2 , . . . , H~CG--E'o)+~CG--Eo ') + . . . . H~[u(~ _ e o 3 ] = /1

H ~ [(E - - U E~)] ~ (1) (E). Hence II

/I

(9) and (10) imply that O(E) is a measure for sets E E r .

2.9. Let n~, ~ , 7:3 be the orthogonal projections of $3 onto the x2x3-plane ,

x~x~-plane, x~x2-plane respectiveIy. For the class of H~-measurable sets r we set

r, - - [ e l t r e r , z . ,~,(~ - - o], i = ~, z, 3, r , = r , n r , , r~ = r , l n r~,

Page 7: Lebesgue area and Hausdorff measure

LEBESGUE AREA A N D I~kU-qDORI~ MEASURE 211

Each of these classes of sets satisfies the condition of the lemma in w 2.8.

Hence

H l ( E ) = g r . l . b . H Z ( E - - E l ) for E, Er , , i ~ - 1 . . . . , 7,

are measures over the class of sets EEI' . EEP t implies that H i ( E ) = 0, EEP

implies that there an ~ E P , , E~ C E such that for ff~ = E - - E~, Hi(Eta) = HZ(E~,).

We also have the relations that H~(E)>~L,%(E) for i = 1, 4, 5, 7, HI(E)>~L2~(E)

for i ~ 2 , 4 , 6 , 7 a n d HI (E) >t L2 ~3 (E) for i = 3 , 5, 6 ,7 . For EEI', xES 3we set

h,(x, E) = l i m s u p n l [ E n s ( x , r)]/zcr ~, D;(E) = [ x l x E E , hi(x, E) > 0]. r-~O

By the lemma in w 2, 7, EEl ' implies that O;(E)er, H,[D;(E)]=H,(E)=H~[D;(E)] and H~ [E - - D; (E)] = 0.

3. ESSENTIAL MAXIMAL MODEL CONTINUA.

3.1. Let

T: x = xl (u, v), x , = x , (u, v), x , = x,(u, v), (u, v)e Q,

be a continuous mapping from the unit square Q: 0-.~<u-~<l, 0-~<v-~<l into $3.

For the orthogonal pr%z:~ions ~t, ~2, % of $3 onto.lthe x2x3-plane, x3xl-plane,

xtx2-plane respectively we set

TI = = t T: X~= X'(U' V)' T 2 = n , T : x3 : x~(u, v), x,=x~(u, v), x,=x,(u, v),

For i : 1, 2, 3 let E~ be the union of all

T 3 = % T: x, = xt (u, v), x2 = x2 (u, v),

(u, v ) eQ .

the essential maximal continua

(e.m.m.c.) under the plane mapping Tt (see Rado [14], pp. 281-282). We define

following sets.

e~=E~--(E~UEs'), e2=E~--(E;UE~, e~=E~--(E~UE~,

e I = ( E ; A E ~ - E ; , e ~ = ( E ; M E ~ - - E ; , e ; = ( E ; M E ~ - - E ; , e;-----E~AE;ME;.

Each of the sets el, . . . , e~ is the union of m.m.c.'s under T, e~ne~ = 0 for

i # j and e[O . . . Ue;=E~UE~UE~. 3.2. An m.m.c, y under T will be called an essential maximal model con-

tinuum (e.m.m.c.) under T if there exists an i, I -_< i.-.< 7, such that (see w 2.9)

TCe~ and T(T)ED;(T(e~), i.e., if T C e ; and ht[T(T), T(e~] > 0. For E C Q we

set K(x, T, E) to be equal to the number (possibly + oo) of e.m.m.c.'s T under

Tsuch that TAEMT - t ( x ) # r

The definition of an e.m.m.c, is used in two different senses in the pre-

ceding definition and the definition for plane mappings. For T as given in w 3.1

Page 8: Lebesgue area and Hausdorff measure

212 E A R L J . M I C K L E

assume that x 3(u, v ) = 0 . Then E ~ = E~=~I , e ; = l J for i = 2 , . . . . 7 and

e ~ = E~. Then (see Rad6 [14], pp. 292-293) T(e~)= TI(E~)is an open set in

the plane n1($3) and hence D~[T(e~)] (q T(e~)= T(e~)= Tt(E~). Thus the two

definitions are equivalent for plane mappings.

3.3. For T as given in w 3.1, let T----lm, m: Q->-M, 1: M->.S3, be a

monotonelight factorization of T (see Rad6 [14], pg. 53) where m is a continuous

monotone mapping from Q onto a metric space M and l is a continuous light

mapping from M into $3. Since E~, E~, E~ are Borel sets (see Rad6 [14], pg. 296)

and the union of m.m.c.'s under T, e~, . . . , e; are Borel sets in Q and

m(e~),.. . , m(e;) are Borel sets in M. For E C M let N(x, l, E) denote the

number of points in EN l-t(x). By a well known result in measure theory

N*(x, T, e~)= N(x, 1, m(e~)) is an Hrmeasurable and an H2-measurable func-

tion. Since D~(T(eD)EF,

(1) K(x, T, e;)= I N* (x, T, e~) for x E D~ (T (e~) ),

0 for xSD;(T(e~),

i s a non-negative H~-measurable function. Thus

(2) K(x, T, Q ) = K ( x , T, e;) + . . . + K(x, T, e;)

i s a non-negative H~-measurable function.

LEMMA. We have the following relationships

(3) fK(x, T, e;)dH~= fN*(x, T, e~dHl, i = 1 . . . . . 7,

(4) fK(x , T, Q)dH~ fN*(x, T, e:)dH, + . . . + fN*(x, T, e;)dH,.

Proof. All the integrals involved in (3)and (4) exist either finite or infinite.

Since (4) follows from (3) and (2) we need to prove only (3).

Case 1. Hl[O;(T(e;))] =H~[D;(T(e;))] = co. Since K(x, T, e ; )>~l and

N*(x, T, e~)>/ 1 for xED~(T(e~) both of the integrals in (3) for the given i are co.

Case 2. Hi[D;(T(e~) ] = H~[D;(T(e~)] < co. For EED;(T(e~), EEF, we have

that H i ( E ) = H~(E). Thus, since Hi[T(e~--O;(T(e~)] = 0 (see w 2.9), for

D; = D;(T (e;))

(5) f T, e~dHt = f N*(x, T, eDdHi ----- f N*(x, T, e~) d H 2 . D~ D i

(3) thus follows for the given i from (1) and (5).

Page 9: Lebesgue area and Hausdorff measure

LEBE.~GUE AREA A N D HAUSDORFF MEASURE 213

3.4. LEMMA. If T as given in w 3.t is such that A(T) = co then

A(T) =fK(x, T, Q ) d H ~.

Proof. A ( T ) = co implies that

f A r* (x, T,, ~ ) d Lz = co (6)

for i = 1, 2, or 3, say i = 1 (see Cesari [2] and Rad6 [13]). Thus, from (6) for

i = 1, the Lemma in w 3.3 and the remarks in w 2.9 7

fK(x, T, O)aH~= ~fN*(x, T, e;)dH,>~ i=1

>~ ~ fN*(x, T, e)aH, >~ fN*(x, T,, g;)dL~ = co. i=1 , 4, 5, 7

3.5. LEMMA. Let E be a Borel set in E~ U E~ U E; such that Lz T~ (E N E~)= 0

for j = 1, 2, 3. Then T(EGe~)EF and Ht[T(ENe;)] = O for i = 1, . . . , 7.

Proof. Since E is a Borel set, E O e ~ is a Borel set and T ( E A e l ) E I ~,

i = 1 , . . . , 7.

Case 1. i = 1, 2 or 3. Then EAe~CEAE~ and L ~ t T ( E A e ) = L2 Ti(Ene~)~<

LzT , (EN E;) = 0. Thus T ( E A e;)~P, and H , [ T ( E A e)] = 0.

Case 2. i = 4 , 5 or 6. Then there are integer il, /2 such that 1 ~ i 1 < i ~ < 3 and r , - = Fe, n Pc=. Then e~C El, n E~2 and L=% T(ENe~) = L2T~n(EA e) ~< L ~ T , . ( e n ED = o, n = 1, 2. Thus T ( e n 4 ) e r , and Ht[T(EAe;)] = O.

Case 3. i = 7 e; = E~ n E~ n E~, rr = I'~ n I~2 n Pa and L2~ T(E n e;) =

L,T,(EAe;)~<L, T , ( E A g ) = 0 , j = l , 2, 3. Thus T(ENe;)EP, and I-IT[T(EAe;)]=O.

3.6. A continuous mapping T as given in w 3.1 is called eB V (essential

bounded variation) if

fN*(x, Ti, E ) d L ~ < co for j = 1, 2, 3,

and is called eA C (essential absolute continuity) if it is eB V and ifor every

set E of L2-measure zero in Q, L 2 T i ( E N E ) = O , j = 1, 2, 3 (see Rad6 [15]).

Let E * = e ; U . - . U e ; , let E be the union of the m.m.c.'s under T in E*

that reduce to a single point and let E=, E2, Ea be the union of the e.m.m.c.'s

under 7"i, T=, T a respectively that reduce to a single point.

LEMMA. If T is eA C then T[(E* - - E) n 4]Er and H~[T(E* - - E) n e~))] = 0

for i = l, . . . , 7 . Proof. It is easily shown that E*- -E is a Borel set and hence T[(E*--E)Ae)]EP

for i = l , . . . , 7. It follows from Rad6 [15] that L 2 Ti(~--Ej)=O for j = 1 , 2, 3. Since

Page 10: Lebesgue area and Hausdorff measure

2 1 4 , , , t , , , , J', M I C K L E

(E*-- E) O E~j C E)-- Ej, it follows that L, TIt(E* - - E) O E~J--.<L~ Tj(E~-- Ej) = 0 , j = 1, 2, 3. By the lemma in w 3.5, H,[T((E* - - E) O e))] = 0, i = 1, . . . . 7.

3.7. LEMMA. Let T as given in w 3.1 be eA C. Then, for i = l, . . . , 7,

(7) f N*(x, T, eDan, = f N(x, T, e?dH,.

Proof. N*(x, T, e~)-.< N(x, T, e3, for x ES3. Set

E, = [xilN*(x, T, e3 < N(x, T, e3].

Then EIEI' and (7) will follow if H t ( E ) = O. Now xEE~ implies that there is

an m.m.c, under T in e~n T- ' (x) that does not reduce to a point. Hence (see w 3.6) E,C T [ ( E * - E ) n e~] and by the l emma in w 3.6, H , ( E ) = O.

3.8. LEMMA. Let T be a continuous mapping as given in w 3.1 for which the

Jacobian J (u, v) exists everywhere on Q except on a set of L~-measure zero, is summable on Q and

A(r) = fJ(a, v)aL,. q

I/ 13, is a Borel set such that J (u, v) exists everywhere on Bo and L, (Q -- Bo) = O, then 1-~ [T(Bo n e,')] = 1-1, [T(Bo n e;)] < oo, i = 1, . . . . 7.

Proof. We have (see w 1.2)

m[r(Bo)] -.< f N(x, T, Bo)dH ~ = fj(u, v)aL, < co. B o

Let Bt be a Borel set such that B, CT(BoOe~, B, EP,. Then B~----- T-'(B,)A e~OBo is a Borel set and T(B~)= B~. Then

/-/~(B,) -.< f N(x, T, B)dl - f 2 = f j(u, v)aL, B;

(8 ) 3 a

flL(u, v)laL,----~-~ N(x, T,, B;OE',)eL,. ..< n=: B; ~='

Now each of the three integrals at the end of (8) is 0 either because N(x, Tn, B;A~)-- - -0 or because L, T n ( B ) = 0 . T h u s / - / 2 ( B ) = 0 . Since H~[T(BoOe)] < oo it follows that /-F2(E~)= 0 for E~C T(BoOe~), E~EP,. Thus

/-/2[T(B0 N e~)] = H,[T(B o O e,')] < oo for i = 1, . . . , 7.

3.9. THEOREM. Let T be a continuous mapping as given in w 3.1 for which the Jacobian J (u, v) exists everywhere on Q except on a set of Lcmeasure zero, is summable on Q and

A(T) = fJ(u, v)aH,.

Page 11: Lebesgue area and Hausdorff measure

L.EBESGUE AREA AND HAUSDORFF MF_.ASUItE 215

Then

A (7") = f K (x, T, Q) d m. (9)

Proof. There is a Borel set Bo such that J(u, v) exists everywhere on Bo

and L , ( Q - B o ) = 0. Then (see w 1.2)

(10) A(T)= fJ(u, v)dL,---- fN(x , T, Bo)dHE Bo

For B = B o-(e~U . . . LJe;), N(x, T~, B A ~ ) = 0 for n----l, 2, 3 and hence

3

f N(x, r, B)aw = f j(u, v)aL,-.< X f lJ.(., v)ldL, B n = l B

(11) s

~E IN(x, T,, B N E'.)dL, = O. n=l

From (11) it follows that

7

(12) fro(x, r, eo)am= fm(x, r, BoN(e:t.3 ... Ue;))dH'=~fN(x, T, BoNe;)dI-I z.

For E = (Q - - Bo) (q (El LI E~ LJ E;), L, (E) = 0 and since T is e A C (see Rad6

[14], pp. 502-503) L, Tn(ECqE',)=O, n = 1, 2, 3. By the lemma in w 3.5,

H~[T(E(~ e;)] = 0 for i = 1, . . . , 7. This fact and the lemma in w 3.8 gives us

7 7 7

(13) E fN(x, r, e;)dn,= X fN(x, r, Bone;)an,= 2 fN(x, T, Bone,~an2. l ~ l t = 1 l = 1

(9) then follows from (10), (12), (13), the lemma in w 3.7 and the lemma in

w 3.3.

4. FRI~CHET EQUIVALENT MAPPINGS.

4.1. For a metric space M we denote~by d(y,, Y2) the distance between

two point y, and Y2 in M and by d(A, B) the distance between two sets A

and B in 34. For a point y in M and a real number r > 0 we denote by s(y, r)

the set ot points y ' in M for which d(y, y') < r. 4.2. LEMMA. Let T, T r be two Frdchet equivalent continuous plane mappings

(see Rad6 [14], pg. 57) from Q: 0 ~ _ ~ u ~ l , o - ~ v ~ l , 0 : 0 ~ u - ~ 4 , 0 ~ v ~ l

respectively into the xlxz-plane $2 and let

T = l m , m: Q-~M, l: M-~S2; T = l m , m: Q-~M, l: M-~S~,

Page 12: Lebesgue area and Hausdorff measure

216 ~.ARL ~. ~,ICKL~

be simultaneous monotone-light factorizations of them such that m and m are

Fr~chet equivalent (see Youngs [17], pp. 329-330). Then an m.m.c. T under T is

an e.m.m.c, under T i f and only i f m- lm (7) is an e.m.m.c, under T.

Proof. If 7 is an m.m.c, under T then T = m -1re(T) is an m.m.c, under T--.

Hence it is sufficient to show that if T is an e.m.m.c, under T then T is an

e.m.m.c, under ~ Assume 7 is an e.m.m.c, under T. Let (~o be an open set

such that T C O o C Q ~ where 2 ~ is the interior of Q. Let O be the union of

all the m.m.c.'s under T that lie in 60. Then O, re(O) are open sets and

T C O C O o C Q ~ For y = m ( y ) = m ( y ) , set ~ = d [ y , M - - m ( O ) ] > O . Then

0 = m -~[s(y, ~/2)] is an open set such that "( C O C QO. Since T is an e.m.m.c.

under T there exists a Jordan region/? with interior R ~ such that T C R ~

and /? is an indicator region for T(T) under T (see Rid6 [14], pp. 281-282).

Let w, w represent points in Q and Q respectively. Since m and m are Fr6chet

equivalent there exists a sequence of homeomorphisms ~v= h~(w) from Q onto

such that d[m(w), mh~(w)] < 1/n for wEQ, n = 1, 2, . . . . We assert that there is an n o such that for n > n o

a) T C h~ (/?o) C h~ (R) C O,

b) h~(/~) is an indicator region for T(T) under T,

and hence that T is an e.m.m.c, under T. Now there exists an n~ such that for n > n~ (b) holds. Since the union of the m.m.c.'s under T that lie in R ~ is

an open set which contains T and whose image under m is an open set,

d[y, m(Q - - R~ > 0. Hence (b) holds for n > no where

no = n~ + 2/~ + 1/d[y, m (Q - - R~

We first show that h~(R) c O - for n > n o . Now w~R, hn(w)~(Q - - O)

implies that d [y, m (w)] < ~/2, d[y, m hn (w)] > ~ and hence d[m (w), m hn (w)] >~

d[y, mh~(w)] - - d[y, re(w)] > ~ - - ~/2 = ~/2 > 1/n o . This is a contradiction

and hence h~ (R) C ().

Finally we show that T-Ch~(/~~ Now w E Q - R ~ h~(w)ET implies that

d[m(w), mh~(w)]=d[m(w) , y]>~d[m(Q--R~ y] ~ 1~no. This is a contradiction

and hence T-Ch~(R~

4.3. Let T, T be two Fr6chet equivalent continuous mappings from Q:

0 ~ < u ~ l , 0 ~ v ~ < l , Q: 0 - ~ < u ~ l , 0 - ~ < v ~ l respectively into Ss and let

T = l m , m: Q ~ M , I: M-~S3, 7"=Ira , m: Q-~M, l: M~.S3 ,

be simultaneous monotone-light factorizations of them such that m and m are

Page 13: Lebesgue area and Hausdorff measure

L E B E S G U F ~ A R E A A N D H A U S D O R F I ~ M E A S U R E 2 1 7

m m

Fr6chet equivalent. Let e~, . . . , e; and e~, . . . , e; be the sets defined in w 3.! with respect to T and 7 ~ respectively.

LEMMA. Under the above conditions e; ~ m-l m (e;), T(e;) ~ T(e;), N*(x, T , e;) =

N*(x, T, e;) for i = 1, . . . , 7.

Proof. To prove the statements of the lemma it is sufficient to show

that if T is an m.m.c under the T and TCE~ for some j, 1 -~< j~<3 , then

= m-lm(,;) C E~ (since y is an m.m.c, under 7). Let

~ l -~- Itmj, mr: M ~ M t, It: Mj-~ ~j(S3),

be a monotone-light factorization of %.I. Then

Tj ---- 7~tT = l~(mtm), mtm: Q-~Mj , It: Mt->.7~t(S3),

= %.7" ~ lt(mtm), mjm: Q ~ . M t , lj: Mt-~.7~t(S3) ,

are simultaneous monotone-light factorizations of Tj and Tj such that, since m

and m are Fr6chet equivalent, mjm and mjm are Fr6chet equivalent. Hence, if "f

is an m.m.c, under T and T C E~ there is an e.m.m.c T* under T t such that

y C y * C E ~ and hence y m- lm(y) C - - I -1 ~, = m mj mtm(y* ) ~- and by the lemma

in w 4.2, y* is an e.m.m.c, under L - H e n c e y C T-* C E~.

4.4. LEMMA. Let T and 7" be Frdchet equivalent continuous mappings from

Q an 0 respectively into $3. Then

K(x, T, Q ) = K ( x , T, Q) for x E S 3.

Proof. This follows directly from the definition o[ K(x, T, Q), K(x, T, Q)

and the lemma in w 4.3.

4.5. THEOREM. Let T be a continuous mappings as given in w 3.1 Then

(1) A(T) =fK(x, T, Q) d/-/~.

Proof. Case 1. A ( T ) : ~ . (1) follows from the lemma in w 3.4.

Case 2. A ( T ) < c~. Then (see w 1.2) there is a continuous mapping ?"

from the unit square Q: 0 ~ u - ~ < 1, 0~<v-~< 1 into $3 such that 7" and T are

Fr~chet equivalent, .](u, v) (the Jacobian of T--) exists everywhere on Q except on a set of L2-measure zero and

A(T) : f ] ~ , v )dH, . O

By the theorem in w 3.9 and the lemma in w 4.4 we have

A(T)= aCr)=fK(x, T, Q)dm=fK(x, T, Q)dn . Columbus, Mars 1955

Page 14: Lebesgue area and Hausdorff measure

2 1 8 , ~ A a , S. M z C x , ~ S

BIBLIOGRAPHY

I. A. S. Besicovitch, On the definition and value of the area of a surface, Quart. J. Math. Oxford Set. 16, 86-102 (1945).

2. L. Cesari, Caratterizzazione analitica delle superficle continue di area finlta secondo Lebesgue, Annali della R. Scuola Normale Superiore di Pisa, (2), vol. I0 (1941), 253-294 and

vol. II (1942), 1-42.

3. L. Cesari, Sulle trasformazionl continue, Ann. Mat. Pura Appl. (4) 21, 157-188 (1942).

4. L. Cesari, On the representation of surfaces, Amer. J. Math. vol. 72, 335-346 (1950). 5. H. Federer, Surface area II, Trans. Amer. Math. Soc., vol. 55, 44-76 (1944).

6. H. Federer, Coincidence functions and their integrals, Trans. Amer. Math. Soc., vol. 59, 441.-400 (1940).

7. H. Federer, the (~, k) rectifiable subsets of n space, Trans. Amer. Math. Soe., vol. 62 114-192 (194~).

s. H. Federer, Measure und area, Bull. Amer. Math. Soc., vol. 58, 306-378 (1952).

9. R. O. Helsel and T. Rad6, The Cauchy area of a rr~chet surface, Duke Math. j., vol. 15 159-167 (1948).

10. E.J. Mickle and T. Radd, A new geometrical interpretation of the Lebes~,e urea of a surface, Duke Math. J., vol. IS, 109-180 (1948).

11. E.J . Mickle and T. Rad6, On a theorem of Besicovitch in surface area theory, Rivista di Matematica della Universith di Parma 2, 19-45 (1951).

12. A. P. Morse, A theory of convering and differentiation, Trans. Amer. Math. Soe., voi. 55, 20,5-235 (1944).

13. T. Rad6, Two-dimensional concepts of bounded variation and absolute continuity, Duke Math. J., vol. 14, 587-608 (1949).

14. T. Radd, Length and area, Amer. Math. Soc. Colloquium Publications, vol. 30, 1948.

15. T. Rad6, On essentially absolutely continuous plane transformations, Bull. Amer. Soc., vol. 55, 629-632 (1949).

16. T. Radd and P. V. Reichelderfer, On generalized Lipschitzian transformations, Rivista di Matematica della Universit~ di Parma 2, 289-301 (1951).

17. J. W. T. Youngs, A reduction theorem concerning the representation problem for Fr~chet varieties, Proc. Nat. Acad. Sci. U. S. A., vol. 32. 328-330 (1946).