ACCESSIBILITY IN EUCLIDEAN N-SPACE
WITH APPLICATION TO DIFFERENTIABILITY THEOREMS
DISSERTATION
Presented in Partial Fulfillment of the Requirements
for the Degree Doctor of Philosophy in the
Graduate School of the Ohio State
University
By
ALBERT GEORGE FADELL, B.A., M.A.
The Ohio State University
19$h
Approved by:
'"’’V a W Y j o u M"
Adviser Acknowledgement
The writer is indebted to Professor T* Rado for his valuable suggestions in the preparation of this dissertation
» i — Table of Contents
Page
I. PRELIMINARIES
1.1, Euclidean n-space ...... • 1
1.2, Lebesgue Measure ...... 5
1.3, Sections ...... 1°
I.iu Lebesgue Density...... 21
1,5* Real-valued Functions ..••«••..••• 36
II, THE SECTIONAL DENSITY THEOREM...... kZ
in. THE ACCESSIBLE S E T ...... 52
17. THE RADEMACHER-STEPANOFF THEOREM...... 63
V. THE STEPANOFF T H E O R E M ...... 85
- 11 - INTRODUCTION
Frequently, theorems are stated and proved for the case of the
plane, and the validity of their n-dimensional analogs is asserted
without formal proof. Most often the n-dimensional analogs are
indeed valid, as asserted, and moreover admit of a proof differing
from the planar case only in notation. Nevertheless, there arise
planar theorems for which either (a) the n-dimensional analog is
false, or (b) the n-dimensional. analog is in fact true, but requires
a new method of proof. Situation (a) is best exemplified by an
outstanding theorem concerning the existence almost everywhere of
a regular approximate total differential (see Rado [7]: numbers in
[ ] refer to the bibliography at the end) which is true for the
plane but is known to fail for the higher dimensions* On the other
hand, situation (b) is brought out by the Rademacher-Stepanoff
Theorem (see IV.10) concerning the existence almost everywhere of
a total differential. For both Rademacher’s proof [6] of his
original restricted planar form of the theorem and Stepanoff's
proof [9 ] of the present complete planar form utilize crucially
a unique property of the plane not possessed by the higher
dimensions. Yet both Rademacher and Stepanoff, overlooking
this fact, assert that their proofs readily extend to the n- dimensional case. One purpose of this dissertation, therefore,
is to present in full a formal proof of the Rademacher-Stepanoff
Theorem for Euclidean n-space. To this end we have introduced the concept of accessibility (see 111*1 ), the study of which is the
subject of chapter III* The need for making £ -estimates along
lines parallel to the coordinate axes motivates the notion of
accessibility* Our fundamental theorem is the so-called
Accessibility Theorem (see 111*10), the proof of which requires
the self-important Sectional Density Theorem (see 11*18), the
latter theorem being the n-dimensional analog of the Linear Density
Theorem (see Saks [8], Chapter 9, Theorem 11.1)« With the aid of
the Accessibility Theorem as a basic tool, ve are able to prove
in a rather natural way the Rademacher-Stepanoff Theorem for
Euclidean n-space* Having thus proved the Rademacher-Stepanoff
Theorem we apply it together with the Accessibility Theorem to obtain a proof of the Stepanoff Theorem for Euclidean n-space
(see V*l;8 ), which characterizes the existence almost everywhere of an approximate total differential (see V.32)* We should remark that the proof of the planar Stepanoff Theorem (see Stepanoff [9] and Saks [8], chapter VII, Theorem 12.2) can be modified readily by an inductive process to take in the n-dimensional case.
Nevertheless, our approach yields in addition to the Stepanoff
Theorem other pertinent theorems of independent interest (for example, V.10, V.31, and V.39), and utilises in a rather natural way the already known Rademacher-Stepanoff Theorem* I. PRELIMINARIES
1.1 Euclidean Spaces
1.1.1. Notation. Given two arbitrary sets of elements S^ and S^
we write S , C S , , or S„”^ S , when every element of S. is an 1 2 2 1 1 element of S . When S C S and 3 C S , we write S * S • 2 12 2 1 12 Also, x e S means x is an element of the set S. By the empty
set we mean the set without any element, and we denote it by 0 .
We define S^ - to be the set consisting of all elements x
wliich belong to S^ but not to S^. If ^ is any finite or
infinite collection of sets we define the union iJ S to be the
set of all x contained in at least one of the sets S, and the
intersection O S to be the set of all x belonging to all of
the sets S. Given an arbitrary sequence of sets S^, i = 1,2,..., we denote by lim sup the set consisting of all points x which belong to S for infinitely many indices i. If Si<— • S , i. i+1 i - 1,2,,.. , we set li^i j and if ,
i » 1,2,..., we set lim S. » O s . • i x ± x
1.1.2. Continuation. For each positive integer n we denote
Euclidean n-space by Rn , whose elements (points) are ordered 1 n n-tuples (x ,..,x ) of real numbers. Usually we abbreviate by 1 n writing x for (x ,...,x )• A standing notational convention will be the following. If any letter, with or without affixes, is used to denote a point in RH the n numbers defining the point
-1- will be denoted by the same symbol (with the same affixes if any) with superscripts \ n . Thus, if we speak of pQ in Rn, we mean the ordered n-tuple ( p ^ , . , p n) « Moreover, for a given o o positive integer j < n, the symbol p^ denotes the j-ih coordinate C of the point p (identified by the context), o
n 1,1.3* Definition, Given two points p, q e R we define the distance
li P - q U from p to q by the formula
U p - q u - [ (p1 - q1)2 r i«l
1,1, U, Theorem, The distance function satisfies the following four properties:
(i) p - q tt > 0 for all pairs p,q s R*1 . n (ii) Given p,q e R , then p - q 4 « 0 if and only if p ■ q ,
(iii) |\ p - q U - \\ q - p U for all pairs p,q e r” •
(iv) Given p,q,r e Rn » there follows the "triangle inequality"
It p - q u ♦ itq - ru > lip - r n •
Proof: See McShane [§*]» section 2,
1.5. Definition. A point p e Rn is termed the limit of a sequence of points p.^ s Rn if ti P - P ^ U -^ 0, We write p » lim p^ , or p± - - P as i / • •
1 n 1.6. Theorem, A point p ■ (p »,..,P ) is the limit of a sequence of I n j j points p. » (p. »•••»? ) if ^ d only if p — ^ p as i jr « i 1 i for j " l,2,«..,n *
Proof: See Caratheodory [1], llh, Satz 5*
T.1.7* Definition. A point p is called an accumulation point of n a subset S ofF. if p is the limit of a infinite sequence of distinct points p e S • Then closure S is the union of S and the set of accumulation points of S, A set S is closed if S « closure 3; that is, S is closed if every point which is the limit of a sequence of points of S must belong to S. A set S is open if Rn-S is closed*
1*1*9. Theorem* The class of closed subsets of Rn is closed under the operations of finite unions and arbitrary intersections. The class of open subsets of Rn is closed under the operations of finite intersections and arbitrary unions. The closure of a set is a closed set.
Proof: See McShane C*J. section 2*
1.1*9. Definition. Given a finite sequence of real numbers a^, b_^, a^, b^ such thatl^ - a_^ -Ib^ - aji, i, j » l,...,n , the set ^ x \ a^ < < b^, i ■ l,...,n ^ is termed an open oriented n-eube. If all of the signs < are replaced by the sign < , the set is called a closed oriented n-cube* If all, some, or none of the signs < are replaced by < the set is called simply an oriented n-cube* The numberlb^ - a^l is called the side-1ength of the
-3- n-cube.
1*1*10, Theorem. If we define an isolated point of a set S to be n a point in R which is not an accumulation point of S, then the set
of points of S which are also isolated points of S is countable*
Proof: We merely sketch the proof. Note first that the class of
'•rational1' n-cubes Q » < x1 < | , where are required to be rational numbers, is countable and covers Rn. With each
isolated point p e S we associate a rational n-eube Q such that p
is the only point of S contained in Q. Since p is isolated, such
a Q exists. Thus the isolated points of S are in a one-to-one
correspondence with a subset of a countable set and hence themselves
constitute a countable set,
1,1,11, Theorem, Let T denote a one-to-one transformation (function) from an arbitrary set X onto an arbitrary set Y. Then the following formulas hold:
(i) T(CJs) - Vj TS,
(ii) T(f\s) - 'I t s ,
(iii) T ^ - S g ) - TS1 - TS? , where S , S 41 X and the unions and intersections are taken over 1 2 any non-empty class of subsets S of X,
Proof: See Kuratowski (3]> ^ 3, sections I and II,
-U- 1.2. Lebesgue Measure
1*2.1. For the following definitions and theorems of this section
1.2 the reader is referred to Saks [8], Chapter II,£U,|2 7>
Chapter III, JfJ 1 to jj 6, inclusive.
1.2*2, Theorem. If we denote by the n-dimensional Lebesgue outer measure (see Saks, Chapter II,£ it), then L° is defined on the class of all subsets of R11 and has the following properties:
(i) S is a non-negative real number or + ,
(ii) L° S < L° So whenever S..— S , n 1 - n ^ 1 2 (iii) L° ( \JS.) < i L° S. for all sequences S., i ■ 1,2,. n i-1 1 “ i-1 n 1 1
1*2.3. Definition. A subset S of Rn is termed L -measurable if for _u______every subset E of Rn we have the identity
L® (S = I E) ♦ L® I(Rn-S) f\E] - l£E .
T.2.it. Theorem. The class of L -measurable subsets of Rn is closed n under the operations of countable unions, countable intersections, and differences, and contains the empty set and the whole space
(see Saks, Chapter II, Theorem U.5).
1.2*5. Theorem. If we define as usual the class of Borel subsets Q n of R to be the smallest class of subsets of which (a) contains the closed and open subsets of Rn , and (b) is closed under the operations of countable unions, countable intersections, and differences, then the Borel subsets of Rn are Immeasurable (see
Saks, Chapter II, Theorem 7.U)•
1 .2.6. Theorem. Given a sequence of bounded Immeasurable subsets
of Rn , i • 1,2,..., and an arbitrary subset E of R11, we have the relations (see Saks, Chapter II, Theorem U.6).
(i) l£ [ li^i sup ( E H S ^ ) ] > ligi sup 1^ ( E H S^)
(ii) L° [ lim (E O Si)] - li| I°(E 0 S.), if either S . d S ^ , i * or S± i ** 1 >2 j««* #
1 . 2 . 7 . Theorem. Given an arbitrary subset S of Rn there exists a
Borel subset S* of Rn such that (see Saks, Chapter III, Theorem 6.8)
(i) S*-> S,
(ii) L°(S*i I E) • L ° ( S C\ F) for every Immeasurable subset E of
Rn.
The set S* is called a measurable hull for S.
1 . 2.8. Theorem. If ws denote by 1^ the restriction of 1^ to the class of Immeasurable subsets of Rn (that is, I^S is defined only when S is Immeasurable, and takes the value I ^ S ) , then Lr has the additive property (see Saks, Chapter II, Theorem 2**6) c o (i) I^( * 0 Sj ) » i Z H L 3. if SiOiS. - # for i / j, and hence i-1 i-1 11 J the subtractive property
(ii) y s ^ ) - - i*s2 , if s2c l Sl .
-6- 1*2.9* Theorem. Given any L -measurable subset S of Rn, there n exists a sequence of subsets of R , i « 1,2,..., for which the following statements hold (see Saks, Chapter III, Theorem 6«6(iv))
(i) Si is closed, i - 1,2,... ,
(ii) S± C S i+1< ^ S , i-1,2,... ,
(iii) I. (s - G s ) - 0 . i-1 1
1.2.10. Theorem. Given two L -measurable subsets S., of Rn, n 1 2 we have the formula
I (S, H S-) + I. (S ^ S j - L S ♦ L S . n 1 Z n 12 nl n2
Proofr We first express S ^ U as the disjoint union of Im measurable sets as follows:
s1U s 2 - (siO s 2)U[si-(sirvs2)]0[s2-(s1O s 2)) ,
Then by 1.2*8 we obtain
* v s!ns5) * V i - v sin y *
♦ L nS ? . V S 1OS2, , and so by conjoining terms the formula is proved*
1*2*11. Definition. Let S be an arbitrary subset of Rn. Then a property P is said to hold at (or for) almost every point of S if * there exists a subset S of S such that (ii) The property P holds at each point of S*.
1.2.12. Theorem. Let S be a bounded Ln-^neasurable subset of Rn , and let P denote a property defined at each point of S. Suppose that for any assigned € >■ 0 there exists an Immeasurable subset
E = E( 6 ) of S such that (a) Ln(S - E)-<6 , and (b) the proper ty P holds at almost every point of E. Then P holds at almost every point of S.
Proof: In accordance with the hypothesis select for each posi tive integer j a subset Ej of S such that
(1) Ln(S - Ej) *< 1 / 2 j ,
(2) P holds at almost every point of Ej .
Now by (2) (see 1.2.11) there exists for each j a subset E*j of Ej such that
(3) Ln(Ej - E * j ) = 0 ,
(4) P holds at each point of E*j .
Also, for each positive integer k we have the obvious inclu sions By (5), (3)» and (l) it follows that
q O Od o O (6) LAS — O E* ) i 2 ( S -E) -V 2 L ( E -E*) ^ ^ J21 y j*kv y j.-kv j j
-< 1/2 k_1
Noting that (6) holds for each k, we have that
LJ (7) Ln'S - jsl y “*1 j ) = 0 • 00 ll ^ But by (4 ) property P holds at each point of . . E and so J J with (7 ) the theorem follows.
-9- 1*3* Sections
1,3*1* Notation. For a given positive integer n » 2 let w ■
denote a proper subsequence of ^l,...,n^ , and * * i let it » j^,*..denote the complementary subsequence. Note
then that 1 < k < n-1 and once n is assigned the associated positive integer k is also assigned. We shall consider tt and v
X ti 21 as point functions in the following sense: if p ■ (p ,...,p) R , then
tip - (phl,...,phk) , «*p - (ph¥,...,ph*n*k ) •
Thus t t , tt* are functions defined on the space Rn and having their
images in the spaces R^, Rn'*k respectively. We call i*p the «• k ... projection of p into R , and tt p similarly*
n k 1*3*2* Continuation. Given a subset S of R and any point R # -l (see 1.3*3 below) we term the set w (S ^ v”Ax. ) the (w«xy) - section of S. Note that the set s O tr "Sc^. consists of points of S which have hj-th coordinates the same as the i-th coordinates of
Xjj,* Then v*(S f V n"*^rjc) is the aggregate of the ^-projections of these points, and therefore is a subset of R11”^. The reader will find it instructive to apply this notation in the case of the plane
(n - ?) *
1.3*3* Continuation* Given a positive integer k < n we will subscribe the k to a small Latin letter to denote a point of R^, and we will
- 10- superscribe the k to a capital Latin letter to denote subsets of
Rk. For example, ■€: Sk C_ . This notation will occur only
in this part 1.3 and in Chapter II, and is not to be confused with the usual notation for infinite sequences*
I*3«li» Continuation. Define for each positive integer 3, 2 < 3 < n
1 n Thus in accordance with 1.3*3, given p » (p ,...,p") we have
*jP ■ ■ (p1 j • •«jP*^**1) ♦
Also, given any subset S of Rn we will use the notation S(p^,...,pn) as defined by the formula
S(pj,*..,pn) - , if 2 < 1 < n * 3 3 3
Thus S(p^,*..,pn) denotes the set of points (a^,**.,x^) <=. such that
(x ,*..,x , p*^,»«»,p ) S *
Therefore the following formula is easily verified:
( S(pn ) ) (p^,*..,?1**1) - SCp*5,...,?11) if 2 < >J < n-1 .
The above notation will find application in Chapter III. But for
Chapter V, as well as for JJ1$. -, where line sections are needed we define the projections ita , ir* by the formulas where the symbol 3 3-s used to indicate the deletion or the singling out of the index i* Thus
/ I i+1 Ux * j W a p ■ (p »*«*»P » P ) > ®* P P • v 3 3 * The symbol p will always indicate p , that is, the j-th coordinate of p, and similarly for any snail Latin letter having j as a superscript*
1*3*?* Agreement* For the sequel of this part 1*3 let a subsequence i . * n of { l,...,n Y be thought of as fixed, and let s be the complementary* sub a equence.
T1 JC 1*3*6, Theorem* If C Sg <- R , then given any point p^^ R we have the inclusion
w " 2^ ) t ^ n ^ ) *
Proofs This follows at once from 1*3*1*
* v n»k 1*3*7* Theorem* Given i», « as in 1*3*1 let S , S be subsets of R^ and R0"^ respectively* Then
(i) «(«_1Ek 0 ir*"1E“-k ) . sk
(ii) .‘ ( n - y H V V * ) - a“-k .
Prooft We prove (i), and (ii) follows in a similar way* The inclusion (1) «(n"Jsk n » * " 1sn’k) c sk
- I k v is obvious, since nn S ■ S . Now assert that
(2) Sk C- 8 ( 1 1 ^ 0 s * " 1^ ) .
To prove (2) take any point (p*’,.*.,^) (q^,#..,qnF*k)ts s“-k • Then the point r * jr11) defined by the requirements hi i hi i r ■ p , i ■ l,...,k and r - q , i ■ l,...,n-k , 1 k lc & T v obviously has the properties wr«(p,...,p)tzS,«r « (q »»*«,qn ) B**k fc=. S . Thus (2) is true, and with (1) the theorem follows* o n i.3.c* Theorem. Given any sequence p ^ R •then p — p if w m and only if * * wp ■— *> »p , » PM —* * P # m n Proofx This follows at once from 1.1.6 . V 1.3.9. Theorem. Given any point p^ R the projection tr is a -1 k one-to-one mapping from n onto R • k it Proof: Let p^t: R be given. By 1.3.1 the projection * is single valued on Rn and hence on the subset *~^p^ of Rn. To show that ff* is univalent on tr"*^pjc, let p, q be two distinct points such that —1 # p,q€. » pjj , in other words, up * ffq * p^ • Then since » is complementary to w it is clear that it p and w q are distinct, -13- since otherwise the stated condition ffp ■ «q would require that p • q • To see that w is onto, take any point p ^ ^ (z. R * Then the point * p having v -coordinates the sane as the coordinates of p^jj, and u- coordinates the same as the coordinates of p^, surely has the properties wp • p^, s*p • P^jr • Thus, since the statement *p « p^ is equivalent to p t w"*^PkJ ^ e Projection a* is onto, and the theorem is proved# v 1*3*10* Theorem* Given any point p^tc R we have the following formulass (i) «*k u s> r\ - u**(sc\ *-\) , (ii) »*[( ^ S) (\ * " V k J " ^ H*(s ^ » (iii) «*C(sr s2) n *’^ 3 » **(s1 '\ 9m\ k ) - «*(s2r\ , where in (i) ,(ii) the 0 , fl are taken over any class of subsets S in Rn. Proofs Ve first note that by the distributive and associative lavs of sets we have the identities (1) ( Us ) n « U (S (\ w - ^ ) , (2) (fis) n »f\(sfi«*1pk) , o) (^ - s2)Pi e - 1^ - ( \ ^ « ~ \ ) - (s2 n « - \ k) . Now since by 1*3*9 the projection tt is one-to-one on p^, it follows from 1*1*11 that -lk- do »*[u (s n ."Sp j - j «*(sn«mlpk) , (5) **i n (s a - V'CSjO it*1^ ) - «*(S2n "“S^) , Now (1), (1*) Imply (i)j (2), (5) imply '(ii)} and (3),(6) imply (iii). Thus the lemma is proved* I*3*11* Theorem* Let S be a closed subset of Rn, and let there be lc given a sequence p ^ <= R , i ■ 1,2,..., with pj^ — =*■ as i — ^ lim sup » (S Pi w"*pj^ ) •— v (S \ • Proof t Take any point (1) P^jj <= lim sup if*(S 0 • Then (see 1*1*1 ) there is an infinite subsequence ^ of the positive integers i such that <2 > P n - k e ** In view of (2) there exists a sequence of points (3) S such that -15- U ) - p ^ , (*) » Pa-k • * But by hypothesis p j ^ > p^ » and bo Py-^ r * Therefore from (1;) we have (6) nqi^ ^ Pk • Consider now the point r ^ Rn defined uniquely (see 1*3*9 ) by the conditions (7) «*r - , (8) irr - pk * Relations ( 5 ) , (6) imply that (see 1*3.8 ) converges and indeed by (7)* (8) we have (9) r * • But since by hypothesis S is closed it follows from (9), (3) that (10) res* Now relations (7)j(8)»(10) require that (11) p <= A s H v ^ a ) n-k K Thus (1) implies (11), and so the theorem is proved* - 16- I *3• I t* Fubini Theorem. Let S be a bounded L -measurable subset n * 1 11 of R • Then the section « (S C\ i T Is ^-measurable for almost every point t t > the function L° , [«*(S \ ) ] is L « * n-k K n integrable, and si° t«*(s a ] d - i^s Rk Prooft The proof (which here we omit) nay be accomplished by applying the technique used by Saks [ 8 ] ^ 8, pp. 76-62 • 1.3*13. Corollary* Under the hypothesis of 1.3*12, I^S ■ 0 if and only if (S H a \ ) 1 - 0 for almost every point x ^ fe R * Proofr This follows at once from 1.3*12 • I.3*lU* Theorem* Let S be an Immeasurable subset of R®. Then for almost every point p € Rn the section «*(S f\ wT^Wp) is I ^ k ” measurable. Proofr Suppose, contrary to the theorem, that there exists a n subset E of R such that (1) L E > 0 , ' n ^ ]L (2) p t E implies that w (S C\»" wp) is not I^-measurable* Prom (1) and 1*3*13 there follows the existence of a subset of Rk - 1 7 - such that (3) ^ ^ > 0 (h) E^ implies that L^^[«*(E Oiw^p^)] » 0 * Let there now be given a point k (£) P t E * k From (U) we know that » (E \ w p^) f fjf and so there exists a point (6) p fc E such that (7) «p - p • k Now (6) and (2) imply that w*(S n \p) is not I^^-aeasur^le, and ®o hy (7) we have that * **3' (8) «w ( s f U p^) is not L^^-neastcrable • Thus (5) implies (8), which by the Fubini Theorem (see 1*3*12) is impossible in view of (3)* Hence the existence of a set E satisfying (1),(2) is impossible, and so the theorem is proved* 1*3*15* Theorem* Let S be a subset of H11. Then given any point p <=. Rk the section A s ^ w " 3^ ) is a Borel, closed, or open subset of Rn"k according as S is a Borel, closed, or open subset of Rn* - 18- Proof t If S is closed^ the section tr*(sO can be easily shown to be c]dosed using 1.3*8* If S is open, then Rn-S is closed) «nd by 1*3.10 s*[(Rn-S)n * ~ \ k ] * A r 11^ - A s n ) - - R ^ - s*(S \ « - \ ) , * *1 and so n (S 11 t> p ) is open* For the case of S being a Borel set k * n ve consider the class B of those Borel subsets of R having the ^ Ip property that « (S f\ n~ p^) is a Borel subset of Rn” • By 1*3*10 ** and 1*2*6 the class B is closed under countable unions and differences, and (as we have just shown) contains the closed and open sets. Now B being a subclass of the class of Borel subsets n of R , the theorem follows fdtwri 1*2.5 • 1.3*16* Definition. Let Sk , S1** be subsets of Rk, R0"* respectively* We define the Cartesian product Sk X s°“k to be a subset of R11 defined by the fommla Sk X s““k • ^p |p - (Pj^P^fc) » Pk *= Sk , P^j, vs Sn“k i . 1*3*17* Theorem* If S^, are respectively I y measurable> Jf y. If measurable sets then S X S is Lyineasurable and Ln(Sk x S®“k) - (I^Sk) . (ik.kS®’*) . Proofs See MacShane [5s], Theorem 25*7, p* lii5 * -19- k 1.5.18 Theorem. The set S X S is a Borel. closed, or open subset of R11 acoording as Sk , Sn”k are Borel, closed, or open subsets of Rk , R11-^ respectirely. Proof* See McShane [5a], Theorem 25.7 (in the proof), p. 145 for the fact that Sk /c Sn ”k is a d o s e d or open subset of B?1 according k k _n-k as S , S are closed or open subsets of R , K respectirely. k & . Consider now the class B of the Borel subsets of R . Let B denote the subclass of B defined by the property that Sk € B* implies that Sk a R^"-^ is a Borel subset of R11 . Observe that B* is closed under the operations of countable unions, countable intersections, and differences « by reason of 1.2. 4 and the following easily verified formulas: (i) (U sM x h " - ^ U ( S k ^ R a-k ), (ii) ( O s k ) = f\ (sk x ), (iii) (Sk - s £ ) « R n”k = (Sk X &-*) - (s| x ) . Thus, sinoe by our initial statement B* contains the open and closed subsets of R^", we hare that B = B* - that is, SkxRP“k is a Borel subset of R“ if Sk is a Borel subset of Rk. By a similar argument I ^ x S n"k is a Borel subset of R11 if Sn_k is a Borel subset of R?1”^. Accordingly, in view of 1.2.4 and the obvious formula Sk X Sn *k - (sk < R11^ ) H (Rk x sn *k ) we have that SkASn”k is a Borel subset of R11 if S^, Sn “k are Borel subsets of R^, Rn_lc respectively. Thus the theorem it proved. -20- I*U Lebesgue Density n I*l;*l« Definition* Let S be an arbitrary subset of R , and let a point n p t R be given. Then p is termed a point of outer density for S if the following condition is satisfied; for any assigned tr > 0 there exists a £ - S(€ ) > 0 such that for any oriented n-cube Q of side-length s the inequality (sf\ Q) > (1 - u) sn holds if p w Q and a < 5 * , n I,U,2* Notation# We denote by *_XS the set of points of R which are points of outer density for S# n I*lu3* Lebesgue Density Theorem* Given an arbitrary subset S of R , we have the relation L° (S -z_^S) - 0 . n Proof: See Saks [ 8] , Theorem 10.^ page 129 * I» k ,h * Theorem* Given S-C. S <— R11 • we have the inclusion / S s J - / ^ S « 1 2 1 2 Proof: This follows at once from I*lwl, since L° is monotone* n I*U*5* Theorem: Given any L -measurable subsets S.., S of Rn we have n «*• z the formula - 21- Proof: The inclusion (1) sg) c. A s 1 rA A s 2 follows at once since A is monotone (see I*iuU) • Assert now that (2) a S j ' l *->Sg A ( s 1 ^ S? ) * To prove (2) take any point of p t ^ such that (3) p ^ , p c= ^S2 , and assign any <=. > 0 • Then (see I*U*I ) there exist, numbers 5- , ^ such that* for 1 2 any n-cube Q of side-length s, if p Q there follow the inequalities (U) 1 (S-Pl Q) > (L.( t/2) ) sn if s < 5 n 1 -»• (5) Ln (S2 f\ Q) > Define S - *in 5., ) , and let there be assigned an n-cube Q of side length s < c5 • Then Inequalities (10 and (5) hold* But application of 1*2*10 to the pair Q , S j O Q yields Sg 1 q) ♦ y t ^ a q) u (s2a ♦ Ife (S2 ^ Q) , -22- and so Ln(Sin Q) - Q) + ^ Q) * V * Using (li) and (5) we obtain L (S. i S 1 Q) > (1 - ( c/2) ) sc ♦ (1 - U / 2 ) ) s* - en - n -l ■ ( 1 - o)sn • Accordingly, (see I.U.l) since € was arbitrary,we have (6) P 6 ^ ( s 1 n s 2), and so since thus (3) iaplies (6) assertion (2) is true* Now (1) and (2) complete the proof* I«lu6. Lenina. Let Q be any closed oriented n-cube of side-lengtb s. Then there exists a sequence of n-cubes Q^, k * 1,2,..., for which the following statements hold: (i) is open, (i d (in) a - ii= u k - co , (iv) s ■ li» s^ as k ^ . Proof: Let Q - ^ x I ^ < a?" < jj Define for each positive integer k the n-cube \ • N^x I a± - (1A) < x1 < \ + (1A) | -23- Then clearly each is open and (see I.1.1 ) Q * lim ^ , as k -—f * Also since Q has side-length s - bA - a. and Q. has side-1 ength s^ ** b i * *1 + (2/k) it is clear that ■?s as k -? c-xj * Thus the lemma is proved* . n n 1*4*7* Theorem* Let S be any subset of R , and let p R • Then p S (see I*lu2) if and only if for any assigned ^ > G there exists a > » ^ ) > 0 such that given any open oriented n-cube Q of side-length s the statements -5 < and p Q Imply L® (S \ Q) > (1 . fc)sn * Proof: The necessity follows at once from the definition I.lul since the class of open oriented n-cubes is a subclass of the class of all oriented n-cubes. To prove the sufficiency} let c > 0 be assigned. Then by hypothesis there exists a .5” ■ ^"( t ) > 0 such that given an open oriented n-cube Q of side-length s we have (1) ij (S \ Q) > (1* c/2)sn if s < 5 , pc- Q . Now take any oriented n-cube Q of side-length s, and assume that (3) a < S, (J0 pcQ* — 2U— Let Q denote closure Q, whence (see 1.1.8) Q is closed. In accor dance with 1.4.6 let t k=l,2,..., be a sequence of oriented n-cubes of side-length sk having the following properties (5) - (8): (5) Qk is open, (6) 0 Q 13 Q , k 1 * (7) Q — lim Qk as k 0 0 , (8) S =r lim s^ as k o o . By (4 ), (6) we have that p 6. for each k ; and (3), (8) imply that sk -C S for k large enough, say k i kQ . Accordingly, in view of (5) it follows that (9) ^°(S ^ ^k^ ^ (l “ (^/2)) sn for k "Z. kQ . k Now, inasmuch as the sets form a descending sequence of Ln-meas- urable sets with limit T , we may pass to the limit in (9) (in view of 1.2.6) and obtain (see (8)) (10) L°(S O Q) i (1 - (6/2)) sn < (1 - e ) s11. Since the boundaryof n-cubes is easily shown to have ^-measure zero, the relation s HQ - ( s D Q) U £s 0( Q - Q)] yields (11) L°(sOQ) £ L°(S0 Q) + L°[s f|(Q - Q)| = L0(sfU) . n n n »• j n But the inclusion Q d Q requires (12) L°(S f\ Q) ■£= L°(sfI Q) . n n - 25 - Since XH).tlad (12) Imply* L°(S H $) - L°(S \ Q) , (30) yields n a (1 3 ) l (s i i q) > (i - e)sn , n and so since (3), (U) imply (13), application of I*i*,l requires p t A S , Accordingly, the theorem is proved, I,fc,8* Theorem, Given n > 2 let S be a I^j-neasurable subset S"1 i n. n of R and let p * (P'S***? ) <=> Bt such that * 1 n~l *np ** (p *,**»P ) *= S Then (see 1,3,16 ) p v_- (S*1"1 K R1 ) t ^ Tla»X Proof t Let > 0 be assigned. By 1,1;, 7 * cAS implies the existence of a number "S ■ S( o) > 0 such that given any open iwl oriented (n-l)-cube Q of side-length s ve have the inequality (1) 1 . (s”-1 ! cT 1) > (1 - J S - 1 if . < Y , »*p <= Q1^ 1 . n»i n Let (2) Q • ^ x \ a^ < at* < ^ be an open oriented n-cube side-length s , (3) p c Q , a < S , Then (see 1,3*4 for notation) by (2) it is clear (since a^ < pn < bn) that the section -26- (U) QCp 0) is an open oriented (n»l)-cube of side-length s, and by (3) ve note that (5) «Jp <=. fl(p“) , Then, in view of (3)> (h) , (5), it follows from (1) that (6) ^ [ S 1*-1 * QCP11)] > fl - O s * " 1 . But dearly we have the identity (7) [S^fiQCpP) } X I - ( S ^ X R1) ^ Q where I is the open interval of real numbers ^ with an < < bn « n-1 Noting (6), (7) and the hypothesis that S is I^j-measurable, we apply 1*3* 17 and thus obtain (8) Ita[(Sn “1 X R1) '1 Q ] - ys*-1 ] * % I > > (1 - Cr)a0-1. S ■ (1 •<= )sn • Briefly, (8) says (9) y (S®"1* R1) fI Q] > (1- (=)sn * Accordingly, since thus (2), (3) Imply (9) , it follows from I*li*7 that p & A ( s “J,x a1) , and so the theorem is proved* n I.lu9* Definition. Given any point p d H and any real number r > 0 we set K(pjr) ■ ( x i x t Rn , \\ p - x tt < r^. Then K(pjr) is termed the closed spherical neighborhood of p with radius r . I.U.10. Lemma. Every closed spherical neighborhood K(pjr) of a n point p is a closed (hence Immeasurable) subset of R , and I^KCpjr) - c rn , where c is a constant depending only upon n . Prooft See Mayrhofer [4 ], pp. 119-122. I.lull. Theorem. Let S be any subset of Rn, and let p t A S , Then for any assigned > 0 there exists a 5" - £>(<=)> 0 such that given any closed spherical neighborhood K(pjr) of p we have the inequality l£[S r \K(pjr)) > (1- e) LK(p;r) if 0 < r < 5 “ . Proof: Let c be the constant in the formula I^K(p,r) * cr* » and denote by Q(p,s) the closed oriented n-cube with center p and of side-length s . By hypothesis (1) p c a s , -28- Let t > 0 be any assigned nuifcer* Then by (1) and there exists a ■ S*( ) > 0 such that (2) l£ [ S U Q ( p j s ) ] > (1 - c2*,U ) a11 if s < 3 • Define (3) 5 - c* /2 , and take any ntufcer r with (1*) 0 < r < J . Let <3(p,s) be the (unique) closed oriented n-cube defined by s « 2r , and note that (3) ,(10 imply (5) s < S * • From (5), (2) it follows that (6) Lf[snQ(p,s)J > (1 - c2"°^ ) sn . n For brevity let K - K(pjr) , Q * Q(pjs) • Then in view of the obvious- inclusions S P i Q C (s n K) J (Q - K) ,KCQ, and the fact that Q,K are Ln<-seasurable, we have (7) l £(s a q ) < l® (s n k ) ♦ y q . k ) • • l£ (S f\ K) ♦ I^Q - y - -29- - l£ (sf\K) ♦ 2V - c r“ Applying (7) and (6) we obtain (recall a » 2r ) (8) I g ( s a K) > t®(S f\ > (1 - c2”n*i ) 2”^ - 2nrn ♦ er11 - - (1 - t) crn - (1 - K • Accordingly) since (U) implies (8), the theorem is proved* 1*1^12* Lemma* Let S, K^, Kg be subsets of R® , and let (i) K p Kg are L^-measurable j (ii) j (ill) L K g > j (iv) I§ (S Vi 1 ^) > (1 -**) 1 ^ . Then L® (S H Kg) > 0 Proof: Suppose, contrary to the lemma, that (1) Ln (sf U 2) - 0 B y (i), (ii) we have (see 1*2*8 ) (8) V i * - *h<*rt> • But in view of the obvious Identity S P\ (t^Kg) - S f 4 - s ' \ Kg we obtain the inequality (3) l£[S f \ (Ki-Kg) ] > l£(S rk Kj) - L°(S \ Kg) . But because of (1) it follows from (3) that (U) l£[S i A (I^-Kg)] - L°(S ,1^) • Application of (U), (ir),(iii) yields (5) I?[S (K^)]>(l-t)L°K1 - LnK1 - 2 Vi ' Vi • Now it follows from (2) and (5>) that (6) - ycs - > yts A (y*,)l » -Vi~ Vt 4 But (6) is Inpossible, and so (1) cannot hold* According!/, the leraa follows* n I*JLi*13* Lemma* let p, q fc R , and let ^ be any assigned number with 0 < ^ < 1 • Then if we set h * U p - q i| the following statements hold (see I.U.9 for notation) t (i) K(qj i^h) CL. K(pj (1+^) h ) j (ii) L K(qj Ah ) > y? 2*n 1 K(pj (!♦ >!) *» ) ♦ n • n -31- Proof: To prove (i) we take any point x t K(qj ^ h) » and note that therefore by I.U.9 we have U q - x il < *\ h • Then using the triangle inequality we obtain tip - x H < ftp - qil ♦ ttq - xn < h + ^ h - (1 + t) h , and so indeed x ^ K(pj (l-*^)h ) • Thus (i) holds. To prove (ii) we apply I.Jj.10 and the bound ^ < 1 to obtain LK(q; ,^h) - c /“hn - “ 2-n c 2“hn > ® 2"n c(l+ ^)”hn - - ^(pjd+.^h ) , and so (ii) holds and the lemma follows* n I.li.lli. Theorem* Let S be any subset of R , and let p <-r S • Then for any assigned number ^ > 0 there exists a T « 5(^ ) > 0 such that to each point q t r f 1 with kip - qu < 3 there corresponds a point q^u: S satisfying the inequality UP - qU . Proof: Clearly we may assume without loss that < 1 • Then since by hypothesis p q iis there exists (see 1 *14*1 1 ) a £> » e: ( ^) > 0 such that for any closed spherical neighborhood K(p;r) of p with radius r > 0 we have - 32- (1) Ig [S H K(pjr)] > (1- I^K(p|r) if r Define a nnriber 5 by the equation (2 ) (1 ♦ ^ - S* Now take any point (3) q ^ with p - q tl < 5 • Upon setting (U) h - i[P - q u we note that (see I«U«13) since »j < 1 we hare the relations (5) K(q; -Vh) - K ( p j ( l ^ ) h ) , (6) LK(qj ^h) > °2’n D/CpjCl* -\) h) . But (2), (3)> (U) imply that (levph < S * , so that from (1) it follows that (7) l£[s OK(p;(l+'»\)h] > (1 - ^ “2-n) I^(pjO>*\)h) . Now let us put K^ • K(p;(l+V|)h), Kg ■ K(qj v^h) , c m ^ 2*n • Then noting (5), (6),(7) we may apply 1*1^12 to obtain I^£S< \K(qj^h)] > 0 and so there exists a point (8) S ( \ K(q; v^h) • But (8) implies that (9) \ u S and (with (10 ) that (10) W % - < -\h • ^ ii P - q .1 • Thus, (3) implies (9) and (10), and so the theorem improved. I.U*l5. Theorem. For fixed n > 1 there exists a number °\ (depending only on n) such that 0 < 4 < 1 and given any two points p^, pg tr we have (see I*lt«9 ) « V i iCPj* UPj-Pg** )f\ X(P2> fePj-Pgd ) 3 V*(Pi» aPi-Pgkl) + V ^ * ui^-Pgii) Proof* The proof (which we omit) follows essentially from the fact (see Mayrhofer (4 ], ^ 19) that under a linear transformation the L -measure of a set is multiplied by the determinant of the n transformation* I«lt*l6* Theorem* Let S^,Sg be two L^-measurable subsets of Rn , and let p^, Pg ^ R*1 • P®* (®ee 1*^*9 ) (i) - Ktpp Ul^-PgU ) , Kg » . *C P2, tt P^PgU ) $ (ii) -3U- and suppose (iii) y s ^ A > ( i - u ) ^ , y s 2 i \ k2) > (1 -«<) LbK2 • Then there exists a point p i Sg such that d P - P^n < 1 Pi - P2 it » d P - P2 li SwPi “ • Proof: Let (1 ) ®x-sx('Ki ,E2 -s2r^x2 . Then we have (see 1.2*10 ) (2) A e2) ♦ y o Eg) - 1^ ♦ 2y , (3) y ^ \ Kg) ♦ Ln(Kx j Kg) - y x ♦ y 2 , and (since E jl J Eg w ^ J Kg) (H) y ^ jig) < y K x j K g ). Successive application of (2), (iii), (U),(3)» and (ii) yields \(\(\ E2) - y 2 - y Ex J Eg) > > ( i - c O y x ♦ ( i ^ ( A ) y 2 - y i ^ j K g ) - - CL-oOCyyV 2) ’ I W 1V[2‘i“(!ci 1'1 V ] * - I* ( K ^ Kg) - c x f y ^ ♦ L^Kg) - 0 Thus E^P\ Eg^t and so by (1) and (i) the lemma follows* I.Uol7« Corollary. Let S,, S be two L,-measurable subsets of the 1 2 l -35- real-line R^, and let P^,P2 ^ 4et ^2 ^eno^e ^'wo intervals of length s= 2 |\ px - p2 |l with centers at p p P2 respectively, and suppose l a ^ O I])^(3/4)S , \ ( s 2 C\ i2) > ( 3 / 4 ) s . Then there exists a point p^ £ Sj n s2 such that 1 p^-p#\ — 1 P1 “ p2 1 , 1 P2 - P*\ |Pl - P2 I . Proof: Assuming without loss that I-., I are closed, we apply 1.4.16 with n s 1, I 5 K , I = K , and o( - 1/4. 1 1 2 2 1.5 Real-valued Functions on Rn 1.5.1 Agreement. For the following definitions and theorems let f be a given real-valued function defined on a given bounded subset S of Rn. (We shall understand the term real-valued to mean finite real-valued.) 1.5.2 Definition. Given a point p £ S, f 3s said to be continuous at p if lim f(x) exists and is equal to f(p) as x approaches p through S. If f is continuous at each point of S, then f is said to be continuous on S. 1.5.3 Theorem. If S is closed and f is continuous on S, then for every number c the set ^ P | P ^ S, f(p)c | is closed. -36- Proof: Given any sequence of points p. e S with >PQ and f(Pi) < c we have pQ t S since S is closed, and f(pQ) * lim f(Pi) < c since f is continuous at pQ. Thus the theorem follows* I*5.it* Theorem. If S is closed and f is continuous on S, then f is uniformly continuous on S, in the sense that for any assigned <= > 0 there exists a *''(-)> 0 such that If(x^) - whenever x , x c S and H r • x il < r « 1 2 l 2 Proofs See Graves [2], Chapter IV, Theorem 23* 1.5*5* Theorem. The stun, difference, and product of two continuous functions on S are continuous functions on S. Also, the absolute value of a continuous function on S is a continuous function on S« Proof: See Graves [2], Chapter IV, Theorem 16. n 1*5*6* Definition. Given that S is an L -measurable subset of R , n then f is termed an I^-measurable function if for every number c the set ^ p \ f (p) < c is I^-raeasurable* 1.5*7* Theorem. The sum, difference, and product of two Im measurable functions on S are I^-measurable functions on S* The absolute value of an Immeasurable function on S is an Im measurable function on S. Proofx See Saks [8], Chapter 8, Theorems 8.2, 8.3» and 8*1;* 1.5.8. Lusin's Theorem. If S is an I^-measurable suboet of Rn and f is an Immeasurable function on S, then for any assigned G ■>■ 0 there exists a set E ■ E( G ) for which the following statements hold: (i) E d S, (ii) E is closed, (iii) Ln(S-E) < (iv) f is continuous on E * Proof: See Saks [8], Chapter III, Theorem 7.3* I*5*9. Corollary. Under the hypotheses of 1.5.8, assume that a property P holds at almost every point of S(see 1.2.11). Then for any assigned (i) E C S , (ii) E is closed, (iii) y s - E ) < , (iv) f is continuous on E, (iv) P holds at each point of E • Proof: By hypothesis there exists a subset S of S such that I (S-S*) « 0 and property P holds at eachpoint of S*• Let n € > 0 be assigned. Then applying Lusin's Theorem to the set S* we obtain a closed subset E of S* such that L^S^-E) < -38- Thus properties (i) - (v) are clearly satisfied, and so the corollary is proved* 1,5.30. Definition* Let a positive integer j < n be assigned. Then f is said to have an S-relative j-partial dexrivative at a point p c S if there exists a (finite) number A such that lim exists and equals A. We write f (p) * A 1*5.31. Theorem. If S is an I. •measurable subset of Rn and f n is continuous on S, then for each positive integer j < n the S-relative j-partial derivative f is an Immeasurable function on the subset of S consisting of all points p for which f^(p) exists. Proof: See Caratheodory [1], ^ 557, Satz 1 • 1*5.12. Definition, f is termed Lipschitzian on S if there exists a (finite) constant M > 0 such that the inequality IfCxj) - f(x2) \ < M U x1 - x2 U holds whenever x^, 1.5.13. Theorem. If f is Lipschitzian on S, then f is continuous on S. Proof: If lf(p) - f(x) \ < M l\p - x U , then I f(p) - f(x) i < whenever 11 p - x ii < and so the continuity of f on S is immediate* I*5*111* Theorem* If f is Idpschitzian on a finite interval [a,b] of the real line, then f is of bounded variation on I, in the sense that there exists a (finite) constant N such that for any sequence of non-overlapping intervals ^ [a,, b ^ J with end points in [a,b] we have | fCb^ - f(ai) \ < N . Proof: By hypothesis there exists a constant M such that I f(x^) - f(x?) V < K I x] - x? \ whenever x,, x„ CT fa,fc] . Then given any sequence of non-overlapping intervals ^ ^ai’^i^ ^ with end-points in [a,b] we have Z i I f(bt ) - f ( ^ ) [ < K 2 I bj- a1 1 £ M 1 b • a \ . Thus, if we set N • M \ b-a\, the theorem follows* 1*5*35* Theorem. If f is Lipschitaian on an n-cube Q, then for t each positive integer j < n the j-partial derivative f exists at J almost every point of Q* Proof: Since f is Lipschitaian on Q it is lipschitaian and therefore (by 1*5.35) of bounded variation with respect to each variable. The theorem therefore is a corollary of Caratheodory [1], ^ 559 > Satz 3* I*5*l6* Definition* Given any subset E / 0 and any point p e E , f is said to have an approximate E-relatlve limit at £ if there exists a (finite) number A such that for any assigned € > 0 (see 1 *14*2) we have p We write ap lim f(x) ■ A xeE, x ~?p 1*5*17* Theorem* If for a givensubset E of S, f has an approximate E-relative limit at p e E, then f has an approximate S-relative limit at p* Proof: This follows at once from I*5*l6 and I*li*ii* 1*5*18* Theorem* If p c E fl A E and f has an ordinary E-relative limit at p, then f has an approximate E-relative limit at p, and lim f(x) • ap lim f(x) * xeE,x-^p xeE, x-^p Proof: See Saks [8], Chapter VII, £ 3, remark 3*2* II. THE SECTIONAL DENSITY THEOREM 11.1. Definition# Let ^h^, ... » ^ be any proper subsequence of ^ | and let it* be the conpLelementary subsequence (see 1*3. I )• Then for each subset S of B® we define a subset D(ff,S) of BP as follows (see I«lu 2 ): D(i»,S) » ^ p \ p 6 R“ , flt*p ^ «*(S f\ w " 1!*?)}. 1 1 . 2 . Lemma. Under the hypotheses and notation of I I . l we have the formula D(w,S) - Vj ^[w*~1 A. «*(S C\ nC^Xjj) ] C\ xk where the union is taken over all points Proof* Let p €. D(«,S)• Then by I I . l we have that rf1 and w*p ^ A * * ( S f\ u^^iip). Then if we put p^ ■ up it is clear that P 6 « * ( S (\ w * 3^ ) ] 0 u*3^ , P j ^ e Rk , so that since p was arbitrary we have the inclusion (1) D(u,S) C_ U w*(S O W 3^)] H w"3^ # *k v Now suppose for some p ^ t E we have p «*(S H w“,1pjc) ] 0 w*3^ Then clearly p <=. tf1 > «p • > and -1*2- n*p (a Av*(S n n^p) , Accordingly (see 11*1) , p ^ D(*,S), and so (2) D(w,S) U [w*-1A **(S ( \ «-ax )] Pi ♦ *k With (1),(2) the lemma is proved* 11*3* Lemma . If r “, then C- ©(w,^) * Prooft This follows at once from 11*1, since ** and A are monotone (see 1*3* ^ and I*.U,4 )• II.Ii* Agreement* For paragraphs 11*5 to H.ll* inclusive assume that n * •••» 1 is a prop®? subsequence of ^1, • n| and S is a given subset of H?. Accordingly, for brevity we agree to write D instead of the more complete notation D(«,S)# It *5* Lemma. Given any point p^ Cs , then «*(D (1 m A w*(S 0 Proof* We first note that since « is single-valued, if ^ p^ we have * 0 $ so (see 11*2) ® ^ *"lpk “ 'l ^ [ n ^ A **(S ft ff"^) ] H w-1^ . \ 0 s " ^ • l x k • U [ w ^ A w*(S a u \ ) ] a * \ [ \ nm\ - xk » a «*(s aw"^) ] ft * —1*3— We now apply 1*3*7 (where we put » A**(sO^ ' - V v and thus obtain \[■*j a«*(sa < r \ ) ] a ^ - » Aw^sHe?*2^) * whence the lemma is proved* u 11*6* Lemma* Qiven any point p ^ G R , we have w*[(s-j5) n n"2^ ] * As 0 *r\) - A n H A\) Proof; Apply 1*3*10 with ■ S, Sg • D * II*?* Lemma* Given any point p^<= , we have ^w k \«*f(S-D)n«*Sk] \ - 0 Proof; By 11*5 we have the formula (1) **(D H w*2^ ) « Aw*(S O w ^ ) * From (1) and the Density Theorem (see I*U*3 ) there follows the equalities (2) C\ w*1^ ) • «*(D a w- 3 ^ ) ] - • r\ «*,1pk) • a . w*(s r\ * o « Now 11*6 , (2) imply Ln-k \ ^ trlpk3 \ * 0 » -UU- and so the lemma Is proved* 11*8* Lemma. If S*D is I^«aeasttrable, then yS-D) - 0 • Proof: Since by hypothesis S-D is Immeasurable, the lemma follows at once from II,7 and the Fubini Theorem (see 1*3*13 ) 11*9* Definition* For each pair of positive integers i, j we define a subset of Rn, depending upon « and S, as follows: p & D ^ if xwSc and only if pfc rf* and for any open oriented (n»k)-cube Q of side* length s the inequality holds whenever s < 1 and «*p <= Q53"^ * 3 11*10* Lenaa* In the notation of 11,9, we have D « ' • O D. i - i 1 J Proof: Assert first that o a (1) D C (I U i-1 J-l J To prove (1) take any point (2) p 6 D • Then by II.l (3) p C - J 11, -4*5- (It) «*p € A**(s 0 witfp) * Assign an arbitrary positive integer 1* Then (It) and the definition of outer density (see I*lt*l ) require the existence of a positive integer 3 such that (5) for any (hence any open) oriented (n-k)-cube of side»length •i »» Hwk s the statements s < =■ and u p c Q imply L°n.ktQn-kr\ «*(s a > (i-(i/ijsD"k. From (3), (5) and 11*9 it is clear that for the assigned i and the corresponding j we have p £= , and so since i was arbitrary it follows that O O o o (6) p c n u d i-i j-i 13 Mow since (2) implies (6), assertion (1) is true* Assert now that o o orO (7) n U D-. d - D * i-1 3-1 To prove (7) take any point O O ck , (8) p H U • i-1 j-l J Froa (8) it is clear that (9) p^R® , and 0 be any assigned number, such that the following conditions holdt