Scribd Logo
Upload

Welcome to Scribd!

  • Mate

    Uploaded by

    JosueAlmeyda
    11 views43 pages
    matematica

    Original Title

    mate

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    matematica

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    11 views43 pages

    Mate

    Uploaded by

    JosueAlmeyda
    matematica

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    of 43
    NOTAS DE MATEMATICA Ne 14 COLECAO PUBLICADA SOB A DIREGAO DE L. NACHBIN DECOMPOSITIONS OF THE SPHERE BY DJAIRO GUEDES DE FIGUEIREDO FASCICULO PUBLICADO PELO INSTITUTO DE MATEMATICA PURA E APLICADA. DO CONSELHO NACIONAL DE PESQUISAS RIO DE JANEIRO 1958 PREFACTO As presentes Notas constituiram a tése de Master of Science que o autor apresentou ao Institute of Mathematical Sciences da New York University. Bsse é 0 motivo pelo qual elas aparecem em inglés, Trata-se de um trabalho de natureza exposi, téria, No primeiro capftulo, apresentamos um resultado eldssico de Hausdorff, dando uma formilag&o ligeiramente mais completa que aquela originalmente dada por Hausdorff. .Nos ca- pitulos seguintes, seguimos a orientagSo de um artigo de R.M. Robinson, Introduzimos, porém, uma modificagdo no sentiao de que, na classe de movimentos rigidos, permitimos também refle- xdes. Deixamos assinalados os nossos agradecimentos ac Prof. W.M, Hirsch, pela orientacdo que nos dev durante a lei- ture dos diversos artigos que conduziu ao presente trabalho. fTambém agradecemos ao Prof, Leopoldo Nachbin, por incluir o nosso trabalho nas "Notas de Matemftica", Djairo Guedes de Figueiredo New York, Novembro de 1958 CONTENTS Introauction I, A free non-Abelian subgroup of rotations wissseseeees 8 Application: The Hausdorff Paradox ssssscsessseeeeees 1B II, A free group of mappings in the sphere 1, A free group of mappings in the sphere ......4.4. 16 2, Classes of equivalence on the sphere ..sssseeeees 17 3, Representation of elements in each class .. III. Decomposition of S according to relations Ly Relations soe se ceececececcecereesecentcestecenees 22 2, A theorem of decomposition sesesssceveevccesevves 26 IV, Decomposition of S according to congruences ..+e+0e+ 30 V. Applications 1, A sharper form of the Hausdorff Paradox vee 35 2, Theorem DS ciseseressensneeneeseeeeservecsenses 35 3. A particular decomposition seressessseeeseeeeeaes 36 4s Theorem DS! cisvessesecrscecceccevsecevecveccace 37 been teeeeeeee eeee 40 Bibliography sesseseseaee In this paper we deal with some striking statements concerning the congruence between a set and a proper subset of it. We begin by presenting some simple examples in order to get acquainted with this sort of ideas. 1) On the unit-circle we consider the set A of points whose angles with the x-axis are ma , n= 0,1,2,.00, where a is such that a/j is not a rational number, let B be the subset of na for which n= 1,2)... . If we rotate the set A by anangle a then it will coincide with B. 2) In the complex plane we consider the set A of points given by polynomials in e* Bylet) = eglet)Peog y(et) Mes etopette, 4 meOply very where the coefficients are non-negative integers. Let A, be the subset of A consisting of all points such that ce, = 0. Let Ay = AnAy It is easy to see that the set A coincides with A, by rotation of one radian around the origin (ive. multiplication by et ), On the other hand the set ‘A coincides with A, by translation of 1 in the positive direction of the real axis (i.e, by addition of +1 to the points in A). 3) The set A is countable in both of the preceding examples, Now we give an example in which the set A is not countable, Again let us consider the complex plane and to each complex number 2% we associate the set E(z) of complex numbers of the following form Ry(e*).2 +R le"), (2) where R,(e*) and R,(e*) are rational functions of e* with integral coefficients and R,(e*) 40 . We can easily see that if zz! , then either E(z) = E(z') or E(z)n NE(z') = 9 . So the complex plane is divided into equivalence classes EZ. let E, be the cass thet contains 0. We form aset B by choosing a point from each of the remaining classes, In doing so we are using the Axiom of Choice, We observe that the set B is not countable. Let @ be the function that associates 2 +1 +o each point 2 of the complex plane, Let R be the function that associates z.et to z. For cach 2 we define a set C(z) as the set of the images of z by all the products of @ and R such that the last factor is 1. A similar definition for D(z) , being R the last factor. Then the following inclusions are true Cz) c {we w= Py(e*).2 + Poet) , P,(0) £0} D(z) c {we w= Py(e!).2 + P(e) , Blo) = 0}, where P,(c*) and Pj(e*) run through all the polynomials with non-negative integral cocfficients and P,(e*) #0. Now we take c= Uo, v= Ubcey zeB zeB and let A=B+C+D. We claim that CND=@ . According to (2), both O(z) and D(z) are contained in E(z) , for each 2%. So we have only to show that C(z)MD(z) = % . Suppose not. Then Py(o).z + Po(e*) = Qler)ez + Oleh) , where P,(0) #0 ana Q,(0) = 0. This last condition implies that P(e") # (et) . So Q(e*) ~ Po(e*) Pye) - Ale which implies that 2B, . This is not possible because 2B . It is easy to see that BCA) R(A) 0 Summarizing we have presented an example in which the sect A is not countable and contains two disjoint proper subsets cach one being congruent with A itself, We observe that both sets © and D are not bounded in the usual R?metric. A historic question is the so-called "measure problem at large". That is, whether it is possible or not to define a real function f for all bounded sets A of an Euclidean space R” such that: i) £(A)20 for all A. ii) £(A,)>0 for some set A, in RP. iii) f(A+B) = £(A) + £(B) if ANB=G, iv) £(A4) =2(B) if A~B. ~ - means congruence (*). Such a function f is called a finite measure, Property iii) is known as finite additivity. In 1923 Banach [2] gave an affirmative answer to the problem for the case n= 2. In 1914 Hausdorff [6] obtained a decomposition of the surface S of the sphere into four disjoint subsets A, B, C, and D such that (I) S=A+B40+D (II) a R BeC , A=Beeo , where D is countable. Here the congruence ~ means only a rotation about the center of the sphere. Hausdorff then gave a negative answer to the measure problem for the spherical surface S$ , basing it on (*) The concept of congruence will be made precise on pages 3038 For the time being we understand that A*B means that A can be superimposed upon B by rotation and/or translation. Introduction 5 the above decomposition, Indeed it is easy to see that if such a.fumction f exists then f(D) must be equal to zero. This follows from properties iii) and iv) and from the fact that D is countable. Now congruences (II) implies that £(S) = 3f(a) , £(8) = ef(a) . Then £(S) is equal to O and so property ii) cannot be fulfilled, The impossibility of defining a measure (even finitely additive) for ell the subsets of the surface S of the sphere is closely related to the existence of a free non- ~Abelian subgroup of the group of all rotations of the sphere. The negative answer to the measure problem for $ implies a similar negative enswer for R™ ,n>2. The decomposition of the surface of the sphere given by Hausdorff is known as the Hausdorff Paradox, It is the source of subsequent amazing facts presented by Banach and Tarski in 1924, Banach and Tarski [3] using Hausdorff Paradox and an extension of the Schroder-Bernstein Theorem for an equivalence relation that they introduced, diery e- quivalence by finite decomposition, showed that any two bounded sets (*) in R? are equivalent by finite decomposition, (*) With interior points, That is, there exists at least one point of the set such that a neighborhood of it is entirely contained in the set, Introduction 6 That is, if A and B are bounded sets in B then there exists an integer n such that Ash thy tae tA, 4 B= BL +B tee. +B, ANA, =8 , if i > BAB=%., if§ and The. key point of the proof of the above statement is the following theorem: "If S, and S, are congruent spheres in R3 then S, is equivalent by finite decomposition to 8145)". In other words, Theorem DS': "There exist three decompositions of the three- dimensional sphere S! into disjoint sets Lt Ab tee + At 2) St = Bl + By + eee + BY 1) St= 3) St= Bey tere + By such that AL& B 2 =1,2,.0-,ke0. 0 ios In their paper Banach ana Tarski did not say any~ thing about the values of the integers k and ‘(, Sierpinski 9 in 1945 showed that k equal to 3(or 2) and l equal to 5(or 6) are a solution of the problem. The question received a complete answer in 1947 when Robinson [7] showed that k = 2 and (=3 are the smallest integers that solve the problem. The greatest part of ovr work is to establish the Theorem DS: "There exist three decompositions of the surface S of the three-dimensional sphere into disjoint sets: + Acee 2) S = By + Bo tae. + By 3) Se Bey tees + Bie VD) SHA tant such that + L=1,2,...,k+0. 9 Then it will be quite easy to prove the theorem DS' . We follow Robinson's paper [7] . In it congruence means superposition by rotations or translations, Here we admit any rigid motion, that is, reflections are also possible. Although some of our theorems are simpler than those of Robinson we do not improve his values k= 2 and ?=3 when he admitted only rotations and translations. I, A FREE NON-ABELIAN SUBGROUP OF ROTATIONS Let G be the group of all rotations about the origin in the three-dimensional Euclidean space, Such e group G can be represented by the multiplicative group of all 3-by~3 orthogonal matrices (*) with determinant equal to +1. We will show that there exists a free non-Abelian subgroup H of the group G of rotations. (A group is said to be free if it has a basis, i.e., a set of elements such that any element of the group can be uniquely expressed as a product of elements of the basis with integral exponents.) Let us consider the rotations g and ‘Y given by ~cos@ 890 sin® hn 0 g= 0 -1 0 , Ys u A 0 , sine ° cos® I 0 ° 1 where A= cos 21/3 = -1/2, w= sin 27/3 =/3/2 , @=2. Looking at these matrices we see that ‘Y is a rotation of 21/3 about the z-axis in the positive direction and » is a rotation of " around an axis H which is in the xz-plane and makes an angle of 6/2 =1 radian with the z-axis , (*) A veal matrix A is orthogonal if AA* = A*A = 1, where A* is the transpose matrix of A. Lemma 1 - The rotations g and ‘are independent except for the relations o@ =Y3=1, Proofs We want to prove that for every positive integer nthe following expressions are different from 1: a= oY Tey em a= Py ot y= ov te 2, ie? 9 wa mut Ly 6 = PID Ae , where ™,M,+.+,", , are equal to 1 or 2, (Observe that tag, whey? , yay) ’ We need only to show that the products of the a form are different from 1, Because, if a 8 was equal to 1, then an a would be equal to 1, namely gfy. Ifa § was equal to 1 we can easily see that an a’ should be equal to 1. Finally if a y wes equal to 1, then a § would be equal to 1. Let N be the point (0,0,1) in the surface of the unit-sphere. If we show that for every a, Na #N, we will be through. In order to prove that, we begin by showing that N is transformed by a into a point P with coordinates Ta nel x = sin@ (a,cos "6 + 44, + a) (1) y = noine (b,cos™10e + 4., + b,) 7 x, Z = 0,008"0 + +4 + 6 ’ A free non= where x/sin® and y/usin® are polynomials of degree a-1 in cos@ with rational coefficients, and z is a similer polynomial of degree n. The proof of this will be by induction over the number n of double factors in a which ave of the form gtlor oly? ). The statement is true for n=1 , because the point N will go into Ney (or NY? ) with coordinates ° [asino | - fo Asin Ya 6 x |o = | sino » yx ob x{o| = |-nsine | (*) 1} | cose La L cose Let us suppose now that after n steps the point N goes into a point P= (x,y,z) where x,y,2 are given by (I). Then it is easy to verify that Pe'Y or Pyty? have coor~ dinates : 7 n, x sind (ai, cos"® +... + af) yi usin (bi, cos +... + b4) : 1 n+1, ' z Ob eos @ + eee + Ch where x'/sin@ and y'/ysin® are polynomials of degree n ana have rational coefficients and 2! is a similar poly- nomial of degree n+l . @hen the point N is transformed by a into a (*) Observe that we 2 lop aA oo point P with coordinates given by (I). Now we prove thet, 0, is different from zero for each integer n. The coefficients aj, and of, are given by Bhar = My ~ AB, = Moy ~ 8) Snel = Cn ~ An This implies ot el 7 Maa = CE - ANCey = 8) + Since for n= 1 o,~a, = 1-2, we see that oa, = (a-a) and so ey = (2 - aR? = (3/2)2# Ten, for each n, the coordimte s is given by a = (1-A)™eosor0, jco8 Ter .40, 5 where are rational numbers. Since cos2 isa ele tres So transcendental number it camnot be a root of an equation with rational coefficients (ay Mee, pte te-2 = 0. ‘ne 0 Then N is always different from N, i.e., no a is the identity rotation of the group G. CED 2-let 9 and ‘y be the two rotations defined before, Then the rotations n= gyoy and p = gy-gty® are independent. Proof: Let us suppose by contradiction that where ij, ip,.+., i,, are integers,” We assume without loss of generality that i,,...; ip, are different from zero. Now we replace p and p by their expressions in terms of 1 2 uy 2 -1 . =Yey*o and ep =Woyo. Then we effect all the possible simplifications in terms of gp and y . We observe thet py g and yy , It is clear that there is no simplifications of g 's and ‘y's in the following double products: 4, pp, no, on, wtp, ptt, atu, ovtont, upd The other ones have the following simplifications: ute = (owe (oy—y?) =yee yoy? , oth = (Wo¥e(oyoy) = Yowroy , ou = (9 Ye Y*)(Y2—¥%e) = gry®pyg yy -. So we see that in-every case the first two factors (4y? is counted as a single factor) and the last two factors remain untouched. Then, if i, 40 , the first two factors in wt remain untouched. If i, = 0 , the first two factors in, p'2 remain untouched. The same reasoning for the last term. So we obtain ean expression in one of the forms a, 8, y or § (see p.9), But such an expression cannot be equal to 1 as we have already proved (lemma 1), QED Lemma 2 assures us that there exist at least two independent rotations in G. (*) Some of them being different from zero. Let » and p be two independent rotations, Then so are \) = HELE, Ayo = 107 L0? sve ay yy = He™He™ for any positive integer m. Proof: The proof is similar to the proof of lemma 2. We do not give it here. 20 Given any integer m we can find m independent rotations, namely, \y') aor +Yq + Such rotations generate a free non-Abelian subgroup of the group G of rotations, APPLICATION: THE HAUSDORFF PARADOX. Let us consider the subgroup F' of G spanned by and Y. According to lemal @ and \) are independent except for _9* = 4 = 1. Therefore every element (AL) of G' 4s in one of the four forms «, B, y or 8. We note that the number of elements in ¢' is countable. Lemma - It is possible to divide G!' into three disjoint subsets A’, BY, and ©! such that i) given two rotations « and ag (*) one belongs to A' and the other to B' + Ct 5 14) given three rotations a, aly, and, a® (**) one and only one belongs to each of the sets A', Bt end C', (*) In this product, a cannot simplify with ». (**) In this product, a camnot simplify with Y or ‘v2, IW A free non-Abelian subgroup of rotations Proof: let us start by alloting 1 %o A', y to B!, \ to Bt, W® to ©! , and so on, using the two above rules i) ana it). We proceed by induction. Let us assume we have done this for any element with n factors or less. (Y* is counted as a single factor). An element of n1 factors is in one of the forms ag (*), aly, atv’ (**), where a is an clement of n factors. Then, if aeA',B' or O', ag will be alloted to B',A' or A'(resp,); if aeA',B! or Ct, aty (or a'y*) will be alloted to BY,C' or A! (or C', A’ or B')(resp.). QED Bach rotation around the origin is equivalent to a rotation about an axis through the origin. So in each rotation two points of the surface § of the sphere remain fixed, Let D de the set of such points for rotations (41) in G@! , The set D is countable. Now for each xeS-D we define the set B(x) = {ys y= xa, aec'} It is easy to see that if x,yeS-D then either E(x) = H(y) or E(x)NE(y) =@. Then S-D is divided into equivalence classes, By choosing a point from each class (axiom of choice) we forma set M. It is obvious that sp= \) ma. sets (*),(*) See foot-notes, p.l3 _A_free non-Abelian subgroup of rotations 23 Let us consider the following three sets ; a=\) wm, Bell m, c=) m . aca! aeB! aec! It is easy to see that the three sets are pairwise disjoint and that A+B+C=S-D., It also follows from the construction of A',3' and C! that Ag = BHC AyeB , ay® ao, Therefore A=B+C , AxB2C . QED Essentially the division of G! was done in such a way.as to obtain the desired congruences. The process was simplified extremely by Robinson, He introduced certain relations corresponding to the given congruences and showed how to obtain a decomposition of the surface of the sphere in parts satisfying those congruences. We will return to this later, It, A FREE GROUP OF MAPPINGS IN THE SPHERE 1 So far we have been concerned only with rotations. Now we introduce reflections of the surface S of the sphere. A reflection of S with respect to its center is the mapping W that transforms each point of. S into its diametrically opposite point. It is obvious that W commtes with any rotation of G. In this paper we use the term "mapping" +o meen any finite product of rotations and reflections in S$ . These mappings are isometric. That is, if x ana y belong to S and their images, by a mapping g, are xp and yo, then 4(x,y) = a(xy,yp) We have proved in the preceding section that the group G contains any finite mumber of independent rotatione(") So let ,Yyy+e+y Wm » be independent rotations of “e. They generate a free subgroup H . We define the following mappings: % = Yo me Ye 8 te (*) Dekker and De Groot recently showed that there exist N independent rotations in ¢ Nis the cardinal number of the real numbers. . A free group of mappings in the sphere _ THEOREM 1 - The mappings qo) Oyreres Oy 1 are independent generators of a free group $ of mappings. Proof: Let us assume by contradetion that a dj L s pr eee ty Fig where O). Then y can be represented as y= xa' , where co! = 0a , It is obvious thet o! begins neither with @ nor with 9;'8. s Now we assume y = xa , where a begins with a « Ten y=xa! , where a! = @c and a! does not begin either with the bloc @ (observe that 6 simplifies) or -j with 9, *. In order to see that such a representation is 8 unique, let us assume xa = x6. Then x(ag7) =x. According lemma 3 it follows that and so ~ If n>0, since ©"p does not simplify because 6 does not begin with a we conclude that a begins with the bloc 8, which is impossible, If n Rk,t) =1. In short we say: 9 is compatible with R. 1-I£ © is compatible with R, then g? is compatible with RB, ~ Let us suppose that 4 gtoK AB. Then Fe (hy TAOBAO OA, 6 Since » is-compatible with R it follows that LE R(yyt)sl. So el(l,) =2 . om Louma 2- If 9, 4s compatible with Ryy 1 = O,lyyeeym, ay js, . then 9 = 93) +++ 93° is compatible with . 2 8 j i 7 R=R,".., B)°, where the j's are equal to #1 ana the a 8 i's ere integers between 0 and m. Proof - The proof is by induction on the numbers s of factors in p-. For s=1, the statement is precisely lemma 1. et us assume that it is true » with s factors. We claim that gp) is compatible with RR} . Indeed, let us. suppose A Coop) Oa Ao. Then (agg) nay A Gs which means that there exists an x in some Ay such that © xpleAy . Therefore Ri(p,t) =1. On the other hand, since xeA, and EAL y it follows that R(k,p) =1. Then RRg(k, 2) =2. QED the m+1 canonical relations RUM Let RoyererBy y defined on page 25 and goysrey%q 9 DE the m+) independent mappings that were introduced in §II. (See page 16), Then we can decompose the surface S of the sphere into n disjoint parts 4,,...,4, , in such a way that each mapping 9, is compatible with R, , for such a decomposition, Proof - We have proved (SII) that the group § determines a decomposition of the surface § of the sphere into equivalence classes .E . Now we have to prove that we can allot the points of each class to sets A,,...,4, , in such a way that the mappings 9, (i = 0,1,...,m) are compatible with R, (i = 0,1,...,m) , for the decomposition of S$ into these sets. Classes of non-fixed points: From each class of non-fixed points choose a point x (axiom of choice), As it was proved in § II, all the elements of a class of non-fixed points can be represented as xa , where a runs through all the mappings of 9. Let us start by allot x to any Ay. By induction we suppose that xa was assigned to A, . Then xe} (shere gJ docs not simplify with a ) wowla be put in A, 80 that Ri(x,t) = 1. Classes of fixed points: let # be a class of fixed points, Let @ be a shortest mapping of @ which has a fixed point x in the class EB. As before we denote 6 by e= oe aa as end the only closed cycle in B will be We have no trouble with the points of EB which do not belong to the cycle. They can be alloted to the classes AjyeoesA, y in the same way as we did for the non-fixed points. But we must be careful with the points in the cycle. They cannot be put arbitrarily in the sets A, . Let us see why. Suppose J Jee. 4 we put x inaoclass A, and seit a ha imaclass Ay so that : J Ig: (at REG, L) a2 1 8-1 4 3 j Now (aot wee aes =x must be in 4, . But if we have i j 4d RC L,x) = 0, then eis is not compatible with Re . We avoid this by proceeding as follows. ay The relation R = Ry 1 J yee RU® has a fixed point k , because 8 it is a relation corresponding to a mapping with a fixed point (lemma 5, p.25). hen there exist integers ky =k, Kyyeeerk, = kK , between 1 and n such thet J, 3 fl rT . 1 ~ Gg) = 1. If we allot x toa, end x93 1g” to Ae, (v = l,os+y8-1) we will be through. QED 30 IV, DECOMPOSITION OF S ACCORDING TO CONGRUENCES DEFINITION: Two sets A and B on the surfae2 8 of the sphere are said to be congruent if there oxists a mapping 9 (in the sense of p.16) which transforms A onto B, We write A“B for Ag=B. The following properties are obvious: i) Awa, di) AB then Bea, iii) A~B, BYC then AG So = defines an equivalence relation between the subsets of S . Such a relation is called congruence, Let Ayyeery4, » be @ decomposition of S$ into nm? n disjoint parts, A congruence Ap tees tA we A ee A aa Be a ls (where 1

    You might also like