3 73 PI MU EPSILON JOURNALThis can be written more compactly by lettingFrom the Taylor series for e x about x = 0,converges to 1 - 1/e as n - m. It should also be noted that 0 2 a^a< 1 for all k andn, and that for each k lim = 1. These facts taken together would suggestn--(- l)k+lthat for n large, pQ = -, and therefore that lirn p = 1 - lle. Ak=1 k! n-00careful proof of this result is somewhat technical, and is deferred until the end ofthis note.If the 2n chairs are arranged in a circle, a couple occupying chairsnumbered 2n and 1 would then be in adjacent chairs. The event that at least onecouple occupies adjacent chairs in this situation can be thought of as occurring inone of two mutually exclusive ways: (1) at least one couple occupies adjacentchairs other than the two numbered 2n and 1, or (2) a couple occupies chairs 2nand 1 with no other couple occupying adjacent chairs. The first of these eventshas probability pn. For the second event we note that the probability of being inthe adjacent chairs 2n and 1 is l/(2n - I), and given that a couple is seated therethe probability that no other couples occupy adjacent chairs among the remaining2(n - 1) seats is 1 - p,,.). Therefore, the probability, fin, of at least one coupleoccupying(A)adjacent chairs when the chairs are arranged in a circle is given byPn = pn + (1 - fiom which it follows that fin also approachesBYRNE, MATCHING PROBLEM 3741 - 1/e as n - m.It is interesting to note that the congruence of both pn and 3 to 1 - 1Ie*is related to that of -( ' , which itself is the answer to a weii knownk=1 k!probability matching problem: if n married couples are randomly paired up for adance, what is the probability that at least one pair is a married couple? Thisproblem is sometimes cast in terms of matching hats to men, letters to envelopes,or positions in two decks of cards (see for example [I], [2] p. <strong>10</strong>7, and [4]). We5 à converges to e, we can choose Nl so thatnow present the proof that lim pn = 1 - lle. Let e > 0 be given. SincenÑk=O k!1- e 1=1 e (_1)k+lk=0 k kq+l k! 3 k=, k!2 - < - for all n t N). Since -converges to 1 - lle, we can choose N2 so that1~(- 1lk+l- for all n t N2. Let N3 = max(Nl, N2).T - ( l - : ) lSince lim = 1, for each k = 1, 2 ,..., N3, we can chooser so thatn--1-a^ c e#e for all n s n,;. Let N4 = max(n,, %,..., s). <strong>No</strong>w supposen s max(N3, N4). We haveThe first term on the right side of (6) issince 1 ~
375 PI MU EPSILON JOURNALA Generalization of a Dimension Formulaand an "Unnatural" IsomorphismDaniel L. Vim(the first inequality holds because n s N4). The last term on the right side of (6)is less than € sinceN3 2 N2. Therefore we have shown that for allReferences1. B.R. Grain, "On the Matching Problem in Probability," <strong>Pi</strong> <strong>Mu</strong> <strong>Epsilon</strong>Journal, 9 (Fall 1992), pp. 448-450.2. W. Feller, An Introduction to Probability Theory and Its Applications,<strong>Vol</strong>ume 1, 3rd Edition, Wiley, 1968.3. R. Honsberger, ed., Mathematical Plums, The Mathematical Associationof America, 1979.4. E. A. Tams, 'The Answer is 1-1/e. What is the Question?," <strong>Pi</strong> <strong>Mu</strong> <strong>Epsilon</strong>Journal, 8 (Spring 1987), pp. 387-389.Let V be a finite dimensional vector space with subspaces V, and V,.Then it is well known thatdim(V, + V,) = dim V1 + dim V2 - dim(V, n V,)where V1 + V, = {x + y: x e V and y e V, 1. Anyone who has studied combinatoneswill immediately recognize that this is similar to the principle ofinclusion/exclusion for two sets (PIE from here on). That is, if Sl and S2 are twofinite sets and lSil denotes the number of elements in S,, thenIS1 u S2l= lS,l + IS21 - IS, s21 -<strong>No</strong>w the PEE generalizes to n sets. Hence a natural question to ask is whether wecan find a similar formula for n subspaces. In this paper we present one suchpossibility and what it translates to in terms of quotient spaces.First of all, the obvious first guess would be to write down the generalformula for the PIE and replace Si by Vi and u by +. Unfortunately, this doesn'twork. For example, the formula for three subspaces would be3dim(Vl+V2+V3)= xdimVi-dim(VlnV2)-d.im(VlnV3)-dun(V2nV3)i=l+ dim(Vl n V2 n V3)which isn't true in general. For Example, in R 2 let V, be the x-axis, V-, be they-axis, and V3 be the line y = x. The left hand side of the above formula is 2 whilethe right hand side is 3. You might ask yourself why the inductive proof of thePIE doesn't work for subspaces. The reason is that in the inductive step of thePIE you need to know that n distributes over u. Unfortunately, n does notdistribute over +. However, we can prove the following,Theorem 1. Let Vl, V2, - ,Vn be subspaces of a finite dimensional vector space.ThenadimVi-dim(VlnV,)-dim(V3n(Vl +V2))dim(Vl +-+v,)=Xi=lProof. We proceed by induction on n, the number of subspaces. If n = 1 then. thestatement is trivial and if n = 2 it is the well known dimension formula. supposethat the formula is true up to n - 1. Then viewing Vl + - + V,., as a single
- Page 2 and 3: - Another Matching Problem with aMa
- Page 4 and 5: Red Light, Green Light:A Model of T
- Page 6 and 7: BENNINK. RED LIGHT, GREEN LIGHT 357
- Page 8 and 9: BENNINK, RED LIGHT, GREEN LIGHT 361
- Page 10 and 11: 365 PI MU EPSILON JOURNALROUSSEAU,
- Page 12 and 13: 369 PI MU EPSILON JOURNALwe havec(n
- Page 16 and 17: VIAR, DIMENSION FORMULA 377subspace
- Page 18 and 19: TOMFORDE, SELF-SIMILARITY 38 1m, n
- Page 20 and 21: TOMFORDE, SELF-SIMILARITY 385Note t
- Page 22 and 23: TOMFORDE, SELF-SIMILARITY 3 896. G.
- Page 24 and 25: 393 PI MU EPSILON JOURNALA Particul
- Page 26 and 27: 397 PI MU EPSILON JOURNALy = ax - b
- Page 28 and 29: 40 1 PI MU EPSILON JOURNALFigure 8P
- Page 30 and 31: The Power Means Theorem via the Wei
- Page 32 and 33: MISCELLANY 409The 1995 GameIn die F
- Page 34 and 35: 0. Co-author of convexity theoremP.
- Page 36 and 37: Pi MU EPSILON JOURNALii) JK is para
- Page 38 and 39: PROBLEMS AND SOLUTIONS 42 1Henry S.
- Page 40 and 41: PROBLEMS AND SOLUTIONS 425865. [Fal
- Page 42 and 43: PROBLEMS AND SOLUTIONS 429868. [Fal
- Page 44 and 45: BENNINK, RED LIGHT, GREEN LIGHT 3 6
- Page 46 and 47: PROBLEMS AND SOLUTIONS 437Solution
- Page 48 and 49: Editor's NoteThe Pi Mu Epsilon Jour