The Second Strong Law of Small Numbers Author(s): Richard K. Guy Reviewed work(s): Source: Mathematics Magazine, Vol.

63, No. 1 (Feb., 1990), pp. 3-20 Published by: Mathematical Association of America Stable URL: http://www.jstor.org/stable/2691503 . Accessed: 13/04/2012 15:51
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to Mathematics Magazine.

http://www.jstor.org

A RTIC LE S
The Second StrongLaw of Small Numbers
RICHARD K. GUY
The University Calgary of Alberta, Canada T2N 1N4

You have probablyalreadymet The Strong Law of Small Numbers, eitherformally [15, 21, 22] Therearen'tenoughsmall numbers meetthemany to demands madeofthem or in some frustrated semiconscious and formulation that occurredto you in the of mathematical rough-and-tumble everyday enquiry.It is the constantenemy of at mathematical sensiblestatement withspuridiscovery: once the Scylla,shattering ous exceptions, theCharybdis capricious and of coincidences, causingcarelessconjectures: the dilemmato search for proofor for counterexample. fooled Fermat It (Example 1 of [21]) and we'll meet Euler's memorable exampleat the end of the article. It's timeto introduce The Second Strong Law of SmallNumbers: Whentwonumbers lookequal, it ain't necessarily so! "How can thispossibly be?" I hearyouask.By way ofanswerI invite youto examine Example 36 Evaluate the polynomial - 6n3 + 24n2 - 18n + 24)/24 for n= (n4 1,2,3, ... more. Examples 1 to 35 are in [21]; therefollow forty-four In each, you are invited is the of and to guesswhatpattern numbers emerging, to decide whether pattern will are but theorems mingled are persist. Many oftheexamples fraudulent, somegenuine or in, to keep you on yourtoes,and theremayeven be an unsolved problem two. Pascal's triangle. Examples37 to 40 involve Example37 Pascal's triangle (modulo2) has been a perennial topic.But have you triedreading the rowsas binarynumbers? 3, 5, 15, 17, 51, 85, 255, 257, 771, 1285,3855, 4369, 1, 13107, 21845, 65535, 65537,... . Remember thatthereare zerosoutsidethe triangle as well, so you can also include theirdoubles, 2, 6, 10, 30, 34, 102,..., their quadruples,4, 12, 20, 60, 68,..., and so on, as well,if you like. Do you recognize thesenumbers?

MATHEMATICS MAGAZINE

101 1 10

0 10 10 o 101 o 01

1 0 0 0 0
I 1 1 0 0 0 0 01

0 0 1 1 1 01 1 01

00 o O o 1 0 1 1 01 O 1 10 00 1i 0o 0 O 0 1 0 10 0 1 10o0 0 0 11 0 0 0 1 o 01 0 011 1 01

11

01

1f

1 10000000000001 o o o o oO 0 oOoo 0 00 0 0 00 0

0 001

1 1
o

o1

1 01

1 0 001 1 0 0 0 0 00 0 0 0 0 0 1000 0 10 I 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 I 10 1 iIiII o o o o 0o I I I I I I I o 1 o 0 1011 o o o o 11oo o 111111 1 0 1 1 00000 0000001 1 0 10 0 0 0 0 1 o 00000 O 1 oI 1 o o o o o o 01I 0OO 1 0 I 1 0 O oo oQO oo 0 OOo 0 o f 01 01 0 0 0000oo o c; o1111 1 o C) 11 0 01 10000 t1o o o 71111
I 1 1 o o o o o o I o I o I1 1 o o co o o Io 1 f 0 I c0 o o o o o t o o o I o o o

1 t

001 1l 1
I1 0 O il

10 0
1 0 1

1
0 1

1 0
I1 o

1
I o

0 01
o i 1

1 0
0 1

0
0

1 0 01
i 0 1 0 f

1
1 0

0
1

1
0

1
1

11

1 111

11

111111

11t111111

11

11 1

off with each row starting two Example 38 Here we've drawn Pascal's triangle i.e. start, with (") in row n and column2n + r. of places to the right the previous
0 0 1

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
1 1 1

23

24

2 3 4 5 6 7 8 9 10 11

1 2 1 1 3 3 1 1 4

6 1

4 5

1 10 10 5 1 1 6 15 20 15 6 1 1 7 21 35 35 21 7 1 1 1 8 28 56 70 56 28 8 1 9 36 84 126 126 84 1 10 45 120 210 11 55 1

a and in by We've printed entry bold ifit's divisible itsrownumber, we've printed an in bold justif all theentries thecolumn bold.Whatare thesebold in are columnhead columnheads? to again, but in contrast the previous Example 39 We've drawnPascal's triangle a in bold justifit's not squarefree, just ifit contains i.e., example,we've put an entry 1. than greater squarefactor

VOL.

63, NO.

1, FEBRUARY

1990

5 1 1 1 1 2 1 1 3 3 1 1 4 6 4 1 1 5 10 10 5 1 1 6 15 20 15 6 1 1 7 21 35 35 21 7 1

0 1 2 3 4 5 6 7 8 9 10 11 12

1 8 28 56 17 56 28 8 1 1 9 36 84 126 126 84 36 9 1 1 10 45 120 210 252 210 120 45 10 1 1 11 55 165 330 462 462 330 165 55 11 1 1 12 66 220 495 792 924 792 495 220 66 12 1

What are the row numbers (printed bold) of the rowswhichcontainat least one in bold entry: 4,6,8,9,10,12,...? Example 40 We finally drawPascal's triangle with(n) column: 0 1 2 3 4 5 6 7 9 10 11 8 0 1 11 1 2 1 2 1 3 1 3 3 1 1 4 6 4 1 4 1 1 5 10 10 5 5 6 1 6 15 20 15 6 1 7 21 35 35 7 1 8 28 56 8 1 9 36 9 10 1 10 11 1 12 in the nthrowand (n + r)th 12 13 14 15 16

1 1 21 7 70 56 28 1 8 84 126 126 84 36 45 120 210 252 210 11 55 165 330 462 1 12 66 220 495

The column totals are 1,1,2,3,5,8,13,21,34,55,89,1445.... These numbersalso as seem to appear in the nextthreeexamples, well as in Examples70 and 80. not Example 41 The ceilingof,leastinteger less than,e (n-1)/2, for n= is 0 1 1 1 2 2 3 3 4 5 5 8 6 13 7 21 8 34 9... 55....

of Example 42 If a(n) is the sum of the divisors n, then a(n)/n measuresthe n abundancyof n. Everynumber withabundancy> j musthave at least k factors, where,for 2 k= 2 3 3 4 4 5 6 6 9 7 14 8 22 9... 35....

The differences thislast sequenceare 1,1,2,3,5,8,13,.... of

MATHEMATICS MAGAZINE

Example 43 You mayhave suspectedthatsome of the sequencesin the last three of examples are manifestations the ubiquitousFibonacci numbers(uo = 0, u1 = 1, to 2 = Un+ I + un). According theLucas-Lehmer theory therankof apparition [33] Un + (the least n forwhich p divides un) of a primep in the Fibonacci sequence is a 0 divisorof p - (p15), where(pH5)is theLegendresymbol, forp = 5, and + 1 or - 1 For accordingas p + 1 or +2, mod5, otherwise. example,the rankof apparition forthe first primes few is p= 2 3 5 7 11 13 17 19 23 29 31 37 3 4 5 8 10 7 9 18 24 14 30 19 41... 20....

When a primedoes first appear,does it alwaysoccurto the first power? suchthat Example 44 Definea sequenceby cl = 1, c2= 2 and c, + I theleastinteger - Cn-I differs all earlier 1 < i < j < n, e.g. differences from positive cn+I Cj-Ci, {c1, c2} = {1,2}

1 difference
1,2 differences

C3- C1=

c3 = c4 =

3 5

{ C1,C2,C3} =
{C1
...,

{1,2,3}

C4- C2 = 3

c4} = {1,2,3,5}

differences 1,2,3,4
differences 1,2,...,7

C5- C3= 5
c6 - c4 =

C5=8
c6= 13

{c1,.. ., c5} = {1,2,3,5,8}

Example 45 In thefollowing arrangements pennies, of each rowforms contiguous a rowtouches block,and each pennyabove thebottom twopenniesin therowbelowit. Count sucharrangements thetotalnumber pennies: by of

0 00 000&
1 1 2

0000&0
5

000

(L%0 OC%0 0(?)

8

cH
oa

YOCC x o oaoo co9o oocx

in Example 46 Alternatively, could countthe arrangements the previous you examof in ple by thenumber pennies thebottom row.

0
1

00

ooo

13

VOL. 63, NO. 1, FEBRUARY 1990

Example 47 The numberof rooted trees with n vertices, just one of which is labelled.

tLL
L 1

L~~~~~~~
L L

~~~~~~25

L
L 13

LL

_ 0t_ 1

of withn + 1 vertices. Example48 The number disconnected graphs

*-O0

@@*S

*AL 11

6 0/

*A

S,_TKA,
13

A%x'\

of on withjustone cycle. Example49 The number connected graphs n + 2 vertices

*13

For many otherexamplesinvolving graphs,see [22], which does not, however, includeExamples47-49. in Example 50 The coefficients the powerseriessolution x2 x3 2X4 5 y= 1 + x2 + ! + i + 5 !x of the differential equationD2y = exy.

13x6
6!

MATHEMATICS MAGAZINE
= an1

Example 51 The sequence an

nan2

(n
2- =

1) witha1 =ao= 1

1/2

al

1 2 +1

a2= 1 +2x a3=2+3X a4=5+4X2=

2=2 1=5 13.

Example 52

The sequence bn= (n

1)2n-2 + 1, n > 1.

bl=0x2-'+
b3=2x2'+

I =1
1=5

b2= 1 x20+ 1 =2

b4=3 x22+ 1 = 13.

ordinal Example 53 How manydistinct sums, f(n), may therebe of n different numbers?Obviously,f(l) = 1. However,f(2) = 2, because ordinaladditionis not 1 commutative. example, + X = X 0 X + 1. You might For guessthatf(3) couldbe as sums of 3 different large as 3! = 6, but in factyou can't have morethan5 distinct ordinals. The answers forn= are f(n) = 1 1 2 2 3 5 4 13 5 33 6 81 7 193 8... 449...

perhapsthe same sequenceas Example52. Or perhapsnot. Example 54 The valuesof the polynomial - 231n + 1523 forn = 0,1,2, ... are 9n2 1523, 1301, 1097, 911, 743, 593, 461, 347, 251, 173, 113, 71, 47, 41, 53, 83, 131, 47n2- 1701n+ 10181. 197,.... Try also the polynomial in Example 55 What are the nextthreeterms the sequence (1),2,3,5,7, 11, 13, 17, 19,23,29,31,37,41,43,47,53,...?

Example 56 The integer partof the nthpowerof 3/2 n (3/2)n 0 1 0 1 1.5 1 2 2.25 2 3 3.375 3 4 5.0625 5 5 7.59375 7 6 11.390625 11

VOL. 63, NO. 1, FEBRUARY 1990

at of Example 57 The number treeswithn edges,and height most2.

11

of of Example 58 The number partitions n n= p(n)= 0 1 1 1 2 2 3 3 4 5 5 7 6 11 7 15 8 22 9... 30....

function: of differences thepartition successive Example59 Ifwe form 1 1 2 3 5 7 11 15 22 30 42 56 77 101 135 176 231 297 385 490 627 0 1 1 2 2 4 4 7 8 12 14 21 24 34 41 55 66 88 105 137 ... 1 0 1 0 2 0 3 1 4 3 7 3 10 7 14 11 22 17 32 ... 7 -3 11 -5 15 ... -1 1 -1 2 -2 3 -2 3 -2 5 -4 7 -3 in alternate sign. differences we see thatthe third-order Example 60 If you expand the product(1 get,successively 1-x
1
-

x)(1

- x2)(1

x3)(1

- X4)*,

you

x -x2

+ X3 _x6

1--x2?+X4+X5

1-x-x2

+ 2x5- x8 -x9 + x10

3 2 and a coefficient has appeared. Indeed, at stage 10, a coefficient appears. calculation appearsto cancel theseout,leaving However,further
1-X-X2 +X5+x7-xl2-xl5+

result? otherthan0, ? 1 in the final Are thereany coefficients

10

MATHEMATICS MAGAZINE

m Example 61 For each integer exponent, is therean integer > 1 such thatthe n, sum of the decimaldigitsof m' is equal to m? 21, 92 = 81, 83 = 512, 74 = 2401, 285= 17210368,186 = 34012224,187 = 61222032, 468, 549, 8210,981",10812, 2013, 9114, 10715, 13316 8017, 17218, 8019,9020, 9021. Example 62 A Nivennumber been defined one whichis divisible the sum has as by of its decimaldigits, suchas 21 and 133. Is n! alwaysa Nivennumber? 4! = 24, 5! = 120, 6! = 720, 7! = 5040, 8! = 40320,9! = 362880, 10! = 3628800. Example 63 Can you choosea sequenceof real numbers from interval 1) so the (0, thatthe first lie in different two the halves, first threein different thirds, first the four in different quarters, so on? For example, and 0.71, 0.09, 0.42, 0.85, 0.27, 0.54, 0.925, 0.17, 0.62, 0.355, 0.78, 0.03, 0.48,.... If you runintodifficulty, are allowedto adjustearlier you members thesequence,if of you like. Example 64 Surely everyodd number (greater than1, ifyou don'twantto count1 as a prime)is expressible a primeplus twicea square? as 3 + 2 02 3 + 2 12,5 + 2 12,7 + 2 12,3 + 2 22, 11 + 2 12, 7 + 2 22, 17 + 2 02, 11 + 2 22, 3 + 2 32, 5 + 2 32, 23 + 2- 12... Indeed, some numbers, suchas 61, have severalsuchrepresentations. Example 65 Is n! alwaysexpressible thedifference two powersof2? as of 0! = 1! = 2' - 2? 2! = 22 2', 3! = 2 - 2', 4! = 2 - 23, 5! = 2 - 23

Example 66 It's wellknown that4! = 52 - 1, 5! = 112-1 and 7! = 712-1, but not so well knownthatifyoutakethe nextsquarebigger thann! thedifference always is a square:
6! = 272 - 32, 8! = 2012 10! = 1905292,

9! = 6032 -272,
182I...

152, 11! = 63182-

Example 67 The valuesof sin2(k7/12), k = 0, 1,... 6 are for k=

sin2 (k7/12) =

6 1

0 (2-F3)/4

1/4 1/2 3/4 (2 + 3)/4

It's also well knownthat

J/ sin2nxdx= (2n- 1)(2n-23) ..3

VOL. 63, NO. I, FEBRUARY 1990

11

rule,using6 equal subintervals, you by If you calculate the integral the trapezoidal will get the answer
~22?-1? (2? )? (2
3)n

+ 3n + 2n +

1}7/12

4n,

whichis exact forn= 1, 2, 3, 4,5,6 and 7. fraction q 2leYis for Example 68 The continued
7T2 1 1 1 1 1 1 1

5? 1+

1? 5?

T 1? 1? 5?

1? 1?
=

are 32 P(P(n - 1)) + P(n - P(n - 1)). The first terms 1, 1, 2,2,3,4,4, 4, 5, 6, 7, 7, 8, 8, 8, 8, 9, 10, 11, 12, 12, 13, 14, 14, 15, 15, 15, 16, 16, 16, 16, 16. Note thatP(2) = 1,
P(4) = 2, P(8) = 4, P(16) = 8, and P(32) = 16.

Example

69

Define a sequence by P(1) = P(2) = 1, and for n > 2, P(n)

withQ(1) = Q(2) = Q(3) = 1, and the same sequence starts Example 70 A similar
34 recurrence for n > 3, Q(n) = Q(Q(n - 1)) + Q(n - Q(n - 1)). The first terms are 13, 13, 13, 13. Notice that Q(2) = 1, Q(3) = 1, Q(5) = 2, Q(8) = 3, Q(13) = 5, Q(21) = 8 and Q(34) = 13.

1, 1, 1, 2, 2, 3, 3, 3, 4, 5, 5, 5, 5, 6, 7, 7, 8, 8, 8, 8, 8, 9, 10, 11, 11, 12, 12, 12, 13, 13,

of Examples40-52 and 70 perhapscontainmanifestations the Fibonaccinumbers. (2n)!/n!(n + 1)!, are Almostas ubiquitous the Catalan numbers, 1, 1,2, 5, 14, 42, 132,429, 1430,48625.... How manyof Examples71 to 79 are genuine? and n rangesyou can drawwithn upstrokes of Example 71 The number mountain downstrokes:

~~~~~~~~~~~~~~~~2

--

AA

14

12

MATHEMATICS MAGAZINE

Example 72 The numberof ways of makingn foldsin a stripof n + 1 postage stamps,wherewe don't distinguish betweenfront and back, top and bottom, left or and right:

0
1

01 1

012 021 2

0123 0132 0231 0321 1032 5

01234

01243 01432 02431 03214 03421 04123 04321 10243 10342 1043214032 01342 02341 14

Example 73 The number different of groups, to isomorphism, order2n is, up of forn = = no. ofgroups 0 1 1 1 2 2 3 5 4... 14....

Example 74 The number ways 2n people at a roundtable can shakehands in of their pairswithout handscrossing.

14

of withn + 4 points. of Example 75 The number triangulations thesphere

We leave the readerto verify that thereare just 14 distinct of triangulations the spherewith8 points.

VOL. 63, NO. 1, FEBRUARY 1990

13

planetreeswithn edges."Plane" meansthatwe of Example 76 The number rooted and right. betweenleft distinguish

14

n objectsin indistinguishof Example 77 The number waysofdistributing different objectsin a box. able boxes,withat mostthree

LiS L
1 I 1 l

2~~~~~~~~11 ~~~~~~~~~2

~LJIkLLLL
D
a

14

f(k), withdomain for Example 78 The probability, n = 0, 1,2,. .., thata function
k = 0,1 ..., n and range [0, 1], is convex, is

1
(0!)2

1
(1!)2'

2
(2!)2

5
(31)2'

14
(4!)2 2**

Example

79

The incomplete Bessel function of order one has power series expansion
2 l(2x )/X 3 I + X

2!

2X2

4!

5X6

6!

14 x8

Examples 37 to 40 involved Pascal's triangle, whose entries are the Example 80 binomial coefficients. We can use a similar array to expand (1 + x + x2),, giving trinomial coefficients. Each entry is the sum of the three nearest in the previous row. 1 1 1 1 2 2 3 1 1 6 3 7 6 3 1 1 10 4 16 16 19 4 10 1 1 5 15 30 51 45 45 30 5 15 1 1 6 21 50 90 126 141 126 50 90 21 6 1 1 7 7 28 77 161 266 357 393 357 266 161 77 28 36 112 266 504 784 1016 1107 1016 784 504 266 112 36 8 1 156 414 882 1554 2304 2907 3139 2907 2304 1554 882 414 156 45 9 1

1 1 8 1 9 45

14

MATHEMATICS MAGAZINE

The centraltrinomial coefficient, an, 1,1,3,7, 19,51, 141,393,1107,3139,... almosttreblesin size at each step: ifwe calculate3an- an+1 we get 2,0,2,2,6,12,30,72,182,... whichare pronicnumbers, m(m + 1), form = 1,0,1,1,2,3,5,8,13,.... Answers
(n 36. This is the polynomial 1) + (n 1) + (n 1 + (n 1) + (n ) of Example5 of [21], and represents numberof pieces you can cut a circular the cake into by n slicingbetweenevery pairof pointschosenfrom aroundthe circumference. It of that4-dimensional is also the number regions space is choppedintoby n - 1 hyperplanes general in position. The sequenceis #427in [45]: 1, 2, 4, 8, 16, 31, 57, 99, 163, 256, 386, 562, 794, 1093, 1471,.... for 37. This verybeautiful setting Example 1 of [21] was observed20 yearsago by William Watkins, now co-editor Coll. Math. J. Gauss has told us that the of numberof sidesin a regular polygon whichcan be constructed withstraightedge and compassis ofshape2m17JFn, wherethe Fnare distinct Fermatprimes + 1. 22n Only fivesuch,0 < n < 4, are knownand somepeople believethatno others will everbe found. thepattern So breaks downat row32. Fermat thought 232+ 1 that was prime, but Eulerdiscovered factorization X 6700417. the 641 if 38. This is the Mann-Shanks test Can you primality [36]. Surprising, not practical. proveit? 39. This is an observation Gerry of thatthebold numbers thecomposite are Myerson: numbers. However, breaks this downin row 13, because(11)=3211. 13 and

223. 11* 13 arenotsquarefree. and Fibonacci numbersis 40. This well-known relationbetween Pascal's triangle in is two sinceeach entry thesumoftheentries theprevious easilyseen to persist, columnsof the previous totals. row,so each totalis the sum of the two previous 41. This is adapted from an inequalityof Larry Hoehn, of ClarksvilleTN. The coincidenceis quite surprising, since rei 1.64872 and the golden ratio close. For n = 10,11,12,... the terms (1 + V5)/2 = 1.61803are not remarkably 91, 149,245,... beginto diverge the from Fibonaccisequence89,144,233,.... 42. In [32], RichardLaatsch showsthatthe sequence continues 55,89,142,230,.... withdifferences 20 30 53 88 143 236 387 641 1061 17632737 4903 8202 1375023095... whichstayclose to theFibonaccinumbers 21 34 55 89 144 233 377 610 987 15972584 4181 6765 10946 17711... morerapidly. forawhile,but eventually tendto infinity 43. See sequence X912 in [45]. This is still a notorious open question:thereare of extensivetables [30, 34, 49, 50]. During revision this article,Dick Lehmer with no on ran kindly a program a 75 Vax,and found counterexample p less than a million. 44. The sequence continues 17,26,34,45,54,67,... and is denserthantheFibonacci The sequence thereis misleading. sequence. It is #254in [45],but the reference E1910 [1966, 775; partialsolution doesn't solve Amer.Math. Monthly problem

(l~)=

VOL. 63, NO. 1, FEBRUARY 1990

15

1968, 80-81] because the differences not unique: e.g., 17 - 8 = 26 - 17 = are } 54 - 45. Nor is it theauxiliary sequence{ r,2 = {4, 5, 9, 10, 11, 16, 18,22, 23, 24, 25, 27, 28, 29,... }, used to construct Sierpinski the sequence, #425 in [45]. There's another open question here:find smallest the possibleasymptotic growth such thateverypositive fora sequence of integers integer occurs uniquelyas a difference. 45. This also failsto continue withthe Fibonaccisequence.The numbers arrangeof are ments with7,8,9,... pennies 12,18,26,.... These arrangements studied were by Auluck[2]; see sequence#253in [45],and compareExample34 in [21]. 46. These are indeed the odd-ranking Fibonaccinumbers, -1' sequence #569 in u2 [45], whichhave theproperty
U2n-1 = U2n-3+ 2U2n-5 + 3U2n-7 + + (n

ul +1

47. 48. 49. 50. 51. 52. 53.

which can be seen to be the number ways thata row of n penniesmay be of with n - k in its bottomrow,in any one of k surmounted an arrangement by wherek = 1,2,. .., n - 1 or it's not surmounted all (k = n). possiblepositions, at These are not the alternate Fibonaccinumbers, e.g., the numbers such trees of with 5,6,7,... vertices 35,95,262,... See sequence #570 in[45] or p. 134 are in [43]. Nor are these.The nextfewmembers the sequenceare 44, 191, 1229, 13588, of 288597,.... See sequence#574in [45],or [24]. Neitheris this the sequence of alternate Fibonacci numbers, but continues 33 240 89 (one short!), (correct!), (7 too many),657, 1806, 5026,.... See sequence #568in [45] or page 150 in [43]. Nor is this, which continues36,109,359,1266,4731,18657,77464,...; see sequence #572in [45] and Tauber'spaper [48]. Nor again, since a5 = 13 + 5 x 5 = 38, a6 = 116, a7 = 382,...; see sequence #573in [45]. Neither these,b5 = 4 23 + 1 = 33, b6= 5 24 + 1 = 81, b7 = 6 25 + 1 = 193, are but theydo feature(for a Fibonacci numbers, b8 = 7 26 + 1 = 449, alternate while)in the nextExample: whichI got from John Conway.If g(k) = k *2k-I + 1, then f(n)=
O<k<n

max f(n-k)g(k),

the and, for n < 8, f(n) is indeed equal to g(n - 1). Thereafter situation gets for morecomplicated, a simpleruleeventually but emerges: n = 9,10,11,12,13, f(n) = 332,33 81,812,81. 193,1932,and, forn > 14, f(n) = 81ff(n 5), except that f(19) = 1933. 54. Severalreadersof [21] said thatI shouldhave includedEuler's famousformula, for thatExample21 was n2+ n + 41, whichgivesprimes 0 < n < 39, notnoticing 41 the just that,except forthe disguiseof omitting tell-tale (n = 0). For some see astonishing examplesof The StrongLaw in thisconnection, the papers of is of Stark[46, 47]. The present polynomial a slight adaptation one due to Sidney n and is found replacing in Euler'sformula 38 - 3n. Surprisingly, Kravitz, by by thisstillgivesprimes 0 < n < 39, although for thirteen themare notamongthe of n original forty; = 40 and 41 give6683 = 41 X 163 and 7181 = 43 X 167. The polynomial 47n2- 1701n+ 10181 was discoveredrecently Gilbert by Fung. If you workmodulo p forprimes2 < p < 43, you'llfindthatit's never divisibleby such primes.It takes primevalues for0 < n < 42, beatingEuler's of recordby two. Noticethatthe discriminant Euler'spolynomial - 163, and is

16

MATHEMATICS MAGAZINE

55.

56. 57. 58.

59.

60.

that of Kravitzis - 32 X 163, while Fung's polynomial to have a positive has discriminant, 979373. Such questionsare hardly fair, since arguments be advancedforcontinuing can sequences in any way you wish. Some answersare moreplausiblethan others, and theone thatPersiDiaconishoped you wouldmissis 59,60,61,..., however, the ordersof thesimplegroups! Another futile attempt foolyouintothinking the primes. to of The nextmember is 17, then25,38,...; see sequence #245in [45]. This is not the same sequenceas the previous but example, see thenext! To see the correspondence betweenthisand the previous example, notethatthe number of'vertices height is the number parts, at one of and thevalencesofthese vertices thesizesoftheparts.Sequence#244in [45]; see also page 122 in [43] are and page 836 in [1]. This examplewas sentby Gerry It Myerson. can be provedthatthedifferences of anyorderare positive from somepointon,butthatpointrecedesrather rapidly as you take higherorderdifferences. nextfew thirddifferences - 4, 17, The are -2,24, -4, 32, 1,38,5,... and are positive from now on. The fourth differences alternate signuntilthe67th,after in whichtheyare positive. This is Euler's famous pentagonal numbers theorem:
n=1

J7(1- -

00

00

n)

k=-oo

(_ I)kXk(3k-l)/2

See theorem 353 in [26],forexample. 61. NormanMegill of Waltham, MA, finds such m foreach n < 104. For n = 105, no however, such m exists. 62. This questionwas askedby Sam Yates.Carl Pomerance suggested thatcounterexamples mightbe expectedby the time n has reached500, and indeed Yates foundthat432! is nota Nivennumber, sincethesumofitsdigits 32 X 433, and is 433 is prime. 63. The given sequence can be continued, 0.97,0.22,0.66,0.32,but Berlekamp and Graham[3] have shownthatno suchsequenceexists withmorethan17 members! 64. This special case of the Hardy-Littlewood problemwas mentionedby Ron of Ruemmler Edison,NJ,who believesthatthefirst exception 5777, and asksif is it is also the last! It is knownfromthe workof Hooley [27], Miech [37], and Polyakov[42] thatthe density exceptions zero. of is 65. Ignace Kolodner thisfrom got HaroldN. Shapiroin an NYU Problem Seminar in 1949. It's leftto the readerto provethat n! is neveragain the difference two of powersof two. 66. This was observedby Larry Hoehn of Clarksville, TN. It fails for 12! but 13! = 789122 - 2882, 14! = 2952602- 4202, 15! = 11435362- 4642, 16! = 45741442- 18562. It's doubtful thisoftenoccursfrom if here on (note thatyou musttake the nextsquarebiggerthann!), but it maybe hardto proveanything. 67. This is also correct n = 8, 9, 10,and 11,butforn = 12 we get(1352079X)/224 for insteadof (1352078v)/224,out by 3 partsin fourmilion! The trapezoidal rule givesthe right answerifyou use k subintervals, provided is less than4k: see 2n [28],forexample.David Bloomsuggested that"fourmillion" shotldread"sixteen million":I intendedthe relativeerror, 2.958/4000000: the actual erroris 2.996/16000000:moreexamples the Strong of Law! 68. If this pattern, noticedby JamesConlan [8], were to continue, would have we (5 + 37 )e- = 27T2.Close,but no cigar!
-

VOL. 63, NO. 1, FEBRUARY 1990

17

69. The sequence thathit the national presseson bothsides of the Atlantic, [6], e.g. encounter.I have an earliermanuscript publicizingthe Conway-Mallows of Conway in which he has written anothernotation)"P(2k) = 2k- 1 (easy), (in of P(2n) < 2P(n) (hard),P(n)/n -2 ' (harder)."It was theproof a preciseform of thislast statement thatalmostwon Mallowseven moremoneythanConway this intended.Papersmentioning sequenceinclude[16, 35]. 70. Yes, the Fibonaccipattern continues [40]. David Newmanshowedthisto David Bloom as a conjecture 1986. in Nine of the final examples intended look like the Catalan numbers; ten are to sequence #577 in [45]. At first is a matter some surprise it of that = I 12n)

is always an integer. connection In withsome recentcorrespondence [41], John Conway makesthemoregeneral that observation
(m, n)(m + n - 1)!

m!n!

is an integer, where(m, n) is the g.c.d. of m and n, because in(m?n-l)! m!n! I (+n m - I1 and n n(mn+n-1)! m!n!
=)

n-I n- 1

71.

72. 73. 74. 75.

are bothintegers. This also answers question B33 of [20],whereNeil Sloane in a gave the examplen = 4m + 3. Catalan numbers occur in manywidelydifferent see [18], lookingcontexts: 500 references, [31],witha listof 31 structures, obtainable withnearly and both from H. W. Gould, Departmentof Mathematics, West VirginiaUniversity, Morgantown, WV, 26506. An articlewith a good bibliography [5]. Several is " proofs without the of words," showing equivalence severalofthe structures, will appear in [9]. This is a genuineexampleof the Catalan numbers. The mountain rangesare the same as paths from (0,0) to (n, n) whichdo not cross y = x, or incoming tied ballotsin whichone candidate neverbehind, sequencesof zerosand ones,or is or of + Is, subject to appropriate sum conditions, e.g., randomone-dimensional walksin whichyou nevergo to the leftof the origin; [13]. see This sequence,#576in [45], is not,and continues (not 38, as statedin [14]), 39 120, 358, 1176,3527, 11622,36627, 121622,389560,..., see [29]. of The numbers groupsof orders and 26 are 51 and 267 [23]. This sequence, 25 2328 [51], 56092 [52]. #581in [45], continues This is genuineCatalan again: see [39]. But thisone, sequence#580in [45],has been calculated onlyfour for moreterms [4, 12, 19]. Of the 1,1,2,5,14,50,233,1249,7595 triangulations,

only0,0,1,1,2,5,12,34,130 containno vertex valence3. of 76. is a genuiine manifestation the Catalan numbers 25], but of [7, 77. is not:sequence #579in [45] continues 166,652, 2780, 12644,61136,312676, 46, 1680592,... [38]. 78. The probability generaln is indeed cn/(n!)2 [10]. for 79. In [10] we asked whatwas the exponential generating function the Catalan for

18

MATHEMATICS MAGAZINE

numbers. Louis W. Shapiroobserves that

oo
X2n Cn (2n)!

=11(2x)/x

whereI is themodified Besselfunction order of 9.6.10. on page one: see formula for 375 of [1]. In the paper [44] he obtainsresults latticepathswhichstaybelow withincreasing analogivenpoints,arranged abscissasand ordinates, somewhat of gous to the convexfunctions [10]. Beforewe say goodbyeto the Catalan numbers, here's an observation which may not be widelyknown.It originated a discussion in withJohnConwayonly six monthsago. What is well knownis thatthe Catalan numbers associated are withparenthesization. thatmostpeople mean the numbers possibleorders By of of n nonassociative operations, usually indicated n - 1 pairsof parentheses: by a n=0 n = 3 ((ab)c)d n = 4 (((ab)c)d)e ((ab)c)(de) a((b(cd))e) n=l ab (a(bc))d ((a(bc))d)e (a(bc))(de) a((bc)(de)) n=2 a((bc)d) (a((bc)d))e (ab)((cd)e) a(b((cd)e)) (ab)c or a(bc) a(b(cd)) (ab)(cd) (a(b(cd)))e ((ab)(cd))e a(((bc)d)e) (ab)(c(de)) a(b(c(de)))

and so on. But they are also the numbersof ways of arrangingn pairs of as parentheses a pattern, fortheir just own sake:
n=O n( ((3)())) )0
n= ( )()())(( )() (())

n=1 (()))
( )() ( ))

()
(
()( )) ( )( )) (() ) ) ( )()( ))

n=2 00(0)
) ) ( (( (( )() ( ))

(()

or 0

()
( ))
))( )) )

(()

) ( )()

)( )

An examination the symmetries the two cases makesit unlikely of in thatyou'll finda directcombinatorial comparison. One-one correspondences between the former manifestation other and Catalanmanifestations wellknown. are The latter are easilyseen to be in correspondence withthe pairsof people shaking handsin Example 73, and withthemountains Example70. in 80. Jack Good [17] has given an asymptotic formulafor the central trinomial coefficient:
an -3

2~7Tn

2+ 1;+ (1-3:1 5122

0(n-3)

whichshowsthatthe leftside of the "identity"

e

3an-an+?1=un1(un?1)

for good to the nearest integer quite largevaluesof n. This example was sent by Donald Knuth.Euler [11] was one of the earlier discoverers The Strong of Law of SmallNumbers, and called this memwrabile exemplum inductionis fallacis. On the same page he givestheFibonacciformula that'soften attributed Binet. to

3a9 - a10 = 464, while u8(u8 + 1) = 21 x 22 = 462. The asymptotic formula is

growslike c x 3nX n-3/2, whereas right the side grows like 'r2n/5,whereT is the golden ratio, T 2= (3 + F5 )/2. Further calculation shows that a10= 8953,

VOL. 63, NO. 1, FEBRUARY 1990

19

Coda I showed this example to George Andrewsduring the recent Bateman Retirement Conference Allerton at Park,Illinois. Half-an-hour he came back with later what Euler reallyshouldhave said. He defines trinomial the coefficients centrally by
(1x?+x2) E ( n

and provesthat,if Fn is the nthFibonaccinumber, then

Fn + 1)=OX n

(1

1 )2

OA + 2 )2)

For - 1 < n < 7, the onlynonzerotermon the rightis X = 0, which accountsfor Euler's observation, since

3(0)2_ n (2

)=2(

0)-2(

) 1

will publishthe q-analog thistheorem Andrews of shortly.

R E F E R E N C ES Nat. Bureau StanFunctions, 1. MiltonAbramowitz and Irene A. Stegun,Handbookof Mathematical dards,Washington, D.C., 1964; Dover,New York,1965. [Exx.58, 79] 2. F. C. Auluck,On some new typesof partitions associatedwith generalizedFerrersgraphs,Proc. Cambridge Philos. Soc. 47 (1951), 679-686. [Exx.45, 46] in of sequences,J. Number 3. E. R. Berlekamp and R. L. Graham, Irregularities the distributions finite Theory (1970), 152-161. [Ex. 63] 2 21 4. RobertBowen and Stephen Fisk,Generation triangulations thesphere,Math. Comput. (1967), of of 250-252. [Ex. 75] Amer.Math. Monthly (1965), 72 5. W. G. Brown,Historical noteon a recurrent combinatorial problem, 973-977. [Exx. 71, 74, 761 The New YorkTimes88-08-30, 6. MalcolmW. Browne, swift response, Intellectual duel: brashchallenge, page B5. [Ex. 69] A 7. N. G. de Bruijnand B. J. N. Morselt, note on plane trees,J. Combin.Theory2 (1967), 27-34. [Ex. 76] Coll. Math. 1. 20 (1989), 50. [Ex. 68] 8. JamesConlan, A sad noteon inductive mathematics, W. Scientific AmericanLibrary, H. 9. JohnH. Conway and RichardK. Guy, The Book of Numbers, Freeman,New York,1989. [Exx.71, 74, 76] is K. to 10. RogerB. Eggletonand Richard Guy,Catalan strikes again! How likely a function be convex?, thisMAGAZINE 61 (1988), 211-219. [Ex. 78, 79] 11. L. Euler, Opera Omnia, Ser. 1, vol. 15, 50-69, esp. p. 54. [Ex. 80] 12. P. J. Federico, Enumeration polyhedra:the numberof 9-hedra,J. Combin. Theory7 (1969), of 155-161. [Ex. 75] vol. Wiley,New 13. W. Feller, An Introduction Probability to Theory and itsApplications, I, 2nd edition, York,1957, pp. 70-73. [Ex. 71] 14. MartinGardner, in mathematics, Sci. Mathematical games:permutations paradoxes combinatorial and Amer.209 #2 (Aug. 1963), 112-119, esp. p. 114; see also #3 (Sept.) p. 262. [Ex. 72] in law Sci. 15. , Mathematical games:patterns primes a clue to thestrong ofsmallnumbers, Amer. are 243 #6 (Dec. 1980), 18-28. 16. SolomonW. Golomb,Discretechaos:sequencessatisfying Amer.Math. Monthly recursions, "strange" 97 (1990). [Ex. 69] and random Philos. Soc. 54 (1958), 17. I. J.Good, Legendrepolynomials trinomial walks,Proc.Cambridge 39-42. [Ex. 80] of 18. Henry W. Gould, Bell and Catalan Numbers: ResearchBibliography Two Special NumberSe1985. [Exx.71, 74, 76, 78, 79] quences, Morgantown, WV, 6thedition, 19. BrankoGruinbaum, ConvexPolytopes, Interscience, 1967,p. 424. [Ex. 75] in Problems Number 20. RichardK. Guy, Unsolved 1981, p. 49. [Ex. 70] Theory, Springer,

20 21. 22. 54]

MATHEMATICS MAGAZINE 95 , The Strong Law of SmallNumbers, Amer.Math. Monthly (1988), 697-712. [Exx.36, 37, 45,

, Graphsand the Strong Law of SmallNumbers, Proc.6th Internat. Conf. Theory Appl. Graphs, Kalamazoo, 1988. 23. MarshallHall and J. K. Senior, The Groupsof Order (n < 6), Macmillan, 2n New York,1964. [Ex. 73] 24. FrankHarary, The number linear, of directed, rooted, and connected graphs, Trans.Amer.Math. Soc. 78 (1955), 445-463. [Ex. 48] 25. FrankHarary, The number plane trees,Indagationes of Math.26 (1964), GeertPrinsand W. T. Tutte, 319-329. [Ex. 76] 26. G. H. Hardyand E. M. Wright, Introduction theTheory Numbers, edition, An to of 4th Oxford, 1960. [Ex. 60] 27. Christopher Hooley,On the representation a number the sum of two squaresand a prime,Acta of as Math. 97 (1957), 189-210; MR 19, 532a. [Ex. 64] 28. Lee W. Johnson and R. Dean Riess,Numerical Analysis,2nd edition, Addison-Wesley, 1982, p. 343. [Ex. 67] 29. JohnE. Koehler,Foldinga strip stamps,J. Combin.Theoay (1968), 135-152. [Ex. 72] of 5 30. Maurice Kraitchik, Recherches la Theorie Nombres, 1, Gauthier-Villars, sur des vol. Paris,1924, p. 55. [Ex. 43] W. 31. Mike Kuchinski, Catalan Structures Correspondences, and M.Sc. thesis, Virginia Univ.,1977. [Exx. 71, 74, 76] 32. Richard Laatsch, Measuringthe abundancyof integers, this MAGAZINE 59 (1986), 84-92; MR 87e:11009. [Ex. 42] 33. D. H. Lehmer,An extended theory Lucas functions, of Ann. Math. 31 (1930), 419-448. [Ex. 43] 34. Douglas A. Lind, RobertA. Morris, and LeonardD. Shapiro,Tables of FibonacciEntry Points,part 20 two,editedBro. U. Alfred, FibonacciAssoc.,California, 1965; see Math. Comput. (1966), 618-619. [Ex. 43] 35. Colin L. Mallows,Conway'schallenge 97 sequence, Amer.Math. Monthly (1990). [Ex. 69] A 36. HenryB. Mann and Daniel Shanks, necessary and sufficient condition primality, its source, for and Ser. J. Comitbin. Thzeory A. 13 (1972), 131-134. [Ex. 38] 37. R. J.Miech,On theequationn = p + x2, Trans.Amer.Math. Soc. 130 (1968), 494-512; MR 42#1775. [Ex. 64] 38. F. L. Miksa,L. Moserand M. Wyman, Restricted of partitions finite sets,Canad. Math. Bull. 1 (1958), 87-96; MR 20#1636.[Ex. 77] 39. Th.S. Motzkin, cross Relations betweenhypersurface ratios, and a combinatorial formula partitions for for and Bull. Amer.Math. Soc. of a polygon, permanent preponderance, fornon-associative products, 54 (1948), 352-360; MR 9, 489. [Ex. 74] E 40. David Newman,Problem 3274, Amer.Math. Monthly (1988), 555. [Ex. 70] 95 41. News & Letters, thisMAGAZINE 61 (1988), 207, 269. [Ex. 70] 42. I. V. Polyakov, in and thesquareof an of of Representation numbers theform sumsofa primenumber integer (Russian) Izv. Akad. Nauk. UzSSR Ser. Fiz.-Mat.Nauk 1979,34-39, 100. [Ex. 64] 43. J. Riordan,An Introduction Combinatorial to Analysis,Wiley,New York,1958. [Ex. 47, 49, 58] in 44. Louis W. Shapiro, latticepathlemmaand an application enzyme A and kinetics, Statist.Planning J. 14 Inference (1986), 115-122. [Ex. 78, 79] 45. Neil J. A. Sloane, A Handbook Integer of Sequences,AcademicPress,New York,1973. [passim] 46. Harold M. Stark, explanation some exoticcontinued An in of fractions foundby Brillhart, Atkinand AcademicPress,1971,21-35. [Ex. 54] Birch(eds), Comptuters Number in Theory, 47. _ all of Proc. , Recent advancesin determining complexquadraticfields a givenclass-number, Symp. Pure Math. XX, Amer.Math.Soc., Providence, 1971,401-414. [Ex. 54] RI, 48. S. Tauber, On generalizations the exponential of Amer.Math. Monthly (1960), 763-767. 67 function, [Ex. 50] 49. Marvin Wunderlich, Tables of Fibonacci EntryPoints, edited Bro. U. Alfred,Fibonacci Assoc., California, 1965; see Math. Comput.20 (1966), 618-619. [Ex. 43] 1 in of 50. Dov Yarden,Tables of ranks apparition Fibonacci'ssequence(Hebrew), RiveonLematimatika (1946), 54; see Math. Tables and Aids Comput.2 (1947), 343-344. [Ex. 43] 51. Eugene Rodemich, The groupsof order128, 1. Algebra67 (1980), 129-142; MR 82k:20049. 52. E. A. O'Brien,The groups orderdividing of 256, Bull. Austral.Math. Soc. 39 (1989), 159-160.