PI in the SKY

ISSUE 13, FALL 2009

IN THIS ISSUE:
THE BOARD GAMING GENIUS: REINER KNIZIA ELEVATOR RIDES IN PURGATORY HEXAGONAL CODES THE ABORIGINAL GAME OF LAHAL

Pacific Institute for the Mathematical Sciences

On the Cover
The Egyptian game Senet, one of the oldest board games, dates back as early as 3500 BCE. This board, from the Brooklyn Museum, is inscribed with the name, Amunhotep III. John Bowman (University of Alberta) Tel: (780) 492-0532, E-mail: bowman@math.ualberta.ca Murray Bremner (University of Saskatchewan) Tel: (306) 966-6122, E-mail: bremner@math.usask.ca John Campbell (Archbishop MacDonald High, Edmonton) Tel: (780) 441-6000, E-mail: jcdotcalm@shaw.ca Florin Diacu (University of Victoria) Tel: (250) 721-6330 , E-mail: diacu@math.uvic.ca Sharon Friesen (Galileo Educational Network, Calgary) Tel: (403) 220-8942 , E-mail: sfriesen@ucalgary.ca Gordon Hamilton (Masters Academy and College, Calgary) Tel: (403) 242-7034, E-mail: gamesbygord@gmail.com Klaus Hoechsmann (University of British Columbia) Tel: (604) 822-3782, E-mail: hoek@pims.math.ca Dragos Hrimiuc (University of Alberta) Tel: (780) 492-3532, E-mail: dhrimiuc@ualberta.ca Michael Lamoureux (University of Calgary) Tel: (403) 220-8214, E-mail: mikel@math.ucalgary.ca Mark MacLean (University of British Columbia) Tel: (604) 822-5552, E-mail: maclean@math.ubc.ca Patrick Maidorn (University of Regina) Tel: (306) 585-4013, E-mail: Patrick.Maidorn@uregina.ca Anthony Quas (University of Victoria) Tel: (250) 721-7463, E-mail: aquas@uvic.ca Wendy Swonnell (Greater Victoria School District) Tel: (250) 477-9706, E-mail: wswonnell@shaw.ca

Pi in the Sky is aimed primarily at high school students and teachers, with the main goal of providing a cultural context/landscape for mathematics. It has a natural extension to junior high school students and undergraduates, and articles may also put curriculum topics in a different perspective.

Submission Information

For details on submitting articles for our next edition of Pi in the Sky, please see: http://www.pims.math.ca/resources/publications/pi-sky

Editorial: David Leeming, Managing Editor, Pi in the Sky

Editorial Board

We welcome our readers to another issue of Pi in the Sky. In addition to being on-line at www.pims.math.ca/pi , we have about 1200 subscribers in 56 countries. Once again, we were faced with the challenge of trying to publish articles that can be understood by capable high school and College-level students. In this context, we are constrained by the articles we receive from authors. This issue contains articles on mathematical games, geometry, face recognition, the Laws of Physics and elevator rides (in purgatory!). We also have a book review written by high school students. In this issue, we announce the winners of the Math Challenge posed in Issue #11 (Spring 2008). One of our Editors, Volker Runde, has resigned. Volker has served the magazine well for a number of years and we thank him for his dedicated service to Pi in the Sky. We welcome Murray Bremner, University of Saskatchewan, to the Editorial Board. This is my last issue as Managing Editor of Pi in the Sky. I have thoroughly enjoyed my five years in the position. I would like to acknowledge the support of my Editorial Board for reviewing articles submitted and for their efforts in proofreading drafts of pending issues. I leave the magazine in the capable hands of Anthony Quas, University of Victoria, who will become Managing Editor of Pi in the Sky starting with Issue #14.

Table of Contents
The Aboriginal Game of Lahal by Peter J. Dukes and David J. Leeming ...................................... 1 Brad Pitt gets Angelina Jolie and I get ... Math? by P H. Trinh ............................................................................... 4 . Hexagonal Codes by Anthony Quas and Peter J. Dukes .......................................... 7 Snell on the Beach by Klaus Hoechsmann ............................................................... 13 Genius Board Game Inventor: Reiner Knizia by Gordon Hamilton ................................................................. 15 Higher Derivative Equations and the Laws of Physics by Neil Barnaby ......................................................................... 18 Heaven to Hell: Elevator Rides in Purgatory By Gordon Hamilton ................................................................. 22

Editorial Coordinator

Adam Wojtowicz (604) 822-0402, E-mail: media@pims.math.ca

Contact Information
PIMS University of Victoria Site Office, SSM Building Room A418b PO Box 3060 STN CSC, 3800 Finnerty Road, Victoria, BC, V8W 3R4 T/(250) 472-4271 F/(250) 721-8958, E-mail: pims@math.uvic.ca

Pi in the Sky

Significant funding for Pi in the Sky is provided by

INNOVATION AND SCIENCE

Pi in the Sky is a publication of the Pacific Institute for the Mathematical Sciences (PIMS). PIMS is supported by the Natural Sciences and Engineering Research Council of Canada, the Province of Alberta, the Province of British Columbia, the Province of Saskatchewan, Simon Fraser University, the University of Alberta, the University of British Columbia, the University of Calgary, the University of Lethbridge, the University of Regina, the University of Victoria and the University of Washington.

Book Reviews ............................................................................ 24 Sacred Mathematics: Japanese Temple Geometry Reviewed by Tom Archibald Uncle Petros and Goldbach's Conjecture Reviewed by Clara Park Pi in the Sky Math Challenges ................................................... 27

the row above. So, for instance, we have, in round 5 (row five - we count the first row as row zero) the entry 10 in column k=5. This is the sum of the numbers 4 and 6 above it in row 4. This corresponds to ten (5, 7) stick arrangements, coming from four (4, 8) arrangements and six (6, 6) arrangements in the previous round. The array of numbers in Table 1 is known as Pascal's Triangle. Later, we will describe another way to find the entries in Pascal's Triangle using binomial coefficients. Note that the sum of all the entries in each row is a power of 2, more precisely, in row n the sum of the entries is 2n. To calculate the probability of holding k sticks after n rounds, n 6, we divide the entry in the kth position by 2 n. For example, after six rounds, since there are 2 =64 different ways to arrive at one of the seven outcomes, so the probability of winning all six sticks from the pointer (or losing all six sticks to the pointer) is 1/64 and the probability that each player still holds six sticks is 20/64 = 5/16.
6

A Simple Version of Lahal

By Peter J. Dukes and David J. Leeming1

In [1], the authors state that \Mathematics is a subject where Aboriginal students can feel particularly isolated and alienated." There are, however, some examples of Aboriginal games and activities that provide opportunities for exploration of the underlying mathematical ideas. One such game is Lahal, also known as `the bones game.' It is usually played by two teams and each team could have anywhere from three to twenty members. The game consists of bones (typically two male, two female) and sticks (six or more). An excellent description of this and other versions of Lahal is contained in [1]. Here, we will present a simple version of Lahal for two players. Each player begins with six sticks and one bone. They takes turns with one player (the hider) hiding the bone in one of his hands and the other player (the pointer) guessing which hand contains the bone. If the pointer guesses correctly he gets a stick from his opponent. If he guesses incorrectly, the hider gets a stick from his opponent. The game continues until one of the players has all twelve sticks. Naturally, we assume a 50% probability that the pointer can correctly guess the hand containing the bone. The outcomes of the first six rounds of this game are shown in Table 1. Specifically, the entry in the kth column of round n indicates the number of possible game sequences which arrive at the outcome of one player holding k sticks. It is easy to compute each successive row in the table as it is the sum of the entries to the right and left in
1

Note that it takes an even number of turns to win a game. (Why?) After six rounds, the hider could be holding 0, 2, 4, 6, 8, 10 or 12 sticks. After six rounds, the game may be over, with the hider winning or losing. In case the game is not over, we cannot rely entirely on Pascal's Triangle any longer to calculate the probabilities of a win or a loss. Let us assume now that the game proceeds to eight or more rounds. The number of ways of reaching one of the outcomes after seven, eight, nine or ten rounds is shown in Table 2 below the line. To calculate the probability of a particular outcome, for example, that each player has six sticks after eight rounds, we find the sum of the numbers in row eight which will be the denominator of the fraction we use. That sum is 248, so the probability that each player has six sticks is 70/248 = 0.282. Using the same reasoning (you can fill in the steps) 1
http://brownsharpie.courtneygibbons.org/

Both authors are with the Department of Mathematics and Statistics, University of Victoria, Victoria, B.C., CANADA V8W 3R4 e-mail: dukes@uvic.ca, leemingd@uvic.ca

is 70/248 = 0.282. Using the same reasoning (you can ll in the steps) we see that after ten y read n choose k) give us rounds,1 the probability that each player has six the n + entries in the nth rowthat after ten rounds, the of Pascals Triangle. A very useful we see sticks is 252/944 = 0.267. Calculate the probaidentity calculating the binomial coecients n inn probabilityn+1. See bility can prove thisplayer has six sticks after twelve that each player has is k + k+1 = k+1 if you that each n! six using the factorial rounds, = 0.267. identitysticks is 252/944 n = k!(nk)! . fourteen rounds. What is hapform k or after n actually counts the From this, we see that probability these probabilities as the number of Calculate the k pening to that 10 number of k-element subsets of an n-element set. rounds increases? each player has six sticks after For this reason, there is a standard convention n twelve deemed equal We could ncontinue this process to nd the that k is rounds, orto afterif fourteen zero k > or krounds. What isgives us a triangular the number of sticks held by the < 0. Pascals Triangle probabilities for happening to display of the nonzero binomial coecients (see these probabilitieshider and the pointer after any even number of as the number Table 1). rounds. However, there are some obvious quesof rounds increases? tions that any mathematician reading this would 2 We could and Generalizations Analysis continue this process be asking by now.

4 back 5 forth 7 gamesimply keep track of the and forth thatmany timesversus how how time and 6 that 8 9 10as 11 12 back two players X wehas won, Y track won, we such track of times and lahal. Call how many how Y the line keep has (Inmany onth on or below simply y = x. of lahal, o X and y-axes of a versusFigure 1:inVarious minim (a, many the many times 3 hasYwon, versus how manymany times how many times how has won.Cartesia times x- Y has Each point X b) times Start 1 Y , and associate them with versus has won.won, is always winning throughou the means one player mea X has won. EachCartesian grid. SupposeX has won. Each b) a and b are the Cartesian point (a, b) in the Cartesian plane, where in the Cartesian plane, integer point (a, b) the nonnegative subdiag in number of Round 1 1 1 instead thesticks trading Eachof game.) Counting point (a, the a plane, where a and1b are nonnegativewe simply keep track of how wins for player X and b win plane, corresponds to nonnegative integers, Round 2 1 2 back and forth that integers, whereand b are are nonnegative integers,proble nal a and where a minimal blattice paths is a famous y = nal or have been a+b rounds wins how to a . There player below the line Round 3 1 corresponds to3a wins for player has and b versus for player Y times on turns out thatwins numb 3 many 1times Y X won,corresponds manywins for It forXplayer X playe in combinatorics. means oneand b the in co corresponds to a wins player is always for player Y . There 4 has been a+b rounds played so Y . There each player hasrounds played or los have won. Each point (a, b) in far, and have been a+b either gained for player the Cartesian 1 4 6 1 Round 4 X of paths on or below line y = Counting the n and |a b| sticks. player Y. Therex,have (0, 0)p b wins for In this game.) the from of equilibrium so nonnegative player has either gained or lost Round 5 1 5 so far, and each player has either and b are lost and n), equals the nth Catalan number, or 10 10 plane,5 where a gained or far, (n, eachintegers, the model, 1 been a+b the this y played soequilibrium has th rounds model, where each player (n, n far, and nal minimal lattice paths is |a this model, the wins for b| state is In wins Round 6 1 6 15 b| sticks. In corresponds to a equilibrium sticks.and b line = x, the 20 15 6 1 |a player X lost each line y has either combinatorics. It turns o in state is the line y = x, whereYeach player has thea+bplayer = x,of sticks.gained orhas the 2n for player 0 .to 6 There havestate is same number where each player . been therounds played = 1 Table 1: Pascals Triangle:numberversus rounds, sticks. In thisunfolds or path in y of same sticks of sticks. and each player has |a-b| Each gamelost lahalpaths on as abelow line th same number of sticks. Cn of n + model, the 1 n so far, either gained or `equilibrium' state (n, as to y=x, Each game of lahal unfolds as a path model, the equilibrium from n), equals (a, b), the cur the game of grid, is the 0)line the nth Catala |a b| sticks. In this in Each Cartesianlahal unfolds (0, a path in the where each player has the alongnumber Cartesian grid, from (0, 0) to (a, b),x, where each player fromthe Moves same [2]thesemore detai grid, has state is the 9 y = the cur- rent situation. (0, 0)reference the cur- 1 Se line 10 11Cartesian See the excellent to (a, b), for paths ar 1 2 3 4 5 6 7 8 of either rent situation. Moves number these paths are sticks. east or north, according ton whether m on minimal lattice paths and theC are on X Catalan num same along of sticks. rent situation. Moves along these paths = Round 5 1 5 10 10 5 1 1 or either east or north, accordingof lahal unfolds Eachpath in of according to givenas aXn +Suc to whether X as aYor(respectively) wins a whether bers. north, either east game theLahal unfolds round. bers Each game 6 Round 6 6 15 20 15 or Y (respectively) wins a given round.(0,Such(respectively) cur- alahal. the excellent reference Y Cartesian 21 grid, from 6 or 0) to paths, which wins backtrack are Such gener (a, in the Cartesian Suppose called Le b), return to See round. in `path' Lets the never given grid, from a min Round 7 6 21 35 35 paths. paths, which never backtrack are Movespaths, which neverJust as on our early lattice paths aft called mini- mal game backtrack are called mini- laha lahal paths with kin minimal player, thatther rent 56 situation. 27 along 0) to (a, b), are currentper discussion, an (0, these the sticks situation. 27 56 70 Round 8 mal Just aseither east or north, according2n Just asone our early discussion, there and at paths. in our early discussion,mal are turns pathsplayer (0, 0)kto (a, sticks, is th there to a+b paths. whether Xfrom has + r b). This 2n t in a bers. Moves along these paths are either east Round 9 27 83 126 126 83 27 a+b of Pascals Triangle are a paths from Y (respectively) wins a givenpaths from(a + b)th (a,<return to ath othe (0, 0) to (a, b). This areatha entry in theSuchsticks, row r < k. is lahal. Su is a+b other has k (0, 0) to 0 b). This r or 209 round. or north, according toLets rows X or whether begins Round 10 110 209 252 110 entry in the (a + b)th row of Pascalsbacktrack where indexing of ways Pascals with kissticks Th Triangle,in are called b)th rowlahalandcan occur inwithpe The + mini- of game Triangle, one-t0 entry the (anumber entries this paths, which never (0, where indexing entries 10 rows as in with 0. Figure 1 shows and rows the number of has k one correspondence with begins player where discussion, there 2n minimal paths 0. minim Table 2: sticks versus rounds, mal paths. Justbegins our earlyindexing entries various turns one withfrom one 0 5 to and minimal 2 Figure 1 shows various a+b paths from (0, 0) to (a, b). This is ath (0, 0) paths fromsticks, 0latti paths fromFigure 1lattice 10). from other to (n+r,rnr) which r (0, 0) to (14, paths minimal has k (0, 0) < d shows various are a

Yand associatepointer the x- and y-axes of a , and the them with after any even Figure 1: Various minimal lattice paths (and read `n k) give us the n the Cartesian grid. Suppose instead of sticks trading read n choose choose k') give us+ 1 entries i number of rounds. keep track of how there th back and forth that we simplyHowever, choose k) give us the nn + entries in in y nth row of Pascals Triangle. A very usefu y 2 read n + 1 (In read n the choose k) give us the n + 1 entries on or below the line y = x. 1 entriesthis the n row of Pascal's Triangle.in lahal, many some obvious versus how many times mathematician are times Y has won, questionsnth row of Pascals Triangle. A winning useful identity in calculating the binomial coecient that anymeans one player is alwaysAvery throughout the very the nth identity Pascals Triangle. A binomial useful row of in calculating the very useful X has won. Each point (a, b) in the Cartesian number of subdiagoreading this wouldnonnegative integers,calculating Counting thecoefficients is is ncalculating the binomial coecients be identity by now. game.) the binomial coecients in + n = n+1. See if you can prove th asking in the See if you plane, where a and b are identity k+1 n nal lattice paths is a famous n nk n+1 k+1 minimal problem n! correspondsthere a formula X n computing n+1 number and b = the . n if you can that is number k+1 1 = using in factorial form form See It turns k) prove kthis nidentityk+1 See if factorial n = 1. Is to a wins for playerisfor + wins in combinatorics.choose can givethe this+identityentries .the the you can prove thisk!(nk) prove the + using y out k k+1 k+1 read us k n for player Y . There have been a+b rounds played n n n!(0, 0) to n! of sequences resulting identity lost number onnthbelow line y = = from . From this, factorial that in a given of paths or row factorial form k x, k!(nk)! . using this, we see form = k!(nk)! . actually Triangle. A the we see so far, and each player has either gained or using the theof sticks of Pascals identityFrom very useful that k k actually counts th n (n, n), equals thenth Catalan number, or n |a b| sticks.player?model, the equilibriumwe see identity inactually counts the this, we of k-element actuallyof an n-element se for each In this 10 number of that counts Fromnumber see k-element subsets of the 10 the From this, that k calculating the binomial coecients k subsets countsan n+1 state is the line y = x, where each player has the n standard conventio number of k-element na + n k+1 1= 2n . . Seeset. Fork-element this thereana a standard number of this reason, this C 2. Willofthe game almost surely endis subsets=of an1 n-element set.you can n reason, there isis n-element set. in k finiten + n-element if For prove subsets of k+1 same number sticks. n deemed equal zero if if k For this reason, there is a using the factorial form reason, deemed equal to to convention > n o standard convention n is there is a standard zero k > For that k = k!(nk)! . thisthat k is n! Each game of lahal unfolds as a path in the identity convention number of turns? Cartesian grid, from (0, 0) to (a, b), the cur- deemed equal to zero if k > ndetails is deemed equal to zero gives us n triangula See the excellent reference [2] for more or n k < 0. Pascals Triangle if k > a or that n is k n or kCatalanPascal's Triangle gives us a triangular < 0. numk From this, we see that that actually counts the 10 k rent 3. If not, is along these paths Pascals Triangle the us athe situation. Moves there a limitingonvalue to gives and triangular display of the nonzero binomial coecients (se minimal lattice paths k < 0. are display k <the nonzero Triangle gives us a triangular of 0. Pascals binomial coefficients (see either east or north, according to whether X bers. number of k-element subsets of an n-element set. probabilities? display Such Lets return to lahal. coecients a general the 1). display of or Y (respectively) wins a given round. of the nonzero this reason,Suppose 1). a(see Table convention For binomial Table in standard nonzero binomial coecients (see there is n k sticks per player,Tableafter Table 1). paths, which never backtrack arequestions requires probing that zero The answers to these called mini- lahal game with is deemed equal to 1). if k > n or that mal Just as in our early discussion, there 2n turns one k paths. player has k + r sticks, and the deeper into the mathematics behind this simple k k 0. Pascals Triangle gives us a triangular are a+b paths from (0, 0) to (a, b). This is ath other has< r sticks, 0 < r < k. a game.the (a + b)thpursue theseTriangle, Thedisplayofof in this can occurbinomial coecients (see We will row of Pascals and other questions the nonzero is in one-toentry in number ways where indexingof thisand rows begins with 0. one correspondence with the number of minimal is helpfuldiagram a diagram which depicts Section 2 entries article. Table 1). It is helpful It draw a to draw which depicts a 14 x to Figure 1 shows various minimal paths from (0, 0) lattice paths from (0, 0) to (n+r, nr) which do game xk and gameto draw a diagram which X players It helpful such as lahal. Call the two and to (14, 10). not coefficients, If you are familiar It is helpful to draw a diagram which =suchisasy Lahal. Call the two players depicts a X an with binomial intersect either of the lines y depicts a = game reaches lahal. the the two restrictionX andthis andlahal. Callthem and y-axes X and Call next players game such as associate x- with the x- of y-axes of One player wins when the games path such as x + k. Like the Catalan Y, and associate ,them with thethe two players and a above, Y then you may wish to go directly to a boundary, which is a certain diagonal line. is another with the x- and y-axes, of a associate them with the minimal lattice paths Figure 1: Suppose instead of sticks Y , and associate them problem on lattice paths, now with Supposegrid.Various x- and y-axes of a tradin Y and Section. If you are boundaries further. two linear boundaries. Cartesian grid.Cartesian instead of sticks trading We must now discuss these not, then this brief introduction See Figure 3 for n = 12, Cartesian grid. Suppose instead of sticks trading back and forth that we simply keep track of how Cartesian and forth a instead of sticks trading back grid. Suppose k is A minimal lattice path is subdiagonal if it lies r = 2, It = 6.helpful to draw a diagram which depictsthat we simply keep

The answers to these questions requires probing deeper into the mathematics behind this simmay be useful to these and other ple game. We will pursue you in reading queswhat follows. tions in Section 2 of this article. If you First, let us with binomial coecients, are familiar examine factorials. then you may non-negative integerSection 2. If For any wish to go directly to n, the you are not, thenfactorial whose symbol be number n this brief introduction may usefulis n! is in reading what= 1, 1! = 1, to you defined by 0! follows. First, let us examine factorials. For any nonn! = n(n - 1) . . . (2)(1), so for negative integer n, the number n factorial whose instance 4! = 24, 9! = 362880, symbol is n! is dened by 0! = 1, 1! = 1, and so on. n! = n(n 1) . . . (2)(1), so for instance 4! = 24, 9! = 362 For and so on. 880, nonnegative integers n x 14 to helpful to draw a diagram which depicts a find the probabilities for the It is Forand k, 0 k n, the binomial n, nonnegative integers n and k, 0 k number lahal. Call the 1. Is therehider Figurecomputing the number paths binomial coecients, denotedby n (and game such asof sticks held by the X and two players by k a formula for 1: Various minimal lattice the coefficients, denoted

Analysis Analysis and Generalization 2 and Generalizations Analysis and Generalizations 2 Analysis and Generalizations

Analysis and Generalizations

outcomeCn = delivers boundaries same number of sticks. g n+ the binomial coecients other, no matter h . See if you can prove this Each game of lahal unfolds as a path in the n! 2n CartesianBy results in [3] 0) to (a, b),of such paths is the excellent refere grid, from (0, the number the cur- 2n See 2n 2n 2n 2n ctorial form n = k!(nk)! . k + + . rent situation. 12i + 12i paths are +on minimal lattice paths n + 12i + 1 n + 12i + 2 n +Moves 6alongn these + 7 + n +8 12i References counts the at n actually i= n + 12i10 2n k 2naccording to whether X bers. either east or north, subsets of an n-element set. Figure 2: f(n), the total number of non-ending sequences after 2n rounds rounds + 1)k . + r + Such n + wins a or Y (respectively) r + 2ik givennround. (2i Lets[1] T. Pelton, L. F return to lahal. i= e is a standard convention Yzero if k > n or Figure 2: f (n), the total number of non-endingbacktrack aftercalled mini- almostgameterms k stick (respectively) wins a given round. Such paths, Note that the sumare finite, since lahal all Exploring prob paths, which never sequences is 2n rounds with equal to Note that the sum is nite, since almost all terms Vector, BCAM which a triangular mal Just as in paths. are angle gives us never `backtrack' are called minimal paths. zero. zero. our early discussion, there 2n turns one player has are are a paths our case, to (a, This is ath other k=6, and Just as y our early discussion, there are a+b In our from (0, 0) 6 endsb). formulaformula for has k r sticks, 0 in (see o binomial coecients Incase, we apply this this for one 6, [2] The case r = we apply game, and k = cannot R. Brualdi, In the entry in (a + b)th row of Pascals of paths from (0, 0) to (a, b). This is ath entry in thetheexplicitly..apply 5obtain the Triangle, r=0, . to to obtain in this case. The numberPrenticeth the distribution for the ways H and r = 0, 1, . .., 5 , the formula distribution for ed. th where indexing entries and rows begins with 0. one correspondence with 2n possible sequences for 2n coin (a+b) row of Pascal's Triangle, where indexing 2nth round. 2For instance, thethe number of game There are the 2nth round. For instance, number of Figure 1 minimal paths from (0, 0) entries and rows begins with 0. Figure 1 showsshows various which resultthe two players lattice paths from (0, 0) t ips (or wins/losses by in two pointer). Some sticks game sequences which resultin lahal players with 6 S. Dua, S. Kh with [3] sequences nd Generalizationspaths from (0, 0) to (14, 10). 10). number f (n) of these sequences result in the ed random w to (14, not intersect either of the various minimal 6 sticks each (i.e. = = 0) after 2n= 20 rounds is r each (i.e. r the after path = 20 rounds is wins when 0) 2n One player14 not being games the reaches x + k. Like the Catalan x over by round. This 2n diagonal line. when game's pathareaches game is a certain 2n 2nth is another reecting barr the 2n a diagram which10 wins 2n depicts 2n boundary, 20 which 2n20 problem One player a 2n 2n 2n 2n f (n) can 2n 20 the2n 066. above, and 169175. on l + + . formula 175 Call the a `boundary,' andn + is a certain 12i + 1 n + We + 2 discussbe found using += further. . two linear 8 two players X which 12i + n + diagonal line.must now n + 12i4 6 n 12i + 7 + 12i + n + 12i + boundaries. S Wen + 12i + 6 10 n + 12i + 7 16 n + 12i + 8 these boundaries n + 12i i=n + 12i + 1 n + 12i + 2minimal lattice paths fully in Figure 2. (Who would have is shown i= m with the x- and y-axes of a theseFigure 1: Various must now discuss boundaries further. A minimal lattice path is subdiagonal if it lies r = 2, k = Pascals Trian [4] 6. thought that afterodd number 2n 2n + such The The behaviorsuch a an odd number+ 1 to 1 of rounds behavior after an simple game leads of ose instead of sticks trading A minimal lattice path is subdiagonal if number of non-ending sequences probability of the game en.wikipedia it lies a long be determined from round rounds can formula!?) The after 2n rounds total can be after 2n using we simply keep track of howf (n),Figure 2: f (n), theof non-ending sequencesdetermined from round 2n using the `sum Figure 2: line ythe below number this means ending on or before the 2nth round is total the line y = x. (In lahal, this 2n rounds 3 on or below the on or x. (In Lahal, = the sum rule which generates Tables 1 and2. rule' which generates Tables 1 and 2. on, versus how many times means one player is always winning throughout one the Cartesian oint (a, b) inplayer is yalways winning throughout the game.) The case rr = 1 2 the game, and one cannot 6 game, y Counting the number of subdiagonal minimal of subdiago- rP== 6ends2n f (n). the game.) Counting the number r = 6 The case game, ends one cannotand one cannot ends the and The case b are nonnegativeintegers, 2n 2n nal minimal lattice paths is 2n famous problem apply the formulain this case. a explicitly apply in formula in this case.4 lattice and b 2n a famous2n problem in2n combinatorics. explicitly the formula the ns for player 12i paths 12i + 1 + n + 12i + 2 n + 12i + 6 explicitly apply 12i + 8 . are 22nthis case. sequences for 2n coin X + n + is wins n+ This possible in number of paths turnsnout that the + actually tends to 1 as n2n coin Therefore, combinatorics. It on or + 12i + 7 are n 2n There sequences for . number i= below There 2 possible 22n possible sequences for 2n coin There are ave beenIt turns outplayed the a+b rounds that a game of lahal (almost by the lahal end. will ips y14= x, from (0, 0) (or wins/losses surely) Somepointer). Some th line y x,or lost (0, paths on orbelow line the (or wins/losses to = from of 0) to flips is wins/lossespointer). Lahal tossed by er has either Figure 2: f (n), the number of(n, n), equals ips x rounds This by the lahal thesesequences result in the total n), equals the nth Catalan number, or (or rather intuitive. If a fair coinpointer). Some n f (n) of the gained is non-ending sequences after 2n number (n, 2n 2n the Catalan2n number f (n) of these you 2n number, or 2n number notsequences resultthe2n result. in run game f(n) being over ofshould sequences these by in 2nth round. This the is model, the equilibrium + + repeatedly, game n + 12i + 7 expect+eventually a 10 n + 12i n+ + gamen + 12i + 6 2 by y10 , where each player has the n + 12i + 1 The case12i16 ends the not being (n) can beaby2nthusing n 12i a This f(n) can be notoneovereventhemillion) heads This row. This and being over found 2nth the in + 8 above, the round. formula round. i= Figure 3: Two minimal lattice rpaths 2n game, of 12 (or = within and f cannot Cn = . f formula in this case. using the formula above, and is shown fully explicitly apply the (n) can be foundusing the formula above, and s. found shown n+1 n . outcomecoin delivers the win for one player over the is in Figure 2. (Who would have boundaries There are 22n possible sequences for 2nFigurefully (Who would have is shown fullyFigure 2. (Who would have thought that such in 2. l unfolds as a path in the in sequencesthat how many sticks are in play! such other, no matter suchrounds thought after 2n a simple game leads to ips number of non-ending Figure 2: f (n), the total(or wins/losses by the lahal pointer). Some thought that such a game leads leads to simple game See (n) these sequences simple the m (0, 0) to (a, b), the curSee the excellent freference [2] for moreadetailsformula!?) to such such of the game number the ofexcellent a result long in The probability By results in [3] the numbernot being overlongisformula!?) The probability of the a long formula!?) The of such paths the 2nth round. This a moreCatalan numgame game ves along these paths are on minimal reference [2] for by lattice paths and the probability of the game ending on is before the 10 ending on or before the 2nth round or y f foundending on or before the using the above, X bers. (n) can beon minimalformulath r = and 2nth round is References The 2n case 6 is , according to whether details fully Figure 2. (Who roundends the game, and one cannot 2n is shown 2n . game leads to such latticelahal.in and explicitly would have formula = 1 2case. (n). wins a given round. Such+ r + 2ik return + r that such Suppose in a apply the 2n P r in this 2n f Lets general thought + (2i + a simple n to paths 1)k the n f sequences for 2n coin Pthe= 1 possible (n). r 2n i= a long formula!?) The player,[1]are after backtrack are called mini- lahal game Catalan numbers. Therethat Pelton,2L. Francis-Pelton and L. Williams, with k sticks per probability ofT. 2game ending on or before the 2nth round isExploring probability with aboriginal games, our early discussion,that the2n turns one since almost k + ips (or wins/losses bytends to 11 as n . Therefore, a This as sticks, and the This actuallyas tends to Therefore, . Note there sum is nite, player has This ractually tends actuallythe lahal pointer). Some Therefore, all 1 Let's returnterms (n). f (n) oftothese n Spring result in the tof Vector, BCAMT, . surely) will end. 2006. 2n sequences (0, 0) to (a, b). are zero. This is ath other has k r sticks, 0 r = r number a game of lahal (almost P < 1 <2k. game of game of Lahal (almost surely) will end. (almost 14 Lahal. x thisacangame xlahalone-to- surely) will end. tossed h row of Pascals In our case, we apply this formula for to in a is inR. This is rather intuitive. If a fair coin is4th Triangle, The number of14 Supposek 1= 6, not being over by the 2nth round. This ways tends occur [2] Therefore, 10 Brualdi, Introductory is tossed This actually . general with(almost fas is can This is rathera intuitive.Combinatorics, is tossed Lahal This n rather intuitive. If the formula If a eventually a run game will end.found using fair coinexpect fair coin above, repeatedly, Hall, New es and rows begins with 0, 1, .one 5 to obtain the distribution (n) of be Prentice you shouldJersey, 2004. minimal a game of surely) and r = 0. . . , correspondence lahal the numberrepeatedly, you should expect run and for ed. repeatedly, you 14 x sticks is shown ofshould expect eventually a eventually a run of This tossed us minimal pathstheminimalround.paths minimal latticeto (n+r,Ifnr)coin is 12in Figure a million) would in a row. This from (0, 0) 3:lattice pathswith iskthe number within which do even 2. (Who heads have from (0, 0) pathspera Figure latticeTwo instance, rather intuitive. (or fair fully (or heads in a row. This 2nth For Figure 3: Two minimal lattice paths within of 12 of even a million) Figure 3: Two repeatedly, you should expect eventually even a million) heads in a row. This outcome 12 (or a run outcome not either y = in row. withinboundaries which result in twothe million) heads [3]that the= S. for onethe win overone such modi- the boundariespaths within player, even a lines thought S. Dua, delivers player for to game sequences intersectof 12 (or of that after xka and such a simple game leads the player over players with delivers y win Khadilkar and K. Sen, A outcome deliversThis win for one player over the other, no Figure 3: Two minimal lattice boundaries the 2n turns onerestriction above, this Thewalk how many the game in play! playerone player over the matter in the presence ofare ed random probability of sticks partially a other, hen the games path reaches (i.e. k. = 0) after Catalan the win foris formula!?)no the 2n = 20 6 sticks each x + r Like outcome delivers roundslong boundaries other, no matter play! many sticks are are in play! how play! matter how many sticks in is Prob. 13 (1976), how in has diagonal By resultsanother other, on -latticeending are or before the 2nth round has is a certain k + r sticks,and in other no mattersuchmany stickson reecting barriers. J. Appl. line. is the the numberkof r sticks, is now with paths, [3] problem paths 20 such linear 20 By results further. number results < k. theseBy0 < r in [3] in [3] the two paths boundaries. 066. Figure for 169175. boundaries the number of 20 ofissuch paths is See 3 n = 12, = 175 2n References f (n). 10 2n= 4 path is The numberlies ways2, k= 6.References 2n subdiagonal2n it of r 2n 2n 16occur is in References P r = 1 2 References if 2n this. can one-to- . [4] Pascals Triangle on Wikipedia: http:// n + rn ++ 1)k 2ik [1] + (2i r + n + r + 2ik n + .r L. (2i + 1)k and L.[1] + Williams, i= number Francis-Peltonactually T. Pelton, L. n . Therefore, of one correspondence after ran (2i +T. Pelton,of 2n +T.with aboriginalFrancis-Pelton Francis-Pelton and L. Williams, minimal n + + number 1)k n The + i= + r behavior with theodd Exploring probability1 Pelton, [1] T. Pelton,1L. Francis-Pelton and L. Williams, 2ik tends en.wikipedia.org/wiki/Pascals triangle [1] This L. games, to asand L. Williams, i= Note that the sum is nite,can 3 determined from round 2n using since almost all rounds frombe 0) terms(n + r, n - r) which2006. of lahal (almost surely) will with aboriginal games, Exploring probability end. lattice paths that(0, sum is nite, since almostExploring probability with aboriginal games, Vector, BCAMT, Spring a terms Exploring probability with aboriginal games, Note the to allgame are zero. 14 sum either of almost all y = x Note that the sum is rule which the lines Tables 1- and 2. This is rather4th do case, we apply thisnite, since = 6, [2] R. Brualdi,x k Vector, Vector,Vector, BCAMT, Spring 2006. not the are zero. intersect formula for k generates terms Introductory Combinatorics, BCAMT, Spring 2006. tossed and BCAMT, intuitive. If a fair coin is In our Spring 2006. are zero. . , to Like the Catalan restriction above, this and ry = 1, . .+5k. obtain the distribution for = 0, x ed. Prentice Hall, New Jersey, 2004. you should expect eventually a run repeatedly,[2] R. Brualdi, Introductory Combinatorics, 4th In our case, we apply this formula for k = 6, [2] R. Brualdi, Introductory Combinatorics, 4th the In our case, we applythe number of 2nth round. For instance, this formula for k = 6, [2] R. Brualdi, Introductory Combinatorics, 4th a Figure 3: and r = on lattice obtain the with two 4(or evened. million) heads in a row. This Two minimal 5 to [3] S. Dua, within of 12 gameis another problem0, 1, . . . ,lattice paths S. distribution K. Sen, A modi-Prentice New Jersey, 2004. sequences which result in two players with paths, now Khadilkar and for Prentice Hall, Hall, New Jersey, 2004. and boundaries. , 5 to obtain the distribution for walk in outcome delivers the win for one player over the r = 0, 1, . . ed. Prentice Hall, New Jersey, 2004. ed random12, r= 2, ed. of partially the presence 6 sticks each (i.e. r = 0) after 2nSee rounds For instance, the number of linear boundaries. = 20 Figure 3 for n = the 2nth round. is reecting barriers. J. Appl. Prob.matter how many sticks are in play! the 2nth round. For instance, the number of other, no 13 (1976), [3] S. Dua, S. Khadilkar and K. Sen, A modik = 6. 20 game 20 20 sequences which result in two players with [3] S. Dua, S. KhadilkarAand K. Sen, A modified 169175. game 10 4 16 = 175 066. two players with [3] S. Dua, S. Khadilkar and K. Sen, the modisequences which result in ed random walk in of partially reflecting By results in [3] the (i.e. r of ofafter paths ed such paths is rounds random walk in the presence presence of partially 6 is By resultssticks eachnumber= 0) Pascals2n = 20 on random http:// 6 sticks each (i.e. in= 0)the number[4] rounds Triangle is Wikipedia: walk in the presence of partially r [3] after 2n = 20 such is reecting barriers. J. (1976), 169-175. The behavior after an odd number 2n 1 of barriers. J.J. Appl.Prob. 13 Appl. Prob. 13 (1976), Appl. Prob. 13 (1976), + en.wikipedia.org/wiki/Pascals triangle reecting barriers. References 2n 20 2n 20 20 rounds can determined from round using be 169175. 20 20 20 2n = 175 066. Pascal's Triangle on Wikipedia: the sum rule which generates 10 and 2. Tables + r + 2ik 1 = 4 r 066.16 1)k . 169175. [4] n 175 + (2i + + n 4 [1] T. Pelton, L. Francis-Pelton and L. Williams, triangle 10 16 i= http://en.wikipedia.org/wiki/Pascal's http:// [4] Pascals Triangle on Wikipedia: [4] Pascals Triangle on Wikipedia: http:// The behavior after 4 odd number 2n +Exploring probability with aboriginal games, triangle an 1 of en.wikipedia.org/wiki/Pascals The Note that the sum isodd numberalmost allof behavior after an nite, since 2n + 1 terms en.wikipedia.org/wiki/Pascals triangle Vector, rounds can be determined from round 2n using BCAMT, Spring 2006. 3 rounds can be determined from round 2n using are zero. the sum rule which generates Tables 1 and 2.

Who in the Land is Fairest of All?

MyHeritage.com is an internetbased company that offers you the chance to see which By P. H. Trinh celebrity you most resemble. Remember how in Snow Oxford University White, the queen has a magical mirror which provides her with uninhibited flattery? This is the same, but like, tons better. After a free signup, you upload a large-ish jpeg of your mug, then let the software crank away. My personal resemblance results were: Brad Pitt (71%), Keanu Reeves (63%), Luke Perry (63%), and Matt Damon (63%). Brad Pitt? Really? Matt Damon? Really? Who wouldathunk? But y'know, as I gaze into the mirror... well... yes, I see it now. Definitely. We're practically brothers! How does it all work? Is this actual science or just deceptive flattery? To understand how facial recognition works, we're going to have to delve into the mathematics behind the algorithm.

Seventh Grade Blues

Back in the seventh grade, one of the girls told me I looked like Keanu Reeves. Seriously, I was hanging upside-down on the jungle gym, minding my own business, and she just walked over and told me like it was no biggie. Then she giggled like a moron and ran away. Now while this singular moment of brazen flattery would become the highlight of my paste-eating academic career, I was also torn. On one hand, I wondered how anyone could confuse Keanu (black shades, gothic trench coat, totally awesome) with me (pubescent, angst-ridden, gawky)? Was this How might all some awfully cruel and we go about sadistic joke girls liked to identifying that play on unsuspecting boys? person from a On the other hand, maybe large database | maybe she was on to of faces? something. Maybe somewhere | somehow, behind all that bad acne and ruffled hair, my hidden Keanu-like features beckoned faintly, like some distant lighthouse A more obscured by fog. computationally Today, however, thanks to the latest advances in facial recognition, I no longer have to wonder: she was right.

Recognizing Faces
Suppose we were given someone's picture. How might we go about identifying that person from a large database of faces? One way we can go about it is by identifying the characteristics of the subject | perhaps the person has small lips, or a pointed chin, or distinctive eyes. From here, we then consult the database, going from picture to picture, each time isolating the features of the faces and

be to examine these faces as a statistical whole rather than as the sum of their parts.

One way we can go about it is by identifying the characteristics efficient way would of the subject

checking for a match. Whilst this might work, it would also be a lot of work; As algorithms would need stored to be defined to analyse in a each desired feature. computer, a picture is nothing Imagine having to do Now more than a great this for each of the big grid of dots thousands of faces in the (or pixels). that stream pass the abstract gates at a football match theory of or in a busy airport. linear algebra, these grids of A more computationally pixels are called efficient way would be to examine vectors . these faces as a statistical whole rather than as the sum of their parts. This is similar to the difference between identifying a city by its landmarks and identifying the same city by the density of its roads, the clusters and heights of its buildings, its downtown areas, and so on.

higher HIGHER DIMENSIONAL COORDINATES IN A dimensional plane. PLANE.

Figure 2: 2: FACES CAN BE IDENTIFIED AS FIGURE Faces can be identified as coordinates in a

as well as a direction. Moreover, they can be Like vectors representing force or motion, added, subtracted, multiplied, and manipulated these new face vectors have a | the like most other mathematical quantities `magnitude (an overall they inhabit some higheronly difference is that brightness), as well as a direction. Moreover, they can be added, subtracted, dimensional face space, rather than the two or multiplied, and manipulated like most other three dimensional physical world we the only difference mathematical quantities live in.

Whats Your Eigenface Basis? or threeface space, rather than the two
However, face spaces are complicated affairs | they're high dimensional boxes stuffed with a large number of faces, each face containing thousands of pixels. It would thus be foolish to try and
dimensional physical world we live in.

is that they inhabit some higher-dimensional

A Picture is Worth a Thousand Digits

Snap! really? But what are pictures,

As stored in a computer, a picture is Think of how, when we analyse nothing more than the motion of a ball flying through the a IGURE 1: A grid of F great big Whats Your Eigenface Basis? air, we break the motion into its horizontal PICTURE pixels). If the picture dots (orIS and vertical components. These two components However, are complicated affairs NOTHING MORE is greyscale, each pixel is Face Spaces fundamental basis capable of describing provide a theyre high dimensional boxes stuffed with THAN A LARGE associated with a number any arbitrary motion. a large number of faces, each face containing GRID fromOF to 255 representing its 0 NUMBERS. thousands of pixels. Similarly, once the eigenface basis is found using brightness, from pitch black linear algebra, each face in the database can then (0) to pure white (255).would thus be foolish to try and compare It be expressed using certain percentages of each of Now in the abstracteach face pixel by pixel; instead we look to theory of linear algebra,construct a small group of pictures these representing the general facial patterns of the grids of pixels are called database. vectors. You've probably This small but crucial group is called the eigenface basis. encountered vectors before in Physics class and in fact, these `face vectors' are quite similar. Like vectors representing force or motion, these new `face vectors' have a magnitude (an overall brightness), Figure 1: A picture is nothing
more than a large grid of Now in numbers. the abstract theory of linear algebra,

...we look to compare each face pixel by pixel; instead construct a small we look to construct a small group group of pictures representing the general of pictures representing the general facial patterns of the database. This facial patterns of the database. This small but small but crucial group is called the crucial group is called eigenface basis. the eigenface basis.

these grids of pixels are called vectors.

http://brownsharpie.courtneygibbons.org/

database. This small but crucial group is called the eigenface basis.

FIG

eigenface components.

, . n

, . , r e l -

Now the problem of recognising the subject becomes as simple as finding the shortest distance (or In optimal closest match) between conditions (with our subject and the good lighting, faces in the database, a representative a process aided database, front-facing pictures, etc.), a simple enormously by the eigenface routine might fact that each face FIGURE 3: VECTORS CAN BE DECOMPOSED INTO produce accurate (FA Figure 3: Vectors can be decomposed into is now represented BASIS ELEMENTS. elements. AT& basis readings of up to 90%. by only a handful of the building blocks. For the motion Think of how, when we analyse example, of eigenface components. a we may say thatthe air, we break the ball flying through a picture is composed its horizontal first motion into of 10% of the and vertical The Future and You eigenface, 25% of components components. These two the second, provide a Unfortunately, fundamental basis capable of describing any is never real life 4% of the third, and so on. But really, just how that simple, and arbitrary motion. accurate are these eigenface The beauty of this one must contend algorithms? treatment the that Similarly, once is eigenfaceeven is found basis with a multitude of in a large database, in the noisy In optimal conditions using linear algebra, each face each database factors. unique can then face can be expressed certain be expressed using (with good lighting, a Figure 4: A database of faces Dist A Problem of very simply using its eigenface percentages of each of the building blocks. For representative database, can be used to construct an eigenface basis. example, we may We no longer haveis to express front-facing pictures, etc.), Now imagine Afterwards, in each face decomposition. say that a picture a face under scrutiny can be composed of 10% of the first eigenface, 25% now, like a simple eigenface routine represented in terms of each face using thousands of pixels; decomposed into different ofa simple recipe the third, and so on. the second, 4% of in which the eigenfaces are the percentages of to coordinate might produce accurate percentages, akin each higher-dimensional plane. O eigenface (Faces courtesy of key ingredients, the entire database can be readings of up to 90%. The beauty of this treatment is that even in a (which may or may not lie in AT&T Cambridge) reconstructed as it was before. large database, each unique face can be Unfortunately, real life is then projected onto this plane expressed very simply using its eigenface never that simple, and one in terms of the eigenface comp decomposition. We no longer have to express must contend with a multitude of `noisy' factors. each face usingA Problem pixels; now, like thousands of of Distance These include variance in pose (person facing a simple recipe in which the eigenfaces are the at an angle), obstructions (sunglasses or other key Now imagine entire databasein the database, ingredients, the each face can be reconstructed as in was before.its eigenface percentages, people), resolution, lighting, and so on. Despite represented it terms of this, however, the science of facial recognition has akin to coordinates lying in some highersteadily improved to the point where today, it is dimensional plane. Our test subject (which may becoming a standard for many military, security, or may not lie in the database) is then projected and commercial applications. onto this plane by expressing it in terms of the

DAT FAC CAN CON EIG AFT FAC SCR DEC DIF PER EAC

FIGURE 5: FINDING THE CLOSE DOWN TO CALCULATING THE SH DISTANCE BETWEEN POINTS.

http://xkcd.com/

in terms of the eigenface components.

But I digress. You see, the whole point of this article was that the inner workings of facial recognition is nothing but science and mathematics. That's right. It's not halfdrunken, hand-wavy speculation that I resemble Brad Pitt and Keanu Reeves. I really, really do. It's backed up with science and everything.

drunken, hand-wavy speculation that resemble Brad Pitt and Keannu Reeves. really, really do. Its backed up with scienc and everything. FURTHER READING

[1] M. Turk and A. Pentland. Eigenfaces FURTHER Cognitive for recognition. Journal of READING Neuroscience (1991) 3 (1), 71-86.

[1] M. Turk and A. Pentland. Eigenfaces for

recognition. Journal of Cognitive Neuroscience [2] W, Zhao and R. Chellappa. Face (1991) 3 (1), 7186. Processing: Advanced Modeling and Methods,Zhao and R. Chellappa. Face Processing: Academic Press, 2006. [2] W, Advanced Modeling and Methods, Academic Press, 2006.

closest match boils down to FFigure 5: Finding theTHE CLOSEST MATCH BOILS IGURE 5: FINDING
DOWN TO CALCULATING THE SHORTEST DISTANCE BETWEEN POINTS.

calculating the shortest distance between points.

By

The idea for this article Anthony Quas and arose from a simple magic Peter J. Dukes trick based on error-correcting University of codes. This activity has been Victoria used in Math Mania, an outreach event in which students and staff members at the University of Victoria visit local elementary schools for hands-on math activities with students. In the trick, the young student is presented with a rectangular array of cups, some face up and some face down. While the magician looks away, the subject is asked to invert (i.e. flip) any cup. Looking back at the cups, the magician then is able to determine which cup the student inverted, without any memorizing! To do this trick, the cups are initially put into a configuration in which each row and each column contains an even number of cups that are the right way up, as in Figure 1. After a cup is inverted, the column in which it sits contains an odd number of cups the right way up (there is either one more Figure 1: Rectangular array of
cups used in the magic trick.

or one less than before); and similarly the row in which it sits contains an odd number of cups the right way up. This allows the magician to figure out which row and column the cup that the student inverted is in. Of course this is enough to figure out which cup was inverted. In this article, we will discuss the connections between this magic trick and error-correcting codes, and investigate an extension where hexagonal arrays replace the rectangular array of cups. It is shown that the hexagonal pattern corresponds to a more robust code (and hence a flashier magic trick!) More on this later; we now comment on codes and their uses. In a basic model of a communication system there is a source that produces some sort of data that is to be sent to a receiver. A typical first step in this process is for the data to be transformed into binary digits (bits - 0s and 1s), with strings of several bits representing one piece of data. If we want to use strings of k bits to represent pieces of data then there are 2k different strings. This means that if there are N different pieces of data, then we need to choose k with 2kN so that all of the possibilities can be encoded. For example, if the data that we want to transmit are letters of the alphabet, then since 24 < 26 25 we need to use 7

The minimum distance of a code is important with respect to error detection and correction since it is the 0 1 1 0 0 1 1 0 minimum number of errors that must occur in order to the error can 1 0 1 0 1 0 1 0 strings of at least 5 bits to codeword into be fixed). code C can transform one another. A encode the letters of the 0 0 0 0 0 0 0 0 Now let or return to the magic trick with cups. detect all combinations of t us fewer errors if and only calculate 1 0 0 1 1 0 1 1 alphabet. We canthe minimum Instead of thinking of cups1. To right way up and if distance of C is at least t + the see the number of bits required the wrong way up, we write a 1 for a cup the right 1 0 1 0 1 0 1 0 this, note that if we start from a codeword and make to be able to encode N way and a 0 for a cup the wrong way up. With 1 1 1 1 1 1 1 1 t errors or less, then we cannot end up at another different possibilities as this view point, the student is presented with an codeword. This means that we can detect that an ergure 2: corresponding array of bits, with one N](where Figure 2: Corresponding array [log2flip. [x] means array of bits ror in transmission has takenwith an even case we of 1s in each place. In this number of bits, with one flip x rounded up to the next row and column. Let's say that a row or column say that C is t-error-detecting. Even though you can integer). rve that 2 errors cannot always be corrected identify that anwith an even number of 1s (cups the right way up) error has taken place, it may not be s procedure. If thecode letters in binary (`A' = 00001, fix the error. balanced. Inverting the cup corresponds If we magician sees two odd rows possible to is called d two odd columns c , c, he or she is notetc.), then we illustrate error detection with a simple ex- thus making sure `B' = 00010, `C' = 00011 to corrupting a single data bit, We will positions (r , run into problems flipped, or aample related to the cup row and Consider starting often c) and (r, c) were if there is noisy some game. column unbalanced. Noting Error s (r , c)channel where some of the wonder with all 32 words of length 5. This would lines is a way of correcting and (r, c) were flipped. We bits may get correction these unbalanced be sufficient , using a hexagonal array instance, consider strayencode the English letters, plus some punctuation. of cups (where there to corrupted. For the error. Translating the configuration of means that ections to check in rather than 2), it is possible CD single error detection, we can 1 into 0s check bit at we obtain the radio interference, a scratch on a For you know there Figure append a and 1s, ately correct twoetc. maybe more) errors. strings ROM, (or Notice that the Be- the end oferror word, creating array in Figure 2. Flipping the bit in was an each and you left words of length 6. The cussing the cup trick and hexagonal arrange- rule for the check bit is that the resulting weight must 00110 00001 01100 01100 and 00110 can work out what row 4, column 3 leads to the right array rther, we give a brief introduction to the theory be even. For instance, 00101 becomes 001010 after 01001 01100 01100 differ in a single bit was there before the in Figure 2. The unbalanced row 4 and -correcting codes. the error occurred but represent the words `FALL' and check bit (zero) is appended. As another example, column 3 identify the corrupted 10101 becomes 101011 bit. the check bit (one) is after `FILL' respectively. nology and facts for codes appended. We will think about the code you get conError-detecting and error-correcting Observe In this errors cannot rd of length n is a finite sequence of bits v = sisting of all words arising this way.that 2 code, there codeseitherdesigned code is a way to words, each of be correctedthe with this are 0 or 1. A to give are still 32 always , where each bit b i , is length 6, and distance n encode data some code we between any procedure. If words is at least two Codes words. The words belonging toso that if not too many bits pair of different the magician seestwo. get corrupted, it is We will (Verify this for odd rows r; r' and two odd columns ng about are usually called codewords. possible to detect that yourself: why cant two codewords, with of there has been an error in check bits included, he or she is one?) Consequently, concerned with codes having codewords all of transmission c; c' be at distancenot sure whether this type e length n. This is the length of better) to correct it. Note positions (r; c) and (r'; c')is not erError or (even the code. this code is 1-error detecting. To see that it were are used in Hamming weight of a word v, written wt (v), ror-correcting, flipped, thatpositions (r; c') and detection that error detection means that you imagine or we receive the codeword assigning ISBN umber of times the digit 1 occurs in v. been an error, but (r'; c) were tell whether it was a cormeans that know there has For 001101. There is no way to flipped. We wonder e, wt (110101) = 4 and wt (00000) = 0. what the errorof 101101, 011101,using a 001001, 001111 numbers to you can't tell ruption was whether, 000101, hexagonal there has been books and v be words of the same lengtherrorThe or 001100. whereas n. correction means array of cups (where there are 3 an error, butthem, denoted d (u, v) is you ng distance between that you know that there was directions to check in rather than

upted bit.

can't tell what the Addition out an error and you can work modulo 2 just meansis+ possibleand 0 + 0 = 1 + 1 = 0 correct 2), it 0 1 = 1 + 0 = 1 to accurately error was  what was there before the error two (or maybe more) errors. Before discussing the occurred. cup trick and hexagonal arrangement further, we Doing this requires that we don't use give a brief introduction to the theory of errorthe minimum possible number of bits to represent correcting codes.
1

the data. The aim is to design error correcting codes which are simple to use, don't involve adding too many extra bits to the strings, and are able to detect and correct most commonly occurring errors. Codes of this type are used in assigning ISBN numbers to books (so that if someone makes a mistake with one of the digits then computers can detect the error rather than bringing up the wrong book), recording data on CDs (so that if the CD gets some dirt on it then the music can still be read) and in computer memory (so that if an electronic glitch corrupts a memory location, 8

Terminology and facts for codes

A word of length n is a finite sequence of bits v = b1b2... bn, where each bit bi, is either 0 or 1. A code is a set of words. The words belonging to some code we are talking about are usually called codewords. We will only be concerned with codes having codewords all of the same length n. This is the length of the code. The Hamming weight of a word v, written wt(v), is the number of times the digit 1

with all 32 words of length 5. This would be sufficient to encode the English letters, plus some punctuation. For Let u and v be words of the The single error detection, we can same length n. The Hamming minimum append a check bit at the end distance between them, denoted distance of of each word, creating words d(u, v) is the number of positions in a linear code of length 6. The rule for the which u and v disagree. C is the smallest check bit is that the resulting If u and v are words weight of a non- weight must be even. For The of length n, we define zero codeword instance, minimum 00101 becomes u+ v to be the word obtained distance of a in C. 001010 after the check bit (zero) by componentwise addition code is important is appended. As another example, modulo1 2. For example, with respect to error 10101 becomes 101011 after the check 01101 + 11001 = 10100. detection and correction bit (one) is appended. We will think about the since it is the minimum You can check that number of errors that must code you get consisting of all words arising this d(u,v) = wt(u + v), occur in order to transform way. In this code, there are still 32 words, each because u + v has a of length 6, and the distance between any pair one codeword into 1 in precisely those of different words is at least two. (Verify this for another. positions in which u yourself: why can't two codewords, with check and v differ. For example, bits included, be at distance one?) Consequently, d(01101, 11001) = 2 = wt(01101 + 11001) = this code is 1-error detecting. To see that it is wt(10100). not error-correcting, imagine that we receive the It is not hard to see that Hamming distance codeword 001101. There is no way to tell whether is symmetric, i.e. d(u, v) = d(v,u), and it was a corruption of 101101, 011101, 000101, obeys the so-called triangle inequality: 001001, 001111 or 001100. d(u,w) d(u,v) + d(v,w). We use this later A stronger condition is that a code C be such without commenting on it - see if you can that whenever a codeword is subjected spot where. For a code C having at least two to a combination of t or fewer errors, codewords, the minimum distance of C is the we can guarantee to correctly recover smallest of the numbers d(v,w) over all pairs the codeword. For this, we require v,w of distinct codewords in C. For example, let the minimum distance of C to be at The C = {0000, 1010, 0111}. The minimum distance of least 2t + 1. The idea is that if t rule for C is 2. (or fewer) positions change in v, the check The minimum distance of a code is important the resulting `noise corrupted bit is that the with respect to error detection and correction since word' v is still closer to v than resulting weight it is the minimum number of errors that must occur to any other codeword. (The in order to transform one codeword into another. distance from v to v is at most must be even. A code C can detect all combinations of t or fewer t so the distance from v to any errors if and only if the minimum distance of C is other codeword is at least t + at least t + 1. To see this, note that if we start 1.) The way to decode a word that is from a codeword and make t errors or less, then we received across a noisy channel is just to replace cannot end up at another codeword. This means it by the closest codeword. As we've just seen, that we can detect that an error in transmission has provided the code has minimum distance at least taken place. In this case we say that C is t-error- 2t + 1 and no more than t bits change, we are detecting. Even though you can identify that an guaranteed to get back to the original word. error has taken place, it may not be possible to fix Thinking back to the cup game, we think of all the error. the rectangular arrays of 0s and 1s with an even We will illustrate error detection with a simple example related to the `cup game'. Consider starting Addition modulo 2 just means 0 + 1 = 1 + 0 = 1 and 0 + 0 = 1 + 1 = 0
1

occurs in v. For example, wt(110101) = 4 and wt(00000) = 0.

d(0,u) = w and 0 and u are both in the code, we see that the minimum distance of the code is less than or equal to w (i.e. m w). On the other hand, since m is the minimum distance, there must be words v and w in the code with d(v, w) = m. Since the code is linear, we must have v + w d C. We now have d(v, w) = wt(v + w) from earlier. However, we also have d(0, v + w) = wt(v + w) so that d(0, v + w) = m. It follows that C contains a codeword (v + w) with weight m so that the minimum weight of a non-zero codeword is less than or equal to m (i.e. w m). This shows that w = m. It turns out that the rectangular cup code that we described before is a linear code. We can see this as follows: suppose u and v are two codewords (so that each has an even number of 1s in each row and column). Let's set w = u + v and let's think about any row of w. Suppose that u had 2s 1s in the row and v had 2t 1s in the row. Let's also assume that there are exactly r places where both u and v have 1s (r can be an even number or an odd number). Now, thinking about how binary addition works, we see that w has a 1 in a spot exactly when u had a 1 there and v had a 0 or vice versa. There are 2s - r places in the row where u has a 1 and v has a 0, and 2t - r places v had a 1 and u has a 0. This means that the total number of 1s in the chosen row in w is (2s - r) + (2t - r) = 2(s + t - r) - an even number. This argument works for every row and every column so that we see that w is in the code. It follows that the code is linear. We now use this together with Theorem 1 to work out the minimum distance of the code. This is the same thing as the minimum weight of a non-zero element. To be a non-zero element, it must contain a row with at least two 1s in (remember that rows have to contain an even number of 1s). The columns containing the 1s need an even number of 1s in them, so there must be at least one further 1 in each of these columns. This shows that the minimum weight must be at least 4. In fact the minimum weight is exactly 4 as we can think of the configuration with 1s in all four corners and 0s everywhere else. This proves that the minimum distance is 4 as we claimed before. Every linear code has a dimension, which 10

http://twistedpencil.com/

number of 1s in each row and column as the set of codewords in a code (we could write them out as a long string instead of as a rectangular array if we wanted). We pointed out before that the rectangular cup code is 1-error-correcting. It turns out that the minimum distance of the code is 4 (we will see this below) so that the code is 3-errordetecting and 1-error-correcting.

Linear Codes
A code C is called a linear code if u + v is in C whenever both u and v are in C. Observe that u and v are allowed to be the same! In this case, u + u = 000 ... 0 is in C. Therefore, linear codes always contain the zero word, which we usually abbreviate simply as 0. Structure like this makes linear codes an important branch of coding theory. For example, rather than checking all pairs of words to find the minimum distance, things are a bit simpler for linear codes. Theorem 1: The minimum distance of a linear code C is the smallest weight of a non-zero codeword in C. Our strategy to prove this will be the following: if we let m stand for the minimum distance of the code and w stand for the smallest weight of a non-zero codeword, we will show first that m w and secondly that w m. Of course this proves that they are equal. This method for showing that two numbers are equal (prove that the first is no bigger than the second; then prove that the second is no bigger than the first) is used all the time in higher mathematics. Proof. Let u be a word in C of weight w. Since

For example, our earlier code C = {000, 101, 110, 011} has information rate 2/3. At one extreme, we have the code {000 ... 0,111...1} with one basis element, and information rate 1/n. At Figure 3: 2-hexagon with associated the other extreme, we have the binary word code consisting of all words of Consider the following process length n, having rate n/n = 1. But the first code applied to a linear code C. First, set S = . We now keep repeating has distance n (and can detect n- 1 errors), while the following step until the second code has no error detection at all, since span(S) = C: Notice that its minimum distance is 1. Both of these codes span(S) is a subset of C. If are rather useless. Coding theory aims for nice span(S) C we add compromises between rate and distance. we now describe. Given a set of words S, the linear span of S, span(S) is defined to be the set of all words that can be obtained as a sum of elements of S (counting the 0 word as the sum of none of the elements of S). any element of C that is not in span(S) to S. When we have finished doing this we end up with a set S with span(S) = C. The set S is then called a basis for C. For further reading on coding theory, consult the web or reference [1].

4 Hex-arrays and correcting two errors

By an n-hexagon, we mean a pattern of cups (or data positions) arranged so that each interior cup is surrounded by six neighbouring cups forming a regular hexagon, and the boundary cups trace out a regular hexagon with n cups per side. For instance, a 1-hexagon is just a single cup, and a 2-hexagon has seven cups as in Figure 3 (a). The translation from cups to binary words is illustrated in Figures 3 (b) and (c). The array for a 5-hexagon is shown in Figure 4. By the way, it is a fun exercise to get a formula for the number of cups in an n-hexagon. Notice that there are several geometric differences from the rectangular array. First, the rows go in three directions: horizontal, northwest, and northeast. Also, the number of cups in a line is not constant: it depends on how close the line is to the centre. Figure 5: A weight 6 codeword However, we can still insist that a line is balanced if the number of cups in it which are the right way up is even. Going back to the magic trick, this time on a hexagonal array, the argument we gave earlier shows that the set of codewords (the arrangements that are balanced in each of the three directions)

Figure 4: 5-hexagon

For example, suppose our code is C = {000, 101, 110, 011}. Start with S = and add 110 to S. Now span(S) = {000, 110}. We add 011 to S so that S = {110, 011}. Now span(S) = {000, 110, 011, 101} = C so that {110, 011} is a basis for C. Observe that all four words in C are found by summing either none, one of, or both of these basis words. In general, the sum of any subset of the basis words gives a unique codeword in C. Theorem 2: Suppose there are k codewords in a basis for C. Then C has 2k codewords. Corollary: Every basis for C has the same number of codewords. In other words, the order in which words are crossed off in the process does not affect the number of remaining elements. This common number of basis elements for a linear code is what we mean by its dimension. If a linear code with dimension k and length n is used, we say it has information rate k/n. Thus, in the Example above, C has a dimension 2. This is because if the dimension of the code is k then there are 2k codewords which means that sending a single codeword carries the same amount of information as k bits. Since the actual length of the word is n bits the information rate is k useful bits per n bits sent. 11

is a linear code. In fact, the dimension of this code is the number of cups in an (n-1)-hexagon see below for more on this. More importantly though, we want to find the minimum Figure 6: A 3-hexagon of cups, distance in the code. By balanced in every line Theorem 1, this is the same as the minimum weight of a non-zero configuration, balanced in every line. One possible configuration is shown in Figure 5. Check that all lines are balanced, so that the minimum weight in this code is no more than 6. On the other hand, we can see that the minimum weight is at least 6 as follows. In order to be a nonzero configuration, there must be a horizontal line which is not all 0s. If there is just one horizontal line that is not all 0s then we're in trouble because a diagonal line that goes through one of the 1s would have just a single 1 on it (an odd number). This shows that there must be at least two horizontal lines with 1s on them. Similarly for the northwest and northeast directions. If there are 3 or more lines in one of the directions with 1s on then since each line contains at least two 1s, we'd have at least six 1s in total. The only possibility to rule out is that there are exactly two lines in each of the directions and each line has exactly two 1s so that the total number of 1s in the configuration is 4. Considering the two diagonal directions first, the only possibility would be that the 1s form a parallelogram with 1s at the corners and edges parallel to the two diagonal directions. It is then easy to see that the top and bottom of the parallelogram are in lines with a single 1. This shows that the minimum weight is at least 6 and so the minimum distance of the code is exactly 6. According to our observations from before, this means that we can detect that up to 5 errors have been made (but not which errors they are), but since 6 2 x 2 + 1, we can correct any 2 errors! Finally we should do what a good magician should never do and tell you how to perform the magic trick in this case. First you need to set up a hexagonal array of cups so that in each direction there are an even number of cups the right way up. This isn't quite as easy as it sounds 12

but quite a good way to do it is to put the cups in an (n-1)-hexagon in any arrangement at all. It turns out that if you then try to extend the configuration to an n-hexagon, there's always exactly one way Figure 7: Two flips on different lines. to do it. Another way to proceed is to start with all cups down, and repeatedly perform six flips as in Figure 5. (There is also a way to wrap around the boundary by flipping eight cups). Next you should invite your victim to turn over any two cups. As the magician you have to figure out which two cups were inverted. Look along the horizontal and diagonal rows and see which rows have odd numbers of upturned cups. There are 2 basic possibilities. In the first case (see Figure 7) there are two rows in each of the three directions that have odd numbers of upturned cups. In this case there will be exactly two spots where three lines cross. These spots are where the cups were inverted. In the second case (refer to Figure 8) two cups were inverted on the same line so that this line ended up with an even number of upturned cups. In this case there are two directions that have two unbalanced lines and in the third direction all Figure 8: Two flips in the same the lines are balanced. line. The unbalanced lines taken together trace out a parallelogram. One of the diagonals of the parallelogram is in the third direction. The two corners of the parallelogram on this line are the cups which were inverted.

REFERENCES:
Pless, V. 1982, Introduction to the Theory of Error-Correcting Code, John Wiley & Sons Inc.

watching her little brother Ivan University of (I) play in the water. It was only British Columbia knee-deep, and she was not worried about him at all. He always enjoyed clowning around, C grimacing and gesticulating. When he suddenly cried \Help!" Carole just waved at him, but when his head disappeared under water and F stayed there, she jumped up and ran to help.

By Klaus Carole (C) was sitting on the beach Hoechsmann

answer is: very little. For now, let us just say that Grant was ogling Judy through his binoculars. This whole story could be told entirely without them, except that Judy stubbornly insisted on being part of this excursion, and as she went so did Grant. We shall say more about them at the end of this tale.

Fred's real name was Fernando, and at school they D H G called him \Ferd the nerd" because he liked mathematics. When Carole had told him Fred (F) was impressed to about Snell's Law, his first see her veer toward Henry reaction was to pull out a I (H), almost stepping on piece of paper and scribble. Donna (D), instead of going \It's easy, Carole," he said, J straight for Ivan (see Figure \.. but, hm, yes, I am using 1). But, Carole was a certified derivatives. There's got to be life guard and knew that she another way: that Dutchman would get there faster by this Snell died in 1626 long little detour, and Fred knew before Newton was even born." Figure 1 why this was so. She was used to that kind of stunt from him, but Since she ran twice as fast on the sand as through this much detail astonished her. \How come you the splashing water, she had learned that the cosine remember that?" she wanted to know. \Ten years of the angle FDC should be twice as big as that of after Shakespeare," was his answer. HDI. The general rule was that these cosines should be in the same ratio as the respective speeds. The instructor had said that this was just a fact, called Snell's Law, and that you could not understand it unless you knew Calculus. The attentive reader might wonder what Grant (G) and Judy (J) are doing in the picture, and the
Figure 2

The next day, he was back with two neat drawings, the first of which is shown here in Figure 2. That was his style: only rarely did he use algebra with Carole because he knew it did not convince her. She had taken an \interdisciplinary" course, where she learned to speak about derivatives and integrals but not to work with them. Fred spent hours with her using reams of graph paper, until one day she exclaimed: \You mean, the steeper the curve, the faster it's moving away from the x-axis? And that steepness, is the derivative? Wow!" Their conversation about the new diagram was rather private, so we'll paraphrase it. If you go 13

back to the beach picture and imagine D moving from F to H, you'll easily see that CD would be increasing while DI would be decreasing. As The you slide your pencil-tip from left general to right along the horizontal rule was line in Figure 3, its distance that these from the upper green area is cosines should increasing (like CD), while be in the same ratio its distance from the lower as the respective gray area is decreasing (like speeds. DI). As this drawing is not to scale anyway, we might as well regard the lower distance as representing not DI itself but mDI, where m is a positive numerical factor reflecting the ratios of the speeds on sand and through water. Then CD + mDI would be represented by the vertical distance between the green and gray regions, right?

where their boundaries have the same steepness." Back to the beach scene. As D moves toward D, the triangle FDC gradually morphs into FDC, and we wish to compare the growth of the hypotenuse (CD to CD) with that of the side (FD to FD). If we mark D on CD so that CD always equals CD in length (i.e. the green triangle remains isosceles throughout), the changes in hypotenuse and side are represented by DD and DD, respectively. Now, as D slides toward D (in synchronisation with its twin D), the green base angles approach 90 degrees, and the quotient DD/DD approaches hold your breath the cosine of the included angle FDC. So, now you have all the ingredients to roll your own proof of Snell's Law. But how Grant and Judy fit into this ? Well, it so happens that in issue Number 7 [1] some graphics expert thought it would be pretty if the whole picture were inside a circle, with Donna in the centre and Ivan on the rim (where Judy is now). With CD = DI of equal length, CD + mDI was constant, and the argument was sunk. But who needs a theory when you have facts?

That distance is minimal at the place where they would just barely touch if you slid them together vertically. \Yes," said Carole, \the red points would be kissing." \And the chaperone could separate them with one straight thrust of her cane," Fred added. Exercise \They'd share the same tangent line, you mean," C Carole sighed, \(you are so romantic, If v and w are Carole's speed on land and water, Fred) if you don't wish to slide respectively, what is the total time it takes her to them together, let us say, it's reach Ivan? What is the meaning of the factor m? How does it show up in the segment FG?

D F D D

References
[1] Judith V. Grabiner, How Looking for the Best Explanations Revealed the Properties of Light, Pi in the Sky, September 2003, 20-22.

Figure 3

http://xkcd.com/

14

Reiner - I want you to imagine a triangle with the points labeled \Entertainment Businessman," \Creative Artist" and \Scientific Researcher" "Genius is 1% inspiration (researching \What is fun?") Where would and 99% you place yourself? perspiration. Least on the Scientific Researcher side and probably in the middle between the Entertainment Businessman and the Creative Artist - because it takes both. I think this is one of the main success criteria in game design: that you can handle both. There are lots of artists out there who are very good artistically, but they are chaotic and don't get themselves organized and if you organize yourself, but have no artistry (and game design is an art) then you are also kind of empty handed. Getting both of them combined in your life is the big challenge. Artists are sometimes called upon to protest with their art. Do you ever feel the urge to protest through your art? The honest answer is probably `No.' I think games are there to entertain people, and to bring enjoyment to people. I think there is no mission beyond that. However there are nuances: We have educational games where you engage kids emotionally so that they are open to learning.
C Ar reat tis ive t
t men rtain Ente ssman e Busin

Interview by Gordon Hamilton

Reiner Knizia has published hundreds of board games including eleven in the top 100 on www.boardgamegeek.com

Where are you on mean Thomas Edisons 99 and 1 Scale? is probably a

Number 2, Pages 375-394]. Do you think games are under-utilized in mathematics classrooms? It is difficult to say. I think

little bit extreme... you can teach math without I think you actually need both. You need to using games at all, so it's not have a good mind and you a necessity; there are many need to have the capability to do things, but then it very much different methods of exciting depends on discipline and kids and making them the willingness to put in the work to produce a really want to learn. I think the good product.

important thing in learning is that you have a positive emotional experience; If the emotions are not there the learning is very flat. If you can create positive emotions in the learning process you get by far the best results and this is where games can help a Scien t lot. If the learning Rese ific arche r content is almost in the background, almost unnoticeably presented, then I think we have an ideal situation. The learning needs to become fun. I think games are an ideal tool for this they can be used much more widely than today, but they are not the only tool. Reiner, to avoid becoming biased, you don't play other people's games. You're right - When you design games you have to take lots and lots of minute decisions. All these decisions have to be taken, and if you already know a decision from somebody else, it is much

Last year, a randomized controlled trial found some games improved mathematical ability [Child Development, March/April 2008, Volume 79, 15

And one of the great mentors was my professor with whom I did my thesis. He was a professor in Germany and he was a professor in America, and he was working in Analysis so naturally I drifted towards this. I learned a lot from him. With live Ernest As a Masters and PhD student respect to mathematics, but also how to approach Hemingway did you also protect your mind from problems generally. I could probably have worked in many fields, but I knew him from very early bias? stages in my studies so we kept working together. No I did not do that in mathematics I did my diploma with him and then my doctoral there were many things to learn and I had thesis with him. great fun in learning them and I didn't try to Apart from your mentor, who are you standing re-invent the wheel so I was quite happy to take in the knowledge from other people and because in on the shoulders of or is game design too young mathematics it is not about solving the problem for that? in a different way once it has been solved it's In game design I probably don't stand on solved it's about finding new knowledge and anybody's shoulders. I think I have very much the more you know, the more new knowledge you developed my own style and learned by doing and can create. I think game design is a very peculiar profession in Let's go back to some of your high school a way because you can't really study it. experiences with math. Can you think of an If you can't study it, and you don't play other instance where you thought \Mathematics is really people's games - how do you stay fresh? beautiful!" My teacher always said: \If you don't read your I remember looking at mastermind, and working daily newspaper you will always be uneducated out an algorithm: What do you need to ask, and and stupid." I have always been uneducated and when you have the answers what do you need to stupid because I've never read it regularly in my do to actually solve it? I tried to work out the life. (laugh-out-loud) I have a different philosophy. complete algorithm and therefore prove that you I'm not interested in day to day news which is could always do it in so many steps. I applied that relevant today to talk to people and next week approach to quite a number of logic puzzles so is totally irrelevant. I prefer to read books rather I was always fascinated by combinatorics how than newspapers because they have long term many different ways are there to do this or that. relevance. The most fascinating questions in mathematics for the general public are probably coming from number theory: \Okay, so I understand what a prime number is. So how do I know that there are an infinite number of prime numbers? How many pairs of prime numbers [x, x+2] are there?" You can ask very simple number questions which are unsolved in mathematics and people can really understand. That is rare most of the questions in mathematics today are complicated and you first need years to understand the question. It is very important to take in the world. Hemingway said so nicely \In order to write you have to live" and the same is true for game design. Games are a mirror of our times. We have a very hectic life and a small attention span - so the dynamic of the game has to be much much greater today than 50 years ago. Half an hour playing time is almost too long. There are a lot of things that are influenced by our times and I need to take these in if I am to remain successful. So, as well as reading, I stay in contact with people - particularly I was perplexed that you chose the area of with the younger age group. I learn what the kids mathematics that you did. You just named like and then I build that into my design process. combinatorics which was where I would have You`re quoted as saying \Other people steal my pegged you. Why not? ideas before I've ever had them." I think apart from finding a mission in life it is extremely important to find good mentors. 16 Yesabsolutely!

more difficult to think of a new one. Once you know how a light bulb works this is In the way to do lights and if you don't order know you might come up with a very to write, you have to different solution

area? Well then our some holes in there? Yes it is, because I do believe there normal understanding If you are lots of games in the universe (like of area starts to fail, dont read your daily mathematics) and I believe that lots but if the process still newspaper - you will always be of these games should be mine, and leads to a result we uneducated and belong to me (I'm kind of smiling when still have something stupid I say this)... and the panic always grabs tangible to talk about. me when I walk around a fair and I look What I worked on in at games and I say \hmmm yes that's one my thesis was a very I should have developed that's the one I should general mathematical PhD Mathematics have found - that should have been mine, I'm too process (which we late there and somebody else stole it." I mean this call the integral) to calculate the \area" for very Interview by Gordon Hamilton is meant to be funny, but there is a sense of urgency very crazy shapes where the normal processes that calculate area no longer work. that overcomes me in these fairs... Is this a big motivator?

What happens if I put

The Board Gaming Genius:

Reiner Knizia

You did Reiner want you to imagine a triangle it ;) There are the right times for- Icertain things with the For labelled One last when the technology becomes ripe. pointsexample, Entertainment quote: \Life offers so many good Businessman, Creative Artist and Scientic I will soon have a game on the market that works choices, you can never take them all, and that's Researcher (researching What is fun?) with transparent cards... I was very anxious to get good, because it makes life so rich." That's by Where would you place yourself? it out quickly... I might have been the first one to Reiner Knizia... Least on the Scientic Researcher side design a game with transparent cards, but it took Absolutely that is one of the absolute between the a while to get it to market and probably see the middlestatements I don't want to be bored I think I and now I in one or Entertainment Businessman and the Creative Reiner Knizia has two other products coming Artist which means I'm can only do one or two things in life really properly. publis out - because it takes both. I think this is board games only second or third. It's this urgency. When thecriteria in game many other routes I could have taken: including one of the main success There are 100 on www.boardg time is ripe you need to do design: that you canneed I enjoyed being a board member, and running it and then you handle both. There are lots of artists out there who are very to be fast. financial companies, and doing mathematical You haven't asked medont getmost difficult research I enjoyed all that, and it was a great the themselves organized and satisfaction, but I don't think in the end it was question, which was to explain my PhD thesis to but have no if you organize yourself, artistry (and game design is an art)my mission. then a high school audience.
Other you are also kind of empty handed.
good artistically, but they are chaotic and

Well - I subsequently looked at your thesis, combined in As a young person, you need to find your Getting both of them people steal mission in life and then to have the courage and thought that it would your too istough a be life the big challenge. my ideas before to follow it. I only found my mission when question! ...but go for it if you wish! Artists are sometimesIve ever had I got to 40, and I jumped and said \now I called upon ...There is this funny thing calledtheir art. Dothem to protest with an you ever integral that we use to calculate the area of through do games full time" I've never regretted it feel the urge to protest and I've found my niche. I know I miss out on your art? certain shapes. Of course we all know very nice Last year, you say \you can almost everything in life, but as a randomized controlled trial f formulas that tell us how to calculate the area of some games improved mathematical ab e anything, but you can't have everything." The honest answer is have specific shapes: The area of a rectangle... length ativ ist Development, March/April 2008, Volume probably No. I think e t Pages take Do you these times width; the area of triangle... lengthtotimesCr Its r lots of choices we have to375-394]. and if think games ar A games are there are rich choices and you have lots ofclassrooms? in mathematics alternatives height divided by two. It gets more people, and entertain difficult when to bring a circle but I think there great life. you want to calculate the area ofenjoyment to people. then it's a is no mission

beyond that. However there are nuances: We reflects into games. A good game And this also have using games at all, so its not a necessity there is a method there as well and then you ask, educational games where you engage kids emotionally me rich choices. I do not \What happens if I put some holes in there? Can I for me is one that givesdifferent methods of exciting kids and ma so that they are open to learning. to learn. I think want to have still have a process to calculate the area?" If I have want to find the least bad option. I the important thing in lea so have a good options. I want many positive emotional experience; twenty holes can I do it? If I are not there the learning is very at. If to sit there biting my finger proces have a thousand holes can positive emotions in the learning nailsbest results and this is where ga I do it? Yes because I can the and hoping that when it comes learning to me that lot. If the round content is almost in t still count them, but what if almost unnoticeably presented, then I th option is still there. I have uncountably many? ideal situation. The learning needs to bec Can I still calculate the

en t rtainm Ente essman Busin

It is difcult to say. I think you can teac

games are an ideal tool for this they ca more widely than today, but they are not

Higher Equations and the Laws the Laws Higher Derivative Derivative Equations andof Physics of Physics
60 St. George 60 St. George Street,M5S 1A7, ON M5S 1A7, Canada Street, Toronto, ON Toronto, Canada Email: barnaby@cita.utoronto.ca NeilNeil Barnaby Barnaby Email: barnaby@cita.utoronto.ca Canadian Institute Theoretical Astrophysics, University of Toronto for Theoretical Astrophysics, University of Toronto Canadian for Higher Derivative EquationsInstitute George Street, PhysicsON M5S 1A7, Canada and the Laws of Toronto, February 26, ON 60 St. 60 February 26, 2009 Toronto,2009 M5S 1A7, Canada St. George Street, Email: barnaby@cita.utoronto.ca Email: barnaby@cita.utoronto.ca Neil Barnaby February 2009 26, February 26, By 2009 Canadian dynamical for Theoreticalare described at any time tof at any time t is described by = function x = described The Institute The dynamical laws of physics are described is Toronto by the function x the laws of physics Astrophysics, University theoretical 60 St. George Street, Toronto, ON As a1A7, Canada x(t). Velocity is the rate of change of position Neil by dierential by dierential equations. M5Sx(t). Velocity is the rate of change of position equations. As a theoretical physicist,barnaby@cita.utoronto.ca respect to with respect to time a job is solving those is with big part of the job solving those time Barnaby Email: physicist, a big part of the Neil Barnaby Higher Derivative Equations and the Laws Toronto Higher Derivative Equations and of University of of Physics Canadian Institute Astrophysics, Astrophysics, Toronto Canadian Institute for Theoreticalfor Theoretical University the Laws of Physics Neil Barnaby

Canadian equations indynamical laws ofand,are described at any any time described by the function x = x = The The and,laws of physics sometimes, dynamical sometimes,physics are described at time t is t is described by the function equations dynamicalsettingsvarious settingsdescribed by dx The in various by up withphysics 2009 that might de- Institute dx Velocityvis the rate of change of (2) laws of new equations. As a theoretical x(t). are for position dierential equations , x(t).=Velocity is the= rate of change of position (2) by dierential that might de- a theoretical Theoretical , equations. As coming up withcoming February theoretical physicist, new equations a 26, v differential scribe nature in some unexplored regime. Forthose with respect to time (2) equations. Asa big part of the job is solving physicist, part of the job is solving those with respect to time dt dt scribe nature in physicist, a big regime. For some unexplored a big part ofequations is solving those equations in the Astrophysics and the theequations in various settings and, sometimes, job in acceleration is the rate of me, inventing various settings physics is me, inventing new theories ofnew theoriesthe and, sometimes, and acceleration change rate dx change of the physics is of and acceleration is the rate of is of of dx the , various physics are describednew new equations that might velocity sometimes, coming up might most comingwithwith the time tvelocity one the excitingup Of of equations that with decourse, at any one x = v = the dynamical laws ofsettingsofand, job.part course,job. Ofis described byde- functionchange of v = , velocity coming up (2) (2) most exciting part the new equations thatnature inwritingunexploredis theregime. change of position dv dt 2 xdt might some Velocity in some For nature regime. scribestart describedown equationsrate of For dv nature in some unexplored hapd 2x erential equations. cantajustdown equations hapscribe theoretical x(t). d cant just start As writing a= = (3) . (3) hazardly inventing new theories some these to play of physics dt2 st, a big unexploredjob is solving thoseinventing newto timerole the aanddt = acceleration is the ratechange of the the part ofand expect me,For me, some role in physics is in the and dt2 . isdt the regime. and expectwith respect theories is = acceleration me, these to play theories of inventing new the rate of of change of (3) hazardly nature.sometimes, of of job. job. course, velocity There part of physics are a mostcriteria a lot we criteriaOf Of Of ns in nature. There is mostmostexciting part of of thethat we must one velocity could replace the acceleration, a, various settings thelotexcitingare thatpartmust job. course, one Using (3) we and, of exciting the the Using (3) dx acceleration,2a, 2 Using replace could we cant justde-writing to come upequations could (3)by a the(2) replace dthedacceleration, a, in writing down with we up with new equations thatjust startattempts down equations new ,hap-(1) course, one can't just our start with down dv dv second derivativex x . = a second cant might use to guide ouruse to guide start writing new in (1) vby hap- in derivative derivative of =of position with (3) attempts to come up ofaposition with . 2 = a == position with respect (3) (1) in dtrole by a second physical theories.any candidate naturephysical theories. Forregime.and For thesetheseanysome role in respect to time. We can dt2 dt something in some unexploredhazardly and expect these to play some equations haphazardly For expect to play candidate hazardly and expect example, respect to time. We can also do something do dt dt also example, to We can force, something similar with new venting newplay law for role in law the physicslotarecriteriaiswe must time. Using we coulddo F . The particular form the of a nature.isThere are a acceleration that we must change of also with theparticular form nature. There of studied of similar with the of are lot forto new some physics must beaand of must be studied in similar F . Thethe could replace acceleration, a, a, a theories of physics nature. Therecriteria that the rate Using (3) (3) we replace the the acceleration, in use to guide attemptsits predictions dont newF in (1) by on second derivative of F depends on ourthat to come up withwith of force, F. The particular form of position with attempts to come up new force, detailthat we sure use xciting part to of criteria toto makeone velocityto of the job. Of thatits our course, predictions dont depends under derivative of position with use a lot make sure guide must detail in (1)systemsecond consideraby a a the system under consideraof In consideration. something theories. For knownF depends on the In conict with hap- For example, any candidate tion. respect time. force acting do do body already dv ust startguidewith what equations come is about na- this about na-d2 x respectsystem the We We can on some In general writing down is physicalwhat up withexample, any candidatethe general under can also also something physical theories. conict our attempts to known already to to time. general . a (3) for thisrole in physics the things studiedtodt2 thethe force acting on some. bodyparticular form a new law onephysics article, =want= in might depend onon some body x (formight depend of of must tion.Ibedt must In studied in force acting the position, The example similar the force, F y andture. Inphysical for one of the things I want to be expect thesearticle, aIn play new law of similar withwith force, F . The particular form new this toture. some For example, theories. article, might position, x (for detail is make Using sense the depend on theon depends on example example, restoring accomplish ato to give whata its itscould replace restoring force on spring),system under consideraits of x or There are a lot ofis to detailyouwe must ofsure thatone ofof whatdontthe acceleration,of athe system under consideracriteriafor to make sure that (3)things Ipredictions dont Fof Fthe position,the (for on the velocity, thata new law you its we want predictions force a depends a, accomplish give sense any candidate of restoring velocity, conict develop new theories of about na- of of spring), or the viscous oractingaon some body like toup withwith what is already known about vAs position with spring), drag onthe velocity, v try to new of already accomplish is physics. na- tion.general a force acting as some body uide our attempts to come new theories is physics.a second derivative tion. In In of the the force on force general to (for example body is like to try to developstudied what in (1) by known example the viscous drag as a body is physics mustconict In this detail onegivethe(for be ture. in a of v Beforecandidate article, equa- things I Ican to might depend on the viscous x (for example l theories. For we can tryanycome can withto come the withanewwant to also do through on the position, x means that, body example, to In we up try respecttoup time. senseequa- moved depend molasses). position, drag as a (for example the This (for example ture. its predictions of of whatmovedtothrough might something means that, this article, new to you things want molasses). This one We Before theoretical its like to make sure that need tions accomplishin to you you a existing what its restoring specifying the or we on the velocity, ew law of physics must be studiedto isgive give laws totheof what its even a physical velocity, sense force, similar a the sense of F . physicist, without moved through tions we conflictaccomplish isthe understand try even without The particular form a of a spring), orthe system, This need to with what is to existing with develop laws specifying the physical system, onmolasses). understand already restoring is force don'tthat its of physics. Perhaps the most famous example part of the (forforce of spring), like to to develop F depends on the systemwe can write means that,viscouswithout specifying try to develop new theories of physics. v considera- the even drag as a body is new big example o make physics. Perhaps thetry dont of example theories of sure predictions famous new theories of physics. v (for example the viscousin trueas a body is Newtons law drag dierential like to most under of known we can write Newtons law in true dierential nature. na- try physics. job with is Newtons about Before we can trytion. to come up the force isequa- on some through molasses). This means write what is alreadyis Newtons law In this In general with new acting those known Before we canto come up with new equa-solving moved bodymolasses). system, we can that, about form as through physical This means that, law the moved form existing laws 2 article, of the theweI needneed to understand the as equations in various without specifying tions want to understand the on the laws2 x x 1 example 1 we I might even d x = F law physical differential system, n this article, one one ofthingsthings to want todepend existingposition, sometimes, dxNewton's x, dx the true system, tions d even(for settings and, without specifying the in. physical (4) F = ma (1) (1) accomplish sense= physics. aPerhaps most famous spring), or= we can write m most famous example the x, dt2 law in dierential lish is to give you a is to of mayouits sense of ,what it's a example 2 we mnewvelocity,. Newtons(4) true true dierential of what coming up with F write Newtons law dt of F give , Perhaps the the force of physics. restoring dt on dtform can in might like to try tois Newtons law law (for example the equations that as a body is as develop is Newtons v try to develop new theories of usnew theories of physics. viscous dragany given choice of F we can nd a unique physics. form as that tellsof mass, a body of mass, m, accelerates choice form we can nd a unique how m, accelerates For any given For of in as dx 2 describe nature F that tells come up bodynew equa1 dx 2 d x e we can try to us how a can try to an applied force, F new with molasses). solution This means d x initial Beforetowe applied force, come up =through Sometimes the initial oncethat, = 2x(0)conditions x(0) = x0 (4) to F . moved with ,solution once unexplored conditions 1= =Fx x, . . the some (1) F m 0 dt x, (4) (4) into understandin response laws F =F ,ma . response an existing Sometimes ma (1)regime. e need equations we need to understand without specifying it and v(0) =system,provided. Note that only two the dt2 dt m dt even the equation when we are rst introduced toexisting (1) vthe physical vNote that only two 0 are when we are rst famous example introduced it and v(0) = 0 are provided. ics. Perhaps thephysics. Perhapstolikemost famous example law in true conditions are needed F weof F nd ato find most made to lookusequation (1)algebraic m, accelerates For For any given choice can we can we body dierential a relation, Newtons laws of look isthatthat tells thea body of mass, initial conditionsFor any any given choice of to nd a unique given choice of F we can tells us algebraica can write m, accelerates initial needed corresponding corresponding unique how how simple of mass, relation, are is made to like a simple ons law solution (4) initial conditions conditions form as equation! .To see unique solution isonce initial initial x(0)x(0) 0 x0 a is it really is but response to a toapplied force, the Sometimes the fact that once the the conditions2 = x = Newton's ain it really is an an applied force, fact Sometimes second order in second order in derivatives law in response dierentialsee F . F that (4) solution once the derivatives is 2x but dierential equation! To 1 d x = v (4) = 0 higher that dt ). (it this, when we thatrstthe mass d to equationx, dx= and v(0) = v0provided. provided. 2Note two suppose are the position equationmass (1) itcontains derivatives d2 ). are when position x of introduced 2 = (1)F derivatives 0no higher are provided. Note that only this, suppose that thewe are rst introduced to x ofcontains x(0)and.v(0)and 0v(0)that vnoareNote that only two that (it the (1) it dt F = ma , (1) 2 dt m only two initial conditions are needed corresponding is made to like like a simpledt algebraic relation, initial conditions are needed corresponding is made to looklook a simple algebraic relation, initial conditions are needed corresponding to to that tells mass, m, accelerates isFordierential choice of Fto the fact fact that isissecond order in derivatives us how a it really mass, m, accelerates To see nd a unique (4) second order in derivatives body of a any given equation! we can fact that (4) is second order in derivatives ls us how a body of but but it really is a dierential equation! To see the the that (4) 1 1 in applied force, Fthis,applied force, F.position xthe mass contains derivatives no higher that that dt22 ). response this,an suppose that position xthe initial (it (it contains = x0derivativeshigher that d2 ). d to . suppose that solution once of of the mass (it x(0) derivatives no no higher the Sometimes conditions contains onse to an Sometimes the dt2 when we are equation (1) it and equation (1) it e are rst introduced to first introduced to v(0) = v0 are provided. Note that only two Newton's law is not the only example of a to look is made to algebraic relation, initial conditions are needed corresponding to like a simple look like a simple algebraic relation, 1 1 differential equation in physics. There are the eally is but it really is a differential equation! (4) is see a dierential equation! To see the fact that To second order in derivatives d2 this, suppose x of the position x of derivatives Navier-Stokes equations that describe the dynamics ppose that the position that the mass (it contains the mass no higher that dt2 ).

at any time t is described by the function x = x(t). Velocity is the rate1 of change of position with respect to time

of fluids, the Maxwell equations that describe the laws of electro-magnetism, the diffusion equation that describes the propagation of heat through a

18

librium position as an approximation because namics of uids, the Maxwell equations that ten equation (5) by x(t). As long as x(t) is not too large, then the restoring force nonlinear describe the laws of electro-magnetism, the dif- at larger displacements there will be is proportional to the displacement fusion equation that describes the propagation corrections like x3 , x5 , etc. Newtons law of heat through a medium, the Einstein equa- for this system now takes the ,simple form (5) F k x(t) (5) tions that describe how space-time curves in 2x where k is a where k 2that a constant that characterizes the of the constant is characterizes the d Figure 1: A plot response to matter or energy, the Schrodinger + (6) 2 stiness ofisthe dtthe only = 0 , spring. a(Don'tinlet the minusbother you, it spring. x the the sign stiffness of(Dont let of aminus in this equation bother you, i this equation sign Newtons law example Newtons a equation that describes the law is notmechan- example ofisanot thisinonly example ofyou, sign it simply rereNewtons law is quantum only example of lawsign not the equation bother not the the only sign this There are botherit re- ussimply reminds always equation the you, it Often the force is you, Newtonsthe only example of a 1 equationin this equationyou, it simplysimply that in physics one woul bother in sign Newtonsmicroscopic system, ... Thedierential in this equation There are not dierential a caveat physics. you ical state dierentialisequation in physics. There Thereequation inushere usbotherthe is the isminds us that the force is always of a law list where areishave in physics.the force force higher that pulling the are There mindsdenedthatthat could obtain minds pulling the the the minds = k/m inalways to order pulling dierential equationThere are the mindsequationsthe forceforce thealwaysalways the mass x system. We in physics. order we us us taking the is always dy- mass total energy of a = dierential equation on! by decoupling Navier-Stokes mass back that describe isbydy- mass back towards = 0.) Th Navier-Stokes equations thatderivatives the 0.) generalthe could go on and on and in physics. that describe make equationsthatthattowards x or Thepulling have writ- writ- xback0.) We Navier-Stokes equations the dy-equations.massreferringtowardsthe most0.) We have theback towards xsimpler.to We havefun- back towards equation lookdescribe = x = We writ- energy: kinetic and pote Navier-Stokes equations that describeofthe dy- towards Im back 0.) that namics mass ten Maxwell equations that ten equation that ten of namicsa set uids, the Maxwell is =only INavier-Stokes of uids,uids, describeequationsof uids, equationequation (5) as not approximation because an approximat want to draw equations the Maxwell the dy- ofthat of the ten So (5)equations an trying toequation (5) as (5) approximat as We have written equation namics your attention to a Maxwell equations formulation. (6)x=0.)an approximation because (5) as an as damental namics ofMaxwellthe common solution that equation an approximation because displacements there will b ergy ten equation approximation the dif- at larger displacements namics all uids, dierent equations. Indescribethe laws of electro-magnetism, the dif- at larger displacements there will b of these the equationsall that the laws of an (5) as Im denitelybecause at larger is generically given by describe dif- higher order equations are there feature ofdescribe the laws of electro-magnetism, claim that at larger larger displacements useful in nonlinear be the the dif- electro-magnetism, never there will be nonlinear at displacements 5 describe the laws of electro-magnetism, larger displacements there will bewillcorrectionslike x33,, x5 , etc. N vpropagationcorrections like x3 x55 etc. Ne describe the of physics, the underlyingthe physics, only thatdescribes the propagationstand-in electro-magnetism, fusion equation thatthere they be 0 it as a corrections like x fusion at x(t) = describes + 3 ,sin(t) . 5 Newtons x, 1 2 where like appear known theories laws ofequationdescribes thedy-dif-equation correctionswillthelike xis3,, etc.nonlinear law law fusionfusion equation that that propagation that xcorrections nonlinear (7) 0 cos(t) x 3 fusionequations that describes thedescribesheat propagation etc. Newton's, law equa- , for this system now takes = mv equation contain at most two timethe through fundamental set, of Einstein for thisetc.law system now takes the simp propagation through a medium, thex5 etc. equa- for this now takes K the simp x x for corrections like x a more a lower-order equations.system of heat equa- for medium, thenow takes Newtons form Einstein the simple Newtons of the 2 namical of heat through a medium, the Einstein this system of heat throughthemedium, the that describe a for nowin this equation in the you, it simply reNewtons a isfundamen- that describe how this takes curves in not the only example systemsign system now takes of heat through a medium,lawofEinsteintionsEinstein equa- simple form the simple form simple form equa- for this of how space-time curves bother space-time tionscurves solution (7) describes a function that osderivatives when written in describe space-time The curves in tions that describe how how space-time in d x tions that terms d22xthe 2 dierential equation incurves in There are the minds d2 physics. to response 2 frequency d22 x tal tions that freedom.1how matter or energy, the Schrodingeror with x thexSchrodinger , is always pulling 22 + 2 x = 0 ,, degrees of describe Its space-time the Schrodinger in time energy, us thatthe= 0the be- 2 x force , just 2 response to matter kind of surprising cillatesmatter or energy, the Schrodinger or energy, response to matter (6) dt + x = 0 d response to 22 + dt + x (6) back Navier-Stokes the Schrodinger that describe response to matter equationsequations equation that describes the + x towards x = 0 , energy, mechandt that so many dierent or that describes theequation that the dy- massquantum dtthe end = 0.)(6)We have writ- (6) haviour mass = 0 2, equation that describes that quantum mechan- describes the quantum mechan- of a the dt2equation equation of uids, thedescribe state of a we expect asystem, on The an quantum where we have dened namicsregimes all share such state ofmechanMaxwell equation that describes the quantum mechan-equations that ten system, ...(5) as list approximation because = k/m ical microscopic ... list where ical spring!a This nature in ical state of a microscopic system, ... The list microscopichave dened The in Figure order to dened = k/m very dierent of http://brownsharpie.courtneygibbons.org/where we have = ical state the laws of electro-magnetism, the dif- where we illustrated k/m in will in order to make microscopic system, on andlist function is displacements there webe nonlinear ical state of mathematicalaproperty. ... The list describeonwhere largerhavedefined in =make the equation look simpler. a describe microscopic system, In fact, go ... The we haveat we have dened order to in order to could 1 for the special case where=mass is releasedk/m equation look simpler. T go on and on and on! and dened on! a fundamental theonEinsteinand on! on! that where make the equation lookk/m medium, could go on and equationscould could go and on make the simpler.5 The etc. that describes the to draw your correctionssimpler. 3 , simpler. general propagation make look look simpler. ,The Newtons law of could on and on equationon and general draw the equation a common this how go I wantIiswantfundamental that it Ito common orsoyour attentionequationislookx general Theequation (6) is commonalityfusionand on! inattention toiswant make the equationthe 0.to like x Thecheck general solution is v = tosocurves space-time draw your your attention from common thatattention toExercise: solution of equation (6) response Iawant atorestsolution of 0equation a common solution of matter (6)now takes the simple form to draw a medium, the toof all these equation equations. In all for this of your want to drawheat attention this seems to a common of all these equation (6) equation (6) solution of is feature In all really dierent (6) is solution often Ioverlooked. of At throughequationfeature Einstein equa-does solve system alsoallis x0 , feature all of all these these dierent equations. describes of dierent equations. In that energy, the Schrodinger v v0 feature rst glance dierentthat that (7) In of physics, the (6) and equations. tions that describe how known theories allin space-time curves v feature of all these dierent is there some- all theories of physics, the underlying dy- v equations. known In x(t) = x cos(t) + 0 sin( like the quantum conspiracy; of state of the microscopic a remarkable mechanical physics, a underlying dyv0 appear interpreted 0 2 0 known theories of physics, the underlying dy- correctly to= x0 cos(t)2+ dy-initial . x(t) = x00cos(t) + sin( x(t) be underlying as sin(t) (7) (7) v0 d x (7) knownphysics, matter or energy, the Schrodinger= x at most 0 cos(t) known theories higher-orderthe underlying dy- equations containcos(t) = xtwo time+ = sin(t) . of theories sin(t) (7) contain x(t) + two + (6) namical thing wrong withresponse to contain couldnamical equations velocity. at most dt2time . x 0 , position x(t) namical equations system,contain atlist at twoat quantum mechan- 0 The describes that most two time go on namical ... that equationsmost two time on written solution of fundamen- a function that oscillates a funct equations most thetime and and The in terms (7) describes The solution (7) describes contain derivatives when equation a fundamental derivatives when written in terms of fundamen- The solution (7) describes a functi namical equations prevents them and on! from when written in terms role of fundamen- solution (7) describes a function that osderivativesplaying of The derivatives of a 1 of in terms written talfundamen- solutionThe solutionof a describes cillates in time with function ical derivatives when state when microscopic system, ... freedom.11wherekind of(7) dened that os- thein order tofrequency , in terms describes surprising degrees of cillates inIts with surprising ,=just in that tal degrees of freedom.time kind frequency ,a justk/m be- osin Its with frequency behaviour we in physics? degreeswrittenof freedom.1fundamen-of The The list (7) timewe have function cillatesthe time with frequency , tal tal degrees of freedom. Its kind of surprising kind surprising could go to consider Itsex- so cillatesain timemake mass on describe the , we expect a mass talanswer this freedom.1 on and on of surprising many dierentcillates in time with frequencybe- just This function on th degrees of question, letsdraw youron! so many dierent equations that describe of a spring! the be- a mass on th with the equation look simpler. The general frequency , just haviour I want Its kind and that describe haviour equations that massend haviour we aexpect attention to that To an that that so many many dierent equations that describe expect aexpect a thea mass end of we on the thatIso dierent equations nature to very want focus a common haviour we expect the spring! This of a that so many commonequations that attention in very dierent regimes of inshare such end on theThis function is illustrate dierent feature a describe regimes share such spring! end plicit example. nature inweto draw yourallnature inshare dierent solution mass Figure(6)for theaspecialfunction is illustrate natureSuppose dierent on of massthese different we spring! a allequation is illustrated in Figure where in very very dierent regimes all haviour is expect function on illustrated in Figure case regimes allm share such such illustratedall function 1 is of spring! This This is nature around feature ofregimes alldierent equations. of This function is illustrated in 1 for the special case where mass very on a spring. In denoteathe dierent all these share fundamental mathematical property. In fact, Figure a such fact, In all equations. We all known In spring!1 for the is property. In fact, so is the =0. bouncing ain In a fundamental mathematicalfundamental mathematical released from rest 1 for released fundamental mathematical property.theories fact, mass special case where where vthat v0special case where mass mass mass 0 the special mass is the a fundamental known theories ofunderlying commonality is so fundamental that it is released isso that v00 = 0. Exer mathematical property. equi- dynamical dy- 1 for case x(t) xcase is from rest so In fact,underlying where all the mass away fromphysics, property. In this rest released physics, the the this that 1 is from rest displacement ofthis commonality fundamental commonality isspecial so that v=thatcos(t) from sin(t) .that (7) = 0. Exer this commonality is so is so fundamentalit for the is so fundamental =0 v it= 0.+ Exercise: checkv 0. Exercise: check 0 that rest sofrom rest = 0.this seems that (7) really does solve (6) and a ittime that v0 so that 0 known namical equations at this position by x(t). As long as x(t) thatat isoverlooked. At rst glance so fundamental isoften most two it from Exercise: check Exercise librium commonality isequationsrst containglanceseemstwo At rst glance this seems that (7) really does solve (6) and a not most often often overlooked. containoften overlooked. overlooked. At that really does (6) and also theories derivatives when written in terms of fundamen- that (7)solve (6) and also (6) a x that xthat x0osAt glance this remarkable conspiracy;reallysolve some- thatappear correctly to be interpret rstlike a this seems (7)The solution doesdescribes and, also 0 ,that , often overlooked. restoring force is thislike a remarkablereally does is there some- v00 appear correctly to be interpret At rst glance proporseems that (7) conspiracy; is there solve v function (7) 0 too large,like a like a remarkable conspiracy; is there sometime derivatives when written remarkable conspiracy; is there some- v0 appear correctly to be interpreted as initial initial of then the physics, appear equationsto frequency , and the becorrectly that position as 0cillates be tal conspiracy; is thereIts kind of surprising vCheckto in time with be interpretedjust velocity. like to the displacement of freedom.1 thing wrongappearhigher-orderthatinterpreted asdoes solve (6) and also a remarkabledegrees some- v0 with correctly thing wrong with higher-order equationsreally position and velocity. that initial underlying tional thething thing wrong higher-order of fundamental wrong with with equations that that position and velocity.(7) in terms that many dierent equations that describe position fundamental a thing wrong with higher-order higher-order equations from playing a fundamental role mass on the end of a equations that them from playing a and velocity. haviour prevents 1 dynamical equations sofrom playing a fundamentalposition and velocity. we expect role prevents them prevents them them, from playingfreedom.roleIt's that x0, v0 appear correctly to be interpreted as Fnature indegrees of regimes all share such spring! This function is illustrated in Figure k very dierent in physics? (5) prevents x(t) a fundamentalfundamental role a role prevents at mostfrom playing contain them two in physics? in physics? position and in when physics? kind of surprising that so initialfor consider velocity. time derivativesa fundamental mathematical answer In question, the special an exin physics? constant that characterizes the property.thisfact, plotletsconsider an ex- where mass is released case To where k is To answer this question, lets consider an exa To answer this question, letsthe solution (7). Figure 1: A 1 of written in terms ofanswer this question, letsex-equations it is many different To this commonality is sominus consider an Suppose we focus on a mass m= fundamental thatexTo answer this question, lets considerplicit example. Supposefrom focus on a mass 0m 0. Exercise: check an example. we rest so that v stiness of the example. (Dont let the plicit mass m spring. plicit degrees Suppose we focus on a fundamental plicit example. Suppose wenature in mass m that describe bouncing around on a that very focus seems spring. inWe like to know (6) plicit of freedom. Suppose we focus on arst glance around physics one(7)We denote one would like to know the example.often overlooked. At mass mon athis in on a Often would denote the the and also that x0 , spring. really does the bouncing the 1 physics solve bouncing around on a spring. higher There is a caveat here that you could regimes all We Often the bouncingremarkableobtain We denotethereofsome- v0 appear correctly to spring. like a different Wedecouplingtheis denote the total energy of a areequi- kinds bouncing around on a aroundoron conspiracy;share energy ofmass away from the equi- be There are two kinds of spring. by a denote total suchthe mass away Therethe two interpreted as initial displacement displacement a system. from system. order equations by taking derivatives displacement of the of the away from the equi- of mass a fundamental librium position that position as velocity. mathematical thing wrong higher-order of energy: kinetic As potential. is not displacementdisplacement away from the equi- position by x(t). andlong as x(t) Kinetic enof referring with mass away from the equia set of equations. Imthe massonly to the mostlibriumequations by x(t). As longandx(t) is not funenergy: kinetic and potential. Kinetic energy is librium position by x(t). As long as x(t) is not librium position by not x(t)toofundamental role property.In fact, Im them from playingtonotas is is the restoring by so damental formulation. bythis commonality As isa large,x(t) generically given force is proporprevents denitely x(t). istoo fundamental librium position So x(t). As long as trying long ergy then not restoring force is proportoo large, large, the are never useful inislarge, then thegenerically given by then then the restoring force is proporrestoring force proporclaim that it is often the restoring At first glance this seems order equations that higher too physics? in too large, then overlooked. force is propor-to the displacement tional tional tional tional to appear it is as a stand-in to the displacement 1 to the displacement physics, only that where they the displacement tional a remarkableofconspiracy; is there something ex- K = oscillator the poten-(8) to the displacement like fundamentalTo answer this question, lets consider an harmonic mv2 . restoring force (5) with our mod (8) for a more set lower-order equations. For our simple k x(t) ,, (5) 2 F k x(t) , x(t) ,focuspreventsF k x(t) (5) (5) oscillator the poten- restoring force (5 plicit example.FSuppose wethat energymassFor our simple harmonic(5) on a isF m wrong with higher-order ,equations tial law (12) we force F k x(t) k (5) For our simple harmonic oscillator the poten- restoringnd (5) with our the plot (12) where where k denote tial Figure 1: potential solutio themwhere where ak constant that athat characterizes the For that characterizes the oscillator (12)A plot of thewe nd from playing is fundamentalspring. Weis a constantenergy characterizes the a a constant role 2 kphysics? the our is plot harmonic in tial a Figure 1: Awelaw of the solutio kbouncing around on characterizesisthe constant that simple of the solution (7). is Figure law Figure the solution where k is a constant that of the massstiness of energy isenergy 1:2 A let theplot of the solution (7). d4 x nd 2 x characterizes the fromthe spring. A(Dont2 let A minus(7). 1 plot of 1: the minus d is Figure 1: displacement away the equistiness of the spring. (Dont 24 of the (9) g + dxx stiness spring. (Dont minus the (Dont an Tostinesslibriumof the spring. thelet 1the minusminusU = m x . U = 1 m 2 x2 . Often in physics onedt2 + like= stinessanswer this question, let's consider let as x(t) is not 2 1 of the spring. (Dontby x(t). As long explicit 2x 4 x would like position let (9)dt4 one would g 4 t Often in physics d 1 There is a caveatOften in you couldone would like to know the d herethat you could obtain higher that physics 2 obtain higher 1 is a caveat here Often 2 There 1 a caveat focus on a massTherehigher is dt example. Suppose we here that you could obtainbouncing in physics one would like towould like to knowathe m 2 (9) g + 1 There is a then the obtain could you equations propor- derivatives or xby.decoupling total energy of system. There too large, caveat here restoringm force Oftentaking U = ofin physics one is higher order equations energy is given a or by sum of know the equation is not (9)dt2 + ar There is aequations bythat you could that or higher obtainbytotal energy 2 bysystem. There total energy of a system. There ar are two order caveat here taking derivatives order decoupling by takingtotal energy thedecoupling the This kinds kinds dt4 much hard by The by decoupling derivatives of a system. There are two total around on ordertaking derivatives or derivativesof equations. Im referring only to the most fun-twoof energy: ofkineticThis equation is equations displacement set or a order equations ational to Im by taking by decouplingequations. Im referring only to the mostare by ofkinds byspring. We denote the the mosttotal energy of a system. There fun- the sum kinetic and potential. a displacement The total energy is given energy: the and potential. a set of equations. thereferring only to set of fun(6). potentialtotal energy: kinetic and potential. Kinetic enand energies notis a setIm equations.only toequilibriumThe most of kinetic energy: potential.sum by the sum ofequationsolution looks of referring Im referringdamentalthe energy: The total energy theandto of the This general is not much h fun- kinetic and kinetic given potential. The the potential aof the mass formulation. So the denitelyonly to formulation.potentialdenitelyby trying Kinetic en- generically (6). The general set of damental away from Im the most fun- position energy is and kinetictrying to ergy is generically given by equations. damental formulation. SoofIm denitely not energies ergy is Kinetic enof given by not trying to ergy So Im given given by is genericallynever useful in by damental formulation. So k denitely not trying to equations are never given damental formulation. as order equationsIm trying, that ergy theand equationsenergies useful in (6). The general solution B claim that higher Im denitely tooclaim to potential order kineticgiven by never then by x(t). As long So higher is F not x(t)arethat higher orderand kinetic1 energies x(t) order equations useful in is genericallyis generically not are claim never useful in ergy 2 are 2 2 large, higher (5) 1 x(t) = A+ cos(+ t) + lo claim that + claim that higher order equations are never it physics,stand-in where= mv + it m x stand-in 1 useful in onlythat where they appear is as a physics, E that 1 1 (10) physics, only is proportional to the displacement U they appear itis as a stand-in the 2solution (7). that where constant that a only= K + is as x(t) where k is athey appear appearcharacterizes the 2of lower-order1: 1 .plot of restoring that 2 Figure equations. physics, only that where they stand-inis fundamental set of lower-order equations.. it K (10)22 = A 2 for equations. a more fundamental physics,for a more where they appear it is as a for a more as a stand-in setE = K + K = A 2 + m x2 . only force (10) K =cos( . + B+ mv 2 (8)+ (8)= 2 mv t). + t) + 1 U 12 K2 = 1 mv 2 . A A+mv 1 fundamental set of lower-order x(t) = 2 cos( forstinessfundamental set of lower-order equations. + a more of the spring. (Dont . (8) = 2 mv 2 . E=K + A for a more fundamental set of lower-order equations. let the minusU = Kmv + m x 2 2 (10) 2 right one would like to know+ A cos( t) + 2 is2in physics expresTo check that this really that this really is the Often 1 To check the where we expression,the freq dened There is a caveat here that you could obtain higher check that this really is 2the right righthave the 2 sion,equations that denedenergy thea right expres- conserved. That is, have d notice that notice that Eisdefined by (10) is are two kinds E total by is system. 2 To this really (10)of conserved.There expres- where we 1 There is a caveat here that you could taking higher orderToby decoupling order equations by obtain derivatives or check sion, notice that E dened by (10) is conserved. 2 2 where we That is, most take the derivative youll nd by taking derivatives or by decoupling a set of equations. Imtheif youfun- of energy: kinetic and potential. Kinetichave dened the f a set of equations. Im referring only sion, notice that E dened by (10) is conserved. find that ento referif you take the derivative you'llyoull = 1 1 1 1 That is, if you take the derivative nd ring only to the most fundamental formulation. So denitely not not Im definitely trying to damental formulation. So Im that 2g 1 = ergy That usefulyou take is generically given by trying to claim that higher order equations are never useful neveris, if that dE the derivative youll nd claim that higher order equations are in physin = =0 (11) ics, only that where they appear it is where they appear it isfundaas a stand-in for a more as a stand-in (11) 1 dE that physics, only that dt dE =01 (11) (8) 2g 2 The solution contains four arbi mental set of lower-order equations. . for a more fundamental set of lower-order equations. dt K = 2 mv A , B rather The solution con =0 than two becaus implying that E is a constant in time. Exer-(11) The solution contains four a dt B implying is conserved by taking time. in time derivatives. t 19E really that E is a constant in order Exer- A , thanrather T cise: check that E is a constant in time. Exer- A , B rather two bec implying that 2 check that E really is conserved by be related to the initial c cise: could taking order in time d

0 restoring force not with modied . For our energy is U = mthe sumthing wrong with higher-order + our+ arbitrary constants 4 equations to solve than and (9) The total simple harmonic by= x . the poten- This equation is g(5)contains fourthat 0position (13) velocity. 2 given2 oscillator of the (11) The solution 4much harder x = Newtons 2 dt 1 lawfromUB nddt xthanlooksbecause (12) is fourth g d x + d x + 2 x = 0 . (12) playing a 2 . dttwo like we rather tial energy is energies prevents themThe generalm 2fundamental role (9) (6). A , = solution potential and kinetic 1 2 implying dt2 1 energy is ag constant in time. of the This equation is not much harder to solve than dt4 1 E is . (15) sum Exer = The totalthat 4 2given by thein physics? derivatives. These constants 2g check that E really2is conserved by taking order ind4time d2 x 1 cise: x sin( + cos( t) + 2 go (6). = begiven byto + B an through (13) potential = 1 mv 2 energies 2 . The U E m 1 m (9) implying and kinetic a 2constant time.total x(t) The Arelatedconsiderx =of conditions x(0), + 2 the looks t) that (10) EtheK + Uthat= 2is + and. using in To answer this question,4letsneed to initial .the This equation is not much harder to = derivative of (10) x 2 x equation (6). energy is general solution + ex-like could g don't + the sum 0 + dt 2 2 olution contains four arbitrary constants plicit example. v(0), etc.effortwe on+ B our point. The general solution looks like Suppose dt cos( dont mass to got) ,(14) The we focus t) a need m through that (6). + A potential and kinetic energies make sin( to Now, lets see what would happen if we de1 1 is fourth 2bouncing x(t) But cos( harder + solve than 2 rather than two because (12) Exercise 2 of the around on a spring. +We point. The plotted (14) is Thiseort to solution (14) + the solution in equation= not much + t) isBto sin(+ t) is A our denote The total energy really is by right x is make To check E = K + U = 2 mvthe the sum . that modify given + 2 m expres- (10) to this some we mass away the frequencies dened in time cided and kineticNewtons laws by addingEwhere+ U havein Figure22witht)g 2B 0.01 andx(t) = A+ cos(+ t) + B+ sin( derivatives. These constants displacement of The= 1 Figure1cos(looks = + + m (6). the general solution the equipotential that E dened by (10) is conserved. = K plotted mv 2 A from2 x2 . + like sin(andaarepenergies (10) t) ,(14) sion, notice derivative terms. This might sound like higher initial conditions x(0), librium position by x(t). As long aswith is not 2 2 be related Check that thisreally is conservedexpres- resentative choice ofx(t) to the initial conditions. ComThat To funny thing thedo (after the right by that where we representative choice of initial + A cos( t) + B sin( is, check that E really is all, derivative youll know taking the nd 1 a if you take go through that A+ dened the + tc. But derivative U = to dened m(10)weconserved. that restoring is1 the t)1+ Bfrequencies we dont + that 1 mv 2 + using x2 tooTo check x(t) thishavecos(Figure 4 2 g . + t) need to E and 1 by 2 equation (6). expresof is large, then the = 2g to+ right 1. sin( (15) verify this conditions. propor. (10) E=K plot Exercise: that sion, notice law(10) perfectly ne!) but its a pare = really force isCompare this plotwe have dened the frequenci Newtons where 2solution derivative youll nd displacement to makeThat point. you worksthe 2(14) is tional tonotice that E + A cos(t) + sin( t) ,(14) our is, if Thetake the that (14) solves equationB dE is conserved. 1 dened by (10) (13). 2 Figure 1. worthwhile exercise. and a rep- can sion, a lot learn =0 (11) 2 = 0.01 Often, we 4 g . to= four 1 1 constants (15) d in Figure 2 with g dt The F take x(t) , (5) To that that this really of nature takeThat form solution contains2g arbitrarynd check why the laws is the right expres-is, if you kthe derivative youll Figure1 A plot of the 1: about initial conditions.would happen if we decidedhave dened the frequencies the 1 1 4 2 g = dE Now,that Esee what (10) is conserved. where we ative choice that let's a constant in time. Exer- A , B rather than two because (12) is fourth of E is Comthat (11) implying do by sion,they notice dened by 0 Exercise solution (7). 2g = restor For our simple harmonic oscillator the poten- (10) wi Figure 1: hiscise: to modifyEimagining howverifybywhere ksome constantdE contains four arbitrary constants Athe totalthe solution (7 plot check that 1.really is conserved adding beis a The time that characterizes theconstants to Figure Newton's laws by would order higher derivatives. These Exercise: things taking dif- in solution dt derivative youll nd noting that plot of energy law (1 That is, if if they took some other form. So, of the spring. (Dont let the minus2 you take the 1 0 tial energy is = (11) stiness lets 14)the derivative of (13). andThis might time. Exer- a A , related the 1 1 4 g(12)The conserved if the motion of x(t) i solvesferent that E is a using equation (6). like equation two fourth = derivative (10) initial conditions is(15) implying terms. constant by the following be B rather than thatbecause . x(0), solution contains four arbitrary that boldly consider replacing (4) in sound could funny dt to Verify (14) solves be in physics one would like to kn equation (13). 2g 1 order is that you could obtaingo throughOftenB 2rather than two because (12 inawe dont needtime. Exer- constants 2 (13) rather than (6). time derivatives.higher These that equation A1 There is a thingcheck whatdE all, we know that taking that law cise: to do (after really is conserved by Newton's here by, Now, lets see that Ewould happen if weimplying caveatEBut constant in to de- v(0), etc. equation =0 U = m (9) These order (11) eortcould a derivatives or thedecoupling constants in x . taking point. The solution order expression for the dt laws by adding some (6).The to make related to four initial conditions x(0), of a system. There are tw works perfectly equations by solution ouris conserved by taking(14) energy time derivatives. kinetic e fine!) but it'sEbe contains by arbitrary total is cise: check that really a2 new cidedthe modify Newtons and using equationequations. Im referring only to the most funto derivative of (10) a set of plotted in Figure than two because (12)through thatrelatedandthe initial condi potential. Kine go and derivation A , B (10) To dont2 = 0.01 (6).ofa repNow, letsterms.constant inhappen theweformulation.we Im 2 with equation wrongenergy: kinetic with we implying that d4 x what worthwhile exercise. Often,of etc. But we see notneed toto is could equationto(13), the This e is higher derivativeEsee a d2This might time. Exer- de- v(0), rather denitelygwhat's energy fourthbe by is pretty complicated a x would sound if derivative So and using 1 dx like trying is given the dont need to ergy (14) g 4 + 2 = laws x, damentalsome eort what would happen if we solutionis generically given by F by adding resentative to make our The conditions.constants ofistheoretical machinery go th , Now, lets see choicederivatives. total de- Com- etc. But we sum of (12) point. The v(0), initial These cided to that doreally isall, learn a lot that higher order equations are never useful the energy. It turns out in our beyo modify Newtonsm we know taking order in time of to compute in a lot conserved dt that by about why the need a cise: check to E (after can funny thing dt claim dt (6). T 2 potential and kinetic energies x(0), physics, pare where related laws initial 1 cided some eort thehigher derivative terms.usingofmight its take couldform to Figureis 1. addingtheory and a tolevel. Let gives2the derivative of (10) andlaws Newtons law works perfectly This equation (6). to thatplotted inhigher thewith Exercise: verify rep-makestillme simply tellsolut ne!) nature aonly modify Newtonsto derivative = 0.01 equation (9) our point. The you but sound like this plot Figure it by a gconditions school the be they appear 2 as stand-in more di2 mv . 2 v(0), how potential initial awhere g is a what would happenby we de- that resentative equation need to however, the simple K = with (8) funny see Often,they all, weif a imagining etc. set of lower-order(13). conditions.that2 worthwhile exercise. to do (after of natureahavingderivative terms. This might soundthrough1 Com- 1 expression g = 0.01 we do for higher fundamental But choice of equations. lot Now, letsthingnew constantcan learn know that (14) solves we dont energy,go like plotted in Figure 2 2 x( . (10) E 1. solution= mv + m 2 x of initial conditio = K + U (14)verify choice dx d3 x mg d2 plot Figure Newtons of works things adding some a0 pare make ourto kinetic energy is resentative funny thing about mensionslaw secondsperfectly ne!) form its=eort to this for theall, weThe Exercise: is K valid.mg easiest cidedwhymodify Newtonssquared. wouldbut gdifferenttoifdo (after point. know that no 2 to the laws of nature take If we a be laws by the take longer 2 = The (14) solves with equation (13). 2 to dt 3 then obviously how This might sound other plottedSo, Figure 2A plot of2 butsolution a will worthwhile exercise. things would usual lot ne!) = 0.01 noting this they do by imagining this reproducescanbeNewtons law works perfectlysee g theis its and pare that plot total energy Exerc difhigher derivative terms. Often, took the learn aforce that in noting that the that energy (10)rep- no longer Figure 1. 2 they we some like form. Figure 2: to To check that a (14). is the right dta way 1850 Russian d m total by this really the expres(14) solves 2 law (4). took do (after nature know that resentative Often, we can conditions. ComFor laws of all, we So, lets form aboutthing togsome other form. takeexpect that exercise.choice of initial learn a lot that is describedequation (13).where why the suciently small we worthwhile the ferent if they a funny m dx of able gradski conserved nothe motion conserved if theismotion was if let's boldly consider replacingbe (10) willsion, longerthatof x(t) be Exercise: dened they do by works perfectlythings butabout why this by ofto Figure 1.notice form E verify by (10)+ conserved. . laws equation take the boldlyequation (12) will give basically would beadif-equationseex(t) is described by than (6). So we rather thandt consider imagining how by thethe same pare Newtons law replacing (4) by ne!)following pre- theTo plotwhats wrong with equation (13), we need eral theorem [1]. O its 2 (4) the following that (14) solves nature (13) ratheryou take the derivative youll (6). equation (13) Figure 2: A plot of the That nd equation is, if dictions as like a lot a lets in ferent they took some other how things energy. kinetic equation if(14). (4). Often, weseemslearn So,decent imagining compute the(13). be dif-turns energy. The fundamental laws worthwhile exercise. So (12) canform. they do by need to newwe need would theIt a So expressionnew expression out the kinetic energy. for solutionof the solution (14). a for thattheory equation (9) still new expression for kin gure 2: boldly consider for our rst attempt at modify- our higher derivative A plot point replacing (4) by starting some other pretty complicated and involves der dierential eq about why the laws of nature take the ferent if they tookderivation is form. So, lets Using this form thefollowing dE The derivation is pretty complicated and involves 2x 4x gives the potential the following beyond 0 high kinetic energy if the = equation laws of nature. dx boldly +U theying d imagining how things would be dif- consider replacing (4) by energy, however, the total energy E = K(11) will bes do the + d = 1 F x, by a lot of theoretical machinery dt simg 4 with2 equation (13), we, (12) see whats wrong dt even simpler, lets focus again ple expression (8) for the kinetic energy is no (12) a lot of theoretical machinery the beyond (13) to describe th the high The Tothey took some otherdt dtmake life m equation ferent if form. So, lets Let Let that tell you use equation rem precludes , B 4 to compute the energy. d2 xturns out indx Using the longerschool level. implying meto E isthat theby thetime. Exer- Athe a constant school level. me simply see you is answer: answer: be. valid. The easiest way simply tell you in it tive equations pla on our g d + It = 1 by the mass-on-a-spring example. , of x(t), just as should boldly considerxreplacing (4) F x, following where new is a new nature having di- (12) d4 x d2 x constant of nature having gher derivative theory equationm stilldt 4 2 2 cise: check that E really is conserved by taking order where g is a dtconstant of (9) 1 dx dt2 3 equation physics. g =0 then 2 = K =the derivative of (10) d x using equation (6).complica + F x, mg dx d x mg Aside from being very Moreove , (12) dimensions however, If simthemensions of seconds seconds the we take g = take dt4 potential energy,of squared. squared. If we 0 and (16) could dt m dt 3 2 something deeply troublingofabou thethe usual force law (4). having di- 3 dt lets see dt cessful law v(0), e phy pressionobviouslythisdreproduces of dx (8) for is newreproduces usual 2 Now, dt 2 if we then where g 4thethis x constant isnature force obviously a kinetic energy no Figure 2: A plot of the solution (14).what would happenenergyde- be ne d x 1 (16): 2 has invol g g 4ofway 2 = small x,we expect(12)= 0equation +seconds squared. expecttake that is a new constant of naturem modify Newtonsthe kinetic to date can eort F we by we thatg g , where mensions For dt suciently m valid. (4). For sufficiently small is If having cided to dx dilaws by adding some law The easiest dt to see that dt . gardless of the sign of g we can alwa (16) derivatives. + then will give basically the same usual force of see Figure 2: A plot we the solution terms. obviously this reproduces the predictions seconds wrong If of take dt = 0 (14). g (12)(12) will give basically the same mensionsTo as (4). squared. with equation (13), we This might sound like plotte higher derivative equation pre2 whats some function x(t) that makes this law (4). new like a decent having obviously this then funny turns (after 2: we know that resent dictions g is(4). Sosuciently small starting di- needfor our reproducesathe usual forceoutless than all, A plot Ostrogradksis s So as a seems (12) seems like we expect that where (12) For g constant of naturea decentpoint to compute the energy. Itthing to doinFigure zero. This of the solution unlike any Using thiswe with equation (13), we energy new expression for energy very pare Using this new expect that forstill kineticne!) alsoisits a deep. t expression(9) kinetic equation for at give basically laws = nature. suciently small Newtons law but starting point seconds rst attemptthe modify-(4). For gTo see whats wrong equation works perfectly to in physics. At rst g mensions of (12)our squared. If we the same 0 our higher derivative theory g of first attempt will modifying at takelaw preenergy E=K+U we arein we can looke E = K It willwill used dictions as this decent the to 2:the totalworthwhile exercise.be see conserved compute the however, turns To be whats learn a have that Often, ing the obviously (4). reproduces the usualaforce gives need potentialplot of the solution the sim- conserved howerwise lotsense of( then laws of nature. So (12) seems like equation (12) will the total energyenergy. +U(14).atout obvious if wrong with equatio Figuregive basically the same preTo make life even simpler, let's modify-again higher Ayou use theory to describe the dynamics andmake serves focus our onyou use energy, equation (13) (9) describe the to it also if derivative kinetic decent laws of to compute the energy. It tu toall nature take the form equation (13) energy is no still dictions expression (8) seems about a the need n- restoring forceg(5) withour rst attempt at that pleas (4). So (12)for the like whyequation netic energies. In physics, negative lawstarting point for our small we expect (4). life suciently modied Newtons To makeFor even simpler, lets focus again our the laws (5) with example. Using mass-on-a-spring ify theory restoringnd of nature. our modied Newtons restoringourx(t), just they do by imagining we things force oten-equation (12) will give basically the samethe the point see the dynamicsas itx(t), just asisit should be. would be dif- equatio of equation ingmass-on-a-spring example. Usingstarting Tovalid.of potential energy, modify- (13), how gives whats easiest way at see that our higher derivative the laws of na for The wrong withshould be. the simrst attempt to however, by law (12) on our we (5) with restoring force(5) with our modied Newtons pre- longer we Aside from being very complicated, theres isinterpreted as s are force as we nd even modifiedlike a decent laws of expression (8) for theifkinetictook ergies no potential energy, however our simpler, lets focus the need tonature. Newton'sagain (12) compute the ferent they energy inthe usually So, lets law ple not widely app lawTo (12) ing gives some dictions make life 2 Aside energy.being very instability.2 form.Often, from It turns out is other complicated, law (12) we nd (4). Sox(12) seems find point for d rst2 attempt at modify-make longer valid. The easiest way again replacingby by Physically this The something lets focus to see(9) still equation physicists. is un deeply consider thatexpression (8) for the kinetic en troubling ple is (4) about d4 x on our To our higher derivative theory equation simpler, boldly following starting g mass-on-a-spring = 0 . ) we can + our + example. Using3 the life even there's something deeply troubling themass bobbing aro (13) we know that a theorem d4 x d2 x x (16): example. however, be negative! The easiest is to see kinetic Using can longer 4 xdt4 nature. (9)ing the laws ofd2 xdt2 2 2 + 2 x = 0 . on our mass-on-a-springthe energy, energy thethe sim- valid. Re- learn a lot way so fun g + (13)gives the potential equation (16): the kinetic d about equation dt g 4 + dt4 + x = 0 . (13) about granted gardless of thesign of g we can end of a spring is forcompletely stab always imagine awhy the that no To make life dt2 much harder to solve than even simpler, lets focus again ple expression (8) forthe be negative! Regardless lawstonature takeabout th 3 (13) energy can kinetic4energy is Its important of be clear dt e This equation is not e 2 this some function x(t) that x that x byexpression ond order in deriv makes is 1 d dx is is not f theonThis equationsolution looks like harder the solve than The easiest way todsee + 3 our equation not muchmuch (6). The general much harder harder to solve to longer valid. of the sign of g we canthe problemthe formequation (13) is Thismass-on-a-spring example. Using than =systemx, withto they(12) F that , askdo by This equation is general solution looks likethan to solve why. less than zero. This isdt4 unlikedt2 always imagining how things any m dt (6). The not (6). The solution looks like looks like general solution perfectly imagine to in physics. At rst glace its not beYou might nd some function x(t) that well-dened mathematica (6). The general A cos( t) + B sin( t) would different we are used x(t) = + + + + 3 ) However, if they took some where g make sense of than it ki- an acceptable x(t) = A+ cos(+ t) + B+ sin(+ t) (14)at makes this expression constant of nature having di- to m all obvious how tois a new less negative isnt rst attempt law (10) x(t) = + +A cos( t)B+B sin( t) ,(14) A cos(+ t) + + sin(+ t) because it mensions of seconds squared. predicts g = 0 zero. This In physics, negative kineticIfen- othersomething - na is unlike any system we takeform. unsuccessful. But netic energies. + A cos( t) + B sin( t) ,(14) s+ havecos( t) the frequencies ,(14) A dened + B sin( t) reproduces that we then obviously in as signalingenergy - that is not of ph are used to thistive kinetic an new force ergies are usually interpreted physics. At the usual theories comp F where we pres- where we have defined the frequencies 2 law (4). For gall obvious small to make sense So natu the world we see around us. we a d. suciently how we expect that where we have dened the frequencies first glace it's not at instability. Physically this is unacceptable: patible with rved. we have dened frequencies where d 1 the equation (12) will give In our the same preTo weof negativea kinetic energies. basicallyrst attempt any new know that mass bobbing discard physics, perform to modif around on the negative 2g . (15) = 1 1 1 4 nd (15) dictions are (4). Sostable seems likeare decent need a as usually interpreted as signaling (12) system. a still (13) as 2g 1 1 4 2 g . end of a spring is a completely law and regard equation useful be (15) = kinetic energies 1 1 2g 1 4 2 g . (15) = But what about ture complicated starting point about rst attempt moreattempts. h 2 ) Its important to be clearfor our exactly what at modify- ourIt this 2g The The solution contains four arbitrary constants A, an instability. Physically Whatis unacceptable: we (11) solution contains four arbitrary constants ing the laws of (13) is. if we added even Ostrograd gives the problem with equationnature. This is a hole to more deriv The solutionthan two four arbitrary fourth B rather contains because (12) constants is - A , B rather than two because constants fourth order inperfectly well-dened mathematical equation. focus this higher derm 2 The solution contains than arbitrary(12)(12) is fourth four two because is To make life even simpler, lets to add instability again ple ex We didnt really see , B derivatives. These constants could time rather Exer-B Arather than two because These constants be related 2 g order in time derivatives. (12) is fourth A , We didnt really see this instabilityin our toyphysics toy modeldidnt inclu However, it on our mass-on-a-springof itself3in our Using the longer isnt an acceptable manifest model because we running i law example. without order in time conditions aking in bethe derivatives.initial conditions constants But we because we didnt include nonlinearterms like x3 in equation (5). to could time initialderivatives. These x(0), etc. order could related to the theThese x(0), v(0), x(0), constants because it predicts something -terms like nega-equation (5). it requ namely x in However, be we dont (6). beetc. But to the initial conditions x(0), v(0), relatedrelated to needinitial through that to go conditions ecould v(0), etc. But we dont tive kinetic energy - that is not compatible with mathematical obj 3 to go through e de- etc. to make dontpoint. toneedsolution that isthat 20 e eort But we our need The through (14) v(0), eort to make our point. The solution (14) is go the world we see around us. So we are forced to with innitely m some plotted in Figure 2 with g 2 = 0.01 and a rep-

know that a mass bobbing around on the end of a spring is a completely stable system. It's important to be clear In physics, about exactly what the problem negative with equation (13) is. This kinetic is a perfectly well-defined energies are usually interpreted mathematical equation. as signaling an However, it isn't an instability. Physically this is unacceptable: we know that acceptable law of physics a mass bobbing around because it predicts on the end of a spring something - namely negative is a completely stable kinetic energy - that is not system.

experiments. Failed theories are still useful because they help to guide future attempts. It turns out that there is a loophole to Failed Ostrogradski's theorem that allows us theories to add higher derivatives to the laws are still useful because of nature without running into trouble they help to with instabilities. However, it requires guide future attempts. recourse to some exotic mathematical objects (for example equations with infinitely many derivatives) and a complete discussion is beyond the scope of this article.

References
Ac. St.

compatible with the world we [1] M. Ostrogradksi, Mem. see around us. So we are forced Petersbourg VI 4 385 (1850). to discard our first attempt to modify Newton's law and regard equation (13) as a dead end. But what about more complicated equations? What if we added even more derivatives? In 1850 a Ostrogradski showed that Russian mathematician named Ostrogradski any fundamental was able to prove an amazingly general laws of physics theorem [1]. Ostrogradski showed that any involving N-th order fundamental laws of physics involving differential equations th will lead to negative This n order differential equations will kinetic energy if powerful lead to negative kinetic energy if N > 2. theorem n > 2. This powerful theorem precludes the precludes the possibility of higher possibility of higher derivative equations derivative equations playing any playing any fundamental fundamental role in physics. role in physics. Moreover, it Moreover, it explains why every explains why every successful successful law of physics that law of physics that has been discovered to date has has been discovered to date has involved no more than involved no more than two time two time derivatives derivatives. Ostrogradksi's theorem is very general and also very deep. It explains what might otherwise have looked like a strange coincidence and it also serves as a guide for trying to modify the laws of nature. However, this theorem is not widely appreciated by most practicing physicists. The reason, I think, is that this theorem is so fundamental. Most of us take it for granted that the laws of nature must be second order in derivatives without ever stopping to ask why. You might find it a little depressing that our first attempt to modify the laws of physics was unsuccessful. But don't despair! Indeed, many new theories of physics are found to be incompatible with nature even without the need to perform any new 21

http://brownsharpie.courtneygibbons.org/

Domenico di Michelino, La Divina Commedia di Dante (Dante and the Divine Comedy). Fresco in the nave of the Duomo of Florence, Italy - the terrace purgatorio lies over his right shoulder

Since his death in 1968, George By Gamow has been living in purgatory Gordon in a seven floor terraced apartment Hamilton building with penthouse garden Masters Academy on top. Scummy tenants reside on & College the lowest floor, but the residents become increasingly sophisticated as one moves up. George's apartment is on the third floor.

likely to be on any floor. It goes at a constant speed when moving, and when it has no requests, it just waits where it is. When the elevator is going past a floor, souls like George can jump on or off without altering the final destination of the elevator.

In his 1958 book, Puzzle Math, George showed the impulsive answer of is wrong. He argued that the elevator's last stop before visiting his In Each day George wakes up, dresses, eats, Julian floor was equally likely to be 1 or 2 (below), locks-up, walks down the corridor to the Havils or 4, 5, 6 or 7 (above) so elevator and impetuously pushes both the Impossible? the elevator is twice as Paradiso (reviewed in up and down buttons. When the elevator Volume 12) Georges likely to have descended arrives he compulsively steps in original calculations from a sweet smelling before checking whether it's ascending are extended to include from the nicotine miasma of the lower multiple elevators. Although floor than to have 7 he does not state it, Julians ascended from the floors or descending from the perfumed multiple elevators are stench below. That 6 fragrances of the upper floors. assumed stationary was great news. It 5 when George pushes If it's ascending from the lower floors, the buttons. meant that George George gets a nasty elevator ride, and is off could expect to have a 4 to a bad start in the day. George bad elevator ride only 1/3rd of 3 Gamows What goes up must come down - so the first the time. impulse is to think that George is going to have Sure enough, when George 2 a bad start to the day about the time. arrived in purgatory in 1968, 1 Is that right or wrong? Before we try to before the population explosion, solve the problem, we need to understand he confirmed that his day-toexactly how the elevator operates: day elevator rides were a nasty

...the elevator almost always appeared as it was going down, and so the journey [up to see his friend] would be a frustrating down and then up .. .

experience only about 1/3rd of the Inferno The elevator goes time. up and down picking up But over the last few years, and dropping off souls George has realized that something is on different floors. The wonderfully wrong. He is now getting elevator's destination bad elevator experiences a lot less is always randomly frequently than 1 in 3 and his luck determined - is especially good during morning Julian Havil in Impossible? so when the rush-hour. Find an explanation for (reviewed in Pi in elevator is at George's statistical joy. (Spoiler alert: the Sky - #12) George Gamov rest, it is equally answer on the next page.)
1904 1968

22

George finally figures out that before, when he pushed the buttons, the elevator was at rest. The probability was 1/3rd of a bad elevator experience. Now, durThe only thing elevators are moving compared to ing rush-hour, the that is different now when George 1968 is thatbuttons. Does that make a difference? The pushes the there are more souls after the population explosion. The with a thing different between rushelevator travels only constant speed, so all we need hour is to find all possible origin-destinationare more to do and non-rush-hour is that there pairs and souls using the elevator at rush-hour. Using these find out how much of each is below Georges floor.

two clues, George finally figures out that before, when he pushed the buttons, the elevator was at rest. The probability was 1/3rd of a bad elevator experience. Now, during rush-hour, the elevators are moving when George pushes the buttons. Does where 0 p 1. He quickly figured Inferno that make a difference? The elevator travels with out that his probability of having a 0 a constant speed, so all we need to do is to find all smelly elevator in the morning is: p if he presses the buttons when the elevator is possible origin-destination pairs and find out how stationary. p if he presses the buttons when the elevator much of each is below George's floor. is stationary. The total length beneath George p22(3-2p) ififhehe presses the buttons when the presses the buttons when the p (3-2p) Gamow's is 16 floor heights out of a total of elevator is moving. George used integration to used integration 21 + 15 + 10 + 6 + 3 + 1 = 56 floor heights. That elevator is moving. Georgealso have used the to find find this, but he might limit but is, the probability that a moving elevator this,of thehe might also have used the limit of the ratio above: Paradiso is below George when he presses the ratio above: Probability (elevator below) probability (elevator below) buttons is 2/7 which is less than 1/3. That is why, as more and more souls 7 = lim (pF - 1) (pF) (pF+1) + (F - pF) (pF - 1) (pF) moved in and started increasing F 6 2 elevator usage George got The total length beneath George Gamows is 16 floor (F - 1) (F) (F + 1) heights out of a total of luckier and luckier 56 floor 21+15+10+6+3+1 = as his 6 probability of getting heights. That is, the probability that a moving elevator a elevator is below George when he presses smelly the buttons is 2/7 George 3 which is less than 1/3. That is why, as morefrom more = lim (pF-1) (pF) (pF+1) + 3 (F-pF) (pF-1) (pF) dropped and 1/3 Gamows F (F - 1) (F) (F + 1) to 2/7. souls moved in and started increasing elevator usage
1
George got luckier and luckier as In probability of his general, if getting a smelly elevator droppedGeorge isto 2/7. from 1/3 on floor In general, if George is on floorf inin building with F f a a building floors: The total length of possible floor-to-floor comwith F floors: mutes is a tetrahedral number: (F-1)(F)(F+1)/6 and the The total length total length of these below Georges floor is another of possible floortetrahedral number (f-1)(f)(f+1)/6 plus a triangularto-floor commutes is a tetrahedral number: prism number (F-f)(f-1)(f)/2.

an average lifespan), the transcendental administration decided last month to scrap the1outdated terParadiso raced landscaping that Dante visthe transcendental administration ited, and replace it decided last month to scrap the with a modern outdated terraced apartment complex that has each landscaping that soul allocated to their own unique Dante visited, and replace it with floor. a modern apartment complex that Earlier this own has each soul allocated to theirweek, George learned George p that he will be allocated floor p Gamows unique floor. where 0p1. He quickly figured Earlier this week, George learned out that his probability of having a that he will be allocated floorinp morning is: smelly elevator the

= (pF) =

+ 3(F - pF) (pF)2 (F)3

Inferno

p2 (3-2p)

(F - 1)(F)(F + 1)/6 and the total length of these below George's floor is another tetrahedral number (f - 1)(f)(f + 1)/6 plus a \triangularprism" number 1 (F - f)(f - 1)(f)/2. probability (elevator below) Probability (elevator below)

In Julian Havils Impossible? (reviewed in Volume 12) Georges original calculations are extended to include multiple elevators. Although he does not state it, Julians multiple elevators are assumed stationary when George pushes the buttons.

(f - 1) (f) (f + 1) + (F - f) (f - 1) (f) 6 2 (F - 1) (F) (F + 1) 6

In an attempt to deal with the purgatory population In an attempt to deal with the purgatory explosion (which lags behind the population explosion (which lags behind the earth's earths population explosion by populationan average lifespan), the transcenexplosion by an average lifespan), dental administration decided last 23 month to scrap the outdated terraced landscaping that Dante vis-

http://twistedpencil.com/

became very common in the Edo period, and We usually think of By were usually painted on wooden shingles in ink. mathematics as something that Fukagawa The most common images were lucky animals, is largely independent of an Hidetoshi and such as horses, but the images associated with individual culture. However, Tony Rothman Princeton University Press when a society is isolated Reviewed by ISBN 978-0-691-12745-3 mathematical problems, called sangaku, filled the dual role of showing reverence to the god and from the rest of the world, Tom Archibald Simon Fraser demonstrating the mathematical prowess of the mathematical thought will University individual who made the offering. The problems continue and sometimes were also collected in books and apparently were develops in a direction that is independent of what is happening elsewhere. widely popular, and there are also manuscripts That is exactly what took place in Japan in the containing such problems that have not been Edo period, when the feudal government of the found posted at temples. The problems range from Tokugawa clan decreed that Japan should be easy to very difficult. closed to the outside world. This situation started Hidetoshi and Rothman here present a in the early 1600s and was to continue until selection of the problems, graded the country was opened up more or according to difficulty, with less forcibly to the west, notably hints and solutions. The with the Meiji restoration of 1868. aim is to give the flavour During that long period, wasan, of these works, and this is Japanese mathematics, went its quite successful. High school own way. This has been very or beginning university students nicely discussed in an excellent interested in problems can read and book by Annick Horiuchi, but work on these with profit, since the methods for the you have to read French. 1 most part don't go beyond elementary calculus. The present book, in The traditional solutions are often surprising. English, also discusses this The very first of the ninety problems cited states: period, but rather than looking \There are 50 chickens and rabbits. The total at research mathematics, number of feet is 122. How many chickens and it presents a remarkable and how many rabbits are there?" Most students intriguing fad: the posting of now would approach this with algebra, but the mathematical problems, particularly in geometry, traditional solution begins: \If rabbits were on pieces of wood outside temples as a kind of chickens the total number of feet would be 100 offering. This is the \sacred mathematics" of the ..." A problem that has a more Japanese flavour book's title. Offerings of images at Japanese temples involves drawing figures on a folding fan. Unfolded, the Japanese fan is a sector of a circular annulus 1 or ring. Suppose the sector has radius R, and draw A. Horiuchi, les mathmatiques japonaises lpoque dEdo. Paris: Vrin,
1994.

24

tablets. This doesn't detract from the interest and two circles of the same radius r that touch overall usefulness of the volume, especially for the one another and also the outer and inner student reader, and it should serve to stimulate edges of the fan. we draw a chord length d further reading on the mathematics of these to the outer edge tangent to the tops of the cultures. The illustrations, some in full colour, will two circles, and add a small circle radius also entice the reader. t tangent to the chord and the fan, as There shown. Given d and the diameter of the Students and teachers will read and work on are 50 small circle, we are invited to find the the problems in this book with pleasure. They chickens and diameter of the large circles. The give an interesting glimpse of another rabbits. The proposer gives the answer (3.025 culture, one in which the value of total number of if d=3.62438 and 42t=3.025) mathematical problem solving merits but not the method. Indeed, we feet is 122. How many a kind of religious recognition. An chickens and how Most don't know his method, since the inscription on one of the sangaku many rabbits are students answer is incorrect. The authors gives a hint of why: \Confucius now would there? don't provide an explanation for approach this says, you should devote all of with algebra, the error; nor do they suggest any your time to study, forgetting to but the traditional reasons for these rather surprising values. solution begins: If have meals and going without But they do give a solution, involving constructing rabbits were chickens sleep. His words are precious the total number of some auxiliary lines and using the Pythagorean to us. Since I was a boy I have feet would be 100 ... theorem. been studying mathematics The authors also give some basic historical background and examples of the Chinese sources of Japanese mathematical work. Hence the book gives a nice glimpse of East Asian mathematics of several centuries past. Not all of the historical details are as carefully written as would be ideal, for example on the subject of the language of the and have read many books on mathematics. When I had questions, I visited and asked mathematician Ono Eijya. I appreciate my Master's teachings. For his kindness, I will hang a sangaku in this temple." These days, such recognitions are unusual.

Faber and Faber ISBN-13 978-0571205110

By Apostolas Doxiadis

Reviewed by Clara Park

Goldbach's conjecture is easy to state: \Every even number greater than 2 is the sum of two primes."

4=2+2 6=3+3 8=3+5 10=3+7=5+5 12=5+7 14=3+11=7+7 We can conclude easily that this theorem is

CheongShim International Academy

This statement seems to be straightforward as we can attempt proving it straight away. 25

The main character, Petros Papachristos, is a mathematical genius but considered a complete failure in life. Recklessly, he abandoned a successful career in academics, shutting himself up away from family and friends in order to devote his every waking hour to proving Goldbach's conjecture, a theorem that mathematicians have been unable to solve for almost three centuries. As foolish as any of us might believe such an action might be, through the eyes of an understanding nephew, we, or background in Thanks to today's too, soon come to understand Petros' passion and mathematics. technology, mathematicians realize he is deserving of our respect. We empathize discovered, with the help of with Petros' bleak hope of solving the problem, supercomputers, that the conjecture holds true and we are just as exhilarated as Petros' nephew for all even numbers up to 400,000,000,000,000 when he learns that his uncle has finally cracked (four hundred trillion). Faced with the reality that the secret. \numbers are infinite," however, they were once The story is not about problem solving. It is again depressed since they are still not able to rather a story of challenge, persistence, and provide a general proof. devotion. Petros' obsession with the problem suggests that the result is not the only thing Books on mathematics are often `howthat matters in life. to' guides on problem-solving or earnest treatises on some area of mathematics, I strongly recommend this book to other The story is not but Apostolas Doxiadis' Uncle Petros readers. It will appeal to almost any about problem and Goldbach's Conjecture takes the student, even students with no previous form of a fictional story that can solving. It is rather interest or background in mathematics. a story of challenge, entertain and inform a wide range of They will be stimulated and entertained persistence, and readers regardless of their mathematical as well as delighting simply in the spirit devotion. knowledge. Readers may feel the same of determination of all those people, sort of enjoyment as watching a movie, including mathematicians, who risk all in such as Forrest Gump. attempting to do the seemingly impossible.

plausible! But, seemingly, this simple theorem, known as Goldbach's Conjecture, has remained one of the most tormenting problems in It the field of mathematics, along will appeal to with Fermat's Last Theorem. almost any For centuries, many talented student, even mathematicians were enticed students with no to this myth but were unable previous interest to prove the theorem.

Math Challange Winner Announced

Obtaining solutions to four Math Challenge Problems posed in Issue #11 (Spring 2008) proved challenging to our readers. At the time of publication of Issue #12 (Fall 2008) we had not received any solutions, so the deadline for submissions was extended to February 28, 2009. I am delighted to announce the decision of our panel, who decided to split the prize between two very talented students. Each submitted three correct solutions to the Challenge Problems, however, neither submitted a solution to the fourth one - Problem #6. The two winners, who will split the $300 prize money, are: Joshua Lam of The Leys School, Cambridge, England and Rati Gelashvili, of Tbilisi, Georgia, currently studying at the Georgian Technical University. Congratulations to Joshua and Rati. Solutions to the four Challenge Problems appear in this Issue of Pi in the Sky. We have included some solutions by our two Prize winners.

26

Solutions to problems 4,5, 6, and 7 published in Spring 2008 and published Sky Solutions to problems 4,5, 6, Issue7of Pi in thein the Spring 2008 Issue of Pi in the Sky

Problem 4. Prove that any positive integer n coprime to 10 SolutionsSolutions to problems 4,5, published publishedin to Solutions to 4,5, 6, and4,5, 6,and 7 in the Solutions to problems 7 and published in i of a repunit.problems problems 4,5,6,6, and77published div Problem 4. Prove that any positive Problems n is a Math Challenges integerthecoprime to 10Sky Spring 2008 and 2008 PiIssue ofPi in the Sky Spring 7 of Issue of Pi in the Spring 2008 in Spring published Pi Solution by Rati Issue2008 IssueofinSkyin the Sky Gelasvili: Solutions to problems 4,5, 6, the of a repunit. Since by Rati = 1 4, 5, 6 and 7 (9n, 10) Gelasvili: in the Sky Spring 2008 Issueaccording to Euler-Fermats theorem 9n is of Pi Solution 10(9n)= 1 according to Euler-Fermats coprime 9n10 is a to1 Problem Problem99...99. Consequently, n is theorem tocoprimedivis 4. 1 = thatProve that any positive integer n 11...11. Prove Since (9n, 10)Problem4.4. Provethat any positiveainteger nof coprimetoto Problem4. any positiveany positivedivisor n coprime Prove that integer n integer is a diviso Published in (9n) 1 = 99...99. Consequently,by is a divisor of 11...11. Alternative solution n of a repunit. aaarepunit.positive integer nJoshua Lam is a divisor 10 4. Prove that any ofof repunit. of repunit. Problem coprime to 10 SolutionLet msolutionRati Gelasvili: Pi in the Let Rn be the r by be by Rati Gelasvili: AlternativeRati aby by Joshua Lam Solution (Spring 2008) of Solution positive Gelasvili: Solution by Rati Issue 11 Gelasvili:integer coprime to 10. Sky of a repunit. n10) = (9n, 10) =1 to coprimetoEuler-Fermats9n theorem 9n 1s. Consider Since there are innitely ma SinceLet m Since(9n, 10) each according m). Euler-Fermats theorem 9n w (9n, Sincea positive integernEuler-Fermats theorembe the a divisori is repunit be 1(9n, 10) 10. Let Rn Since according R (mod according Solution by Rati Gelasvili:== 11 accordingtoto Euler-Fermatstheorem 9n is (9n) only(9n)dierent residues n is a divisor of 11...11. m (mod m), is divisor (9n) == 99...99.Consequently, are innitely many R 1 10 99...99.to=n99...99.Consequently, n n isaadivisorofof11...11. = (9n)11Consequently, Since theren9nais a divisor11...11. n 10 1s. Consider each 1REuler-Fermats theorembyisPigeonholeofPrinciple, (mod Consequently, 99...99. m). divisor of 11...11. 10 Since (9n, n = 1 10 10) according Solutionsonly problemssuch that Risolutionby Joshua Lami Rj we (mod to m i > Alternative by7Joshuaby m). Then RPrinciple 0 can in jAlternative solution Lam the Alternative 4,5, 6, and is aR m), by Pigeonhole (9n) 1 = 99...99. Consequently, n (mod j (mod Joshua Lam dierentsolution solution by Joshua Lam residues Alternative published 11...11. divisor of 10 j of in the (m, Let be ijIssue andPi(mod integer we conclude that R be he positive Prove that any positive integer n coprime mm|Rthat mm beaaajsince coprime = coprime Rto10. 0 the RRn ijthe iSpring 2008 Let10i Joshua Laminteger 1coprimetoto 10. Let m|Rbe , th > j solution R Rpositive integer 10. iLet j be (modnm) wi Let m be positive 10) to R positive Sky 10. Let n be Alternative such aLet bybe integer m). ThencoprimeRn Let repunitthe n be a Consider each R coprime m).1(mod therebe that thereij with 1s. positive1s.Consider(mod RR10. Let m). Sincethere are innitely b since 10) n (mod m). are m|R are to 10 is a divisor of a repunit. m m|Rij 101s.integer n (m,each= nnSince m). Since innitely many Rn , m n n 1s. Considereach to (mod Rn Since there are innitely m nj and Consider each R we conclude the repunit innitely Let ProblemLetLet n bepositive integer and let (n) denotes 5. a positive integer and let (n) only m dierent n residues are innitely Pigeonhole Principle dierent be a coprime m), by Principle Principle residues (mod Pigeonhole Princip ProblemConsider each only mm dierent(modthere (modton thataare coprimewe can Pr Prove dierentpositive integer or equal tom), is bydivisor any residues residues (mod m), by Pigeonhole to n 1s. 4.only mthat only(mod m). Since nm), by Pigeonholemany Rn , but n. of Rn 5.R denotea the m). integerR10 let (n)integers the, num positive integers positive Then of positive 0 denotes less number and R (mod m) hen Problem Let Rj R RR (mod m). Then R R be that Ri i >in thei ii > jjj suchnthat(modPigeonhole m). Then RRRnd000 (mo j such that such that Rby Rjj(mod Principle wei ican Rj (mod i i j (mod m). Then i (m >> such Solutions toSolution by Ratiandrepunit.m dierent residues (mod m), i i problems 4,5, 6, of a 7 only published j jj Gelashvili integers less or n that that less R since and Then R 10) are 0 (mod n. Provem|Rij or m|Rij Ranyandij 10integers m and10) =1 wethatto m) that m|R that 10j positive and equal to nn m|R m|Rij equal andsince (m,concludeweconclude that m|Rj 10 to since we 10)j coprime to hence Spring 2008 Issue of inithejSky of positivem|Rij 10j jj m). since (m,i are ==11weconcludeijn. Prove ij blems 4,5, 6, and 7 published in thePiSolution such that Gelasvili:(mod(m, 10) = 1(m, Rcoprimeconclude,thatm|Rij > by Rati i any positive integers m and n (mn) 2008 Issue of Pi in 10) = 1 according to (9n, 10) = 1j according to Euler-Fermats theorem 9n is (m2 )(n2 ). Since Euler-Fermat's since (m, 10) = positive integers m|Rijn a Since (9n, the Sky m|Rij 10 Problem 5. Let n be a positive integerm anddivisor of and that for any 1 we conclude that and let (n) denotes the numb Problem a divisorbe positiveinteger and let (n) denot Problem 5.Let n of 11...11. integer and let (n) denot Problem Let be positive integer be positive (9n) 1 = 99...99. Consequently, 10(9n) 1 = 99...99.10 is a divisor n is 5.5. Letn n aaa(m2 )(n2 ). and let (n) denote theorem 9n that divisor of integer - coprime to (mn) Problem 4. Prove is a any positive 10 n of positive integers lessintegerslessororthat(n)coprime arecoprimetoton.n. P Solution: or equal to n let are nthat are coprime positive Lam equal that to n. Prove n. f 5. Letofpositive integers less andequaltotonthat are coprime tothatP Alternative solutionofofbe a positive bypositive integers lessorequal tondenotes Joshua Consequently, n is a divisor is a11...11. Problem positivenintegers mthe integerfactors of m and n, qthe number factor of divisor of aany positive integer n coprime to 10 at repunit. any Solution: positiveintegers mmandn n and that mand n n areand Let any pi be integers any equal j the less coprime to prime Rcoprime repunit with prime Let of positive integers any orpositiveintegersLet n be theto n. Prove that for m be a positive integerpositive to n 10. Solution by Rati Gelasvili: Solution many R22, but22 the prime factors m and2 n, q and rthe prime factors ofof n only. 2 the prime factors of m o any positive Let pi m k there are (mn) )(n ). )(n2 ). (m by Consider 9n integers be and n (mn) j (m )(n (mn) (m Since (9n, 10) Alternative Solutionn 1s.Joshua Lam is n (mod m). Since (mn) innitely (m2n)(n ). ). = 1 according to Euler-Fermats theoremeach R a divisor of elasvili: We prime factors n only. 2 of and rk the(mod have by Pigeonhole Principle we can nd m), (9n) 1 = 99...99. Consequently, n is only m dierent residues a divisor 10 to Euler-Fermats theorem 9n is a divisor ofof 11...11. Solution: (mn) prime factors of m and n, ing Let pi be the (m2 )(n ). 1 We Solution: Solution: 1 q the i coprime to 10: j (mod m). Then Ri Let a solution by Joshua Lam j such that RiLet Rhave Solution: = Rj 10(mod m),1 hence j 1 1 m be a positive integer > Alternative sequently, n is divisor of 11...11. mn and r Let pi primeLet i i(mn) m only, n, km the primej the primemfact be the factorsfactors of mfactors qmandprime factors of facto j and Let i be of prime factors pi be the n,q the1prime ron = prime the prime i1 and of theand j jfactors conclude k R be positive integer coprime m|RijLet Solution: (m, 10)Let1pppbe the prime factors ofofjmandn,n,qqj theprime facto Let m n be the repunit with n 1s. Consider each Rn sincerepunit with we mnof n 1that m|Rij 1 j1 to 10. 10Rn be (mod the prime factors the n by JoshuaaLam k (mn) the = only. rk n only. and only. prime factorsof n only. andfactors prime factors Letbutbe manyof andrrk theprime and pij ofn nprime qj of1m only, piand the prime rkkthe of m factors q oftheonly. factors rk n, m). Since Rn (mod infinitely many R only n 1s. Consider to 10. Let Rn be the repunit with innitely We have n , but nteger coprime each there are m). Since there are n, 5. Let n be aR j Problem the prime factors Wehave positive have iand let (n) denotes the number integer We n only. 1 k 1 and Pigeonhole can Wehave rk Weofhave 2 2 m different are (mod m), m); n , the only m dierentthere residues (mod by Pigeonhole Principle we nd mod m). Since residues innitely many Rby but integers less or equal to n that are coprime n. Prove for (m ) = mto 1 1 that of positive We have 1 1 p hence (mn) ==mn 111 111 1i 11j11 111 1 qj 111 11 > j such that Pigeonhole(mod i > jwe canthatRRiRj (mod m) (mn) = mn2 Principle Rj Principle such positive 0 (mod (mod m), by Riwe can find m). Then Ri nd j integers m and, n (mn)1 mn 1 any (m ) = m2 mn (mn) p 1 1 q 1 r = i j and since (m, 10) = 1 we conclude that m|R i k q qj j rrr m|Rij m). Then Ri i Rjj 0 (mod m) , hence 10 1i 1 i ppj j j jj 1j qqj k kkk (mod m). Then R - R 0 (mod m), hence kk ij j p pi i i i i ii 1 1 (mn) = mn 1 1 2 )(n22 )1 n2 (mn) (mi (n ). qj rk = 1 , 10) = 1 we conclude that m|Rij p 1 j m|RLet 10 a positive (m, 10) = 1 we conclude i j k rk111 1 1 1 pi 11 1 Problem 5. i-j n be and since integer and let (n) denotes the number 2 (n2 m(m )))=mm 1 1 i (m2 ) ) = (m222==222 111 1 k = n2 1 1 (m m 111 that m|Rless or equal to n that number Solution: of positive integer and let (n) denotes the are coprime to n. Prove that for integers i-j a qqj 1ii p i or pppiqrkj jj i j j isi replaced byqj 1 if t Notice 2 that each product pii, i i 1k j 2 Let (m = 1 prime i any positive that are coprimento n. Prove that for be the prime factors) of m and n, qj the j k m only, integers m and equal to nAlso solved by Simon Morrispi or 1 factors of by if there are 1 p q , Notice n only. 1 i and rk the prime factors of that containing pi , qj222 j k 21is jreplaced factors each 2product i ,or r . 22 111 nd n j = 2 )(n2 ). (n ) n2 (n ))1=nk = (ni =n 1 111 111 (mn) (m ) Now the ,required (n n pi equivalent to 111rrrk .i factors p q R and r are respectivelyWe have the radii of the containing j ,or rk 1inequality iisi 1 pppirk kkk ii kk 2 ) = n2 i i k (mn) (m2 )(n2 ). (n 1 1 Now that each product pi to by 1 if inequality or is equivalent 1 1 Solution: spheres circumscribed about, and inscribedNotice the Notice that each 1product ,rreplaced is replacedthere 111 if in requiredithat each ,product 1 k is is1replaced by are i replaced by if by Notice k (mn) = mn 1 Notice that each product , , oror 1 is or 1 1 Let pi be a tetrahedron. Prove that n, q the prime factors of mNotice that each qij j 1 i i, rjj ork1kis replaced the prime factors of m and R j 3r. only, i product i j k k k p i j or k qj by1 1 if there are no rk Notice that eachfactorscontainingk p. p, ,is,or rrr . . . product , qj ,or r factors containing there, are1noqqfactors containing1p , q , nor and rk of mprime factors of n only. factors containing k i jq ,or kkk 1 byfactorspcontainingp,ij1 jreplaced k 1 if i i j i j ,or ctors the and n, qj the prime factors of m only, i j Now the. q Now therequired inequality is istok required. inequality isinequalityisequivalent toto the 1 qj equivalent r We have Now Now required equivalent of n only. Solution 1 j k factors containing pik, 2 ,or rk therequired inequality equivalent to 2 r j 1 1 (m) = m p Now inequality required 1inequality 11 is 1 1 the required Now the iequivalentj 111 equivalent 1 q to is 111 to i j 1 = mn 1 the tetrahedron having the (mn) consider 1 1 111 1 1 111 r 11 1 1 Let us 1 p 1 qj rk qqqj k rrrk k 11 qj j jj 11 jj 1 vertices 1 the icentroids ofj the faces of the given 1 i kk j k 2 2 at qj kkk 11 1 11 pi rk (n ) = n i j k qpi rrk j k tetrahedron. 2 This 2tetrahedron ji 1 is similar to the 1 kk (m ) 1 1 11 given 1 = m ratio 1of similitude qj eachIts one and the1 pi Notice is 1/3 . product is, obviouslyreplaced by equality1occursno or is true. The 1 if there are only if that 2 that i j =m 1 1 i j k circumscribed sphere is therefore of radius R/3. This 1 pi qj i j facesof the tetrahedron, ,ormkand n have the same prime factors. r . 1factors containing pi , qj 1 sphere (n2 ) = n2 the1 intersects all Now1therrequired inequality is equivalentsolution was submitted by A similar to 1 its 1 p therefore radius should be igreater or equal than k 2 i k =n 1 1 Joshua Lam Rati Gelashvili p the radiusk of the rk sphere. inscribed is replaced by 1 if there are 1 and 1 i no Notice that i each product , or 1 1 1 k qj rk is replaced by i 1j if there are no duct , or j k actorsi containing pi , qj ,or rk . 27 j k r rkNow the required inequality is equivalent to .

Solution: Let us consider the tetrahedron having the vertices at the centroids of the ces of the given tetrahedron. This tetrahedron is similar to the given one Prove that in any convex polyhedron (a) By using these inequalities and Euler's polyhedral d the ratio of similitude is 1/3 . Its circumscribed sphere is therefore of dius R/3 . This sphere intersects al the faces of or atetrahedron, therefore we get there is either a triangle face the vertex at formula radius should be grater or equal than and radius of the inscribed sphere. which three edges meet the (b) there is a face

Let F1 , ...,Ff, V1, ...Vv be the faces and the vertices i=1 i=1 i=1i=1 i=1 i=1i=1 i=1 , ..., Ff V1 , ...Vv be the faces and the verices of the and Let F1 of the ,polyhedron, fi the number of edges of Fi polyhedron, fi Also, any V . Also, e numberi of edges of Fi of edges issuing fromedgesAlso, in any polyhedroninwe have issued from polyhedron we we polyhedron we have v the number and vi the number of Vi. Also, in in any ipolyhedronany have have (a) Assume the contrary, that there is no triangle face and there is no v v vv (a) Assume the contrary, that there is no triangle vv v v rtex at which three edges meet, i.e., 3 3v hence 2e 3v. hence face and there is no vertex at which three edges v v 3 = = hence 2e 2e 2e = vii i 3 = 3v 3vhence 2e 3v.3v. 2e= = i i=1 i=1 i=1i=1 i=1 i=1i=1 i=1 meet, i.e. fi 4, vi 4 f 4, vi 4
f i=1

which Problem 7. Prove that in any convex polyhedron (a) there is either a is a contradiction. Therefore any convex angle face or a vertex at whichSolution meet and (b) there is a face three edges polyhedron should have at least a face with three ving less then six sides. edges or Therefore vertex at which three edges which is a contradiction. at Therefore any convex polyhedron should have at which a a contradiction. least a any convex polyhedron should have at which is is contradiction. Therefore any convex polyhedron should have at least a face meet. If v, e, f denotes the number of vertices, aa face with three edges or orat at least vertex at which three edges meet. least edges with three edges or a at least vertex at which three edges meet. least face with three edges a a least vertex at which three edges meet. Solution: Assume contradiction that every face has more then ve edges. (b)(b)we (b) Assume by contradiction that every face has more then ve edges. and respectively faces of the polyhedron thenAssume by by contradiction that every face has more then ve edges. (b) If v, e, f denotes the number of vertices, edges and respectivelyhave: ofAssume by contradiction that every face has faces As As above we have: above we have: As have then we polyhedral formula formula above we more then five edges. As above we have: Euler's have Eulers polyhedral e polyhedron ff ff f f v-e+f=2. f f 6 6f hence e 3f. hence 2e 2e = fii i 6 = 6f 6fhence ee 3f.3f. 2e= = f f 6 = = hence ve+f =2 i

having less then six sides.

4 = (2f - e) + (2v - e) 0

hen we have:i

OnOn the other hand the using Eulers polyhedral Euler's polyhedral On the other hand by by other hand, by using formula we get the other hand by using Eulers polyhedral formula we get On using Eulers polyhedral formula we get

Then we have:
fi
f i=1

formula we get
v i=1 v

2e =

4 = 4f and 2e =

nce e 2f hence e 2f and e 2v. and e 2v. By using these inequalities and Eulers polyhedral formula we get A similar solution was submitted by Joshua Lam. The probRemark: similar solution was submitted Remark: A similar solution was submitted by Joshua Lam. The probRemark: should have at most five edges.by Joshua Lam. The probA lem was also solved by by Rati Gelasvili. lem was also solvedsimilar solution was submitted by A lem was also solved by Rati Gelasvili. Rati Gelasvili. 4 = (2f e) + (2v e) 0 2 Math Challenges Problems

vi 4 = 4v which combined with the above two inequalities leads to to contradiction. which combined with the above two the above leads a contradiction. which combined which combined with inequalities two to a a contradiction. with the above two inequalities leads inequalities i=1 Hence least one face a should have most ve edges. Hence at at least oneto should have at at most ve at least one face Hence at least leadsface should have at most ve edges. one face contradiction. Hence edges.

6 = (3f - e) + (3v - 2e)

6 (3f e) + (3v 2e) 6 = (3f(3f e) + (3v(3v 2e)2e) 6 = = e) +

Joshua Lam. The problem was also solved by Rati Gelashvili.

Published in Issue 12 (Fall 2008) of Pi in the Sky

Prove that the equation: x5 + y5 + 2 = (x + 1)5 + (y + 2)5 does not have integral solutions

Prove that pn > 2n for every n 5 where pn denotes the nth prime number (p1=2)

Solution by Edward T. H. Wang

If x and y are of the same parity, then the left side of the equation is even while the right side is odd. If x and y are of opposite parity, then the left side is odd while the right side is even. Also solved by Rati Gelashvili, and Joshua Lam 28

Solution by Rati Gelashvili

Let's prove by using induction. The assumption holds for n = 5. If p33- 13 > 2(n- 1) then pn- 1 2n n 1 hence pn > pn- 1 2n- 1. Since pn 2n we must have pn> 2n Also solved by Edward T. H. Wang, Joshua Lam and Andrew J. Pai

the by side square 0.1 can be one square contains at least t We rst Since a square ofRati Gelasvili: be least completely coveredsquares of si The Principle Solution Pigeonholelength can at partitioned in 100 little by a circle rk that remark that 3 3 2 2 Sinceconclude thatsidePrinciple little squares square lengthsconta a square of there in 1000.1circle be radius 0.1 contains at B The square the Pigeonhole length at least one of side which 0.1. a3 + b3 = (a + b)(a2 ab + b2 ) (a + b)ab radius 0.1 we can be partitioned exists a can of completely covered by 3 3 2b = (a + b)(a ab + b ) (a + b)ab at least three of the given points. 2 radius Principle at least one square contains be least three poin 0.1 we square of side length 0.1 can at completely wh the Pigeonhole Since a conclude that there exists a circle of radius 0.1cove a + b = (a + a + ab + b ) (a + b)ab b)(a hence Inside a square with side length 1 there are square radius 0.1 we cgiven points. there exists a circle of a circle0 at least three a, the conclude that numbers covered by radius Since a of Let length 0.1 can be completely such that side of b, be positive hence 201 points. Prove that there exists a radius 0.1 we conclude 18.b,Prove exists anumbers such that abc = 18. Pr circle abc = a, cthere that circle of radius 0.1 which conta Problem 4. Let that be the given points. at least three of positive of radius 0.1 which contains at least three of threeProblem 4. points. b, c be positive numbers such that abc that at least of the given Let a, 3 3 3 (a + b)ab c3 a 3 + b3 c3 c these(a + b)ab 3 points. that a + b + c ab + a,+ ba + c + cnumbers such tha Problem 4. Let c b, c be positive a + b (*) 3 b3 + 3(a + b)ab + c3 2 3 c (a + b)ab c3 (a b)ab 3 = cabca + b. + a (a + b)abc c + a3 + b3 + c3 2 Solution 2 Rati c 22 a + b. 22 22 + + 2 + by2 = Gelashvili c abc a + b. 4. Let 3a, b, c be positive a b + c + b athat + c = 18. Pro abc 2 = Problem that numbers such + c abc a + b 3 + b3 + c3 2 2 2 2 2 2 2 2 2 2 2 22 2 3a thatSolution by Rati Gelasvili: Solution by Rati Gelashvili ba + c + ca + b+c+ 3 + b3 + c 3 Similarly, The square can be partitioned in 100 little 33 2. If amultiply the left side of a 18 Solution by abc Gelasvili: we Since abc = then Rati =b c + b c + ca + b Similarly, a + a+ squares of side lengths 0.1. By the Pigeonole Since and 18 by 18 then = 2. If multiply we 3 Since abc =the right abcby 3 abc then multiply the wh inequality by 3 2abc = then side Gelasvili:If we the inequality left c + a Principle at3 b3 +one aabcc + b least 3 square contains at least inequalitySolution and Ratiright side by abc then the inequ 3 + b3 a3 replaced that the that Since abc by the left side If we multiply th Solution by is by 3 of2 the 18 c3 three points.2Since a2 aabcc + b + b3 a3 c Weshould be provedrst remark = inequality by 3 2. and the rst remarkWe Rati Gelasvili: then abc = + a 2 + c square of side length 0.1 can shouldinequality is replaced2. thewe multiply abc then of t abc + b be proved 3 =andbyIf need the inequality:side the 23 right side by left the abc 318 by 0.1 right side 3 abc then+we 2be2 completely covered2by a circle with radius Since 3 = then by 2 2 2 3 inequality a + b + c be b)(a2 isside c b)(aby +b)ab b2 )a + b). () by 3should the3abc(a =replaced abc then right 2 2 c + by b inequality b3 +c a3that3 there exists a circle of radius 0.1 2 a 3 +32=and+ 3proved3 abb(ab+)+ (aa+ ab + cthe (a + b)ab whi b (a + we 3 conclude 3 3 3 2 +b a +b + by + c b3 b3 a 2 at + b the givenc +c a3 which contains 2 + 2 3 ofabca + c points. should be proved isareplaced c abc(a b + c + b a + c + c a + b). 3 abc a + + b abcleast Webrst remark that a+c (using the fact 3(a -3b)2 0 for any a and b). that hence 2 three 2 2 2 is2 obtained 2 adding side by side the above 3 hence 3 2 a + b + c abc(a b + c + b a + c + c a The inequality () Also solved by by Edward T. H. Wang, 2 a + b3 + c abc(a b +4c + b a + c + c a + b). () The inequality () is obtained by adding side 3by b3 = (a +above Using 2this (a + b)ab we have y () is Joshua Lam andside by sideJ. Pai. Wethree remark2 that + b ) inequality obtained by adding Andrew the aboveside the b)(a three equalities. a + rst ab 4 We a3 + remark3that(a + 3 rst 3 equalities. c (a c c3 c3 c + b)ab Let ABCD be a convex quadrilateral, Mbthe a3 3 (a + b)ab + b)ab 3 = (a + b)ab 3 b. 4 hence quadrilateral, the midpoint b + c 2 2 2 + + + Problem 5. Let ABCD be a convex midpoint of 2M2 Ifa3 + b3 = (a + b)(a 2 + ) 2(a2 b)ab2 c abc a + = c ab midpoint ofBCbe a convex and N the CD. +2 2 2 2 2 2 22 2 2 ab 2 b2 4 2 2 2 Problem Let ABCDtheLet ABCDquadrilateral, + ANthe 1 then thea3that3the a AN M the midpoint BC and N 5. be +convexof1CD. Ifprove quadrilateral, Mprovemidpoint (a + b)(a ab + b ) (a + b)ab midpoint= then AM that = area of the + b = AM 1 then eBC and quadrilateral is less than = AM + AN =hence prove that the the eamidpoint of CD. If AMof CD. If3 1 then prove thatSimilarly, 3 of the N the midpoint + AN than 3 1/2. Similarly, quadrilateral is less a + b 1/2.c3 henceb)ab (a + b)ab c3 (a + c ea of theis less than 1/2. less than 1/2. + ilateral quadrilateral is Solution: = c abc a + b. + 2 Solution 2 2 2 2 2 2 2 Solution: c32 b3 +2 a3 c3 b3 a3 + We have c3+ (a abc +cc+ b a abc c + b 3 + b3 3 a + b)ab 3 a3 c3 (a + b)ab 2 3 We have We have 2 2 +2 b)ab 3 2 = cabc 2 b. c a b3 + (a + b)ab + 2 (a the inequality x +a + xy c + c 2 Similarly, SABCD = SABC + SADC = 2SAM C +22 C 2 Notice + 2we 2 2SAN + 22 that used 22 2 = c abc y a + b. 2 2 2 2 2 is valid 2 b3 a3 c3 b2 2 which 2a3 + c3 for any non-negative numbers x and y. 3 SABCD = = 2S + SADC AN C AN + BCD = SABC + SADC SABCAM C= 2S = 2SAM C + 2S. C 3 + b + c b abca + c = 2SAM CN + 2S N + 2SCM N 3 + b abc a + AM c a3 Similarly, a abc 2 = 2S N CN = 2S . 2 c 2 2 2 = 2SAM CN = 2SAMAM+ 2SCM NAM N + 2SCM N . + Similarly + b 2 Similarly, 2 2 SinceN ) < dist(A,MN)) <the quadrilateral 2 (the dist(C, M N ( dist(A, MN) is convex) we conSince dist(C, M The inequality c3 +is 3obtained () adding side by adding side by threethe () The inequality by is obtained by side the above side a3 Since dist(C, M N ) ) ( the MN) ( convex) we c3 b + a M N ) < quadrilateral dist(A,convex)thewe AM N . inequalities. b3 con- b3 3 aabcc + b convex) wea is that + HenceSABCD is4S include 3 + ude that dist(A, M NAM N . is quadrilateral< quadrilateralcon- c3 SCM N < S < + inequalities.a 2 a abc c + b b22 + + c abc SAM ude SAM Nother << SAMNN .Hence SN . < 4SAMN.N . < that S.CM N hand . < 4SAM ABCD < 4SAM HenceSABCD HenceS ABCD On the SCMN 2 2 2 2 2 2 On handthe other handother hand On the Problem 5. Let + c3 a quadrilateral, quadrilateral, M Problem 5. Letconvex above a convex M the midpoint t a3 ABCD b3be sideABCD be three 1 1 The inequality () is obtained by adding3 sideb3 a3 + by the + AN SAM N = 1 AM AN sin M AN 1 AMBC and N the BC c +ofthe bIf abca +CD.= 1 AM + AN = 1 then pr of AN AM of midpointN CD. b abc a + c If then prove that the and + 2 midpoint of c 1 S 1 22 2 AM AN sin AM AN sin AMMAN AM N = M AN inequalities. 2 AM AN quadrilateral is quadrilateral is less than 1/2. AN 2 the less 2 2 MN = area of the than 1/2. 2 1 1 1 2 1 inequality area of 2 2 = 1 AM (11 AM ) 1 1 =TheSolution: () is obtained by adding side by side the above three 1 . Solution: 1 1 (1 Problem 5.4 Let1 .ABCD be a() Adding these adding the midpoint the above three is obtained by inequalities side 8 AM =The inequality convex quadrilateral, M side by together, we get = AM (1 = 2 AM = . ) 2 inequalities.have AM ) 2 2 of BC8and N the midpoint of CD. If We have = 1 then prove that the 2 4 8 We inequalities. 4 2 AM + AN Therefore Therefore area of 1 quadrilateral is less5. Let1/2. 3 + b3 be ca )convex quadrilateral, M the midpoint the 1 Problem than ABCD S 3 + S Therefore 4(a + SABCD = beSABCD ADC = ABCc+ SADC AN C midpoint + M the S1 ABCconvex= S 2SAM C b 2S c 2S a + b). Problem a abc(aAN+= + then prove that the b 1 SABCD < 4 1 = 1 . of BC and N 5. Let ABCD CD. If AM quadrilateral, a += + cAM C + 2SAN C Solution: of 1 . SABCD < 4 = ABCD < 4 8 = 2 . of BC and N the midpoint= 22SAMIf AM + AN + 2Sthen . the midpoint2of =1 that the 2S We 8 have2 area of the quadrilateral is less CD. CN = =+ 2SAM CN CM2Sprove + 2SCM N . 8 2 than 1/2. AM N = N AM N Dividing through by abc = 3 2 and rearranging area of the quadrilateral is less than 1/2. Also solved by Joshua Lam. Solution:dist(C, M N )thedist(C, M N ) (*)dist(A, M N ) ( the convex) we con-con Since ( Since inequality <the gives < dist(A, Solution: SABCD = SABC + SADC = 2SAM C + 2SAN C quadrilateral is quadrilateral is We having There are given five line segmentshave SCM N < Sthat SCM N < ABCDN<HenceSABCD < 4SAM N . clude that clude AM N . HenceS SAM . 4SAM N . We = 2S have 5 2SAM solved AM the property that any three of them can be the CN = Also N + 2SCM N . by Joshua Lam using On the other hand the other hand 5 5 SABCD On SABC + SADC = 2SAM C + 2SAN C techniques + S would not + 2S sides of a triangle. Prove that at least one SABCD = SABC that = 2SAM C be standard for a of = quadrilateral ADC AN Since dist(C, M N ) < dist(A, M N ) (= 2S audience. convex) we 1 . C the con1 1 general 1 AM AN is M AM AN these triangles must be acute. SAM N <2SAM CN. =N2SAM N + 2SCM N .sin M AN AM AN = 4S CN = 2SAM N + 2SCM N sin AM AN AN SAM = = AM N clude that SCM N < SAM N . HenceSABCD AM 2 2 2 2 Solution 1 1 On the other hand dist(C, M N ) < dist(A, M N ) ( the quadrilateral is convex) we con1 1 1 1 1 1 Since AM (1 = Since dist(C, MN) = Let a b c d e be lengths of the that SCM NM N )AM N= HenceSABCDAM )2 2 . 4 AM ) 2 we= 8 . segments. < S < dist(A,AM (1 ( the quadrilateral is. convex) 4 con2 1 8 . < 4SAM N clude clude that AM < SAM Assume by contradiction that all AM NOn the1SCM N hand M.AN ABCD AN AM N . S the constructed AN sin N HenceS AM < 4S = other Therefore On the2other hand Therefore 2 triangles are right or obtuse. 1 1 1 1 1 1 S 1 1 ABCD S .1 = AMAM N AM ) 1 AM AN sin< 4AN ABCD <AN = . (1 = 1 = . M 8 = 2 1 AM 4 S Then 8 2 SAM N = 2 AM 4AN 8 M AN 2 AM AN 2 2 sin 2 1 1 12 1 e2 d2 + c2, d2 c2 + b2, c2 b2 + a2 Therefore = 1 AM (1 AM ) 1 1 = 1 . 1 =1 2 AM (1 AM ) 2 4 = 8 . and therefore SABCD < 4 = . 2 2 4 8 8 2 2 2 2 2 2 2 2 2 Therefore 2 5 5 e d + c c + b + b + a b + 2ba + a Therefore 1 1 SABCD < 4 1 = 1 . From this inequality we obtain that e a + b, SABCD < 4 8 = 2 . 8 2 a contradiction. 5

29
5

Problem 1
Find all positive integers n such that log2008n = log2009n + log2010n.

Problem 2
Find the smallest value of the positive integer n such that (x + y2 + z2)2 n(x4 + y4 + z4) for any real numbers x, y, z.
2

Problem 3
Let a be a positive real number. Find f(a) = maxx {a + sin x, a + cos x}.

Problem 4
Prove that the equation x2 - x + 1 = p(x + y) where p is a prime number, has integral solutions (x, y) for infinitely many values of p.

Problem 5
Find all functions f : such that 3f(n) - 2f(n + 1) = n - 1, for every n . (Here denotes the set of all integers).

Problem 6
In ABC, we have AB = AC and BAC = 100o. Let D be on the extended line through A and C such that C is between A and D and AD = BC. Find DBC. .