This action might not be possible to undo. Are you sure you want to continue?

Foreword Elwyn Berlekamp and Tom Rodgers

Ü ½ ¿ ½¿ ½ ¾½ ¾¿ ¾ ¾ ¾ ¿ ¿

I Personal Magic

Martin Gardner: A “Documentary” Dana Richards Ambrose, Gardner, and Doyle Raymond Smullyan A Truth Learned Early Carl Pomerance Martin Gardner = Mint! Grand! Rare! Jeremiah Farrell Three Limericks: On Space, Time, and Speed Tim Rowett

II Puzzlers

A Maze with Rules Robert Abbott Biblical Ladders Donald E. Knuth Card Game Trivia Stewart Lamle Creative Puzzle Thinking Nob Yoshigahara

v

vi

Contents

Number Play, Calculators, and Card Tricks: Mathemagical Black Holes Michael W. Ecker Puzzles from Around the World Richard I. Hess OBeirnes Hexiamond Richard K. Guy Japanese Tangram (The Sei Shonagon Pieces) Shigeo Takagi How a Tangram Cat Happily Turns into the Pink Panther Bernhard Wiezorke Pollys Flagstones Stewart Cofﬁn Those Peripatetic Pentominoes Kate Jones Self-Designing Tetraﬂexagons Robert E. Neale The Odyssey of the Figure Eight Puzzle Stewart Cofﬁn Metagrobolizers of Wire Rick Irby Beautiful but Wrong: The Floating Hourglass Puzzle Scot Morris Cube Puzzles Jeremiah Farrell The Nine Color Puzzle Sivy Fahri Twice: A Sliding Block Puzzle Edward Hordern Planar Burrs M. Oskar van Deventer

½ ¿

½¼¿ ½¼ ½½ ½¾ ½¿½ ½¿ ½ ½ ½ ½ ¿ ½

Contents

vii

Block-Packing Jambalaya Bill Cutler Classiﬁcation of Mechanical Puzzles and Physical Objects Related to Puzzles James Dalgety and Edward Hordern

½ ½ ½ ½ ½ ½ ½ ¿ ½ ½ ¾¼¿ ¾½¿ ¾½ ¾¿ ¾ ½ ¾ ¾ ¿

III Mathemagics

A Curious Paradox Raymond Smullyan A Powerful Procedure for Proving Practical Propositions Solomon W. Golomb Misﬁring Tasks Ken Knowlton Drawing de Bruijn Graphs Herbert Taylor Computer Analysis of Sprouts David Applegate, Guy Jacobson, and Daniel Sleator Strange New Life Forms: Update Bill Gosper Hollow Mazes M. Oskar van Deventer Some Diophantine Recreations David Singmaster Who Wins Misère Hex? Jeffrey Lagarias and Daniel Sleator An Update on Odd Neighbors and Odd Neighborhoods Leslie E. Shader Point Mirror Reﬂection M. Oskar van Deventer How Random Are 3x + 1 Function Iterates? Jeffrey C. Lagarias

Forward

Martin Gardner has had no formal education in mathematics, but he has had an enormous inﬂuence on the subject. His writings exhibit an extraordinary ability to convey the essence of many mathematically sophisticated topics to a very wide audience. In the words ﬁrst uttered by mathematician John Conway, Gardner has brought “more mathematics, to more millions, than anyone else." In January 1957, Martin Gardner began writing a monthly column called “Mathematical Game” in Scientiﬁc American. He soon became the inﬂuential center of a large network of research mathematicians with whom he corresponded frequently. On browsing through Gardner’s old columns, one is struck by the large number of now-prominent names that appear therein. Some of these people wrote Gardner to suggest topics for future articles; others wrote to suggest novel twists on his previous articles. Gardner personally answered all of their correspondence. Gardner’s interests extend well beyond the traditional realm of mathematics. His writings have featured mechanical puzzles as well as mathematical ones, Lewis Carroll, and Sherlock Holmes. He has had a life-long interest in magic, including tricks based on mathematics, on sleight of hand, and on ingenious props. He has played an important role in exposing charlatans who have tried to use their skills not for entertainment but to assert supernatural claims. Although he nominally retired as a regular columnist at Scientiﬁc American in 1982, Gardner’s proliﬁc output has continued. Martin Gardner’s inﬂuence has been so broad that a large percentage of his fans have only infrequent contacts with each other. Tom Rodgers conceived the idea of hosting a weekend gathering in honor of Gardner to bring some of these people together. The ﬁrst “Gathering for Gardner” (G4G1) was held in January 1993. Elwyn Berlekamp helped publicize the idea to mathematicians. Mark Setteducati took the lead in reaching the magicians. Tom Rodgers contacted the puzzle community. The site chosen was Atlanta, partly because it is within driving distance of Gardner’s home. The unprecedented gathering of the world’s foremost magicians, puzzlists, and mathematicians produced a collection of papers assembled by

ix

x

FORWARD

Scott Kim, distributed to the conference participants, and presented to Gardner at the meeting. G4G1 was so successful that a second gathering was held in January 1995 and a third in January 1998. As the gatherings have expanded, so many people have expressed interest in the papers presented at prior gatherings that A K Peters, Ltd., has agreed to publish this archival record. Included here are the papers from G4G1 and a few that didn’t make it into the initial collection. The success of these gatherings has depended on the generous donations of time and talents of many people. Tyler Barrett has played a key role in scheduling the talks. We would also like to acknowledge the tireless effort of Carolyn Artin and Will Klump in editing and formatting the ﬁnal version of the manuscript. All of us felt honored by this opportunity to join together in this tribute to the man in whose name we gathered and to his wife, Charlotte, who has made his extraordinary career possible. Elwyn Berlekamp Berkeley, California Tom Rodgers Atlanta, Georgia

Also below are two quotes showing a strong childhood interest in puzzles.] *** [1934] “As a youngster of grade school age I used to collect everything from butterﬂies and house keys to match boxes and postage stamps — but when I grew older . *** “I have recently read an article on handwriting and forgeries in which it is stated that ink eradicators do not remove ink. I am deeply interested in the success of the organization. and I’ve always talked clearly when people interview me. I don’t think my life is too interesting. Can you give me a description of this process.Martin Gardner: A “Documentary” Dana Richards I’ve never consciously tried to keep myself out of anything I write. The second was the “New Color Divination” in the magic periodical The Sphinx.” [2] *** An able cartoonist with an adept mind for science. [1932 yearbook caption. puzzles. while a sixteen-year-old student at Tulsa Central High. and writing were to stay with him. what chemicals are used. The ﬁrst two times Gardner appeared in print were in 1930. and how it is performed?” [1] *** “Enclosed ﬁnd a dollar bill for a year’s subscription to The Cryptogram. magic. there are various interviews and articles about Gardner. That is. It’s lived mainly inside my brain. Instead of a true biography. The early interest in science. The ﬁrst. and 3 . we give a collection of quotes and excerpts.. we present here a portrait in the style of a documentary. was a query to “The Oracle” in Gernsback’s magazine Science and Invention. a month later.. having been a fan for some time. quoted below. [21] While there is no biography of Martin Gardner. without narrative but arranged to tell a story. and that ink so bleached can be easily brought out by a process of ‘fuming’ known to all handwriting experts. but merely bleach it. I sold my collections and chucked the whole business.

For about a year I actually attended an Adventist church.. was a pantheist. My conversion to fundamentalism was due in part to the inﬂuence of a Sunday school teacher who was also a counselor at a summer camp in Minnesota where I spent several summers. My father.... I learned later. Knowing little then about geology. never missed a Sunday service unless she was ill..... as far as I know. [19] *** “My fundamentalism lasted.” [22] *** Gardner was intrigued by geometry in high school and wanted to go to Caltech to become a physicist. At that time. Thus it was several years ago I decided to make a collection of mechanical puzzles.. then as now a citadel of secular humanism.. I became convinced that evolution was a satanic myth.. He was the chief detective in a series of short stories that ran many years ago in one of the popular mystery magazines. There was no particular day or even year during which I decided to stop calling myself a Christian. RICHARDS began to look for something new to collect. It wasn’t long until I discovered Dwight L.. That institution in the 1930s was under the inﬂuence of Robert Maynard Hutchins. Caltech accepted undergraduates only after they had completed two years of college. incredibly. a lapsed Baptist who became a wellknown humanist. “The ﬁrst and only puzzle collector I ever met was a ﬁctitious character....... Throughout my ﬁrst year in high school I considered myself an atheist. I can recall my satisfaction in keeping my head upright during assemblies when we were asked to lower our head in prayer.” [3] *** “My mother was a dedicated Methodist who treasured her Bible and. [and] Seventh-Day Adventist Carlyle B. I was one of the organizers of the Chicago Christian Fellowship. so Gardner went to the University of Chicago for what he thought would be his ﬁrst two years.. which wound up displacing his interest in physics and Caltech. The erosion of my beliefs was even slower than my conversion. however. Moody ..4 D. A major inﬂuence on me at the time was a course on comparative religions taught by Albert Eustace Haydon.. Haynes. Gardner. who had decreed that everyone should have a broad liberal education with no specialization at ﬁrst. but at any rate I have found the hobby equally as fascinating.” [22] . through the ﬁrst three years at the University of Chicago. Personally I can’t say that I have reaped from my collection the professional beneﬁt which this man did. thus prevented from pursuing math and science. took courses in the philosophy of science and then in philosophy...

Martin disposed of it all. divesting himself of all material things to which he might be forced to give some consideration.. [9] *** [1940] A slim. and he has spent the last halfdozen years of his life eliminating both from his consideration. are now his most precious.He tired of visiting oil companies every day. He is doing his stint again this season.” Gardner said of his reporting stint. with ﬁfty or a hundred dollars at the eleventh hour — the result of having sold an idea . and took a job . Those hands can at any time be his passport to fame and fortune.“Real dull stuff. facing eviction from the Homestead Hotel. and almost only possession. together with the summarized total of his knowledge..MARTIN GARDNER: A “DOCUMENTARY” 5 “After I had graduated and spent another year at graduate work. The son of a wellto-do Tulsa. his simian stride and posture is contrasted by the gentilityand ﬂuent deftness of his hands. after having ﬁrst cut out from the important books the salient passages he felt worth saving or remembering. In a civilization of property rights and personal belongings. and somewhat rareﬁed library. in Chicago. and triumphantly turning up. Gardner broke from established routine to launch himself upon his self-chosen method of traveling light through life. But to fame Gardner is as indifferent as he is to fortune. Those cards. Chicagoans who are not too stultiﬁed to have recently enjoyed a Christmas-time day on Marshall Field and Company’s toy ﬂoor may remember Gardner as the “Mysto-Magic” set demonstrator for the past two years. Martin Gardner is a Robinson Crusoe by choice. I decided I didn’t want to teach. ﬁlling some twenty-ﬁve shoe boxes. The rest of the year ﬁnds him periodically down to his last ﬁve dollars. family that is the essence of upper middle-class substantiality. middling man with a thin face saturnined by jutting. upon a series of thousands of ﬁling cards. The card entries run from prostitutes to Plautus — which is not too far — and from Plato to police museums. Desperate Desmond fashion. diversiﬁed. jetted eyebrows and spading chin. These clippings he mounted. Possessor a few years ago of a large.” [24] *** Gardner returned to his home state after college to work as assistant oil editor for the Tulsa Tribune. [17] *** He returned to the Windy City ﬁrst as a case worker for the Chicago Relief Agency and later as a public-relations writer for the University of Chicago. I wanted to write. for competent magicians consider him one of the ﬁnest intimate illusionists in this country today. Oklahoma.

and occasional dabblings in writings here and there have made him even more well known as an “idea” man for small novelty houses and children’s book publishers.. the University of Chicago. accept Mr. but he doesn’t like being branded. then based in Chicago.. The editor invited the starving writer for lunch at a good restaurant. Gardner’s four or ﬁve hundred? [4] *** He was appointed yeoman of the destroyer escort in the North Atlantic “when they found out I could type.. Christ. and the night is large and full of wonder. but it has taken world chaos to make his father say that his oldest son is perhaps the sanest of his family. Those mental plots evolved into imaginative short stories that he sold to Esquire magazine.” said the soft-spoken Gardner.6 D. He arrived with a large suitcase full of puzzles! Puzzles had been a hobby with him. He tours the world pulling rabbits out of hats.” “I amused myself on nightwatch by thinking up crazy plots. Gardner was hanging around his alma mater.” [5] *** Martin Gardner ’36 is a professional [sic] magician. and he thinks Plato.. but where to park them while he was peregrinating over the globe was a problem.Still ﬂush with musteringout pay. Those sales marked a turning point in Gardner’s career. . writing and taking an occasional GI Bill philosophy course. His breakcame when he sold a humorous short story called “The Horse on the Escalator” to Esquire magazine. His personal philosophy has been described as a loose Platonism. [18] *** His career as a professional writer started in 1946 shortly after he returned from four years on a destroyer escort in World War II.. “pitchmen’s” novelties.He called up Mr. who had the largest collection he had ever heard of. If he were to rest his thoughts upon one quotation it would be Lord Dunsany’s: “Man is a small thing. RICHARDS for a magic trick or a sales-promotion angle to any one of a half-dozen companies who look to him for specialties. straight magic and card tricks.Would Mr. might object with sound reason.During the past few months a determined outpouring of ideas for booklets on paper-cutting and other tricks. Christ and asked if he might come out to Christ’s home. too. When Professor Jay Christ (Business Law) was exhibiting his series of puzzles at the Club late last Fall Gardner chanced to be in town and saw one of the exhibits.. To Gardner’s family his way of life has at last become understandable.

I wrote about those three things in an article for the Antioch Review. then expanded that article into a book by adding chapters on dowsing.. almost every night. Paul Curry. Judge George Starke. I suppose I really didn’t get into it seriously until I wrote my ﬁrst book.” [15] About 1947. I was inﬂuenced by the Dianetics movement. Children’s Digest. “I was brought up on John Martin’s magazine. I remember the hatcheck girl looking askance when I handed her the ﬁlthy rag. he has a good background for the task. It took a long time for the book to start selling. I’ve been fascinated by crazy science and such things as perpetual motion machines and logical paradoxes. Frank Baum “Oz” books. and Bill Simon. It was Simon who introduced Martin and Charlotte (Mrs. Fads and Fallacies in the Name of Science. he moved to New York where he soon became friends with such well-known magic devotees as the late Bruce Elliot. Gardner needed a regular income in those years and with his usual serendipity found a job that was just right for him: contributing editor for Humpty Dumpty’s Magazine. now called Scientology.. and regards Baum as “the greatest writer of children’s ﬁction yet to be produced by America. early ESP research. another magic friend. ﬂying saucers.. the . Gardner) and served as best man at their wedding. the hollow-earth theories.“Those were good years at Humpty.” Gardner recalls. I had friends who were sitting in Wilhelm Reich’s orgone energy accumulators. Piggity’s.MARTIN GARDNER: A “DOCUMENTARY” 7 “The only coat I had. I’ve always enjoyed keeping up with those ideas. which was then promoted by John Campbell in Astounding Science Fiction. He designed features and wrote stories for Humpty. the book would be mentioned on the show by some guest who was attacking it. Persi Diaconis. and Polly Pigtails. He says that he is a great admirer of the L. and so on.” [20] *** Their ﬁrst son was born in 1955 and their second three years later. Dai Vernon. and one of the greatest writers of children’s fantasy in the history of world literature. Clayton Rawson. Atlantis. pyramidology.” He adds. “was an old Navy pea jacket that smelled of diesel oil.” [15] *** Although Gardner is a brand-new children’s writer. I was astonished at how rapidly the thing had become a cult. performed the ceremony. too. having read all of them as a child.And the Immanuel Velikovsky business had just started. but it really took off when they started attacking it on the Long John Nebel Show. For about a year. [12] *** “Ever since I was a boy.

but. But it was a lucky idea because that’s been the best seller of all my books. because that’s all the mathematics Gardner knows. When I ﬁrst met Asimov. but the mathematics in them has never gone much beyond second-year college level. So it wasn’t really too far aﬁeld from recreational math. because the two books are ﬁlled with all kinds of mathematical jokes. “But I knew of some famous math books. One of them intrigued him with a so-called hexaﬂexagon — a strip of paper folded into a hexagon.. Through the years they have grown far more sophisticated in logic.“I’ve never made a discovery myself. crammed with math journals in every language. unless by accident.” Gardner says he never does any original work.” Gardner reminisces. I didn’t own a single math book..” he explains with a shrug and a gesture toward the long rows of bookshelves. the column was read mostly by high school students (he could tell by the mail). of course. as he studied the enormous literature on recreational math and learned more about it. you fool people. all doing magic tricks.” [14] *** At ﬁrst. and Gardner was asked to do a monthly column.. Newman’s The World of Mathematics had just been published. “I’m really a journalist. gradually. [16] *** “The Annotated Alice. I asked him if he was a professor at Boston University. James R.He said no and . because Lewis Carroll was. I was lucky there in that I really didn’t have anything new to say in The Annotated Alice because I just looked over the literature and pulled together everything in the form of footnotes.”His ﬁrst columns were simple.“There would be 50 magicians or so. “At the time. as you know. and I jumped at the chance. which led to an article on combinatorial geometry in Scientiﬁc American in December 1956. where graduate students had invented it. does tie in with math. Gardner drove to Princeton. If you write glibly. demonstrating the appeal of math for the masses. RICHARDS inﬂuence of which can be seen in some of the activity pages which I am contributing to Humpty Dumpty. which was a big part of the puzzle page at Humpty. [23] *** He got into mathematics by way of paper folding.” he recalls.Fascinated. he watched his readers become more sophisticated. “This kind of just happened. Gardner says.8 D. he simply popularizes the work of others. asked me .” [6] *** Every Saturday a group of conjureres would gather in a restaurant in lower Manhattan. that line one alcove in his study. which turns inside out when two sides are pinched. a professional mathematician. A friend showed him a novel way to fold a strip of paper into a series of hexagons.

His wife interjected: “The fact is he doesn’t want to do . Professors at Stanford University have just programmed a computer to carry to the millionth digit.” [20] *** His “Mathematical Games” column in Scientiﬁc American is one of the few bridges over C. would be the number 5.. He was asked how he knew he was ill at ease if he had never done it.Let me say that I think not knowing too much about a subject is an asset for a journalist. For example. I think of myself as a journalist who knows just enough about mathematics to be able to take low-level math and make it clear and interesting to nonmathematicians. I have also managed to provoke some outspoken enemies.D. [8] *** “I particularly enjoy writing columns that overlap with philosophical issues. The late Jacob Bronowski was a devotee. whom he calls “an invented philospher.” Nevertheless. P.” [14] *** “I’m very ill at ease in front of an audience. and that stumped him for a moment. “not all my readers are fans. Matrix that the millionth digit of . I said I didn’t have one and he looked startled. ‘you just read books by the professors and rewrite them?’ That’s really what I do. and “proof” by a ﬁctional Dr.MARTIN GARDNER: A “DOCUMENTARY” 9 where I got my Ph.” In the forefront are the credulous victims of Gardner’s recent hoaxes: an elaborate treatise that demonstrated the power of pyramid-shaped structures to preserve life and sharpen razor blades.” Gardner said. not a liability. poet W. Snow’s famous “gulf of mutual incomprehesion” that lies between the technical and literary cultures. In his novel Ada. I did a column a few years ago on a marvelous paradox called Newcomb’s paradox. as the mathemagician admits.’ he said. in decision theory.. It’s a very intriguing paradox and I’m not sure that it’s even resolved.” [11] *** “I can’t think of any deﬁnition of ‘mathematician’ or ‘scientist’ that would apply to me. To everyone’s surprise — especially the hoaxer’s — the number turned out to be 5. The last scoop that I got was when I heard about a public-key cryptography system at MIT. Vladimir Nabokov pays a twinkling tribute by introducing one Martin Gardiner. ‘You mean you’re in the same racket I am. And then every once in a while I get a sort of scoop. Auden constantly quoted from Gardner.. if it were ever computed. The great secret of my column is that I know so little about mathematics that I have to work hard to understand the subject myself. Maybe I can explain things more clearly than a professional mathematician can. I realized what a big breakthrough this was and based a column on it. H. and that was the ﬁrst publication the general public had on it.

A hobby I acquired late in life is playing the musical saw. I have written a number of small books that are sold only in magic shops.” he remarks. and H.” Gardner says. and other books about math and science. G. and enumerates them. “From Chesterton I got a sense of mystery in the universe.” [23] *** Gardner takes refuge in magic. during which he gave his computer to one of his two sons. the wise visitor will keep his money in his pocket. From each Gardner has culled some wisdom. why anything exists. for that matter. My second major hobby as a child was chess. Wells.” he expounds.. I much more enjoyed writing the book with Carnap. and most enjoyed playing was tennis. “From Wells I took his tremendous interest in and respect for science.” [19] *** “My earliest hobby was magic. there are G. Gardner peers at the world with such wide-eyed wonder as to inspire trust in all who meet him. RICHARDS it the same way he doesn’t want to shop for clothes. K. A lot of it is a waste of time.. “Then one day I was doing dishes with my wife. Rudolf Carnap. The sport I most enjoyed watching as a boy was baseball. Miguel de Unamuno. I would have had little time for anything else. a fax or answering machine). and I have retained an interest in it ever since.” [23] *** “In a way I regret spending so much time debunking bad science.10 D. or The Ambidextrous Universe. [10] *** “Certain authors have been a big inﬂuence on me. but I stopped playing after my college days for the simple reason that had I not done so.” [26] . Although I have written no general trade books on conjuring. To my knowledge he’ll shop only for books. and I continue to contribute original tricks to magic periodicals. He once did — and got hooked playing chess on it. Charles S.” . The retinal retention lasted about a week. William James. “I don’t believe God interrupts natural laws or tinkers with the universe. Chesterton. “I’m a scissors-and-rubber-cement man. “I don’t think there’s any way to prove the existence of God logically. From James he derived his notion that belief in God is a matter of faith only. at which he probably is good enough to earn yet another living.” [13] *** Gardner himself does not own a computer (or. Besides Plato and Kant. But when Gardner brings out his green baize gaming board.” he recalls. Peirce. and I looked down and saw the pattern of the chessboard on the surface on the water.

28.” [25] *** “There are. October 1940. One of the most damaging examples of pseudoscience is false memory syndrome. [2] Martin Gardner. He was poor for a very long time until he ﬁt into something. No. p. and he starved. The Cryptogram. He knew what he wanted to do. [6] LaVere Anderson. “Escape to Bohemia.’ ” [7] *** “In the medical ﬁeld [scientiﬁc ignorance] could lead to horrendous results. The mysterians are not an organized group or anything.” Chicago Daily News. 27–29. p. [8] [Stefan Kanfer]. 3. 2. vol. p..” [21] *** “Martin never sold out. 4.” Time.” Diaconis said. pp. April 28... [Letter]. “He would never do anything that he wasn’t really interested in. . [7] Bernard Sussman. April 1930. 63. educated people can succeed in shedding light into these areas of prejudice and ignorance. April 1932. p.” [19] References [1] Martin Gardner. I’m on the board of a foundation exposing this problem. and always have been. 1939. vol. We don’t hold meetings. for as Voltaire once said: ‘Men will commit atrocities as long as they believe absurdities. [3] Martin Gardner. Hopefully.” Pulse [University of Chicago]. 1. August 22.” Southwind [Miami-Dade Junior College]. Fall 1968.People who don’t understand the difference between a controlled experiment and claims by some quack may die as a result of not taking medical science seriously.. 7–11. destructive pseudo-scientiﬁc notions linked to race and religion. these are the most widespread and the most damaging. “The Mathemagician.” Tulsa World (Sunday Magazine). pp. “Martin Gardner’s Tongue-in-Cheek Science. It really is wonderful that he achieved what he achieved. [4] Tower Topics [University of Chicago].. p. April 21. Mysterians believe that at this point in our evolutionary history there are mysteries that cannot be resolved.. 16–17. 1. no. no. [9] Betsy Bliss. 1975. “Under the Reading Lamp. 1975. 7. 8. “Now It Is Now It Isn’t. 2. Sharpless Hickman. p. [5] C. 1119.” Science and Invention. September 1934. “A Puzzling Collection. pp.” Hobbies. 1957. “Exclusive Interview with Martin Gardner.MARTIN GARDNER: A “DOCUMENTARY” 11 *** “As a member of a group called the mysterians I believe that we have no idea whether free will exists or how it works.

1976. 1997. 1979. pp. March/April 1998. Material taken from the Afterword. [23] Philip Yam. vol.” Time-News [Hendersonville.” Washington Post.” Bookletter. [17] Sara Lambert. “Martin Gardner: A Writer of Many Interests. 5. pp. pp. 2. p. 3. “A Conversation with Martin Gardner. [15] Rudy Rucker. 4. 1976. [25] Michael Shermer. 16. [20] Scot Morris. 32–37. [21] Lawrence Toppman. vol. December 1995.12 D. vol. 80–86. 58. pp.” The Charlotte [North Carolina] Observer. [11] Sally Helgeson. 10. no. 4. December 6. 40–41. 3.” Scientiﬁc American. pp. 1. 1981. 1981. [16] Jerry Adler and John Carey. “Mastermind. “Every Day. pp.” Los Angeles Times. 1B–2B.” Skeptic. April 1978. Dover. “The Magician of Math. 66–69. Impresario of Mathematical Games. “The Annotated Gardner. 89. The Flight of Peter Fromm. “A Great Communicator of Mathematics and Other Games: A Conversation with Martin Gardner. 36–40. pp. 2. December 12. 34–39. [22] Martin Gardner.” Something About the Author. vol.” Skeptical Enquirer. NC]. no. no. 6. pp.” Omni. “Martin Gardner. 1979. 1E. [18] Lynne Lucas. p.” Two-Year College Mathematics Journal.” Newsweek. 47– 48. 10–21. vol. [13] Anne Commire. no. pp. pp. pp. 1993. . 6E. p. 38.” Mathematical Intelligencer. [14] Anthony Barcellos. 4.” Linking Ring. [26] Kendrick Frazier. 19. July/August 1981. 117–119. January 1982. [19] Lee Dembart. 1–10. vol.” Science 81. vol. “The Math-e-magician of Hendersonville. 1981. 1994. “[Martin Gardner]. vol. “The Puckish High Priest of Puzzles. “Magician of the Wonders of Numbers.” The Greenville [South Carolina] News. “The Mathematical Gamester. pp. [12] John Braun. December 9. “A Mind at Play. 56–61. pp. 1981. November 16. December 5. “Interview: Martin Gardner. 4. 232–244. 101. June 20. 8. March 11. no. [24] Istvan Hargittai. “Martin Gardner. RICHARDS [10] Hank Burchard. no.

Ambrose. this fact has often puzzled me! How could anyone with the brilliance to write the Sherlock Holmes stories ever get involved with spiritualism—and in such a crazy way? Ambrose: You mean that you have doubts that Doyle wrote the Holmes stories? Byrd: Of course not! That thought has never crossed my mind! All I said was that I ﬁnd the situation puzzling. and of course I’ve heard of Martin Gardner. I guess the answer is that Doyle went senile in his last years? Ambrose: No. Professor Ambrose: Have you ever read the book Science: Good. He 13 .The year is 2050 A. rational.D. He was a very famous science writer of the last century. no! Gardner correctly pointed out that all the available evidence shows that Doyle remained quite keen and active to the end. Bad. and Doyle Raymond Smullyan SCENE I . Byrd: Oh? Ambrose: The chapter is titled “The Irrelevance of Conan Doyle”. Byrd: To tell you the truth. and Bogus. by Martin Gardner? Professor Byrd: No. scientiﬁc mind to write the Sherlock Holmes stories could possibly have spent his last twelve years in a tireless crusade against all rationality—I’m talking about his crazy involvement with spiritualism. Gardner. Why do you ask? Ambrose: Because the book contains one weird chapter — It is totally unlike anything else that Gardner ever wrote. He actually advances the thesis that Conan Doyle never wrote the Sherlock Holmes stories—that these stories are forgeries. Byrd: That is weird! Especially from Gardner! On what does he base it? Ambrose: On absolutely nothing! His whole argument is that no one with the brilliant. I’ve heard of it.

his disturbance went much deeper. Byrd: But now something else puzzles me: Martin Gardner was no fool. . Byrd: All tight. it’s just as implausible as Gardner’s idea that Doyle never wrote the Holmes stories at all. but now your explanation of the apparent contradiction between Doyle the rationalist and Doyle the crank makes some sense. Don’t you know that he believed that the famous Harry Houdini escaped from locked trunks by dematerializing and going out through the keyhole? What’s even worse. Byrd: I just thought of another idea! Perhaps Doyle was planning all along to foist his spiritualism on the public and started out writing his rational Holmes stories to gain everybody’s conﬁdence. whamo! Ambrose (After a pause): That’s quite a cute idea! But frankly. Doyle insisted that Houdini was lying! If that’s not psychotic paranoia. So senility is not the explanation. he was surely one of the most interesting writers of the last century. Now. he absolutely refused to believe Houdini when he said that there was a perfectly naturalistic explanation for the escapes. how do you explain the mystery? Ambrose: The explanation is so obvious that I’m amazed that anyone can fail to see it! Byrd: Well? Ambrose: Haven’t you heard of multiple personalities? Doyle obviously had a dual personality—moreover of a serious psychotic nature! The clue to the whole thing is not senility but psychosis! Surely you know that some psychotics are absolutely brilliant in certain areas and completely deluded in others. when the public was convinced of his rationality. SMULLYAN also pointed out that Doyle’s interest in spiritualism started much earlier in life than is generally realized. I never had the slightest doubt that Doyle wrote the Holmes stories. As I said. how could someone of Gardner’s caliber ever entertain the silly notion that Doyle never wrote the Holmes stories? Ambrose: To me the solution is obvious: Martin Gardner never wrote that chapter! The chapter is a complete forgery. Ambrose: I’m glad you realize that. then. What better explanation could there be? Byrd: You really believe that Doyle was psychotic? Ambrose: Of course he was! Byrd: Just because he believed in spiritualism? Ambrose: No. Then.14 R. what is? Byrd: I guess you’re right. I have no idea who wrote it.

The title is ”Gardner and Doyle”. But that one chapter sticks out like a sore thumb—not just with respect to the other chapters. I’ll send you a copy. the more I think about it. whatever remains. A person of Gardner’s caliber could never have written anything like that! Byrd: Now just a minute. We can only hope that future research will answer the question of how that strange chapter ever got into the book. In fact. nobody in his right mind could believe that Gardner actually wrote that chapter. And as Holmes wisely said: Whenever we have eliminated the impossible. AND DOYLE 15 but it was certainly not Martin Gardner. your theory strikes me as most improbable! Ambrose: I agree with you wholeheartedly. the theory is most improbable. you must be right! It is certainly not conceivable that anyone as rational as Gardner could entertain such a stange notion. are you talking about the whole book or just that one chapter? Ambrose: Just that one chapter. . But now I think you’ve made a very important historical discovery! Why don’t you publish it? Ambrose: I am publishing it. Now. Byrd (After a long pause): I guess you’re right. however improbable must be the truth. he couldn’t possibly have written such a chapter. And so I am forced to the conclusion that Martin Gardner never wrote that chapter. GARDNER. but completely out of the question. but an isolated chapter of a book? How could the chapter have ever gotten there? Could Gardner have hired someone to write it? That seems ridiculous! Why would he have done a thing like that? On the other hand. Will you please explain one thing: How did the chapter ever get there? No. No. but he certainly never wrote that chapter. It will appear in the June issue of the Journal of the History of Science and Literature. But the alternative that Gardner actually wrote that chapter is not just improbable. I believe that he did exist.AMBROSE. why would Gardner have ever allowed the chapter to be included? Or could it possibly have gotten there without his knowledge? That also seems implausible. they are perfectly consistent in spirit with all the sensible things that Gardner ever wrote. All the other chapters are obviously genuine. I don’t go as far as some historians who believe that Martin Gardner never existed. I can see how an entire book by an alleged author might be a forgery. I don’t see how there can be the slightest doubt that this chapter is a complete forgery! Byrd: But that raises serious problems! All right. but in relation to all of Gardner’s writings. But surely.

“The Irrelevance of Conan Doyle”? Cranby: Oh yes! As a matter of fact that is the strangest chapter of the book and is quite unlike anything else Gardner ever wrote. Broad: And do you recall the chapter. Of course. and Bogus? Cranby: Oh. then I won’t get it for a couple of days. come on now! That’s no objection! It’s obvious that Doyle. I’m familiar with much of Ambrose’s excellent work. had an insane streak that simply got worse through the years. are you familiar with the Ambrose paper. where did you send it? Broad: To your Connecticut address. and Doyle”? Professor Cranby: No. with all his brilliance. Broad: Do you believe that Doyle wrote the Holmes stories? Cranby: Of course! Why should I doubt it for one minute? Broad: Then how do you answer Gardner’s objection that no one with a mind so rational as to write the Holmes stories could possibly be so irrational as to get involved with spiritualism in the peculiarly anti-rational way that he did? Cranby: Oh. Cranby: Good God! That’s ridiculous! That’s just as crazy as Gardner’s idea that Doyle didn’t write Holmes. Of course Gardner wrote that chapter! Broad: Of course he did! . SMULLYAN SCENE II . What is it about? Broad: You know the twentieth century writer.16 R. Science: Good. Doyle wrote the Sherlock Holmes stories! Broad: I heartily agree! Cranby: The one thing that puzzles me—and I remember that it puzzled me at the time—is how someone like Martin Gardner could ever have believed such an odd-ball thing! Broad: Ah. certainly. Gardner. Bad.One Hundred Years Later Professor Broad: Did you get my paper. Why do you ask? Broad: Well. What is it about? Broad: Well. you remember his book. ”Ambrose. ”Gardner and Doyle”? Cranby: No. that’s the whole point of Ambrose’s paper! His answer is simply that Gardner never wrote that chapter—the chapter is just a forgery. He seriously maintained that Conan Doyle never wrote the Sherlock Holmes stories. but not this one. Martin Gardner? Cranby: Of course! I’m quite a fan of his. I think I have just about everything he ever wrote. Cranby: Oh.

so well known and highly respected for the mathematical games column he wrote for years for Scientiﬁc American. Whys and Wherefores (University of Chicago Press. The Hunting of the Snark. his numerous puzzle books. How could he ever believe anything that bizarre? Broad: Ah. is one of Gardner’s pseudonyms. by the way. GARDNER.” . a discovery of a fatal ﬂaw in the theory of relativity. he reported the discovery of a map that required ﬁve colors. that’s where my paper comes in! I maintain that Ambrose never wrote that paper—it must be a complete forgery! SCENE III . and a lost manuscript proving that Leonardo da Vinci was the inventor of the ﬂush toilet. AND DOYLE 17 Cranby: But what puzzles me is how such a sober and reliable historian as Ambrose could ever believe that Gardner didn’t write that chapter.AMBROSE. The Ancient Mariner. an opening move in chess (pawn to Queen’s rook four) that guaranteed a certain win for white. and his Whys of a Philosophical Scrivener—how come he wrote such a crazy chapter as “The Irrelevance of Conan Doyle”? This troubled me for a long time. 1989). in his April 1975 column in Scientiﬁc American. In Martin’s book.A Hundred Years Later (To be supplied by the reader) Discussion: How come this same Martin Gardner. The Flight of Peter Fromm. is reprinted a scathing review of his The Whys of a Philosophical Scrivener by a writer named George Groth. and Casey at the Bat—not to mention his religious novel. until Martin kindly informed me that the whole thing was simply a hoax! Martin is really great on hoaxes—for example. The review ends with the sentence “George Groth. his annotated editions of Alice in Wonderland.

Technical proﬁciency is a worthy goal. But Martin Gardner. From Martin Gardner I learned of logical and language paradoxes. both groups being beer lovers. The competitions led me to believe I had a talent. The former was ﬁlled with one dreary numerical problem after another. with the national mood for education reform. It was true. mathematics is fun. This fundamental truth was learned from Martin Gardner when I was young and impressionable. and for an adolescent unsure of himself and his place in the world. the techniques of integration for a later course. In fact. I did get to the fun stuff.A Truth Learned Early Carl Pomerance It was in high school that I decided to be a mathematician. and it is a truth I carry in my heart. perhaps. I surely try not to. I learned of hexaﬂexagons (I still have somewhere in my cluttered ofﬁce a model of a rotating ring I made while in high school). through his books and columns. but now I see another side of the story. this was no small thing. I learned sneaky ways of doing difﬁcult computations (a round bullet shot through the center of a sphere comes to mind). say. Today. I ﬁgured that if I could just stick it out long enough. Do I duplicate the school experiences I had with my students? Well. sooner or later I would get to the fun stuff. led me to the more important lesson that. I would be remiss if I didn’t cover the topic. it seems the rest of the country is ﬁnally learning this truth too. The credit (or. and when my students need to know. blame!) for this can be laid squarely on mathematical competitions and Martin Gardner. But I know also that the driving engine behind mathematics is the underlying beauty and power of the subject and that this indeed is the reason it is a subject worth studying. A good part of my job now is being a teacher. and I learned of islands populated only by truth tellers and liars. such as the condemned prisoner who wasn’t supposed to know the day of his execution (I don’t think I understand this even now!). Welcome aboard. there seemed to be two completely different kinds of mathematics: the kind you learned in school and the kind you learned from Martin Gardner. while the latter was ﬁlled with ﬂights of fancy and wonderment. above all else. The contrast with my teachers in school was striking. 19 . This colorful world stood in stark contrast to school mathematics.

Martin Gardner = Mint! Grand! Rare! Jeremiah Farrell I was not surprised to discover the wonderful equation in the title (that so aptly describes the person) since logology and numerology are — according to Dr. Six is the ﬁrst perfect number. Matrix notes that 13 is an emirp since its reversal is also prime. while seven is the only odd prime that on removal of one letter becomes EVEN. Dr. Matrix — two faces of the same coin. Matrix has informed me that a better statement is “Martin Gardner: a man and rarer enigma. (It can be no coincidence that the even number SIX.” (He also found that the square had contained a prediction for the 1980 presidential election: “Reagan ran. an enigma. upon subtraction of the same letter becomes the odd number IX. His brilliant future as the world’s premiere mathematical wordsmith had been fated since the day he was christened. The ﬁrst and last names have six and seven letters. 21 .” Not to be outdone (as usual). In!”) Figure 1. page 88) that the ¿ ¢ ¿ word square in Figure 1 spells out with chess king moves the laudatory phrase “Martin Gardner. 1981. For instance: ´½µ There are 13 letters in MARTIN GARDNER. It was Mr.) ´¾µ I had remarked in an issue of Word Ways (May. Since then Matrix and I have marveled over the inevitability of Gardner’s career choice. Gardner who introduced me to the wiley doctor over ten years ago.

Dr. but the point is clear. ´¿µ . not often found in squares composed from names. Choose three letters. FARRELL Figure 2.22 J. A common English word will be the result. This tribute could be continued. exactly one from each row and one from each column. A different arrangement of the nine letters produces the square shown in Figure 2. Matrix claims this to be an unusual property. Martin Gardner is indeed Mint! Grand! Rare! He has my love.

seen as an Army Sergeant Major. Friday night At a minute to twelve Eve spin. right? A week = 7 days corresponds to 14 Billion Years 1daycorresponds to2BillionYears(USA) 1 minute corresponds to 2 Million Years 1 Millisecond is 23 Years 23 . Time. Bang!. our Galaxy — done! ½¼ ¼ meter = 1 meter = a Human Body ½¼ meters = Earth’s diameter ½¼½ meters = outer Solar System ½¼¾½ meters = Galaxy’s diameter ½¼ ¾ meters = Universe. Light Sun and Earth. form up.Three Limericks — On Space. and a bit more ½¼ meters = Nucleus of a human cell ½¼ ½ meters = Atom’s nucleus Time The Creator. the Earth Sun’s System. and Speed Tim Rowett Space Seven steps each ten million to one Describe the whole space dimension The Atom. Cell’s girth Our bodies. You. barks out his orders for the week. Adam delve In the last millisecond. First thing on Monday.

roughaveragespeedofEarth 700.000mph.000 mph turning speed of Galaxy 1.24 T. ROWETT Speed A child cycles ‘round the schoolyard Which lies on the Earth turning hard The Earth rounds the Sun As Sol does “the ton” And our Galaxy ﬂies — Gee! I’m tired 7 mph — child cyclist 700mphEarth’ssurface(NorthAfrica) 70.4 million mph Galaxy’s speed through the debris of the Big Bang .

and you can only tip it onto squares that contain letters. then you can tip it onto a square with an Ä. If a 1. This is my manuscript version of the maze. 3. or 5 is on top. If a 4. The next page shows one of the mazes from my book. These columns presented rolling cube puzzles by Roland Sprague and John Harris. Mad Mazes. but it takes 66 moves. I got the original idea for this maze from remembering columns that Martin wrote in December 1963. In my maze. but later I realized they were more like mazes. I wrote half the dopey stories and I sort of like some of them. At the time I thought these were puzzles. Around 1980 I started creating more of these things (which I now think could best be described as “mazes with rules”). and March 1975. As Martin’s columns said. place a die on the square marked START. you can tip it onto a square with an À. What you have to do is tip the die off the starting square. He used another of my puzzles in the November 1963 column — this one involved traveling in three dimensions through a ¢ ¢ grid. you should think of a cube as a large carton that is too heavy to slide but that can be tipped over on an edge. or 6 is on top. November 1965. and Ú Ò. 4. you can tip the die onto a square with an . and in 1990 I had a book of them published. or 6 is on top. or 3 is on top of the die. then ﬁnd a way to get it back onto that square. (Actually. If a 2. The letters stand for ÐÓÛ. 27 . Ó . I won’t give the solution.) I chose this particular maze because it illustrates the cross-fertilization that Martin’s columns created. If (and only if ) a 1. 5. Position the die so that the 2 is on top and the 6 is facing you (that is. . before my publisher added art work and dopey stories. You can tip the die from one square to the next. 2.A Maze with Rules Robert Abbott In his October 1962 column. The puzzles involved tipping cubes from one square to another on a grid. Martin Gardner presented a puzzle of mine that involved traveling through a city that had various arrows at the intersections. you can tip the die onto a square with an Ç. the 6 faces the bottom edge of the page).

Since then the concept has grown. While you’re there. but you can also download it off my website. ABBOTT Maze Addendum. The diagram should be at least 6 inches square to have a die roll across it. several of the mazes were built as adjuncts to large cornﬁeld mazes. Oops! My diagram is too big for this book. In the summer of 1998.28 R." The ﬁrst of these walk-through mazes appeared at the Gathering for Gardner in January 1993. . Go to ØØÔ »» ÓÑ º ØØºÒ Ø» ÖÓ Ø ÓØØ»ÖÓÐÐº ØÑÐ. December 1998. You might try enlarging it on a copier. I have a long write-up (with pictures) of something called “walk-through mazes-with-rules. check out the rest of the site.

. . ( 1 Corinthians 15 : 7 ) ( Luke 10 : 20 ) ( 1 Kings 7 : 33 ) ( 2 Samuel 22 : 5 ) ( Judges 8 : 30 ) ( Exodus 1 : 14 ) ( Genesis 47 : 28 ) Puzzle #1. . made their . . .) 29 . . because your . When the . Ë Ë. But the tableau lists only the verse numbers. . Ë Ë. . aka Lewis Carroll. bitter with . . . never backtracking in Biblical order. not the words. . . And Jacob Â Å Æ Å Æ Î Ï Î ÏÁÎ ÄÁÎ ÄÁÎ Ë. . Many people consider the Bible to be a story of transition from ÏÊ ÌÀ to ÁÌÀ. . in which one word metamorphoses into another by changing a letter at a time. . had many . seen of . the words should all be present in the Bible that was used in Dodgson’s day. . . . is a six-step sequence that we might call Jacob’s Ladder. invented a popular pastime now called word ladders. what are the missing words? (Remember to use a King James Bible for reference. ÌÀ Æ. . was strictly decreasing. . . in the land . . Knuth Charles Lutwidge Dodgson. Dodgson also was quite familiar with the Bible. . that there’s a Biblical word ladder corresponding to such a transition. Here. and their . the sequence of verses in this ladder is strictly increasing through the Old Testament. Jacob’s Ladder. More precisely. And his . on the other hand. the King James translation. ÌÀ Ì. As an ordained deacon of the Church of England. not a newfangled translation! Incidentally. So let’s play a game that combines both activities: Let’s construct word ladders in which all words are Biblical. ÌÀÁÆ. . Ë then of all . for example. and their . We can go from ÌÀÁË to ÌÀ Ì in three such steps: ÌÀÁË. because ‘James’ is a form of ‘Jacob’: . . . in fact.Biblical Ladders Donald E. of death . . . are written . The following tableau shows.

. . . . Go from ËÏÇÊ to (plow) ËÀ Ê in four steps: . charity shall Æ Ã .. in eight steps? A suitable middle verse is provided as a clue... . .. .. . ÇÎ Ê the multitude . .. .. if you have a good concordance or a King James computer ﬁle.. So neither words nor verse numbers will be given this time...... by his D. ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ÁÌÀ.. . .. ... they sware. ººº ººº ººº ººº ººº ººº ººº ( Genesis 3 : 7 ) ( ) ( ) ( ) ( Luke 17 : 27 ) ( ) ( ) ( ) ( 1 Peter 4 : 8 ) . . .... E... ËÀ Ê . . .. . ººº ººº ººº ( Micah 4 : 3 ) ( ) ( ) ( ) ( 1 Samuel 13 : 20 ) (Hint: In the time of King James.. ..... . . . .. .30 . Yea also. . people never swore. .. .. Of course #2 was also pretty easy... .. that his .. and his .. . . ( Genesis 39 : 19 ) ( Genesis 40 : 2 ) ( Exodus 24 : 4 ) ( Exodus 34 : 1 ) ( Leviticus 13 : 3 ) ( Leviticus 14 : 46 ) ( Leviticus 25 : 29 ) ( Deuteronomy 22 : 21 ) ( Joshua 11 : 4 ) ( 1 Samuel 13 : 20 ) ( 1 Samuel 15 : 3 ) ( 1 Samuel 26 : 13 ) ( 1 Kings 10 : 15 ) ( Psalm 10 : 14 ) ( Isaiah 3 : 17 ) ( Isaiah 54 : 16 ) ( Isaiah 54 : 17 ) ( Habakkuk 2 : 4 ) Puzzle #2. . KNUTH ÏÊ ÌÀ was kindled... How about going from Æ Ã to ÇÎ Ê. . . .. . ... . . . Of course #1 was too easy. . . .. and they .. . . . . . . .. ... every man his ËÏÇÊ against . not lift up a . they were ... ..) Puzzle #3.. .. . .

. . seventy times .) Matthew 11 : 11 Judges 9 : 9 Mark 12 : 42 Ecclesiastes 9 : 3 Luke 9 : 58 Exodus 28 : 33 Lamentations 2 : 13 1 Peter 5 : 2 Genesis 34 : 21 Acts 20 : 9 .... . .. . . . ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ( Matthew 18 : 22 ) ( Matthew 13 : ) ( John 4 : ) ( Numbers 33 : ) ( Deuteronomy 29 : ( 1 Samuel 15 : ) ( Job 16 : ) ( 1 Kings 7 : ) ( Romans 9 : ) ( Leviticus 9 : ) ( Acts 27 : ) ( Jeremiah 10 : ) ( Matthew 13 : ) ( Ezekiel 24 : ) ( Matthew 18 : ) ) Puzzle #7. Find a Biblical word ladder from ÀÇÄ to ÏÊÁÌ... . . Complete the following Biblical ladder.. Here’s one that anybody can do. . using only words from the King James Bible verses shown. .) Construct a 12-step Biblical ladder from ÇÊ Ë (Judges 3 : 28) to ÊÇÄÄË (Ezra 6 : 1)... Puzzle #6.... Puzzle #5. ... . with only a Bible in hand.. Ë Î Æ. (The words will read the same down as they do across... Therefore . . ... ... a change of pace: Construct ¢ word squares. .. unless you have special resources. which “comes back on itself” in an unexpected way. Of course #5 was too hard... (For worshippers of automobiles. .. ...BIBLICAL LADDERS 31 Puzzle #4. Finally.

. Other words can now be ﬁlled in. . . . . First observe that the word in Luke 17 : 27 must have at least one letter in common with both Æ Ã and ÇÎ Ê. the LORD. and the step after Æ Ã must be Ï Ã . Yea also. year after . . . . . . . in her . not lift up a . . . . against two . . . ËÅÁÌ ËÅÁÌÀ Ë ÁÌÀ ÁÌÀ. . and I will . . on the . all the . . Here’s a strictly decreasing solution: . . ÏÀÁÄ ÏÀÇÄ ÏÀÇÊ ËÀÇÊ ËÀ Ê . which came . since neither Ã nor ÆÁÃ nor Æ Î nor Æ Î Æ is a word. . Puzzle #3. being between . . . that it is . . by his ÏÊ ÌÀ ÏÊÇÌÀ ÏÊÇÌ ÏÊÁÌ ÏÀÁÌ . in multitude. . . . . beside Eloth. . . And Moses . ËÈ Ê ËÈ ËÈÁ ËÈÁÌ . . So it must be ÏÁÎ Ë or ÁÎ Æ.. . . The Lord GOD hath . .32 D. to requite . and his . . . . . . . . . . . . the sea . a great . . . Lord will . . . . man his . have. . . with a scab . Thus the middle word must be ÏÁÎ Ë. against . that bloweth . still avoiding forward steps. and 1 Kings 4 : 29 could also be used for ËÀÇÊ . was kindled. . . is of me. merchants. sheep that are even . . . created the . . . . upon these . and his . of the Red . play the . . . . by himself. . . . ËÀÇÊ ËÀ Ê . . . all the . . . . . . . E. . Pharaoh was . . . ( Micah 4 : 3 ) ( Amos 6 : 8 ) ( Song of Solomon 4 : 2 ) ( 1 Kings 9 : 26 ) ( 1 Samuel 13 : 20 ) There are 13 other possible citations for ËÏÇÊÆ. . is turned . . . . . . . . of the . . ( Genesis 39 : 19 ) ( Genesis 40 : 2 ) ( Exodus 24 : 4 ) ( Exodus 34 : 1 ) ( Leviticus 13 : 3 ) ( Leviticus 14 : 46 ) ( Leviticus 25 : 29 ) ( Deuteronomy 22 : 21 ) ( Joshua 11 : 4 ) ( 1 Samuel 13 : 20 ) ( 1 Samuel 15 : 3 ) ( 1 Samuel 26 : 13 ) ( 1 Kings 10 : 15 ) ( Psalm 10 : 14 ) ( Isaiah 3 : 17 ) ( Isaiah 54 : 16 ) ( Isaiah 54 : 17 ) ( Habakkuk 2 : 4 ) Puzzle #2. . KNUTH Answers Puzzle #1. mischief and . . every man his ËÏÇÊ ËÏÇÊÆ ËÀÇÊÆ. . and the . . and ÁÎ Æ doesn’t work. within a . . . . . . them not. and . and . . . . .. . .

À Ï Ê of thy wood . forth. I have . . and they . charity shall 33 . . . . . . . . Therefore . . .) Puzzle #6. . . . seventh hour the . . they married . till seven Ë Î Æ. . Ë Ï with saws. as a . . they were . . . . À Ï Agag in pieces . Ë. . looked in the . . which is . . times . Gather up thy . . . . . . Ï Î for a wave . . . Ë for the name . Ï Î Ë. Acts 2 : 45. Ë Î : For he will . (See also Genesis 27 : 44. . . . . . . . . remnant shall be . . . . the mighty . from the . ( Genesis 3 : 7 ) ( Zechariah 4 : 1 ) ( Exodus 29 : 27 ) ( Psalm 93 : 4 ) ( Luke 17 : 27 ) ( Acts 15 : 26 ) ( Ezekiel 21 : 21 ) ( Titus 1 : 8 ) ( 1 Peter 4 : 8 ) Æ Ã Ï Ã Ï Î Ï Î ÏÁÎ ÄÁÎ ÄÁÎ ÄÇÎ ÇÎ (Many other solutions are possible. . again. . a . . .) Puzzle #4. hazarded their . . . . . (But ‘writ’ is not a Biblical word. for example. . Psalm 145 : 15. Ï Ê ye shall give . . ( Matthew 18 : 22 ) ( Matthew 13 : 49 ) ( John 4 : 52 ) ( Numbers 33 : 54 ) ( Deuteronomy 29 : 11 ) ( 1 Samuel 15 : 33 ) ( Job 16 : 15 ) ( 1 Kings 7 : 9 ) ( Romans 9 : 27 ) ( Leviticus 9 : 21 ) ( Acts 27 : 41 ) ( Jeremiah 10 : 17 ) ( Matthew 13 : 25 ) ( Ezekiel 24 : 23 ) ( Matthew 18 : 21 ) . . ÌÁÅ Ë? Jesus saith . . . . Ruth 3 : 16. Job 40 : 17. . Genesis 32 : 15. . . shoulder Aaron . . . Psalm 84 : 6. Ecclesiastes 12 : 11. . . Ì Ê Ë among the . . . . . And your . . and to the . . . . . John 18 : 18. . Matthew 25 : 33. . in Revelation 3 : 11. Ê the multitude . and sowed . . . me. And the soldiers’ . . Ë of the sea. violence of the . . . Ë Î Ê the wicked . .) Puzzle #5. . . . . . Suitable intermediate words can be found. 2 Corinthians 11 : 19. . they were . Matthew 6 : 28. At his . Î Ê left him. Ezra 7 : 24. . . and . . use intermediate words found in Matthew 24 : 35. ÌÁÊ Ë shall be . Ê. . For example. and which . . and . but none are strictly increasing or decreasing. Ê of good men. . . . . . . Ezra 34 : 25. . . . . hospitality. Ë Ï sackcloth . Ï Ê Ë out of . Daniel 3 : 21. . . hewed stones. . Numbers 1 : 2. And Samuel . . Micah 1 : 8.BIBLICAL LADDERS .

1996). The Universe in a Handkerchief: Lewis Carroll’s Mathematical Recreations. KNUTH Puzzle #7. ØØÔ »» Ø ÜØºÚ Ö Ò º Ù» Úº ÖÓÛ× º ØÑÐ [online text of the King James Bible provided in searchable form by the Electronic Text Center of the University of Virginia]. .34 D. Chapter 6. Games. E. and Word Plays (New York: Copernicus. ÏÇÅ Æ ÇÄÁÎ ÅÁÌ Ë Î ÆÌ Æ ËÌË ÄÄË ÉÍ Ä ÄÍ Ê Ä Ê ËÄ È References Martin Gardner. Puzzles.

Hearts). are all twosided card games. two-sided playing cards and games were invented by Stewart Lamle. (Cards. 16th Century: The four suits were created to represent the ideal French national. Diamonds. MaxxÌ Å . Clubs.Card Game Trivia Stewart Lamle 14th Century: Decks of one-sided Tarot playing cards ﬁrst appeared in Europe. uniﬁed (feudal) society as promoted by Joan of Arc: Nobility. “Finally. 20th Century: After 600 years of playing with one-sided cards. 19th Century: The Joker was devised by a Mississippi riverboat gambler to increase the odds of getting good Poker hands. Peasants. and BettoÌ Å . like other forms of entertainment and gambling. competed with Holy services. Over 100 million decks of cards were sold in the United States last year! 35 . They were soon banned by the Church. Aristocracy. 18th Century: Symmetric backs and fronts were designed to prevent cheating by signaling to other players. you can play with a full deck!”—Zeus Ì Å . the Church (Spades.) Card-playing spread like wildﬁre.

” “An odd number times an odd number is an odd number.” “An even number plus an even number is an even number. and an even number plus an odd number is an odd number.Creative Puzzle Thinking Nob Yoshigahara Problem 1: “An odd number plus an odd number is an even number. and an odd number times an even number is an even number.” “Then an even number times an even number is an odd number. Problem 3: What number belongs at ? in this sequence? 13 24 28 33 34 32 11 17 16 18 14 36 35 46 52 53 22 ? 33 19 24 37 . OK?” “Yes. OK?” “No! It is an even number. OK?” “OK. OK?” “Of course.” “No! It is an odd number! I can prove it!” How? Problem 2: Move two matches so that no triangle remains.

Problem 6: Which two numbers come at the end of this sequence? 2. 3:59:53. (2) Find the two palindromes whose difference is closest to 12 hours. 44. 13:22:31. 50. 32.38 N. 1:01:01. y Problem 7: The ﬁgure shown here is the solution to the problem of dividing the ﬁgure into four identical shapes. Can you divide the ﬁgure into three identical shapes? ´Ü µ´Ü µ´Ü µ ´Ü Ýµ´Ü Þ µ Problem 8: A 24-hour digital watch has many times that are palindromic. 64. 52. 34. 23:55:32. 54. (3) Find the longest time span without a palindromic time. 4. 6. (1) Find the closest such times. 46. (Ignore the colons. x. For example.) These curious combinations occur 660 times a day. 36. 40. etc. 60. 42. 66. . 2:41:42. YOSHIGAHARA Problem 4: Arrange the following ﬁve pieces to make the shape of a star. 56. 30. 62. Problem 5: Calculate the expansion of the following 26 terms.

4. and 5 matchsticks long is divided into two parts with equal area using 3 matchsticks.CREATIVE PUZZLE THINKING 39 Problem 9: A right triangle with sides 3. . but you may rotate them in the plane. You are not allowed to turn over any of the pieces. Can you divide this triangle into two parts with equal area using only 2 matchsticks? Problem 10: Cover the 7 ¢ 7 square on the left with the 12 L-shaped pieces on the right.

How many pieces do you need? . divide the left-hand shape and rearrange the pieces to make the right-hand shape. YOSHIGAHARA Problem 11: Using as few cuts as possible.40 N.

**Number Play, Calculators, and Card Tricks: Mathemagical Black Holes
**

Michael W. Ecker

The legend of Sisyphus is a lesson in inevitability. No matter how Sisyphus tried, the small boulder he rolled up the hill would always come down at the last minute, pulled inexorably by gravity. Like the legend, the physical universe has strange entities called black holes that pull everything toward them, never to escape. But did you know that we have comparable bodies in recreational mathematics? At ﬁrst glance, these bodies may be even more difﬁcult to identify in the world of number play than their more famous brethren in physics. What, after all, could numbers such as 123, 153, 6174, 4, and 15 have in common with each other, as well as with various card tricks? These are mathematical delights interesting in their own right, but much more so collectively because of the common theme linking them all. I call such individual instances mathemagical black holes.

**The Sisyphus String: 123
**

Suppose we start with any natural number, regarded as a string, such as 9,288,759. Count the number of even digits, the number of odd digits, and the total number of digits. These are 3 (three evens), 4 (four odds), and 7 (seven is the total number of digits), respectively. So, use these digits to form the next string or number, 347. Now repeat with 347, counting evens, odds, total number, to get 1, 2, 3, so write down 123. If we repeat with 123, we get 123 again. The number 123 with respect to this process and the universe of numbers is a mathemagical black hole. All numbers in this universe are drawn to 123 by this process, never to escape.

Based on articles appearing in (REC) Recreational and Educational Computing. 41

42

M. W. ECKER

But will every number really be sent to 123? Try a really big number now, say 122333444455555666666777777788888888999999999 (or pick one of your own). The numbers of evens, odds, and total are 20, 25, and 45, respectively. So, our next iterate is 202,545, the number obtained from 20, 25, 45. Iterating for 202,545 we ﬁnd 4, 2, and 6 for evens, odds, total, so we have 426 now. One more iteration using 426 produces 303, and a ﬁnal iteration from 303 produces 123. At this point, any further iteration is futile in trying to get away from the black hole of 123, since 123 yields 123 again. If you wish, you can test a lot more numbers more quickly with a computer program in BASIC or other high-level programming language. Here’s a fairly generic one (Microsoft BASIC):

½ ÄË ¾ ÈÊÁÆÌ Ì ½¾¿ Å Ø Ñ Ð Ð ÀÓÐ » ´ µ ½ ¿¸ Öº Åº Ïº Ö ¿ ÈÊÁÆÌ ÈÊÁÆÌ Á³ÐÐ × ÝÓÙ ØÓ ÒÔÙØ ÔÓ× Ø Ú Û ÓÐ ÒÙÑ Ö ÒÓÛº ÈÊÁÆÌ Á³ÐÐ ÓÙÒØ Ø ÒÙÑ Ö× Ó Ú Ò Ø×¸ Ó Ø×¸ Ò ØÓØ Ðº ÈÊÁÆÌ ÖÓÑ Ø Ø Á³ÐÐ ÓÖÑ Ø Ò ÜØ ÒÙÑ Öº ËÙÖÔÖ × Ò ÐÝ¸ Û ÐÛ Ý× ÈÊÁÆÌ Û Ò ÙÔ Ö Ò Ø Ñ Ø Ñ Ð Ð ÓÐ Ó ½¾¿ººº ÈÊÁÆÌ ÇÊ Ä ½ ÌÇ ½¼¼¼ Æ Ì ½¼ ÁÆÈÍÌ Ï Ø × ÝÓÙÖ Ò Ø Ð Û ÓÐ ÒÙÑ Ö Æ° ÈÊÁÆÌ ¾¼ Á Î Ä´Æ°µ ½ ÇÊ Î Ä´Æ°µ ÁÆÌ´Î Ä´Æ°µµ ÌÀ Æ ½¼ ¿¼ ÇÊ Á ÁÌ ½ ÌÇ Ä Æ´Æ°µ ¼ ° ÅÁ °´Æ°¸ Á ÁÌ¸½µ ¼ Á ° ÌÀ Æ ¼ ¼ Á Î Ä´ °µ»¾ ÁÆÌ´Î Ä´ °µ»¾µ ÌÀ Æ Î Æ Î Æ · ½ ÄË Ç Ç · ½ ¼ Æ Ì Á ÁÌ ¼ ÈÊÁÆÌ Î Æ¸ Ç ¸ ÌÇÌ Ä ¼ ÆÍ° ËÌÊ°´ Î Æµ · ËÌÊ°´Ç µ · ËÌÊ°´ Î Æ · Ç µ ½¼¼ ÈÊÁÆÌ Î Æ Ç Î Æ · Ç ¹¹¹ Æ Û ÒÙÑ Ö × Î Ä´ÆÍ°µ ½½¼ ÈÊÁÆÌ Á Î Ä´ÆÍ°µ Î Ä´Æ°µ ÌÀ Æ ÈÊÁÆÌ ÓÒ º Æ ½¾¼ Æ° ÆÍ° Î Æ ¼ Ç ¼ ÇÌÇ ¿¼

If you wish, modify line 110 to allow the program to start again. Or revise the program to automate the testing for all natural numbers in some interval.

**What Is a Mathemagical Black Hole?
**

There are two key features that make our example interesting:

MATHEMAGICAL BLACK HOLES

43

1. Once you hit 123, you never get out, just as reaching a black hole of physics implies no escape. 2. Every element subject to the force of the black hole (the process applied to the chosen universe) is eventually pulled into it. In this case, sufﬁcient iteration of the process applied to any starting number must eventually result in reaching 123. Of course, once drawn in per point 2, an element never escapes, as point 1 ensures. A mathemagical black hole is, loosely, any number to which other elements (usually numbers) are drawn by some stated process. Though the number itself is the star of the show, the real trick is in ﬁnding interesting processes. Formalized Deﬁnition. In mathematical terms, a black hole is a triple U is a function, all (b, U, f), where b is an element of a set U and f: U satisfying: 1. f (b) = b . 2. For each x in U, there exists a natural number k satisfying f (x) = b. Here, b plays the role of the black-hole element, and the superscript indicates k-fold (repeated) composition of functions. For the Sisyphus String½ , b = 123, U = natural numbers ¸ and f (number) = the number obtained by writing down the string counting # even digits of number, # odd digits, total # digits. Why does this example work, and why do most mathemagical black holes occur? My argument is to show that large inputs have smaller outputs, thus reducing an inﬁnite universe to a manageable ﬁnite one. At that point, an argument by cases, or a computer check of the ﬁnitely many cases, sufﬁces. 999, then In the case of the 123 hole, we can argue as follows: If n f (n) n. In other words, the new number that counts the digits is smaller than the original number. (It’s intuitively obvious, but try induction if you would like rigor.) Thus, starting at 1000 or above eventually pulls one down to under 1000. For n 1000, I’ve personally checked the iterates of f (n) for n = 1 to 999 by a computer program such as the one above. The direct proof is actually faster and easier, as a three-digit string for a number must have one of these possibilities for (# even digits, # odd digits, total # digits):

½For generalized sisyphian strings, see REC, No. 48, Fall 1992.

44

M. W. ECKER

(0, 3, 3) (1, 2, 3) (2, 1, 3) (3, 0, 3) So, if n 1000, within one iteration you must get one of these four triples. Now apply the rule to each of these four and you’ll see that you always produce (1, 2, 3) — thus resulting in the claimed number of 123.

Words to Numbers: 4

Here is one that master recreationist Martin Gardner wrote to tell me about several years ago. Take any whole number and write out its numeral in English, such as FIVE for the usual 5. Count the number of characters in the spelling. In this case, it is 4 — or FOUR. So, work now with the 4 or FOUR. Repeat with 4 to get 4 again. As another instance, try 163. To avoid ambiguity, I’ll arbitrarily say that we will include spaces and hyphens in our count. Then, 163 appears as ONE HUNDRED SIXTY-THREE for a total count of 23. In turn, this gives 12, then 6, then 3, then 5, and ﬁnally 4. Though this result is clearly language-dependent, other natural languages may have a comparable property, but not necessarily with 4 as the black hole.

**Narcissistic Numbers: 153
**

It is well known that, other than the trivial examples of 0 and 1, the only natural numbers that equal the sum of the cubes of their digits are 153, 370, 371, and 407. Of these, just one has a black-hole property. To create a black hole, we need to deﬁne a universe (set U ) and a process (function f ). We start with any positive whole number that is a multiple of 3. Recall that there is a special shortcut to test whether you have a multiple of 3. Just add up the digits and see whether that sum is a multiple of 3. For instance, 111,111 (six ones) is a multiple of 3 because the sum of the digits, 6, is. However, 1,111,111 (seven ones) is not. Since we are going to be doing some arithmetic, you may wish to take out a hand calculator and/or some paper. Write down your multiple of 3. One at a time, take the cube of each digit. Add up the cubes to form a new

MATHEMAGICAL BLACK HOLES

45

number. Now repeat the process. You must reach 153. And once you reach 153, one more iteration just gets you 153 again. Let’s test just one initial instance. Using the sum of the cubes of the digits, if we start with 432 — a multiple of 3 — we get 99, which leads to 1458, then 702, which yields 351, ﬁnally leading to 153, at which point future iterations keep producing 153. Note also that this operation or process preserves divisibility by 3 in the successive numbers.

½¼ ¾¼ ¿¼ ¼ ¼ ¼ ¼ ¼ ¼ ½¼¼ ½½¼ ½¾¼ ½¿¼ ½ ¼ ½ ¼ ½ ¼ ½ ¼ ½ ¼ ÄË ÈÊÁÆÌ Ì Ñ Ø Ñ Ð Ð ÓÐ ½ ¿ººº ÈÊÁÆÌ ÈÊÁÆÌ Ë ÑÔÐ Ø ×Ø ÔÖÓ Ö Ñ ÓÖ ×ÙÑ Ó Ù × Ó Ø×ººº ÈÊÁÆÌ ÈÊÁÆÌ ÓÔÝÖ Ø ½ ¿¸ Öº Åº Ïº Ö ÈÊÁÆÌ ÇÊ Ä ½ ÌÇ ¾¼¼¼ Æ Ì ÁÆÈÍÌ ÆÙÑ Ö ØÓ Ø ×Ø Æ ÈÊÁÆÌ Á Æ»¿ ÁÆÌ´Æ»¿µ ÇÊ Æ ½ ÌÀ Æ ÈÊÁÆÌ ÈÓ× Ø Ú ÑÙÐØ ÔÐ × Ó ¿ ÓÒÐÝ ÇÌÇ ¼ Á Æ ½¼¼¼¼¼¼¼ ÌÀ Æ ÈÊÁÆÌ Ä Ø³× ×Ø ØÓ ÒÙÑ Ö× ÙÒ Ö ½¼¼¼¼¼¼¼ ÇÌÇ ¼ Æ° ËÌÊ°´Æµ Ä Ä Æ´Æ°µ ³ ÓÒÚ ÖØ ØÓ ×ØÖ Ò ØÓ Ñ Ò ÔÙÐ Ø Ø× ËÍÅ Í Ë ¼ ³ÁÒ Ø Ð Þ ×ÙÑ ÇÊ Á ÁÌ ½ ÌÇ Ä Î´ Á ÁÌµ Î Ä´ÅÁ °´Æ°¸ Á ÁÌ¸½µµ ³ Ø Ú ÐÙ Ó Ø Í´ Á ÁÌµ Î´ Á ÁÌµ¶Î´ Á ÁÌµ¶Î´ Á ÁÌµ ³ Ù Ø ËÍÅ Í Ë ËÍÅ Í Ë · Í´ Á ÁÌµ ³Ã Ô ÖÙÒÒ Ò ØÓØ Ð Ó ×ÙÑ Ó Ù × Æ Ì Á ÁÌ ÈÊÁÆÌ Ì ×ÙÑ Ó Ø Ù × Ó Ø Ø× Ó Æ × ËÍÅ Í Ë Á ËÍÅ Í Ë Æ ÌÀ Æ ÈÊÁÆÌ ËÙ ×× ÓÙÒ Ð ÓÐ ÈÊÁÆÌ ÇÌÇ ¼ Æ ËÍÅ Í Ë ÇÌÇ ½½¼

This program continues forever, so break out after you’ve grown weary. One nice thing is that it is easy to edit this program to test for black holes using larger powers. (It is well known that none exists for the sum of the squares of the digits, as one gets cycles.) In more formal language, we obtain the 153 mathemagical black hole by letting U = 3 · = all positive integral multiples of 3 and f (n) = the sum of the cubes of the digits of n. Then b = 153 is the unique black-hole element. (For a given universe, if a black hole exists, it is necessarily unique.) Not incidentally, this particular result, without the “black hole” terminology or perspective, gets discovered and re-discovered annually, with a paper or problem proposal in one of the smaller math journals every few years.

f (n) each value eventually is “pulled” into the black hole of 153. by cases or computer check. but be sure to sandwich the column that contains the chosen card between the other two columns. ECKER The argument for why it works is similar to the case with the 123 example. At the end. It’s a classic card trick. for suitably small numbers. W. yet meets the two criteria for a black hole. There are two ways to prove this. Now ask him (or her) which of the three columns contains the card. there are some).. for large numbers n. the card chosen must be in the center of the array. and then test to see whether other numbers are drawn to it. 1¿ + 3¿ + 5¿ = 153. card 11.46 M. but the easier way is to draw a diagram that illustrates where a chosen card will end up next time. Repeat asking which column. I’ll omit the proof. This is the card in the fourth row and second column. regrouping cards with the designated column being in the middle. First of all. But for those who enjoy programs. you will need ﬁrst to discover a number that equals the sum of the fourth (or higher) power of its digits. and re-dealing out by rows. Even Here’s an example that sounds a bit different. try this one from one of my readers. Ask somebody to think of one of the cards without telling you which card he (or she) is thinking of. Arrange them in seven horizontal rows and three vertical columns. Now re-lay out the cards by laying out by rows (i. To ﬁnd an analagous black hole for larger powers (yes. for the black-hole attraction. Regroup the cards by picking up the cards by whole columns intact. n. laying out three across at a time). Repeat one last time. which is to say. Card Tricks. Remove 21 cards from an ordinary deck. note that. ½¼ ¾¼ ½¼¼ ½½¼ ½¾¼ ½¿¼ ½ ¼ ½ ¼ ÄË ÈÊÁÆÌ Ì ¾½ Ö × ¹ ÈÖÓ ÈÊÁÆÌ ÔØ Ý Öº Åº Ïº ÁÆÌ ¹ ÁÅ Î´¾½µ¸ ´ ¸¿µ ÁÌ Ê ¼ ÇÊ Æ ½ ÌÇ ¾½ Î´Æµ Æ Æ ÈÊÁÆÌ ÈÊÁÆÌ È Ö ¸ ÔÐ Á ÁÌ Ê ¿ ÌÀ Æ ÈÊÁÆÌ ÈÊÁÆÌ Ö Ñ Ý Ë ÐÐÝ Ö ÞÞ Ö ÓÖ Ê Ì Æ × ººº ÈÊÁÆÌ ÓÙÖ Ö × Î´½½µ Æ . Then. Second.e. so 153 is indeed a ﬁxed point.

involve numbers. the smallest permutation (allowing initial zeros as digits). and subtract. don’t take one of the nine numbers with four identical digits). ¾ Kaprekar’s Constant: What a Difference 6174 Makes! Most black holes. but this trick. Within the total of seven steps. write down the largest permutation of the number. as with the sisyphian strings. .e..MATHEMAGICAL BLACK HOLES 47 ½ ¼ ½ ¼ ½ ¼ ½ ¼ ¾¼¼ ¾½¼ ¾¾¼ ¾¾½ ¾¾¾ ¾¾¿ ¾¿¼ ¾ ¼ ¾ ¼ ¾ ¼ ¾ ¼ ¾ ¼ ¿¼¼ ¿½¼ ¿¾¼ ¿¿¼ ¿ ¼ ¿ ¼ ÁÌ Ê ÁÌ Ê · ½ Æ ¼ ÇÊ Á ½ ÌÇ Æ Æ · ½ ´Á¸Âµ Î´Æµ Æ Ì Â Æ Ì Á ÈÊÁÆÌ ÈÊÁÆÌ ÇÊ Á ½ ÌÇ ÈÊÁÆÌ ÍËÁÆ Æ Ì Á ÈÊÁÆÌ ÁÆÈÍÌ ÓÐÙÑÒ Ó Á ½ ÇÊ ÇÊ Ã ½ ÌÇ ¿ Á ¾ ÌÀ Æ Æ ¼ ÇÊ Ã ½ ÌÇ ¿ Â Ç´Ãµ Æ Æ · ½ Î´Æµ ´Á¸Âµ Æ Ì Á Æ Ì Ã ÇÌÇ ½ ¼ ÇÊ Â ½ ÌÇ ¿ ´Á¸½µ ´Á¸¾µ ´Á¸¿µ Ö ´½¸ ¾¸ ÓÖ ¿µ ¿ ÌÀ Æ ¾ ¼ Ç´Ãµ Ã Æ Ì Ã ËÏ È Ç´ µ¸ Ç´¾µ ÇÊ Á ½ ÌÇ Perhaps it is not surprising. Fall 1992. ¾REC. At that point. Rearrange the digits of your number to form the largest and smallest strings possible. generalizes somewhat. nonetheless. further iteration with 6174 is pointless: 7641–1467 = 6174. That is. Apply this same process to the difference just obtained. Take any four-digit number except an integral multiple of 1111 (i. you always reach 6174. No. 48.

Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . we would have obtained the reciprocal of phi as the limit. 3.6 or so. the smallest is 0288. 2. 8.). 34. phi. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors. Now take the sum of the digits of these divisors. ´½µ ´Ò ½µ · ´Ò ¾µ for integers Ò ¾. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ Ä¾ Ä ½ ¼ . The ﬁrst two numbers numbers: 1. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. etc. 5. 13. Your own example may take more steps. More formally. but fewer authors use that approach. but you will reach 6174. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò call it Ä — as Ò increases.618 . This one may defy not only your ability to explain it. That is.48 M. and ´Òµ preceding two elements. but it is also that much more strange at the same time. but your very equilibrium. ECKER Example: Start with 8028. Had we used the ratios ´Òµ ´Ò · ½µ instead. and the difference is 8532. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. Divide both sides of the equation ´Òµ ½µ · ´Ò ¾µ by ´Ò ½µ. and these digits sum to 15. 55. including 1 and the number itself. The black-hole number this time is 15. and 15. 1. but without using a program. ½ . This one is a bit more tedious. W. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. 3. Its divisors are 1. The largest permutation is 8820. In fact. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. 21. Its value is roughly 1. The ratios seem to be converging to a number around 1. To get a quick and dirty feel for why phi arises. Iterate until a number repeats. 89. 5. Repeat with 8532 to calculate 8532–2358 = 6174.

The ﬁrst two numbers need not even be whole numbers or positive. if you take any additive sequence (any sequence — no matter what the ﬁrst two terms — in which the third term and beyond are obtained by adding the preceding two). In closing out this brief connection of Fibonacci numbers to the topic. had we formed the ratios ´Òµ ´Ò · ½µ. ½¿ . that we must similarly rule out ¿ . the second golden ratio. This. The limiting ratio. must also be eliminated from the universe. The solution to the equation ´Üµ ½ · ½ Ü ¼ is just Ü ½. in succession. etc. . each number is the negative reciprocal of the other. could not use an input Third. Each is also one minus the other. I selected (or rather stumbled across) this function because of the simple argument above that gave phi as a limit. note that our last function of 0 because division by zero is undeﬁned. But there is still another black hole one can derive from this. Notice that these fractions are precisely the negatives of the ratios of consecutive Fibonacci numbers in the reverse order that we considered. more generally. Thus.MATHEMAGICAL BLACK HOLES 49 Solving by the quadratic formula yields two solutions. ´Üµ ½·½ Ü. too. Ü may not equal ½ either. For ´Üµ ½·½ Ü. one gets the same result: convergence to phi. More about the second solution in a moment. Second. one of which is phi. We have not ﬁnished. we . One obtains convergence to phi — but still no appearance of the second number that is the solution to the quadratic equation above. All of these must be eliminated from the universe for . ´ ´Üµµ ¼ and then ´ ´ ´Üµµµ is undeﬁned). then ¿ . Thus. if we extend the deﬁnition of black hole to require only that the iterates get closer and closer to one number. If we continue now working backward with function ’s preimages we ¾ ﬁnd. I would be remiss if I did not follow my own advice on getting black holes by looking for ﬁxed points: values Ü such that ´Üµ Ü. we would have obtained the absolute value of the second solution instead of the ﬁrst solution. Start with a seed number Ü and iterate function values ´Üµ ´ ´Üµµ. is easy to test in a program that you can write yourself. etc. along with 0. Solving ½ · ½ Ü ½ gives Ü ½ ¾. Notice that the above plausibility argument did not use the values F (1) and F (2). We must now avoid ´Üµ ½ (otherwise. Indeed. Consider the function ´Üµ ½·½ Ü for nonzero real numbers Ü. we have a black hole once again. Is there a connection between the two solutions of the quadratic equation? First.

you get 16. take half. Now try some examples. and conversely. our black hole called phi. then 4. then 2. As things are set up. a cycle a cycle of length three. All real inputs except zero. ECKER obtain our two golden numbers. Pick the largest odd factor. W. iteration of matrix powers draws one to a result that represents long-term stability independent of an input . because the largest odd factor in 4 is 1. Under certain conditions. triple and add one. dating back to the 1930s and still an open question (though sometimes also identiﬁed with the names of Hailstone. which really are cycles of length one. Close Other examples of mathemagical black holes arise in the study of stochastic processes. and ¿¢½·½ . Hmm now only in black holes. This answer is the next iterate. You should ﬁnd that you keep getting 4. Unsolved Problems as Black Holes Even unsolved problems sometimes fall into this black-hole scenario. this problem has the paradoxical property that. (Just multiply all the odd prime factors together. Deﬁne the process instead by taking the starting number and breaking it down completely into factors that are odd and even. Success! In fact. although one of the hardest to settle deﬁnitively. We’re interested that repeats ad inﬁnitum. then 1 again. then 8. the numbers ´Ò ½µ ´Òµ. let’s be creative and ﬁx this up by modifying the problem. is among the easiest to program (a few lines). If you do reach 1 — and nobody has either proved you must.50 M. while ¼ ½ is a repeller. Keep iterating. In the example that would be ¿¢ ¾½. Because this is so easy to program. you stay at 4. the number ½ ½ is an attractor with the black-hole property. For instance. Consider the Collatz Conjecture. Once you hit 4. I will omit a program here. and the second golden number lead to the attractor. The only non-odd prime is 2. then 1. and Syracuse). Must you always reach 1? If you start with 5. then 2. Ulam. 84 is ¾ ¢ ¾ ¢ ¿ ¢ . If even. Anybody who proves the Collatz Conjecture will prove that my variation is a mathemagical black hole as well. So. Start with a natural number.) Now triple the largest odd factor and then add 1. If odd. nor shown any example that doesn’t — then you next get 4.

. Editor Recreational and Educational Computing 909 Violet Terrace Clarks Summit. Recreational and Eductional Computing. Pennsylvania 18411-9206 USA Happy Black-Hole hunting during your salute to Martin Gardner. whom we are proud to make a tiny claim to as our Senior Contributing Mathematical Editor. This is an ongoing pursuit in my own newsletter. Ecker.MATHEMAGICAL BLACK HOLES 51 vector. I invite readers to send me other examples or correspondence: Dr. too. Michael W. in which we’ve had additional ones. the teasers and problems here are intriguing in their own right. the real value for me is in seeing the black-hole idea as the unifying theme of seemingly disparate recreations. Moreover. Quite apart from any utility. however. There are striking resemblances to convergence results in such disciplines as differential equations.

Bob Wainwright. James Dalgety. Easy Problems E1. Nick Baxter. After going 10 miles in a straight line he stops for lunch. These problems are for the enjoyment of the solver and should be passed on to others for their enjoyment as well. where I worked for 27 years. Brian Barwell. especially. The problems are arranged in three general categories of easy. embellishments or adaptations of such problems. A sources section provides information on the source of each problem as far as I know. Inc.. The regular problem column appeared from 1984 to 1994 and was called “Puzzles from Around the World”. a company newsletter published at Logicon. I would be delighted to hear from anyone who can improve the solution approach. Nob Yoshigahara. Clayton Dodge. After lunch he gets 53 . Karl Scherer. Source information begins on page 82. A man has breakfast at his camp. Martin Gardner. other solvers through word of mouth.Puzzles from Around the World Richard I. Dieter Gebhardt. Allan Gottlieb. the solution section gives an approach to answering each of the problems. and hard . Harry Nelson. Dario Uri. medium. David Singmaster. or entirely new problems of my own creation. Naoaki Takashima. and. Hess Introduction Most of the puzzles in this collection were presented in the Logigram. Laurie Brokenshire. He gets up and travels due North. add information on the source of the problem. Yoshiyuki Kotani. it consisted of problems from a number of sources: other problem columns. I thank the following people for providing problems and ideas that have contributed to the richness of the problems involved: Leon Bankoff. or offer more interesting problems for future enjoyment.

54 R. You have containers that hold 15 pints. measure exactly two pints for yourself to drink and end up with 8 pints in the 10-pint container and 5 pints in the 6-pint container. 0).5 nautical miles from Los Angeles to Honolulu. 63 9 38 33 32 12 88 25 16 18 15 Ü 5 E3. reverse the digits of the result. 5. 2 and 1. 6. 0. Calculate the expansion of this 26 term expression: ´Ü µ´Ü µ´Ü µ´Ü µ ´Ü Ýµ´Ü Þ µ . Divide the number by 3. 8. After going 10 miles in a straight line he ﬁnds himself back at camp. What number belongs at Ü in this sequence? 34 32 36 46 64 75 50 35 34 16 18 14 22 Ü 40 35 15 20 12 E4. My regular racquetball opponent has a license plate whose three-digit part has the following property. Find the rule for combining numbers as shown below (e. E9. 10 pints. 4. It is approximately 2244.. g. 6. The 15pint container starts out full. Through transferring liquid among the containers. I. 3. 8. Find the most efﬁcient solution. 3. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu. 7. 2. How long does it take? E5. How many transfers of liquid will it take you to get a volume of liquid in one container that is within one percent of exactly one liter? E7. E6. subtract 1 and you produce the original number. 9 in some perE mutation such that when digits are removed one at a time from the right the number remaining is divisible in turn by 8. Where on earth could he be? E2. and the other two start out empty: (15. HESS up and travels due North. 4. 7. What is the number and what is the next greater number (possibly with more than three digits) having this property? . Find a nine-digit number made up of 1. You are at a lake and have two empty containers capable of holding exactly ´ ¿ ½ ½ µ and ´ ¾ ½ ¾ ½ µ liters of liquid. 5. 25 and 9 combine to give 16) and use the rule to determine Ü. and 6 pints.

Each man walks at a different speed: A takes 1 minute to cross. Place the numbers from 1 to 15 in the 3 ¢ 5 array so that each column has the same sum and each row has the same sum. E11.. and 9) with a sum of 45 and a product of 9 362.PUZZLES FROM AROUND THE WORLD 55 E10. 8. E14. If it cannot move to a square previously visited. A maximum of two people can cross at one time. 6. and the bridge will hold up to three men at a time. It is night. how? Try these other problems. A knight is placed on an inﬁnite checkerboard. A pair must walk together at the rate of the slower man’s pace. E13. how can you make it unable to move in as few moves as possible? E12. (a) There are six men with crossing times of 1. Can all four men cross the bridge? If so. 7. Find nine single-digit numbers other than (1. C takes 5 minutes. The Bridge. 3. Four men must cross a bridge. 4.880. and they need their one ﬂashlight to guide them on any crossing. and 9 minutes and they must cross in 31 minutes. . 8. B takes 2 minutes. They all start on the same side and have 17 minutes to get across. 2. and they must cross in 25 minutes. 2. (b) There are seven men with crossing times of 1.. 9. 3. Divide the ﬁgure below into (a) 4 congruent pieces and (b) 3 congruent pieces. . and 10 minutes.. and D takes 10 minutes. 6.

120. one having an 11-pint container and another having a 2Npint container. M4. Potato Curves. N thirsty customers arrive. Find a simple continuous function to generate the series and compute the surprise answer for Ü. It has a symmetry property in that the burn rate a distance Ü from the left end is the same as the burn rate the same distance Ü from the right end. . The lace burns irregularly like a fuse and takes 60 minutes to burn from end to end. Shoelace Clock. Three points have been chosen randomly from the vertices of a cube. 450. You were playing bridge as South and held 432 in spades. and diamonds. Despite your lack of power you took 6 tricks. Suppose a clock’s second hand is exactly on a second mark and exactly 18 second marks ahead of the hour hand. 45. You have a 37-pint container full of a refreshing drink. (b) a right triangle? M3. How will you most efﬁciently measure out 1 pint of drink for each customer to drink in turn and end up with N pints in the 11-pint container and 37-2N pints in the 37-pint container if (a) N = 3. In clubs you held 5432. 1260. 60. From England comes the series 35. x. 180. making a 4-club contract.56 R. and a pair of scissors. You are allowed to draw a closed path on the surface of each of the potatoes shown below. hearts. You are given some matches. HESS E15. (b) N = 5? M2. Can you draw the two paths so that they are identical to each other in three-dimensional space? E16. Produce the other hands and a line of play that allows this to occur. What is the minimum time interval you can measure? Medium Problems M1. 744. I. What time is it? E17. What is the probability that they form (a) an acute triangle. 280. a shoelace.

Find the smallest prime number that contains each digit from 1 to 9 at least once. and [(13). (13). What is the least value for L that allows a successful plan for crossing the river? M9. how many legal combinations would there be? M13. Show that no more than two of its vertices lie on grid points. What is the longest Lyran musical composition? M12. . (34)] = [(21). M10. (14)]. [(134)]. [(1). (3)].PUZZLES FROM AROUND THE WORLD 57 M5. (a) How many legal combinations are there? (b) If a sixth button were added. (4)] are legal combinations while [(1). You wish to cross from the south to the north side and have only two thin planks of length L and width 1 ft to help you get across. (34)]. Prove there are no M8. Music on the planet Alpha Lyra IV consists of only the notes A and B. From the USSR we are asked to simplify Ü M7. M11. The river shown below is 10 feet wide and has a jog in it. (14)] are not. Many crypto doorknob locks use doors with ﬁve buttons numbered from 1 to 5. Also. Thus [(12). Ô ¾ · ¿½¼¿ · Õ ¿ Ô ¾ ¿½¼¿ . Legal combinations allow the buttons to be pushed in speciﬁc order either singly or in pairs without pushing any button more than once. (a) Determine the weight of each package with a two-pan balance and four weights of your own design. 26 packages labeled A to Z are known to each weigh whole numbers of pounds in the range 1 to 26. and [(2). A regular pentagon is drawn on ordinary graph paper. Find all primes others. it never includes three repetitions of any sequence nor does the repetition BB ever occur. Õ ¿ Ô such that ¾Ô + Ô¾ is also prime. How many ways can four points be arranged in the plane so that the six distances between pairs of points take on only two different values? M6. (b) Now do it with three weights.

The band is initially 100 feet long and is instantaneously and uniformly stretched an additional 100 feet at the end of each minute. I. A solution with eight triangles is known. M17. Dissect a square into similar rectangles with sides in the ratio of 2 to 1 such that no two rectangles are the same size.01). It crawls at a rate of 1 foot per minute. M15. (b) all three regions are of different size. You are told that the three numbers are primes and that they form the sides of a triangle with prime perimeter. (a) If the ant starts at point A. which point is the greatest distance away? (It is not B. What number is written on your forehead? M18. HESS M14. A solution with nine rectangles is known. d. . between these points is greater than 3. Dissect a square into similar right triangles with legs in the ratio of 2 to 1 such that no two triangles are the same size. both of whom state that they cannot deduce the number on their own foreheads.58 R. You see 5 and 7 on the other two people. An ant crawls along the surface of a 1 ¢ 1 ¢ 2 “dicube” shown below.) (b) What are the two points farthest apart from each other on the surface of the dicube? (The distance. Divide an equilateral triangle into three contiguous regions of identical shape if (a) All three regions are the same size. M16. You and two other people have numbers written on your foreheads. A snail starts crawling from one end along a uniformly stretched elastic band. and (b) farthest from the far end of the band? M19. (c) two of the regions are the same size and the third region is a different size. At what points in time is the snail (a) closest to the far end of the band. The snail maintains his grip on the band during the instant of each stretch.

) . And I ought to mention that in no two subjects did I get the same mark. I’ll put it another way. But what has that got to do with liking them? When I qualiﬁed as a Good Egg — many. “I can’t. The number of other Good Eggs who qualiﬁed when I did. “Good at sums. “I must think.” He composed himself for a nap. The number of subjects was one third of the number of marks obtainable in any one subject.” said Humpty Dumpty. It was a number equal to four times the maximum obtainable in one subject. Alice was almost in tears. Well.PUZZLES FROM AROUND THE WORLD 59 M20. interested. “So I am. or something I’ve never learned?” Humpty Dumpty opened one eye.” “But you haven’t told me . and that is also true of the other Good Eggs who qualiﬁed.” said Alice. Humpty Dumpty. But to make it as easy as I can for you. Isn’t it differential equations. “I know I haven’t told you how many marks in all one had to obtain to qualify. “Ah!” said Humpty Dumpty. gives a product equal to half the number of marks obtained by each Good Egg. None of us did as well in arithmetic as in any other subject. “Don’t be a fool.” she said.” “How many subjects were there?” said Alice. Similar Triangles. And we all just managed to qualify.” began Alice.” said Humpty Dumpty. How many ways can you ﬁnd to do this? (Nineteen solutions are known.” “But how many . “You don’t like arithmetic. “How many of us were there? Well.” said Alice. multiplied by the number of subjects (I’ve told you about that already). child.” What was Humpty Dumpty’s mark in Arithmetic? Hard Problems H1. Not that that’s saying a lot. The ﬁgure following this problem shows a 30–60–90 triangle divided into four triangles of the same shape. And now you can ﬁnd out all you want to know. child? I don’t very much. I’ll tell you now. “Anyone ought to be able to do it who is able to count on ﬁve ﬁngers. “I’m coming to that. that was — I got a better mark in arithmetic than any of the others who qualiﬁed. “But I thought you were good at sums.” he said crossly. oh certainly. many years ago.” said Humpty Dumpty. when I tell you that no two of us obtained the same assortment of marks — a thing which was only just possible — you’ll be well on the way to the answer. “do any of it.

” She pointed to a pile of wooden stakes. How many stakes should he take with him to get as much land as possible? H3. P. on the exact top is resting on the horizontal plane.) H5. “Take some of these stakes with you. HESS H2.” (The Queen had read a little mathematics.902.60 R. A billiard ball with a small black dot.851 ft for the radius of the earth. I. Given that the width of the building is 125 ft and the rubber band must stretch an extra 10 cm to accommodate the building.) Assume that it takes Sir George 1 minute to pound a stake and that he walks at a constant speed between stakes. “To reward you for killing the dragon.” the Queen said to Sir George. An irrational punch centered on point P in the plane removes all points from the plane that are an irrational distance from P. What is the least number of irrational punches needed to eliminate all points of the plane? H4. “I grant you the land you can walk around in a day. All the land in the convex hull of your stakes will be yours. “Pound them into the ground along your way. and be back at your starting point in 24 hours.” she continued. Where is the black dot when the ball returns to its initial resting place? . how tall is the building? (Use 20. Imagine a rubber band stretched around the world and over a building as shown below. It rolls without slipping or twisting so that its contact point with the plane follows a horizontal circle of radius equal to that of the ball.

3). (7. 6). (11. For what values of p and score situations during a set can the player ahead according to the score have less than a 50% chance of winning the set? . and parentheses. The server in any game wins each point with a ﬁxed probability p. (c) Show that ½ can be approximated with arbitrary accuracy using any two digits if the rules of part (b) are followed. are all equal to 1. (8. decimal points. (12. You are given two pyramids SABCD and TABCD. where 0 p 1. H11. What is the smallest value of n for which you can ﬁt 2n + 1 such circles into a 2 ¢ n rectangle? H10. (a) Find the best approximation using two digits of your choice. (b) The same as part (a) except square roots are allowed.n) = (a) (b) (c) (d) (e) (f) (g) (3. taken from vertices S and T. Tennis Paradox. (a) What region inside a unit square has the greatest ratio of area to perimeter? (b) What volume inside a unit cube has the greatest ratio of volume to surface area? H9. 4). H8. 4). No square roots or other functions are allowed. 4). Two evenly matched tennis players are playing a tiebreak set. exponents. (9. A good approximation of ¿ ½. You may use · ¢ . Locate m points in the plane (perhaps with some exactly on top of others) so that each of them is a unit distance from exactly n of the others for (m.PUZZLES FROM AROUND THE WORLD 61 H6. 5). (12. 2). using just two digits is H7. The altitudes of their eight triangular faces. Prove or disprove that line ST is perpendicular to plane ABCD. As shown below it is easy to place 2n unit diameter circles in a 2 ¢ n rectangle.

then (2) wears any matching pair and returns the odd sock to the drawer or (3) returns the three socks to the drawer if he has no matching pair and repeats steps (1) and (3) until he completes step (2). where he (1) picks out three socks at random.0107 . b. their ages are ” H14. two brown) and ends up with four socks each Saturday evening.0107 0. Minimum Cutting Length. A number may be approximated by a /b where a and b are integers. Deﬁne the goodness of ﬁt of a /b to x as g (a. I. I’ll let you off the hook. ) 0. where e = H16. (a) The ﬁgure shows two Pythagorean triangles with a common side where three of the ﬁve side-lengths are prime numbers. into ﬁve parts of equal area? (c) an equilateral triangle into ﬁve parts of equal area? H15. d. x) = a bx . six black. Since Smith missed the problem tonight and Jones missed it at the party two years ago. b. with g (355. HESS H12. given the sum and product of their ages. An approximation to is 355/113.71828 . The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem.a . (a) Find the minimum a and b so that g (a. 113. My uncle’s ritual for dressing each morning except Sunday includes a trip to the sock drawer. ) = 0.What is the minimum cut-length needed to divide (a) a unit-sided equilateral triangle into four parts of equal area? (b) a unit square. and proceeds to explain: “I normally ask guests to determine the ages of my three children. (a) On which day of the week does he average the longest time to dress? (b) On which day of the week is he least likely to get a pair from the ﬁrst three socks chosen? H13. (b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes? . b) Find the minimum c and d so that g (c.0107. The drawer starts with 16 socks each Monday morning (eight blue.62 R. . Find other such examples.” Your response is “No need to tell me more. e ) 2.

On Lyra III one is a student until reaching a special year immediately following a special year.. The common axis of these layers is the planet’s rotational axis. Cosmic. Inc. where Ô and Õ are different prime numbers. just as one observes for Sol III. the little that is known about the krypton layer was received via sub-etherial communication from a Venerian pioneer immediately prior to its demise at the hands (some would say wattles) of a frumious snatcherband.PUZZLES FROM AROUND THE WORLD 63 H17. The pioneer reported with typical Venerian obscurantism that the ratio of the volume of the smallest sphere that could contain the planet to the volume of the largest sphere that could be contained within the planet is 1. and the next inner layer is an anhydrous fromage. Its axis of rotation coincides with the spheroid’s small axis. being made of a number of coaxial right circular cylindrical layers.) Unfortunately. however. The ﬁrst few such special years are 12. The inhabitants of Lyra III recognize special years when their age is of the form Ô¾Õ. It (the Venerians are sexless) also reported that the straight line distance between it and its copod was 120 spandrals. By nature the Venerians can only survive at the krypton/fromage boundary and by tragic mistake the two copods had landed on disconnected branches of . one then becomes a master until reaching a year that is the third in a row of consecutive special years. each of homogeneous composition. is contemplating mining the outer layer. The outermost of these layers is nearly pure krypton.331 to 1. is unique. The planet Alpha Lyra IV is an oblate spheroid. and the company’s ﬁnancial planners have found that the venture will be sound if there are more than one million cubic spandrals of krypton in the outer layer. 18 and 20. (a) (b) (c) (d) When does one become a master? When does one become a sage? How long do the Lyrans live? Do ﬁve or six special years ever occur consecutively? H18. ﬁnally one becomes a sage until death. It’s internal structure. (A copod corresponds roughly to something between a sibling and a rootshoot. (A spandral is the Lyran unit of length. which occurs in a special year that is the fourth in a row of consecutive special years.) The reporting Venerian was mildly comforted because the distance to its copod was the minimum possible distance between the two.

Five students of Taurus and its ancient culture were discussing their ﬁeld work one day when the following facts were brought to light: ¯ ¯ ¯ ¯ ¯ The ﬁrst student had dug the entire length of one of the channels in search of ancient daggers.64 R. undertake the mining venture? H19. a moon of a-Lyra IV (hieronymous). Taurus. . was also tired because he had thoroughly dug the entire length of the largest possible channel. A curious feature of these channels is that each is a complete and perfect circle. was occupied by a race of knife-makers eons ago. An even more curious fact is that Taurus is a torus (doughnut shape). How long was the channel which the braggart haggard dagger digger dug? H20. The third student had explored a channel 50 spandrals in length. Inc. Should Cosmic. (b) Find ´½¼¼¼µ to three signiﬁcant ﬁgures. I. lying along the intersection of a plane with the satellite’s surface. He found nothing but the fact that the length of the channel was 30 spandrals. a rather lazy fellow (a laggard dagger digger?). The fourth student. HESS the curve of intersection of the krypton/fromage boundary and the planetary surface. (c) Find the smallest Ñ such that the magnitude of ´Ñµ is less than the magnitude of ´Ñ · ½µ. a rather boastful sort. swearing at the difﬁculties he had in crossing the channel of the haggard dagger digger.. He had dug the entire length of a longer channel but never crossed the path of the ﬁrst dagger digger. The second student was very tired from his work. Deﬁne ´Òµ where Ò ¾Ò · ¾ ¿ Ò Ò ½ ´ ¼ Òµ ½¾¿ and ¾ ½ ¾ ½ (a) Prove that ´Òµ goes to zero as Ò goes to inﬁnity. crossing the channel of the haggard dagger digger. they dug a series of channels in the satellite surface. Before they were wiped out by a permeous accretion of Pﬁster-gas. The ﬁfth student. had merely walked the 60 spandral length of another channel.

4. 10. Anywhere within 10 miles of the north pole. 0). 6. (5. Instead. so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. (3. 5). ½ ¼½ or ¼ and ½ ¼½. E2. 0. 6). 6. E9. 15 moves: (15. (3.425.741. 6). But this is not right because 18 is not the difference of 38 and 16. 2. 4. 0). 2). 2). it appears that the rule might be subtraction. E7. 2. 0). 2). so E8. 10. 2. (7. 6). (11. In each case the numbers combine by summing the products of their digits. 9. . 0). E3. E10. 0. Ø¾ ¾¾ hours. and Ü ½ . (1.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 5. 6). Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. (5. 6). E6. ¿ ½ ¼¼¼ are the smallest values satisfying the above equations. (9. (3. The goal is to ﬁnd the smallest integer values of and so that Ò is a minimum and ¼ By numerical search we ﬁnd ¾´ · ½µ ½ ¼¼ ¼¾ . (11. 10. Thus Ü ½½. 0. Thus and Ò ¾½¼ transfers is the minimum. 6. (3. 8. The license number is 741. 0). the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. 4. E4. ¼. The trick here is that 1 knot = 1 nautical mile per hour. 4. . 6. 0). 9). (1. One of the terms is ´Ü Üµ. (7. The number is 381. At ﬁrst glance. (0. the next greater number is 7. (1. 6). 7. (9. ¾ gives Ø E5. 8. 0.729.654.

1 comes back. C and D go across. 9. 2 comes back. E12. (a) 1 and 3 go across. 3 comes back. 1 and 4 go across. A and B go across.66 R. E15. Imagine intersecting the potatoes with each other. 1 comes back. 1 and 2 go across. The ﬁgure below shows how to trap the knight after 15 moves. A and B go across. See the array below. A comes back. The path of their intersecting surfaces is a desired path. B comes back. 8. and 10 go across. (b) 1 and 2 go across. The solutions are shown in the following ﬁgures. HESS E11. 1 and 3 go across. 8 and 9 go across. 6. 1 comes back. E14. I. 1 comes back. Yes.and 7 go across. 1 and 6 go across. 1 comes back. . E13. 1.

See the ﬁgure. 11. 0. Burn A from both ends. 3. (19. 2. 126. (29. (21. 4). 0. 1. 2. (31. 6). 11. 0. 0). 6. (35. (18. 10). (28. 146. 6). 0. 0. 11. (33. (23. 10). 6. 0). 6). (11. 11. (25. 0. 145. (18. 10). 11. (37. (30. (37. 1). 124. (36. (19. (29. 4. (30. Cut B at that corresponding point. (26. (24. (35. 11. 1. 11. 0. 1. 0). 148. 1). the probability of a right triangle is ½ ¾½ or . 1. 128. 0. (31. 6). (25. (21. as seen from the list below. 0). Cut the lace in half. 0) (b) 40 moves. 4. (25. (24. 0. 138. 0). (22. 127. 156. 0. 0. This is the shortest time interval that can be measured. 0). (35. 10). 6). 147. When C is consumed. (30. 1. 134. (24. 5. (24. (24. 0). 3. (34. 11. (31. (36. 0). 0). 10). Solutions to Medium Problems M1. At the same time. (33. (25. 6). (27. 2). (18. 0). In 3. 178. 158. 5). 0. (31. 0. 11. 0). 0). The probability of an acute triangle is ¿ ¾½ or ½ . 1. 3). (29. (22. 0. D will ﬁnish burning. If the hour hand is exactly on a second mark then the second hand will always be on the 12. (36. 6. 6). (30. 10). (25. 136. (31. 1. (29. 1). (23. There are 7 ¢ 6/2 = 21 choices using vertex 1. (26. 11. 10). (29. 168. (26. 5). 10). 1). 11. (24. 10. 6). 9. E17. 11. 6. 10). (28. 4. (29. 0). 125. noting the point that burns last. 10. 1. 11. (30. 0). (17. 5. (23. (20. 6). 0). 4). 0). M2. 0). producing pieces A and B. 11. 6. 5. 0). 0). 0). and the time is 8:24. (19. 0. 0). 10). 0). (19. (22. 0. 0. 0.PUZZLES FROM AROUND THE WORLD 67 E16. 4. (23. 0). 0). (17. 9. ½¿ ½¿ ½ Ê 123. 1). (18. 11. 0. 4). start burning C from both ends and D from one end. (23. 0. 6). 11. 3. 0. 0). 0). 1). (30. 3). For the second hand to be 18 second marks ahead of the hour hand the hour hand must be at the 42 nd second mark. (25. 167. 3. producing pieces C and D. 2). 6.75 minutes. 0. 1. (34. (20. 0). (23. 0). (a) 32 moves. 4). 11. 10). 5. 10). 1). 10). 10). (25. . 0). 6). light the other end of D.

Then ¾Ô ¿Ò ½ and Ô¾ ¿Ñ · ½. The ﬁnal two tricks are won by South’s long diamonds. East playing the 9. Ü¿ ¾Ü ¼. ¿. HESS M3. we take the limit of ½¾¼´¾Ò ½µ Ò ÓÖ Ò ´Òµ as Ò goes to 0. Ô M7. For ¾. M6. Ë ¹ À ½¼ Â Ë À ¿¾ ¿¾ ¿¾ ¿¾ Ë ÃÉÂ½¼ À ¹ ¹ ÃÉ Ë ¹ À ÃÉÂ½¼ ÃÉÂ ½¼ M4. The ﬁrst ﬁve tricks are alternate spade and heart ruffs by North and West. I. North next leads a heart which West trumps with the queen. The next four tricks are won by South with the 5432 of clubs. North and East discard all their diamonds on these tricks. There are six ways as shown below.68 R. so that ¾Ô · Ô¾ is always a Ô Ô ¿ we have Ô · ½ or . with East underrufﬁng North each time. The only real root is Ü ¿ we get the prime 17. North wins the sixth trick with the 10 of diamonds. For multiple of 3 for Ô ½. ¿½ ½¾¼ ÐÒ ¾ ¼ Ü M5. The hand below does the job if the play goes as follows. The series can be expressed by a simple continuous function: ´Òµ To get Ü.

3. The triangle they form will always include a vertex with an angle of 36 degrees as shown below. (b) Weights of 2. Assume that three vertices do fall on grid points. 5S. P3S. 9. P. Note that a 1-pound package is the only one that wont balance or lift a 2-pound weight. (a) For 5 buttons there are the following types of combinations where S is a single button pushed and P is a pair of buttons pushed: S. If these lines are interpreted as theÔ diagonals of the planks then ½ ¿ ¿ ¼ feet.PUZZLES FROM AROUND THE WORLD M8. M10. Then Ó×´¿ Æ µ ½· Ô Ô But is irrational and therefore cannot be the ratio of two integers. 4S. P2S. 6. M11. PS. In the ﬁgure below the oblique lines each have a length of ´¾¼ ¾µ ¿ ¾ ¼ feet. The longest musical composition without three consecutive repetitions of any sequence are AABABAABABAABAAB and its reverse. 2S. and 27 do the job. (a) Weights of 1. 2PS. the planks can be marginally longer than Ä Ô 69 M9. 3S. and 18 do the job. 2P. This contradiction proves that three vertices of a regular pentagon cannot lie on grid points. M12. Place that vertex at the origin and suppose the other two vertices of the triangle are at grid points ´ µ and ´ µ. Corresponding to each of these types the Ô · ´ ¾ · ¾µ´ ¾ · ¾µ ¿· Ô ´ · µ¾ ¾ · ¾ µ´ ¾ · ¾ µ ´ .

I. 5S.70 R. The number of combinations is · ¾¡ ¡ · ¡ · ¿¡ ¡ · · · ¾¡½ ¾¡¼ ¾¡¾¡¾ ¾¡¾ ¾¡¾ ¾¡¾¡¾ ¿ ¾ ½ ¼ · · · · · · ¾¡ ¿¡ · · ¾¡¿ ¾¡¾ ½ M13. P4S. Shown below. Each number is the length of the shorter side of the rectangle. 3S. 6S. and 3P.123. 2P.789. M15. PS. P. . A nine-rectangle solution is shown below. 2P2S. P3S. M14. 2PS. 1. 4S. P2S. 2S.465. HESS number of combinations is · ¿ · ¾¡ · ¾ ½ ¼ ¾¡¿ ¾¡¾ ¿¡ ¡ · ¿¡ · · ¾¡½ ¾¡¼ ¾¡¾ ¾¡¾ · · · · ¿ (b) For 6 buttons the types are: S.

7) and concludes he has 3 or 5. then B would conclude he has a 5. The snail eventually arrives at the far end when ½· ½ · ½ · · ½ ½¼¼. 3) and would know his number is 7. the 100-foot stretch causes the far end to move 1 foot farther from the snail. 3) would immediately know he had 5. Since A doesn’t know his number. (b) When the snail has traversed 99 % of the band. (3) If you have a 7 then person B with a 5 sees (7. An eight-triangle solution is shown below. That happens for Ø ½ ¼ ¾ ¢ ½¼ ¿ ¾ ¿ Ø minutes. Since A doesn’t draw this conclusion you know you don’t have a 5. This 1 foot is overcome by the snail’s normal progress during that minute. A would conclude that he has a 7. Since B doesn’t draw this conclusion you know you don’t have a 7. The farthest is at point P one quarter of the way along the diagonal on the 1 ¢ 1 face from B. But if he has 3 then he reasons that person B sees (5. M18. 3) and would know his number is 5 or 3. or Ù ¾¾ ¢ ½¼ ¾ Ø . . (a) During the Òth minute the band is ½¼¼Ò long and the snail travels a fraction of the band equal to ½ ½¼¼Ò. B can eliminate 3 because anyone seeing (3. ¿ M19. 7 or 11 on your forehead. 5) and concludes that he must have 3 or 7. Therefore you have 11 on your forehead. M17. (1) The three possibilities are that you have 5. Since B doesn’t know his number. (2) If you have a 5. But if he has 3 then he reasons that person A with 7 sees (7. (a) Unfold the box and examine shortest path distances from A to various points. This happens ½ ½ when ½ · ¾ · ½ · ¡ ¡ ¡ · Ù . then person A with a 7 sees (5.PUZZLES FROM AROUND THE WORLD 71 M16.

12. (15. Solutions to Hard Problems H1. 12.0119 . 9) and (14. 14. 12. The points are ´ ¿ ½µ ¾ of the way along each diagonal for a distance of 3. 14. 11. I. 13. 13. 11. 13. 12. 12. 14. (15. M20. (15. The number in each triangle indicates its area relative to the other triangles in the ﬁgure. Shown below. 11. 10.72 R. There are 5 subjects with 15 marks possible in each. 9). . 6). Humpty Dumpty got a 10 in arithmetic. 8). 12. (15. HESS (b) The two points are along the diagonal near P and its opposite Ô point on the ½ ¢ ½ face that includes A. 14. 14. The scores for the Good Eggs are (15. 13. (15. 11. 10). 7). 8). 10. There are 7 Good Eggs.

where × is the side length and is the distance from the center of the polygon to the center of a side: ´¾ µ × ÓØ´½ ¼ Òµ. If Sir George travels at speed Ú then × Ú ´½ ¼ Òµ Ò where 1440 is the number of minutes in a day. BCB rolls along the plane. To maximize the area is to ﬁnd the Ò that maximizes Ú¾ ÓØ´ The area is largest for Ò ½ ¼ µ´½ ¼ Òµ¾ ´ Òµ Ò ½ . where . The distance from B to the distance BC. Let Û width of the building. 0. For a full rotation of the ball on the circle the point P will execute ¾ rotations about the axis. as shown below. with 159. Three equally spaced punches in a straight line do the job if the square of the spacing is irrational. H5. H4. Then the ¾Ø Ò Û Ö· Ö ÖØ Ò Û ¾Ö · ¾ · ¾ ¿ ½ ¾Ö Ø Ò · Û ¾Ö´ · µ These equations can be iterated to produce ft.PUZZLES FROM AROUND THE WORLD 73 H2. Its area is Ò× ¾. the point P will have coordinates È Ô Ô Ö × Ò ¾ Ö´ Ó× ½µ ¾ Ö´ Ó× · ½µ ¾.1968 Ú ¾ H3. Thus B will be in contact with the axis AC is ¼ of the Ô horizontal plane when ¼ rotations about the axis AC have occurred.300. TheÔ cone. For Ö ½È ´ ¼ ¿ ¾ ¼ ¾ ½¼ ½ ¼ ¼ ¼ ½ µ its initial position was (0. 1). Imagine a regular polygon of Ò sides after Sir George’s trip. Point P will return to the top again as Ô well. . and stretch in the band due to the building. Relative to the center point of the ball. Ö earths radius. height of the building. Consider the ball rotating about the axis AC without slipping.

(a) (b) ¼Æ ¼Æ ½ ¼Æ ¾ ¼Æ ÓÖ ¿¼¼Æ. ÓÖ ¿¼¼Æ . HESS H6.74 R. The solutions are shown below. I. (c) ¼Æ ¼Æ ÓÖ½ ¼Æ (d) and · · ¼Æ ¼Æ ¼Æ ½ ¼Æ ¾ ¼Æ.

(a) (b) ½ ÖÕÔÔ Ô½ ¿ ½ ¾¾ ¿ ½ ½ ¾¾ ´Òµ Ô ´Òµ ´Ñµ can be made arbitrarily close to 0. which contains ½ . The perimeter of (c) Deﬁne ¡ ¡ ¡ Ô ¡ ¡ ¡ where there are Ò square roots. H8. Pick and as any digits other than 0. ´ µ can be put in the interval ¼ ¾ ¼ . Ó×´¾ µ · Ó×´¾ µ ¾ Ó× · ¾ Ó×´ µ Ó×´ and µ. The area of such a shape is ½·´ µÖ¾ . (g) H7. (a) Imagine an expanding soap bubble in the unit square.PUZZLES FROM AROUND THE WORLD 75 (e) (f) · ¼Æ ¼Æ. . Clearly . For the right . There are an inﬁnite number of choices of Ñ and Ò followed by the appropriate to make an inﬁnite density of ´ µ’s in the ¼ ¾ ¼ interval. Its form will take on the shape of four quarter circles of radius Ö at the corners as shown in the ﬁgure. or ½¾¼Æ.

There are 7 circles on each end with 105 sets of 3 circles in the middle. H9. The smallest rectangle found to date containing 329 circles has 13 circles on each side of 101 sets of 3 and measures 2 ¢ 163. The maximum ratio È Ö ¼¾ ¼ . Raise S and T appropriately above the . I. Now consider two such circles of different sizes as shown. They deﬁne the four tangent lines drawn to produce ABCD as shown.25 for the entire square or a circle inscribed in µ the square. HESS the shape is È Ô ½ ´¾ · ¾ Ö · Ö. The ratio È takes on a maximum when Ö ¼¾ ¼ . For Ò ½ there is just enough room for 329 circles. The point S may be raised above the plane to give unit altitudes if the inscribing circle has radius ½. A view of the pyramid looking down from S perpendicular to the plane ABCD will look like the ﬁgure below.9973967 . The circles need to be packed as shown below. H10. (b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved. Each has a radius ½.76 R. Note that the ratio È is 0.

7892 0. Clearly ST is not perpendicular to the plane. · È ´ ¼ ¿¼µ Õ È ´ ¼ ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ .2270 1.2069 1. ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ . Ô Õ At a score of Ò games to Ò (Ò ¼ to 5). È ´ ¼ ¿¼µ Ô · ÕÈ ´ ¼ ¼µ. ¾ ¾ Õ´ È ´¿¼ ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ . My uncle is expected to take the longest time to dress on Saturday and on Friday he is least likely to get a pair from the ﬁrst three socks chosen.2511 1. Let Ô the probability of the server winning a point. È ´¼ ¼µ È ´ ¼ ¿¼µ when Ô¿ Ô¾ ¾Ô ½.7805 0 .8163 0. Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1.2801 1. A computer program was written to calculate the probabilities shown in the table below. At Ò games to Ò.8027 0. and let Õ ½ Ô. which leads to Ô ¼ ½½ ¿ ¾ .7849 . Let È ´ µ be the probability of the server winning the game when the server has points and the receiver has points. ¾ H12.3074 1. H11. which leads to Ô ¼ ½ ¿¿ ¼ .8286 0. È ´ ¼ ¼µ ÔÈ ´ ¼ ¿¼µ · ÕÈ ´¿¼ ¼µ. È ´¼ ¼µ È ´¿¼ ½ µ when Ô ¾ Ô ¿.3586 Pairing probability 0. This leads to ¾ È ´ ¼ ¼µ Ô¾Ô Õ¾ . ·¾ Ô ´½ ½ Õ µ È ´¼ ¼µ .PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit. È ´¿¼ ¼µ ÔÈ ´ ¼ ¼µ.

Using a multiprecision continued fraction program. . H15.546.321 is the ﬁrst of 5 in a row. . ´ ¾ ½µ ¾. (d) Yes. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô . The product and sum two years ago could be achieved by (2.444 years. HESS H13. prime. The product and sum could be achieved by (4. we get (a) (b) H16. 15). 12). (a) 45 years. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½.641. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. 7.042. (b) Length ¾ ¼¾½½¾ ¿ as shown. (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ . We look for cases where (1) the oldest child is under 20. ¼ ¼½¼ . (b) 605 years. 8. 16). (a) Length ½ ¿ ¾½ ½ ¼ as shown. H17. which must have been guessed by Smith. ´ ¾ · ½µ ¾ prime. I.093. ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . (2) the younger two children have different ages. · ½. 6. which must have been guessed by Jones two years ago. There is only one set of ages that accomplishies this: (5. ¿ ½¾¼. ´ prime. ¼ ¼¼ ½ ¼. ¾ ½µ ¾. (c) 17. . a=10.613. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago. Solutions occur for (b) ¾ ½.512. (c) Unknown. · ½.78 R. H14.

Let Ö and Ê dug on its surface. There are three classes of channels that can be (a) It is clear that many channels with radius Ö are possible. H19. which is the intersection of the plane shown by the slanting line and the torus. H18. (b) Channels with radii between Ê ¾Ö and Ê are possible. From the descriptions of the ﬁrst two students they must have dug channels of type (b). . (1) Deﬁne Á ´Ò Üµ Clearly Ò ´ Òµ ´ ÒµÜ Ò ½ ´Òµ ¾ ¾Ò · ¿ Á ´Ò ½µ ¼ ¼ Ü ´ ÒµÜ . the other case gives Ê Ö ¿¼ and Ö ¾ . The third and fourth students must have explored channels of type (a) and (c) in some order. H20. The venture should be undertaken since the volume in cubic spandrals can be determined using calculus to be Î ¿ ´½ ¿¿½µ¾ ¿´ ¼µ¿ ½¼ ¾ ¾¼ It is noteworthy that the volume doesn’t depend on the polar or equilateral radii of Alpha Lyra IV. One case gives Ê Ö ¾ and Ö ¿¼. The ﬁgure below shows a cross-section of the torus. The ﬁrst case is not possible because Ê ¾Ö ¼. (c) A less well-known third type of circular channel with radius Ê Ö is possible. Thus the second case applies and ¾ Ê ½½¼ .PUZZLES FROM AROUND THE WORLD 79 Six in a row is impossible because three consecutive even special years cannot occur.

The only pole of Õ´Þ µ ´Ò·½µÞ within is a double pole at Þ ¼. and by Cauchy’s Theorem we have ½ ¾ Thus ÞÕ´Þ µ ´Ò·½µÞ Þ Þ ¾ Õ´Þ µ ´Ò·½µÞ Þ ¼ ¾ ´¾Ò · ¿ µ ´Òµ ¾½ ÞÕ´Þ µ ´½ · Þ µÒ·½ To evaluate ´Òµ.80 R. Thus Þ ¾ Õ´Þ µ ´Ò·½µÞ is regular within . (3) The summation over can be carried out explicitly to give ´Òµ ¾ ½ ¾Ò· · ¿ ¾ Þ Þ ½ ´½ · Þ µ Þ ´½ · Þ µÒ·½ (4) Deﬁne Õ´Þ µ ½ ½ ´½ · Þ µ Þ . HESS (2) Since ´ ÒµÞ is an entire function in the complex plane. which still avoids enclosing the poles of Õ´Þ µ. it follows from Cauchy’s Theorem that ´Òµ where must enclose Þ ¾Ò · ¾ ¿ Ò ½ and will be taken to be Þ ´Ò·½µÞ ½ ½ ´½ · Þ µ Þ ¾ ½ ¾ ¼ Þ ´ ÒµÞ ´½ · Þ µ ·½ ¾. I. we expand and deform to a contour the ﬁgure. Thus ¼ like that in ¼ gets arbi- ´Òµ ¾½ ¾½ Þ ¼ ÞÕ´Þ µ ´½ · Þ µÒ·½ ÞÕ´Þ µ ½ · ´½ · Þ µÒ·½ ¾ Ê ½ ½ ÞÕ´Þ µ ´½ · Þ µÒ·½ . As trarily large it still encloses the same poles as .

Call the Ø such pole in the upper half-plane Þ Ü · Ý . The table below gives numerical values of the ﬁrst few. which reduces to ´½ · Ü · Ý µ´Ü Ý µ ´ Ó×´Ò · ½µÝ × Ò´Ò · ½µÝ µ where ´Ü¾ · Ý¾ µ ´Ò·½µÜ .5012690 3. Then Ý approximately equals ½·ÐÒ´ µ . For larger deﬁne ´¾ ·½ ¾µ .6745053 3. A pole of Õ´Þ µ occurs Þ . (5) On Þ Ê for sufﬁciently large Ê.8221528 Ý 7. Ü 1 2 3 4 5 6 7 (6) The contribution to 2. where Þ at ½ · Þ Ü · Ý. ¼ for all . Since the equation in Ý is even.1512074 45.PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ. Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ.6406814 3.4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ Þ µÕ´Þ µ ´½ · Þ µÒ·½. ¾ × ´Ý · µ Ü Ý .8790560 20. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to inﬁnity.0262969 3. not including its pole at Þ ¼.8505482 39. The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ.2916783 3.0888430 2.2238350 26. Combining the poles in the upper and lower half-planes and some rearrangement ﬁnally produces ´Òµ where ½ ½ ÒÜ × Ò´ÒÝ µ (a) Since Ü Ü¾ · Ý ¾ Ü¾ · Ü · Ý¾ Ø Ò´Ý · µ Ý ÒÜ goes to zero as Ò goes to inﬁnity.5432385 32.4614892 13. × ÒÝ Ý we can look for poles in the upper half-plane only and reﬂect each one into the lower half-plane.

“Puzzle Corner. E5. From Nobuyuki Yoshigahara (private communication). E11. M5. Thus ´½¼¼¼µ is very nearly ½ ½¼¼¼Ü½ × Ò´½¼¼¼Ý½ ½ µ. 8. From David Singmaster (attributed to Roger Penrose.82 R. M6. Schoenberg. Submitted by Lawrence Kells. Reprinted with permission of the Canadian Mathematical Society. Vol. p. E7. E1. E13. edited by Alan Gottlieb. 3. 1992. All rights reserved. E14. E15. From Nobuyuki Yoshigahara (private communication). . M3. E16. From David Singmaster (private communication). Unknown.” Crux Mathematicorum. E10. Problem 3. (a) and (b) original problems. E4. M4. From Nobuyuki Yoshigahara (private communication). M2. ´ ¼¼µ Sources ¼ ½ ¢ ½¼ ¾ ´ ¼½µ ½¾¾ ¢ ½¼ ¾ All problems written by the author unless otherwise indicated. All rights reserved. 302. May–June 1990. MIT 55. Problem 1. 1985. All rights reserved.” in MIT’s Technology Review. Reprinted with permission. No. giving ´½¼¼¼µ ¢ ½¼ ¼ (c) From observing the behavior of × Ò´ÒÝ½ ½ µ one can determine the smallest Ñ where the magnitude of ´Ñµ is less than the magnitude of ´Ñ · ½µ. It is at Ñ ¼¼.” in MIT’s Technology Review. “Olympiad Corner. 10. “Puzzle Corner. HESS ½ ½ (b) For Ò ½¼. p.” in MIT’s Technology Review. From Yoshiyuki Kotani (private communication). E8. From Dieter Gebhardt (private communication). Reprinted with permission. MIT 55. Problem 3a. page 178. E12. All rights reserved. Problem 7 from the 1980 Leningrad High School Olympiad. MIT 59. October 1992. Reprinted with permission. p. E3. J. E2. Unknown. 1990. Submitted by S. October 1987. 9. 1983. private communication). Unknown. edited by Allan Gottlieb. From Nobuyuki Yoshigahara (private communication). Unknown. “Puzzle Corner. Reprinted with permission. edited by Alan Gottlieb. I. “Puzzle Section. No. p. ´Òµ is dominated by the ﬁrst pole and its reﬂected pole. Vol. Modiﬁcation of Problem 1. J.” Pi Mu Epsilon Journal. From David Singmaster (private communication). Einhorn and I. Unknown. All rights reserved. 1987. From Nobuyuki Yoshigahara (private communication).

M14. p. Pi Mu Epsilon Journal.” in MIT’s Technology Review. MIT 47. Purdy. Vol. Akiyama. All rights reserved. From Yoshiyuki Kotani (private communication). 12. “Puzzle Corner. Submitted by David Singmaster. No. (b) original problem. 1994. From Karl Scherer (private communication).” in MIT’s Technology Review. All rights reserved. p. May–June 1994. All rights reserved. Problem 1995. 8. H6. p. All rights reserved. Problem 2. January 1989. p. Crux Mathematicorum. 155. 2. edited by Allan Gottlieb. From Robert T. Reprinted with permission of the Canadian Mathematical Society. 10. 1988. 282. Wainwright (private communication). Reprinted with permission of the Canadian Mathematical Society. 142. 1994. Problem 944. Crux Mathematicorum. p. 1995.” MIT’s Technology Review. Part (d): Problem 7. February–March 1997. M11. Problem 1289. Submitted by Bob High. 1987. From Karl Scherer (private communication). Part (a): Problem 870. Submitted by Jerzy Bednarczuk. p. Unknown. Vol. Problem 1194. Vol. From Robert T. No. MIT 55. Submitted by Sydney Kravitz.” Pi Mu Epsilon Journal. Crux Mathematicorum. M16. MIT 53. Vol. Reprinted with permission of the Canadian Mathematical Society. H11. “Puzzle Corner. 10. 13. Crux Mathematicorum. 526. H8. Reprinted with permission. H1. Submitted by Joe Shipman. 86. Reprinted with permission. Crux Mathematicorum. No. “Puzzle Corner. 10. Vol. M18. Crux Mathematicorum. 180.” in MIT’s Technology Review. H5. edited by Allan Gottlieb. Erdos and G. 1989. “Puzzle Corner. 7. 3. p. 1986. M15. 8. Reprinted with permission of the Canadian Mathematical Society. Reprinted with permission of the Canadian Mathematical Society. Submitted by Cops of Ottawa. MIT 53. 1984. From Robert T. Reprinted with permission. H10. edited by Alan Gottlieb. p. H12. All rights reserved. H3. A special case of a problem by P. 20. H9. p. M17. All rights reserved.PUZZLES FROM AROUND THE WORLD 83 M7. Wainwright (private communication). Reprinted with permission. 1987. 285. No. All rights reserved. 10. (a) Unknown. Problem 4. 13. First published as Problem 3. 12. Vol. M19. Submitted by Carl Sutter. All rights reserved. Submitted by Nobuyuki Yoshigahara and J. No. (a) From Yoshiyuki Kotani (private communication). H2. “Puzzle Section. 290. All rights reserved. Vol. p. Problem 3. Wainwright (private communication). 1988. 9. Reprinted with permission. 1997. p. 1986. Part (b): Problem 1225. All rights reserved. 5. and then as Department Problem 860. No. All rights reserved. M20. Vol. October 1988. No. No. edited by Allan Gottlieb. Reprinted with permission. .

All rights reserved. Problem 1231. 9. Submitted by Richard I. p. 1986. H18. All rights H17. 118. Unknown. 13. Part (a): Problem 1190. H20. 12. Unknown. Vol. 242. Wainwright (private communication). HESS 1987. From Robert T. I. p. Crux Mathematicorum. reserved. Hess from Albert Latter (private communication). H19. Vol. 4. No. Reprinted with permission of the Canadian Mathematical Society. No. Reprinted with permission of the Canadian Mathematical Society.84 H14. R. . Crux Mathematicorum.

It soon became necessary to devise a classiﬁcation scheme. Or try using 0. The next solution in my collection is Figure 7 by Mike Guy (March 1960). then there are 19 shapes (Figure 1). Question: Will the 19 shapes cover the ﬁgure? It took O’Beirne some months to discover that the answer to the question is “Yes!” Figure 3 was discovered in November 1959. since it was not easy to decide whether a solution was new or not. 6. One of these is the regular hexagon. Rearrange pieces 1. among the shapes that can be formed by adjoining six equilateral triangles. 85 . and in January 1960 he found the solution shown in Figure 6. and 8 from Figure 3 and replace them in a different way. while seven did not. 5. Guy Tom O’Beirne is not as celebrated a puzzler as he deserves to be. and . O’Beirne thought that the result would be more pleasing if the Hexagon were in the center. Martin [3] uses O’Beirne’s names for the shapes but does not distinguish between reﬂections so his problems only involve 12 shapes. 3. ½. [1] and the not very accessible reference Guy [2]. He showed us many remarkable puzzles. Remove pieces ½. and in Figure 7. ﬁve had reﬂexive symmetry. I was sure that I would ﬁnd references to the Hexiamond in Martin Gardner’s column. and in Figure 5. 6. Nor is it mentioned in O’Beirne’s own book [4]. When I began to write this article. but the one that grabbed us the most was the Hexiamond. and ¿. since everyone wanted to try it at once. No one went to bed for about 48 hours. In the interim he had found solutions with the Hexagon in two other of its seven possible positions (Figures 4 and 5). giving a ﬁgure (Figure 2) having the same total area as the 19 shapes. but I haven’t found one yet. particularly on this side of the Atlantic.O’Beirne’s Hexiamond Richard K. It was soon after this that O’Beirne visited the Guy family in London. But it does appear in his column in the New Scientist. It must have been in 1959 that O’Beirne noticed that. which can be surrounded by six more hexagons. and then by twelve more. 1. . 3. ¿. 7. We became adept at ﬁnding new solutions based on the old. Maybe the only other places where it has appeared in print are Berlekamp et al. If we count reﬂections as different. 5. and several copies had to be manufactured.

with its handle on the left. to the Signpost. as do pieces 1. to the Crook. which appears to be turning to the left. GUY Figure 1. a parallelogram drawn in the opposite way from the usual textbook fashion. The symmetries of the Hexagon are those of the dihedral group D6. with its hook on the left. By subtracting 10 from the Hexagon. . and 9. Crown. 7. and Lobster each have a single symmetry of reﬂection. . exist in enantiomorphous pairs. and the Sphinx. Pieces with odd-numbered labels have no symmetry and. together with those that have rotational symmetry only. Notice that 0 has the greatest symmetry. K. ½.86 R. . We don’t know how O’Beirne decided whether to give a piece a number less than or greater than 10. but our mnemonics are as follows. . to the Snake. Pieces 4 and 8 have only rotational symmetry and exist in enantiomorphous pairs. sailing to the left. with its pointer on the left. which have no symmetry. Negative (bar sinister?) labels are given to the Bar. and the other even numbers have reﬂective symmetries. The Chevron. and subtracting each of the labels 11 to 19 from 20 we arrive at the labeling in Figure 1. and each is positioned by its apex. The nineteen Hexiamond pieces. The Butterﬂy has the same symmetries as a rectangle: Its position is described by that of its body. to the Hook. O’Beirne had already suggested a numbering of the pieces from 1 to 19. . whose head is on the left. to the Yacht. the multiples of 4 have rotational symmetry through 180 degrees. ¿. . 5. 3.

Hexagon. Hexiamond board. Crown. and Lobster. This doesn’t completely describe the solution. 1959). ¿ 60-03-27 We classify solutions with a string of ﬁve labels. Figure 4.O’BEIRNE’S HEXIAMOND 87 Figure 2. O’Beirne’s ﬁrst solution (November 30. giving the positions of the Chevron. Butterﬂy. ¿ Ã¾ Ð 60-01-05 Figure 5. Ü Í¿ 60-01-21 Figure 7. but gives enough detail to enable comparisons . Ô Ü 60-01-10 Figure 6. Figure 3.

0. 2 or 4 for a subscript. Then append a subscript ¼ ½ Ü indicating its position on the clock. piece number 0. We won’t count a solution as different if it is just a rotation or a reﬂection. 4. If the Chevron is central (positions b to g in Figure 8). B. A to G. K. 8. A. 2. reﬂect the board left to right in order to bring it onto the right half. shown in Figure 10.) Next use Figure 9 to describe the position of the Hexagon. Read off the lowercase letter at the apex. s. reﬂect the board if necessary to bring the Butterﬂy into the right half. Coding the Hexagon. If both Chevron and Hexagon are symmetrically placed. f and m are equidistant from opposite sides of the board. Then. l. piece number 2. The subscripts a. Note that the letters a and n are missing. reﬂect the board if necessary to bring the Hexagon into the right half. If all three are symmetrically placed. reﬂect if . the center. so ﬁrst rotate the board so that the Chevron. it is possible to place the Chevron in such positions. as in the ﬁrst solution in Figure 14. and are given only 0. and G. of the Chevron from Figure 8. and t are omitted.88 R. F only even ones. This is indicated by a lowercase letter. piece number ¾. or Ü the side of the board it is nearest to. is unique and requires no subscript. The subscripts on C are about ½ more than their ¾ clock hour. (The exact ﬁgure is 89. 6. A and E have only odd subscripts. e.5%. but it’s clear that they can’t occur in a solution. depending on how far it is from the center. This will be a capital letter. Figure 8. Figure 9. together with an even subscript.” the edge that bisects it. is that of its “body. Coding the Chevron. since placing the Butterﬂy there does not allow a solution to be completed. More than three-quarters of the positions found so far have the Chevron in position h. is pointing upward. if the Chevron is on the left-hand side of the board. The position of the Butterﬂy. D. GUY to be made quickly.

t. these are in the same positions as the capital letters used for the Crown. is given by the capital letter at its apex. Coding the Crown. is located by a lowercase letter in Figure 12. A. Figure 11. in Figure 11. Again the subscript is even or odd according to whether the piece is to the left or right of the axis of symmetry of the board in the direction its tail is pointing. M. The position of the Crown. Positions S and T are symmetrical and carry even subscripts. R are omitted.O’BEIRNE’S HEXIAMOND 89 Figure 9. g. Continued: Coding the Hexagon’s position. The subscript is even or odd according to whether the Crown is to the left or right of the axis of symmetry of the board in the direction in which its own axis of symmetry is pointed. and r don’t lead to legal solutions. Q. Coding the Butterﬂy. piece number 6. Positions C. piece number . F. and u are symmetrical and carry only even subscripts. Positions s. necessary to make the apex of the Crown point toward the right half of the board. c. as they don’t allow legal solutions. The Lobster. Except for a. . Figure 10.

We have found 247 of these cases (no fewer than twenty-six bit the dust during the writing of this article). There are 508 essentially different relative positions for the Chevron and Hexagon that have not been proved to be impossible. and 1 legal positions . 39.90 R. Remember that the code doesn’t specify the solution completely. note that O’Beirne’s ﬁrst solution (Figure 3) is of type Ô Â . although some of these turned out to be so. h. K. there are respectively 21. Continued: Coding the Butterﬂy’s position. which will be deposited in the Strens Collection in the Library at the University of Calgary. Coloring arguments don’t seem to lead to much restriction. With the Chevron in positions g. There are already more than 4200 in the collection. and o. Now that you know how to classify solutions. (Can you ﬁnd a larger class?) How many solutions are there? A wild guess. but perhaps some reader will be more perspicacious. is about 50. GUY Figure 10.000. It isn’t very easy to give a good upper bound. Continued: Coding the Crown’s position. Figure 11. based on how rarely duplicates appear. for example. There are at least ten solutions of class Ô ¼ ¼ .

the discerning solver will soon wish to specialize. suggested looking for solutions on a machine. how symmetrical a solution can you get? Figure 14 shows two solutions. then a graduate student at the University of Calgary. Figure 13 shows two of Bert Buckley’s machinemade solutions. . For example. As we go to press. each with as many as 11 of the 19 pieces symmetrically placed. the ﬁrst found by John Conway and Mike Guy in 1963. and solutions are known in all of these cases. Coding the position of the Lobster. Marc Paulhus has established that there are just ½¿ · ½ · ½ · ½ · ½ · ¾½ · ¿ · ¾ · ¾ · ¿¼ · ¾ · ¿¼ · ½ · ¾¿ · ¿¿ · · ¿¿ · ¾¾ · ¿¾ ¿¿ different relative positions for the Chevron and Hexagon that yield solutions. In 1966 Bert Buckley. but after a few months of intermittent CPU time on an IBM 1620 he found half a dozen or so solutions from which it was possible to deduce another ﬁfty by hand. I didn’t think he would be successful. the second by the present writer. With such a plethora of solutions. for the Hexagon.O’BEIRNE’S HEXIAMOND 91 Figure 12.

Figure 15 shows eight solutions. We have always drawn the hexiamond with one of these directions vertical. Some solutions come in large groups. in Figure 14. Solutions found by machine by Bert Buckley. found by John Conway in 1964. but you may prefer to put one of them horizontal. rearrange pieces ¾¾ or ¿¿ and note that the hexagon formed by ¼½½ has the symmetries of a rectangle. or ½¿ . after which you can swap ¼¾ with ¿ . You can . at angles of 60 degrees to one another. Conway showed that you can’t have more than 15. For example. each with 11 symmetrically placed pieces. respectively 13. Æ Figure 14. K. All edges of the pieces lie in one of three different directions. It seems unlikely that anyone can beat 11. pieces placed symmetrically with respect to the two kinds of axis. GUY ½ ÒÜ ¼ Ó¿ 66-05-30 ½ÕÜ Ó¿ 66-08-15 Figure 13. rearrange pieces ¾¿¿.92 R. ¼Ô Á Notice that there are two different kinds of axes of symmetry. To deduce the others from the single diagram shown.

¼ Ç¼ Ð The most symmetrical solution? rotate ¼¾ ´¿ µ or ¼¾ ´¿ µ and produce two “keystones.O’BEIRNE’S HEXIAMOND 93 Figure 15. Readers will no doubt discover other curiosities and ﬁnd their own favorites. shows that keystones ´ µ can occur internally and don’t have to ﬁt into a corner. found by Mike Guy. and so on . Figure 19 is one of at least eight examples that have six pieces meeting at a point.” which can be swapped. ÔÍ has an inter- . Figure 16 shows a solution displaying three keystones that can be permuted to give other solutions. ½ Ô¼ Ü ×¾ has three Figure 17. Figure 17. nal keystone. here ¿ . In May 1996 Marc Figure 16. keystones. In an earlier draft I wrote that with modern search techniques and computing equipment and some man–machine interaction it has probably become feasible to ﬁnd all the solutions of the Hexiamond. Figure 18 has the Hexagon in position G and the other pieces forming three congruent sets.

94 R. Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to ﬁnd all the solutions. Their numbers. Six pieces meet at a point.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. Ô Ü (3-symmetry). classiﬁed according to the position of the Chevron. for those who prefer their puzzles to have just one answer. Figure 19. GUY Figure 18.717 C 6675 D 7549 E 11.518 A 75.489 B 15.447 F 5727 G 1914 I don’t think that this need take the fun out of one of the best two-dimensional puzzles ever invented. by the Hexagon’s position Ø Ù Total 264 1094 124. K. are 130 195 533 193 377 2214 111. I wrote to Kate Jones. president of Kadon Enterprises and discovered that she markets a set of “iamonds” . there are no fewer than 40 relative positions of the Chevron and Hexagon that determine such a unique solution: Ð × Ü ½ ¼ Ð Ø ¿ ¿ ¼ Ñ Ø Ü ¿ Ô Ô Ø ½ Ø Ü Ü ¾ ¾ ¼ Ô Õ ¼ Õ Ü Ö Ø ½ Ø Ø ¼ Ø Ü ½ ½ ¾ × Ø Ü Ü How long will it take you to ﬁnd them all without peeking at Marc’s database [5]? After writing a ﬁrst draft of this article. On the contrary.

787–788. Conway. doesn’t do justice Figure 20. Berlekamp. The numbers of iamonds. triamonds. Guy. 1982. and R. Figure 20. Kate Jones’ Iamond Ring. diamonds.” For this purpose the iamonds are not counted as different if they are reﬂections of one another. Winning Ways for Your Mathematical Plays. so. London. . if you want to use them to make O’Beirne’s hexiamond puzzle. J. H. you must get two sets. Academic Press. pp. References [1] E. R. K. where the last line counts reﬂections as different: Order 1 2 3 4 5 6 7 # of iamonds 1 1 1 3 4 12 24 with reﬂections 1 1 1 4 6 19 44 The Iamond Ring is a beautiful arrangement. discovered by Kate herself. in which each x-iamond is labeled Ü ½ Ü to the very appealing colored pieces of the professionally produced puzzle. are given in the following table. .O’BEIRNE’S HEXIAMOND 95 under the trademark “Iamond Ring.

Polyominoes: A Guide to Puzzles and Problems in Tiling. submitted. . New Scientist. 144–153.96 R. Soc. II pp. Assoc. Guy. especially pp. A database for O’Beirne’s Hexiamond. 104–106 and 152–153. Malayan Math. 316–317. of America Spectrum Series. Martin. GUY [2] Richard K. Math. Oxford University Press. 44: Pentominoes and hexiamonds. 9 (1960) pp. see Puzzles and Paradoxes No. Some mathematical recreations I. Nabla (Bull. K. 168–170. 97–106. 1991. O’Beirne. [3] George E.). [4] Thomas H. 259 (61-11-02) pp. 1965. pp. Puzzles and Paradoxes. [5] Marc Paulhus.

a court lady of the late 10th and early 11th centuries. a famous puzzlist in Japan. but their 97 . 16 centimeters wide and 11 centimeters long. (b) The Sei Shonagon pieces. She was the author of a book entitled Makura no Soshi (Pillow Book: A Collection of Essays). Sei Shonagon. The introduction bears the pseudonym Ganreiken. was one of the most clever women in Japan. (a) Tangram pieces. Sei Shonagon Chie-no-ita is a 32-page book. so I sent a report about the Japanese tangram to him. There are 42 patterns with answers. a little book about Japanese seven-piece puzzles was published. but nobody knows its real author. The tangram came to Japan from China in the early 19th century. and the Japanese edition of Qiqiaotu Hebi (The Collected Volume of Patterns of SevenPiece Puzzles) (1813) was published in 1839. Japan had a similar puzzle already. In fact. The book was called Sei Shonagon Chie-no-ita (The ingenious pieces of Sei Shonagon).Japanese Tangram: The Sei Shonagon Pieces Shigeo Takagi In 1974. In 1742. I received a letter from Kobon Fujimura. I heard that Martin Gardner had been planning to write about tangrams.

98

S. TAKAGI

shapes are inaccurate. The puzzle was introduced to the world by Martin Gardner in his book Time Travel ½ , and he elaborated as follows. “Shigeo Takagi, a Tokyo magician, was kind enough to send me a photocopy of this rare book. Unlike the Chinese tans, the Shonagon pieces will form a square in two different ways. Can you ﬁnd the second pattern? The pieces also will make a square with a central square hole in the same orientation. With the Chinese tans it is not possible to put a square hole anywhere inside a large square.”

A square with a center hole

I know of two other books that are collections of patterns of the Sei Shonagon pieces. In about 1780, Takahiro Nakada wrote a manuscript entitled “Narabemono 110 (110 Patterns of an Arrangement Pattern),” and Edo Chie-kata (Ingenious Patterns in Edo) was published in 1837. In addition, I possess a sheet with wood-block printing on which we can see patterns of the Sei Shonagon pieces, but its author and date of publication are unknown.

Some patterns formed from the Sei Shonagon pieces

½1988, W.H. Freeman and Co., NY

**How a Tangram Cat Happily Turns into the Pink Panther
**

Bernhard Wiezorke

Do you want to create, or better generate, a new two-dimensional puzzle? Nothing could be easier! Just take one of the many well-known puzzles of this type and submit it to a geometrical transformation. The result will be a new puzzle, and — if you do it the right way — a nice one. As an example, let us take the good old Tangram and try a very simple transformation, a linear extension (Figure 1).

Figure 1. Tangram submitted to a linear extension. The result is a set of pieces for a new puzzle, a Tangram derivative called Trigo-Tangram.

While the tangram pieces can be arranged in an orthogonal grid, the Ô pieces of our new puzzle, due to ¿ as extension factor, ﬁt into a grid of equilateral triangles, a trigonal grid. That’s why I gave it the name TrigoTangram. Puzzles generated by geometrical transformations I call derivatives ; our new puzzle is a Tangram derivative, one of an inﬁnite number. Depending on the complexity of the transformation, certain properties of the original puzzle are preserved in the derivative, others not. Tangram properties preserved in Trigo-Tangram are the linearity of the sides, the convexity of the pieces, and the side length and area ratios. Not preserved, for example,

99

100

B. WIEZORKE

Figure 2. Suitable problems for a puzzle derivative can be found by transforming problems of the original puzzle. For Trigo-Tangram, two problems can be generated by extending this original Tangram problem in two different directions. The transformation of the solution can be a solution for the derivative (right image) or not (upper image).

are the angles and the congruence of both the two big and the two small triangles. For our derivatives, we do not only need the pieces, we need problems as well. Problems? No problem at all! We just take some ﬁgures for the original puzzle and submit them to the same transformation as the puzzle itself. The images of the ﬁgures will be suitable problems for the derivative. In Figure 2 this is done for a Tangram problem and its solution. As you can see, the corresponding linear extension can be performed in two different directions, which gives us two different problems. And you can see another important fact: The transformation of the solution does not always render a solution for the derivative. Figure 3 shows some problems for Trigo-Tangram. So, take a piece of cardboard, copy and cut out the pieces, and enjoy this new puzzle. Here is another way to have fun with a puzzle and its derivative: Take the original Tangram, lay out a ﬁgure, and try to make a corresponding ﬁgure with the pieces of Trigo-Tangram. In many cases, you get funny results. Figure 4 shows how a Tangram bird stretches its neck and how a Tangram cat turns into the Pink Panther.

TANGRAM CAT HAPPILY TURNS INTO PINK PANTHER

101

Figure 3. Some problems for the Trigo-Tangram puzzle.

Transforming two-dimensional puzzles provides endless fascination. Just look what happens if you rotate the Tangram in Figure 1 by ¼ Æ and then perform the transformation. The square and parallelogram of the original are transformed into congruent rhombi, which makes the derivative less convenient for a puzzle. You obtain better results by applying linear extensions diagonally (Figure 5). Since in this case the congruence between the two large and two small Tangram triangles is preserved, the two puzzles

Figure 4. A Tangram bird stretches its neck, and a Tangram cat turns into the Pink Panther.

102

B. WIEZORKE

Figure 5. Transforming the Tangram by diagonal extensions leads to two derivatives which are more Tangram-like than the one shown in Figure 1.

generated are even more Tangram-like than the Trigo-Tangram. To create problems for these derivatives is left to the reader. One set of problems will serve for both of them. I hope I have stimulated your imagination about what can be done by transforming puzzles so you may start your own work. Take your favorite puzzle, and have fun generating individual derivatives for yourself and your friends!

She wrote back with profuse thanks for such an “elegant” solution. Frankly. She had one large stone that was perfectly square. His wife Polly is an avid gardener. In order to meet Polly’s required dimensions. and her sarcasm rather put me on guard. She presented me with the following problem. 103 . I was surprised that Polly would bother to write about such a trivial dissections problem. She asked me if it were possible to cut the stone into four pieces that could be arranged to form a somewhat larger square perimeter enclosing a square hole having sides one quarter those of the original stone. She and Paul have a bent for geometrical recreations. The simple solution to this problem which I then sent to her. You may remember meeting Paul in The Puzzling World of Polyhedral Dissections. so they are always on the lookout for creative and original solutions to their various landscaping projects.Polly’s Flagstones Stewart Cofﬁn I wish to report on some recent correspondence with my good friends the Gahns in Calcutta. but I should have known better. Polly places precisely ﬁtted ﬂagstones around various plantings in her garden in order to suppress weeds. the sides are divided in the ratio of 3 to 5. is shown below. a classic dissection of the square.

I came up with the scheme shown below. but after tinkering for a while with paper. which could then be rearranged to create a square hole within a square enclosure or any one of three different rectangular holes within rectangular enclosures. how about a scheme whereby the stones could be rearranged to form any one of three different-sized rectangular openings. which she proposed that we cut into four pieces. The smaller and larger holes have dimensions that are irrational. to allow for more ﬂexibility in landscaping. It was then I realized that from the start she . One of these rectangular holes will have the same shape as the original rectangle and will be surrounded by a rectangle also having that same shape. She decided that she preferred mostly rectangular rather than square planting spaces. each one enclosed by a rectangular perimeter. pencil. but this time to form a rectangular perimeter enclosing a rectangular planting space.104 S. If you think that Polly was satisﬁed with this solution. cut into four pieces. Suppose we were to consider starting with a large rectangular stone. This required a bit more reﬂection than the ﬁrst problem. then you don’t know Polly very well. Almost any rectangle dissected symmetrically by two mutually perpendicular lines that touch all four sides will produce four quadrilaterals that can be rearranged to create the required three different rectangular holes enclosed by rectangles. And for good measure. the medium-sized rectangular hole will have dimensions exactly one quarter those of the original stone. COFFIN Now she posed another problem. She immediately wrote back and suggested that it was a shame to cut up two beautiful large stones when one might do. and scissors. With the dimensions shown.

Don’t ask my why this works. Locate the midpoint E of side BC. Draw EF. by return mail arrived her solution. a solid trapezoid (two solutions). so I wrote back and asked what made her so sure it was even possible. The actual arrangements of the pieces for the four different solutions. the square hole will be one foot square. but it does. Bisect EG to locate H. For added recreation. except that one of the trapezoids becomes a triangle. are left for the reader to discover. making one of the pieces triangular. let the original square stone ABCD be four feet on a side. Sure enough. and draw JHK perpendicular to EHF (most easily done by swinging arcs from E and G ). but for this example let it be one foot from A.POLLY’S FLAGSTONES 105 had just been setting me up. shown below. Now what sort of scheme do you suppose Polly will come up with for that other stone. An interesting variation is to let points A and F coincide. and use that distance from F to locate point G. the rectangular one? . and a different trapezoid (one solution). The location of point F is arbitrary. note that the pieces can also be arranged to form a solid parallelogram (two solutions). For the purpose of this example. With these dimensions. Subtract length CE from length DF. one square and three rectangular. I decided to play into it. The same solutions are possible. One of the rectangles has a pair of solutions — the others are unique.

The subscription expired after one year. Clarke. Thwarted from including pentominoes in the movie 2001: A Space Odyssey. and I was not able to renew it. Figure 1. in his mostly autobiographical Ascent to Orbit. based on Solomon Golomb’s presentation in 1953 to the Harvard Mathematics Club. Arthur C. That column was inspired by a 1954 article in the American Mathematical Monthly. In late 1976. as the game HAL and Bowman play (the ﬁlm shows them playing chess). created an ever-widening ripple effect. Later editions of the book actually show a pentomino rectangle on the ﬂyleaf. It expired one month before the May 1957 issue. a group of expatriates stationed in Iran took a weekend trip to Dubai. declares himself a “pentomino addict”. focusing especially on the pentominoes. In 1956 I received a gift subscription to Scientiﬁc American as an award for excellence in high school sciences. so I did not get to see the historic column introducing pentominoes (Figure 1) to a world audience.Those Peripatetic Pentominoes Kate Jones This recital is yet another case history of how the work of one man—Martin Gardner— has changed the course and pattern of a life. Golomb’s naming of the “polyomino” family of shapes and their popularization through Martin Gardner’s beloved column. a paperback 107 . As we loitered around the airport newsstand. crediting Martin Gardner’s column as the source. Clarke wrote pentominoes into his next science ﬁction book. My favorite part of the magazine was the “Mathematical Games” column by Martin Gardner. Imperial Earth.

Exploration of the pentominoes’ vast repertoire of tricks was a ﬁne way to spend expatriate time. and inevitably a dominotype game idea presented itself to me. Well. and soon friends got involved. . A longtime Clarke fan. to be shared with friends. Super Quintillions. I bought the book and soon caught pentomino fever. having sold my graphics business and with no career plans for the future. JONES rack with a copy of Imperial Earth caught my eye. Playing ﬁrst with paper and then with cardboard was not enough. when the Iranian revolution precipitated the rapid evacuation of Americans. By fall of 1979 a wooden prototype was in hand. and after some preliminary doubts we went for it. and the pieces turned out to be thick enough to be “solid” pentominoes. I commissioned a local craftsman to make a set of pentominoes in inlaid ivory. but idleness was not my style. Fast forward to December 1978.108 K. My husband’s job took care of our needs. A friend’s suggestion that we “make and sell” that game I had invented popped up just then. It was a thrill to ﬁnd mention of Martin Gardner in the back of the book. a specialty of the city of Shiraz. This magniﬁcent set invited play. of course that was the way to go! A little research turned up the fact that “pentominoes” was Figure 2. Back home I was in limbo.

and the product names are trademarks of Kadon. Inc. keeping the checkerboard pattern (Figure 3). These alone or combined with the 12 Quintillions blocks can form double and triplicate models of some or all of the 29 pentacube shapes (see Figure 2).½ Super Quintillions: 17 non-planar pentacubes (plus one duplicate piece to help ﬁll the box). these are products made by Kadon Enterprises. Here (chronologically) are the many guises and offspring of the dozen shapes that entered the culture through the doorway Martin Gardner opened in 1957. ½All . We had much to learn about marketing. And so Quintillions begat a large number of kindred puzzle sets. The puzzle challenge: Change any one into another in the minimum number of chess knight’s moves. The most important lesson was that one needed a product “line”. Figure 3. checkerboarded format.. joy and reverence that we sent one of the ﬁrst sets to the inspiration of our enterprise. not just a single product. Martin Gardner. Leap: a ¢ grid whereupon polyomino shapes are plotted with black and white checkers pieces in a double-size.THOSE PERIPATETIC PENTOMINOES 109 a registered trademark of Solomon Golomb. so we’d have to think of another name. and most of them sneak in some form of pentomino or polyomino entity among their other activities. Fives quints quintillions! It was with great pride. And it was encouraging to us neophyte entrepreneurs when Games magazine reviewed Quintillions and included it on the “Games 100” list in 1980.

or diagonal. What is the minimum number of moves to exchange the congruent groups of black and white knights? Figure 4. Colormaze: Square tiles in six colors form double polyomino shapes with no duplicate color in any row. The pieces can form duplicate and triplicate models of themselves. . Figure 5. In another use.110 K. formed with checkers (Figure 4). with checkerboarded arrangements (Figure 5). column. JONES Void: a ¢ grid on which a “switching of the knights” puzzle is applied to pairs of polyomino shapes from domino to heptomino in size. Quintachex: the pentominoes plus a ¾ ¢ ¾ square checkerboarded on both sides (different on the two sides). double polyominoes can be formed with ¾ ¢ ¾ quadrants of each color and then dispersed through a maze-like sequence of moves to the desired ﬁnal position (Figure 6).

. Figure 7.THOSE PERIPATETIC PENTOMINOES 111 Figure 6. Poly-5: two-dimensional polyominoes of orders 1 to 5. Sizecompatible with Poly-5. Sextillions: the hexomino shapes can form double through sextuple copies of themselves and various enlargements of the smaller polyominoes.and double-size polyomino shapes (Figure 8). Snowﬂake Super Square: the 24 tiles are all the permutations of three contours—straight. They can form various single. convex. in a four-way symmetrical tray of 89 unit squares (Figure 7). concave—on the four sides of a square.

The pieces can form diagonally doubled pentominoes and hexominoes with colormatching adjacency of tiles (Figure 9). Multimatch I: The classic MacMahon Three-Color Squares have been discovered to form color patches of pentominoes and larger and smaller polyominoes within the ¢ rectangle and ¢ square. Lemma: A matrix of multiple grids lends itself to polyomino packing with checkers in three colors. JONES Figure 8. including partitions into multiple shapes (Figure 10).112 K. Triangoes: orders 1 and 2 polytans permuted with two or more colors (square. triangle. Figure 9. parallelogram). .

Gallop: On a ¢ ½¾ grid. double checkerboard hexomino shapes deﬁned by pawns are transformed with knight’s moves (borrowed from the Leap set). Figure 12. Another puzzle is to change a simple hexomino made of six pawns into as many different other hexominoes as possible by moving the pawns with chinese-checker jumps (Figure 12). Multimatch II: The 24 tri-color squares with vertex coloring (each tile is a ¾ ¢ ¾ of smaller squares) can form polyominoes both as shapes with colormatching and as color patches on top of the tiles (Figure 11).THOSE PERIPATETIC PENTOMINOES 113 Figure 10. . Figure 11.

Figure 14. Rhombiominoes: A skewed embodiment of pentominoes. Quantum: Pentomino and larger polyomino shapes are created by movement of pawns (the puzzles to be published as a supplement to the existing game rules). The 20 distinct pieces form a ½¼ ¢ ½¼ rhombus (Figure 14). JONES Figure 13. Throw a Fit: Mulit-colored cubes form pairs of pentominoes with colormatching. Perplexing Pyramid: A Len Gordon invention. The T and U puzzles are actually dissected pentominoes (Figure 13). . Tiny Tans: Four triangle-based pieces can make some pentomino shapes. the six pieces comprising 20 balls can make most of the pentominoes in double size.114 K. This is a limited-edition set. where each square is a rhombus the size of two joined equilateral triangles.

humor and special themes of these two games set them apart from all others. using Martin’s game rules.THOSE PERIPATETIC PENTOMINOES 115 Heptominoes: These sets of the 108 seven-celled polyominoes are produced by popular demand. and we are dedicated to their care and continuance. In retrospect. Octominoes: The 369 eight-cell shapes exist in limited-edition sets. sized to Sextillions and Poly-5. A vast territory just begging for exploration. The charm. polytans. the inspiring and nurturing presence of Martin Gardner has been there. Game of Solomon. I cannot imagine any other career in which I would have found greater satisfaction. The Hexacube: The 166 planar and non-planar hexacube pieces plus 4 unit cubes form a ½¼ ¢ ½¼ ¢ ½¼ cube. this set is sized to Quintillions. polyiamonds. All these are ﬁnding homes in the Kadon menagerie. then. fulﬁllment and never-ending challenge than in the creation of Figure 15. sized to smaller members of the family. and always helpful information. my personal career and unusual niche in life came about directly as a result of the intellectual currents and eddies created by one mind—Martin Gardner’s—whose ﬂow I was only too happy to follow. besides polyominoes: polyhexes. . with kind and encouraging words in correspondence. in the reprints of his columns. through articles in various publications. A natural side-effect of his raising the consciousness of the public about the joys of combinatorial sets has been the proliferation of other sets. We were honored that Martin selected us to produce his Game of Solomon and the Lewis Carroll Chess Wordgame. Throughout the years of Kadon’s evolution. The above concepts arose and became possible because Martin Gardner fostered a love of mathematical recreations among his readers.

Martin Gardner.116 K. if there is a 90-degree angle or parallelogram to be found. Thank you. . JONES Figure 16. the pentominoes will announce themselves in yet another manifestation. And assuredly. Lewis Carroll Chess Wordgame. beautiful playthings for the mind.

**Self-Designing Tetraﬂexagons
**

Robert E. Neale

Flexagons are paper structures that can be manipulated to bring different surfaces into view. The four-sided ones discussed here are “ﬂexed” by folding them in half along an axis, and then opening them up a different way to reveal a new face. Sometimes the opening is in a different direction (mountain or valley fold), sometimes along the other axis, and sometimes both. Although usually constructed from a strip of paper with attachment of the ends, ﬂexagons seem more elegant when no glue or paste is required. Interesting designs of the faces can be found when the material is paper, or cardboard, with a different color on each side (as in standard origami paper). I call these “self-designing.” What follows is the result of my explorations. Although the text refers only to a square, the base for construction can be any rectangle. There are ﬁve different starting bases (paper shapes; see Figures 1–5). There are three different ways of folding the bases. Warning: Not all the results are interesting. But some are, my favorites being three: the ﬁrst version of the cross plus-slit (for its puzzle difﬁculty); the ﬁrst version of the square slash-slit (for the minimalist base and puzzle subtlety); and the third version of the square cross-slit (for its self-design). Note that the text assumes you are using two-colored paper, referred to, for simplicity, as black and white.

Figure 1. Square window.

Figure 2. Cross plus-slit.

117

118

R. E. NEALE

Figure 3. Square plus-slit.

Figure 4. Square slash-slit.

Figure 5. Square cross-slit.

Square Window

The ﬁrst ﬂexagon I saw that did not need glue was shown to me by Giuseppi Baggi years ago. It is a hexa-tetraﬂexagon made from a “window” — a square (of 16 squares) with the center (of 4 squares) removed (Figure 1). It has six faces that are easy to ﬁnd. ½ I have recently used this window model for a routine about the riddle of the chicken and egg: “What’s Missing Is What Comes First.” This involves decorating the paper with markers, so is a separate manuscript. It should be noted, however, that there are two ways of constructing this ﬂexagon, both of which are discussed below. Same Way Fold. As the dotted line indicates in Figure 6, valley fold the left edge to the center. The result is shown in Figure 7. As the dotted line indicates in Figure 7, valley fold the upper edge to the center. The result is

**½Directions for constructing the hexa-tetraﬂexagon are found on pages 18–19
**

of Paul Jackson’s Flexagons, B.O.S. Booklet No. 11, England, 1978.

SELF-DESIGNING TETRAFLEXAGONS

119

shown in Figure 8. As the dotted line indicates in Figure 8, valley fold the right edge to the center. The result is shown in Figure 9.

Figure 6.

Figure 7.

Figure 8.

Figure 9.

The ﬁnal fold is very tricky to communicate, but not at all tricky to do when you understand it. The goal is to make this last corner exactly the same as the other three. As you would expect, the bottom edge will be valley folded to the center. The right half of this lower portion falls directly on top of the portion above it. The left half of this lower portion goes underneath the portion above it. (So Figure 9 shows a valley fold on the right half and a mountain fold on the left half.) To make this actually happen, lift up the upper layer only of the left half of the lower portion, and then fold the entire lower portion to the center. The result is shown in Figure 10.

120

R. E. NEALE

Figure 10.

Note that the model is entirely symmetrical. The four corners are identical, and both sides of the model are identical. (If this is not the case, you have made a mistake, probably on the last move.) This method amounts to folding the edges to the center, one after another (proceeding either clockwise or counterclockwise around the square), all of the folds being valley folds. This produces a Continuous Flex, moving in a straightforward manner from face 1 to 2 to 3 to 4 to 1, 5, 6, 1. The designs procured are merely three black faces and three white faces, as follows: black, black, white, white, black, black, white, white, black. Alternating Way Fold. As Figure 11 indicates, mountain fold the left edge to the center. The result is shown in Figure 12. As the dotted line indicates in Figure 12, valley fold the upper edge to the center. The result is shown in Figure 13. As the dotted line indicates in Figure 13, mountain fold the right edge to the center. The result is shown in Figure 14.

Figure 11.

Figure 12.

SELF-DESIGNING TETRAFLEXAGONS

121

Figure 13.

Figure 14.

Before you make the ﬁnal valley fold of the bottom edge to the center, pull the bottom layer at the lower left corner away from the upper layer. Make the valley fold, and allow that bottom layer to go back to the bottom of the lower left corner. The result is shown in Figure 15. Note that opposite corners are identical, and both sides of the model are identical. This noncontinuous Puzzle Flex can be ﬂexed continuously through four faces only: 1 to 2 to 3 to 4 to 1, etc. Finding the other two faces is easy in this particular case, but backtracking is required. The four faces are identical: a checkerboard pattern of two black squares and two white squares. The other two faces are a solid black face and a solid white face.

Figure 15.

Note the two arrows in Figure 16. Orient them just as indicated. they are turned over so you see the 5 cells. This puzzle is interesting for two reasons: It is tricky to construct. When you pull them down and away from each other. exactly as indicated in Figure 16. The other cells come together to form two boxes open at the top. and two slits in the center that intersect in the shape of a plus sign (see Figure 2). one corner in each hand. very quick. One result was a puzzle made from a “cross” — a square (of 16 squares) with the four corner squares removed. There is a 6 cell at the bottom of each box (see Figure 16. NEALE Cross Plus-Slit The window model inspired me to think of other ways to form a ﬂexagon without having to attach the ends to each other. You will make two. the other two faces are quite difﬁcult to discover. The other two faces are solid black and solid white. while four of the faces are easy to ﬁnd. and the second ﬂattens it again. being black and white checkerboard patterns. E. involving changing the ﬂat ﬂexagon into a three-dimensional ring and back again. The four faces are continuous and identical. They indicate that you are to make the base 3-D by pushing two opposite corners down and away from you. The ﬁrst renders the base 3-D. moves. the 3-D Tricky Way. . So reach underneath and hold the two free corners of the 1 cells. (Note: The trick fold for constructing the ﬂexagon can be done in a slightly different way that has the four faces not continuous.) 3-D Tricky Fold In order to follow the instructions and understand the diagrams. This base is manipulated in a third way. and. The numbers in parentheses are on the back side of the base.122 R. number the base from one to six.

Now open the model into a tube — a cube open at both ends. and then make the tube move again to return to face 4 ¾Other directions for constructing this ﬂexagon are on page 27 of Jackson’s Flexagons. The model is completed.) Now push down on these corners. move the lower inside packet of squares (with 2 and 3 on the outside) to the left. the left and right halves going back away from you. Once this happens. the base forming a ﬂat square of four cells.SELF-DESIGNING TETRAFLEXAGONS 123 Figure 17. Freeman & Co. You can ﬁnd the faces numbered 1 to 4 by the usual ﬂexing. then 3. and at the same time. or a 3 instead of a 4. Figure 17). Check both sides. Change your grip so you are holding the two inside corners of the 5 cells. Life and Other Mathematical Amusements. pp. and four 4 cells on the other side.. Begin with face 1 on the top and face 4 on the bottom. Now the base should be ﬂattened into a compact model with four cells showing on a side.H. and the upper inside packet of squares (with 2 and 3 on the outside also) to the right. 1983. however. The result is shown in Figure 18. Do not open the model as you do when ﬂexing. and in Martin Gardner’s Wheels.¾ Flexing. 64–68. Mountain fold the model in half on the vertical axis. pull them away from each other. Faces 5 and 6 are found by the following procedure. (These are the corners opposite the ones you were holding. a 2 instead of a 1. You can correct this easily by tucking the wrong cell out of sight. Collapse the tube in the opposite way. Sometimes. you will ﬁnd that another number has appeared. Rather. replacing it with the proper one. you should have four 1 cells facing you. Flex it in the usual way to show face 5. New York: W. The 3-D boxes will collapse. This creates a new arrangement. .

but they are not continuous. For best results. More recently. with a single slit. The designs of the four faces are two solid black and two checkerboard. . use the Alternating Way fold. and repeat the moves just given. solid white. solid black. To ﬁnd face 6. and then face 1. (Alternating Way gets you nowhere. Square Slash-Slit (Repeat “slash-slit” out loud quickly at your peril. Indeed. Two faces can be found only by forming the rings.) Four faces are easily found. I discovered a way to make a ﬂexagon without removing any portion of the square. begin on face 4. I have worked out some variations on the theme of removing no paper. The other two faces are solid white. as mentioned above. Another version of this ﬂexagon can be made by using the Same Way Fold. These follow. Square Plus-Slit About a year or so ago. Flexing is continuous. but they are not continuous. (The goal was inspired by my desire to make a ﬂexagon from paper money with minimal damage.) It is pleasing that a ﬂexagon can be constructed from a square. The Same Way Fold can be used also. and four faces can be shown: checkerboard. checkerboard. or rectangle. All faces are of the identical solid color. NEALE Figure 18.124 R. with face 1 at the back. but with poor results. Using the 3-D Tricky Fold gives the same results as the Alternating Way Fold. E. both the third and fourth are derived from the ﬁrst face. Four faces can be shown.) It is made from a square with two slits intersecting the center in the shape of a plus sign (Figure 3).

etc. The faces are these: solid white. The designs are these: checkerboard. leaving a white square in the center. two white triangles arranged as in face 1. solid black. Use of paper money is recommended. two white triangles. in the following manner: face 1. 3. The result ﬂexes exactly the same as the one above. Here are four of them. Square X-Slit These models are made from squares with two intersecting slits making an X (Figure 5). but with different designs: eight triangles of alternating color. but pull gently as you do it. do the folding. 4. Being sure you have the right axis on which to fold the model in half. The obvious way is to open up each side of the slit. Then ﬂexing is possible. 2. solid white. 1. This does not appear to offer as much variation due to the repetition involved. Or. solid black. ﬂexing is not continuous. However. circling about the center. the second allowing some presentational by-play. However. All have six faces. with the inner ﬂaps arranged differently: two opposite ﬂaps creased one way. triangle in each corner. the move can be done quite secretly in the course of doing the ﬁrst ﬂex. two white triangles opposite each other. solid black. arranged differently from the previous two. The Same Way Fold can also be used. once you understand what is required. ﬂexing and unfolding. 5. the other two creased the other way. Use of the Same Way Fold shows four faces continuously. and the little ﬂaps will be adjusted automatically and properly. arranged as in face 2. a second solid white. 2. and best of all. There are two versions. two black triangles in a different arrangement. 4.SELF-DESIGNING TETRAFLEXAGONS 125 It must be a slash along the diagonal. then challenging the knowledgeable paperfolder to duplicate the action. you can create some mischief by folding. and the other two easily. the ﬁrst face takes on a quite . checkerboard. Obviously. 6. two black triangles opposite each other. however (Figure 4). to be seen only at the very end of the ﬂexing. Using the Alternating Way Fold. ﬂexing is not possible without an additional move. Many versions of these are possible. you can tuck the ﬂaps properly as you make the valley folds. The faces are as follows: two black triangles opposite each other. solid black. and then move the two little ﬂaps inside the slit to a position outside it. and two of the sides are just a little tricky to ﬁnd. 1. same as the second face. two black (and six white) triangles opposite each other. The ﬂexing is continuous. Using the Same Way Fold. The model will show four faces. solid white.

For example. as the eight triangles are rearranged in an interesting way — four triangles of alternating color circle about the center. And so on. but written more than a year or two before that. triangles in each corner so the center forms a white square. The designs on one pair of faces are different. with one triangle in each corner of the model. The second and third showings of the face reorient parts of it. but differing in both color and arrangement. and use the Same Way Fold. Four faces are shown continually. although no movement to a ring is required. two triangles on each face. and the other two are deceptively hard to ﬁnd.) The Alternating Way Fold can be used with the inner ﬂaps arranged differently. One of the items was taught at the Gathering and appeared in the Martin Gardner Collection. checkerboard of four squares. when a half of a corner is removed. these reveal seven different designs: solid color. one triangle (out of eight). four large triangles alternating in color. instead of all or none of it. For another example. Use of the Alternating Way Fold creates a puzzle version. Additionally. NEALE different appearance when ﬂexed to the back of the model. solid black. E. checkerboard square. (This is my favorite of these four because the designs are handsome and the two solid color faces are a little tricky to ﬁnd. the methods given above can be combined to create still different design possibilities. This manuscript was submitted for possible publication after the Gathering for Gardner in January of 1993. This uses the fact that all ﬂexagons show some faces more than once. the design changes.126 R. four faces each with different arrangements of three triangles. solid white. I constructed one to reveal the maximum of variety in design. . So although my multiversion shows only six faces. combine a half-cross base with a slash-slit. Possibilities Other ﬂexagons can be constructed from the X-slit approach. also. The designs are these: checkerboard square. For example.

103 . so I will not waste space on them here. which almost by deﬁniton do not require close tolerances and are just about the easiest to fabricate. The three rather unexciting puzzles that came out of this (Lamplighter. and Bottleneck) are described in the 1985 edition of Puzzle Craft. This led to a couple of not very original or interesting string-and-bead puzzles (the Sleeper-Stoppers). with wire passing through the loops in various ways. I may have been inspired in this by a neat little puzzle known as Loony Loop that was enjoying commercial success at the time. thus: Then bend the wire into some animated shape. What I lacked from the start was a simple interesting puzzle that could be produced and sold at low cost to curious children (or their parents) who could not afford the fancy prices of those AP-ART sculptures. Take a foot or two of ﬂexible wire and form a loop at both ends. To ﬁll this need. I turned my attention to topological puzzles. except to describe their common design feature. Liberty Bell. hand-crafting interlocking puzzles in fancy woods and selling them at craft shows as fast as I could turn them out.The Odyssey of the Figure Eight Puzzle Stewart Cofﬁn I became established in the puzzle business in 1971. I also tried to come up with something in string and wire that could be licensed for manufacture. such as in the very simple example below.

I had continued to . somewhat to my surprise. none of these ideas enjoyed any success. knotted a cord around it. which then left them utterly bafﬂed as to ﬁnding it. I received a seven-page document from someone in Japan. full of diagrams and such. The puzzle editor made the surprising observation that it was topologically equivalent to the Double-Treble-Clef Puzzle made by Pentangle and therefore solvable. But a careful check showed that the two were not quite equivalent. however. Royce Lowe of Juneau. and I did not spend a lot of time trying to digest it.) The story does not end there. Alas. Some readers misread that purposely vague write-up and assumed that it must have a solution. Next it appeared in a 1976 issue of a British magazine on puzzles and games. Alaska. I soon became convinced that this was impossible. When some of his customers started asking for the solution. decided to add the Figure Eight to the line of puzzles that he made and sold in his spare time. Over the intervening years. they reported that topological puzzles in general were their least favorite of any puzzles I had produced. I was at a loss for any sort of formal proof. To add to the confusion. I published this simply as a curiosity in a 1974 newsletter (later reprinted in Puzzle Craft). during an idle moment one day I formed some wire into the ﬁgure-eight shape shown below. Unlike the simple example shown above. the puzzle appeared in Creative Puzzles of the World by van Delft and Botermans (1978) with a farce of a convoluted “solution” thrown in for added amusement. purporting to prove that the puzzle was unsolvable. The proof appeared to be a rather complicated. knot a sufﬁciently long loop of cord around some part of the wire and try to remove it (or try to put it back on after it has been removed). some of these puzzles can be quite bafﬂing to solve. he begged me for help. since the Pentangle puzzle was. COFFIN Finally. and tried to remove it. (In a recent survey of my puzzle customers. For a slight variation of this.104 S. but being a novice in the ﬁeld of topology. More recently.

THE ODYSSEY OF THE FIGURE EIGHT PUZZLE 105 received numerous requests for the solution. it is often the simplest things that prove to have the greatest appeal. Or will it mischievously rise again. Whoever would have guessed that this little bent piece of scrap wire and loop of string would launch itself on an odyssey that would carry it around the world? I wonder if this will be the ﬁnal chapter in the life of the infamous Figure Eight Puzzle. and by that time I was rather tired of the whole thing. probably not even realized at the start. perhaps disguised in another form. When it comes to puzzles. as topological puzzles so often do? .

puzzles entertain. those who like puzzles generally like to solve just about any problem. The majority of mechanical puzzle solvers take the puzzle apart through a series of random moves with no thought given to the fact that this way they have only half-solved the problem. Unlike magic or illusions with misdirection and hidden mechanisms. magic. About twenty years after my introduction to those ﬁrst little wire teasers. My own interest in puzzles began in early childhood. In fact. the search for answers drives many of us on.Metagrobolizers of Wire Rick Irby Many people love the challenge of solving a good puzzle. mystify and educate. Regardless of one’s ability to solve them. all parts exposed ready for minute examination and scrutiny. with everything visible. Be it a paradox. the solutions can elude even the sharpest and quickest minds of every discipline. The random-move method will sufﬁce for easy to medium puzzles but will do little or no good for solving the more difﬁcult ones. with the small packaged and manufactured wire puzzles available at the local 5 & 10-cents store. In spite of this.” on the other hand. Wire disentanglement puzzles are topological in nature and can vary widely in both difﬁculty and complexity of design. and the search for puzzling challenges will undoubtedly continue. Puzzles can go beyond an understanding of the problem and its solution. or a puzzle. a mathematical problem. a back injury from an auto accident and lots of encouragement from puzzle collectors brought the following and many other puzzle ideas to fruition. mechanical puzzles are an open book. will analyze the problem with logic and stratagem. A “puzzle crazy. Wire lends itself very easily to topological problems because of its inherent nature to be readily 131 . then reason out the solution to include returning the parts to their original starting position. and here is where the separation between the common puzzler and (to borrow a phrase from Nob Yoshigahara) a “certiﬁable puzzle crazy” lies. they were never quite enough of a challenge to satisfy my hunger. Somewhere in the back of my mind I knew I could come up with better puzzles than were currently and commercially available. Although entertaining.

Unfortunately. The Bermuda Triangle Puzzle . for asking me to participate in “Puzzles: Beyond the Borders of the Mind. It is my suspicion that the subconscious mind is constantly at work attempting to ﬁt pieces of countless puzzles together.” and for presenting me with the opportunity to meet Martin Gardner. either you become good at it or you are devoting your mind to the wrong thing. I neither know or understand the process of any creativity than to say that it just happens. Many thanks must go to Martin Gardner as an inspiration to the millions enlightened by his myriad works. Often it is easier to come up with and develop a new puzzle idea than to give it a good and catchy name.132 R. The Bermuda Triangle Puzzle Knowing the fascination that many people have with the somewhat mysterious and as-yet-unexplained disappearances of various airplanes and ships in the area known as the Bermuda Triangle and the Devil’s Triangle made naming this puzzle relatively easy. IRBY and easily formed into whatever permanent shapes may be necessary to present a concept. This puzzle idea came to me as I was driving to San Francisco to sell my puzzles at Fisherman’s Wharf. sometimes it succeeds! If you devote your mind to something. A couple of examples of ideas that have “popped” out of my mind at various times are explained and illustrated below. I really can’t answer. Thanks also to Tom Rodgers for his support of my work. in 1971 or 1972. I am frequently asked to explain the thought process involved in coming up with a new puzzle.

taking the UFO with it as it moves. although moderately difﬁcult for the average puzzler to solve initially. the UFO being a ring with an abstract shape mated to it. this one is relatively easy to remember once the solution has been seen. as I sat playing with it. The Nightmare . The solutions to many puzzles can be elusive until the puzzle has been manipulated many times. We didn’t have any nightmares about it. this one came to me fully formed and complete.METAGROBOLIZERS OF WIRE 133 The object of the Bermuda Triangle is to save the UFO that is trapped in the puzzle. The Nightmare Puzzle The Nightmare Puzzle The Nightmare puzzle was conceived as Johnny Carson was delivering his monologue during the Tonight Show one night in 1984. There are several places where the UFO may be separated from the triangle but only one place where the separation will allow the solution to be executed. I keep tools and wire handy for just such events having learned that three-dimensional ideas are difﬁcult to decipher and duplicate from two-dimensional scribbling on a piece of paper. As with most of my puzzle ideas. The Bermuda Triangle rates about a medium level of difﬁculty. The puzzle is generally set up with the ring around the Bermuda Triangle. my wife joined me in critiquing my latest design. The triangle can be moved over the entirety of the larger conﬁguration. which is a triangular piece. Sometime later one of us stated that we probably would have nightmares about it that night. After making the prototype. but the name stuck. Most of the large conﬁguration to which the triangle is attached is there simply to bewilder the would-be solver.

after the cord is removed. I rate the Nightmare an 8. A cord encircles the two inner loops of the puzzle. In addition to the difﬁculty in conceptualizing the convoluted shape of the puzzle. the ﬂexibility of the cord allows one to make mistakes not possible with rigid pieces. and the object is to remove the cord completely from the puzzle. quickly compound into a tangled mass of knots soon precluding any progression toward the solution. with the wire ends making small rings that are wrapped around the wire in such a manner as to eliminate any usable ends. There are two outer and two inner loops. The difﬁculty may be increased by. On a scale of 1 to 10. The Nightmare puzzle is made from one continuous strand of wire. Any wrong moves. adding a ring to the cord that will not pass through either of the small end rings then attempting to replace the cord. IRBY has more than lived up to its name with a convoluted three-dimensional shape that exacts extreme effort and concentration from all who attempt it.134 R. not promptly corrected. .

I was going to write a column on “Games” so I made a pilgrimage to Hastings-on-Hudson to visit the Master. when the answer came. the hourglass remains at the bottom of the cylinder until a certain quantity of sand has ﬂowed into its lower compartment. and then as an editor of Psychology Today. ﬁlled with water. and corresponded with him occasionally from 1963 on. just like Martin said it would. The next month. I was delighted. so absurdly obvious. I read Martin Gardner’s columns religiously. It was so simple. Can you guess the simple modus operandi ? I gave the problem some thought but couldn’t come up with any good theory. If the cylinder is inverted a curious thing happens. There on a shelf was the infamous Floating Hourglass itself. I could ﬁnally try out the curious toy I had read about so many years before. The effect was like seeing a good magic trick or hearing a good joke. was to be launched. Omni. It seems impossible that a transfer of sand from top to bottom of the hourglass would have any effect on its overall buoyancy.Beautiful But Wrong: The Floating Hourglass Puzzle Scot Morris The Beginning One of the problems in Martin Gardner’s August 1966 “Mathematical Games” column was The Floating Hourglass. An unusual toy is on sale at a Paris shop: a glass cylinder. as a graduate student. I assumed it had to do with some law that I had forgotten since high school. I turned it over and the hourglass stayed at the bottom. In 1978. I should have. I had the pleasure of ﬁnally meeting the Master Explainer. that I not only could have thought of it myself. as a college student. 135 . and at the top an hourglass ﬂoats. in the months before a new science magazine. Then it rises slowly to the top.

had visited Martin in early 1966. Tim Rowett had brought one from England. He didn’t bring one back.136 S. I saw my second Floating Hourglass and knew I could ﬁnally write about it. For years I have itched to tell this story. When Martin allowed me to look through his ﬁles. but no magazine article could possibly contain it. but his description was clear enough. a treasure trove of letters and drawings. made by Ray Bathke of London. My September 1992 column introduced the Floating Hourglass 25 years after I ﬁrst heard about it. the Danish sculptor/inventor (the Soma Cube. I immediately ordered some. MORRIS At the 1991 Puzzle Collectors Party in Los Angeles. The First Theories Piet Hein. I found a thick folder on the Hourglass. This book ﬁnally gives me the chance to tell the history of the Floating Hourglass Puzzle. Gardner knew how it must work and wrote about it in his August and September columns without even having . the Super Egg) and poet/artist (Grooks). The results appeared in the January 1993 issue. I asked my clever readers to submit theories to explain it. since I only published puzzles in Omni that I knew my readers could ﬁnd. Hein told him about a toy he had seen in the Paris airport.

He drew a cartoon of himself on an elevator with Einstein. For him this meant that the ﬁnal proof was not yet in. pondering an oversized hourglass.” he wrote “and all that was needed to make my intellectual headache come back much worse than . There it stands at the bottom and changes its weight!.” but he also knew that Martin had never examined an actual hourglass sample. just enough to see that it worked. the glasses stay in place at ﬁrst and then one rises while the other sinks.What keeps the hourglass down is the falling of the sand. don’t weigh the package while they’re running. but Pien wasn’t convinced. the best criterion for a theory is its beauty.. an ‘effective change in mass’: “The hourglass is heavier while the sand is falling. He knew of Gardner’s “answer. He knew immediately that his impact theory was doomed. Hein tried the double-glass in the shop a few times. Since the glasses change their weight whenever the sand falls inside. Hein issued a mock warning: When mailing hourglasses. 1966.THE FLOATING HOURGLASS PUZZLE 137 seen one. He felt there must be something more. break it open or learn how it is made. but because the rate of falling sand has decreased. “That was all I wanted to know. In a shop there he saw a double-glass: two cylinders side by side. And Martin had already admitted that Piet’s impact theory was beautiful. Hein was obsessed with hourglasses. He designed an hourglass-powered perpetual motion machine and created a fantasy ocean full of bobbing hourglasses. The hourglass rises not because there is little sand left in the top chamber. He thought the inside of the cylinder he had seen was too smooth to offer much resistance. A Painful Paradigm Shift Just a couple of days after writing the letter and drawing the cartoon. a ﬂoating hourglass in one and a sunken hourglass in the other. Martin’s theory relied on friction between the glass and the cylinder. In the absence of knowing the truth. Until you could see and touch one. or you’ll have to pay a higher postage. “Imagine if the hourglass were opaque and you didn’t know it were an hourglass at all. A “sunken” hourglass that stays in place at the top of a tube can’t be explained by sand grains falling in the opposite direction. Hein believed the impact of the sand grains hitting the bottom of the glass exerts a downward force. This seems to solve the whole problem. Italy.. When turned over. not the amount of sand that has fallen or is left.” Hein wrote Gardner on September 16. Hein had an agonizing experience in Milan.” Note that he wrote all this after the September Scientiﬁc American was out. something to do with the falling sand. all theories were valid contenders.

like a small ﬁsh on the center of a huge platter. The hourglass puzzle is caused by an hourglass effect. the heavier liquid is on top pushing down. and he found it painful. not in the falling of the sand but in the ﬂowing of the liquid.” The Solution Is in the Solution The key to the puzzle isn’t in the glass but around it. He knew that the sinking glass directly refuted his theory. Hein realized. The falling sand is just for misdirection. a tabula rasa. When the cylinder is inverted. There must be two liquids of different speciﬁc weights that don’t mix completely. with the exception that the sand was running downward in both hourglasses. I must admit. MORRIS the ﬁrst time. but let us be honest and not rate it any lower just because it is false. but what a clever ruse it is.” Then. but he didn’t buy one to take home. “somewhat worried and depressed on behalf of my beautiful theory. there was a lull in my mind. He rationalized in jest: “My theory works. “So do I. “This is not a question of fumbling one’s way to a solution. I admit.” Hein was forced to make a paradigm shift. it’s very simple. but of thinking. being false makes it less right. My dear Watson. he was somber. “I am glad you think my explanation is beautiful.” He couldn’t bear to abandon his pet idea completely. “I couldn’t think of anything else but the principle. “I should like to meet the person who invented this effect and designed it so as to hide the solution so elegantly for us.” he explained.“At an altitude of 8500 meters. exactly half moon (D for dynamics) were left behind us and all was dark.” When Piet Hein left Milan on the afternoon of September 21. But it does not make it any less beautiful. but the relevant displacement is in the liquid. Hein had a sudden crystallizing insight .” he wrote on September 22. It really hurt. And on that tablet lay.” he wrote. Only when enough of it has seeped down does the glass begin to rise. the solution of the hourglass mystery. as the plane soared over Mont Blanc. .138 S. and are indistinguishable in color and transparency. not the sand. but it would be easier to explain the symmetry of the phenomenon if it were falling upward in one of them!” Hein acknowledged Gardner’s appreciation of beauty in a theory.” He later wrote about the “Aha” experience. I found myself in possession of the solution. “When the Alps in the lite of the (superelliptic) setting sun and the growing. Hein was awed by the brilliance of it all.

Albert Altman wrote from the U. and the other on the principle I suggested. Naval Ordnance Lab. On September 6th. The height decreases due to the buildup of sand on the bottom of the hourglass and at a critical value the net force on the hourglass acts upward and rises.THE FLOATING HOURGLASS PUZZLE 139 Gardner’s reply on October 5th was a gentle letdown: “You said you would like to meet the clever fellow who thought of the ‘two-liquid’ principle. and to suggest that you not publish a correction. where is the rate of the ﬂow of the sand. the acceleration of gravity.” Reid went on to show mathematically how the impact of sand hitting the glass’ bottom is exactly balanced by the loss of the sand’s mass while it is in free fall. Gardner was beginning to wonder if his friction explanation told the whole story. Well.S. but did not bring one back with him. Could temperature and sand impact also be factors? Would he have to print a correction? He replied to Altman on September 12th: I am embarrassed to admit that your explanation may be right. I am sure that your explanation was correct. at this point I am hopelessly confused. or perhaps still another one. In short. also from the U. The principle is simply too clever to be true. Reid later adapted his letter for publication and his short article “Weight of an Hourglass” appeared in American . One man wrote that he owned one of the Paris cylinders. Reid. and the height through which the sand has fallen. having relied (unfortunately) on an account given to me by a friend who examined the toy in Paris.” The Hourglass Letters While Hein was having his epiphany in Europe. It is possible that one version of the toy works on the principle you mention. Gardner was getting letters in response to the September issue. A letter dated September 29 came from Walter P. perhaps depending on the temperature. You are the inventor. and it didn’t last long. but reported that his glass sometimes ﬂoated and sometimes sank.S. I have not yet seen the toy. Hopelessly confused? Martin Gardner?! That is something that surely doesn’t happen very often. Maryland: Another solution to the hourglass science teaser is that the momentum carried by the falling sand causes the hourglass plus sand to weigh more than its static weight by an amount Ô¾ . all you have to do is shake your own hand. Naval Ordnance Laboratory at Silver Spring. “I am writing to put your mind at ease (on the impact theory posed by Altman).

so there is no visible division between the two. the other down. 1967. Set into type and slated to appear in a summer 1967 column was this recreation of the moment of truth: “When Hein stated his two-liquid theory. Gardner wrote to Dietermann but received no reply.” . which eventually sends one hourglass up. He decided he had to write a correction after all.L. “I ﬁlled the cylinder with water.” he wrote to Hein. but which do not separate completely. and he conﬁrmed Hein’s two-liquid theory. “then I wrapped copper wire around the middle of the hourglass. MORRIS Journal of Physics. Mr.140 S. “Red” Stong. The glasses do tip to the side when inverted. but that is just a holding operation until the liquids have had time to change places. Gardner was playing amateur scientist himself. He turned out to be a Czechoslovakian glassblower named Willy Dietermann. It was the ﬁrst time his secret had been guessed: two liquids of different density. It worked just like the big version. By snipping off the ends of the wire I was able to make its weight such that it slowly rises in the cylinder. Hein wrote that he went back to the shop in Paris where he ﬁrst saw the glass. to my knowledge. completely invisible. He bought a cheap egg timer. and found a transparent cylinder into which it would ﬁt. He also reported it had a freezing point of eight degrees below zero. Stong shone polarized light through the cylinder and found that the refractive index of the liquid did not change from top to bottom. the inventor staggered as if struck. but then in the spring of 1967 the glass rose again. April. the original “Amateur Scientist” columnist of Scientiﬁc American. Gardner had his own hourglass tube. Dietermann explained it was 90 to 95% water and 5 to 10% a combination of alcohol and formic acid. Hein and the Horse’s Mouth All was settled for a few months. He concluded the liquid inside was homogenous. Meanwhile. Gardner cites it in the hourglass puzzle reprinted in Mathematical Circus. He loaned it to C. When Hein asked about the liquid. and tracked down the maker. The main principle is the liquid hourglass effect. 35(4). the only scientiﬁc writing on the subject. This remains. freed the glass from the wooden frame. In his lab. A slight difference in refractive power is concealed by the curvature of the glass cylinder.” Proof Positive By this time.

Canada. sorting them into different piles. “When enough sand falls into the bottom chamber. More than 50 readers thought that the hourglass was ﬂexible. The Dane and the Czech had to speak German because Dieterman knew no Danish or English. G. increasing the hourglass’s volume. and Hein’s French was worse than his German. it ‘bubbles’ the bottom end out. Final Thoughts What fascinated me about the hourglass puzzle was how it led a mind like Piet Hein’s to come up with such brilliantly incorrect theories. Falling sand generates heat. Ontario. The other 60% broke down into about 15 different theories. Q. We may never know what Hein got from the “horse’s mouth. of Richmond Hill.THE FLOATING HOURGLASS PUZZLE 141 Sanity was restored and the correction was never printed. I found the largest single category was always the “correct” theory. Some even used mathematical formulas to show how much force a sand grain exerted. or perhaps there was an honest misunderstanding about the use of different liquids. the hourglass widens and wedges itself into the cylinder. Hein wanted to think the problem through. that the hourglass ﬂoates up with the pocket of warm liquid surrounding the glass’s neck. but they were creative products of human thought.” wrote B. Some reasoned that when the sand presses down from the top. California. This proportion stays at about 40% with each new batch of mail. beautiful explanations. and his German was better than his French.” reasoned D. He wanted to expand his “perpendicular thinking. He wanted to search for alternate. and deserved to be prized for that alone. Some argued that this warms the surrounding liquid so the hourglass stays down until the liquid cools again. About 40% of those readers with incorrect answers cited heat as a factor in the hourglass’s behavior.” I received over 1200 letters about the hourglass after publishing the puzzle in Omni. ﬁrst on the bottom of the glass and later on a . “The top and bottom of the hourglass are so thin as to sag under the weight of the sand.” Perhaps Dieterman was leading him along. making it expand slightly and then rise. There certainly was plenty of room for misinterpretations. As I read them. They may have been wrong. But most in this category thought the heat warms the air in the glass. they said. Others decided that the hourglass is ﬂexible only at the ends. others. Let others measure a refractive index or a freezing point. of Los Altos Hills. Many correspondents blamed the “impact” of falling sand for keeping the glass down.

The theory may be correct. A surprising number thought it was all an illusion. Many believed the air at the top of the hourglass lifts it to the top of the tube. When the air rises to the top half. They concentrated on the air bubble that constantly rises.142 S. California. “It just takes a long time for the hourglass to get started. T. It eventually seeps down below the hourglass and buoys it up. while the weight of the falling sand is required to release the mechanism. and air in the bottom of the hourglass became compressed until “a jet of air shoots into the top half of the hourglass.” explained D. N. the two tiny opposing forces exactly cancel each other out. “The hourglass.” argued P. If the liquid is naturally cooler and denser at the bottom. About 4% of those who wrote in thought that the shape of the hourglass affected its buoyancy. W.” . ﬁrst in the hourglass and then in the tube. wrote one reader. Correct answers to this puzzle aren’t very interesting because they’re all virtually alike. MORRIS mound of other sand grains. Movements within the system don’t alter the weight of the system. of Glendale. I awarded books for incorrect answers only: ½º ¾º Pete Roche of Chicago foresaw this and sent both a correct theory and this incorrect alternate: “The hourglass has a small clasping mechanism at each end. Oklahoma. I promised copies of the book Omni Games to the ﬁve “most interesting” entries. of Tillamook. Therefore. Another line of reasoning put the emphasis on the liquid: “The solution is in the solution!” wrote H. Gagne of Eglin Air Force Base. A.” wrote T. supposed that sand blocked the hole. When the air is in the bottom half. Florida. as a system. “It’s the same principle that causes a snow cone to pop out of its cup when you squeeze the bottom. of Chapel Hill. of Coweta. is rising from the moment the column is inverted. a greater surface area.” perhaps because the liquid is very viscous. H. Oregon. The momentum of rising provides enough energy to engage the mechanism as it reaches the top of the cylinder. North Carolina.” John J. the water below the glass can push up only on the circular end of the glass. imparting just enough lift to overcome inertia and start the glass moving up. “When enough air reaches the top chamber and exerts its pressure there. then the denser liquid is at the top when the tube is turned over. the hourglass begins to rise. water can push up all around the inverted cone. but the calculations have to consider the amount of time each grain of sand is falling and weightless.

Ph. Arizona. Shoreham.S. as Martin Gardner wrote it in 1966: When the sand is at the top of the hourglass. Dinger. therefore. New York. Physics Teacher. the loss of friction enables it to rise. a high center of gravity tips the hourglass to one side. then my name is John Holzapfel and I teach chemistry. Ph.THE FLOATING HOURGLASS PUZZLE 143 ¿º º º “The hourglass is composed of a ﬂexible material such as Nalgene.D. The resulting friction against the side of the cylinder is sufﬁcient to keep it at the bottom of the cylinder.” He added. . theorized that the tube’s caps are hollow and that the ﬂuid must ﬂow from the tube through a hole into the cap below before the glass can rise. goes to Holzapfel. Finally.” The ﬁfth book. and Daniel E. New York. Cliff Oberg of Clarkdale. “P. After enough sand has ﬂowed down to make the hourglass ﬂoat upright. Watson Research Center in Yorktown Heights.D.” wrote Timothy T. The “correct” answer. Shoreham-Wading River High School. Edelstein. a theory about the density gradient of the liquid was signed “Bob Saville. The two share a prize for having the conﬁdence to submit their incorrect theory on company stationery: IBM’s Thomas J.. If this is wrong.

In fact he has recommended that every teacher of algebra construct a set of eight pieces for classroom use. a single cube). With proper insight it can be solved in a few minutes. if colored as above. combinatorial puzzle can be constructed by acquiring twenty-seven uniform cubes of size ¢ ¢ and gluing them together to form the following eight pieces: One “Smally” of size ¢ ¢ (i.e. This puzzle. Three “longs” of size ¢ ¢ . Now color each piece with six colors according to the scheme in Figure 1.Cube Puzzles Jeremiah Farrell Binomial Puzzle A new. For more advanced students it would be reasonable to ask in how many essentially different ways can the cube be constructed from the eight (uncolored) pieces. This will depend on just what is meant by the words “essentially 145 . Most Martin Gardner aﬁcionados will recognize that the eight pieces model the situation represented by the binomial expansion ´½µ ´ · µ¿ ¿·¿ ¾ ·¿ ¾· ¿ Gardner calls such models of mathematical theorems “look-see” proofs. Before proceeding with this discussion the reader is urged to build a puzzle and attempt its solution. Usually there is an “aha” reaction when students see that a cube can actually be constructed from the pieces.. One “Biggie” of size ¢ ¢ (where ¾ ). rather amusing. Three “ﬂats” of size ¢ ¢ . Without this insight it typically takes several hours to arrive at a solution if one can be found at all. The problem then is to arrange the eight pieces into a cube so that opposite faces have the same color. has exactly two solutions — each with a different set of three colors.

C. and 3 are any six distinctive colors. The combinatorial puzzle uses only six colors in its construction. Therefore. different” but one interpretation could be to orient the cube to sit in the positive octant in space with Biggie always occupying the corner (0. ½º Call the six colors A. Trial and error is not a very fruitful way of trying to solve this puzzle. Notice that the eight pieces must form the eight corners of the completed cube — one piece for each corner.½ ¼. 2. If.½½ . FARRELL Figure 1. If you cannot. then there are ´ ¿µ´ µ´¿ µ¾ . 0. There are 93 solutions in this case. A. 2. each . These hints are to be regarded as progressive. B. 1. 1. Solution Hints. after reading a hint.½ distinct ways of constructing the cube. B. 0). and the solver has the additional clue that a face is all of one color. additionally. B. proceed to the next hint. each of the 48 faces of the pieces is colored with a unique color. and 3. A. and C will turn out to be a solution set.146 J. That is. but there still remain a great many cases that are nearly right. C. try again to solve the puzzle.

B–2. When eliminating possibilities. Place Biggie as in the diagram so that A–B–C is a corner. These three ﬂats must cover all or part of the colors 1. It is easy to visualize exactly where a particular ﬂat . and 3 of Biggie. 1–B–C. Of the 20 possible sets of three colors (out of six) we are left with only eight possibilities: A–B–C.) Figure 2. 2 opposes A. 2. A–2–C. A–B–3. A–2–3. ¾º ¿º º To determine the colors A. we may choose another piece. colors that oppose each other cannot appear in the same solution. say a ﬂat. and C. One or two more tests with other pieces lead to A–B–C (or 1–2–3) as a candidate for a solution. Likewise B and 2 cannot. it may happen that on that ﬂat. We would then know to eliminate anything with A and 2. Of course. nor can C and 3. keep A opposite A. and C are showing. etc. and 1–2–3. B. take any piece (Biggie is a good choice). using Biggie as a guide. This means that A and 1 cannot appear together in a solution. 1–2–C. it is convenient to turn the three ﬂats to allowed colors. B. and C–3 oppose each other. B. and notice that the three pairs of colors A–1. Since it is also true that. that four pieces will have the counterclockwise order A–B–C and the other four the order A–C–B. and use it to reduce further the possible solution colors. Place all three ﬂats so that A. This would force A and B to be together. and C in some order. (It is a fact. on any other piece. but not necessary for the solution. For instance. 1–B–3.CUBE PUZZLES 147 piece must have on it a corner with the three colors A.

B. column. FARRELL must go by looking at its A–B–C corner. main diagonal (upper left to lower right). There is another solution to this puzzle using the colors 1– 2–3. one that gives only one solution. Then place the longs and. and C (or two of 1.148 J. We like to have three Smallys prepared: one that yields two solutions. or off-diagonal around all four lateral faces is always 42. This will change the parity on one corner of the cube so that a solution is impossible using that set of three colors. That is. take any piece and interchange two of the colors A. The solver will notice that the completed cube has a “fault-free” property. ﬁnally. This will assure that only one solution is attainable with the colors A–B–C (the proof is left to the reader). . and 3). If only one solution is desired. The Magic Die has the amazing property that the sum of any row. no seam runs completely through the cube in any direction and therefore no rotation of any part of the cube is possible. Smally. 2. Figure 3. and one to slip in when our enemies try the puzzle that is colored so that no solution is possible! Magic Die Figure 3 shows the schematic for a Magic Die.

and. you can copy Figure 3 and paste it onto heavy paper to make a permanent Magic Die. There will be only one solution to this puzzle (with magic constant 42). (Alternatively.CUBE PUZZLES 149 You can construct a Magic Die puzzle by taking twenty-seven dice and gluing them together into the eight pieces of our combinatorial cube puzzle— being sure the dice conform to the layout shown in Figure 3. even with the schematic as a guide.) . it will be extremely difﬁcult to ﬁnd.

has numerous variations and solutions. and requires logical decision-making. combinatorics. is a manipulative puzzle. and. Figure 1. Typical solutions are shown at the end of this paper. A typical set is shown below in Figure 1. I was intrigued by the puzzle’s characteristics but did not. with each cube a different color.” I learned of the puzzle in 1977 from Reference 2. and twelve different dicubes with each cube of a dicube a different color.The Nine Color Puzzle Sivy Farhi The nine color puzzle consists of a tricube. the nine color puzzle was ﬁrst introduced during 1973. until recently. and isomorphisms. has an easily understood objective. The object of the puzzle is to assemble the pieces into a cube with all nine colors displayed on the six faces. analyze it thoroughly. parity. This puzzle is of particular interest because it is of simple construction. A complete analysis requires the use of a computer. sets. Altogether there are three cubes each of nine different colors. Included in its analysis are the concepts of transformations. using the name “Nonahuebes. where each number represents a color. but after reading this article it should not be 151 . ordered pairs. as “Kolor Kraze.” included it among the puzzles I produced in New Zealand under the aegis of Pentacube Puzzles. According to Reference 1. in Canada. backtracking. Ltd.

only thirty dicubes are required.677. is that there are numerous solutions.700 (36!/12!24!) different ways of which. The question then arises: How many solutions are there? Most intriguing about this puzzle. The individual dicubes can be colored thirty-six different ways. Twelve of these can be selected in 1. it becomes apparent that most of the 133. and only 10. Thus. as is shown in Section 5.152 S. but not in the center of a face. Fortunately. when two cubes of the same color are in position.251. It becomes necessary. obtained after some experimentation. The third observation. The ﬁrst observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube.691 combinations need to be sorted into disjoint sets. Since one person can assign the fourth and ﬁfth colors to orange and red and another person to red and orange. none of the nine planes may contain two cubes of the same color. Since during the course of trying to solve this puzzle a conﬂict often occurs. The second observation is that for a cube to be assembled with the nine colors on each face. some with the tricube on the edge and some with the tricube through the center of the cube. . only 133.105 combinations are isomorphisms. The previous question now becomes more interesting. the location of the third cube is determined. to separate these cases into disjoint sets of isomorphisms. This was achieved through the use of a computer program. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. then. The method of sorting the 10. A proof is given in Section 3. in which the number of isomorphisms using the set of thirty-six dicubes is also determined. the results are shown in Table 1. A proof is given in Section 2. FARHI a difﬁcult exercise for a moderately proﬁcient programmer to verify the results obtained. as determined in Section 4. this rule then tells the experimenter to backtrack.105 of these combinations meet the three-cubes-of-each-color requirement.691 combinations into 148 disjoint sets is described in Section 6. The ﬁnal part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. and the most difﬁcult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1.

DZppp 360 156 428 24 208 8 36 /90eyz 180 60 16 0 38 32 0 /:0eyz 360 120 80 4 24 4 4 /:Z[yz 720 240 182 8 50 16 24 /:Zpez 360 120 124 4 38 44 16 /:Zppp 180 60 408 0 186 32 64 /COeyz 360 72 0 0 6 4 0 /DXeyz 720 132 18 0 4 2 0 /DZ[yz 720 132 72 0 32 12 0 /DZepz 1439 264 41 0 40 9 0 /DZppp 360 66 80 0 117 28 0 /MNeyz 60 8 0 0 0 0 0 /MOeyz 720 96 8 0 34 8 16 /MXeyz 720 84 32 0 30 12 12 /MYeyz 720 84 26 0 18 6 4 /MZ[yz 1439 168 62 2 70 21 23 /MZepz 1439 168 41 3 35 15 16 /MZpez 1439 168 51 4 57 17 20 /MZppp 720 84 68 4 73 26 14 /NXeyz 720 96 8 0 36 12 12 /NYeyz 1439 192 40 2 64 17 24 /NZmyz 720 96 8 0 10 2 0 /NZnpz 1439 192 53 2 53 11 17 /NZpez 720 96 36 4 48 24 32 /NZpfp 1439 192 10 0 49 34 23 /Ob[yz 1439 168 25 1 81 24 16 /Obmyz 1439 168 98 5 65 18 19 /Obofz 720 84 6 0 18 6 8 /Obpez 1439 168 124 3 53 8 17 .:Zppp 10 10 0 0 0 0 0 . 153 No.DOeyz 720 312 20 0 12 0 4 .:Z[yz 60 60 0 0 0 0 0 .DZpez 720 312 198 12 83 10 16 .DZ[yz 1439 624 96 4 80 6 15 .THE NINE COLOR PUZZLE Table 1.COeyz 360 180 8 0 16 16 0 . 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center .

FARHI No. 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center /Obppp 720 84 38 2 104 46 20 /Oceyz 720 84 18 0 46 6 8 /Ocmyz 1439 168 34 3 27 12 2 /Ocnpz 1439 168 126 4 117 16 19 /Ocnyz 1439 168 40 1 36 11 11 /Ocnzp 1439 168 49 3 45 24 9 /Ocoxz 1439 168 112 5 14 3 2 /Ocozp 1439 168 18 1 35 28 9 /Oewez 720 84 32 2 38 48 4 /Oewnz 1439 168 52 3 20 15 1 /Oewpp 1439 168 17 6 38 12 4 /Oeyxp 720 84 36 0 12 6 4 9DEeyz 180 36 0 0 16 0 0 9DNeyz 30 6 0 0 0 0 0 9DOeyz 360 72 0 0 32 12 4 9DXeyz 720 108 10 0 0 0 0 9DYeyz 720 108 22 2 38 18 6 9DZ[yz 1439 216 69 3 32 17 2 9DZepz 1439 216 71 7 32 16 7 9DZpez 1439 216 55 6 52 13 5 9DZppp 720 108 54 12 72 18 20 9ODeyz 360 36 4 0 4 0 0 9OF[yz 360 36 76 0 72 8 4 9OFepz 720 72 82 0 45 0 6 9OFppp 180 18 208 0 134 8 0 9Ob[yz 720 72 54 4 41 12 2 9Obepz 1439 144 145 7 48 13 5 9Obmyz 1439 144 87 2 31 8 0 9Obnpz 720 72 2 0 18 4 0 9Obofz 720 72 68 4 28 2 0 9Obpez 720 72 22 2 25 4 0 9Obpfp 720 72 174 6 71 10 0 9Obppp 1439 144 30 3 26 5 4 9Odmpz 720 72 186 6 54 10 2 9Odmzp 360 36 16 0 6 12 0 9Odoyz 360 36 76 8 66 16 4 9Oewpp 360 36 40 0 16 4 0 DDDeyz 15 1 0 0 0 0 0 .154 S.

THE NINE COLOR PUZZLE 155 No. 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center DDEeyz 720 36 2 2 2 2 0 DDZQyz 180 7 0 0 0 0 0 DDZ[yz 1439 56 1 3 5 0 1 DDZofz 360 14 4 0 2 0 2 DDZpez 720 28 0 4 4 0 0 DDZppp 360 14 0 12 6 0 4 DEDeyz 360 12 4 4 4 2 0 DEF[yz 1439 48 10 3 15 3 0 DEFepz 720 24 18 6 16 0 0 DEFppp 360 12 0 20 7 6 0 DEOeyz 720 24 0 1 2 0 2 DEPQyz 720 16 6 0 0 0 0 DEP[yz 1439 32 1 0 0 0 0 DEPepz 1439 32 8 3 3 1 2 DEPmyz 1439 32 11 2 1 0 1 DEPnpz 1439 32 23 5 3 0 1 DEPofz 1439 32 13 3 10 1 1 DEPpez 1439 32 17 2 17 1 3 DEPpfp 1439 32 2 1 0 0 0 DEPppp 1439 32 12 8 2 1 0 DEZQyz 1439 48 23 5 6 0 1 DEZ[yz 360 12 0 0 0 0 0 DEZeyz 1439 48 4 6 4 0 1 DEZnpz 720 24 26 2 4 0 0 DEZnyz 720 24 14 0 14 0 1 DEZnzp 720 24 4 4 4 0 2 DEZozp 720 24 10 14 6 0 3 DEeOyz 720 16 6 2 0 0 0 DEePpz 1439 32 17 0 0 0 0 DEeRez 720 16 44 0 0 0 0 DEeRfp 1439 32 16 0 0 0 0 DEeYyz 720 16 0 0 0 0 0 DEeZpz 720 16 22 0 2 0 0 DEeZyz 1439 32 8 3 6 0 0 DEeZzp 1439 32 4 0 0 0 0 DEecpz 1439 32 54 5 9 0 0 DEecyz 1439 32 39 3 48 3 0 DEeczp 1439 32 18 1 10 2 0 .

112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center DEedyz 1439 32 13 1 11 2 0 DEeexz 1439 32 28 0 23 2 0 DEeeyz 1439 32 15 1 2 0 0 DEewez 1439 32 27 2 8 0 3 DEewfp 1439 32 36 0 1 0 0 DEewxp 1439 32 34 1 2 0 1 DEexnz 1439 32 33 1 10 0 1 DEexxp 1439 32 31 1 47 1 1 DEexyy 1439 32 47 0 3 1 1 DZDQyz 90 1 0 0 0 0 0 DZD[yz 720 8 2 2 6 0 0 DZDofz 180 2 8 0 0 0 0 DZDpez 360 4 0 0 12 1 0 DZDppp 180 2 0 8 6 0 2 DZGOyz 1439 16 28 0 17 1 1 DZDQfz 720 8 4 0 2 0 0 DZGRez 1439 16 20 0 11 0 0 DZGRpp 720 8 62 0 74 4 2 DZGYyz 1439 16 15 3 9 2 0 DZGZpz 1439 16 39 3 9 0 0 DZGZzp 720 8 4 2 6 0 0 DZG[xz 720 8 26 0 15 0 0 DZGwez 1439 16 29 2 2 0 0 DZGwnz 1439 16 36 0 8 0 0 DZGwpp 720 8 4 1 3 0 0 DZGwxp 720 8 45 0 15 0 0 DZpO[z 180 2 0 0 2 0 0 DZpOez 360 4 25 0 3 0 0 DZpOnz 720 8 200 6 22 0 1 DZzOpp 1439 16 183 4 51 1 1 DZpQnz 360 4 8 0 0 0 0 DZpQxp 720 8 129 2 10 1 0 DZpZpp 720 8 172 6 61 2 0 DZp[ p 360 4 65 14 16 1 4 DZp[fp 360 4 24 6 12 4 1 DZp[fy 360 4 21 0 4 0 0 Dzp[pf 720 8 137 0 69 1 1 .156 S. FARHI No.

where odd numbers represent one state and even numbers the other. Verifying this would be an interesting exercise for the reader. these two cases have no puzzle solutions. It is not necessary that the tricube be straight. as mentioned in Section 2. 14. Is this a coincidence? 2. There are two rather startling phenomena shown in Table 1. The number of solutions for these cases are included in Table 1. and 19 is referred to as an “edge case” while the tricube location in spaces 5. and. .THE NINE COLOR PUZZLE 157 1.” Other possible locations are reﬂections or rotations of these and are not considered as separate puzzles solutions. While some color combinations within the set of thirty dicubes have up to 720 isomorphisms. Each dicube when similarly checkered has one even and one odd number. no matter how the colors are swapped. map onto themselves. and 23 is referred to as a “center case. Permissible Locations of the Tricube as Determined by Parity Restrictions Figure 2. The tricube location ﬁlling spaces 1. Additional Comments The Kolor Kraze puzzle shown in Reference 1 and the nine color puzzle shown in Reference 2 are isomorphic. Since the cube has fourteen odd numbers and thirteen even numbers. The three layers of the cube are checkered as shown in Figure 2. There are only two possibilities. the tricube must ﬁll two odd numbered spaces. an irregular tricube may have three locations. there are two color combinations (Numbers 73 and 121) which. 10. thus leading me to believe that they originated from the same source. Stranger yet.

and on a centerface such as shown by the letter “B. such as shown by the letter “C. and 3. The “No Two in the Same Plane” Rule Figure 3.158 S. If none of the previous selections include color 3. and 14. and 11. 2.” In each case.” The remaining six corners must have cubes of different colors. the edge case ﬁlls spaces 2. 11. The corner case ﬁlls spaces 1. there are three possible locations. The only way the remaining two cubes of that color can be displayed on all six faces is for them to be located on opposite corners as shown in Figure 3 by the letter “A. If color 2 is selected from the ﬁrst row.” The remaining two colors must all be on edges. accounting for two faces. . FARHI If an irregular tricube is used. then a third dicube with color 2 and any color from the second row is selected. Since the corner colors are displayed on three faces. Determination of the 133. If color 3 has been selected once. 3. One cube is centrally located and thus not visible. then none are selected from the third row. 4. then one color is selected from the third row and if color 3 has been selected twice. then color 3 is combined with two colors from the third row. cubes of the same color are never in the same plane. and 11. 2. 5. The two dicubes with color 1 may use any two colors in the ﬁrst row of the following table. the center case ﬁlls spaces 5. the remaining two cubes must be located on an edge.105 Possible Dicube Color Combinations The tricube colors are deﬁned as colors 1. If color 2 is not selected from the ﬁrst row then color 2 is combined with any two colors from the second row for the third and fourth dicubes.

5. 6. 9. (5. Identiﬁcation of the second color of these ordered pairs is sufﬁcient to deﬁne all the colors. (4. and 9 one can logically deduce the ﬁrst colors by the three . 3). or 5. 8). (6.THE NINE COLOR PUZZLE 159 First Color 1 2 3 4 5 6 7 8 Second Selection 2 3 4 5 6 7 8 3 4 5 6 7 8 4 5 6 7 8 5 6 7 8 6 7 8 7 8 8 9 9 9 9 9 9 9 9 The process continues: color 4 is combined with colors in the fourth row such that there are three cubes with color 4. 7. Using this algorithm. (2. 8. 5. 3. (7. using the same algorithm as before. 9). 6. 6. a computer program established 133. etc. Reduction of the Number of Isomorphisms The table in Section 4 is simpliﬁed by restricting the selection in the ﬁrst two rows as shown below. 8. 9.691 color combinations. (5. 4. (7. Color 1 2 Selection 2 3 4 5 3 4 5 6 7 This can be justiﬁed by deﬁning the colors associated with color 1 that are not 2 or 3 as colors 4 and 5. only 10. (3. This results in requiring only thirty cubes and.105 possible color combinations. 8). 6). 5). Similarly if color 2 is not associated with colors 3. the two new colors are identiﬁed as colors 6 and 7. color 5 is combined with colors in the ﬁfth row such that there are three cubes with color 5. Identiﬁcation and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. For example. 2). 4. given the second numbers 2. 6). 9) and (8. (1. 7). (4. 4). 9).

The process was then continued until the list was exhausted. (5. Thus this color combination is identiﬁed as /:Z[yz. . 90. This process then identiﬁed the 148 disjoint sets of isomorphisms. can be swapped with each other. . Not all of these mappings produce a new isomorphism. 9). i. 35. Identiﬁcation was further simpliﬁed by placing these numbers in pairs (see Note below). (6.:Zppp. (2. (7. four at a time. (4. and 99. 6). and then assigning a printable character to each. the ﬁrst entry in Table 1 is . 6). 68. on the ends of the tricube. The head of the list then became . (1. Decoding a name is quite simple. ﬁve at a time including three at a time with two at a time. 3). 7).:Z[yz. has sixty isomorphisms.160 S. 2). pairs of three at a time and four at a time with two at a time. Color 2. 8). (24). (4. 98.691 cases into 148 disjoint sets of isomorphisms. including pairs of two at a time. (35). Shown in Figure 4 are an edge solution and a center solution for the color combination of Figure 1. then additional criteria for recording solutions are recommended. For example. It is then a simple process to recognize that this represents the dicubes colored (1. which is in the center of the tricube. three at a time. 9). Colors 1 and 3. The solutions are identiﬁed by a three-digit number indicating the color of the buried cube and the two “edge” colors. The ASCII code for 23 is not a printable character. 121. These were removed from the list. FARHI cubes of each color requirement and the restriction that the second number be greater than the ﬁrst. (7. 9). (67). its isomorphisms are determined and removed from the list. and (8. The ASCII codes for these characters are: 46. The use of this six-character identiﬁcation greatly simpliﬁed the process of sorting the 10.e. cannot be interchanged with other colors. 5). but not with any of the other colors. (4. 3). (68). The remaining six colors may be interchanged in numerous ways: two at a time. (5. Any permissible interchange of colors is an isomorphism.:Z[yz. by adding twenty-three to each and using the ASCII character set codes (ASCII stands for American Standard Code for Information Interchange). and (99). Subtracting twenty-three results in: 23. and 122. and six at a time including a triplet of two at a time. Note: The smallest possible number is 23 and the largest is 99. 91. 8). The 10. (98).691 color combinations obtained by the algorithm described in Section 5 were listed in order according to their ASCII characters. If the solutions are to be catalogued. 67. The ﬁrst entry.. An inspection of the ASCII table will explain why it was decided to add 23. 58. Different solutions often have the same three-digit identiﬁcation.

and Botermans. 200 pp. the plane can be translated to the other side resulting in a new solution. Using these transformations. Harry N. . J.org. and Torbijn. New York. softcover. dicubes (4. P. CEDIC. A preliminary investigation shows some merit in utilizing graph analysis to illustrate why some color combinations have no solutions while other color combinations have numerous solutions. 9) in the bottom layer may be swapped with dicubes (4. 9) on the upper right. Abrams. the reader should now be able to determine ﬁve more edge solutions. Jack. hardcover. [2] Meeus. 1977. Polycubes. Whenever the tricube is on an edge and shares a plane with only three dicubes. 6) and (7. Creative Puzzles of the World. all 148 graphs have not been compared. Jerry. Puzzle sets or puzzle solutions may be obtained by contacting the author by E-mail at sivy@ieee. Stan Isaacs has suggested that graph analysis would be useful in determining the number of solutions..THE NINE COLOR PUZZLE 161 Figure 4. The edge solution is interesting in that new solutions are often obtainable by using two transformations. in French. Often two dicubes may be swapped with two others having the same colors. 1978. Distracts 4. Typical solutions. Unfortunately.. Paris.. 176 pp. J. 7) and (6. References [1] Slocum. France. For example.

or no channel at all. but only eight are used in each puzzle. Channels (or grooves) are cut into the bases of some blocks.” The name was chosen because there are two quite different puzzles involving two different blocks numbered 2(a) and 2(b). Figure 1. 163 . Blocks can either have a horizontal channel. allowing them to pass over the pegs in the corners (see Figure 2). There are nine square blocks (including two numbered 2). Figure 2. cannot get there at all. Italy. one being used in each puzzle. in some cases.Twice: A Sliding Block Puzzle Edward Hordern Twice is a new concept in sliding block puzzles: Some blocks are restricted in their movements and can only reach certain parts of the board from particular directions or. one in each corner (see Figure 1). a vertical channel. The puzzle was invented by Dario Uri from Bologna. and these blocks can only reach the corners from a horizontal Written by Edward Hordern and reproduced with the kind permission of Dario Uri. both channels. Description Fixed into the base of the board are four pegs. and was originally issued in 1989 with the name “Impossible!!” Subsequent additions and improvements made over the next couple of years led to a change of name. Blocks A. to “Twice. 7. and 2(b) have a single horizontal channel.

and yet so far! In puzzle 2 it is quite a task to move block A more than a square or two. This causes something of a trafﬁc jam. It is quite an achievement just to solve them. they must only move in a cross-shaped area. The object of the puzzle in each case is to move block A to the bottom right corner. The delight of the ﬁrst puzzle is that it is very easy to move block A to just above the bottom right corner. These blocks. The real “problem” blocks are block 7 in the ﬁrst puzzle and blocks 7 and 2(b) in the second. the second being the harder of the two. The Puzzle Figure 3 shows the start position. or to get anywhere at all! Hints for Solving In both puzzles the “nuisance” blocks are 3 and 6.164 E. the shortest known solutions are 50 moves for the ﬁrst puzzle and a staggering 70 moves for the second. a plan can be made as to which blocks have to be moved so as to allow block A to pass. So near. During the solution they must continually be moved “around a corner” to get them out of the way of another block that has to be moved. however. Blocks 1 and 5 have both channels (in the form of a cross) and can go anywhere. as well as block A. Blocks 4 and 2(a) have a single vertical channel and can only approach the corners from a vertical direction. only to get hopelessly stuck…. The puzzles are both rated as difﬁcult. . and the second puzzle uses block 2(b). can only move freely up and down the center of the puzzle. HORDERN direction. which has to be overcome…. Blocks 3 and 6 have no channels and cannot go into any corner. Figure 3. Since they can’t go into the corners. Both solutions are believed (but not proved) to be minimum-move solutions. Once this has been mastered. The ﬁrst puzzle uses block 2(a). For the expert.

it cannot be realized as a planar burr because in two dimensions each piece should be disconnected.” which requires twelve moves! Recently. until the two smaller pieces are free. Dewdney’s Scientiﬁc American column (January 1986. in Japan. 16). Since about 1985 a “Most Moves Competition” for six-piece burrs has been running. The idea of a two-dimensional burr immediately appealed to me. This puzzle has two congruent large pieces and two congruent smaller ones. an ingenuously designed puzzle that he called “Four Sticks and a Box. Muroi wrote me that his idea was inspired by a design of Yun Yananose. I am not completely aware of the state of the competition.” The puzzle has no more than four movable pieces. with only three moves needed to remove the ﬁrst piece. all interactions between the four pieces take place in one plane. but nevertheless requires twelve moves (!) to get the ﬁrst piece out of the box. The most common are the six-piece burrs. p. so in some sense it might be considered a planar burr. One result is the “Zigzag” planar burr.Planar Burrs M. because these can be so constructed that several moves are needed to separate the ﬁrst piece. depicted in Figure 1. The same movements. Oskar van Deventer Usually “burrs” are considered to be three-dimensional puzzles. Though Muroi’s puzzle is three-dimensional. but as far as I know Bruce Love is the record holder with his “Love’s Dozen. It takes ﬁve moves to separate the ﬁrst piece. True planar burrs are rarely found. and in April 1986 I made some attempts to ﬁnd a design of my own. To solve the puzzle. K. in backward order. Carter’s puzzle has four pieces. Bill Cutler’s “Bafﬂing Burr” and Philippe Dubois’ “Seven Up” were the ﬁrst attempts. depicted in A. which occur in many different designs. The most interesting are those with internal voids. will separate large pieces. who was inspired by “Dead Lock. However. 165 .” a puzzle of mine. the two large pieces move into each other along a zigzag line. I received from Tadao Muroi. The ﬁrst design I saw is Jeffrey Carter’s. The puzzle is not very difﬁcult to solve.

Nine and One-Half Moves.166 M. The three pieces of the puzzle form a square with internal voids.” This is is a true two-dimensional burr of only three pieces and it needs no less than nine and one-half moves to separate one piece from the other two. particularly inspired by Muroi’s “Four Sticks” (so closing the circle of mutual inspiration). O. Zigzag. V AN DEVENTER Figure 1. Some months ago. Figure 2. I took up the challenge again and succeeded in ﬁnding a new design of a planar burr. . which I have called “Nine and One-Half Moves.

we can glue one piece of the puzzle between two square plates as indicated in the top left corner of Figure 2. The ninth move is a slide-plus-rotate move. By using opaque plates the design is hidden as well. In order to prevent us three-dimensional people from cheating. A round hole can be used to hide a coin. and the moves required to separate the pieces are depicted in Figure 2. so I count it as a move and a half. .PLANAR BURRS 167 The pieces.

7. but the “tricks. Presented here are 11 such “block-packing” puzzles. date of design. It is 3-dimensional box-packing puzzles where the box and all the pieces are rectangular solids. The number of such puzzles that I am aware of is quite small. All Pieces the Same “Aren’t these puzzles trivial?. If known. and manufacturer are given.Block-Packing Jambalaya Bill Cutler My primary interest over the years has been burr puzzles. The puzzles are grouped according to whether there are holes in the assembled puzzle and whether the pieces are all the same or different. The following are my favorites: ½º ¾º unnamed (44) (Singmaster–Klarner): Box: ¢ ½½ ¢ ¾½ Pieces: (44) ¾ ¢ ¿ ¢ unnamed (45) (de Bruijn): Box: ¢ ¢ Pieces: (45) ½ ¢ ½ ¢ 169 . the total number of pieces is in parentheses.” you ask. the inventor of the puzzle. Well. No Holes. I know of no other small group of puzzles that encompasses such a rich diversity of ideas. but there is another small category of puzzles that is especially intriguing to me. David Klarner gives a thorough discussion of this case in [5].” or unique features that the puzzles employ are many and varied. but there are some interesting problems. For each puzzle. The tricks to most of the puzzles are discussed here. but complete solutions are not given. you are not far from being completely correct.

(6) ½ ¢ ¾ ¢ ¾ unnamed (18) (John Conway): Box: ¢ ¢ Pieces: (3) ½ ¢ ½ ¢ ¿. ½ ¢ ¿ ¢ ½¼. rigorous analysis of the puzzle can be done by hand in about 15 minutes. . The seven pieces can be placed in the box in ten different ways. ¾ ¢ ¿¢¿¢ ¢ . the three ½ ¢ ½ ¢ ¿ pieces must be used sparingly. Limited Number of Piece Types The puzzles I know of in this category follow a common principal: There are basically two types of pieces — a large supply of one type and a limited supply of another. 9. The small box is very easy. but must be used efﬁciently to solve the puzzle. (1) ¾ ¢ ¾ ¢ ¾. The large box is moderately difﬁcult. The solver must determine exactly where the second set of pieces must be placed. See [2] or [5] for more information. No Holes. The 18 pieces are all different. ¾ ¢ ¿ ¢ . (13) ½ ¢ ¾ ¢ In the ﬁrst of these. it is easy to see that the cubes must be placed on a main diagonal. CUTLER 8.¿¢¿¢ . Box 2: ¢ ¢ ½¼ Pieces: ½ ¢ ¿ ¢ . ¿º º unnamed (9) (Slothouber–Graatsma): Box: ¿ ¢ ¿ ¢ ¿ Pieces: (3) ½ ¢ ½ ¢ ½. Naef): Box 1: ¢ ¢ Pieces: ¾ ¢ ¿ ¢ ¿. and has many solutions. the three individual cubes are obviously easy to place. but it does make a nice set of puzzles. ¢ ¢ Box 3: ¢ ¢ ½¼ Pieces: all 18 pieces from ﬁrst two boxes . ¾¢¿¢ . The pieces of the second type are smaller and easier to use. In the second design. The rest of the pieces. (1) ½ ¢ ¾ ¢ ¾. although not exactly alike. ½ ¢ ¿ ¢ . and then the rest is easy. No Holes. but they must not be wasted. Quadron does not use any special tricks that I am aware of. and the three boxes offer a wide range of difﬁculty. ½ ¢ ¾ ¢ ¿ ¢ . function similarly to the ½ ¢ ¾ ¢ ¾ pieces in the ﬁrst puzzle. ¾ ¢ ¢ .170 B.¿¢¿¢ ¢ . By analyzing “checkerboard” colorings of the layers in the box. A complete. not counting rotations and/or reﬂections. Pieces Mostly Different º Quadron (18) (Jost Hanny. ¾ ¢ ¿ ¢ . The middle-size box is difﬁcult — there is only one solution. ½ ¢ ¿ ¢ . ¿ ¢ ¿ ¢ ¿.¾¢ ¢ .

½ ¢ ¾ ¢ ¾. the rate increases to 40 pieces/second. Bill Cutler Puzzles): Box: ½ ¢ ½ ¢ ½ Pieces: ½¿ ¢ ½½¾ ¢ ½ ½. ¢ ½¼. all very similar. This leaves 4 pieces that might be placed sideways. which is no good for this purpose. ¢ ½ . ¢ ½¿. copied from a model in the collection of Abel Garcia): Box: ¢ ½ ¢ ¾ Pieces: all pieces are of thickness 2 units. º Boxed Box (23) (Cutler. When running these programs on more powerful computers. it is tempting to solve the puzzle by constructing three layers of pieces. The result was a fascinating demonstration of how a computer can be used to solve such a puzzle.450. ¢ ½½. the difference between the three boxes is stunning: The ﬁrst box can be completely analyzed in a small fraction of a second. 1978. ¢ ½ . ½ ¢ ¢ ¼. The solution involves use of the following obvious trick (is that an oxymoron?): Some piece(s) are placed sideways in the box. I painted pieces of an actual model to match the display. . One or two individual layers can be constructed. There are 3. ¢ . The program displayed the status of the box at any instant using color graphics. º Parcel Post Puzzle (18) (designer unknown.BLOCK-PACKING JAMBALAYA 171 Quadron also makes for a nice entrance into the realm of computer analysis of puzzles and the limitations of such programs. ¾¢ . or the equivalent of one year on one machine. ½¼ ¢ ½½. ½ ¢ ¼ ¢ . I did a complete analysis of the third box. but there are more efﬁcient algorithms that can be used for block-packing puzzles. a Commodore 64. not counting rotations and reﬂections.480 solutions. ½ ¢ ¿ ¢ . The total CPU time used was about 8500 hours. and they all have three of these four pieces placed sideways. The second box was analyzed in about a minute of mainframe computer time. Since all the pieces are of the same thickness and the box depth equals three thicknesses. about once a second a piece is added or removed from the box! Using compiled BASIC. ¢ ½¿. ¾½ ¢ ¢ ½ ¼. By the end of the runs. ¢ ¾½. ½ ¢ ½ ¢ ¾ ¢ . but the process cannot be completed. The programmer can use algorithms that are used for pentominoe problems. I wrote such a program on my ﬁrst computer. The analysis was done on about 20 powerful IBM workstations. the machines had constructed 2 1/2 trillion different partially ﬁlled boxes. ¢ . The Commodore 64 is such a wonderously slow machine — when running the program in interpreter BASIC. and ¢ ½¿. Of the 18 pieces. ½½ ¢ ½½ and two each of ¢ . 10 are too wide to ﬁt into the box sideways and 4 are of width 5. In early 1996. ¢ ½ . There are four solutions to the puzzle. ¾¼ ¢ ¼ ¢ ¾. the widths and lengths are ¢ .

½¼ ¢ ½ ¢ ¾ . ¢ ½¼ ¢ ½½. There are 21 solutions. Almost 15 years later. ¿¿ ¢ ¿ ¢ ¿ ¢ ½¾½. ¢ ¢ . I like the dimensions above because it tempts the solver to stack the pieces three deep in the middle dimension. ½½ ¢ ½½ ¢ ¾ .172 B. ¿ ¢ ½½¼ ¢ ½¿ . Bill Cutler Puzzles) Box: ¼ ¢ ¾ ¢ ¾ Pieces: ¢ ½ ¢ ¾ . ½½ ¢ ½¾ ¢ ¾. . ½¼ ¢ ½ ¢ ¿½. They can be any three different numbers. CUTLER ¾¾ ¢ ½¼ ¢ ½¿½. 1986. ½½ ¢ ½ ¢ ¿¼. 10. The dimensions of the pieces can be modiﬁed. º Hoffman Junior (8) (NOB Yoshigahara. ½ ¢ ½ ¢ ¾½ The original design of Cutler’s Dilemma had 23 pieces and was constructed from the above. The net result was a puzzle that is extremely difﬁcult. ½½ ¢ ½ ¢ ½ . Holes. this puzzle still fascinates me. version by cutting some of the pieces into two or three smaller pieces. ¢ ¢ ½½. ¢ ¼. ½¾ ¢ ½ ¢ ½ . ½¼ ¢ ½¾ ¢ ¾ . The pieces ﬁt into the box with no extra space. ¾ ¢ ¿¼ ¢ ¢ ½¿ . The smallest number for which this can be done is 23. Pieces Different ½¼º Cutler’s Dilemma. See [1] or [3] for more information. ¢ ½¼ ¢ ½½ 11. ¾¿ ¢ ½ ¢ ¿. none having any symmetry or pattern. ½¼ ¢ ½ ¢ ¾ . Holes. ½½ ¢ ½ ¢ ¾ . ¿ ¢ ¾ ¢ ½¾ . ½¼ ¢ ½½ ¢ ¾. ¿½ ¢ ¢ . The extra space makes available a whole new realm of possibilities. I will not say anything more about this design except that the trick involved is different from any of those used by the other designs in this paper. 1976) Box: ½ ¢ ½ ¢ ½ Pieces: (27) ¢ ¢ This sounds like a simple puzzle. ½ ¢ ½ ¢ ¾½. Simpliﬁed (15) (Cutler. ¾ ¢ ¿ ¢ . ¢ ¢ The dimensions of the pieces are all different numbers. 1981. See [4]. ¢ ¾¼ ¢ ¾¼. Hikimi Puzzland) Box: ½ ¢ ½ ¢ ½ Pieces: Two each of ¢ ¢ ½¼. Pieces the Same or Similar º Hoffman’s Blocks (27) (Dean Hoffman. ¾ ¢ ¢ ½¾¿. basic. but it is not. ¿ ¢ ¾ ¢ ¼. where the smallest is greater than one-quarter of the sum. The box is a cube with side equal to the sum. ¢ . There are many other 23-piece solutions that are combinatorially different from the above design.

112–117. (No.¿ ¢ ¢ ¢ ¢ . pp. Klarner.BLOCK-PACKING JAMBALAYA 173 12. including the one in your hand. Cutler. “Subdividing a Box into Completely Incongruent Boxes. readers. February 1976. 6(2).¿ ¢ ¢ plus a second copy of ½ ¢ ¾ ¢ . Mathematical Games column in Scientiﬁc American. one of the puzzles listed above is impossible. but rather a tautology to the 99% of the world that would never even have started to read this article. being careful to keep the renegade piece hidden from view. “Packing Problems and Inequalities. Hoffman. 1979–80. [4] D. ¿ ¢¾ ¢ .” J. When the ninth piece is added to the group. I won’t say which one (it should be easy to ﬁgure out). the pieces can be rearranged to make a ¢ ¢ ½¿¿ rectangular solid. pp. (This second construction is a little more difﬁcult. Miscellaneous ½½º Melting Block (8-9) (Tom O’Beirne) Box: ¢ ¢ ½¿¿ Pieces: ½ ¢ ¾ ¢ . You are now prepared for your next encounter with a boring puzzle-nut. Gardner. 212– 225. 12(2). being sure that the unﬁlled space is concealed at the bottom and is stable. Recreational Math. Show off the solved box to your victim. [5] D.. Klarner (Wadsworth International. Spring 1973. pp. Place the box on your puzzle shelf with the remaining piece hidden behind the box. pp.) This is a great puzzle to show to “non-puzzle people” and is one of my favorites. The eight pieces ﬁt together easily to form a rectangular block ¢ ¢ ½¿¾.½ . except one. . “Brick-Packing Puzzles. and then dump the pieces onto the ﬂoor. Mathematical Games column in Scientiﬁc American. 1981). 104–111. By the way. pp.” in The Mathematical Gardner. February 1979. 122–127. into the box.. Pack all the pieces. but seems to the casual observer to ﬁll up the box completely.) Pick up the box and last piece with both hands. [3] M. This ﬁts into the box with a little room all around. It is a valuable weapon in every puzzle collector’s arsenal. The Melting Block is more of a paradox than a puzzle. 20–23. [2] M.” J. edited by D. Recreational Math.¿ ¢¾ ¢ . This should keep him busy for quite some time! References [1] W. this is not another oxymoron. ½ ¢ ¾ ¢ ½ ¢ ¢ . Gardner.

CUTLER Written on the occasion of the Puzzle Exhibition at the Atlanta International Museum of Art and Design and dedicated to Martin Gardner.174 B. .

contrived for the principle purpose of exercising one’s ingenuity and/or patience. A mechanical puzzle is a physical object comprising one or more parts that fall within the above deﬁnition. To provide a logical and easy-to-use classiﬁcation to enable nonexperts to ﬁnd single and related puzzles in a large collection of objects. barrel. etc. A puzzle should be classiﬁed by the problem that its designer intended the solver to encounter while attempting to solve it. (As presented here. but most attempts so far have either been far too specialized in application or too general to provide the basis for a deﬁnitive classiﬁcation. and patents. In cases where it seems possible to place a puzzle in more than For updated information and illustrations. Method. while examples are given for most groups. An interlocking puzzle should be classiﬁed according to its interior construction. or teddy bear may all have similar Cartesian internal construction and so should all be classiﬁed as Interlocking– Cartesian). sphere. a wooden cube.) Deﬁnitions. Consider a three-dimensional (3-D) interlocking assembly in the form of a cage with a ball in the center.. 175 . go to http://puzzlemuseum. rather than its outward appearance (e.com. Many people have provided a great deal of help. A puzzle is a problem having one or more speciﬁc objectives.. The disassembly and/or reassembly of the cage remains the primary function of the puzzle. “Mechanical Puzzles” is the descriptive term used for what are also known as “Chinese Puzzles. some knowledge of the subject is required. Objective. but particular thanks are due to Stanley Isaacs. David Singmaster. The fact that the instructions request the would-be solver to “remove the ball” does not change the 3-D assembly into an opening puzzle. related to such objects.g. books.Classiﬁcation of Mechanical Puzzles and Physical Objects Related to Puzzles James Dalgety and Edward Hordern Background. and Jerry Slocum.” Several attempts have been made to classify mechanical puzzles.

” which. if the route has been found. The fourteen main classes are as follows: ¯ ¯ ¯ Dexterity Puzzles (DEX) require the use of manual or other physical skills in their solution. the exercise of whose ingenuity is entirely incidental to the original warlike intent of the sport. it also excludes the archer attempting to get a bull’s-eye. For example. while perhaps puzzling the infant. The pieces have L-shaped grooves. crossword puzzles). which may be ﬂexible.. usually. thickness of paper or plywood or an operation involving a third dimension such as folding). unlike the parts of an interlocking puzzle. It is based on a ¿ ¢ ¿ sliding block puzzle under a clear plastic top.g. plus up to four characters. however. unless they are on or part of some physical object. then the dexterity and sequential movement must also have been achieved. The full abbreviations consist of three characters. Tanglement Puzzles (TNG) have parts that must be linked or unlinked. Most standard jigsaws are 2-D. A puzzle will be referred to as two-dimensional (2-D) if its third dimension is irrelevant (e.. A few puzzles may have to be cross-referenced if it is absolutely necessary. so they must be classed as 3-D. and a ball must be rolled up a ramp in the lower right onto one of the blocks — the ball must be moved from block to block. Tanglement–Rigid & Semi-Rigid is awaiting a thorough study of the topology of wire puzzles. have signiﬁcant freedom of movement in relation to each other. Routeﬁnding Puzzles (RTF) require the solver to ﬁnd either any path or a speciﬁc path as deﬁned by certain rules. and Route-ﬁnding.176 J. although jigsaws with sloping cuts in fact have a relevant third dimension. It is understood that specialist collectors will further subdivide the subclasses to suit their own specialized needs. it must be classiﬁed in whichever is the most signiﬁcant category. Thus the puzzle requires Dexterity. was contrived only to educate and amuse. It will be noted that the deﬁnition of a puzzle excludes the infant’s “posting box. . It would be classiﬁed as Route-ﬁnding because.” These are the standard abbreviations for the classes that have been chosen for relative ease of memory and conformity with most computer databases. Also excluded are puzzles that only require paper and pencil (e. one category will be dominant. and the blocks themselves must be slid around so that the ball can exit at the top left. DALGETY AND E. HORDERN one category. Sequential movement. A good example of multiple-class puzzles is the “Mazy Ball Game” made in Taiwan in the 1990s. such as “INT-BOX. The linked parts. hyphen.g.

They may clip together but do not interlock in 3-D. nor do they involve general assembly/disassembly of parts that interlock in 3-D. Vessels having a mechanical puzzle or trick in their construction that affects the ﬁlling. magnetic poles. Measuring. etc. and the principle objective is to restore them to their unique original form. and Theoretical Puzzles. undo it. The mechanism of the puzzle is not usually apparent. Folding and Hinged Puzzles (FOL) have parts that are joined together and usually do not come apart. or folding. i.. etc. or a nut and bolt. Logic. a padlock and its hasp. one or more pieces hold the rest together. They are solved by hinging. Puzzles in which something appears impossible or ambiguous. Mystery. Trick. Other Types of Mechanical Puzzles and Objects (OTH). Pattern Puzzles (PAT) require the placing or arrangement of separate pieces of a similar nature to complete patterns according to deﬁned rules.e. Jugs and Vessels (JUG). close it.” Assembly Puzzles (Non-Interlocking) (ASS) require the arrangement of separate pieces to make speciﬁc shapes without regard to the sequence of that placing.CLASSIFICATION OF PUZZLES 177 ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ Opening Puzzles (OPN) are puzzles in which the principle object is to open it. the differences must be sufﬁciently minor so as not to obscure the similarity of the pieces. They usually comprise a single object or associated parts such as a box with its lid. Interlocking Puzzles (INT) interlock in three dimensions. or drinking therefrom. Jigsaw Puzzles (JIG) are made from cut or stamped-out pieces from a single complete object. or otherwise get it to work. provision is made for puzzles pending classiﬁcation. remove something from it. Math. Where the pattern is due to differences in shape. shape. The pattern may be color. . Many clip-together puzzles are “non-interlocking. or the pieces are mutually selfsustaining. ﬂexing. Sequential Movement Puzzles (SEQ) can be solved only by moves that can be seen to be dependent on previously made moves. Cutting. Some have a container and are posed as packing problems. This group is for puzzle objects that do not easily fall into the above categories and cannot be categorized into sufﬁciently large groups to warrant their own major class. Also. faces of a cube. Included in this group are Balancing. texture. pouring. The pattern required may be the matching of edges of squares. Ambiguous Pictures and Puzzling Objects (AMB).

or partly used in solution. 2-D to 3-D. where XXX is the main class and YYYY is the sub-class. 2-D on 3-D. Number of Dimensions. A detailed classiﬁcation with second-level classes is given in the separate table. ribbons. Required for INT ASS PAT. the type of structure. triangular. Examples of approved structure adjectives that may be used include ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ Identical: All pieces identical Cartesian: Three mutually perpendicular axes Diagonal: Pieces rotated 45 degrees along their axis Skewed: Like a squashed puzzle Polyhedral Geometrical: Non-Cartesian but geometrical structures Organic: Amorphously shaped pieces Ball: Pieces made from joined spheres Rod: Square. not needed. strings. Thus a standard six-piece burr uses square rods in a Cartesian structure. Linked: Pieces joined together by hinges. and 4-D. Number of Pieces. We are now working on deﬁning what should be included automatically in the subclasses where they are relevant. Type of Piece Structure.178 J. This has resulted in an unwieldy list. Whether needed. and any group/non-group moves. which. 3-D. and the standard Stellated Rhombic Dodecahedron has a six-piece Diagonal Cartesian structure. Examples of puzzles in each class are given in the right-hand column. while not strictly puzzles. We hope to reduce the number of sub-classes substantially in the near future. Puzzle Class Abbreviations (PZCODE) are standardized to a maximum of eight characters: XXX-YYYY. Required for SEQ. . Required for INT JIG ASS PAT RTF SEQ FOL. etc. 2-D. Group Moves. etc. ¯ Proposed Developments Prior to 1998 the subclasses attempted to incorporate the number of dimensions. need to be classiﬁed as part of the collection. DALGETY AND E. 2-D. 3-D. This category has been included because most puzzle collections include related ephemera. Required for INT JIG ASS PAT. hexagonal. HORDERN Ephemera (EPH).

Generalized Information ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ Generic name of puzzle or objective if not obvious.. solar Patent and markings Notes. Good. **A required for scale What the dimension refers to.CLASSIFICATION OF PUZZLES 179 Guide to Making a Catalogue or Database for Puzzle Collections Headings for cataloguing puzzle collections could include the following items.e. i.. it is probably advisable to be selective and limit the amount of information recorded. but otherwise be in excellent condition. **required Class + Subclass **required Theme or advertisement (subject) Materials Dimensions A. designer Manufacturer or publisher’s name and country Type of manufacturing. i.e. d = diameter. mass produced (over 5000 pieces made). In practice. Poor) Condition qualifying note. if known Number of set in collection A picture or photograph Location/cabinet/bin Acquisition Number Condition (Excellent.. Source Date Cost Insurance value Information Speciﬁc to This Object Information Speciﬁc to This Collection . i. clockwork. electric. C. references. battery. puzzle may have a cracked glass top or be missing one piece. biggest piece..e. assembled puzzle If powered. box.e. homemade or tribal Year of manufacture **required Country of manufacture if different from publisher’s country Manufacturer’s series name Manufacturer’s product name Number in complete set. B. envelope. without paid curators. craft made (commercial but small volume). i. Fair.

visiting places en route.” “Yankee. jumping beans.” “Theo der Turnier. etc.” Tomy’s “Crazy Maze. plain balls into holes Dexterity with sundry objects and/or obstacles Liquid objects Dexterity in liquid Indirect viewing Mechanized Using tools and magnetic tools Route following dexterity Objects concealed from view Electrica and electronic dexterities Pinball-related dexterities Other dexterities.” colour mazes. number totalling mazes. “Pike’s Peak” Icosian Game. Most hedge mazes RTF-2D Route path) mazes 2-D (any . “Le Pendu.” “Tandem Maze” (complex).” Engel’s “Black Box” Bagatelle “Frying Pan. DALGETY AND E.180 J. “Bootlegger” (complex) Ring and hole mazes.” puzzles using tops Pentangle “Roly-Poly” puzzles Ramps. HORDERN Detailed Puzzle Classiﬁcation Class DEXDEX-UNCA Description Dexterity Puzzles Dexterity or other physical skills in their solution Example Cup and Ball. Mercury manipulation Water-ﬁlled puzzles View by mirror Tomy’s “Pocketeers” DEX-BALL DEX-SDRY DEX-LQOB DEX-INLQ DEX-MIRR DEX-MECH DEX-TOOL DEX-RTFL DEX-HIDD DEX-ELEC DEX-PINB DEX-OTHR RTFRTF-CHNG RTF-STEP RTF-UNIC RTF-SHOR RTF-CPLX Dexterity. bridges. pneumatic operation Route-Finding Puzzles Route-ﬁnding with changing path and/or Complex Traveller Route-ﬁnding step mazes Unicursal route-ﬁnding Shortest route Complex route mazes with special objectives Four Generations “Ball in Block. avoiding objects. Konigsburg Bridges “Worried Woodworm.

checkerboards. “Hoffman” cube.” “Margot Cube. pens. Dalgety’s “Devil’s Halo” Leather tanglement puzzles Boxes. Pentominoes. poison rings Nuts and bolts. knives. Oscar’s “Dovetail.CLASSIFICATION OF PUZZLES RTF-2D3D RTF-3D Route mazes 2-D on 3-D surfaces (any path) Route mazes 3-D (any path) Maze on surface of cube Some hedge mazes. puzzle rings Hess wire puzzles. “Mayer’s Cube.” keychain cars “Plato’s Plight” OPN-OTHR INTINT-BOX INT-CART INT-POLY INT-SHAP Interlocking Puzzles Boxes that disassemble Cartesian (has parts along three mutually perpendicular axes) Interlocking polyhedral Other rigid shapes INT-TENS ASSASS-MAT ASS-2D ASS-CART Tensegrity structures in which compression and tension elements separate Assembly Puzzles Match stick puzzles and tricks 2-D Assembly 3-D Cartesian Assemblies Tangram. “T” puzzle Polycubes. cast “ABC. ball in cube of cubelets 181 ¢ TNGTNG-RIGI TNG-R&F TNG-FLEX OPNOPN-BOX OPN-LOCK OPN-HID Tanglement Puzzles Tanglement of rigid and semirigid parts Tanglement of rigid and ﬂexible parts All ﬂexible Opening Puzzles Opening containers Opening locks Opening/ﬁnding hidden compartments not originally designed as puzzles Opening other objects Wire puzzles. Oskar’s keys. purses Padlocks Chippendale tea chests.” “Nine of Swords. O’Beirne’s “Melting Block” .” Chinese rings. “Tak-it-Apart.” Cutler’s burr in a glass Cofﬁn’s “Saturn” JWIP and keychain animals. gears. cutlery. Soma.” “Hazelgrove Box” Strijbos aluminium burr box Burrs. berrocals.

” Gannt’s “Chandy” Can include puzzles double-sided JIGJIG-STD JIG-3D JIG-2DID Jigsaw Puzzles Standard jigsaws 3-D jigsaws 2-D jigsaws with identical pieces for secondary objectives 2-D jigsaws which only partially cover the plane 2-D jigsaws with nonperpendicular/sloping cuts Multiple-layer 2-D jigsaws 2-D jigsaws with parts that can be made into 3-D objects Picture cubes/blocks Pattern Puzzles Pattern arrangements of points. number puzzles Heads and tails “Testa” PAT-STIX PAT-NUM PAT-2DEG PAT-2D PAT-STAK PAT-2D3D Stacking transparent layers. Josephus.” Loyd’s Donkeys. “Lapin. Jensen’s “Tricky Laberint” Magic squares. and weaving 2-D patterns Patterns with 2-D parts on 3-D surface “Shmuzzles” tesselations JIG-PART JIG-SLOP JIG-LAYR JIG-2D3D JIG-BLOX PATPAT-2DPG Bilhourd’s jigsaws “Sculpture Puzzles” “Toyznet” Queens on chess board.” “Phoney Baloney. or pieces. nine-piece ivory cube Pack the Plums.182 ASS-POLY ASS-SHAP 3-D Assembly Polyhedra and Spheres 3D Assembly Other Shapes J. DALGETY AND E. according to predetermined rules Patterns of sticks Pattern: arrangements of number 2-D matching edge patterns Arrangement of 2-D pieces on 2-D surface to make a pattern according to predetermined rules Stacking. Apple and Worms. “Managon. Waddington’s “Black Box” Match puzzles. Squashed Soma. weaving puzzles “Dodeca” .” “Even Steven. overlapping. pegs. HORDERN Ball pyramids. BlackBox.

Fit Pz “Wolf.” “Backspin.” Chinese balls in ball. Tomy’s “Great Gears” Raba’s “Rotascope. Oscar’s “Solar System. Tit-Bit’s “Teasers” “Inversions” PAT-LINK Linked part pattern puzzles SEQSEQ-PLAC SEQ-RIVR SEQ-HOPP Sequential Puzzles Sequential placement Sequential river crossing Sequential hopping and jumping Sequential sliding and shunting in 2-D Sliding and shunting in 3-D (only single piece moves needed) Sliding and shunting with mechanical or rotating parts (single piece moves and group moves needed) Sequential rotating in 2-D (group moves only) Sequential rotating and/or mechanical in 3-D (group moves only) Sequential rolling Sequential miscellaneous mechanical Folding Puzzles Folding single-part puzzles Folding hinged parts in open loop Hinged parts in closed loop Folding hinged parts that separate SEQ-SL2D SEQ-SL3D SEQ-SLRO SEQ-RT2D SEQ-RT3D SEQ-ROLL SEQ-MMEC FOLFOL-SING FOL-HGOP FOL-HGCL FOL-HSEP “Tower of Babel.” Skor Mor “Instant Indecision. Rubik’s “Magic.” Laker Cube Bognar’s planets.CLASSIFICATION OF PUZZLES PAT-3D 3-D pattern puzzles with separate parts 183 “Instant Insanity.” “Jugo Flower. Panel Nine.” “Hexadecimal.” Mobius strips Rubik’s “Snake.” “Flexicube” “Clinch Cube” . “Dodecahedru” — rotating faces of Dodecahedron Psychic Pz.” “Turntable Train”.” “Kaos” Rolling eight cubes “The Brain..” Rubik’s “Clock” Rubik’s Cube.” Waddington’s “Kolor Kraze.” “Spin Out” “Why Knots.” “Orbit. counter and peg moving puzzles 15s Pz.” “Missing Link. Tower of Hanoi. “Masterball. Sheep and Cabbage” Solitaire.” strung cubes Flexagons.

HORDERN Map folding. “Jail Nixon” Strip polyhedra Standard “block holes and suck” solution “Jolly Jugs” Cadogan teapots. DALGETY AND E. 12 Golf Balls.184 FOL-SH2D FOL-SH3D JUGJUG-STD JUG-CPLX JUG-BASE JUG-NLID JUG-OTHR OTHOTH-ELEC OTH-BAL OTH-MEAS OTH-CUT OTH-RIDD OTH-WORD Folding sheets and strips into 2-D solution Folding sheets and strips into 3-D shapes Puzzle Jugs Puzzle mugs standard Complex mugs requiring special manipulation Pour from base Lidless wine jugs Self-pouring and other patents Other Types of Puzzle Electrical and electronic (non-dexterous) Balancing (non-dexterity) Measuring and weighing puzzles Cutting puzzles Riddles Word puzzles J. crosswords on bathroom tissue OTH-MATH OTH-LOGI OTH-TRIK OTH-MAGI OTH-MYST OTH-SET OTH-THEO Mathematical puzzles (excluding number pattern arrangements) Logic puzzles Trick or catch puzzles (solution needs subterfuge) Conjuring tricks presented as puzzles Objects whose function or material is a mystery Sets of puzzles of mixed type Puzzles whose existence is only theoretically possible Cartoon pictures to arrange in order “Infernal Bottle” Disappearing coin slide box Creteco spacers such as 4-D puzzles. Archimedes gold Cork for three holes. or those that can only be represented on a computer . Chinese winepots Royale’s patent Luminations “Columbus Egg” Jugs and liquids. rebus plates and prints. ﬁve square puzzle 19th century riddle prints Anagrams.

“Spot the Difference.” “Find the 5th Pig” needing coloured overlays.” Hooper’s paradox Anamorphic pictures Pictures and objects of one subject made from completely unrelated objects Devinettes (obscure outlines). Oskar’s “Escher Puzzle.” Penrose triangle “Vanishing Leprechauns. impossible dovetails. weight illusions AMB-VANI AMB-DIST AMB-ARCH Distortions Archimboldesque objects AMB-HIDD Hidden image pictures (no manipulation required) Hidden image pictures (manipulation required) AMB-HMAN AMB-TURN Pictures that require turning to show different images Perception illusions Ephemera related to puzzles Wiggle Woggle HTL Micro Printing Moire effect Hold to light Heiroglyphs (non-rebus) Anaglyphs AMB-ILLU EPHEPH-WIWO EPH-MICR EPH-MOIR EPH-HTL EPH-HEIR EPH-ANAG EPH-STRP Strip prints (three views in one frame) Hold to light cards producing movements by shadows Images concealed by extreme smallness Puzzles/effects produced by moire/fringe effects Protean views.” random dot stereograms “Naughty Butterﬂies. hold to light advertisements Obscure non-rebus heiroglyphic prints Requiring red and green glasses for either 3-D or movement effects Framed strip prints showing different views from different directions . courtship/matrimony Optical illusions. OHOs. soot on unglazed part of ashtray Landscape turned to make a portrait. Topsey Turveys.CLASSIFICATION OF PUZZLES OTH-PEND AMBAMB-POBJ Pending Classiﬁcation !!! Ambiguous Paradoxical objects (objects that apparently cannot be made) Vanishing images 185 Puzzles awaiting classiﬁcation Arrow through bottle.

186 EPH-OTHR XXX-XXX DEL Other puzzle-related ephemera Lost records Deleted record J. no longer in collection . HORDERN For database integrity only Puzzle disposed of. DALGETY AND E.

one of which is twice as great as the other..A Curious Paradox Raymond Smullyan Consider two positive integers Ü and Ý. is equal to the excess of y over x. and the excess of Ý over Ü. if Ý is greater than Ü. Ä Ø Ø «ÖÒ ØÛ Ò Ü Ò Ý ÓÖ Û Ø × Ø × Ñ Ø Ò ¸ Ø Ð ×× Ö Ó Ø ØÛÓº Ì Ò Ó Ú ÓÙ×ÐÝ Ø Ü ×× Ó Ü ÓÚ Ö Ý¸ Ü × Ö Ø Ö Ø Ò Ý¸ × ¸ Ò Ø Ü ×× Ó Ý ÓÚ Ö Ü¸ Ý × Ö Ø Ö Ø Ò Ü¸ × Ò º ËÒ ¸ ÈÖÓÔÓ× Ø ÓÒ ¾ × ×Ø Ð × Proof of Proposition 1. Proposition ½.e. The excess of x over y. suppose Ý. if x is greater than y. is certainly 50 (since Ü is then 50). is 100. Now. Then the excess of Ü over Ý. Propositions 1 and 2 can’t both be true! Which of the two propositions do you actually believe? Most people seem to opt for Proposition 2. ËÙÔÔÓ× Ü × Ö Ø Ö Ø Ò Ýº Ì Ò Ü ¾Ý¸ Ò Ø Ü ×× Ó Ü ÓÚ Ö Ý × Ø Ò Ýº Ì Ù× Ø Ü ×× Ó Ü ÓÚ Ö Ý¸ Ü × Ö Ø Ö Ø Ò Ý¸ × Ýº ÆÓÛ¸ ×ÙÔÔÓ× Ý × Ö Ø Ö Ø Ò Üº Ì Ò Ü ½ Ý¸ Ò Ø Ü ×× ¾ Ó Ý ÓÚ Ö Ü × Ø Ò Ý ½ Ý ½ Ýº Ì Ù× Ø Ü ×× Ó Ý ÓÚ Ö Ü¸ Ý × Ö Ø Ö ¾ ¾ ½ ½ Ø Ò Ü¸ × ¾ Ýº Ë Ò Ý × Ö Ø Ö Ø Ò ¾ Ý¸ Ø × ÔÖÓÚ × Ø Ø Ø Ü ×× Ó Ü ÓÚ Ö Ý¸ Ü × Ö Ø Ö Ø Ò Ý¸ × Ö Ø Ö Ø Ò Ø Ü ×× Ó Ý ÓÚ Ö Ü¸ Ý × Ö Ø Ö Ø Ò Üº Ì Ù× ÈÖÓÔÓ× Ø ÓÒ ½ × ×Ø Ð × º Proof of Proposition 2. if Ü is greater than Ý. We are not told whether it is Ü or Ý that is the greater of the two. is certainly 100. Proposition ¾. And isn’t 100 surely greater than 50? 189 . But look. if y is greater than x. is greater than the excess of y over x. if y is greater than x). I will now prove the following two obviously incompatible propositions. The two amounts are really the same (i. say. if x is greater than y. the excess of x over y.

demonstrated the applicability of formal mathematical reasoning to real-life situations with such incontrovertible rigor as evidenced in his syllogism: All Scotsmen are canny. Reprinted from Mathematics Magazine. with special emphasis on obtaining conclusions having political or moral signiﬁcance. Therefore. no incomplete investigation is unbiased. 191 . Theorem 1. p. could be adduced. Vol. All rights reserved. Apathetic people are not human beings. closely following Dodgson’s original mod-el. Every incomplete investigation is a partial investigation. Numerous additional examples. Every unbiased investigation is an impartial investigation. For example. No. our next objective is to broaden the approach to encompass other models of mathematical proof. nor does it shed much light on situations that we are likely to encounter on a daily basis. All apathetic people are indifferent. 383. Reprinted with permission. The purpose of this note is to exploit this powerful proof methodology. introduced by Dodgson. All incomplete investigations are biased. Proof. However. December 1994. Golomb The eminent Oxford don. the conclusion (that no Scotsmen are dragons) is not particularly surprising. 67. no Scotsmen are dragons. Charles Lutwidge Dodgson. While his logic is impeccable. All human beings are different. All dragons are uncanny. Therefore. Proof. 5. to a broader range of human experience. no apathetic people are human beings. Theorem 2. Therefore.A Powerful Procedure for Proving Practical Propositions Solomon W.

GOLOMB it is a well-established principle that a property P is true for all members of a set S if it can be shown to be true for an arbitrary member of the set S. is left as an exercise for the reader. it is sufﬁcient to prove it for an arbitrary government. Proof. W. it is obviously unjust. and extending the method to other models of mathematical proof. it is true for all governments. And since this is true for an arbitrary government.192 S. Theorem 3. Finding further theorems of this type. All governments are unjust. We exploit this to obtain the following important result. To prove the assertion for all governments. . If a government is arbitrary.

a task that last misﬁres for N ¯ ¯ ¯ ¯ ¯ ¯ can be done for at least one earlier Ò (½ cannot be done for Ò Æ . meaning simply that the task can be performed for all higher integers. involve an equation with poles or zeros in just the right places.. not from a generative construct such as “divide Ò things into groups of 17 and 5. 6. e. divide it into n non-overlapping subsquares). for example. I thought. but the bottom of my basket ended up barely covered and.e. There are mathematical tasks like that: Parameterized in n. where 5 is “missing” from the sequence. after some last “misﬁring” value. to be sporting.. each can be accomplished for some hit-and-miss pattern of integer values until. Martin Gardner gave me some help. When a cold engine starts. This can in fact be performed for n = 4. One such task is to partition a square into n subsquares (i. 8. others scraped together a token from here and there. and. it must be positively stated. it often misﬁres a few times before running smoothly. using all of its area. 7. I browsed. except for the accumulated webs and dust. not for example “prove the impossibility of ”. A devilishly difﬁcult kind of mathematical problem. it must at least seem to have originated innocently. visiting non-innumeretic friends asking for contributions of a very particular kind. things run smoothly.” 193 . not as much as I expected. etc. At this point.g. I pass it around once more for contributions. the task must not be designed in an ad hoc way with the desired answer built in — must not. More precisely. might be to pose such a question in reverse. or for n = 9?. I pilfered. Ò Æ ). so it remains. can be done for all Ò Æ .Misﬁring Tasks Ken Knowlton Years ago I went on a collecting trip. What is a task parameterized in n that last misﬁres for n = 6?. before throwing it out.

) Cut a square into Ò acute triangles with clean topology: no triangle’s vertex may lie along the side of another triangle.194 K. Possible for 8. but the stated answer implicitly but erroneously suggests that the task can be performed for any Ò). these are attrociously imprecise speciﬁcations. 5. well known. (Impossible for n = 2. Cut a cube into n subcubes (called the “Hadwiger” problem. KNOWLTON For a mathematician or logician. 11.) What we have here is an inﬁnite number of problems of the form “Here’s the answer. the next have 8. and who did. edges. and 77. One of Martin Gardner’s books contains essentially this problem as a wire-identiﬁcation task. Some of these are well known. But a misﬁre problem. with a rather obvious proof: Partition a square into n subsquares. my guess is that the matter lies somewhere on the non-crispness scale between agreement on “proof” and agreement on “elegance. The tetrahedron has 6 edges. and who then went on to prove that 77 is the largest integer that cannot be so partitioned. ﬁnally clinched independently by Doris Rychener and A. its members do exhibit a special charm. others quite esoteric: ´ µ ´ µ ´ µ ´ µ ´ µ ´ µ ´ µ Stated above. ask themselves “Why is 9 missing?”) Place n counters on an inﬁnite chessboard such that each pair exhibits different numbers of row-mates and/or different numbers of column-mates. Zbinder who demonstrated a dissection into 54 subcubes). even after developing a complete proof. Partition n into distinct positive integers whose reciprocals sum to one. what’s the question?” Usually such a setup is too wide open to be interesting. (Charles Cassidy and Graham Lord. has such a severely constrained answer that it is almost impossibly difﬁcult. The trouble is that a misﬁre problem is . of course. 7. for me at least. a sociological rather than mathematical question. I have answers only for N = 5. lies the intriguing question: Will we agree that one or another task statement lies within the spirit of the game? This is. (Ron Graham is the only person I know who would ¾ ¿ think of such a problem. 12 . proved by Ron Graham.” Meager as the current collection is. or 9. 10. 47.g. 6 since 2 + 3 + 6 = 11 and ½ · ½ · ½ ½. e. 9. tracked for years by Martin Gardner.. But here. such a partition for 11 is 2. 9. 3. (Generally known. But “difﬁcult” may be the wrong word. Design a polyhedron with Ò edges. 6. conjectured by Ken Knowlton. Divide a rectangle into n disjoint subrectangles without creating a composite rectangle except for the whole (Frank Sinden). as I think I have deﬁned it.

Instead of searching directly for a task last misﬁring. Or arguments as to why this is too mushily stated a challenge. except to review all the math and geometry that you already know. it’s much more likely that in the course of your mathematical ramblings you will someday bump into one. say. I invite further contributions to the collection. for Ò ½¿.MISFIRING TASKS 195 a math problem with no implied search process whatever. .

Drawing de Bruijn Graphs Herbert Taylor The simplest kind of de Bruijn graph. When the number of nodes is ¾ Ò we can label them with the integers from ¼ to ¾ Ò ½ and notice that they correspond to all the numbers expressible with Ò bits (Ò binary digits). If ¾Ü or ¾Ü ·½ is bigger than ¾Ò ½. at each node. just subtract ¾ Ò from it to bring the number back into the range from ¼ to ¾ Ò ½. . has. two edges directed out and two edges directed in. Figure 1. Out from the node labeled Ü the two directed edges go to the nodes labeled ¾Ü and ¾Ü ·½. in which the number of nodes is a power of two. Ò 197 .

The picture for Ò suggests 2-pire fashion½ for drawing the de Bruijn graph on the sphere without any lines crossing each other. Vol.198 H. the numbers from ½ to ¾ Ò ¾ in order going south down the Greenwich Meridian. . and down the international date line. because Hal Fredricksen told me he had not seen it in any book. 242. 14–22. The earliest reference I know of to the m-pire (empire) problem is P. while edges out from ¾ Ò ½ to ¾Ò ½ go west. Ò . The scheme would be to put ¼ on the North pole. 2. pp. ¾Ò ½ on the South pole. Heawood’s 1891 paper in the Quarterly Journal of Mathematics. available from Aegean Park Press. “Mathematical Games. Golomb. TAYLOR There is reason to think that the picture for Ò may ﬁll a gap in the existing literature. Pictures up to Ò can be found in the classic “Shift Register Sequences” by Solomon W. J. See Scientiﬁc American. Bigger de Bruijn graphs get really hairy. Edges out from ¼ to ¾ Ò ½ ½ on the meridian go east to the dateline. No. ½In 2-pire fashion each node of the graph is allowed to appear in two places in the picture. Figure 2.” by Martin Gardner. February 1980.

A single existing spot may serve as both endpoints of a curve. A player who cannot make a legal move loses. A sample game of two-spot Sprouts. a spot may have a maximum of three parts of curves connecting to it. either the ﬁrst player or the second player can always force a win. adding a new spot on each curve drawn.Computer Analysis of Sprouts David Applegate. The initial position of the game consists of a number of points called spots. Players alternate connecting the spots by drawing curves between them. a professor of mathematics. regardless of the opponent’s strategy. Furthermore. Guy Jacobson. It was invented in Cambridge in 1967 by Michael Patterson. Impartial games have variants where the condition of victory is inverted: The winner is the Figure 1. and the last player to make a legal move wins. and Daniel Sleator Sprouts is a popular (at least in academic circles) two-person pen-and-paper game. Which of the players has this winning strategy depends on the number of initial spots. with the ﬁrst player winning. Sprouts is an impartial game: The same set of moves is available to both players. 199 . and John Horton Conway. Since draws are not possible. Shown below is a sample game of two-spot Sprouts. then a graduate student. Most people (including us) learned about this game from Martin Gardner’s “Mathematical Games” column in the July 1967 issue of Scientiﬁc American. Each curve must be drawn on the paper without touching itself or any other curve or spot (except at end points).

our program is the only successful automated Sprouts searcher in existence. AND D. n. and that the analysis of eight-spot Sprouts was far beyond the reach of present-day (1967!) computers. Many sprouts positions that occur during the search are the sum of two or more non-interacting games. APPLEGATE.and spaceefﬁciency to solve the larger games. but as the number of spots increases. While Sprouts has very simple rules.200 D. Sometimes it is possible to infer the value of the sum of two games given the values of the subgames. JACOBSON. we managed to achieve sufﬁcient time. and in misère games of up to nine spots. This is the principal reason that we are able to extend the analysis of the normal game further than the misère game. Of course. caching the results . performed by Denis Mollison (to win a 10-shilling bet!). We have written a program that determines which player has a winning strategy in games of up to eleven spots. This is called the losing or misère version of the game. positions can become fantastically complicated as the number of spots. These ideas are not nearly as useful in analyzing misère Sprouts. Our Sprouts Program As far as we know. Our program is successful for several reasons: ¯ ¯ ¯ We developed a very terse representation for Sprouts positions. After many hours of late-night hacking and experimenting with bad ideas. Our representation strives to keep only enough information for move generation. and also increases the number of moves available at each turn signiﬁcantly. computers have come a long way since then. ran to 47 pages. such as cutting off the search as soon as the value is known. We used standard techniques to speed adversary search. SLEATOR player who cannot make a legal move. Our program makes use of these sum identities when evaluating normal Sprouts. The combination of this low-information representation and hashing (whereby the results of previous searches are cached) proved to be extremely powerful. the complexity of the problem overwhelms human powers of analysis. Games with small numbers of spots can be (and have been) completely solved by hand. Each additional spot adds between two and three turns to the length of the game. grows. G. Many seemingly different Sprouts positions are really equivalent. Conway said that the analysis of seven-spot Sprouts would require a sophisticated computer program. The ﬁrst proof that the ﬁrst player loses in a six-spot game.

and an asterisk indicates a new result obtained by our program. This problem seems to lie just beyond our program. and we devised and implemented two methods to save space without losing too much time efﬁciency. Our Results Here’s what our program found: Number of Spots normal play misère play 1 2 1 2 2 2 3 1 2 4 1 2 ¾ £ ¾£ ½£ ½£ ½£ ½ £ ½£ ¾£ ¾£ ¾£ 5 1 6 2 7 8 9 10 11 A “1” means the ﬁrst player to move has a winning strategy.cmu. The n-spot Sprouts positions evaluated so far fall into a remarkably simple pattern. and our ingenuity. The ﬁrst player has a winning strategy in n-spot Sprouts if and only if n is 3. characterized by the following conjecture: Sprouts conjecture. We are still left with the nagging problem of resolving a bet between two of the authors. The only remaining case required to resolve the bet is the 10-spot misère game. The ﬁrst player has a winning strategy in n-spot misère Sprouts if and only if n is 0 or 1 modulo 5. go to http://www. 4. The data for misère Sprouts ﬁt a similar pattern.cs. or 5 modulo 6. To reduce the space still further we used a data compression technique. and he bet Sleator a six-pack of beer of the winner’s choice that one of them would fail on some game up to 10 spots. Misère sprouts conjecture.edu/~sleator . Sleator believes in the Sprouts Conjecture and the Misère Sprouts Conjecture. a “2” means the second player has a winning strategy.COMPUTER ANALYSIS OF SPROUTS 201 ¯ of previous searches in a hash table. The size of the hash table turned out to be a major limitation of the program. our computational resources.½ ½For a paper describing these results in more detail. and searching the successors of a position in order from lowest degree to highest. Applegate doesn’t believe in these conjectures. We discovered that saving only the losing positions reduced the space requirement by a large factor.

found by Hickerson (at U. But the (stationary) right end.Strange New Life Forms: Update Bill Gosper The Gathering for Gardner Atlanta International Museum of Art and Design Atlanta. infecting some of the world’s best brains and machinery. Martin Gardner. Sadly. has to be Time 0: Note that it extends leftward with velocity ½ . it would be nice to report that. and the solution of most of the questions initially posed by the game.C. But if you interrupt it. as of late 1992. of which the weirdest. Back when only the large corporations could afford computers because they cost hundreds of dollars an hour. With the great reduction in computation costs. The toll in human productivity probably exceeded the loss in computer time. alien Life forms. 203 . the Life bug no longer poses a threat. this is not the case. by singlehandedly popularizing Conway’s game of Life during the 1970s. like smallpox. the beam shortens in both directions at the speed of light. new developments in the game are still spreading through computer networks. clandestine Life programs spread like a virus. with human programmers as the vector. sabotaged the Free World’s computer industry beyond the wildest dreams of the KGB. with the left end (found by Hartmut Holzwart at the University of Stuttgart) repeating with period 4. While the Life programs distributed with personal computers are harmless toys whose infective power is 100% cancelled by TV. leading to hundreds of unnatural. Georgia Dear Gathering. The biggest shock has been Dean Hickerson’s transformation of the computer from simulation vehicle to automated seeker and explorer. and then both ends explode. Davis) has period 5! Stretching between is a beam of darts that seems to move rightward at the speed of light.

204 B. GOSPER Times 100–105: .

and myself Oscillator and Spaceship Searches.STRANGE NEW LIFE FORMS: UPDATE 205 So. David Bell. Bill Gosper. Paul Callahan. It found many oscillators of periods 3 and 4 (including some smaller than any previously known.” here is a synopsis of recent developments computer-mailed by Hickerson when the ineffable Creator of the Universe and its Laws himself asked the same question. For a few months I ran it almost constantly. and I’ve run it occasionally since then. Bill Gosper. and Hartmut searches Holzwart Construction of oscillators mostly by David Buckingham and Bob Wainwright Glider syntheses mostly by Buckingham and Mark Niemiec New glider guns by Buckingham. if you’re wondering “Whatever happened to Life?. In 1989 I wrote a program to search for oscillators and spaceships. Hickerson’s Synopsis There have been developments in various areas by different people: Oscillator and spaceship by myself. and myself Large constructions mostly by Buckingham. and a period 4 .

(These are used in some of the large patterns with unusual growth rates mentioned later. inﬁnitely many period 3. JHC!).) Periods ½¿ · Ò and ½ ¼ · Ò use a mechanism developed by Buckingham. not just lcms of smaller period oscillators acting almost independently) of oscillators of periods ½ ½ ½ ¾ ¾ ¾ ¿¼ ¿¾ ¿ ¼ ¾ ¼ ¾ ½¼¼ ½¼ ½¾ · ½¾¼Ò ½¿ · ½¾¼Ò · ¾ Ò ¾ · ¾ Ò ¼ · ¾ Ò ¾¿¼ · ¼Ò ¾ ¾ · ¿ Ò ·¿ Ò ½¿ · Ò ½ ¼· Ò and all multiples of ¿¼ and ½¼¼. speed c/2 orthogonal spaceships. based on construction universality. (Any greater nonmultiple of 4 would also work. For example. and 47 are the only ones we know with the right sparks. a few oscillators with periods 5 (including some with useful sparks) and 6. More recently. inﬁnitely many period 2. Some of the c/3s and c/4s have sparks at the back which can do various things to c/2 spaceships that catch up with them. The · ¾ Ò ¾ · ¾ Ò ¼ · ¼Ò ¾¿¼ · ¼Ò ¾ ¾ · ¿ Ò and · ¿ Ò use a device discovered by Wainwright.e. 6. which implies that all sufﬁciently large periods are possible. Another amusing oscillator uses the glider crystallization that Bill mentioned in his centinal letter a few years ago. (Mine runs only on an Apple II. which reﬂects a symmetric pair of parallel gliders. Due mostly to the work of Buckingham and Wainwright. speed c/4 orthogonal spaceship. he and Hartmut Holzwart have been producing oodles of orthogonal spaceships with speeds c/2. We also have an argument.) Lately. or 47. The turn can take either 64. so we also get glider guns of those periods. The ﬁrst glider to hit them is reﬂected . we put several copies of the B in the track to reduce the period to one of those mentioned. in which a B heptomino is forced to turn a 90 Æ corner. or 73 generations. by combining them we can build a closed loop whose length is any large multiple of 5 or 8. and c/4. of periods 5. Hartmut found another with speed 2c/5. I built a p155 oscillator in which 20 Bs travel around a track of length ¿½¼¼´ ¾¼ ¢ ·¾ ¢ µ generations.206 B. GOSPER resembling your initials. we now have nontrivial examples (i. the 73 gen turn can emit a glider. For the multiples of 8. it consists of a still life (“eater3”) and two spark producing oscillators.) Construction of Oscillators.) The multiples of ¿¼ and ½¼¼ and the · ½¾¼Ò and ½¿ · ½¾¼Ò are based on gliders shuttling back and forth between either oscillators (or periods ½ ¿¼ and ½¼¼) or output streams of glider guns (of periods ¿¼ and ¼¼ · ¾¼¼Ò). 6. speed c/3 orthogonal spaceships. but 5. one period 4. speed 2c/5 orthogonal spaceship. one period 4. c/3.. 65. and one period 5. (No doubt you ﬁgured that out for yourself long ago. Also. David Bell wrote a similar program that runs on faster machines. A period 150 gun ﬁres toward a distant pair of pentadecathlons. speed c/4 diagonal spaceship (not the glider).

The largest by far that’s actually been built is Bill’s period 1100 gun (extensible to ¼¼ · ¾¼¼Ò). Buckingham has also found extremely weird and elegant guns of periods 144 and 216.) It’s also possible to build a gun that produces a glider stream of any period ½ . based on the period 100 centinal. Subsequent gliders grow a crystal upstream. forming gliders that stabilize the ends of a row. Here’s how to turn 2 MWSSs into a glider: In Winning Ways. I’ll just add that some of Buckingham’s syntheses of still lifes and billiard table oscillators are aweinspiring. The gun itself has a larger period.STRANGE NEW LIFE FORMS: UPDATE 207 180Æ and collides with the second to form a honey farm. A honey farm starts to form but is modiﬁed by an eater and a block. New Glider Guns. Rich has already mentioned this. Fortunately.112. If you delete the trafﬁc light. with a total of 36 AK47s. it’s based on the “AK47” reaction discovered independently by Dave Buckingham and Rich Schroeppel. When they’re all gone. they don’t work with period 30 streams. there are other ways to get any multiple of 30. two gliders delete one pair of beehives. and then starts forming another honey farm in a different location. centinal-based ¼¼ · ¾¼¼Ò. A close pair of AK47s can delete each other’s trafﬁc lights. The period 94 gun is 143 by 607. emit gliders that can crash to form MWSSs. Glider Syntheses. I’ve found some eaterassisted kickbacks that give any even period. there’s now a fairly small period 44 gun that Buckingham built recently. Using normal kickbacks. you (JHC) described how to thin a glider stream by kicking a glider back and forth between two streams. the resulting period must be a multiple of 8. (Sadly.584 by 12. its bounding box is 13. For example. and it begins to decay. which hit eaters. The period ½¿ · Ò guns based on Buckingham’s B heptomino turns are also smaller. forms a trafﬁc light. It’s fairly easy to get any period down . Buckingham mercifully outmoded it with a comparatively tiny. It emits a glider. the cycle repeats every ¢ ¾ generations. of size 149 by 149. there’s a period ½ ´ ¢ ¿ · ¢ µ gun that’s 119 by 119 and a period ½¿ ´ ¼ ¢ ¿ · ½ ¢ µ ¾ gun that’s 309 by 277. In the period 94 gun. it uses various mechanisms to interleave larger period streams. based on a period 72 device discovered by Bob Wainwright. the process begins again. The other guns are all pretty big. an eater stops its growth. so we can build a long row of them that is unstable at both ends. When the crystal reaches the guns. In addition to the long-familiar period 30 and period 46 guns. two such rows. 11 gliders add a pair of beehives to the crystal.

and heavyweight spaceships from gliders. . 120. We can also synthesize light-. . its bounding box is 373 by 372. so we have guns that produce these as well. 300. I have a working construction to build a P14 glider stream by inserting a glider between two gliders 28 gens apart! . it escapes: There are also combinations of two period 30 guns that give fairly small guns with periods 90. . . This will make it possible to produce glider streams of any period. Universality implies the existence of Life patterns with various unusual properties. 210. based on the . (At least they seem unusual. 10. and 17. if the glider is delayed by 60 (or.208 B.” (Less than 14 is physically impossible. Here it’s shown deleting a glider. For example. this reaction leads to a period 30 LWSS gun: Large Constructions. and 360. Ò · ) generations.) Returning to Hickerson’s account: Many of the large constructions mentioned later require a period 120 gun. more generally. GOSPER to 18 this way (period 23 is especially simple). 16. The smallest known uses a period 8 oscillator (“blocker”) found by Wainwright to delete half the gliders from a period 60 stream.) New Year’s Eve Newsﬂash: Buckingham has just announced “. and Buckingham has found clever ways to get 15. 150. (I built a pseudo-period 15 gun based on his reactions. middle-.

it eventually becomes periodic. respectively.) Most of my large constructions are designed to achieve unusual population growth rates. Cells along its ﬂight path are occupied with decreasing frequency: The gap between the occupations increases by a factor of 3 or 9 each time. Using this I built a sliding block memory register. But it’s fairly easy to build a pattern for which this isn’t true: Take two glider puffers. Add a glider that travels northwest and southeast.) But some of these properties can also be achieved without universality. If you look at a ﬁnite region in a typical small pattern. Smaller breeders are now known.” versions of which were built independently by Bill and myself. and linear growth with an irrational growth rate. a signal is produced whenever a decrement operation reduces the value to 0. In addition. (Figuring out how to achieve a particular behavior makes a pleasant puzzle. I found a way to pull it one unit with two gliders and push it one unit with three gliders. Instead. such as Ø ½ ¾ Ø¿ ¾ ÐÓ ´Øµ ÐÓ ´Øµ¾ Ø ÐÓ ´Øµ and Ø ÐÓ ´Øµ¾ . and breeders. ﬁring gliders southwest and northeast.STRANGE NEW LIFE FORMS: UPDATE 209 small things we normally look at. Probably the second such pattern is the “exponential aperiodic.) The ﬁrst such pattern (other than glider guns) was Bill’s breeder. These use a glider salvo to push a block (or blinker). I’ve also built some aperiodic patterns with bounded populations. using a kickback reaction each time it hits one of the glider waves. headed east and west. although almost all large patterns grow to inﬁnity. In Winning Ways you (JHC) describe how to pull a block three units using two gliders and push it three units using 30. there are several . Bill also built an “arithmetic aperiodic” in which the gap between occupations increases in an arithmetic progression. so some of us have spent many hours putting together guns and puffers to produce various results. Even inﬁnite growth is rare for small patterns. whose population grows like Ø ¾. (It has one small difference: the “test for zero” is not a separate operation. actually building the thing is mostly tedious. similar to the one you described. puffers. and probably others. Both of these patterns have populations tending to inﬁnity. the reaction also sends back a glider (or two) which triggers the release of the next salvo. This is even true for guns. depending on which cells you’re looking at.

Following Dean’s update. Instead. Warm ones. Thankfully. we were all astounded when Achim Flammenkamp of the University of Bielefeld revealed his prior discovery of Dean’s smallest period 3 and 4 oscillators during an automated. we produce one of period ¾Æ as follows: Arrange 3 period Æ puffers so their gliders crash to form a MWSS moving in the same direction as the puffers. . building such a thing from standard glider logic is straightforward. but harder to describe. there’s already a MWSS in the way. (Of course. The next time the gliders try to crash.) —Dean Hickerson To clarify. (This could be used to get population growth Ø ¾ ÐÓ ´Øµ. whose populations are unbounded but do not tend to inﬁnity. I also built a pattern (initial population ¿¼¼¼) that computes prime numbers: An LWSS is emitted in generation ½¾¼Ò if and only if Ò is prime. the MWSS takes the place of the stable intermediary of the reaction. so I’ll just describe mine. but I haven’t built that. Buckingham’s “awe-inspiring glider syntheses” are the constructions of prescribed. His construction is more efﬁcient than mine. GOSPER “sawtooth” patterns. entirely by crashing gliders together. this happens every ¾Æ generations.210 B. which adds two period 60 input streams and produces a period 60 output stream. providing us with natural syntheses (and probably estimates) of rare objects.) Paul Callahan and I independently proved that arbitrarily large puffer periods are possible. so they can’t produce another one. months-long series of literally millions of random soup experiments. two of them destroy it and the third escapes. Give a glider puffer of period Æ . but they used some very clever ideas to do it more efﬁciently. sometimes even oscillating objects. often large and delicate. The difﬁculty is comparable to stacking water balls in 1G. he recorded the conditions that led to these (and many other) discoveries.) Dave Buckingham and Mark Niemiec built a binary serial adder. (Basically this mimics the way that a ternary reaction can double the period of a glider gun.

STRANGE NEW LIFE FORMS: UPDATE

211

ºººººººººººººººººººººÓººººººººººººººº ºººººººººººººººººººººÓºÓººººººººººººº ºººººººººººººººººººººÓÓºººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººÓºººººººººººººººººººººººººººººº ººººÓºÓººººººººººººººººººººººÓºÓººººº ºººººÓÓººººººººººººººººººººººÓÓºººººº ººººººººººººººººººººººººººººººÓºººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººÓººººººººººººººººººÓººººººº ºººººººººººÓÓººººººººººººººººÓºÓººººº ººººººººººÓÓºººººººººººººººººÓÓºººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººÓÓÓºººººººººººº ººººººººººººººººººººººÓºººººººººººººº ºººººººººººÓºººººººººººÓººººººººººººº ºººººººººººÓÓºººººººººººººÓÓÓººººÓÓºº ººººººººººÓºÓºººººººººººººÓººººººÓºÓº ºººººººººººººººººººººººººººÓºººººÓººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ÓÓººººººººººººººººººººººººººººººººººº ºÓÓºººººººººººººººººººººººººººººººººº ÓºººººººÓÓººººººººººººººººººººººººÓÓº ºººººººººÓÓºººººººººººººººººººººººÓºÓ ººººººººÓºººººººººººººººººººººººººÓºº ººººººººººººººººººººººººººººººººººººº ºººººººººººººººººººººººººººÓÓºººººººº ºººººººººººººººººººººººººººÓºÓººººººº ºººººººººººººººººººººººººººÓººººººººº ººººººººººººººººººººººººººººººººººººº ºººººººººÓÓºººººººººººººººººººººººººº ººººººººÓºÓºººººººººººººººººººººººººº ººººººººººÓºººººººººººººººººººººººººº

Conway called such oscillators billiard tables, and along with the rest of us, never imagined they could be made by colliding gliders.

ººÓÓºÓÓººººº ººÓºººÓººººº ºººÓÓÓºººººº ºººººººººººº ºººººÓÓÓºººº ººººÓºººÓºÓÓ ºÓººÓºººÓºÓÓ ÓºÓºÓºÓºÓººº ºÓÓºÓºººÓººº ºººººÓÓÓºººº ºººººººººººº ºººººººÓÓÓºº ººººººÓºººÓº ººººººÓÓºÓÓº

Finally, after a trial run of the foregoing around the Life net, Professor Harold McIntosh (of the Instituto de Ciencias at Puebla) responded: “The humor in that proposed introduction conjures up images of untold taxpayer dollars (or at least hours of computer time) disappearing into a bottomless black hole. That raises the question, ‘Have other hours of computer time been spent more proﬁtably?’ (Nowadays, you can waste computer time all night long and nobody says anything.) “Martin Gardner is a skilled presenter of ideas, and Life was an excellent idea for him to have the opportunity to present. Not to mention that the time was ideal; if all those computer hours hadn’t been around to waste, on just that level of computer, perhaps the ideas wouldn’t have prospered so well. “In spite of Professor Conway’s conjecture that almost any sufﬁciently complicated automaton is universal (maybe we can get back to that after vacations) if only its devotees pay it sufﬁcient attention, nobody has yet

212

B. GOSPER

come up with another automaton with anything like the logical intricacy of Life. So there is really something there which is worth studying. “It might be worth mentioning the intellectual quality of Gardner’s presentation of Life; of all the games, puzzles and tricks that made their appearance in his columns over the years, did anything excite nearly as much curiosity? (Well, there were ﬂexagons.) “Nor should it be overlooked that there is a more serious mathematical theory of automata, which certainly owes something to the work which has been performed on Life. Nor that there are things about Life, and other automata, that can be foreseen by the use of the theories that were stimulated by all the playing around that was done (some of it, at least).” The Life you save may not ﬁt on your disk. —Bill Gosper

ØØÔ »»ÛÛÛºÑ Ò ×ÔÖ Ò º ÓÑ» Ð Ò »Ð

Many remarkable discoveries followed this writing. See, for example,

» Ò Üº ØÑÐ

STRANGE NEW LIFE FORMS: UPDATE

213

An early “stamp collection” by Dean Hickerson. These are representative oscillators of the (small) periods (indicated by the still-Life numerals) known by the end of 1992. As of the end of 1998, all periods have been found except 19, 23, 27, 31, 37, 38, 41, 43, 49, and 53.

Hollow Mazes

M. Oskar van Deventer

In October 1983 I discovered hollow mazes. A. K. Dewdney presented the concept in Scientiﬁc American, September 1988. In this article I shall give a more detailed description of the subject. The ﬁrst part is about multiple silhouettes in general, the second part is about their application to hollow mazes.

Multiple Silhouettes

Silhouettes are often used in, e.g., mechanical engineering (working drawings) and robotics (pattern recognition). One particular mathematical problem reads: “Which solid object has a circular, a triangular, and a square silhouette?” The solution is the well-known Wedge of Wallis (Figure 1).

Figure 1. Wedge of Wallis.

A silhouette deﬁnes a cylindrical region in space, perpendicular to the silhouette surface. With this meaning the word no longer refers to an object, but only to a region of a ﬂat surface (Figure 2). Two or more silhouettes (not necessarily perpendicular to each other) deﬁne unambiguously a region or object in space that is the cross-section of two or more cylinders. Dewdney introduced the term projective cast for an object deﬁned by a number of silhouettes. Not every object can be deﬁned as the projective

213

214

M. O. V AN DEVENTER

Figure 2. (a) Negative of a silhouette; (b) The cylindrical region in space deﬁned by it.

cast of a (limited) number of silhouettes: One cannot cast a hollow cube or a solid sphere. There are two interesting topological properties that apply to silhouettes, to projective casts, and to mazes and graphs in general. First, a graph may have either cycles or no cycles. Second, a graph may consist of either one part or multiple parts (Figure 3).

Figure 3. Topological properties of mazes and graphs: (a) multiple parts with cycles; (b) one part with cycles; (c) multiple parts, no cycles; (d) one part, no cycles.

Dewdney used the term viable for a silhouette or a maze that consists of one part and has no cycles. Even when silhouettes are viable, their projective cast can still have cycles (Figure 4) or multiple parts (Figure 5). There is no simple connection between the properties of the silhouettes and the properties of their projective cast. The basic problem is the analysis

Hollow Mazes A hollow maze is a rectangular box with six sides (Figure 6). and how can a projective cast be given certain properties by its silhouettes? I do not know whether there is a general answer. Control mazes are always viable because of their construction. The resemblance between multiple silhouettes and hollow mazes is evident. sliding along the slots of the control maze on each side. since a spar cannot jump from one part to another. In this way. Control mazes cannot consist of multiple parts. Figure 5. There can be no cycles since the center of a cycle would fall out. a single three-dimensional maze is produced from three pairs of two-dimensional mazes. and synthesis of silhouettes and projective casts: How can the properties of a projective cast be found without constructing it from its silhouettes. The two control mazes on opposite sides of the box are identical. Each spar passes from one side of the box to the other.HOLLOW MAZES 215 Figure 4. Three viable silhouettes yielding a projective cast with one cycle. Three viable silhouettes yielding a projective cast with two parts. Each side is a two-dimensional control maze: a surface into which slots have been cut. . A cursor consisting of three mutually perpendicular spars registers one’s position in the hollow maze.

e.) Cut and Connect. a square or cubic grid is used for control mazes and their projective casts. their paths have zero width). The number of links can be counted from the control mazes without constructing the projective cast. the control mazes are line mazes (i.) It can easily be tested whether or not a projective cast is viable. Count the links in the hollow mazes of Figure 7. Cut the projective cast several times and link the pieces again (Figure 8). because the control mazes are line mazes. Helmut Honig. a German computer science . There are several ways to construct a hollow maze. viable or not: ¯ ¯ Trial and Error. (Working at random.. a ¿ ¢ ¿ ¢ ¿ projective cast has 27 grid points. assuming that it cannot have cycles: We count the number of grid points of the projective cast. V AN DEVENTER Figure 6. I have developed several restrictions for the investigation of hollow mazes: ¯ ¯ ¯ ¯ There are three (pairs of) perpendicular control mazes. Then there must be exactly 26 “links” between these points for a viable cast. O. It seems that the projective cast cannot have cycles anymore. A simple hollow maze. the total number of links is found after taking all cross-sections. but I lack the mathematical tools. For example. the control mazes are viable.216 M. Now we can take another look at the topological properties of the projective cast. This method guarantees that the eventual hollow maze is viable. (I’m sure that there must be a simple proof of this. The dotted cross-section of Figure 7(a) has six links through it.

viable! not viable. Test of the viability of a projective cast. Figure 8. (b) ½ · · · · ¾ · · · · ·¾ ¾ ½. ¯ Regular Control Mazes. How should identical control mazes of this kind be placed to yield a viable hollow maze? . The cut and connect method for constructing hollow mazes. showed me that not all viable hollow mazes can be constructed by this method. Figure 9 is a counterexample. student.HOLLOW MAZES 217 Figure 7. See Figure 10. (a) ¾· ¾ ½.

There is still much research to be done on multiple silhouettes and hollow mazes. Replace each point by a hollow maze (see Figure 11).218 M. A viable hollow maze that cannot be constructed by the “cut and connect” method. and perhaps some of the problems mentioned above will be solved in the future. . I keep receiving letters from readers of Scientiﬁc American. Figure 10. ¯ Recursion. Construction of a hollow maze by recursion. V AN DEVENTER Figure 9. Regular control mazes. O. Figure 11.

done to illustrate the ideas. the problem of Diophantus’ age gives us Ü · Ü Ü Ü · · · · ½¾ ¾ Ü i. In this case. the equations and their solution are relatively straightforward. and the material goes back several centuries). it is not always straightforward to determine when the problem has solutions or to ﬁnd them all or to determine the number of solutions. I discuss three examples of such diophantine recreations. The simplest are just problems of the ‘aha’ or ‘heap’ type. complicated versions like “How Old Is Ann?” appeared. which contained statements such as “Mary is twice as old as Ann was when Mary was as old as Ann is now. The ﬁrst is a simple age problem.” (See Loyd [14]) 219 . The third is the problem of Selling Different Amounts at the Same Price Yielding the Same.D. compiled around A.Some Diophantine Recreations David Singmaster There are a number of recreational problems that lead to equations to be solved. Other times the equations are straightforward. In a third class of problems. In the late nineteenth century..e. for which I give an algorithm for ﬁnding all solutions and a new simple formula forthe number of solutions. for which I have just found a reasonably simple condition for integer data to produce an integer solution. See Tropfke [20] for this and related problems. where the problem was in understanding the phrasing of the question. A Simple Age Problem Age problems have been popular since at least The Greek Anthology [16]. The second is the Ass and Mule Problem. leading to one equation in one unknown. Ü · Ü or Ü or Ü . 500 (although the date of this is uncertain. but there is an additional diophantine requirement that the data and/or the answers should be integral. Sometimes the difﬁculty is in setting up the equations. with the solution then being easy. such as exploiting the symmetry of the problem. In The Greek Anthology. but the solution requires some ingenuity.

from what I’ve said. 1824 [5]. SINGMASTER Between these are a number of types of problem. We man and wife have been. . albeit less poetic. The solution of these is easily ´¿µ ´ ½µ ´ ½µ assuming .220 D. It is easily seen that the problem leads to the equations ´¾µ where and found to be denote the ages of · and ´ · µ . version. 3.½ When ﬁrst the marriage knot was ty’d Between my wife and me. when are and integers? For example. [6]. more comprehensible. we can give a pretty simple answer: is integral if and only if ½I haven’t actually seen the book. My age was to that of my bride As three times three to three But now when ten and half ten years. I pray. 2) does not have an integral solution. is times as Thus the problem in The American Tutor’s Assistant and Bonnycastle is Problem (3. 1791. reportedly from The American Tutor’s Assistant. what were their ages on their wedding day? We can state the general form of this problem as follows. µ. after years. Now we give a diophantine aspect to our problem by asking the question: If . The type that I want to examine is illustrated by the following. A person. Since is integral if is. at the time he was married. and are integers. but after they had lived together 15 years. is now times as old as . What were our ages when we wed? A more typical. the problem is quoted in Bunt. appears at about the same time in Bonnycastle. he was only twice as old. Problem (4. et al. which is the earliest I have found. Now tell. but with the same numbers. As eight is to sixteen. was 3 times as old as his wife. Problem ´ old as . Her age to mine exactly bears. 2). . 15.

you were 5 times her age. I will have as many as you. I’d have . Ass and Mule Problems The classical ass and mule problem has the two animals carrying sacks. you were 3 times her age. as in the following problem: My daughter Jessica is 16 and very conscious of her age. did you ever consider when I would be half as old as Helen?” Now it was my turn to be worried. she said to me. six years ago. Can you help me out? *** It is also possible to ﬁnd a situation in which one person’s age is an integral multiple of another’s during six consecutive years. one can still get unexpected results. This pretty much settles the problem. as she stomped off to school. four years ago. and went off muttering. soon you’ll be the same age!” Jessica seemed a bit worried. that’s just the limit! By the way.” said Jessica. with ½. We pick arbitrary integers and .” The ass responds: “If you give me one of your sacks. Nonetheless. you’re always older than Helen. I saw her doing a lot of scribbling. The mule says to the ass: “If you give me one of your sacks. you were 9 times as old as Helen.” How many sacks did they each have? The general version of the problem for two individuals can be denoted ´ µ for the situation where the ﬁrst says: “If I had from you.Our neighbour Helen is just 8. then let be any multiple of ´ µ . “Dad. ´ µ This gives us the kind of condition that allows us to construct all solutions. and I teased Jessica by saying “Seven years ago. I will have twice as many as you. and now you are only twice her age.SOME DIOPHANTINE RECREATIONS 221 ´ µ If we let denoted as divides ´ ½µ be the greatest common divisor (GCD) of ´ ½µ. then (3) holds if and only if divides and ½. If you are not careful.” “Don’t be so positive. although one can consider permitting some or all of the values to be rational numbers. and I began muttering — “That can’t be. The next day.

but it is traditional for the parameters and the solutions. which is similar to. Now the last statement of the previous paragraph can be divided by to give us that Ü and Ý are integers if and only if ´ · ½µ´ · µ ½ ´ · ½µ´ · µ ½ ´¿µ ´ · ½ · ½µ ½ divides · This seems to be as simple a criterion for integrality as one could expect. ´ · ½µ´ · µ ´ · µ´ · ½ · ½µ. as before. which I would not have predicted. the age problem. I said I knew of no way to decide this that was simpler than actually seeing if the solutions were integers. Luckily. I’d have times you. I am indebted to S. say Ü Ý. I ﬁnd it particularly striking that and only enter via the sum · . The general problem ´ µ leads to the system of equations ´½µ ´¾µ Ü· Ü · ´Ý µ Ý· Ý · ´Ü µ The solutions are readily computed to be Thus. Hence a natural question is to consider the diophantine question Which integer problems have integer solutions? In the early 1990s [1]. unfortunately. that Ü is an integer if and only if Ý is an integer. This still doesn’t give us a very simple or satisfactory criterion. . assuming ½. to be integers. this false trail led me to the following simple result. Looking for a more symmetric solution. Ü and Ý are integers if and only if the second terms in (2) are integers. One can see from (2). Parameswaran for the letter that inspired this investigation. where ´ · ½ · ½µ denotes the GCD of · ½ and · ½. this doesn’t hold. assuming ½. SINGMASTER times you. and then determines which values of and give integral solutions. The general solution of the problem gives somewhat complicated expressions for Ü and Ý. Consider ´ · ½ · ½µ. which is if and only if ½ divides GCD ´ · ½µ´ · µ. This also divides ´ · ½µ´ · ½µ ´ · ½µ ´ · ½µ ½. The values of Ü and Ý are integers if and only if ½ divides both ´ · ½µ´ · µ and ´ · ½µ´ · µ.222 D. though. but somewhat more complex than. and it is obvious from (1). Here I present a reasonably simple criterion that allows one to generate all the integer problems with integer solutions.” and the second responds: “And if I had from you. The criterion allows us to pick arbitrary and . and I thought that the integrality of this was equivalent to integrality of Ü and Ý. I am also intrigued to see that any and any can be used. so we only need to check one of the second terms in (2). I ﬁrst noted that Ü · Ý has a symmetric expression." Many versions of this problem occur. however.

One can deal with rational Ü Ý by scaling the problem so that one has an integer problem. Æ «. so you don’t know whether the statements are made by the same or different animals. one can easily generate all examples. but by analogy with the argument above. We easily see that ´ · ½ · ½µ. we get a fairly direct way to check whether Ý is integral. Although the problem traditionally has all integral parameters. but we see from (6) that Ü is an integer if Ý is an integer. however. so Parameswaran’s version has integral solutions if and only if µ · ´ · ½µ´ µ ´ µ ´ · ½ · ½µ divides From this. Thus he considered the problem where equations (1) are replaced by ´ µ ´ µ Ü· Ý · ´ · ½µ´ ´Ý µ Ü· Ü ´Ý µ This is easily solved to obtain The expression for Ü is a bit complicated. so it seems most interesting to assume that we have integral and and we want integral Ü and Ý. but doing so doesn’t seem to make the result signiﬁcantly clearer. Thinking about Parameswaran’s version led me to wonder if there was any case where both versions gave integral answers.SOME DIOPHANTINE RECREATIONS 223 The basic role of ·½ and ·½ in the above leads one to ask whether the problem can be recast in some way to make these the natural parameters. That is. It is then quite feasible to consider rational and . Are there problems where the answers are integral. in either case? Perhaps surprisingly. it is direct to show that Ü and Ý are integral if and only if ´ µ ¬Æ «¾ ´¬ · « · µ divides «´ · µ Dr. Setting ¬ «. there are problems where the answers are indeed the same. or even the same. Parameswaran interpreted the problem as though the second statement was also made by the ﬁrst animal. though the converse may fail. possibly even the same answers. suppose you can hear the animals but can’t tell which one is speaking. loses all number-theoretic interest. this is not really essential. . From the ﬁrst equation. Now consider ´ · ½µ. we see that Ý is integral if and only if divides ´ · ½µ´ µ. the admission of irrational values. and I leave it for the reader to discover them. where GCD ´ ¬ Æ µ ½.

50.” This is a different type of problem and belongs in Tropfke’s previous section. not only in Europe but also in India and the Arab world. so that. 2. Alcuin gives an unusual variant of the problem in that he assumes the second person starts from the situation after the transfer mentioned by the ﬁrst person takes place. the second equation of (1) would be replaced by This is our problem ´ µ. 3) as an example. I’d have 5 times you plus 1 more. the problem is of type I. depending on whether ‘you’ refers to all the others or just the next in cyclic sequence. He proposes “to ﬁnd two numbers such that each after receiving from the other may bear to the remainder a given ratio. who typically gives many versions. 2. e. the Ass and Mule problem may not be in any Chinese source. about A.” who is elsewhere listed as probably being Ý · ´Ü · µ . Indian. 2. and Western sources. “If I had half of what you have. The other Chinese reference is a work of about A. from about 300 B.C..224 D. Two versions. He also covers similar problems with three and four persons. That is. including some with ﬁve persons. Arabic. The examples cited in the Chiu Chang Suan Shu [7] state. 3.” He gives (30. 10.” This includes all the relevant sources cited by Tropfke except Abu al-Wafa (only available in Arabic) and “Abraham. appear in The Greek Anthology [16]. and it has remained a standard problem ever since. 485 that has only been translated into Russian and may well be similar to the earlier Chinese example. the problem was a standard.g.. 5) and (2. “If you give me 7. Byzantine. The earliest three-person version that I know of appears in India. I’d have 50.D. 850. By about the ninth century. in the work of Mahavira [15]. an inconsistent example. The problem is readily generalised to more than two participants. The earliest four-person version I have seen is Arabic. SINGMASTER History of the Ass and Mule Problem. 4). in al-Karkhi (aka al-Karaghi) from about 1010 [13].g. The earliest known version of this problem has an ass and a mule with parameters (1. and extended versions where a person says.D. Heiberg’s edition [10] gives the problem in Greek and Latin verses. (10. surprisingly. Below I give a summary of versions that I have noted up through Fibonacci. and the problem is of type II. Diophantus studied the problem in general. 1. but then there are two forms. 1) and is attributed to Euclid. 2. e. Tropfke [20] discusses the problem and cites some Chinese. I denote these as I ´ µ for the case where ‘you’ means all the others and II ´ µ for the case where ‘you’ means just the next person in a cyclic sequence.

2. 3. 2) I-(9. 5. and to the second thirty apples. 300 B. 50. 1. 4.C. 5. 6. 2. Sridhara [18]. 23. 8. 2. although the material is believed to be based on older Arabic sources. 1. 11. 5. 10) (7. hence the above title. 7) II-(1. 4. 2. 2. Now how many sold each of them for a penny? . 850 c. 3. 5) c. 6) (1. done two ways Selling Different Amounts at the Same Price Yielding the Same Although this problem is quite old. 5. 7) I-(7. 4. 7) I-(7. 22. 10. 4. 9. 850 c. so I will ﬁrst describe the later form which is ﬁrst found in Fibonacci [11]. 3) (10. 3. 3. 11. 9. 5) (100. which daughters he delivered. 10. 5. certain apples. 5. there seems to be no common name for it. 6) I-(7. 6) I-(7. (1. 5) (2. 0. 1. with the problem statement giving the 6 as a 7 Fibonacci. 3. and all these three sold in like many for a penny. 9. 3. 2. 250 c.” printed by Wynken de Worde in 1511 [9]. to sell. 11. 8. 3. and Bhaskara [4]. noting that this system is inconsistent Fibonacci. with the problem statement giving the 6 as a 7 Fibonacci Fibonacci. 510 9Ø century c. 9. and to the youngest ten apples. c. I quote a version that appears as No. 11. A man had three daughters of three ages. 2. The problem is treated in one form by Mahavira [15]. 1. 50 in the ﬁrst printed English riddle collection “The Demaundes Joyous. 2. 4) (2. 4. 10. And he took to the eldest daughter ﬁfty apples. 1) (30. 2. 10. 9. but this form is less clearly expressed and has inﬁnitely many solutions. 3. 11. 11. which is rather a mouthful. 1010 1150 1202? 1202? 1202? 1202? 1202? 1202? 1202? Euclid Diophantus The Greek Anthology The Greek Anthology Alcuin Mahavira Mahavira al-Karkhi Bhaskara Fibonacci Fibonacci Fibonacci. 510 c. and brought home in like such money. 6) I-(7. 4.5) I-(25.SOME DIOPHANTINE RECREATIONS 225 from the early fourteenth century and hence after Fibonacci. 5.

I have now arrived at a reasonably simple approach that turns ¾I have not seen these editions. for which he often gives fractional answers. which would produce the positive solutions with the smaller price equal to 1.’ Anyone meeting this problem for the ﬁrst time soon realises there must be some trick to it. 30. which is essentially the same as the ﬁrst edition of 1694. Tartaglia [19] justiﬁes the different prices by having large and small pearls. if the prices are also given. 30). SINGMASTER ‘In like many’ means ‘the same number’ and ‘in like much money’ means ‘the same sum of money. Fibonacci goes on to consider solutions with certain properties — e. distinctly against the spirit of the problem. Although the objects are usually eggs. ﬁnd all solutions with one price ﬁxed. The extensively revised and enlarged edition of 1725 (or 1723) outlines a fairly general method and correctly says there are ten solutions. and he gives ﬁve solutions for his ﬁrst version — but there are 55. which normally are all of the same value. 30. This is historically by far the most common version. or lowering prices late in the day in an attempt to sell their remainders. he added several more versions as well. 40).226 D. Fibonacci is one of the few to give more than a single solution. Labosne revised the material. considers (10. or simply changing prices after lunch. In 1874. According to Glaisher [12]. 30) and gives two solutions.g. I tried several times to ﬁnd an easy way to generate the solutions. of which 36 are positive. This may happen by the sellers going to different markets. 25. dropping Bachet’s rule and replacing it with some rather vague algebra. During the 1980s. The question of ﬁnding all solutions for a three-person problem seems to have ﬁrst been tackled in the 1600s.. but we both conjecture that he was counting the positive solutions. and it is not immediately obvious how to ﬁnd or count all the solutions. or raising them late in the day when few items are left and new buyers arrive. 33). The 1708 English edition of Ozanam [17]. Let us denote the problem with numbers ½ ¾ by ´ ½ ¾ µ. Neither Glaisher nor I have seen the relevant work of De Lagny. the 1612 and 1624 editions of Bachet [3] ¾ give a fairly general rule for this case and apply it to (20. such as the ﬁrst receives twice the second. . it can take a fair amount of time to work out a solution. (12. 50). ﬁnd solutions where the amounts received are in some ratio. or sometimes fruit. He gives a fairly general rule for generating solutions [12]. (12. 32). is there a solution. Even with this trick exposed. contrary to De Lagny’s statement that there are six solutions. Fibonacci only gives two-person versions: (10. Fibonacci has no truck with trick questions and gives the necessary information — the sellers sell part of their stock at one price and then the remainders at another price. of which there are indeed six. so that the above is (10.

and generalizations that it is often difﬁcult to see where he is going — it is not until the 51st page that he gives the solutions for the problem (10.e. we get ´ ½ ¾ µ is a rational number. but did not thoroughly check that all solutions are obtained. Most previous methods yielded solutions involving some choices. special cases. this doesn’t change the problem.SOME DIOPHANTINE RECREATIONS 227 out to be essentially a generalisation and extension of Ozanam’s method. so we want the numbers to be integers. but he is so verbose and gives so many pages of tabular examples. By (4) below. so I later sketch his method in my notation and ﬁll in a gap. Glaisher’s paper contains most of the ideas involved. 50). Ayyangar [2] gives a much simpliﬁed general method for the Indian version of the problem and says many of his solutions are not given by Glaisher’s method on page 19 — although I can’t tell if Glaisher intends this to be a complete solution. and let be the number of items that the Ø person sells at the lower price . which he starts considering on the ﬁrst page. He also assumes that the are an arithmetic progression.. that ´ µ ´ µ ½. We can reorder them so that ½ ¾ ¡ ¡ ¡ Ò . This is equivalent to scaling the prices in equations (2) and reduces the dimensionality of the . whenever Glaisher gives the number of solutions. 30. Normally the objects being sold are indivisible. i. and we assume Ò ½ to avoid triviality.) From ½ · ½ ´ ¾ ½ µ. and we can assume that and are integers with no common factor. Indeed. i.. Hence ¾ · ¾ . We thus expect a three-dimensional solution set. Ø person has items to Consider the problem ´ ½ ¾ Ò µ. (One can reverse and in order to make the s increase. This makes ½ ¾ ¡¡¡ Ò and ½ ¾ ¡ ¡ ¡ Ò.e. subject to some conditions. but I prefer to deal with the original values. he displays or indicates all of them. We have ¾Ò · ¿ unknowns connected by these ¾Ò equations. but the amounts sold at the higher price are the smaller numbers and hence are easier to deal with. Then we have the following: ´½µ ´¾µ · · where is the constant amount received. Because of the symmetry between and . but it is not until late in the paper that one sees that he solves versions that are not in arithmetic progression by the simple expedient of inserting extra values. Ayyangar’s method is quite similar to my method. we can assume and have assumed . Let be the number of items that the Ø person sells at the higher price . the sell. No one else has noticed that the number of solutions in the Western version can be readily computed.

Since ´«¾ ¾ µ ½. and the solution set is again twodimensional. we have «¾ ¾ . This tells us that . where ´ µ ´«¾ ¾ µ ½. this implies that ¾ divides . Let ´«¾ ¾ µ. 50) problem has ¿ ½ — and the usual solutions have each person selling as many groups of ¬ as possible. ´«¾ · ¾ µ «¾ ½ · ¾ «¾. these prices are considered the same as ¾½ ½. in particular ½ ¬ for some integer ¬ — e. Solutions in the literature. ½ with the condition that be integral. the most common solution of the (10. are often given with fractional values. This is also equivalent to scaling the prices. we have ´ µ« ¾ ¾ . and I claim that any such «¾ generates an integer solution of our problem. as solutions that differ only in price scale are really the same. In the present approach. the yield is speciﬁed. We also get ´«¾ · ¾ µ . and (4) gives us ½ µ ´ divides . Glaisher considers ¿.. divides .g. i. From just before (6). «¾ . Hence ¾ . so that µ ´ µ « is a multiple of Now we will take « ¾ as the ﬁrst basic parameter of our solution. In some versions of the problem. so the previous equation becomes ´ µ« ¾ ¾ . From the case ¾ of (4). SINGMASTER solution set to two. Now use equation (1) to eliminate from (2).228 D. 30. however. This is quite reasonable. it follows that (4) has an integral solution for ½ if and only if divides ½ and this holds for each if and only if ½ leaves us Ò ½ equations: ´ µ « ½ divides GCD´ ½µ GCD´ It is now convenient to adopt the following: Thus (5) can be rewritten as for some .. . giving us ´¿µ ´ µ ´ ´ µ Subtracting each of these from the case µ´ ½ µ ´ ½ µ Since ´ µ ½ implies ´ µ ½.e. and they are considered to give the same sets of solutions. and we note that ½ «¾ ½ . unless one imposes some further condition such as being the reciprocal of an integer. we must have ¾ dividing «¾. which restricts the solutions to a subset of those for the case ¾½ ½. From (6). so «¾ ¾ . which is the condition (5) for all the equations (4) to have integral solutions.

hence ¾ ¿ ¾. and there are solutions with « . ¾. . we have ´ µ´ ½ Òµ Ò . we get no solutions. Thus Ozanam is correct. but four of these solutions have either ½ ¼ or ¿ ¼. «¾ gives ½¼ ¢ ¾¼ ½ · ½ ¿. 8. For «¾ ¿. set ¾ Ò «¾ Ò ¾ ´½½µ «¾ . and «¾ gives a negative value. so there are just six positive solutions.e. so we have ¾ ½ «¾ Ò . 9.SOME DIOPHANTINE RECREATIONS 229 As an example. for a given ´ µ «¾Ò ½ · ½ solutions ¾ In the above example. 9. namely (10.. we have ´ ½ µ ¿. For «¾ . ½ ¾ . ½ ¿ . there are «¾Ò ¾ divides Ò . Recalling that «¾ is a multiple of ¾ ½ . . . ½. 6. since «¾. 5.. For «¾ . so that Ò ¼ when ´ µ ½ Ò (This expression is an integer. we can let ½ until Ò becomes zero.e. Then ¾ ½ . From the case (4). 30). i. consider the problem discussed in Ozanam. 7. we have Ò of ´ µ ´½¼µ «Ò ½ Now the largest value «Ò can have is In our example.) Hence. giving us three solutions. We can now let ½ vary as the second basic parameter of our solution and we can let ½ ½¼. From the case Ò ½ ½ ½ of (4). 8. Counting the Solutions. hence ¾ ¿ ½. We can now determine the possible values of « ¾. ¿ ¾¼. and we choose «¾ as a multiple of ¾ ¿. ½ ¿ . we have ´¿ ½ µ ¿. 25. . i. indicating no solutions. we ﬁnd ½ «¾ . and there are solutions with « ¾ . «¾ ¿ gives ½¼ ¿ ¢ ¾¼ ½ · ½ solutions. We get solutions in the above section for each « ¾ that is a multiple of ¾ and in the interval ½ «¾ ½ . 4. For such an «¾. . ½ ¾ ¿. We can now let ½ ½¼. giving us seven solutions.

230 D. perhaps as intended by De Lagny. we have Ã ´ ½ Ò · ½µ ½ Æ Ã ´ ½ · ½µ Ò Ã ´Ã · ½µ ¾ ½¼ ¾ and ¾ ¢ ½½ ¾¼ ¢ ¾ ¢ ¿ ¾ ¢ Most of the earlier writers excluded zero values and only considered positive solutions. then ½ we get Ã ´ ½ ½µ ¾ Æ ´ ¾ ½µ Æ· ´ ½ ¾µ¾ ½ . From (15). Substituting (11) into (8) and summing gives us the number of solutions. 20. in our example. ½ A few longer APs occur — I have seen examples with Ò . i. 90). and then Ã ½ ¾ Æ ¾ Æ· ´ ½ ¾µ¾ .. Looking again at our example. i. Tartaglia’s problem 139 is the one with two sizes of pearls and is (10. which is trivially an AP.. Most of the examples in the literature have the in arithmetic progression (AP). We then have Ã ½ Æ ½´ ½ · ½µ ¾ Æ· ´ ½ ¾µ´ ½ ½µ ¾. Hence we normally get ¾Ã situations with a zero value. All but one of these have ½ even. Subtracting ¾Ã or ¾Ã ½ from Æ yields Æ· . Then we have Ò ´Ò ½µ . In that case. ´½ µ Again. unless it happens that ½ «¾Ò ¾ . . ½ «¾Ò ¾ by (7). SINGMASTER Then (10) becomes ´½¾µ ´½¿µ ½ ½ Ò The upper bound may not be an integer.e. The majority of examples are APs with Ò ¿. when two zero values occur in the same solution. and when Ò ¼. Æ· ½¼ ¾ ¢ ¾ . so we let Ã Æ ½ Ò Ã where [ ] is the greatest integer function. Ò Ò ½: ´½ µ Ã A number of early cases have Ò ¾. we have ¾Ã ½ solutions with zero values. i.. we ﬁnd that this occurs when ½ Ò . when the upper bound for given in (13) is an exact integer.e. we have Ã ½ Æ ¿ Æ· ½. Rearranging. A few cases occur that are not APs (see below). We can count the number Æ · of positive solutions by ﬁnding the number of solutions with a zero value. These occur when ½ ¼. A little thought shows that then is the common difference of the progression. If ½ is odd.e. This led me to ask when a problem admits only one solution or only one positive ½ Ò ½ Æ Ã ´ ½ · ½µ ´Ò ½µÃ ´Ã · ½µ ¾ . ½ ½ . so we can write ½ ·´ ½µ .

we are speciﬁcally given the greater price . given as ´Æ Æ· µ.e. and resold [a part] so: and disposed of the remainder at one for ﬁve panas. so the above gives all cases where Æ · ½ and Æ ¿. The Indian Version As mentioned earlier.SOME DIOPHANTINE RECREATIONS 231 solution. but the capitals (in panas) of each trader. A little manipulation shows that this works if and only if ´½ µ Ò For an AP. (10. where ¬ is an integer.. i. In general. For an AP. noting the analogous cases with 10 and 11 people. We will get Æ ¿ only if Ã ½. bought leaves of betle [or fruit] at an uniform rate.. To illustrate. Ò 20. eight. as in Tartaglia’s example. consider the following problem from Bhaskara. Below I tabulate all the different versions that I have noted. we have Ã ¾ and Æ . so the the numbers of items will be and our equation (1) becomes ´ · ½µ ½ ¾ ´½¼µ · Further. and the number of solutions given in the source. hence Ò ½ or ½ ´ · ½µ ½. It is assumed that you can buy items per pana. and thus became equally rich. 110) has Æ ½. Three traders. Initially I thought that Æ · ½ would imply that Æ ¿. What was [the rate of] their purchase? and what was [that of] their sale? This requires some explanation. Æ· ½ implies that the unique positive solution occurs with ½ and has ½ ½ ½ and Ò ½. we ﬁnd Ã ¿ and Æ . but when Ò ¾.e. Ayyangar’s translation of the problem . along with the ﬁrst known (to me) dates and sources. ½ Ò ¾. and it is understood that the lesser price corresponds to selling ¬ items per pana. . the Indian version is somewhat different and gives inﬁnitely many solutions. Example instanced by ancient authors: a stanza and a half. while when Ò ¿. Glaisher [12] also studied this problem and was intrigued by this observation. this turns out to be if and only if ½ Ò · ½. i. having six. used as an example in Ayyangar. the numbers of solutions. and a hundred. Combining this with (16) shows that this implies Ò ¿. Here the numbers are not the numbers of objects. Tropfke [20] and Glaisher [12] cite a number of sources that I have not seen. ½ ¬ . For example. this is if and only if ½ Ò ½. Æ ½ implies that the only solution occurs with ½ and has ½ and Ò ¼. Similarly. which gives us ½ Ò . for their capitals respectively.

33) (20. 170) (305. 36) (78. 16) (16. 50) (10. . 29. 454. 20) (10. 49) (3. 55) (16. 1 given. 1 given. 1 given. 33. 16) D. none given. 37) (27. Sridhara ﬁnds a one-parameter solution set. 12. 1 given. 50) (30. 196) (45. One can scale the set . 20) (64. 11. 50) (11. 80) (10. 15) (31. (20. 1 given. 1 given. 136) (3. 1300 1300s 1489 1489 1489 c. 30) (18. . so we again have a three-dimensional solution set. 40. 1 given. 603. 90) Date 1202? 1202? 1202? c. 17. 60) . 752. 100) (45. 55) (78. 1 given. 36) (30. 26) (20. none given. 140) . 901) (10. one with fractions. 30. though it is not as easy to see how to do this as in the Western version. 1 given. 36) (27. 32. 1 given with fractions. 30) (12. SINGMASTER Solutions in Source 5 given.4) (70. 1 given. 1) (100.9) (100.6) # of Solutions (55. makes this more speciﬁc by saying they ‘sold a part in lots and disposed of the remainder at one for ﬁve panas. Bhaskara makes the following comment: “This. 30) (8.232 Version (10. At this point. 20. 1 given. . 1500 1513 1515 1521? 1556 1556 1556 1612 1725? 1874 1874 1874 1874 1893 1905 1924 Source Fibonacci Fibonacci Fibonacci “Abraham” Munich codex14684 Widman Widman Widman Pacioli Blasius Tagliente Ghaligai Tartaglia Tartaglia Tartaglia Bachet Ozanam from De Lagny Labosne Labosne Labosne Labosne Hoffmann Dudeney Glaisher (171. 2 given. 35) (11. 119.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 81) (10. 60) (10. 56. 25. I must point out that some Indian versions give and/or as fractions! The Indian authors do not get very general solutions of these problems. 1 given. 6 given. 40. ¬ . 68. 1 given. all given. 33) (10. 48. 1) (7. 81) (55. (20. 30. 32) (12. 60) (117. 40) (18. in some way to reduce the solution set to two dimensions. 40) (10. 82) (17. 21) (225. 1 given. 40. 30.400) (25. 36) (25. 20. 55) (55. 1 given.552. 4 given.

and ´ µ can be a proper divisor of Ö. pointing out an improvement. Eliminating ¬ . the answer must by the intelligent be elicited by the exercise of ingenuity. but in order for «½ Ò ½ .. which is if and only if Ö divides ´ µ. where Ö comprises all the prime powers in whose primes also occur in . Then ´ ¼¼¼ µ is equivalent to ¬ ½ ¼ ´mod µ and this is solvable if and only if ´ µ ½. This gives us a rather peculiar condition. Then is relatively prime to if and only if divides ×. we must have Ò ´ ¼µ ¬ ½ ´ ¼µ « .SOME DIOPHANTINE RECREATIONS 233 ´½¼¼ µ which is instanced by ancient writers as an example of a solution resting on unconﬁrmed ground. which is if and only if Ö divides .e. when the operation.” Amen! For ease of expression. i. we let ¬ ¬ . ´ ¼¼¼µ ´ ¬ ½µ ¼ ´mod µ Let ´ µ. and such a supposition introduced. as has brought out a result in an unrestricted case as in a restricted one.) Both and ¬ can take on an inﬁnite range of values. In the like suppositions. but since his notation is different. so our equations become · ¬¬ ·¬ ´¾¼¼ µ ´ ¼µ ´ ¼µ Ayyangar gives a reasonable solution process for this problem. we get ¬ ¬ ½ ´ ¬ ½µ « divides and the condition for all the equations to have integral solutions is If we set ´ ¼¼ µ then ´ ¼µ gives ´ ¬ ½µ We need to determine which and ¬ will make ´ ¼¼ µ hold. has been by some means reduced to equation. owing to restriction. Write Ö×. then subtracting to eliminate and using the same abbreviations as before. disappoints. (Ayyangar notes only that ´ µ must divide . I will sketch the solution in the present notation.

1624. annotated by David Singmaster and John Hadley. Rigaud. Neun Bucher arithmetischer Technik. 1st ed. 13. Gaz. 1782. Vieweg.C. with Arithmetic and Mensuration from the Sanscrit of Brahmegupta and Bhascara. 150 B. v. Gordon Fraser Gallery. London. trans. 2nd ed. J. PrenticeHall. 18. pp. 2nd ed. “A classical Indian puzzle-problem.470 [4]. Labosne. 4th ed. 5th ed. Braunschweig. London. J. 1817. Krishnaswami. See No. Claude-Gaspar. Indian Math. London. 170. The ﬁrst nine editions appeared “without any material alterations. 54. 21. pp. 5th ed. A. Nos. Arithmetica. Diophantos of Alexandria. Bîjaganita. Facsimile with transcription and commentary by John Wardroper. 214–223.. Translated by John Hadley. Lucas N. nos.. London. 1884. he produced a 10 Ø ed. .so and ¬ ½½¼.. 1968. 12. In: T. 1976. [8] Diophantos. H. with Notes and Observations. Heath. 102– 126. Vogel. 1879. 15. By the way. Math. there are only a ﬁnite number of solutions. “Nine Chapters on the Mathematical Art. [4] Bhaskara. (There have been several reprints. 1150. References [1] Alcuin. [3] Bachet. 86–88. Gauthier-Villars. Paris.900 and of 3294. and Other Places of Public Education. 1964. Johnson. In: Henry Thomas Colebrooke. aka Bhâskara. 3rd ed. Wynken de Worde. 1923–24.” J.. John Murray. P. 134–136. Prob. 76.) Chap.. 10. 106–115. Analysis similar to that done before shows that there are then ½ · ´ ¬ ½ Ò µ ¬ solutions. reprinted by Blanchard. Mar 1992. A. 24. [7] Chiu Chang Suan Ching. Lyon. 3200 and 16. pp. pp. Lyon. Nunn et al.(475). “an entire revision of the work.” which “may be considered as a concise abridgment” of his two-volume Treatise on Algebra. 1959. Bijanganita. 19. pp. Prob. [9] The Demaundes Joyous. and the 13Ø edition. Bhaskara solves his example by taking ¾. Rigaud.” c. SINGMASTER For given and ¬ satisfying ´ ¼¼¼ µ and ´ ¼µ. pp. 1612. An Introduction to Algebra. Book I. 16: “De duobus hominibus boves ducentibus — Two men leading oxen. including Sandig. 1824.. 178–186. 4392.. 24. 1874. giving total numbers of 3295. John. Revised by A. 242–244. Problemes plaisants & delectables qui se font par les nombres. 1910. Prob. [5] Bonnycastle.. 33.. The Historical Roots of Elementary Mathematics. P. pp. so . 15. Translated into German by K. Problems to Sharpen the Young. Wiesbaden. Paris. designed for the Use of Schools.234 D. Algebra. 3292. 1973. [6] Bunt. which may be the same as the 1815 edition. Soc.. 1805. 6. pp. 1813. Oxford University Press.” [2] Ayyangar. London.. 122–126. L. 1511. I examined the 7 Ø edition.” In 1815. et al.. p. reprinted by Dover.

Tricks and Conundrums. Recreations Mathematiques et Physiques. Pinnacle or Corwin. Metrodorus (compiler). 5. L. Jacques. Bonwick. Paris. aka Mahâvîrâ(cârya). MS 952. S. Tartaglia. “On certain puzzle-questions occurring in early arithmetical writings and the general partition problems with which they are connected. Government Press. p.. C. pp. Mass. vol. New edition. 1914. Revised by Kurt Vogel. Teubner. Glaisher. 1857. Menge. 24. R. surviving in a single example in the Cambridge University Library. pp. Euclid. 1556. 286–287. General Trattato di Numeri et Misure. Edited by J. W. 50. 24. Tropfke. book 16. Translated by M. 1982. Abo^ Beqr Mohammed Ben Alhacen. 1708.000 Puzzles. Translated by W.SOME DIOPHANTINE RECREATIONS 235 [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] 1971. Lucknow. Sect. Harvard University Press. 1010. Venice. Sam Loyd’s Cyclopedia of 5. 850. and Helmuth Gericke. Karin Reich. Vol. 1916–18. 900. 6 of the facsimile. Fibonacci. et al. Cambridge.D. 1912. Paris. Jombert. Rome. ex. Opera.. 68–70. Transcribed and translated by K. 1853. Nicolo. p. London. 1976. Untitled manu uscript called “Alfakhrî. I. . Lamb Publishing. VIII. prob. Johannes. prob. Pâtîganita. V. Shukla. 1: Arithmetik und Algebra. Edited into French by Franz Woepcke as “Extrait du Fakhrî. and Heinemann. L. Heiberg and H. In: Scritti di Leonardo Pisano. Ragacarya. Paris. Curtio Troiano. 79–80. pp. Hildesheim. Prob. 298–302. R. c. A. 1916. Vol. edited and published by B.. This is the oldest riddle collection printed in England. pp. Loyd. Reprinted by Georg Olms Verlag. 1980. Lucknow University. Part 1. 1694 (not seen). Leipzig. 90. Madras. Geschichte der Elementarmathematik. Alkarkh. J. 1228. 1725. Arabe de la Bibliotheque Imperiale. 3. 53 and 346. “Ganita-sâra-sangraha. 76–77. 2 Ò edition.” A.” c. 201–210. Sridhara. pp. pp.. 1–131. 1923–24. Paton. The Greek Anthology. four vols. Supp. Paris. pp. Loeb Classical Library. Amsterdam. Berlin. Sam. pp. 1696. (1202).” Messenger of Mathematics. Liber Abbaci. Vol. prob. Boncompagni. Ozanam. 256r–256v. 1959. 53.” L’Imprimerie Imperiale. pp. aka Srîdharâcârya. 28. Reprint. De Gruyter. English version: Recreations Mathematical and Physical. 60-62. 44–49 and 94. 5. aka Leonardo Pisano. 4 Ø edition. 26–27 of the transcription. edited by Grandin. no. 136–139. reprinted 1976. prob. pp. Edited by Sam Loyd II. aka al-Karagi. London. Mahavira.

White’s goal is to put white stones in a set of hexagons that connect the top and bottom of the rhombus. for in any Hex board ﬁlled with white and black stones there must be either a winning path for white.Who Wins Misère Hex? Jeffrey Lagarias and Daniel Sleator Hex is an elegant and fun game that was ﬁrst popularized by Martin Gardner [4]. there must be a winning strategy for either the ﬁrst or the second player. Assume. The game was invented by Piet Hein in 1942 and was rediscovered by John Nash at Princeton in 1948. (This fact is equivalent to a version of the Brouwer ﬁxed point theorem. 3].) Since the game is ﬁnite. for the sake of contradiction. Two players alternate placing white and black stones onto the hexagons of an Æ ¢ Æ rhombus-shaped board. and Black’s goal is to put black stones in a set of hexagons that connect the left and right sides of the rhombus. Gardner credits Nash with the observation that there exists a winning strategy for the ﬁrst player in a game of hex. The proof goes as follows. 237 . or a winning path for black [1. as shown by Gale [3]. First we observe that the game cannot end in a draw. A game of 7 ¢ 7 Hex after three moves. A hexagon may contain at most one stone.

and now imagines that this one is empty. In fact. The purpose of this note is to analyze a variant of Hex that we call Misère Hex. The difference between normal Hex and Misère Hex is that the outcome of the game is reversed: White wins if there is a black chain from left to right. This fact is mentioned in Gardner’s July 1957 column on Hex. LAGARIAS AND D. It sufﬁces to prove the second assertion. the winning strategy will work. the proof of the existance of a winning strategy does not shed any light on what that strategy is. was later developed by Yamasaki [5]. An abstract theory of “Division games. If the strategy requires the ﬁrst player to move in this non-empty cell. Furthermore. Of course this proof is non-constructive. A small step was made in that direction by Ron Evans [2] who showed that for even Æ . on even boards the ﬁrst player can win.238 J. because it shows that the parity of the number of cells on the board determines which player has the winning strategy.” which includes Hex and Misère Hex as special cases. For any winning strategy Ä for È . Theorem: The ﬁrst player has a winning strategy for Misère Hex on an Æ ¢ Æ board when Æ is even. the winner depends on the parity of Æ . and on odd boards the second player can win. As in the case of Hex. Since the extra stone can only help the ﬁrst player. the player simply chooses another empty cell in which to play. The ﬁrst player can make an arbitrary ﬁrst move. our result shows that in optimal play the loser can force the entire board to be ﬁlled before the game ends. either the second player or the ﬁrst player has a winning strategy. In addition to showing who wins. and Black wins if there is a white chain from top to bottom. Here we present an elementary proof showing who wins Misère Hex. then follow the winning strategy (reﬂected) for a second player (imagining that the hexagon containing the ﬁrst move is empty). Contrary to one’s intuition. and let É be the other player. Let È be the player who has a winning strategy. Proof. and the second player has a winning strategy when Æ is odd. Gardner attributes the discovery to Robert Winder. Misère Hex has also been called Reverse Hex and Rex. who never published his proof. the losing player has a strategy that guarantees that every cell of the board must be played before the game ends. SLEATOR that the second player has a winning strategy. Because the game cannot end in a draw. the ﬁrst player can win by moving in an acute corner. and an explicit winning strategy for the ﬁrst player is not known. it is not the case that the second player can always win at Misère Hex. This contradicts our assumption that the second player has a winning strategy. and the ﬁrst player wins.

Suppose that É is the ﬁrst player. Consider a terminal position of a game that is a win for É. in that we will describe a new strategy for É in which (in effect) É makes an extra move and then plays the reﬂected version ÄÊ of È ’s hypothetical winning strategy Ä. and draws a circle around the cell containing this move. except that in the imaginary game the encircled cell is empty. When this happens she has also won the real game. because of the monotonicity property. We are now ready to prove the theorem. The proof splits into two cases depending on whether É is the ﬁrst player or the second. Therefore it is possible for È to move. When the strategy Ä Ê requires É to play in the encircled cell. By deﬁnition. She makes an arbitrary ﬁrst move. while in the real game. because none of these changes would interfere with the È -path. she plays instead into another empty cell (chosen arbitrarily). playing out ÄÊ in an imaginary game. . From now on she applies strategy Ä Ê in what we shall call the imaginary game.) The proof is complicated. when it is É’s turn to move there must be at least three empty cells in the imaginary game. (Note that Ñ´Äµ Ñ´Ä Ê µ. This contradicts our assumption that È has a winning strategy. and draws a new circle around the move just played. Player É begins by encircling Ô¼. Eventually É will win the imaginary game because Ä Ê is a winning strategy. Now. by the fact that it is not clear a priori that having an extra stone on the board is either an advantage or a disadvantage. This relationship will be maintained throughout the game.) Similarly. however. Let Ô ¼ be È ’s ﬁrst move.WHO WINS Misère HEX? 239 deﬁne Ñ´Äµ to be the minimum (over all possible games of Misère Hex in which È plays strategy Ä) of the number of cells left uncovered at the end of the game. supposing that Ñ´Äµ ½. Because Ñ´ÄÊ µ ½. Suppose the position is modiﬁed by ﬁlling in any subset of the empty cells with É’s stones. The state of this game is exactly like that of the real game. and further modiﬁed by changing any subset of É’s stones into È ’s stones. Player É applies the following strategy. The contradiction will be to show that under this assumption É has a winning strategy. that cell contains a É-stone. We shall argue by contradiction. such a position contains a È -path. when it is È ’s turn to move there must be at least two empty cells in the imaginary game. We must show that Ñ´Äµ ¼. erases the circle. and there must be at least one empty cell in the real game. so there are at least two empty cells in the real game. suppose that É is the second player. The basic idea resembles Nash’s proof that the ﬁrst player has a winning strategy for Hex. We shall make use of the following monotonicity property of the game. The position is still a win for É. (We’ll see below that È will not have won the real game. Thus the real game can continue.

7(3). Inst. 337–358. Gardner. and changing the contents of Ô¼ from a É-stone to a È -stone. 1978. New York. Kyoto Univ. The fact that Ñ´Ä Ê µ ½ ensures that both players can continue to move. New York. Worth. Yamasaki. This contradicts our assumption that È has a winning strategy. SLEATOR The imaginary game and the real game differ in up to two places. Res. pp 189–192. pp. M. It is easy to see that the relationship between the real game and the imaginary game is maintained. Publ. The American Mathematical Monthly. The imaginary game is obtained from the real game by ﬁrst changing Ô¼ from È ’s stone to É’s stone. 73–83. Evans. 1969. A winning Opening in Reverse Hex. pp. [5] Y. Summer 1974. Excursions into Mathematics. . 1979.. and transfers the circle from its current location to the new cell. 327–339. If the strategy ÄÊ requires a move into the encircled cell. Gale. References [1] A. Sci. [2] R. 14. Crow. 818–827. Recreational Mathematics. Player É eventually wins the imaginary game. and J. and the monotonicity property ensures that the corresponding position in the real game is also a win for É. The position in the real game is obtained from the position in the imaginary game by putting É’s stone in the encircled cell. Math. LAGARIAS AND D. then É arbitrarily chooses a different empty cell in which to move. pp. Bleicher. as follows. [3] D. The position in the imaginary game is a winning position for É. The Game of Hex and the Brouwer Fixed-Point Theorem. then by erasing the stone in the encircled cell. J. [4] M.240 J. Simon and Schuster. 1959. Beck. 86 (10). pp. Theory of Division Games. The Scientiﬁc American Book of Mathematical Puzzles and Diversions.

Æ . We need to show that this equation has a solution for every adjacency matrix . Let be the Theorem 1. A neighborhood. Æ is odd if there are an odd number of squares in Æ with labels “1". Bryan. The ﬁrst solution of Problem 1 was said to be quite difﬁcult. Later. . if the problem is generalized to all graphs the proof is quite elementary. Then Ý´ · Á µ ¼ ´µ Ý ÝÁ ´µ 241 ÝÌ ÝÁÝÌ ÝÝÌ ½Due to the author’s son. of square . cellular automata ideas were used. while a graduate student. Suppose Ý is a ½ ¢ Ò vector of 0’s and 1’s. of course. is the set of squares sharing an edge with square and does include square . Then the vertices of can be labeled with 0’s and 1’s so that every neighborhood is odd. the integers modulo 2. over ¾. Shader Some time ago (1986–88) this problem appeared in mathematical circles: Problem 1: Can the unit squares of an Ò ¢ Ò sheet of graph paper be labeled with 0’s and 1’s so that every neighborhood is odd? A neighborhood. since we are interested in only odd/even properties. ½ ½ ¼ ½ ½ ½ ½ ¼ ¼ ½ ½ ½ ½ ¼ ½ ½ Figure 1. Proof: ½ Consider the matrix equation ´ · Á µÜ ½ where 1 is a column of 1’s. The linear algebra is.An Update on Odd Neighbors and Odd Neighborhoods Leslie E. However. be a graph with no loops or multiple edges. Figure 1 shows a 4 ¢ 4 example with a desired labeling. Let adjacency matrix of .

Is there an Ò so that the developer can sell all the lots in the Ñ ¢ Ò rectangle and satisfy the politically correct criteria that all neighborhoods have an odd number of odd neighbors? When this paper was originally written. Theorem 2. to all graphs and prove a theorem. so ´ · Á µÜ ½ for every adjacency matrix . If we think of the unit squares in Problem 1 as vertices of and Ú Ú is an edge in ´µ Ú shares an edge with Ú . ÝÝÌ ¼ and rank´ · Á µ rank ´ · Á ½µ. Let be a graph with vertices labeled with 0’s and 1’s in any way. as before. E. SHADER But is symmetric. the answer to the Developer’s Dilemma was unknown. However. . then can be embedded in a graph ¼. But this time the generalization does not yield a solution to the Developer’s Dilemma. Figure 2.242 L. for every graph . The Developer’s Dilemma Suppose a real estate developer has an inﬁnite (at least sufﬁciently long) strip of land Ñ lots wide. Now for some new questions (at least at the time that this paper was submitted). If not all neighborhoods of are odd. so Ý Ý Ì ¼. with all neighborhoods in ¼ odd. It is quite possible that some of the buyers are odd. and some are not odd. We will generalize. a proof that the Dilemma can be solved will also be presented. Ñ Further. then Theorem 1 implies that the squares of the Ò ¢ Ò square can be labeled with 0’s and 1’s so that each neighborhood is odd. we assume all lots in the ﬁrst strip of Ñ lots have been sold. Finally. Therefore. the vertices of can be labeled with 0’s and 1’s so that every neighborhood in is odd.

Let the Ñ ¢ ½ strip of unit squares be horizontal. there are an even number of vertices ¾ labeled 1 and Æ is even in . any 0–1 Ñ-tuple. But.) . Let be the number of vertices labeled 1 and adjacent to vertex ¾ . all neighborhoods are odd. Theorem 3. The Developer’s Dilemma can be solved. Edges from Û to vertices Ú ¾ .ODD NEIGHBORS AND ODD NEIGHBORHOODS 243 It is rather interesting that only one vertex must be added to to prove Theorem 2. Let ¼ be the graph formed from by adding a new vertex. So Æ Û is odd ¾ ¼. Ü½ Ü¾ Ü¿ Ü Ü (b) If not. (See Figure 3(b). and we describe an algorithm that can be used to label the next column so that all Æ ½ are odd for Ú½ and unit square in column 1. ¾ ½ ¼ ¼ ½ ¼ ¼ ½ (a) Figure 3. are also added. Æ is even in ´µ is odd and Æ is odd in ´µ is even. Clearly all neighborhoods ÆÚ with Ú Ú ¾ ¼ are now odd in ¼. We wish to construct a proper labeling of an Ñ ¢ Ò rectangle so that all neighborhoods are odd and the vector Ú½ is the “starter" column. Hence. Û. We call Ú ½ . with ÆÚ even. a “starter" for the Ñ ¢ Ò rectangle we seek. These vertices now are all connected to Û with label 1. Proof. Figure 3 shows that this phenomenon can happen. Label Û with a 1. is even labeled ½ since the sum counts the edges from vertices labeled 1 exactly twice. Perhaps the Ñ ¢ ½ rectangle using the starter Ú ½ has the property that all neighborhoods are odd. in order Proof. We need only to show Æ Û is odd in ¼.

Of course. and each new column will be dependent on just the preceding two. For Ñ . ¼ ¼ ½ ½ ¼ ½ ¼ ½ ½ ¼ ¼ ½ ¼ ¼ ¼ ½ ¼ ½ ¼ ½ ¼ ¼ ¼ ½ ¼ ¼ ½ ½ ¼ ½ ¼ ½ ½ ¼ ¼ ¼ ¼ ¼ ¼ ¼ Figure 5. Ò is independent of Ú. Ú ·½ . We can generate new columns moving from left to right or from right to left. Perhaps the next question might be For Ñ ﬁxed. Figure 5 shows an example where Ú ½ col´¼¼½½¼µ. i. let Ü ¼. otherwise Ü¿ ½. Using the second occurrence of Ú of the ﬁrst occurrence of Ú Ú ·½ . i.. otherwise Ü ½. to show that Ò ¾¿ is a nice exercise for the reader. let Ü¿ ¼. · · is odd. · · is odd. Æ Ò is odd for square Ú in column Ò Ò.e. we can repeat the process for column 2. Ñ is not limited to the value 5.e.244 L. otherwise Ü½ ½. . · · is odd. Each neighborhood If If If If If This process can be repeated as many times as we like. let Ü¾ ¼. let Ü ¼. let Ü½ ¼. Contradiction! Hence there is a column of 0’s and the proof is complete. Let Ú Ú ·½ be the ﬁrst pair that repeats. otherwise Ü ½. Now that every neighborhood for Ú½ is odd. We will have a solution to our problem if the Ò · ½ column has all entries 0. · is odd. and we generate a second Ú Ú ·½ somewhere to the right of the ﬁrst Ú Ú ·½ (see Figure 4). SHADER will now have all but one entry labeled. Ú½ ¡ ¡ ¡ Ú Ú ·½ ¡ ¡ ¡ Ú Ú ·½ Figure 4. and that label can be assigned so that the neighborhood is odd. Since there are only a ﬁnite number of adjacent pairs of the Ñ-tuple. and Ò . is there an Ò such that for every starter Ú the Ñ ¢ Ò rectangle has all neighborhoods odd?. · is odd. E.. either we get a column of 0’s or there must be an adjacent pair that repeats in our process. otherwise Ü¾ ½. column Ú ½ must repeat to the left Ú½ ¡ ¡ ¡ Ú ½Ú Ú ·½ ¡ ¡ ¡ Ú ½ Ú Ú ·½ So Ú Ú ·½ is not the ﬁrst repeating pair.

This suggests an intriguing problem: How must Æ point mirrors be placed and oriented to reﬂect an entering ray of light as often as possible between these mirrors? What is the maximum number of reﬂections for varying Æ ? Upper Limits. P = perpendicularly reﬂecting mirror). half of the potential reﬂections would not be used. Solutions are shown in Figure 1. If we introduce the condition that a ray should reﬂect only one point on the mirror — reducing the mirrors to point mirrors — we ﬁnd a maximum of three reﬂections for two mirrors and seven reﬂections for three mirrors. From Figure 1 we can see that after some reﬂections the ray of light is reﬂected perpendicularly onto a mirror P and returns along the same paths but in the opposite direction. Figure 1.Point Mirror Reﬂection M. if these are suitably placed and oriented. It is obvious that the strategy of 245 . Oskar van Deventer The Problem It is well known that a ray of light can reﬂect many times between two ordinary line mirrors. If the ray would not return along the entry paths. Maximum number of reﬂections for two and three point mirrors (E = entrance mirror. passing once again by entrance mirror E.

O. as shown for the regular pentagon in Figure 3.246 M. so remains valid. . using a perpendicularly reﬂecting mirror P yields the maximum number of reﬂections. Only at mirror P can there be an odd number of reﬂections. which is an odd number. the theoretical maximum number of reﬂections is Æ ´Æ ½µ · ½. At each mirror a maximum of Æ ½ rays can be reﬂected. for even Æ there is a parity complication. see Figure 2. at the other Æ ¾ mirrors (other than the E and P mirrors) there can’t be Æ ½ reﬂections. since the light paths are used in two directions. At mirror P there can be at most Æ ½ reﬂections. from the other mirrors. Number of reﬂections for different mirror types. but only Æ ¾ to make it an even number. At mirror E there can be one additional reﬂection for the entering ray. the angle between all the neighboring pairs of light paths is 180 Æ Æ . Therefore. So the upper limit for an even number of mirrors is Æ · ´Æ ½µ · ´Æ ¾µ´Æ ¾µ or ´½ µ Ê¾ Æ ¾ ¾Æ · ¿ Ê½ For odd Æ the parities are okay if E and P are the same mirror. At mirror E there can be at most Æ reﬂections. At all other mirrors there must be an even number of reﬂections. or ´½ µ Ê½ Æ¾ Æ · ½ However. Æ ½ from the other mirrors and one from the entering ray. going to every other mirror. V AN DEVENTER Figure 2. Mirrors on the Corners of a Regular Polygon Can the theoretical maxima Ê ½ and Ê¾ be achieved for any Æ . However. and what should the positions and the orientations of the mirrors be? A well-known geometrical property may help us: When the Æ mirrors are put on the corners of a regular Æ -gon.

We take Ñ ½ for E (= P) and Ñ ¼ for all other mirrors. since a ray can neither enter nor leave such a loop. Only discrete orientations of the mirror are allowed and are multiples of 90 Æ Æ . Figure 4. Since Æ is prime. after passing mirror E again. Prime Number of Mirrors. makes hops of one mirror. the investigation of various cases is less straightforward. Using such an Æ -gon grid. For prime Æ the maximum number of reﬂections can always be achieved. Draw the Æ -gon and all possible light paths (the Æ -gon grid). two-dimensional space is sufﬁcient to ﬁnd solutions for each Æ . starting at mirror E. The ray of light.POINT MIRROR REFLECTION 247 Figure 3. Omit one or more light paths (lines of the Æ -gon grid). . For Ñ ¼ we have the normal orientation. and adjust the mirror orientations such that symmetry of light paths is preserved at every mirror and such that only at mirror P is there a perpendicular reﬂection. This is illustrated in Figure 4. Add the starting ray at an angle of ½ ¼ Æ Æ with the Æ -gon. A systematical trialand-error search for the maximum number of reﬂections may be performed using the following strategy: Step 1. say Ñ ¢ ¼Æ Æ . Prevent any obvious loops. If the number of mirrors is non-prime. see Figure 5. Step 2. and so on. Rotation of a mirror. Non-Prime Number of Mirrors. it makes hops of two mirrors. Ñ ½ yields a rotation of one step clockwise. and so on. Light paths in a regular pentagon. the ray will pass every mirror once at each tour from E to E.

At least two light paths have to be omitted for Step 2. reject the candidate solution. . Reﬂection diagrams for prime numbers of mirrors. If this is true. there is a loop and the candidate has to be rejected. a choice of a set of 2 points out of 6. and so on. Figure 6 illustrates the procedure for the case of six mirrors. V AN DEVENTER Figure 5. If not. until one or more solutions are found. then for all candidates missing two light paths. Step 3. Figure 6. Searching strategy for Æ = 6. The grid consists of 15 lines. Indeed. we do not count 2 ¢ 13 + 1 = 27 reﬂections but only 19. Count the number of reﬂections by following the entering ray. (a) Hexagon grid plus entering ray. Calculate whether the number of reﬂections is equal to two times the number of remaining light paths plus one. The candidate has 13 light paths. (b) A candidate having 19 reﬂections and a loop. Repeat Steps 2 and 3 for all candidates missing one light path. decreased by 2 to account for the loss of the two vertical lines on the left and right sides. however. there are no loops and all of the remaining light paths are used.248 M. shown separately in (c) and (d). In this case. O.

It is likely that this is true. For all other Æ up to 16 there is only one solution. 8. A general solution of the problem has not been found. Data for composite Æ up to 16. Some results obtained by applying the strategy outlined above are summarized in Figures 7 and 8. in which Ò stands for any divisor of Æ . see Figures 7. m = mirror orientation number. Ñ must be 0 for all mirrors. c = corner number. and 9. the value Ê¾ can be achieved. ¯ Conclusion A problem was deﬁned regarding the maximum number of reﬂections of a ray of light between Æ point mirrors. For Æ composite and odd the theoretical maximum of Ê ½ can never be achieved. but might be possible. entrance mirror E is positioned in corner number one ( ½). which has Ñ ½. ¯ For Æ there are two solutions. For this case a sharper theoretical maximum might be deﬁned. since light paths have to be omitted to account for all Æ Ò loops. ¯ For even Æ this sharper maximum is given by Ê¾. The position of the perpendicular reﬂecting mirror P is marked with an asterisk*. since only the Æ -gon grid seems to have a geometry that suits the problem. except for E. There will always be an Æ Ò loop.POINT MIRROR REFLECTION N 4 6 6 8 9 10 12 14 15 16 249 m R 11 27 27 51 69 83 123 171 199 227 d 1 2 2 3 2 4 5 6 6 7 0 1 1 1 1 1 1 1 *1 1 1 *0 1 0 1 0 1 1 1 0 1 2 0 -1 1 -1 0 -1 -1 -1 0 -1 3 1 -1 *0 0 *0 1 1 0 *0 4 -1 0 1 *1 0 -1 -1 0 0 5 *0 *0 -1 -1 1 *0 1 0 0 6 1 0 -1 1 -1 0 0 7 -1 0 1 -1 1 0 0 8 0 -1 1 -1 0 1 9 0 -1 1 2 -1 10 1 -1 0 1 11 -1 1 -2 -1 12 -1 1 0 13 *0 -1 0 14 0 0 15 0 16 The c N = number of mirrors. Figure 7. For Æ ¾ to Æ ½ . nor a proof that a solution can always be put on our Æ -grid. If no light paths are omitted. . Actual solutions for Æ ½ or Æ ¾¼ have not been found. R = number of reﬂections. d = number of deleted paths. A theoretical maximum was found.

No general strategy was found. O.250 M. solutions were found by a systematic trial and error search using a regular Æ -gon grid. Deleted paths for composite Æ up to 16. which can be reached for prime Æ . V AN DEVENTER Figure 8. For composite Æ up to 16. and many questions remain open: ¯ ¯ ¯ ¯ ¯ Is there a better search strategy for composite Æ ? What are the properties of solutions for Æ ½ and higher? Are there more Æ with multiple solutions? Can a sharper theoretical maximum for odd and composite expressed in a closed form? Are there solutions that use the back side of the mirrors? Æ be .

for example. Reﬂection diagrams for composite Æ up to 9.POINT MIRROR REFLECTION 251 Figure 9. ¯ ¯ ¯ Are there solutions that can’t be put on an Æ -gon grid or that require three-dimensional space? Is there a general theory to the problem? Are there applications to. lasers or burglar alarms? .

Simple probabilistic models appear to model the average behavior of the initial iterates of a “random” positive integer Ò. in a trajectory starting from an integer Ò should be of size Ò ¾´½·Ó´½µµ as Ò and trajectories should be of length at most ½ ÐÓ Ò before reaching ½. The ¿Ü · ½ Conjecture combines simplicity of statement with apparent intractability. In particular. the largest integer occurring . see Oliveira e Silva [19]. etc. 253 ¾ . deﬁned by Ì ´Òµ ¿Ò·½ ¾ Ò if Ò is odd if Ò is even The ¿Ü ·½ Conjecture asserts that any positive integer Ò eventually reaches ½ under iteration by Ì . The predictions of these models are consistent with empirical data for the ¿Ü · ½ function. ¾ ½ 13. This paper surveys results concerning more sophisticated probability models. which predict the behavior of “extreme” trajectories of ¿Ü · ½ iterates. The ¿Ü · ½ problem concerns the behavior under iteration of the ¿Ü · ½ function Ì . it has now been veriﬁed for all Ò ¾ ¼¾ ¢ ½¼½ . Introduction The well-known ¿Ü · ½ problem concerns the behavior under iteration of the ¿Ü · ½ function Ì . For example. deﬁned by ´½¿º½ µ Ì ´Òµ ¿Ò·½ ¾ Ò if Ò is odd if Ò is even The 3x+1 Conjecture asserts that for each positive integer Ò there is some iterate with Ì ´ µ´Òµ ½. Huge numbers of computer cycles have been expended studying it. where Ì ´¾µ ´Òµ Ì ´Ì ´Òµµ. Lagarias Abstract.How Random Are 3x + 1 Function Iterates? Jeffrey C.

´½¿º¾¿µ Ö´Òµ The ¿Ü · ½ Conjecture is made plausible by the observation that ¿Ü · ½ iterates behave “randomly” in some average sense. . Müller [18]. and who has stated that he circulated the problem in the early 1950s [6]. Surveys of the known results on the ¿Ü · ½ problem can be found in Lagarias [13]. [21]). see [22]. each possible binary pattern occurs exactly once in ½ Ò ¾ . LAGARIAS The ¿Ü·½ problem appeared in Martin Gardner’s “Mathematical Games” column in June 1972 [10]. This randomness property can be formalized by taking an input probability distribution on the integers and looking at the probability distribution resulting from iterating the ¿Ü · ½ function some number of times. This basic property was independently discovered by Everett [8] and Terras [21]. i. We let Ø´Òµ denote the largest value reached in the trajectory of Ò. and now more than one hundred papers have been written about the problem. if one takes the uniform distribution on ½ ¾ . i. Prior to that it circulated for many years by word of mouth. For example. Furthermore the resulting pattern is periodic in Ò with period ¾ . It is usually credited to Lothar Collatz. so that ´½¿º¾¾µ Ø´Òµ ×ÙÔ Ì ´ µ ´Òµ Ò ¼ Ó where Ø´Òµ ½ if the sequence of iterates is unbounded. i. The ﬁrst mathematical papers on it appeared around 1976 ([8].e. The total stopping time ½ ´Òµ of Ò is the minimal number of iterates needed to reach ½.e. and Wirsching [25]. The trajectory or forward orbit Ç´Òµ of the positive integer Ò is the sequence of its iterates Ò Ì ´Òµ Ì ´¾µ´Òµ Ì ´¿µ´Òµ Thus Ç´¾ µ ´¾ ½ ¾ ¿½ µ. Ò Ó ´½¿º¾½µ Ì ´ µ´Òµ ½ ½ ´Òµ Ñ Ò ´½¿º¾¼µ Ç´Òµ where ½ ´Òµ ½ if no iterate equals ½.. who studied problems similar to it in the 1930s. To describe iterates of the ¿Ü·½ function we introduce some terminology. Thwaites in 1952. It was also independently discovered by B.. then the distribution of the -vector ½ ½ ´Òµ Ò Ì ´ µ ´Òµ ½ ´ÑÓ ¾µ for ¼ ½ ´Òµ Ó Ú ´Òµ Ò Ì ´Òµ Ì ´ ½µ´Òµ ´ÑÓ ¾µ is exactly uniform..e. Finally we let Ö´Òµ count the fraction of odd iterates in a trajectory up to the point that ½ is reached. C.254 J.

A well-developed theory of pseudorandom number generators shows that even a single bit of pseudorandomness can be inﬂated into an arbitrarily efﬁcient pseudorandom number generator. one expects that ¾ ¾ on average it changes multiplicatively by their geometric mean. [20]. Several heuristic probabilistic models describing ¿Ü ·½ iterates are based on this idea. [24]. Here we survey some recent results on stochastic models. so that the average number of steps to reach 1 should be ´½¿º¾ µ ½ ½ ÐÓ ÐÓ Ò ¾ ¿ ¾½¾ ÐÓ Ò Furthermore. There is excellent numerical agreement of this model’s prediction with ¿Ü · ½ function data. [24]. one expects the iterates to decrease in size and eventually become periodic. The simplest of them assumes that this independent coin-ﬂip behavior persists until ½ is reached under iteration. [5] proposed using the ¿Ü · ½ function as a pseudorandom number generator. [15]. These mathematical models predict that the expected size of a “random” Ò after steps should be about ´ ¿ µ ¾Ò. see [9]. for an ¾ explicit constant ½. On the positive side. Since this is less than ½. which concern extreme behaviors of ¿Ü · ½ iterates. Garner [11] and Korec [12]) they also seem to possess some residual structurelessness. The pseudorandom character of ¿Ü ·½ function iterates can be viewed as the source of the difﬁculty of obtaining a rigorous proof of the ¿Ü · ½ Conjecture. The repeated random walk model results of [15] show almost perfect agreement between empirical data for ¿Ü · ½ function iterates for Ò up to ½¼ ½½. cf. Lagarias [14] and Luby [17]. the number of steps ½ ´Òµ for Ò to iterate to 1 should be nor ¡ ½ Ô mally distributed with mean ½ ÐÓ ¿ ÐÓ Ò and variance ½ ÐÓ Ò. which may be enough for a pseudorandom bit to be extracted. which is ´ ¿ µ½ ¾ . From this viewpoint it becomes interesting to determine how well the properties of ¿Ü · ½ iterates can be described by a stochastic model. et al. In doing so we are in the atypical situation of modeling a purely deterministic process with a probabilistic model. and a branching random walk model for backward iterates of Ì . Cloney. [13]. Since a number Ò is multiplied by ½ if it is even and approximately ¿ if it is odd. Even though ¿Ü · ½ function iterates possess quite a bit of structure (cf. This model has the drawback that . These models consist of a repeated random walk model for forward iterates by Ì .3X + 1 FUNCTION ITERATES 255 One may summarize this by saying that the parity of successive iterates of Ì initially behaves like independent coin ﬂips. see [20]. Lagarias and Weiss [15] studied two different stochastic models for the behavior of ¿Ü · ½ function iterates. obtained in joint work with Alan Weiss and David Applegate.

They compared empirical data with predictions made from stochastic models. Lagarias and Weiss introduced a family of branching random walk models to describe the ensemble of such trees. LAGARIAS it does not model the fact that ¿Ü·½ trajectories are not independent. Various authors have obtained rigorous results in the direction of the ¿Ü ·½ Conjecture. for these probabilistic models the deviation from independence exhibited by ¿Ü ·½ trajectories does not affect the asymptotic maximal length of extremal trajectories. which are deﬁned in Section 3. That is. It remains a challenge to exploit such regularities to obtain new rigorous results on the ¿Ü · ½ problem. consisting of lower bounds for the number of integers Ò below ½. In this model. see [2]. More recently Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees. At present the best rigorous bound of this sort states that for positive ½ or ¾ ´ÑÓ ¿µ one has for all sufﬁciently large Ü. The local structure of backward ¿Ü · ½ iterates can be explicitly described using ¿Ü · ½ trees. Repeated Random Walk Model Lagarias and Weiss [15] studied a repeated random walk model for forward iterates by Ì . indeed actual ¿Ü ·½ trajectories coalesce to form a tree structure. This leads to two conjectures stated in Section 4. More generally. The ¿Ü · ½ Conjecture remains unsolved and is viewed as intractable. These models make a prediction for extreme values of ½ ´Òµ that is shown to coincide with that made by the repeated random walk model. and found small systematic deviations of the distribution of ¿Ü · ½ inverse iterates from that predicted by the branching random walk model above. C. one may a value Ü that have some iterate Ì ´ µ ´Òµ estimate for a positive integer the quantity Ü with some Ì ´ µ´Òµ It has been conjectured that for each positive ½ or ¾ ´ÑÓ ¿µ there is a ´Üµ card Ò Ò ½ Ò Ó positive constant such that ´Üµ Ü for all Ü see Applegate and Lagarias [2] and Wirsching [25].256 J. They observed that the distribution of the largest and smallest number of leaves possible in such trees of depth appears to be narrower than what would be predicted by the model of Lagarias and Weiss. ´Üµ Ü¼ ½ 14. the successive iterates ÐÓ Ì ´ µ´Òµ are modeled .

the limiting graph of these trajectories has a different appearance.1) it is a straight line segment with slope ¼ ¼¾¿½ starting from ´¼ ½µ and ending at ´ ¼µ. For each initial value Ò an entirely new random walk is used. For the extremal trajectories for Ø£ ´Òµ in (14. with probability 1. see Figure 1. The length of such trajectories approaches the limiting value ½ ´Òµ ÐÓ Ì ´ µ ´Òµ ÐÓ Ò ÐÓ Ò ¼ ½ ´Òµ . Large deviation theory also asserts that the graph of the logarithmically scaled trajectories ´½ º µ approaches a characteristic shape. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally. we consider the random variable Ê ¾ ÐÓ ½ ¾ · ¿ ¾ Ö£ ´Òµ ½ ½ ´Òµ Ò Ì ´ µ ´Òµ Ì ´ ½µ´Òµ with ½ ½ ´Òµ Ó Large deviation theory predicts that with probability 1 it satisﬁes Ð Ñ ×ÙÔ Ö£ ´Òµ Ò ½ ³¼ ¼ ¼ ¼ The quantity corresponds to an estimate of the maximum fraction of elements in a ¿Ü ·½ trajectory that are odd. this explains the name “repeated £ random walk model.24)." What is the extremal size of ½ ´Òµ? Large deviation theory shows that with probability one ´½ º½µ The constant ÊÏ is the unique solution with tional equation £ ´Òµ Ð Ñ ×ÙÔ ½ Ò ½ ÐÓ Ò ÊÏ ³ ½ ´ ½ ÐÓ ¿ µ ½ of the func¾ ´½ º¾µ where ´ µ ×ÙÔ ¾ ½ ½ ´½ º¿µ Next. so that Ø£ ´Òµ ÐÓ Ò. let Ø£ ´Òµ equal the maximal value taken during the random walk. The random variable ½ ´Òµ is the step number £ with probability ¾ at which 0 is crossed. For the extremal trajectories for £ ´Òµ in (14. This random variable is ﬁnite with probability one.4).3X + 1 FUNCTION ITERATES 257 ¿ by starting at ÐÓ Ò and taking independent steps of size ÐÓ ¾ or ÐÓ ¾ ½ each. £ and the expected size of ½ ´Òµ is given by (13. Large deviation theory shows that.

Scaled trajectories of Ò maximizing ´ÐÓ ½¼ ·½ (dotted for ½ . solid for Ò ½¼). and ending at ´ ¾µ.258 J. solid for ½¼). and the second with slope ¼ ½ ¿ for ÐÓ Ò ¾½ ÐÓ Ò. Ø´Òµµ ÐÓ Ò in ½¼ Ò . C. maximizing ´Òµ in ½¼ Ò ½¼ ·½ (dotted ¾½ ÐÓ Ò. Figure 2. see Figure 2. starting at ´ ¾µ and ending at ´¾½ ¼µ.5) approximates the union of two line segments. LAGARIAS Figure 1. the ﬁrst with slope ¼ ½¿½ for ½ ÐÓ Ò. starting from ´¼ ½µ. and the graph (14. Scaled trajectories of for ½ .

½ ´Òµ It gives the longest trajectory.91 32. Maximal value of ´Òµ in intervals ½¼ Ò ½¼ ·½.1) with numerical data up to ½¼ ½½ is given in Table 1.5841 0.91 19. Section 5] shows that in ½¼ ½½ trials one should only expect to ﬁnd a largest ´Òµ ¿¼.884.94 31. and the large deviations extremal trajectory is indicated by a dotted line.254 9 4. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ´Òµ ¿ ¾ .456. A comparison of (14.527 Random walks model ½ ´Òµ ´Òµ 70 108 165 214 329 429 592 593 706 755 21.328.77 31.890.38 41.68 Ö´Òµ 0.64 32. an analysis in [15. Vyssotsky [23] has found much larger numbers with high values of ´Òµ. He also found a number around ½¼½½¼ with ´Òµ exceeding ¿ ¼ Ø´Òµ in intervals ½¼ Ò ½¼ ·½ up to ½¼½½ with the prediction of (14.127 8 127. Here Ñ Ü ´Òµ denotes the number of steps taken until the maximum value In Table 2 we present empirical results comparing the maximal value of is reached.511 7 63.180.6091 Table 2. The agreement with the stochastic model is quite good.5927 0.6020 0.769.371. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good. Extensions of the data can be found in Leavens and Vermeulen [16].5857 0. More recently V.5818 0.5741 0.48 18. He found that Ò 12. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1.13 26.527 5 837. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory.5967 0.815 10 13.400. . The large deviations extremal trajectory is indicated by a dotted line.4).728. The deviations from the stochastic model are well within the range that would be expected from the size of the expected standard deviation for such models.527 has ½ ´Òµ ½¾ ½ and ´Òµ ¿ ¾ ½ .24 16.799 6 8. where ´Òµ ÐÓ Ò .6020 0.68 24. ½ ½¼. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale.6026 0.171 4 52.194.266.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6.6030 0.

049.06 4.909 1. 4. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ. LAGARIAS Ò 1 27 2 703 3 9.02 21.790 1.631.65 7.75 5. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿.645 ´Òµ 21.560 1. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ. .83 3. using (15. or 7 ´ÑÓ µ. (See Table 3 in Section 4.000 Ò 13.817.84 19.86 13.511 10 77.511 6 6.59 23. 15.976 1.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½.804.559 Random walks model Table 3.1). Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼. if Ò ¾ or 8 ´ÑÓ µ.24 16.20 6.566.903 2.260 J.897 2. respectively.671 5 704. Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2.528. In this case.86 5.46 2.788 1.362. In iterating the multivalued function Ì ½. there are duplications.32 2.675 7 80. 5. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree.48 12.791 1. where ½ ´½ º½µ Ì ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ. so do not have any signiﬁcant effect on the size of the tree. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth . Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk.391 8 319. In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed.86 7.65 5. these trees have the minimal and maximal number of leaves possible for depth .14 11.03 19. ½ ½¼.663 4 77.) empirically like . C. The set of inverse iterates associated to any Ò has a tree structure.831 9 8.05 19.10 20. In fact.099 1.819 1. see [15] for more explanation.

and where fying £ ´ µ ×ÙÔ ´½ º¿µ ÐÓ ´¾ · ½ ¾ ¿ ¿ ¼ This constant is provably the same constant as (14. and the root node is assigned the value 1. Theorem 4. cf. The random walk at a given leaf node ends at a position Ù on the real line. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½. We assign labels ¼ or ½ to the edges. In order to prove the result (15. where ¬ is the unique positive number satis£ ´¬ µ ¼. where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree. È ÊÏ .1).1]. Lagarias and Weiss [15. satisﬁes (with probability one) ´½ º¾µ ÐÓ Ò£ ´ µ È³ ½ ÐÑ ½ ÐÓ This constant È ¬ ½ . Pruned ¿Ü · ½ trees of depth Ò . Theorem 3.. [15. .e. The branching random walk arises from associating a walk to each path in the tree from a root node to a leaf node at depth .2). starting from root nodes Ò and The problem of determining the largest trajectory to get to 1 under iteration in a ¿Ü · ½ tree is equivalent to studying the leaf in a ¿Ü · ½ tree having the smallest value. where the random walk starts at the origin on the real axis at the root node and takes a step ÐÓ ¾ along an edge labeled ¼ and a step ÐÓ ¾ along an edge ¿ labeled ½. In [15] it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ . it was necessary to determine the distribution of the number of leaves in a tree in the branching process. where the values of the inverse image nodes of a node Ò of value Ò are ¾Ò or ¾Ò ¾¿ depending on whether the tree has one or two branches. and the value assigned to that vertex by the rules above is exactly Ù . i.4] show that the expected size of ÐÓ Ò£ ´ µ.3X + 1 FUNCTION ITERATES 128 64 32 16 8 4 40 20 10 5 13 224 112 56 28 14 7 74 37 261 Figure 3.

the corresponding random walk . 7. 4. they proposed that the normalized extreme quantities remain bounded. C. This can be formalized as Conjecture C . 5. Ê´ µ counts the number of distinct edge-labeled ¿Ü · ½ trees of depth . Let ´½ º¾µ ÐÑ Ò ½ ¼ ´ µ Ò ½ · · Ð Ñ ×ÙÔ · ´ µ ½ Then these quantities satisfy the inequalities ½ Applegate and Lagarias [3] compared this conjecture with predictions derived from one of the branching random walk models of [15]. Since the number of leaves is expected to grow like ´ ¿ µ . Theorem 3. The branching process is a multi-type Galton–Watson process with six types labeled 1. 2. (See Athreya and Ney [4] for a general discussion of such processes. Here the pruned 3x+1 tree Ì ´ µ is the tree of inverse iterates of the ¿Ü · ½ function grown backward to depth . respectively. 8 pictured in Figure 4. Based on this data. from a given root node . LAGARIAS and as ½ the leaf probability distribution is of the form Ï ´ ¿ µ .262 J. Applegate and Lagarias [1] empirically studied the extremal distribution of the number of leaves in a ¿Ü ·½ tree of depth as the root node is varied.) The random walk arises from assigning the labels ¼ and ½ to the edges of the resulting tree. with all nodes that have a label congruent to ¼´ÑÓ ¿µ removed. 16. where ¼ represents the entering vertex being even and ½ represents the entering vertex being odd. Distribution of 3x + 1 Trees Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees. Let Æ ´ µ and Æ · ´ µ denote the minimal and maximal number of leaves. see [15. they studied the normalized extreme values ´½ º½µ ·´ µ ¿ Æ ´ µ Ò ´ µ ¿ Æ ·´ µ Table 3 gives data up to ¿¼.2]. In this table. where Ï is a random variable satisfying ´½ º µ and Û prob Ï Û ´Üµ Ü for ¼ ½ ´Üµ is strictly positive on ´¼ ½µ.

We make Ê´ µ independent draws of a tree of depth generated by this process. (For actual ¿Ü · ½ trees. 1/3 each) 2 8 5 4 1. so that each vertex of the tree corresponds to a random walk ending at some position Ù on the real line. The paper [3] presents more evidence in favor of Conjecture ½These facts are derived using the probability density for Ï given in (15. The number of leaves in actual ¿Ü · ½ trees empirically exhibits less variability than is predicted by this branching random walk model. 4 or 7 1 7 263 (prob. 5 or 8 0 0 0 0 0 0 1 4 7 2 5 8 Figure 4.4). and we consider the value of that vertex to be Ù . ¾ takes a step ÐÓ ¾ along an edge labeled ¼ and takes a step ÐÓ ¿ along an edge labeled ½.) We consider as random variables the smallest and largest number of leaves that occur among this set of trees. choosing the root node uniformly ´ÑÓ µ. respectively largest number of leaves. . Transitions of the branching process [9]. Let Æ ´ µ and Æ · ´ µ denote the expected value of the smallest number of leaves. and it yields a second child by the yields a child by the map Ò ½ multivalued map Ò ¿ ´¾Ò ½µ (edge label 1) if Ò ¾ or 8 mod 9. This value represents the initial value of the ¿Ü · ½ iteration starting from that vertex of the tree. and ¢ for a ﬁxed constant ¢ ½. choosing Ê´ µ Ê´ µ ¿ ·½ so ½ ¢ ¿.3X + 1 FUNCTION ITERATES (prob. The base vertex of the tree is located at the origin. The analogues of normalized extreme values in the models are ´½ º¿µ ´ µ ¿ Æ ´ µ Ò ·´ µ ¿ Æ ·´ µ In [3] it is proved½ that and Ð Ñ ×ÙÔ · ´ µ ÐÑ Ò ½ ´ µ ½ ¼ These results for the branching random walk model do not agree with Conjecture C above. 1/3 each) 2. The parent (bottom) always ¾Ò (edge label 0).

78 2.375 2.633 0.249 0.09 56.255 2.394 0.307 0.83 99.289 0.748 1.778 1.32 17.486 0.500 1.562 0.748 0.360 2.923 1.481 0.762 1.557 1.81 4199.433 10850 20436 38550 72806 137670 260612 493824 936690 1778360 3379372 6427190 12232928 23300652 44414366 84713872 161686324 308780220 100 134 178 237 311 413 548 736 988 1314 1744 2309 3084 4130 5500 7336 9788 114 154 206 274 363 484 649 868 1159 1549 2052 2747 Table 4.34 420.792 1.422 0.33 1.728 1.802 1.500 1. LAGARIAS Ê´ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 µ 4 8 14 24 42 76 138 254 470 876 1638 3070 5766 Æ ´ µ 1 1 1 2 2 3 4 5 6 9 11 16 20 27 36 48 64 87 Æ ·´ µ 2 3 4 6 8 10 14 18 24 32 42 55 74 ¡ ¿ ¿Ü · ½ Function Trees ´ Branching Process µ · ´ µ ´ µ · ´ µ 1.118 2.57 42.742 1.507 0.481 0.211 2.281 0.327 2.375 0.898 1.986 2.292 2.802 1.489 0.310 2.076 2.76 23.390 2.342 0.37 3.491 1.746 1.409 0.898 1.190 2.633 0. C.507 0.054 2.270 2.77 133.481 0.751 1.131 2. Normalized extreme values for ¿Ü · ½ trees and for the branching process.562 0.669 1.486 0.305 0.422 0.62 7.50 315.68 31.271 0.240 0.405 2.03 177.60 747.475 0.688 1.21 5.488 0.738 1.99 13.791 1.490 0.344 2.784 1.753 1.481 0.026 2.273 0.782 1.252 0.911 1.741 1.802 1.501 0.750 1.12 74.534 0.401 0.782 1.747 1.482 0.489 0.268 0. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a ﬁxed node Ò´ÑÓ ¿ ·½µ.869 1.465 0.36 3149.688 1.491 0.758 1.67 0.83 1771.166 2. .488 0.750 0.238 1.300 0.77 2362.75 5599.534 0.419 2.259 0.750 0.738 1.780 1.264 J.746 1.490 0.492 0.534 0.16 4.45 560.352 0.49 9.47 996.232 2.246 0.38 236.451 0.475 0.302 0.265 0.475 0.62 1328.758 1.243 0.255 0.774 1.490 0.

C. [20] D. Krasikov Inequalities. . pp. [16] G. 64 (1995). 371– 384. Lagarias. Mag. Conf. Korec. Lagarias. [13] J. 1996. Merlini. 4. R. Imitation of an Iteration. Applegate and J. (June). Density Bounds for the ¿Ü · ½ Problem I. 55 (1985). [19] T. Monthly 82 (1985). Annals Appl. Muller. [7] J. TreeSearch Method. 231–251. Springer-Verlag. 44 (1994). 42–45. The ¿Ü · ½ Problem. pp. 9–11. in: Cryptology and Computational Number Theory. Complex Systems 1 (1987). pp. 349–360. C. pp. 85–89. pp. (1976). 68. Weiss. Tartani. on Stochastic Processes. 427–438. Pomerance. C. Appl. pp. AMS. Lagarias. Discrete Math. E. 3rd Intl. Athreya and P. Slovaca. Comp.. Lagarias. Proc. 42. Applegate and J. The ¿Ü · ½ Problem: a Quasi-Cellular Automaton. E. Proc. C. Branching Processes. Computers Math. [18] H. and Cellular Automata. A. pp. 1972 Number Theory Conference. Mathematical Games. 25 (1977). Math. of QuFu Normal University. Garner. 115–143. (1995). E. H. Cloney. Density Bounds for the ¿Ü · ½ Problem II. Providence. [6] L. (1999). pp. 58 (1985). Leavens and M. Muriel. Das ¿Ò · ½ Problem. Ney. New York. Ges. pp. pp. 11 (1992). Math. Physics and Geometry-Locarno. [14] J. 12. [9] M. Amer. Applegate and J.. Feix. 2 (1992). 114–118. 1972. Maximum Excursion and Stopping Time Record Holders for the ¿Ü · ½ Problem: Computational Results. Oliveira e Silva. The ¿Ü · ½ Problem and Its Generalizations. 411–426. J. C. Princeton University Press: Princeton. 1990. Vol. [2] D. ¿Ü · ½ Search Programs. Comp. Colorado.. J. 57–64. Prob. On the origin of the ´¿Ò · ½µ Problem (in Chinese). 1990. 3–23. Pseudorandomness and Cryptographic Applications. Scientiﬁc American 226 (1972). [3] D. 64 (1995). [5] T. Hamburg 12 (1991). pp. 3. Luby. Collatz. Y. D. Boulder. and G. 172–176. Generalized Parcal Triangles. Pseudorandom Number Generators in Cryptography and Number Theory. No. pp. of Colorado. Lagarias and A.I. C. Goles. pp. Rawsthorne. 101–117. [17] M. pp. [11] L. [10] Martin Gardner. 24. Math. Mitteilungen der Math.. NJ. Proc.3X + 1 FUNCTION ITERATES 265 References [1] D. [12] I. ed. The ´¿Ü · ½µ ¾ Problem: A Statistical Approach. [4] K. On the Distribution of ¿Ü · ½ Trees. Everett. Comp. Math. B. Math. 79–99. pp. Univ. pp. [15] J. Lagarias. R. Symp. Ò¸ ´¾Ò · [8] C. Advances in Math. C. Iteration of the Number Theoretic Function ´¾Òµ ½µ ¿Ò · ¾. Vermeulen. C. Vichniac. No. On Heights in the Collatz ¿Ò · ½ Problem. 1972. The ¿Ü · ½ Problem: Two Stochastic Models. and R. Experimental Math. Math. 49–52. pp. Natural Science Edition. 229–261. Math. Unpredictable Iterations. Conway. Appl.

Vyssotsky. Acta Arithmetica 30 (1976). The Collatz Problem. pp. [22] B. 72–76. [24] S. Terras. Lecture Notes in Math. No. Math. J. Springer-Verlag. Math. Wirsching. Thwaites. 35–41. 1681. 21 (1985).266 J. [23] V. Appl. [25] G. Bull. LAGARIAS [21] R. 241–252. C. 1998. Inst. . Wagon. New York. pp. A Stopping Time Problem on the Positive Integers. The Dynamical System Generated by the ¿Ò · ½ Function. private communication. My Conjecture. pp. Intelligencer 7 (1985).

Sign up to vote on this title

UsefulNot useful- Classic Brainteasers - Martin Gardener
- Math Puzzles and Games
- Mathematical Magic Tricks
- Math Puzzles Sacred Geometry Mysteries Etc
- Martin Gardner- Wheels,Life and Other Mathematical Amusements
- Martin Gardner - The Colossal Book of Mathematics
- 40 Fun-Tabulous Math Puzzles
- Everybody's Illustrated Book of Puzzles
- GARDNER04
- Martin Gardner - A Die of Another Color
- The Penguine Book of Curious and Interesting Puzzles
- A Friendly Mathematics Competition - 35 Years of Teamwork in Indiana
- Mathematical Puzzles for the Connoisseur
- 54601090 Maths Puzzles
- Maths Puzzles
- Modern Conjurer Magic Trick Ebook
- Mathematics
- Martin Gardner the Last Recreations
- Magic Tricks (Misc.)
- MATHHHHHHH
- Math Puzzles and Games
- Puzzlers' Tribute - A Feast for the Mind - David Wolfe & Tom Rodgers, Eds.
- Test Your Math IQ
- Chinese Maths Olympiad
- The Band
- math
- The Big Book of Mind-Bending Puzzles
- Math Tricks, Brain Twisters and Puzzles by Joseph Degrazia PDF Tam
- Mathemagic - Advanced Math Puzzles

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.