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

# Contents

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

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

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

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

Those sales marked a turning point in Gardner’s career. Those mental plots evolved into imaginative short stories that he sold to Esquire magazine.. “pitchmen’s” novelties. 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.” “I amused myself on nightwatch by thinking up crazy plots. but it has taken world chaos to make his father say that his oldest son is perhaps the sanest of his family... accept Mr. He arrived with a large suitcase full of puzzles! Puzzles had been a hobby with him. RICHARDS for a magic trick or a sales-promotion angle to any one of a half-dozen companies who look to him for specialties. His personal philosophy has been described as a loose Platonism. .” [5] *** Martin Gardner ’36 is a professional [sic] magician. but he doesn’t like being branded.During the past few months a determined outpouring of ideas for booklets on paper-cutting and other tricks. To Gardner’s family his way of life has at last become understandable. but where to park them while he was peregrinating over the globe was a problem.Still ﬂush with musteringout pay.He called up Mr. Gardner was hanging around his alma mater.” said the soft-spoken Gardner. writing and taking an occasional GI Bill philosophy course. and the night is large and full of wonder.Would Mr.. He tours the world pulling rabbits out of hats. and he thinks Plato. His breakcame when he sold a humorous short story called “The Horse on the Escalator” to Esquire magazine. who had the largest collection he had ever heard of.. [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. straight magic and card tricks. 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 asked if he might come out to Christ’s home. might object with sound reason. If he were to rest his thoughts upon one quotation it would be Lord Dunsany’s: “Man is a small thing. The editor invited the starving writer for lunch at a good restaurant.6 D. the University of Chicago. 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.. too. Christ. then based in Chicago.

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

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

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

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

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

80–86. [11] Sally Helgeson. “A Conversation with Martin Gardner. 4. no.” The Greenville [South Carolina] News. January 1982. Impresario of Mathematical Games. “The Puckish High Priest of Puzzles. 38.” Skeptic. 117–119. 3. [21] Lawrence Toppman. [25] Michael Shermer. 3.” Something About the Author. 32–37. “A Mind at Play. [13] Anne Commire. December 6. p. Material taken from the Afterword. pp. 8. 101. 232–244. 1976. 4. pp. no. vol.” Los Angeles Times. April 1978. 10–21. December 5. 1–10. December 1995. “The Magician of Math. vol. March 11. [18] Lynne Lucas.12 D.” Newsweek. vol. vol. 2. 4. no. 1B–2B. March/April 1998. vol. 58.” Skeptical Enquirer. 1976. pp.” Omni. 1993. pp. 89. [26] Kendrick Frazier. no.” Bookletter. [14] Anthony Barcellos. 2.” Linking Ring. pp. 1981. 1981.” Science 81. Dover. “Interview: Martin Gardner. pp. 5. pp. “The Mathematical Gamester. July/August 1981. The Flight of Peter Fromm. p. p. “Every Day.” Washington Post. vol. 1979. 1. 6E. pp. 34–39. 47– 48. [24] Istvan Hargittai. November 16. “Martin Gardner: A Writer of Many Interests. 1997. 40–41. June 20. 1994. 36–40. 1981. “The Math-e-magician of Hendersonville. December 12. “Martin Gardner. “The Annotated Gardner. NC]. 19. [17] Sara Lambert. “[Martin Gardner]. 1979.” Scientiﬁc American.” The Charlotte [North Carolina] Observer. [23] Philip Yam. RICHARDS [10] Hank Burchard. pp. 1E. 1981. pp. [16] Jerry Adler and John Carey. 6. vol.” Mathematical Intelligencer. 4. 10. no. “Mastermind. no. “Magician of the Wonders of Numbers. vol. . [12] John Braun.” Two-Year College Mathematics Journal. pp. [15] Rudy Rucker. [22] Martin Gardner. 66–69. 56–61. December 9. 16. [20] Scot Morris. [19] Lee Dembart. pp.” Time-News [Hendersonville. “A Great Communicator of Mathematics and Other Games: A Conversation with Martin Gardner. “Martin Gardner. pp.

and of course I’ve heard of Martin Gardner. Gardner. Why do you ask? Ambrose: Because the book contains one weird chapter — It is totally unlike anything else that Gardner ever wrote. 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. and Doyle Raymond Smullyan SCENE I . Professor Ambrose: Have you ever read the book Science: Good. He actually advances the thesis that Conan Doyle never wrote the Sherlock Holmes stories—that these stories are forgeries. Byrd: To tell you the truth. 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. rational. He was a very famous science writer of the last century. Bad. He 13 . 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.The year is 2050 A.Ambrose.D. by Martin Gardner? Professor Byrd: No. Byrd: Oh? Ambrose: The chapter is titled “The Irrelevance of Conan Doyle”. and Bogus. no! Gardner correctly pointed out that all the available evidence shows that Doyle remained quite keen and active to the end. I’ve heard of it. I guess the answer is that Doyle went senile in his last years? Ambrose: No.

Byrd: All tight. but now your explanation of the apparent contradiction between Doyle the rationalist and Doyle the crank makes some sense. .14 R. 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. Then. his disturbance went much deeper. I never had the slightest doubt that Doyle wrote the Holmes stories. Ambrose: I’m glad you realize that. Now. 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. he absolutely refused to believe Houdini when he said that there was a perfectly naturalistic explanation for the escapes. 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. Byrd: But now something else puzzles me: Martin Gardner was no fool. So senility is not the explanation. then. SMULLYAN also pointed out that Doyle’s interest in spiritualism started much earlier in life than is generally realized. it’s just as implausible as Gardner’s idea that Doyle never wrote the Holmes stories at all. Doyle insisted that Houdini was lying! If that’s not psychotic paranoia. As I said. when the public was convinced of his rationality. I have no idea who wrote it. 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. what is? Byrd: I guess you’re right. he was surely one of the most interesting writers of the last century. 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.

But that one chapter sticks out like a sore thumb—not just with respect to the other chapters. I believe that he did exist. 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. We can only hope that future research will answer the question of how that strange chapter ever got into the book. they are perfectly consistent in spirit with all the sensible things that Gardner ever wrote. But surely. 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. No. the theory is most improbable. I’ll send you a copy. I don’t go as far as some historians who believe that Martin Gardner never existed. Byrd (After a long pause): I guess you’re right. AND DOYLE 15 but it was certainly not Martin Gardner. he couldn’t possibly have written such a chapter. . A person of Gardner’s caliber could never have written anything like that! Byrd: Now just a minute. And as Holmes wisely said: Whenever we have eliminated the impossible. Will you please explain one thing: How did the chapter ever get there? No. you must be right! It is certainly not conceivable that anyone as rational as Gardner could entertain such a stange notion. but he certainly never wrote that chapter. whatever remains. I can see how an entire book by an alleged author might be a forgery. the more I think about it. 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. But the alternative that Gardner actually wrote that chapter is not just improbable. 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. And so I am forced to the conclusion that Martin Gardner never wrote that chapter. but in relation to all of Gardner’s writings. however improbable must be the truth. The title is ”Gardner and Doyle”. nobody in his right mind could believe that Gardner actually wrote that chapter. In fact. your theory strikes me as most improbable! Ambrose: I agree with you wholeheartedly.AMBROSE. GARDNER. All the other chapters are obviously genuine. It will appear in the June issue of the Journal of the History of Science and Literature. but completely out of the question. Now.

and Doyle”? Professor Cranby: No. He seriously maintained that Conan Doyle never wrote the Sherlock Holmes stories. Cranby: Good God! That’s ridiculous! That’s just as crazy as Gardner’s idea that Doyle didn’t write Holmes. with all his brilliance. Cranby: Oh. certainly.One Hundred Years Later Professor Broad: Did you get my paper. are you familiar with the Ambrose paper. you remember his book. What is it about? Broad: You know the twentieth century writer. Why do you ask? Broad: Well. I think I have just about everything he ever wrote. where did you send it? Broad: To your Connecticut address. Martin Gardner? Cranby: Of course! I’m quite a fan of his. 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. “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. then I won’t get it for a couple of days. What is it about? Broad: Well. Science: Good. Of course. SMULLYAN SCENE II . Bad. Of course Gardner wrote that chapter! Broad: Of course he did! . had an insane streak that simply got worse through the years. Broad: And do you recall the chapter. but not this one. I’m familiar with much of Ambrose’s excellent work. come on now! That’s no objection! It’s obvious that Doyle. 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. and Bogus? Cranby: Oh. ”Ambrose. 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. Gardner.16 R. ”Gardner and Doyle”? Cranby: No.

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

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

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

A common English word will be the result. ´¿µ . exactly one from each row and one from each column.22 J. not often found in squares composed from names. FARRELL Figure 2. This tribute could be continued. Martin Gardner is indeed Mint! Grand! Rare! He has my love. Choose three letters. A different arrangement of the nine letters produces the square shown in Figure 2. Matrix claims this to be an unusual property. Dr. but the point is clear.

Adam delve In the last millisecond. barks out his orders for the week. and a bit more ½¼ meters = Nucleus of a human cell ½¼ ½ meters = Atom’s nucleus Time The Creator.Three Limericks — On Space. form up. the Earth Sun’s System. First thing on Monday. 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 . and Speed Tim Rowett Space Seven steps each ten million to one Describe the whole space dimension The Atom. Light Sun and Earth. Friday night At a minute to twelve Eve spin. You. Cell’s girth Our bodies. Time. seen as an Army Sergeant Major. 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.

4 million mph Galaxy’s speed through the debris of the Big Bang .roughaveragespeedofEarth 700.000 mph turning speed of Galaxy 1. 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) 70.24 T.000mph.

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

While you’re there. Go to ØØÔ »» ÓÑ º ØØºÒ Ø» ÖÓ Ø ÓØØ»ÖÓÐÐº ØÑÐ. Since then the concept has grown. but you can also download it off my website. Oops! My diagram is too big for this book. December 1998. several of the mazes were built as adjuncts to large cornﬁeld mazes. . check out the rest of the site. You might try enlarging it on a copier. I have a long write-up (with pictures) of something called “walk-through mazes-with-rules. The diagram should be at least 6 inches square to have a die roll across it.28 R." The ﬁrst of these walk-through mazes appeared at the Gathering for Gardner in January 1993. In the summer of 1998. ABBOTT Maze Addendum.

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

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

with only a Bible in hand. Therefore . . . . Find a Biblical word ladder from ÀÇÄ to ÏÊÁÌ... .. Puzzle #5. Complete the following Biblical ladder. . . ... (For worshippers of automobiles.....BIBLICAL LADDERS 31 Puzzle #4. (The words will read the same down as they do across..) Construct a 12-step Biblical ladder from ÇÊ Ë (Judges 3 : 28) to ÊÇÄÄË (Ezra 6 : 1).. .. . Finally. 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 . ... .. .. . . a change of pace: Construct ¢ word squares. using only words from the King James Bible verses shown. 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. Here’s one that anybody can do. Puzzle #6. unless you have special resources. which “comes back on itself” in an unexpected way.. .. Of course #5 was too hard...... ... Ë Î Æ. ...

of the Red . . . . . ÏÀÁÄ ÏÀÇÄ ÏÀÇÊ ËÀÇÊ ËÀ Ê . . to requite . . ËÀÇÊ ËÀ Ê . on the . . have. . . ËÈ Ê ËÈ ËÈÁ ËÈÁÌ . that bloweth . . beside Eloth.32 D. . play the . Lord will . of the . merchants. E. . sheep that are even . . KNUTH Answers Puzzle #1. and his . . and . . ËÅÁÌ ËÅÁÌÀ Ë ÁÌÀ ÁÌÀ. . . and ÁÎ Æ doesn’t work. Thus the middle word must be ÏÁÎ Ë. . man his . . . . is turned . which came . . the LORD. . And Moses . . . and . . mischief and . . upon these . . . being between . . . and 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. . . . and 1 Kings 4 : 29 could also be used for ËÀÇÊ . . Yea also. . not lift up a . against two . . . . . within a . . . still avoiding forward steps. with a scab . ( 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. all the . . the sea . . . . . . . . . . The Lord GOD hath . . . Puzzle #3. . . . . . . and I will . . . is of me. . Here’s a strictly decreasing solution: . . a great . . So it must be ÏÁÎ Ë or ÁÎ Æ. against . . . . . every man his ËÏÇÊ ËÏÇÊÆ ËÀÇÊÆ.. . since neither Ã nor ÆÁÃ nor Æ Î nor Æ Î Æ is a word. . in multitude. . was kindled. . . year after . that it is . Pharaoh was . . and his . . . . . by himself. . . created the . . by his ÏÊ ÌÀ ÏÊÇÌÀ ÏÊÇÌ ÏÊÁÌ ÏÀÁÌ . . .. all the . in her . First observe that the word in Luke 17 : 27 must have at least one letter in common with both Æ Ã and ÇÎ Ê. . . . and the step after Æ Ã must be Ï Ã . . . them not. . . . . . . Other words can now be ﬁlled in. . .

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

E. ÏÇÅ Æ ÇÄÁÎ ÅÁÌ Ë Î ÆÌ Æ ËÌË ÄÄË ÉÍ Ä ÄÍ Ê Ä Ê ËÄ È References Martin Gardner. Games. 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]. Chapter 6. 1996).34 D. . The Universe in a Handkerchief: Lewis Carroll’s Mathematical Recreations. KNUTH Puzzle #7. Puzzles. and Word Plays (New York: Copernicus.

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

” “Then an even number times an even number is an odd number. and an odd number times an even number is an even number. and an even number plus an odd number is an odd number.” “An odd number times an odd 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?” “No! It is an even number.Creative Puzzle Thinking Nob Yoshigahara Problem 1: “An odd number plus an odd number is an even number. OK?” “OK.” “An even number plus an even number is an even number. OK?” “Yes. OK?” “Of course.” “No! It is an odd number! I can prove it!” How? Problem 2: Move two matches so that no triangle remains.

6. 66. 3:59:53. . 30. 34. YOSHIGAHARA Problem 4: Arrange the following ﬁve pieces to make the shape of a star. 4. (Ignore the colons. 32. 1:01:01. 50. 60. etc. 62.38 N. 64. 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. 42. 54. 44. 36. (1) Find the closest such times. y Problem 7: The ﬁgure shown here is the solution to the problem of dividing the ﬁgure into four identical shapes. 13:22:31.) These curious combinations occur 660 times a day. (2) Find the two palindromes whose difference is closest to 12 hours. 46. x. Problem 5: Calculate the expansion of the following 26 terms. 2:41:42. For example. 40. 23:55:32. 56. Problem 6: Which two numbers come at the end of this sequence? 2. 52. (3) Find the longest time span without a palindromic time.

and 5 matchsticks long is divided into two parts with equal area using 3 matchsticks. 4.CREATIVE PUZZLE THINKING 39 Problem 9: A right triangle with sides 3. . 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. You are not allowed to turn over any of the pieces. but you may rotate them in the plane.

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

**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.

laying out three across at a time). This is the card in the fourth row and second column. regrouping cards with the designated column being in the middle. Card Tricks. but be sure to sandwich the column that contains the chosen card between the other two columns. for suitably small numbers. Repeat asking which column. Arrange them in seven horizontal rows and three vertical columns. There are two ways to prove this. First of all. card 11. Remove 21 cards from an ordinary deck. which is to say. W. note that. At the end. so 153 is indeed a ﬁxed point. Even Here’s an example that sounds a bit different. Then. Regroup the cards by picking up the cards by whole columns intact. yet meets the two criteria for a black hole. try this one from one of my readers. f (n) each value eventually is “pulled” into the black hole of 153. 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 . Ask somebody to think of one of the cards without telling you which card he (or she) is thinking of. for the black-hole attraction.46 M. by cases or computer check. there are some). but the easier way is to draw a diagram that illustrates where a chosen card will end up next time. ECKER The argument for why it works is similar to the case with the 123 example.. you will need ﬁrst to discover a number that equals the sum of the fourth (or higher) power of its digits. Now re-lay out the cards by laying out by rows (i. Repeat one last time. n. the card chosen must be in the center of the array. and re-dealing out by rows.e. Now ask him (or her) which of the three columns contains the card. To ﬁnd an analagous black hole for larger powers (yes. It’s a classic card trick. and then test to see whether other numbers are drawn to it. But for those who enjoy programs. I’ll omit the proof. 1¿ + 3¿ + 5¿ = 153. for large numbers n. Second.

further iteration with 6174 is pointless: 7641–1467 = 6174. you always reach 6174. Rearrange the digits of your number to form the largest and smallest strings possible. . involve numbers. ¾REC. and subtract. ¾ Kaprekar’s Constant: What a Difference 6174 Makes! Most black holes. nonetheless. generalizes somewhat. Within the total of seven steps. write down the largest permutation of the number. Fall 1992. At that point. No. but this trick. Apply this same process to the difference just obtained. 48. That is. as with the sisyphian strings..e. the smallest permutation (allowing initial zeros as digits).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. don’t take one of the nine numbers with four identical digits). Take any four-digit number except an integral multiple of 1111 (i.

we would have obtained the reciprocal of phi as the limit. That is. This one may defy not only your ability to explain it. but it is also that much more strange at the same time. ECKER Example: Start with 8028. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ Ä¾ Ä ½ ¼ . The Divisive Number 15 Take any natural number larger than 1 and write down its divisors. 1. 21. 34. The ﬁrst two numbers numbers: 1. 89.6 or so. Had we used the ratios ´Òµ ´Ò · ½µ instead. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value.48 M. 2. 55. To get a quick and dirty feel for why phi arises. etc. Your own example may take more steps. including 1 and the number itself. phi. Its divisors are 1. and ´Òµ preceding two elements. Divide both sides of the equation ´Òµ ½µ · ´Ò ¾µ by ´Ò ½µ. In fact. Now take the sum of the digits of these divisors.). Iterate until a number repeats. The black-hole number this time is 15. and these digits sum to 15. 5. Repeat with 8532 to calculate 8532–2358 = 6174. 5. but without using a program. and the difference is 8532. W. but you will reach 6174. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. ½ . 3. 8. The largest permutation is 8820. and 15. 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.618 . are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. but your very equilibrium. More formally. The ratios seem to be converging to a number around 1. the smallest is 0288. 13. Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò call it Ä — as Ò increases. ´½µ ´Ò ½µ · ´Ò ¾µ for integers Ò ¾. This one is a bit more tedious. Its value is roughly 1. 3. but fewer authors use that approach.

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

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

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

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

2 and 1. 6. 6. 10 pints. 3. You have containers that hold 15 pints. 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. 5. 8. 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. It is approximately 2244. My regular racquetball opponent has a license plate whose three-digit part has the following property.. 0. After going 10 miles in a straight line he ﬁnds himself back at camp. I. subtract 1 and you produce the original number. What is the number and what is the next greater number (possibly with more than three digits) having this property? . Find the rule for combining numbers as shown below (e. 63 9 38 33 32 12 88 25 16 18 15 Ü 5 E3. 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. reverse the digits of the result. E9. 0). 25 and 9 combine to give 16) and use the rule to determine Ü. How long does it take? E5. 5. 4. Through transferring liquid among the containers. 8. 4. E6. 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. Calculate the expansion of this 26 term expression: ´Ü µ´Ü µ´Ü µ´Ü µ ´Ü Ýµ´Ü Þ µ . 3. The 15pint container starts out full. and the other two start out empty: (15. Divide the number by 3. 7. Find a nine-digit number made up of 1. 7. g. HESS up and travels due North. and 6 pints.54 R. 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. 2.5 nautical miles from Los Angeles to Honolulu. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu. Find the most efﬁcient solution.

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

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

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

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

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

” 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.) Assume that it takes Sir George 1 minute to pound a stake and that he walks at a constant speed between stakes. “I grant you the land you can walk around in a day. Given that the width of the building is 125 ft and the rubber band must stretch an extra 10 cm to accommodate the building. Imagine a rubber band stretched around the world and over a building as shown below.60 R.851 ft for the radius of the earth. All the land in the convex hull of your stakes will be yours. A billiard ball with a small black dot. HESS H2.) H5.” (The Queen had read a little mathematics. 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. Where is the black dot when the ball returns to its initial resting place? . “Take some of these stakes with you.902. “To reward you for killing the dragon. What is the least number of irrational punches needed to eliminate all points of the plane? H4.” the Queen said to Sir George. on the exact top is resting on the horizontal plane. I. how tall is the building? (Use 20. An irrational punch centered on point P in the plane removes all points from the plane that are an irrational distance from P.” she continued. and be back at your starting point in 24 hours. “Pound them into the ground along your way.

H11. The altitudes of their eight triangular faces. Tennis Paradox. decimal points. (12. (8. 4). (b) The same as part (a) except square roots are allowed.PUZZLES FROM AROUND THE WORLD 61 H6. H8. are all equal to 1. (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. (12. 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? . 4). A good approximation of ¿ ½. exponents. (c) Show that ½ can be approximated with arbitrary accuracy using any two digits if the rules of part (b) are followed. Prove or disprove that line ST is perpendicular to plane ABCD. where 0 p 1. As shown below it is easy to place 2n unit diameter circles in a 2 ¢ n rectangle. (9. using just two digits is H7. 6). You may use · ¢ . What is the smallest value of n for which you can ﬁt 2n + 1 such circles into a 2 ¢ n rectangle? H10. 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. (11. taken from vertices S and T. (7. 3). 2). 4). You are given two pyramids SABCD and TABCD. The server in any game wins each point with a ﬁxed probability p. Two evenly matched tennis players are playing a tiebreak set.n) = (a) (b) (c) (d) (e) (f) (g) (3. and parentheses. No square roots or other functions are allowed. 5). (a) Find the best approximation using two digits of your choice.

A number may be approximated by a /b where a and b are integers. (b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes? .0107 . ) = 0.0107 0.62 R.71828 . (a) Find the minimum a and b so that g (a. Deﬁne the goodness of ﬁt of a /b to x as g (a. b. and proceeds to explain: “I normally ask guests to determine the ages of my three children. Find other such examples. their ages are ” H14. My uncle’s ritual for dressing each morning except Sunday includes a trip to the sock drawer.” Your response is “No need to tell me more. Minimum Cutting Length. An approximation to is 355/113. into ﬁve parts of equal area? (c) an equilateral triangle into ﬁve parts of equal area? H15.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. (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. The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem. b. d. 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). x) = a bx . two brown) and ends up with four socks each Saturday evening. Since Smith missed the problem tonight and Jones missed it at the party two years ago. with g (355. (a) The ﬁgure shows two Pythagorean triangles with a common side where three of the ﬁve side-lengths are prime numbers. where e = H16. ) 0. e ) 2. b) Find the minimum c and d so that g (c. . 113. given the sum and product of their ages. I. I’ll let you off the hook. The drawer starts with 16 socks each Monday morning (eight blue. six black.a . HESS H12. where he (1) picks out three socks at random.0107.

(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. (A spandral is the Lyran unit of length. Cosmic. The inhabitants of Lyra III recognize special years when their age is of the form Ô¾Õ. Its axis of rotation coincides with the spheroid’s small axis. On Lyra III one is a student until reaching a special year immediately following a special year. 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. is unique.331 to 1. 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.) The reporting Venerian was mildly comforted because the distance to its copod was the minimum possible distance between the two. which occurs in a special year that is the fourth in a row of consecutive special years. one then becomes a master until reaching a year that is the third in a row of consecutive special years. where Ô and Õ are different prime numbers.. It (the Venerians are sexless) also reported that the straight line distance between it and its copod was 120 spandrals. It’s internal structure. The outermost of these layers is nearly pure krypton. The planet Alpha Lyra IV is an oblate spheroid. ﬁnally one becomes a sage until death. however. The ﬁrst few such special years are 12. (A copod corresponds roughly to something between a sibling and a rootshoot. 18 and 20. Inc. The common axis of these layers is the planet’s rotational axis. just as one observes for Sol III. is contemplating mining the outer layer. and the next inner layer is an anhydrous fromage. each of homogeneous composition.PUZZLES FROM AROUND THE WORLD 63 H17.) Unfortunately. 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 . being made of a number of coaxial right circular cylindrical layers. 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.

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. (c) Find the smallest Ñ such that the magnitude of ´Ñµ is less than the magnitude of ´Ñ · ½µ. The third student had explored a channel 50 spandrals in length. . was also tired because he had thoroughly dug the entire length of the largest possible channel. undertake the mining venture? H19. 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.. they dug a series of channels in the satellite surface. The ﬁfth student. The second student was very tired from his work. a moon of a-Lyra IV (hieronymous). I. Taurus. (b) Find ´½¼¼¼µ to three signiﬁcant ﬁgures. Inc. Before they were wiped out by a permeous accretion of Pﬁster-gas. Should Cosmic. a rather lazy fellow (a laggard dagger digger?). crossing the channel of the haggard dagger digger. 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. lying along the intersection of a plane with the satellite’s surface. A curious feature of these channels is that each is a complete and perfect circle. The fourth student. HESS the curve of intersection of the krypton/fromage boundary and the planetary surface. had merely walked the 60 spandral length of another channel. 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. He found nothing but the fact that the length of the channel was 30 spandrals.

(3. 0. (11. 6). 6). 8. The license number is 741. 2. E3. (1. so E8. Thus Ü ½½. 5). 4. At ﬁrst glance. 4. E6. 6). 4. 0). (11.654. 0). . Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. 8. (7. 2. Anywhere within 10 miles of the north pole. Ø¾ ¾¾ hours. 0). 0). 6). 0. 6). (0. 10. Instead. 0. 15 moves: (15. (3. ¿ ½ ¼¼¼ are the smallest values satisfying the above equations. 6. 0. 7. One of the terms is ´Ü Üµ. (9.741. 0). 10. (1. 9. (3. The trick here is that 1 knot = 1 nautical mile per hour. 10. the next greater number is 7. ½ ¼½ or ¼ and ½ ¼½. (1. 2). E9. so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. But this is not right because 18 is not the difference of 38 and 16. Thus and Ò ¾½¼ transfers is the minimum. In each case the numbers combine by summing the products of their digits. E7. (7. ¼.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 2). 9).729. 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 . E4. (3. the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. (5. 5. E2. (5. 4. The number is 381. (9.425. 6. 6. ¾ gives Ø E5. E10. 2). 2. 6). 6. . it appears that the rule might be subtraction. 0). and Ü ½ .

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

4. 1). (25. producing pieces C and D. 6). (36. 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. 138. (22. 11. The probability of an acute triangle is ¿ ¾½ or ½ . 0). 11. 0. (31. 5. (a) 32 moves. 0). (19. 0). (19.75 minutes. 1. (31. 3. 3. This is the shortest time interval that can be measured. 0. 11. (26. 1. 6). 0). 10). 156. 6). 11. 4). 2). 0. 0) (b) 40 moves. 1). 124. 0. 0). 11. 4. (30. (18. 6). 0). 1). (24. (23. (23. 10). M2. 0. noting the point that burns last. 10). 0). 10). (30. There are 7 ¢ 6/2 = 21 choices using vertex 1. 158. 2. 0). (30. (29. (19. (33. 10). 1. 0). 126. 0. (36. 0). 9. (29. 10). 3). 10). 0. 11. (36. 0). 0. 148. (28. 1). 134. 1. 0. (34. In 3. ½¿ ½¿ ½ Ê 123. 10.PUZZLES FROM AROUND THE WORLD 67 E16. (21. the probability of a right triangle is ½ ¾½ or . (34. 147. 3. (24. 5. 6). (11. (28. (25. 0. (31. (23. 6. 0). (30. 6). 0. light the other end of D. (21. 11. 11. 0). 4. 1. 10). See the ﬁgure. 0. At the same time. 6). 11. (30. as seen from the list below. 1. When C is consumed. (31. 0. (17. 0). 11. 6). 0). 0). 6. (25. . Burn A from both ends. 6). (18. (29. 3. (30. 0. E17. (24. (33. (20. 0. (37. 0. (22. 0). (26. 9. 11. (24. 146. 10. 6). 10). 2). (24. 1. 5). Solutions to Medium Problems M1. 0). (23. 168. 4). 0). (35. (35. (26. (25. 0. 125. 4. 5. 11. 0. 0. 10). (18. 0). 0). 10). 1). 1. 1. 5). (29. 11. 0. 0). and the time is 8:24. Cut B at that corresponding point. 11. 10). Cut the lace in half. 1). 11. 136. (24. 178. 5. 6. 3). D will ﬁnish burning. 2. start burning C from both ends and D from one end. (25. (35. 6). 0). 10). 4). 0). (29. 0). (31. producing pieces A and B. (18. 167. (29. 145. 6. (19. 6. 0. (23. 0). (27. (22. (23. 6. 1). 128. 11. (25. (20. (37. 11. 0). 4). (25. 0). 0. If the hour hand is exactly on a second mark then the second hand will always be on the 12. 0). 127. (17. 10).

The hand below does the job if the play goes as follows. ¿½ ½¾¼ ÐÒ ¾ ¼ Ü M5. For multiple of 3 for Ô ½. Ô M7. HESS M3. East playing the 9. Ë ¹ À ½¼ Â Ë À ¿¾ ¿¾ ¿¾ ¿¾ Ë ÃÉÂ½¼ À ¹ ¹ ÃÉ Ë ¹ À ÃÉÂ½¼ ÃÉÂ ½¼ M4. we take the limit of ½¾¼´¾Ò ½µ Ò ÓÖ Ò ´Òµ as Ò goes to 0.68 R. ¿. North wins the sixth trick with the 10 of diamonds. The ﬁrst ﬁve tricks are alternate spade and heart ruffs by North and West. with East underrufﬁng North each time. so that ¾Ô · Ô¾ is always a Ô Ô ¿ we have Ô · ½ or . The series can be expressed by a simple continuous function: ´Òµ To get Ü. North next leads a heart which West trumps with the queen. The only real root is Ü ¿ we get the prime 17. The ﬁnal two tricks are won by South’s long diamonds. For ¾. I. Then ¾Ô ¿Ò ½ and Ô¾ ¿Ñ · ½. The next four tricks are won by South with the 5432 of clubs. North and East discard all their diamonds on these tricks. M6. There are six ways as shown below. Ü¿ ¾Ü ¼.

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

Each number is the length of the shorter side of the rectangle. 6S. PS.789. P2S. 2P. Shown below. A nine-rectangle solution is shown below. P4S. 3S. I.123. . P3S. 1. 2PS. 5S. 2P2S. M14. and 3P. M15. P. 2S.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. 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. 4S.465.

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

(15. 13. 13. 12. 13. The number in each triangle indicates its area relative to the other triangles in the ﬁgure. M20. 8). 13. 12. Shown below. The scores for the Good Eggs are (15. 12. 11. 10). 10. 11. (15. 11. 14. 14. 11. 12. Solutions to Hard Problems H1. 12. 12. Humpty Dumpty got a 10 in arithmetic. 14. 8). . 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. I. (15. 7). 6). (15. There are 7 Good Eggs. 9) and (14. 9). 14. (15. There are 5 subjects with 15 marks possible in each.72 R. 10. The points are ´ ¿ ½µ ¾ of the way along each diagonal for a distance of 3.0119 .

BCB rolls along the plane. the point P will have coordinates È Ô Ô Ö × Ò ¾ Ö´ Ó× ½µ ¾ Ö´ Ó× · ½µ ¾. TheÔ cone.300. The distance from B to the distance BC. 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 . height of the building. 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. Relative to the center point of the ball. Three equally spaced punches in a straight line do the job if the square of the spacing is irrational. For a full rotation of the ball on the circle the point P will execute ¾ rotations about the axis. If Sir George travels at speed Ú then × Ú ´½ ¼ Òµ Ò where 1440 is the number of minutes in a day.PUZZLES FROM AROUND THE WORLD 73 H2.1968 Ú ¾ H3. H4. Let Û width of the building. and stretch in the band due to the building. 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. as shown below. Point P will return to the top again as Ô well. 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. 1). Ö earths radius. For Ö ½È ´ ¼ ¿ ¾ ¼ ¾ ½¼ ½ ¼ ¼ ¼ ½ µ its initial position was (0. 0. Imagine a regular polygon of Ò sides after Sir George’s trip. with 159. Its area is Ò× ¾. Consider the ball rotating about the axis AC without slipping. where . H5. .

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

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

They deﬁne the four tangent lines drawn to produce ABCD as shown. (b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved. HESS the shape is È Ô ½ ´¾ · ¾ Ö · Ö.9973967 . H10. Each has a radius ½. There are 7 circles on each end with 105 sets of 3 circles in the middle. H9. I. The ratio È takes on a maximum when Ö ¼¾ ¼ .25 for the entire square or a circle inscribed in µ the square. The maximum ratio È Ö ¼¾ ¼ . Raise S and T appropriately above the . Now consider two such circles of different sizes as shown. 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 circles need to be packed as shown below. 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. Note that the ratio È is 0.76 R. For Ò ½ there is just enough room for 329 circles.

Let È ´ µ be the probability of the server winning the game when the server has points and the receiver has points.3074 1.8163 0. Let Ô the probability of the server winning a point.PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit.8286 0. È ´ ¼ ¼µ ÔÈ ´ ¼ ¿¼µ · ÕÈ ´¿¼ ¼µ.2069 1. È ´ ¼ ¿¼µ Ô · ÕÈ ´ ¼ ¼µ. which leads to Ô ¼ ½ ¿¿ ¼ . ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ . which leads to Ô ¼ ½½ ¿ ¾ . H11. A computer program was written to calculate the probabilities shown in the table below.8027 0. · È ´ ¼ ¿¼µ Õ È ´ ¼ ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . and let Õ ½ Ô. ¾ ¾ Õ´ È ´¿¼ ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ . È ´¼ ¼µ È ´ ¼ ¿¼µ when Ô¿ Ô¾ ¾Ô ½. Ô Õ At a score of Ò games to Ò (Ò ¼ to 5).2511 1. È ´¿¼ ¼µ ÔÈ ´ ¼ ¼µ. È ´¼ ¼µ È ´¿¼ ½ µ when Ô ¾ Ô ¿.3586 Pairing probability 0.7892 0. ·¾ Ô ´½ ½ Õ µ È ´¼ ¼µ .2270 1.7805 0 . Clearly ST is not perpendicular to the plane. This leads to ¾ È ´ ¼ ¼µ Ô¾Ô Õ¾ . ¾ H12. 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. At Ò games to Ò.2801 1.7849 . Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1.

¿ ½¾¼. (b) 605 years. (b) Length ¾ ¼¾½½¾ ¿ as shown. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. Solutions occur for (b) ¾ ½. ´ ¾ ½µ ¾.546. (a) Length ½ ¿ ¾½ ½ ¼ as shown. · ½.444 years. There is only one set of ages that accomplishies this: (5. H14. The product and sum could be achieved by (4. The product and sum two years ago could be achieved by (2. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago.321 is the ﬁrst of 5 in a row. we get (a) (b) H16.042.512. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. H15. ¼ ¼¼ ½ ¼. prime. (c) 17.613. . HESS H13. Using a multiprecision continued fraction program. (c) Unknown. (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ . ¼ ¼½¼ . ¾ ½µ ¾. 15). . which must have been guessed by Smith. (2) the younger two children have different ages. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½. ´ ¾ · ½µ ¾ prime. (d) Yes. H17.093. 16). a=10. 6.78 R. 8. ´ prime. which must have been guessed by Jones two years ago. (a) 45 years.641. I. ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . · ½. . 7. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô . We look for cases where (1) the oldest child is under 20. 12).

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

we expand and deform to a contour the ﬁgure. 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. which still avoids enclosing the poles of Õ´Þ µ. 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. HESS (2) Since ´ ÒµÞ is an entire function in the complex plane. 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 . The only pole of Õ´Þ µ ´Ò·½µÞ within is a double pole at Þ ¼. As trarily large it still encloses the same poles as .80 R. Thus Þ ¾ Õ´Þ µ ´Ò·½µÞ is regular within . I. (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 .

which reduces to ´½ · Ü · Ý µ´Ü Ý µ ´ Ó×´Ò · ½µÝ × Ò´Ò · ½µÝ µ where ´Ü¾ · Ý¾ µ ´Ò·½µÜ . where Þ at ½ · Þ Ü · Ý.4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ Þ µÕ´Þ µ ´½ · Þ µÒ·½. 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.8505482 39.8221528 Ý 7.8790560 20. Then Ý approximately equals ½·ÐÒ´ µ . × ÒÝ Ý we can look for poles in the upper half-plane only and reﬂect each one into the lower half-plane. ¼ for all .6406814 3. not including its pole at Þ ¼.0888430 2.2916783 3.2238350 26.5012690 3. (5) On Þ Ê for sufﬁciently large Ê. The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to inﬁnity. A pole of Õ´Þ µ occurs Þ .0262969 3.1512074 45.6745053 3. Call the Ø such pole in the upper half-plane Þ Ü · Ý . For larger deﬁne ´¾ ·½ ¾µ . Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ.PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ. Since the equation in Ý is even.5432385 32.4614892 13. The table below gives numerical values of the ﬁrst few. Ü 1 2 3 4 5 6 7 (6) The contribution to 2. ¾ × ´Ý · µ Ü Ý .

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

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

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

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

86 R. exist in enantiomorphous pairs. We don’t know how O’Beirne decided whether to give a piece a number less than or greater than 10. . . . Negative (bar sinister?) labels are given to the Bar. The nineteen Hexiamond pieces. GUY Figure 1. with its hook on the left. to the Hook. the multiples of 4 have rotational symmetry through 180 degrees. and the Sphinx. Notice that 0 has the greatest symmetry. a parallelogram drawn in the opposite way from the usual textbook fashion. with its pointer on the left. 3. The Chevron. and each is positioned by its apex. Pieces with odd-numbered labels have no symmetry and. to the Crook. with its handle on the left. to the Signpost. 7. together with those that have rotational symmetry only. and subtracting each of the labels 11 to 19 from 20 we arrive at the labeling in Figure 1. which have no symmetry. ½. sailing to the left. and Lobster each have a single symmetry of reﬂection. By subtracting 10 from the Hexagon. 5. O’Beirne had already suggested a numbering of the pieces from 1 to 19. K. as do pieces 1. but our mnemonics are as follows. and 9. whose head is on the left. . ¿. . The Butterﬂy has the same symmetries as a rectangle: Its position is described by that of its body. Crown. Pieces 4 and 8 have only rotational symmetry and exist in enantiomorphous pairs. The symmetries of the Hexagon are those of the dihedral group D6. which appears to be turning to the left. to the Snake. and the other even numbers have reﬂective symmetries. . to the Yacht.

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

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

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

although some of these turned out to be so. for example. It isn’t very easy to give a good upper bound. (Can you ﬁnd a larger class?) How many solutions are there? A wild guess. Now that you know how to classify solutions. There are 508 essentially different relative positions for the Chevron and Hexagon that have not been proved to be impossible.90 R. Continued: Coding the Crown’s position. but perhaps some reader will be more perspicacious. note that O’Beirne’s ﬁrst solution (Figure 3) is of type Ô Â . based on how rarely duplicates appear. Figure 11. and 1 legal positions . h. There are already more than 4200 in the collection. is about 50. GUY Figure 10. K. We have found 247 of these cases (no fewer than twenty-six bit the dust during the writing of this article). Coloring arguments don’t seem to lead to much restriction. 39.000. Remember that the code doesn’t specify the solution completely. which will be deposited in the Strens Collection in the Library at the University of Calgary. there are respectively 21. and o. With the Chevron in positions g. There are at least ten solutions of class Ô ¼ ¼ . Continued: Coding the Butterﬂy’s position.

Figure 13 shows two of Bert Buckley’s machinemade solutions. the ﬁrst found by John Conway and Mike Guy in 1963. I didn’t think he would be successful. each with as many as 11 of the 19 pieces symmetrically placed. for the Hexagon. the discerning solver will soon wish to specialize. then a graduate student at the University of Calgary.O’BEIRNE’S HEXIAMOND 91 Figure 12. the second by the present writer. 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. suggested looking for solutions on a machine. For example. As we go to press. and solutions are known in all of these cases. . With such a plethora of solutions. In 1966 Bert Buckley. 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. how symmetrical a solution can you get? Figure 14 shows two solutions.

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

found by Mike Guy. here ¿ . and so on . ÔÍ has an inter- . Readers will no doubt discover other curiosities and ﬁnd their own favorites. Figure 17. Figure 16 shows a solution displaying three keystones that can be permuted to give other solutions. nal keystone. ½ Ô¼ Ü ×¾ has three Figure 17. In May 1996 Marc Figure 16. 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.” which can be swapped. Figure 19 is one of at least eight examples that have six pieces meeting at a point. keystones. shows that keystones ´ µ can occur internally and don’t have to ﬁt into a corner.O’BEIRNE’S HEXIAMOND 93 Figure 15. ¼ Ç¼ Ð The most symmetrical solution? rotate ¼¾ ´¿ µ or ¼¾ ´¿ µ and produce two “keystones.

president of Kadon Enterprises and discovered that she markets a set of “iamonds” . Six pieces meet at a point.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. 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. I wrote to Kate Jones. Ô Ü (3-symmetry).717 C 6675 D 7549 E 11.518 A 75. Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to ﬁnd all the solutions.94 R. are 130 195 533 193 377 2214 111. by the Hexagon’s position Ø Ù Total 264 1094 124.489 B 15. Figure 19. K. GUY Figure 18.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. On the contrary. Their numbers. for those who prefer their puzzles to have just one answer. classiﬁed according to the position of the Chevron.

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

1991. II pp. pp. Guy.). Malayan Math. A database for O’Beirne’s Hexiamond. Oxford University Press. Polyominoes: A Guide to Puzzles and Problems in Tiling. of America Spectrum Series. 1965. 104–106 and 152–153. 44: Pentominoes and hexiamonds. New Scientist. [3] George E. O’Beirne. . K. 168–170. Some mathematical recreations I. 9 (1960) pp. see Puzzles and Paradoxes No. submitted. 259 (61-11-02) pp. 97–106. Assoc. Puzzles and Paradoxes. especially pp. Martin. Nabla (Bull. 316–317. [4] Thomas H. Soc.96 R. [5] Marc Paulhus. Math. 144–153. GUY [2] Richard K.

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

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!

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

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

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

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

A longtime Clarke fan. JONES rack with a copy of Imperial Earth caught my eye. My husband’s job took care of our needs. Playing ﬁrst with paper and then with cardboard was not enough. Exploration of the pentominoes’ vast repertoire of tricks was a ﬁne way to spend expatriate time. I bought the book and soon caught pentomino fever. A friend’s suggestion that we “make and sell” that game I had invented popped up just then. a specialty of the city of Shiraz. and soon friends got involved. Fast forward to December 1978. By fall of 1979 a wooden prototype was in hand. This magniﬁcent set invited play. to be shared with friends. and after some preliminary doubts we went for it. It was a thrill to ﬁnd mention of Martin Gardner in the back of the book. Back home I was in limbo. Well. of course that was the way to go! A little research turned up the fact that “pentominoes” was Figure 2. I commissioned a local craftsman to make a set of pentominoes in inlaid ivory. and inevitably a dominotype game idea presented itself to me. but idleness was not my style. Super Quintillions. when the Iranian revolution precipitated the rapid evacuation of Americans. and the pieces turned out to be thick enough to be “solid” pentominoes.108 K. having sold my graphics business and with no career plans for the future. .

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

column. with checkerboarded arrangements (Figure 5). 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). or diagonal. Figure 5. 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). The pieces can form duplicate and triplicate models of themselves. formed with checkers (Figure 4). Colormaze: Square tiles in six colors form double polyomino shapes with no duplicate color in any row. In another use. What is the minimum number of moves to exchange the congruent groups of black and white knights? Figure 4. .110 K.

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

including partitions into multiple shapes (Figure 10). . parallelogram). The pieces can form diagonally doubled pentominoes and hexominoes with colormatching adjacency of tiles (Figure 9). Triangoes: orders 1 and 2 polytans permuted with two or more colors (square. 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. triangle.112 K. JONES Figure 8.

Gallop: On a ¢ ½¾ grid. 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).THOSE PERIPATETIC PENTOMINOES 113 Figure 10. double checkerboard hexomino shapes deﬁned by pawns are transformed with knight’s moves (borrowed from the Leap set). Figure 11. 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).

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

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

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

**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.

Orient them just as indicated. This base is manipulated in a third way.) 3-D Tricky Fold In order to follow the instructions and understand the diagrams. the 3-D Tricky Way. very quick. moves. The numbers in parentheses are on the back side of the base. and the second ﬂattens it again. The other two faces are solid black and solid white. When you pull them down and away from each other.122 R. they are turned over so you see the 5 cells. the other two faces are quite difﬁcult to discover. The ﬁrst renders the base 3-D. while four of the faces are easy to ﬁnd. and. . The four faces are continuous and identical. E. being black and white checkerboard patterns. One result was a puzzle made from a “cross” — a square (of 16 squares) with the four corner squares removed. They indicate that you are to make the base 3-D by pushing two opposite corners down and away from you. involving changing the ﬂat ﬂexagon into a three-dimensional ring and back again. You will make two. exactly as indicated in Figure 16. (Note: The trick fold for constructing the ﬂexagon can be done in a slightly different way that has the four faces not continuous. This puzzle is interesting for two reasons: It is tricky to construct. 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). So reach underneath and hold the two free corners of the 1 cells. 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. one corner in each hand. There is a 6 cell at the bottom of each box (see Figure 16. number the base from one to six. Note the two arrows in Figure 16.

The result is shown in Figure 18. The model is completed. . This creates a new arrangement. Now open the model into a tube — a cube open at both ends.. and the upper inside packet of squares (with 2 and 3 on the outside also) to the right. Faces 5 and 6 are found by the following procedure. the base forming a ﬂat square of four cells.H. Flex it in the usual way to show face 5. pull them away from each other. Life and Other Mathematical Amusements.) Now push down on these corners. 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. and at the same time. 1983. 64–68. Change your grip so you are holding the two inside corners of the 5 cells. Figure 17). Once this happens. (These are the corners opposite the ones you were holding. however. the left and right halves going back away from you. Rather. you will ﬁnd that another number has appeared. You can correct this easily by tucking the wrong cell out of sight. Mountain fold the model in half on the vertical axis. Freeman & Co. You can ﬁnd the faces numbered 1 to 4 by the usual ﬂexing. or a 3 instead of a 4.¾ Flexing. you should have four 1 cells facing you. The 3-D boxes will collapse. then 3. Check both sides. a 2 instead of a 1. Now the base should be ﬂattened into a compact model with four cells showing on a side. Collapse the tube in the opposite way. New York: W. and four 4 cells on the other side. move the lower inside packet of squares (with 2 and 3 on the outside) to the left. and in Martin Gardner’s Wheels. Do not open the model as you do when ﬂexing. Begin with face 1 on the top and face 4 on the bottom. Sometimes. replacing it with the proper one.SELF-DESIGNING TETRAFLEXAGONS 123 Figure 17. pp.

I have worked out some variations on the theme of removing no paper. but they are not continuous. and then face 1. Two faces can be found only by forming the rings. . I discovered a way to make a ﬂexagon without removing any portion of the square. Using the 3-D Tricky Fold gives the same results as the Alternating Way Fold. but with poor results. and four faces can be shown: checkerboard. both the third and fourth are derived from the ﬁrst face. Square Plus-Slit About a year or so ago. but they are not continuous. More recently. and repeat the moves just given. Flexing is continuous. Indeed. Four faces can be shown. or rectangle. solid black. Square Slash-Slit (Repeat “slash-slit” out loud quickly at your peril. For best results. with a single slit. E. NEALE Figure 18. The Same Way Fold can be used also. (Alternating Way gets you nowhere. These follow.) It is pleasing that a ﬂexagon can be constructed from a square. as mentioned above. begin on face 4.) It is made from a square with two slits intersecting the center in the shape of a plus sign (Figure 3). Another version of this ﬂexagon can be made by using the Same Way Fold. To ﬁnd face 6. The other two faces are solid white. with face 1 at the back. (The goal was inspired by my desire to make a ﬂexagon from paper money with minimal damage.) Four faces are easily found. checkerboard. All faces are of the identical solid color. solid white. The designs of the four faces are two solid black and two checkerboard.124 R. use the Alternating Way fold.

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

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

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

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

and by that time I was rather tired of the whole thing. Or will it mischievously rise again. as topological puzzles so often do? . When it comes to puzzles. it is often the simplest things that prove to have the greatest appeal.THE ODYSSEY OF THE FIGURE EIGHT PUZZLE 105 received numerous requests for the solution. 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. probably not even realized at the start. perhaps disguised in another form.

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

in 1971 or 1972. 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. 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. I am frequently asked to explain the thought process involved in coming up with a new puzzle.132 R. I neither know or understand the process of any creativity than to say that it just happens. Often it is easier to come up with and develop a new puzzle idea than to give it a good and catchy name. This puzzle idea came to me as I was driving to San Francisco to sell my puzzles at Fisherman’s Wharf. Thanks also to Tom Rodgers for his support of my work. IRBY and easily formed into whatever permanent shapes may be necessary to present a concept. The Bermuda Triangle Puzzle . either you become good at it or you are devoting your mind to the wrong thing. Many thanks must go to Martin Gardner as an inspiration to the millions enlightened by his myriad works. Unfortunately.” and for presenting me with the opportunity to meet Martin Gardner. sometimes it succeeds! If you devote your mind to something. 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.

the UFO being a ring with an abstract shape mated to it. taking the UFO with it as it moves. this one is relatively easy to remember once the solution has been seen. this one came to me fully formed and complete. The triangle can be moved over the entirety of the larger conﬁguration. Most of the large conﬁguration to which the triangle is attached is there simply to bewilder the would-be solver. as I sat playing with it. which is a triangular piece. 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. my wife joined me in critiquing my latest design. 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. The solutions to many puzzles can be elusive until the puzzle has been manipulated many times. although moderately difﬁcult for the average puzzler to solve initially.METAGROBOLIZERS OF WIRE 133 The object of the Bermuda Triangle is to save the UFO that is trapped in the puzzle. The Nightmare . The Bermuda Triangle rates about a medium level of difﬁculty. but the name stuck. The puzzle is generally set up with the ring around the Bermuda Triangle. We didn’t have any nightmares about it. After making the prototype. Sometime later one of us stated that we probably would have nightmares about it that night. 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.

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

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

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

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

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

but reported that his glass sometimes ﬂoated and sometimes sank. Gardner was getting letters in response to the September issue. 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. Gardner was beginning to wonder if his friction explanation told the whole story. Naval Ordnance Laboratory at Silver Spring. You are the inventor.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. where is the rate of the ﬂow of the sand. Reid later adapted his letter for publication and his short article “Weight of an Hourglass” appeared in American . at this point I am hopelessly confused.” 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. A letter dated September 29 came from Walter P. perhaps depending on the temperature. Reid. Hopelessly confused? Martin Gardner?! That is something that surely doesn’t happen very often. In short. One man wrote that he owned one of the Paris cylinders. or perhaps still another one. It is possible that one version of the toy works on the principle you mention. and the other on the principle I suggested. Naval Ordnance Lab. the acceleration of gravity. and the height through which the sand has fallen. I have not yet seen the toy.S. Albert Altman wrote from the U.S. “I am writing to put your mind at ease (on the impact theory posed by Altman). but did not bring one back with him. 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. Well. The principle is simply too clever to be true.” The Hourglass Letters While Hein was having his epiphany in Europe. and to suggest that you not publish a correction. all you have to do is shake your own hand. On September 6th. 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 . also from the U. I am sure that your explanation was correct. having relied (unfortunately) on an account given to me by a friend who examined the toy in Paris. and it didn’t last long.

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

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

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

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

In fact he has recommended that every teacher of algebra construct a set of eight pieces for classroom use. 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. Three “longs” of size ¢ ¢ . Before proceeding with this discussion the reader is urged to build a puzzle and attempt its solution.Cube Puzzles Jeremiah Farrell Binomial Puzzle A new. a single cube). 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. has exactly two solutions — each with a different set of three colors. rather amusing. Without this insight it typically takes several hours to arrive at a solution if one can be found at all. 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. The problem then is to arrange the eight pieces into a cube so that opposite faces have the same color. Three “ﬂats” of size ¢ ¢ . With proper insight it can be solved in a few minutes. This puzzle. This will depend on just what is meant by the words “essentially 145 . 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 ). Now color each piece with six colors according to the scheme in Figure 1.

2.½½ . If you cannot. 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. try again to solve the puzzle. C. and C will turn out to be a solution set. C.146 J. then there are ´ ¿µ´ µ´¿ µ¾ . and 3 are any six distinctive colors. FARRELL Figure 1. 0. B. but there still remain a great many cases that are nearly right. Therefore. Notice that the eight pieces must form the eight corners of the completed cube — one piece for each corner. These hints are to be regarded as progressive. That is. B. additionally. B. 2. Solution Hints. There are 93 solutions in this case. each . after reading a hint. each of the 48 faces of the pieces is colored with a unique color. ½º Call the six colors A. and the solver has the additional clue that a face is all of one color. and 3. If. A. A. proceed to the next hint. The combinatorial puzzle uses only six colors in its construction. Trial and error is not a very fruitful way of trying to solve this puzzle.½ distinct ways of constructing the cube. 0). 1. 1.½ ¼.

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

This will assure that only one solution is attainable with the colors A–B–C (the proof is left to the reader). . column. FARRELL must go by looking at its A–B–C corner. If only one solution is desired. or off-diagonal around all four lateral faces is always 42. The Magic Die has the amazing property that the sum of any row. 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. Figure 3. This will change the parity on one corner of the cube so that a solution is impossible using that set of three colors. The solver will notice that the completed cube has a “fault-free” property. ﬁnally. B. Smally. 2. Then place the longs and. take any piece and interchange two of the colors A.148 J. one that gives only one solution. and C (or two of 1. That is. There is another solution to this puzzle using the colors 1– 2–3. We like to have three Smallys prepared: one that yields two solutions. no seam runs completely through the cube in any direction and therefore no rotation of any part of the cube is possible. and 3). main diagonal (upper left to lower right).

(Alternatively. and.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. it will be extremely difﬁcult to ﬁnd.) . even with the schematic as a guide. There will be only one solution to this puzzle (with magic constant 42). you can copy Figure 3 and paste it onto heavy paper to make a permanent Magic Die.

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

FARHI a difﬁcult exercise for a moderately proﬁcient programmer to verify the results obtained. to separate these cases into disjoint sets of isomorphisms. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. Since one person can assign the fourth and ﬁfth colors to orange and red and another person to red and orange.251. this rule then tells the experimenter to backtrack. The ﬁnal part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. A proof is given in Section 2. in which the number of isomorphisms using the set of thirty-six dicubes is also determined. It becomes necessary. only 133. The individual dicubes can be colored thirty-six different ways. as is shown in Section 5.691 combinations need to be sorted into disjoint sets. none of the nine planes may contain two cubes of the same color. A proof is given in Section 3. only thirty dicubes are required.700 (36!/12!24!) different ways of which. some with the tricube on the edge and some with the tricube through the center of the cube. 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. The previous question now becomes more interesting.677. Twelve of these can be selected in 1. . This was achieved through the use of a computer program. The question then arises: How many solutions are there? Most intriguing about this puzzle.691 combinations into 148 disjoint sets is described in Section 6. Thus. The second observation is that for a cube to be assembled with the nine colors on each face. it becomes apparent that most of the 133. then. obtained after some experimentation. and only 10. the results are shown in Table 1. when two cubes of the same color are in position. Fortunately. and the most difﬁcult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1. Since during the course of trying to solve this puzzle a conﬂict often occurs. as determined in Section 4. The method of sorting the 10. but not in the center of a face. is that there are numerous solutions.105 combinations are isomorphisms.152 S.105 of these combinations meet the three-cubes-of-each-color requirement. the location of the third cube is determined.

:Z[yz 60 60 0 0 0 0 0 .:Zppp 10 10 0 0 0 0 0 .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 .DOeyz 720 312 20 0 12 0 4 . 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 .THE NINE COLOR PUZZLE Table 1.DZ[yz 1439 624 96 4 80 6 15 .COeyz 360 180 8 0 16 16 0 .DZpez 720 312 198 12 83 10 16 . 153 No.

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 . FARHI No.156 S.

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

and on a centerface such as shown by the letter “B. cubes of the same color are never in the same plane. 4. One cube is centrally located and thus not visible. 5.105 Possible Dicube Color Combinations The tricube colors are deﬁned as colors 1. 11. If none of the previous selections include color 3. the center case ﬁlls spaces 5. .” In each case. the remaining two cubes must be located on an edge. there are three possible locations. Since the corner colors are displayed on three faces. The corner case ﬁlls spaces 1. 2. and 14. 2. 3. then a third dicube with color 2 and any color from the second row is selected.” The remaining six corners must have cubes of different colors. The two dicubes with color 1 may use any two colors in the ﬁrst row of the following table. such as shown by the letter “C. accounting for two faces. The “No Two in the Same Plane” Rule Figure 3. If color 3 has been selected once. 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. 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. the edge case ﬁlls spaces 2. If color 2 is selected from the ﬁrst row.” The remaining two colors must all be on edges.158 S. and 11. then color 3 is combined with two colors from the third row. then none are selected from the third row. Determination of the 133. then one color is selected from the third row and if color 3 has been selected twice. and 3. and 11. FARHI If an irregular tricube is used.

This results in requiring only thirty cubes and. (6.691 color combinations. (1. 6. 4. 5. (5. a computer program established 133. Using this algorithm. Similarly if color 2 is not associated with colors 3. 9. 8. 8. given the second numbers 2. 9. etc. and 9 one can logically deduce the ﬁrst colors by the three . Identiﬁcation of the second color of these ordered pairs is sufﬁcient to deﬁne all the colors. 8). 6). 5). 6. 7. color 5 is combined with colors in the ﬁfth row such that there are three cubes with color 5. 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. 9) and (8. or 5. (2. (5. using the same algorithm as before. (7. For example. 6. 7).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. 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.105 possible color combinations. 9). 8). 9). Identiﬁcation and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. 6). 4). 4. only 10. (3. (4. the two new colors are identiﬁed as colors 6 and 7. (7. 5. (4. 3. 3). 2).

Note: The smallest possible number is 23 and the largest is 99. (98). Different solutions often have the same three-digit identiﬁcation. 98. Not all of these mappings produce a new isomorphism. The 10. For example. 3). (5. pairs of three at a time and four at a time with two at a time. 9). The head of the list then became . Any permissible interchange of colors is an isomorphism. The ﬁrst entry. on the ends of the tricube. (7. 91. its isomorphisms are determined and removed from the list. The ASCII code for 23 is not a printable character. An inspection of the ASCII table will explain why it was decided to add 23.160 S.:Z[yz. (35). including pairs of two at a time. Shown in Figure 4 are an edge solution and a center solution for the color combination of Figure 1. Identiﬁcation was further simpliﬁed by placing these numbers in pairs (see Note below). and 99. (5. the ﬁrst entry in Table 1 is . 68. Color 2. which is in the center of the tricube. 90. 9). The solutions are identiﬁed by a three-digit number indicating the color of the buried cube and the two “edge” colors. 35. Decoding a name is quite simple. The ASCII codes for these characters are: 46. 8). These were removed from the list. (4. Colors 1 and 3. The use of this six-character identiﬁcation greatly simpliﬁed the process of sorting the 10. and (8. (4. (7. 67..e. then additional criteria for recording solutions are recommended. . (1. Thus this color combination is identiﬁed as /:Z[yz. 3). can be swapped with each other.:Zppp. . 6). (24). It is then a simple process to recognize that this represents the dicubes colored (1. 9). (4.691 color combinations obtained by the algorithm described in Section 5 were listed in order according to their ASCII characters. The remaining six colors may be interchanged in numerous ways: two at a time. (68). 121. and six at a time including a triplet of two at a time. FARHI cubes of each color requirement and the restriction that the second number be greater than the ﬁrst. This process then identiﬁed the 148 disjoint sets of isomorphisms.691 cases into 148 disjoint sets of isomorphisms. The process was then continued until the list was exhausted. ﬁve at a time including three at a time with two at a time. by adding twenty-three to each and using the ASCII character set codes (ASCII stands for American Standard Code for Information Interchange). three at a time. (2. but not with any of the other colors. i. Subtracting twenty-three results in: 23. If the solutions are to be catalogued. 6). 8). 2). and 122. 7). (67). 58. cannot be interchanged with other colors. 5).:Z[yz. and then assigning a printable character to each. and (99). four at a time. has sixty isomorphisms. (6.

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. the plane can be translated to the other side resulting in a new solution. in French. Typical solutions. . Puzzle sets or puzzle solutions may be obtained by contacting the author by E-mail at sivy@ieee. Often two dicubes may be swapped with two others having the same colors. Stan Isaacs has suggested that graph analysis would be useful in determining the number of solutions. hardcover.. Polycubes. CEDIC. Distracts 4. all 148 graphs have not been compared.. References [1] Slocum. 6) and (7. Whenever the tricube is on an edge and shares a plane with only three dicubes. The edge solution is interesting in that new solutions are often obtainable by using two transformations. 9) on the upper right. 9) in the bottom layer may be swapped with dicubes (4.org. 1977. [2] Meeus. 1978. P. 200 pp. New York. the reader should now be able to determine ﬁve more edge solutions. Using these transformations. Abrams. Jerry. Harry N. J. For example. softcover. and Torbijn.THE NINE COLOR PUZZLE 161 Figure 4. and Botermans. Unfortunately. J. France.. Jack. 176 pp. Paris. dicubes (4. Creative Puzzles of the World. 7) and (6.

in some cases. Blocks A.” The name was chosen because there are two quite different puzzles involving two different blocks numbered 2(a) and 2(b). Description Fixed into the base of the board are four pegs. cannot get there at all.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. The puzzle was invented by Dario Uri from Bologna. one in each corner (see Figure 1). Figure 2. both channels. but only eight are used in each puzzle. Blocks can either have a horizontal channel. one being used in each puzzle. Figure 1. and these blocks can only reach the corners from a horizontal Written by Edward Hordern and reproduced with the kind permission of Dario Uri. 163 . 7. to “Twice. or no channel at all. Italy. 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. Channels (or grooves) are cut into the bases of some blocks. and 2(b) have a single horizontal channel. allowing them to pass over the pegs in the corners (see Figure 2). a vertical channel. There are nine square blocks (including two numbered 2).

The ﬁrst puzzle uses block 2(a). or to get anywhere at all! Hints for Solving In both puzzles the “nuisance” blocks are 3 and 6. Blocks 3 and 6 have no channels and cannot go into any corner. HORDERN direction. can only move freely up and down the center of the puzzle. It is quite an achievement just to solve them. which has to be overcome…. and yet so far! In puzzle 2 it is quite a task to move block A more than a square or two. however. Blocks 4 and 2(a) have a single vertical channel and can only approach the corners from a vertical direction. 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. and the second puzzle uses block 2(b). These blocks. the shortest known solutions are 50 moves for the ﬁrst puzzle and a staggering 70 moves for the second. only to get hopelessly stuck…. as well as block A. Since they can’t go into the corners. they must only move in a cross-shaped area. . Once this has been mastered. a plan can be made as to which blocks have to be moved so as to allow block A to pass. The puzzles are both rated as difﬁcult.164 E. The object of the puzzle in each case is to move block A to the bottom right corner. So near. For the expert. This causes something of a trafﬁc jam. the second being the harder of the two. Both solutions are believed (but not proved) to be minimum-move solutions. The delight of the ﬁrst puzzle is that it is very easy to move block A to just above the bottom right corner. The real “problem” blocks are block 7 in the ﬁrst puzzle and blocks 7 and 2(b) in the second. Blocks 1 and 5 have both channels (in the form of a cross) and can go anywhere. Figure 3. The Puzzle Figure 3 shows the start position.

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

Some months ago.166 M.” 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. . Zigzag. particularly inspired by Muroi’s “Four Sticks” (so closing the circle of mutual inspiration). The three pieces of the puzzle form a square with internal voids. V AN DEVENTER Figure 1. Figure 2. Nine and One-Half Moves. I took up the challenge again and succeeded in ﬁnding a new design of a planar burr. O. which I have called “Nine and One-Half Moves.

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

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

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

¢ . The total CPU time used was about 8500 hours. The programmer can use algorithms that are used for pentominoe problems.450. it is tempting to solve the puzzle by constructing three layers of pieces. ¢ ½¿. The analysis was done on about 20 powerful IBM workstations. ¢ ½½. ½¼ ¢ ½½. the rate increases to 40 pieces/second.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. copied from a model in the collection of Abel Garcia): Box: ¢ ½ ¢ ¾ Pieces: all pieces are of thickness 2 units. ¢ ½¿. ¢ ½¼. and ¢ ½¿. º Boxed Box (23) (Cutler. The result was a fascinating demonstration of how a computer can be used to solve such a puzzle. I did a complete analysis of the third box. 1978. By the end of the runs. not counting rotations and reﬂections. The solution involves use of the following obvious trick (is that an oxymoron?): Some piece(s) are placed sideways in the box. ¢ ½ . ½ ¢ ½ ¢ ¾ ¢ . ¢ ¾½. There are four solutions to the puzzle. the difference between the three boxes is stunning: The ﬁrst box can be completely analyzed in a small fraction of a second. 10 are too wide to ﬁt into the box sideways and 4 are of width 5. ¾¢ . I wrote such a program on my ﬁrst computer. ¾¼ ¢ ¼ ¢ ¾. When running these programs on more powerful computers. ¢ ½ . The program displayed the status of the box at any instant using color graphics. I painted pieces of an actual model to match the display. a Commodore 64.480 solutions. º Parcel Post Puzzle (18) (designer unknown. ¢ . ½ ¢ ¿ ¢ . all very similar. There are 3. and they all have three of these four pieces placed sideways. about once a second a piece is added or removed from the box! Using compiled BASIC. or the equivalent of one year on one machine. The second box was analyzed in about a minute of mainframe computer time. The Commodore 64 is such a wonderously slow machine — when running the program in interpreter BASIC. which is no good for this purpose. the machines had constructed 2 1/2 trillion different partially ﬁlled boxes. ½ ¢ ¾ ¢ ¾. Of the 18 pieces. ¢ ½ . Since all the pieces are of the same thickness and the box depth equals three thicknesses. One or two individual layers can be constructed. the widths and lengths are ¢ . This leaves 4 pieces that might be placed sideways. ½ ¢ ¢ ¼. ½ ¢ ¼ ¢ . . but the process cannot be completed. ¾½ ¢ ¢ ½ ¼. but there are more efﬁcient algorithms that can be used for block-packing puzzles. In early 1996. ½½ ¢ ½½ and two each of ¢ . Bill Cutler Puzzles): Box: ½ ¢ ½ ¢ ½ Pieces: ½¿ ¢ ½½¾ ¢ ½ ½.

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. 1986. The dimensions of the pieces can be modiﬁed. They can be any three different numbers. The smallest number for which this can be done is 23. ½¼ ¢ ½ ¢ ¿½. Bill Cutler Puzzles) Box: ¼ ¢ ¾ ¢ ¾ Pieces: ¢ ½ ¢ ¾ . ¢ ¢ The dimensions of the pieces are all different numbers. See [4]. ¢ . . Pieces Different ½¼º Cutler’s Dilemma. ½ ¢ ½ ¢ ¾½ The original design of Cutler’s Dilemma had 23 pieces and was constructed from the above. ½½ ¢ ½ ¢ ¾ . Pieces the Same or Similar º Hoffman’s Blocks (27) (Dean Hoffman. 1976) Box: ½ ¢ ½ ¢ ½ Pieces: (27) ¢ ¢ This sounds like a simple puzzle. version by cutting some of the pieces into two or three smaller pieces. ¾¿ ¢ ½ ¢ ¿. basic. ¿¿ ¢ ¿ ¢ ¿ ¢ ½¾½. where the smallest is greater than one-quarter of the sum. The net result was a puzzle that is extremely difﬁcult. ½ ¢ ½ ¢ ¾½. ¿ ¢ ¾ ¢ ¼. Holes. Almost 15 years later. Hikimi Puzzland) Box: ½ ¢ ½ ¢ ½ Pieces: Two each of ¢ ¢ ½¼. ¢ ¢ ½½. ¿ ¢ ¾ ¢ ½¾ . ½¼ ¢ ½¾ ¢ ¾ . ¾ ¢ ¿ ¢ . ½¾ ¢ ½ ¢ ½ . ½¼ ¢ ½ ¢ ¾ . I like the dimensions above because it tempts the solver to stack the pieces three deep in the middle dimension. ½½ ¢ ½½ ¢ ¾ . 1981. ½½ ¢ ½ ¢ ½ . ½¼ ¢ ½ ¢ ¾ . ¢ ¢ . but it is not. There are many other 23-piece solutions that are combinatorially different from the above design. this puzzle still fascinates me. CUTLER ¾¾ ¢ ½¼ ¢ ½¿½. ¢ ¼. ¾ ¢ ¿¼ ¢ ¢ ½¿ . ¿ ¢ ½½¼ ¢ ½¿ . 10. º Hoffman Junior (8) (NOB Yoshigahara. ¢ ½¼ ¢ ½½. ½½ ¢ ½¾ ¢ ¾. There are 21 solutions. See [1] or [3] for more information. The pieces ﬁt into the box with no extra space. The extra space makes available a whole new realm of possibilities. ¢ ½¼ ¢ ½½ 11. none having any symmetry or pattern. ¾ ¢ ¢ ½¾¿. Holes. ½¼ ¢ ½½ ¢ ¾. ½½ ¢ ½ ¢ ¿¼. ¿½ ¢ ¢ . Simpliﬁed (15) (Cutler. ¢ ¾¼ ¢ ¾¼.172 B. The box is a cube with side equal to the sum.

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

. 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.

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. “Mechanical Puzzles” is the descriptive term used for what are also known as “Chinese Puzzles. rather than its outward appearance (e. but most attempts so far have either been far too specialized in application or too general to provide the basis for a deﬁnitive classiﬁcation. The disassembly and/or reassembly of the cage remains the primary function of the puzzle. barrel.. or teddy bear may all have similar Cartesian internal construction and so should all be classiﬁed as Interlocking– Cartesian). while examples are given for most groups.” Several attempts have been made to classify mechanical puzzles. books. sphere. Method. An interlocking puzzle should be classiﬁed according to its interior construction. and patents. David Singmaster. To provide a logical and easy-to-use classiﬁcation to enable nonexperts to ﬁnd single and related puzzles in a large collection of objects.) Deﬁnitions.g. some knowledge of the subject is required. A puzzle should be classiﬁed by the problem that its designer intended the solver to encounter while attempting to solve it. contrived for the principle purpose of exercising one’s ingenuity and/or patience. (As presented here. go to http://puzzlemuseum.. and Jerry Slocum. etc. Objective. a wooden cube. 175 . In cases where it seems possible to place a puzzle in more than For updated information and illustrations. Consider a three-dimensional (3-D) interlocking assembly in the form of a cage with a ball in the center. related to such objects. Many people have provided a great deal of help. A mechanical puzzle is a physical object comprising one or more parts that fall within the above deﬁnition. A puzzle is a problem having one or more speciﬁc objectives.com.Classiﬁcation of Mechanical Puzzles and Physical Objects Related to Puzzles James Dalgety and Edward Hordern Background. but particular thanks are due to Stanley Isaacs.

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

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

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

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

” colour mazes.” “Tandem Maze” (complex). “Pike’s Peak” Icosian Game.” “Theo der Turnier. Konigsburg Bridges “Worried Woodworm. avoiding objects. HORDERN Detailed Puzzle Classiﬁcation Class DEXDEX-UNCA Description Dexterity Puzzles Dexterity or other physical skills in their solution Example Cup and Ball. bridges. jumping beans. “Le Pendu.” Tomy’s “Crazy Maze. etc.180 J. Most hedge mazes RTF-2D Route path) mazes 2-D (any . “Bootlegger” (complex) Ring and hole mazes. 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. 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.” Engel’s “Black Box” Bagatelle “Frying Pan. visiting places en route. DALGETY AND E. 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. number totalling mazes.” “Yankee.” puzzles using tops Pentangle “Roly-Poly” puzzles Ramps.

checkerboards. poison rings Nuts and bolts. knives.” 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.” “Hazelgrove Box” Strijbos aluminium burr box Burrs. Soma. gears. cast “ABC. Pentominoes. “Mayer’s Cube. berrocals. 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. puzzle rings Hess wire puzzles. cutlery.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. pens. “Hoffman” cube. Dalgety’s “Devil’s Halo” Leather tanglement puzzles Boxes. “T” puzzle Polycubes. Oscar’s “Dovetail.” “Nine of Swords. Oskar’s keys. “Tak-it-Apart.” Chinese rings. O’Beirne’s “Melting Block” .” Cutler’s burr in a glass Cofﬁn’s “Saturn” JWIP and keychain animals. purses Padlocks Chippendale tea chests.” “Margot Cube.

nine-piece ivory cube Pack the Plums.” “Phoney Baloney. 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. Josephus. “Lapin. “Managon. weaving puzzles “Dodeca” .” Loyd’s Donkeys.” 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. overlapping. Waddington’s “Black Box” Match puzzles. 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. number puzzles Heads and tails “Testa” PAT-STIX PAT-NUM PAT-2DEG PAT-2D PAT-STAK PAT-2D3D Stacking transparent layers.182 ASS-POLY ASS-SHAP 3-D Assembly Polyhedra and Spheres 3D Assembly Other Shapes J.” “Even Steven. DALGETY AND E. pegs. BlackBox. Apple and Worms. HORDERN Ball pyramids. Squashed Soma. Jensen’s “Tricky Laberint” Magic squares. or pieces.

” Chinese balls in ball.” “Hexadecimal. Oscar’s “Solar System. Panel Nine.” Rubik’s “Clock” Rubik’s Cube.CLASSIFICATION OF PUZZLES PAT-3D 3-D pattern puzzles with separate parts 183 “Instant Insanity. 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. Tower of Hanoi..” “Orbit. Fit Pz “Wolf.” Laker Cube Bognar’s planets.” strung cubes Flexagons.” “Spin Out” “Why Knots. counter and peg moving puzzles 15s Pz. Tomy’s “Great Gears” Raba’s “Rotascope.” “Backspin. Rubik’s “Magic.” “Kaos” Rolling eight cubes “The Brain. “Dodecahedru” — rotating faces of Dodecahedron Psychic Pz.” “Turntable Train”.” Mobius strips Rubik’s “Snake. “Masterball.” Waddington’s “Kolor Kraze.” Skor Mor “Instant Indecision. Sheep and Cabbage” Solitaire.” “Flexicube” “Clinch Cube” .” “Missing Link.” “Jugo Flower.

12 Golf Balls. “Jail Nixon” Strip polyhedra Standard “block holes and suck” solution “Jolly Jugs” Cadogan teapots. 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. DALGETY AND E. HORDERN Map folding. or those that can only be represented on a computer . Chinese winepots Royale’s patent Luminations “Columbus Egg” Jugs and liquids. Archimedes gold Cork for three holes.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. rebus plates and prints. ﬁve square puzzle 19th century riddle prints Anagrams.

” Hooper’s paradox Anamorphic pictures Pictures and objects of one subject made from completely unrelated objects Devinettes (obscure outlines). soot on unglazed part of ashtray Landscape turned to make a portrait. Oskar’s “Escher Puzzle. “Spot the Difference.” Penrose triangle “Vanishing Leprechauns.” “Find the 5th Pig” needing coloured overlays.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. OHOs. 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. Topsey Turveys. courtship/matrimony Optical illusions.” 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 . impossible dovetails.

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

suppose Ý. But look.. one of which is twice as great as the other. is certainly 50 (since Ü is then 50). Now. is 100. The two amounts are really the same (i. and the excess of Ý over Ü. is certainly 100. if y is greater than x). if Ý is greater than Ü. I will now prove the following two obviously incompatible propositions. is greater than the excess of y over x. 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. is equal to the excess of y over x. the excess of x over y. 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 x is greater than y. Then the excess of Ü over Ý. if Ü is greater than Ý.A Curious Paradox Raymond Smullyan Consider two positive integers Ü and Ý. if y is greater than x. 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. We are not told whether it is Ü or Ý that is the greater of the two. say. And isn’t 100 surely greater than 50? 189 . Proposition ¾.e. if x is greater than y. The excess of x over y.

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

it is obviously unjust. Finding further theorems of this type. To prove the assertion for all governments. 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.192 S. And since this is true for an arbitrary government. Theorem 3. W. . it is true for all governments. Proof. If a government is arbitrary. is left as an exercise for the reader. All governments are unjust. and extending the method to other models of mathematical proof. We exploit this to obtain the following important result. it is sufﬁcient to prove it for an arbitrary government.

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

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

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

Out from the node labeled Ü the two directed edges go to the nodes labeled ¾Ü and ¾Ü ·½. . two edges directed out and two edges directed in. Ò 197 . has. in which the number of nodes is a power of two. 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). just subtract ¾ Ò from it to bring the number back into the range from ¼ to ¾ Ò ½.Drawing de Bruijn Graphs Herbert Taylor The simplest kind of de Bruijn graph. If ¾Ü or ¾Ü ·½ is bigger than ¾Ò ½. at each node. Figure 1.

while edges out from ¾ Ò ½ to ¾Ò ½ go west. The scheme would be to put ¼ on the North pole. No. Ò . Edges out from ¼ to ¾ Ò ½ ½ on the meridian go east to the dateline. the numbers from ½ to ¾ Ò ¾ in order going south down the Greenwich Meridian. Heawood’s 1891 paper in the Quarterly Journal of Mathematics. available from Aegean Park Press. Golomb. Bigger de Bruijn graphs get really hairy. Pictures up to Ò can be found in the classic “Shift Register Sequences” by Solomon W. February 1980. 2. TAYLOR There is reason to think that the picture for Ò may ﬁll a gap in the existing literature. The earliest reference I know of to the m-pire (empire) problem is P. . J. “Mathematical Games. The picture for Ò suggests 2-pire fashion½ for drawing the de Bruijn graph on the sphere without any lines crossing each other. pp. Figure 2. 14–22. ½In 2-pire fashion each node of the graph is allowed to appear in two places in the picture. ¾Ò ½ on the South pole.” by Martin Gardner. and down the international date line. 242.198 H. Vol. See Scientiﬁc American. because Hal Fredricksen told me he had not seen it in any book.

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

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

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

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

GOSPER Times 100–105: .204 B.

and a period 4 . Hickerson’s Synopsis There have been developments in various areas by different people: Oscillator and spaceship by myself.STRANGE NEW LIFE FORMS: UPDATE 205 So. and I’ve run it occasionally since then. Bill Gosper.” 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. Bill Gosper. Paul Callahan. and myself Large constructions mostly by Buckingham. For a few months I ran it almost constantly. 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. In 1989 I wrote a program to search for oscillators and spaceships. It found many oscillators of periods 3 and 4 (including some smaller than any previously known. David Bell. and myself Oscillator and Spaceship Searches. if you’re wondering “Whatever happened to Life?.

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

The period ½¿ · Ò guns based on Buckingham’s B heptomino turns are also smaller. two gliders delete one pair of beehives. which hit eaters. 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. two such rows.) It’s also possible to build a gun that produces a glider stream of any period ½ . and it begins to decay. centinal-based ¼¼ · ¾¼¼Ò. the cycle repeats every ¢ ¾ generations. New Glider Guns. It emits a glider.584 by 12. with a total of 36 AK47s. its bounding box is 13. Fortunately. Glider Syntheses. there are other ways to get any multiple of 30. 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 uses various mechanisms to interleave larger period streams.STRANGE NEW LIFE FORMS: UPDATE 207 180Æ and collides with the second to form a honey farm. emit gliders that can crash to form MWSSs. it’s based on the “AK47” reaction discovered independently by Dave Buckingham and Rich Schroeppel. and then starts forming another honey farm in a different location. Rich has already mentioned this. When the crystal reaches the guns. based on the period 100 centinal. The period 94 gun is 143 by 607. (Sadly. Subsequent gliders grow a crystal upstream. In the period 94 gun. 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. A honey farm starts to form but is modiﬁed by an eater and a block. The gun itself has a larger period. For example. based on a period 72 device discovered by Bob Wainwright. It’s fairly easy to get any period down . forms a trafﬁc light. you (JHC) described how to thin a glider stream by kicking a glider back and forth between two streams. an eater stops its growth. of size 149 by 149. When they’re all gone. If you delete the trafﬁc light. Using normal kickbacks. 11 gliders add a pair of beehives to the crystal. they don’t work with period 30 streams. I’ve found some eaterassisted kickbacks that give any even period. forming gliders that stabilize the ends of a row. so we can build a long row of them that is unstable at both ends. the process begins again. Buckingham mercifully outmoded it with a comparatively tiny. there’s now a fairly small period 44 gun that Buckingham built recently. The other guns are all pretty big. In addition to the long-familiar period 30 and period 46 guns. the resulting period must be a multiple of 8.112. A close pair of AK47s can delete each other’s trafﬁc lights.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

ﬁnd solutions where the amounts received are in some ratio. which would produce the positive solutions with the smaller price equal to 1. The 1708 English edition of Ozanam [17]. or lowering prices late in the day in an attempt to sell their remainders. The extensively revised and enlarged edition of 1725 (or 1723) outlines a fairly general method and correctly says there are ten solutions. 25. the 1612 and 1624 editions of Bachet [3] ¾ give a fairly general rule for this case and apply it to (20. or simply changing prices after lunch. Even with this trick exposed. 32). distinctly against the spirit of the problem. Tartaglia [19] justiﬁes the different prices by having large and small pearls. This may happen by the sellers going to different markets.g. (12. . and he gives ﬁve solutions for his ﬁrst version — but there are 55. of which 36 are positive. Neither Glaisher nor I have seen the relevant work of De Lagny. 30). which normally are all of the same value. of which there are indeed six. or raising them late in the day when few items are left and new buyers arrive. considers (10. but we both conjecture that he was counting the positive solutions.226 D. 33). 30) and gives two solutions. 50). Although the objects are usually eggs. According to Glaisher [12]. (12. ﬁnd all solutions with one price ﬁxed. which is essentially the same as the ﬁrst edition of 1694. Let us denote the problem with numbers ½ ¾ by ´ ½ ¾ µ. and it is not immediately obvious how to ﬁnd or count all the solutions. I have now arrived at a reasonably simple approach that turns ¾I have not seen these editions.. The question of ﬁnding all solutions for a three-person problem seems to have ﬁrst been tackled in the 1600s. SINGMASTER ‘In like many’ means ‘the same number’ and ‘in like much money’ means ‘the same sum of money. 30. for which he often gives fractional answers. Fibonacci goes on to consider solutions with certain properties — e. is there a solution. Fibonacci only gives two-person versions: (10. he added several more versions as well. During the 1980s. 30. This is historically by far the most common version. 40). 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. I tried several times to ﬁnd an easy way to generate the solutions. so that the above is (10. In 1874.’ Anyone meeting this problem for the ﬁrst time soon realises there must be some trick to it. contrary to De Lagny’s statement that there are six solutions. Fibonacci is one of the few to give more than a single solution. dropping Bachet’s rule and replacing it with some rather vague algebra. or sometimes fruit. Labosne revised the material. such as the ﬁrst receives twice the second. if the prices are also given. He gives a fairly general rule for generating solutions [12]. it can take a fair amount of time to work out a solution.

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

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

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

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

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

55) (55. none given. 1 given.4) (70. SINGMASTER Solutions in Source 5 given. 33) (20. 140) . 1 given. 603. 20) (10. 12. 1300 1300s 1489 1489 1489 c. . One can scale the set . 81) (55. in some way to reduce the solution set to two dimensions. Bhaskara makes the following comment: “This. 21) (225. 30. 16) D. none given. 100) (45. 29. 1 given. 1 given. . 752.232 Version (10. . 1 given. 49) (3. 50) (11. 90) Date 1202? 1202? 1202? c. 40. 1 given with fractions. 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. so we again have a three-dimensional solution set. Sridhara ﬁnds a one-parameter solution set. 50) (10. 60) . 136) (3. 30) (12. 1 given. 30) (18. 81) (10. 454. 2 given. 15) (31.9) (100. 36) (25. 32. 30) (8. 1 given. 20. 20. 68. At this point. 55) (78. 1 given. 1 given. makes this more speciﬁc by saying they ‘sold a part in lots and disposed of the remainder at one for ﬁve panas. 1 given. 37) (27. 32) (12.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 4 given. (20. 196) (45.552. 60) (10.6) # of Solutions (55. 901) (10. 40) (18. 36) (27. 55) (16. 6 given. 60) (117. 40. one with fractions. (20.400) (25. 33. ¬ . 40) (10. 119. 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. 56. 1 given. 1 given. 26) (20. 40. 82) (17. 80) (10. 33) (10. 1 given. 1 given. all given. 20) (64. 25. 1) (7. though it is not as easy to see how to do this as in the Western version. 48. 170) (305. 16) (16. 50) (30. 1) (100. 1 given. 30. 11. 17. 36) (78. 30. 35) (11. 36) (30.

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

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

prob. “Ganita-sâra-sangraha. prob. aka Leonardo Pisano. Sam Loyd’s Cyclopedia of 5. W. 1982. 1228. pp. London. 5. 53 and 346. The Greek Anthology. Abo^ Beqr Mohammed Ben Alhacen. pp. Alkarkh. 850. pp. reprinted 1976. This is the oldest riddle collection printed in England. Boncompagni. Ozanam. Lucknow University. A. 1: Arithmetik und Algebra.. R. 28. book 16. 4 Ø edition. four vols. MS 952. English version: Recreations Mathematical and Physical. Johannes. aka Srîdharâcârya. Mahavira. 44–49 and 94. L. aka al-Karagi. De Gruyter. 60-62. Glaisher. pp. 1696.SOME DIOPHANTINE RECREATIONS 235 [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] 1971. Bonwick. Reprint. 79–80. Tricks and Conundrums. Cambridge. 256r–256v. pp. Edited into French by Franz Woepcke as “Extrait du Fakhrî. Arabe de la Bibliotheque Imperiale. Opera. Geschichte der Elementarmathematik. 136–139.. 1708. 1914. Paris. C. aka Mahâvîrâ(cârya). and Heinemann. Loeb Classical Library. Tartaglia. edited by Grandin. 5. no. Venice. 2 Ò edition. Sam. and Helmuth Gericke. Vol. I. Vol. 3. (1202). 1–131. Paton. Jombert. Paris.000 Puzzles. Mass. 76–77. Pinnacle or Corwin. p. New edition. surviving in a single example in the Cambridge University Library. Metrodorus (compiler). 201–210. et al. 1725. Recreations Mathematiques et Physiques. 1916. Lucknow. Madras. Curtio Troiano. pp. prob. Reprinted by Georg Olms Verlag. 24. Revised by Kurt Vogel. 1010. Karin Reich. Transcribed and translated by K. Shukla. V. edited and published by B. Tropfke.D. 1857. R. pp. Heiberg and H.. VIII. Prob. Menge. Teubner. Jacques. 900. prob. p. Supp. 1923–24. 298–302. 1976. Harvard University Press. 24. Pâtîganita. Sridhara. Hildesheim. J. L. Euclid. 6 of the facsimile. 1853. . Paris. Government Press.” c. Untitled manu uscript called “Alfakhrî. 68–70. Liber Abbaci. pp. 286–287. vol. Sect. 50. c. General Trattato di Numeri et Misure. Part 1.” A. Translated by M. 1556. 1959. “On certain puzzle-questions occurring in early arithmetical writings and the general partition problems with which they are connected. 90. 1912. Vol. Leipzig.” L’Imprimerie Imperiale. Amsterdam. 26–27 of the transcription. Fibonacci.. In: Scritti di Leonardo Pisano. Paris. Edited by Sam Loyd II. S. 53. Rome. London. Nicolo. pp. 1980.” Messenger of Mathematics. 1916–18. Loyd. ex. Translated by W. 1694 (not seen). Ragacarya. Lamb Publishing. Edited by J. pp. Berlin.

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

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

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

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

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

Let be a graph with vertices labeled with 0’s and 1’s in any way. the answer to the Developer’s Dilemma was unknown. 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 . a proof that the Dilemma can be solved will also be presented. Now for some new questions (at least at the time that this paper was submitted). Theorem 2. we assume all lots in the ﬁrst strip of Ñ lots have been sold. The Developer’s Dilemma Suppose a real estate developer has an inﬁnite (at least sufﬁciently long) strip of land Ñ lots wide.242 L. ÝÝÌ ¼ and rank´ · Á µ rank ´ · Á ½µ. If not all neighborhoods of are odd. It is quite possible that some of the buyers are odd. Ñ Further. so Ý Ý Ì ¼. and some are not odd. However. SHADER But is symmetric. Therefore. to all graphs and prove a theorem. But this time the generalization does not yield a solution to the Developer’s Dilemma. 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. for every graph . E. . the vertices of can be labeled with 0’s and 1’s so that every neighborhood in is odd. with all neighborhoods in ¼ odd. We will generalize. Figure 2. Finally. then can be embedded in a graph ¼. as before. so ´ · Á µÜ ½ for every adjacency matrix . 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.

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

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

if these are suitably placed and oriented. 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). Maximum number of reﬂections for two and three point mirrors (E = entrance 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. passing once again by entrance mirror E. If the ray would not return along the entry paths. 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.Point Mirror Reﬂection M. Oskar van Deventer The Problem It is well known that a ray of light can reﬂect many times between two ordinary line mirrors. Figure 1. It is obvious that the strategy of 245 .

see Figure 2. since the light paths are used in two directions. going to every other mirror. so remains valid. the theoretical maximum number of reﬂections is Æ ´Æ ½µ · ½. At mirror P there can be at most Æ ½ reﬂections. Therefore. At mirror E there can be one additional reﬂection for the entering ray. as shown for the regular pentagon in Figure 3. but only Æ ¾ to make it an even number. Only at mirror P can there be an odd number of reﬂections. for even Æ there is a parity complication. Mirrors on the Corners of a Regular Polygon Can the theoretical maxima Ê ½ and Ê¾ be achieved for any Æ . V AN DEVENTER Figure 2. 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. the angle between all the neighboring pairs of light paths is 180 Æ 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. At each mirror a maximum of Æ ½ rays can be reﬂected. Æ ½ from the other mirrors and one from the entering ray. at the other Æ ¾ mirrors (other than the E and P mirrors) there can’t be Æ ½ reﬂections. using a perpendicularly reﬂecting mirror P yields the maximum number of reﬂections. O. from the other mirrors. which is an odd number. Number of reﬂections for different mirror types.246 M.

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

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

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

which can be reached for prime Æ . O.250 M. No general strategy was found. For composite Æ up to 16. Deleted paths for composite Æ up to 16. V AN DEVENTER Figure 8. 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 . solutions were found by a systematic trial and error search using a regular Æ -gon grid.

Reﬂection diagrams for composite Æ up to 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.POINT MIRROR REFLECTION 251 Figure 9. lasers or burglar alarms? . for example.

Huge numbers of computer cycles have been expended studying it. Lagarias Abstract. For example. the largest integer occurring . Introduction The well-known ¿Ü · ½ problem concerns the behavior under iteration of the ¿Ü · ½ function Ì . The predictions of these models are consistent with empirical data for the ¿Ü · ½ function. The ¿Ü · ½ problem concerns the behavior under iteration of the ¿Ü · ½ function Ì . The ¿Ü · ½ Conjecture combines simplicity of statement with apparent intractability. which predict the behavior of “extreme” trajectories of ¿Ü · ½ iterates. This paper surveys results concerning more sophisticated probability models. 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. etc. where Ì ´¾µ ´Òµ Ì ´Ì ´Òµµ. In particular. 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 . 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. ¾ ½ 13. 253 ¾ . Simple probabilistic models appear to model the average behavior of the initial iterates of a “random” positive integer Ò.How Random Are 3x + 1 Function Iterates? Jeffrey C. see Oliveira e Silva [19]. it has now been veriﬁed for all Ò ¾ ¼¾ ¢ ½¼½ .

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. C. This basic property was independently discovered by Everett [8] and Terras [21]. It was also independently discovered by B. [21]). if one takes the uniform distribution on ½ ¾ .e. and Wirsching [25]. and who has stated that he circulated the problem in the early 1950s [6]. LAGARIAS The ¿Ü·½ problem appeared in Martin Gardner’s “Mathematical Games” column in June 1972 [10]. who studied problems similar to it in the 1930s. see [22]. Prior to that it circulated for many years by word of mouth. It is usually credited to Lothar Collatz. each possible binary pattern occurs exactly once in ½ Ò ¾ . 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.e. The ﬁrst mathematical papers on it appeared around 1976 ([8]. 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. Thwaites in 1952. Furthermore the resulting pattern is periodic in Ò with period ¾ . Surveys of the known results on the ¿Ü · ½ problem can be found in Lagarias [13]. i. We let Ø´Òµ denote the largest value reached in the trajectory of Ò. so that ´½¿º¾¾µ Ø´Òµ ×ÙÔ Ì ´ µ ´Òµ Ò ¼ Ó where Ø´Òµ ½ if the sequence of iterates is unbounded. For example. . To describe iterates of the ¿Ü·½ function we introduce some terminology. and now more than one hundred papers have been written about the problem.e. i. The total stopping time ½ ´Òµ of Ò is the minimal number of iterates needed to reach ½... Müller [18]. Finally we let Ö´Òµ count the fraction of odd iterates in a trajectory up to the point that ½ is reached..254 J. 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. ´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. i.

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 concern extreme behaviors of ¿Ü · ½ iterates. [24]. Since this is less than ½. 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. Garner [11] and Korec [12]) they also seem to possess some residual structurelessness. which is ´ ¿ µ½ ¾ . see [9]. In doing so we are in the atypical situation of modeling a purely deterministic process with a probabilistic model. Here we survey some recent results on stochastic models.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. 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. Since a number Ò is multiplied by ½ if it is even and approximately ¿ if it is odd. This model has the drawback that . The simplest of them assumes that this independent coin-ﬂip behavior persists until ½ is reached under iteration. 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. [15]. et al. Even though ¿Ü · ½ function iterates possess quite a bit of structure (cf. [13]. These mathematical models predict that the expected size of a “random” Ò after steps should be about ´ ¿ µ ¾Ò. for an ¾ explicit constant ½. [20]. 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. These models consist of a repeated random walk model for forward iterates by Ì . On the positive side. Cloney. From this viewpoint it becomes interesting to determine how well the properties of ¿Ü · ½ iterates can be described by a stochastic model. Lagarias and Weiss [15] studied two different stochastic models for the behavior of ¿Ü · ½ function iterates. one expects that ¾ ¾ on average it changes multiplicatively by their geometric mean. see [20]. [5] proposed using the ¿Ü · ½ function as a pseudorandom number generator. Several heuristic probabilistic models describing ¿Ü ·½ iterates are based on this idea. which may be enough for a pseudorandom bit to be extracted. cf. [24]. and a branching random walk model for backward iterates of Ì . one expects the iterates to decrease in size and eventually become periodic. obtained in joint work with Alan Weiss and David Applegate. There is excellent numerical agreement of this model’s prediction with ¿Ü · ½ function data.

More recently Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ 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. 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. The local structure of backward ¿Ü · ½ iterates can be explicitly described using ¿Ü · ½ trees. ´Üµ Ü¼ ½ 14. 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.256 J. which are deﬁned in Section 3. 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]. Repeated Random Walk Model Lagarias and Weiss [15] studied a repeated random walk model for forward iterates by Ì . consisting of lower bounds for the number of integers Ò below ½. see [2]. They compared empirical data with predictions made from stochastic models. and found small systematic deviations of the distribution of ¿Ü · ½ inverse iterates from that predicted by the branching random walk model above. The ¿Ü · ½ Conjecture remains unsolved and is viewed as intractable. the successive iterates ÐÓ Ì ´ µ´Òµ are modeled . indeed actual ¿Ü ·½ trajectories coalesce to form a tree structure. C. for these probabilistic models the deviation from independence exhibited by ¿Ü ·½ trajectories does not affect the asymptotic maximal length of extremal trajectories. More generally. That is. This leads to two conjectures stated in Section 4. 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. Lagarias and Weiss introduced a family of branching random walk models to describe the ensemble of such trees. In this model. It remains a challenge to exploit such regularities to obtain new rigorous results on the ¿Ü · ½ problem.

For the extremal trajectories for £ ´Òµ in (14. This random variable is ﬁnite with probability one. this explains the name “repeated £ random walk model. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally." 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. the limiting graph of these trajectories has a different appearance. £ and the expected size of ½ ´Òµ is given by (13.3X + 1 FUNCTION ITERATES 257 ¿ by starting at ÐÓ Ò and taking independent steps of size ÐÓ ¾ or ÐÓ ¾ ½ each. For the extremal trajectories for Ø£ ´Òµ in (14. Large deviation theory also asserts that the graph of the logarithmically scaled trajectories ´½ º µ approaches a characteristic shape. For each initial value Ò an entirely new random walk is used.1) it is a straight line segment with slope ¼ ¼¾¿½ starting from ´¼ ½µ and ending at ´ ¼µ. see Figure 1.4). let Ø£ ´Òµ equal the maximal value taken during the random walk. so that Ø£ ´Òµ ÐÓ Ò. Large deviation theory shows that. The random variable ½ ´Òµ is the step number £ with probability ¾ at which 0 is crossed. The length of such trajectories approaches the limiting value ½ ´Òµ ÐÓ Ì ´ µ ´Òµ ÐÓ Ò ÐÓ Ò ¼ ½ ´Òµ .24). 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. with probability 1.

starting at ´ ¾µ and ending at ´¾½ ¼µ. and ending at ´ ¾µ. Ø´Òµµ ÐÓ Ò in ½¼ Ò . Scaled trajectories of for ½ . and the second with slope ¼ ½ ¿ for ÐÓ Ò ¾½ ÐÓ Ò. see Figure 2. LAGARIAS Figure 1.258 J. Figure 2. C. starting from ´¼ ½µ. and the graph (14. solid for Ò ½¼). solid for ½¼). the ﬁrst with slope ¼ ½¿½ for ½ ÐÓ Ò.5) approximates the union of two line segments. maximizing ´Òµ in ½¼ Ò ½¼ ·½ (dotted ¾½ ÐÓ Ò. Scaled trajectories of Ò maximizing ´ÐÓ ½¼ ·½ (dotted for ½ .

890.1) with numerical data up to ½¼ ½½ is given in Table 1.77 31.194. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale. Vyssotsky [23] has found much larger numbers with high values of ´Òµ. The large deviations extremal trajectory is indicated by a dotted line.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6. 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. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory. 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.456. ½ ½¼.266.5857 0.527 5 837.13 26.5818 0.127 8 127.24 16.5841 0.371.5741 0.6020 0. He found that Ò 12. where ´Òµ ÐÓ Ò .400.799 6 8.527 has ½ ´Òµ ½¾ ½ and ´Òµ ¿ ¾ ½ .884.815 10 13. 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.254 9 4.180.5967 0. The agreement with the stochastic model is quite good.6091 Table 2.6030 0. and the large deviations extremal trajectory is indicated by a dotted line. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good.91 32. Section 5] shows that in ½¼ ½½ trials one should only expect to ﬁnd a largest ´Òµ ¿¼.68 Ö´Òµ 0. Maximal value of ´Òµ in intervals ½¼ Ò ½¼ ·½. an analysis in [15.68 24. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ´Òµ ¿ ¾ .6020 0. More recently V.91 19. Extensions of the data can be found in Leavens and Vermeulen [16]. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1.94 31.64 32.328. A comparison of (14. ½ ´Òµ It gives the longest trajectory.4).6026 0.38 41.171 4 52.5927 0.527 Random walks model ½ ´Òµ ´Òµ 70 108 165 214 329 429 592 593 706 755 21.511 7 63. .728.769.48 18.

511 10 77.24 16. 15. In iterating the multivalued function Ì ½. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿.631.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½. 4. or 7 ´ÑÓ µ.528.788 1.03 19. these trees have the minimal and maximal number of leaves possible for depth . .65 5.75 5.909 1.86 5.391 8 319. respectively.362. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼.976 1. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ.511 6 6.83 3.819 1. In fact. where ½ ´½ º½µ Ì ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½.817.14 11.671 5 704.645 ´Òµ 21. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth . In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed.10 20. LAGARIAS Ò 1 27 2 703 3 9.32 2. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ.559 Random walks model Table 3. if Ò ¾ or 8 ´ÑÓ µ.1). using (15. so do not have any signiﬁcant effect on the size of the tree.566.049.000 Ò 13.48 12.46 2.099 1.) empirically like . (See Table 3 in Section 4.791 1.86 7.897 2. Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2. The set of inverse iterates associated to any Ò has a tree structure. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch.675 7 80. ½ ½¼. In this case. Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk.02 21. see [15] for more explanation.86 13. there are duplications. 5.903 2.790 1.06 4. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ.65 7. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree.560 1.20 6.804.831 9 8.84 19.59 23. C.663 4 77.05 19.260 J.

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. cf. We assign labels ¼ or ½ to the edges. In order to prove the result (15. È ÊÏ . The random walk at a given leaf node ends at a position Ù on the real line. i. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½. it was necessary to determine the distribution of the number of leaves in a tree in the branching process. Theorem 4. where ¬ is the unique positive number satis£ ´¬ µ ¼. and the value assigned to that vertex by the rules above is exactly Ù . 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. Lagarias and Weiss [15. where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree.e.1).2). Pruned ¿Ü · ½ trees of depth Ò .1]. .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 branching random walk arises from associating a walk to each path in the tree from a root node to a leaf node at depth . and where fying £ ´ µ ×ÙÔ ´½ º¿µ ÐÓ ´¾ · ½ ¾ ¿ ¿ ¼ This constant is provably the same constant as (14.. In [15] it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ . 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 root node is assigned the value 1.4] show that the expected size of ÐÓ Ò£ ´ µ. [15. Theorem 3. satisﬁes (with probability one) ´½ º¾µ ÐÓ Ò£ ´ µ È³ ½ ÐÑ ½ ÐÓ This constant È ¬ ½ .

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

1/3 each) 2. ¾ takes a step ÐÓ ¾ along an edge labeled ¼ and takes a step ÐÓ ¿ along an edge labeled ½. and we consider the value of that vertex to be Ù . (For actual ¿Ü · ½ trees. respectively largest 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. The parent (bottom) always ¾Ò (edge label 0). 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.3X + 1 FUNCTION ITERATES (prob. 4 or 7 1 7 263 (prob. The number of leaves in actual ¿Ü · ½ trees empirically exhibits less variability than is predicted by this branching random walk model. The base vertex of the tree is located at the origin. 1/3 each) 2 8 5 4 1. and ¢ for a ﬁxed constant ¢ ½. We make Ê´ µ independent draws of a tree of depth generated by this process. 5 or 8 0 0 0 0 0 0 1 4 7 2 5 8 Figure 4.4). . The paper [3] presents more evidence in favor of Conjecture ½These facts are derived using the probability density for Ï given in (15. Let Æ ´ µ and Æ · ´ µ denote the expected value of the smallest number of leaves. Transitions of the branching process [9]. choosing Ê´ µ Ê´ µ ¿ ·½ so ½ ¢ ¿. choosing the root node uniformly ´ÑÓ µ.) We consider as random variables the smallest and largest number of leaves that occur among this set of trees. This value represents the initial value of the ¿Ü · ½ iteration starting from that vertex of the tree. so that each vertex of the tree corresponds to a random walk ending at some position Ù on the real line.

486 0.36 3149.534 0.81 4199.342 0.75 5599.264 J.405 2.422 0.738 1.37 3.482 0.475 0.344 2.898 1.489 0.780 1.259 0.534 0.742 1.60 747.669 1.271 0.238 1.33 1.270 2.758 1.746 1. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a ﬁxed node Ò´ÑÓ ¿ ·½µ.307 0.34 420.77 133.750 1.47 996.67 0.78 2.21 5.481 0.375 0.38 236.490 0.433 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 4.490 0.481 0.240 0.246 0.409 0.492 0.750 0.68 31.633 0.738 1.310 2.50 315.753 1.475 0.83 1771.57 42. C.688 1.633 0.500 1.211 2.352 0.451 0.746 1.491 0.83 99.03 177.419 2.422 0.750 0.562 0.252 0.534 0.911 1.748 0.758 1.562 0.481 0.869 1.289 0.688 1.488 0.62 1328.501 0.489 0.500 1.490 0.243 0.778 1.792 1.802 1.507 0.475 0.09 56.16 4.076 2.32 17.491 1.802 1.782 1.488 0.054 2.791 1.026 2.481 0.45 560.741 1.507 0. Normalized extreme values for ¿Ü · ½ trees and for the branching process. .62 7.784 1.166 2.76 23.401 0.390 2.986 2.360 2.923 1.305 0.255 2.751 1.748 1.268 0. 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 1.49 9.302 0.77 2362.375 2.300 0.118 2.292 2.728 1.99 13.190 2.747 1.273 0.762 1.802 1.131 2.249 0.486 0.327 2.557 1.281 0.12 74.232 2.265 0.774 1.465 0.255 0.782 1.394 0.898 1.

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

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