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

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

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

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

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

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

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

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

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

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

D. 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. Byrd: Oh? Ambrose: The chapter is titled “The Irrelevance of Conan Doyle”. and Doyle Raymond Smullyan SCENE I .The year is 2050 A. no! Gardner correctly pointed out that all the available evidence shows that Doyle remained quite keen and active to the end. He 13 . He was a very famous science writer of the last century. Byrd: To tell you the truth. I’ve heard of it. this fact has often puzzled me! How could anyone with the brilliance to write the Sherlock Holmes stories ever get involved with spiritualism—and in such a crazy way? Ambrose: You mean that you have doubts that Doyle wrote the Holmes stories? Byrd: Of course not! That thought has never crossed my mind! All I said was that I ﬁnd the situation puzzling. and Bogus. I guess the answer is that Doyle went senile in his last years? Ambrose: No. and of course I’ve heard of Martin Gardner. 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. by Martin Gardner? Professor Byrd: No.Ambrose. 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. Bad. rational. Why do you ask? Ambrose: Because the book contains one weird chapter — It is totally unlike anything else that Gardner ever wrote. Gardner.

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

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

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

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

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

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

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

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

000 mph turning speed of Galaxy 1.000mph.4 million mph Galaxy’s speed through the debris of the Big Bang . 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.roughaveragespeedofEarth 700.

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

Go to ØØÔ »» ÓÑ º ØØºÒ Ø» ÖÓ Ø ÓØØ»ÖÓÐÐº ØÑÐ. but you can also download it off my website. several of the mazes were built as adjuncts to large cornﬁeld mazes.28 R. December 1998. In the summer of 1998. Since then the concept has grown. check out the rest of the site. 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." The ﬁrst of these walk-through mazes appeared at the Gathering for Gardner in January 1993. You might try enlarging it on a copier. Oops! My diagram is too big for this book. While you’re there. ABBOTT Maze Addendum. .

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

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

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

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

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

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

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

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

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

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

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

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

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

618 . but you will reach 6174. 89. Now take the sum of the digits of these divisors. That is. Iterate until a number repeats. including 1 and the number itself. ´½µ ´Ò ½µ · ´Ò ¾µ for integers Ò ¾. 5. 21. and 15. To get a quick and dirty feel for why phi arises. The largest permutation is 8820. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ Ä¾ Ä ½ ¼ . 13. but your very equilibrium. In fact. Repeat with 8532 to calculate 8532–2358 = 6174.48 M. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. and ´Òµ preceding two elements. The ratios seem to be converging to a number around 1. 1. 5. but it is also that much more strange at the same time.). 2. etc. More formally. 34. and these digits sum to 15. phi. but without using a program. The black-hole number this time is 15. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò call it Ä — as Ò increases. we would have obtained the reciprocal of phi as the limit. W. Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . This one may defy not only your ability to explain it. ½ . The ﬁrst two numbers numbers: 1. This one is a bit more tedious. Divide both sides of the equation ´Òµ ½µ · ´Ò ¾µ by ´Ò ½µ. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. the smallest is 0288. 55. Its divisors are 1.6 or so. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. 3. 3. Its value is roughly 1. Your own example may take more steps. and the difference is 8532. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. but fewer authors use that approach. 8. ECKER Example: Start with 8028. Had we used the ratios ´Òµ ´Ò · ½µ instead. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors.

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

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

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

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

4. Find a nine-digit number made up of 1. 5. It is approximately 2244.. g. 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. 7. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu. My regular racquetball opponent has a license plate whose three-digit part has the following property. Find the most efﬁcient solution. You have containers that hold 15 pints. Calculate the expansion of this 26 term expression: ´Ü µ´Ü µ´Ü µ´Ü µ ´Ü Ýµ´Ü Þ µ . and the other two start out empty: (15. 8. 10 pints. E6.54 R. 0. The 15pint container starts out full. 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. 7. After going 10 miles in a straight line he ﬁnds himself back at camp. Divide the number by 3. 63 9 38 33 32 12 88 25 16 18 15 Ü 5 E3. and 6 pints. E9. 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. 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. HESS up and travels due North. 2. 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. 8. 0). How long does it take? E5.5 nautical miles from Los Angeles to Honolulu. Find the rule for combining numbers as shown below (e. reverse the digits of the result. Through transferring liquid among the containers. Where on earth could he be? E2. subtract 1 and you produce the original number. I. 4. 3. 2 and 1. 3. 25 and 9 combine to give 16) and use the rule to determine Ü. What is the number and what is the next greater number (possibly with more than three digits) having this property? . 6. 6. 5.

Find nine single-digit numbers other than (1. A knight is placed on an inﬁnite checkerboard. 2.880. 8. 4.. and D takes 10 minutes. It is night.. Each man walks at a different speed: A takes 1 minute to cross. and 9) with a sum of 45 and a product of 9 362.. and the bridge will hold up to three men at a time. Divide the ﬁgure below into (a) 4 congruent pieces and (b) 3 congruent pieces. and 10 minutes. 9. and 9 minutes and they must cross in 31 minutes. (a) There are six men with crossing times of 1. Four men must cross a bridge. 3. how? Try these other problems. If it cannot move to a square previously visited. E11. The Bridge. 6. E14. . E13. A pair must walk together at the rate of the slower man’s pace.PUZZLES FROM AROUND THE WORLD 55 E10. 3. 8. C takes 5 minutes. Can all four men cross the bridge? If so. (b) There are seven men with crossing times of 1. and they must cross in 25 minutes. 6. A maximum of two people can cross at one time. B takes 2 minutes. and they need their one ﬂashlight to guide them on any crossing. 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. how can you make it unable to move in as few moves as possible? E12. 2. 7. They all start on the same side and have 17 minutes to get across. .

You were playing bridge as South and held 432 in spades. 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. . HESS E15. You are allowed to draw a closed path on the surface of each of the potatoes shown below. 45. In clubs you held 5432. and a pair of scissors. Produce the other hands and a line of play that allows this to occur. 1260. I. Shoelace Clock. a shoelace. 744. 280.56 R. From England comes the series 35. x. hearts. Three points have been chosen randomly from the vertices of a cube. 120. and diamonds. What time is it? E17. Can you draw the two paths so that they are identical to each other in three-dimensional space? E16. 450. (b) a right triangle? M3. 180. 60. You are given some matches. Suppose a clock’s second hand is exactly on a second mark and exactly 18 second marks ahead of the hour hand. Find a simple continuous function to generate the series and compute the surprise answer for Ü. 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. What is the minimum time interval you can measure? Medium Problems M1. The lace burns irregularly like a fuse and takes 60 minutes to burn from end to end. Potato Curves. making a 4-club contract. N thirsty customers arrive. M4. Despite your lack of power you took 6 tricks. What is the probability that they form (a) an acute triangle. one having an 11-pint container and another having a 2Npint container. (b) N = 5? M2. You have a 37-pint container full of a refreshing drink.

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

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

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

” the Queen said to Sir George. I. and be back at your starting point in 24 hours. What is the least number of irrational punches needed to eliminate all points of the plane? H4.” She pointed to a pile of wooden stakes.” she continued. 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. All the land in the convex hull of your stakes will be yours. “To reward you for killing the dragon. “Take some of these stakes with you. P. how tall is the building? (Use 20.) H5. How many stakes should he take with him to get as much land as possible? H3. An irrational punch centered on point P in the plane removes all points from the plane that are an irrational distance from P. A billiard ball with a small black dot.) Assume that it takes Sir George 1 minute to pound a stake and that he walks at a constant speed between stakes. “Pound them into the ground along your way. HESS H2. on the exact top is resting on the horizontal plane. “I grant you the land you can walk around in a day.851 ft for the radius of the earth.60 R.” (The Queen had read a little mathematics.902. Where is the black dot when the ball returns to its initial resting place? . 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.

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

Since Smith missed the problem tonight and Jones missed it at the party two years ago. I’ll let you off the hook. given the sum and product of their ages. and proceeds to explain: “I normally ask guests to determine the ages of my three children. two brown) and ends up with four socks each Saturday evening. An approximation to is 355/113. b) Find the minimum c and d so that g (c. their ages are ” H14. ) = 0. A number may be approximated by a /b where a and b are integers. (a) Find the minimum a and b so that g (a. six black. (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? H13.62 R. d. e ) 2. The drawer starts with 16 socks each Monday morning (eight blue.0107 0. HESS H12.71828 . (a) The ﬁgure shows two Pythagorean triangles with a common side where three of the ﬁve side-lengths are prime numbers. where e = H16. b.a . x) = a bx . . Minimum Cutting Length.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. ) 0.” Your response is “No need to tell me more. I. with g (355. where he (1) picks out three socks at random. My uncle’s ritual for dressing each morning except Sunday includes a trip to the sock drawer.0107 . 113. The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem. into ﬁve parts of equal area? (c) an equilateral triangle into ﬁve parts of equal area? H15. Deﬁne the goodness of ﬁt of a /b to x as g (a.0107. b. 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). (b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes? . Find other such examples.

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

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

8. E3. 6). the next greater number is 7. But this is not right because 18 is not the difference of 38 and 16. ¼. Ø¾ ¾¾ hours. (3. 2. 4. In each case the numbers combine by summing the products of their digits. 0). 4.654. 6). Thus and Ò ¾½¼ transfers is the minimum. (11. (3. so E8. E10. 0). it appears that the rule might be subtraction. (1. At ﬁrst glance. (9. 6. 2.729. (1.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 10. (3. 5). The number is 381. The trick here is that 1 knot = 1 nautical mile per hour. 9. and Ü ½ . E9. 0). One of the terms is ´Ü Üµ. 7. 10. The license number is 741. 4. (11. 0. 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 . (3. (7. E7. (0. (1. Instead. 9). 0. 6. 2). 6. E2. E4. 0). (5. (7.741. 15 moves: (15. ¾ gives Ø E5. the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. 10. 8. 0). (5. . 0. 2. 5. 6). Anywhere within 10 miles of the north pole. 0. 0). E6. Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. 6). so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. 2). 2). . ½ ¼½ or ¼ and ½ ¼½. Thus Ü ½½. ¿ ½ ¼¼¼ are the smallest values satisfying the above equations. 4. 6. (9.425. 6). 6).

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

1. 4). 11. (27. 10). 0). as seen from the list below. 6). 0) (b) 40 moves. 1. noting the point that burns last. (23. (17. 148. 0). (37. 6). (30. 6. 11. 1). 2. 127. (28. 1). (23. 6). 9. When C is consumed. (31. 124. (37. 0. 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. 146. 0. 0). 4. 0). (19. the probability of a right triangle is ½ ¾½ or . 0. 3. 0). 10). (24. (19. 0. 0. 0. 3. 10). 4). 1. 4. (35. 0). (24. 11. (28. (35. 6). 0. 0). 147. (31. 1. (31. If the hour hand is exactly on a second mark then the second hand will always be on the 12. light the other end of D. 4. Cut the lace in half. (30. and the time is 8:24. (30. 3). 0). 0). (31. 0. M2. (23. 6). 10). 11. 0). Burn A from both ends. 11. 11. (21. 0). 0). 3. 1). Cut B at that corresponding point. 10. 5). 0. 6. At the same time. 6). 0). 1. 10). 10). 10). (25. (36. (22. 0). 0). 6. producing pieces A and B. 9. (17. (24. 2). 0. 0. 1). 0. (29. 0). (21. 11. . 10. (34. (18. (29. 10). 3). 136. 1). The probability of an acute triangle is ¿ ¾½ or ½ . (29.75 minutes. (23. 11. 1. 1. 10). 0). (29. 0. 0. (25. 6). (36. (33. E17.PUZZLES FROM AROUND THE WORLD 67 E16. ½¿ ½¿ ½ Ê 123. This is the shortest time interval that can be measured. 1. 0. (20. 11. 5. 178. 6). 0. 10). 5. 4). (36. 0. 0). 0). 0). 145. (19. (33. 0). 5). (29. 168. (22. (26. 0. (24. 6. 126. There are 7 ¢ 6/2 = 21 choices using vertex 1. 2. (20. 1). (34. 11. (18. producing pieces C and D. (25. 11. 11. 128. (31. 11. 2). (25. (24. 0). 0). 5. (25. 4). In 3. 11. 10). 6). 0). (25. (26. (23. Solutions to Medium Problems M1. (11. 6. 4. (19. (22. 5. (18. 167. 0). 3. start burning C from both ends and D from one end. (18. 0. 138. 0. See the ﬁgure. (30. (24. D will ﬁnish burning. 1. 0). 0. 6). 11. 6. 0). 10). 11. (35. 158. (a) 32 moves. 156. (30. 0. 1). (30. 11. 0). 6). (26. (23. 134. 10). 0). 10). (29. (25. 125.

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

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

Shown below. P4S. P3S. 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. . I.789. PS. M14.465. 4S. 3S. Each number is the length of the shorter side of the rectangle.70 R. 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. 2P. 2PS. 6S. 2P2S. 1. P2S. P. A nine-rectangle solution is shown below. M15. 2S. 5S. and 3P.123.

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

11. HESS (b) The two points are along the diagonal near P and its opposite Ô point on the ½ ¢ ½ face that includes A. 11. 8). 12. 13. 14.0119 . M20. (15. The scores for the Good Eggs are (15. 13. 14. 6). 9) and (14. There are 5 subjects with 15 marks possible in each. 7). 10. I. 12. 14. 8). 9). 14. There are 7 Good Eggs. 12. 13. (15. 11. (15. 10). 14. The points are ´ ¿ ½µ ¾ of the way along each diagonal for a distance of 3. 10. Solutions to Hard Problems H1. Shown below. 12. 13.72 R. Humpty Dumpty got a 10 in arithmetic. (15. 12. The number in each triangle indicates its area relative to the other triangles in the ﬁgure. (15. 12. 11. .

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

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

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

The maximum ratio È Ö ¼¾ ¼ . They deﬁne the four tangent lines drawn to produce ABCD as shown.76 R. 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 ratio È takes on a maximum when Ö ¼¾ ¼ . H9. For Ò ½ there is just enough room for 329 circles. Note that the ratio È is 0. Now consider two such circles of different sizes as shown. The point S may be raised above the plane to give unit altitudes if the inscribing circle has radius ½. (b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved. There are 7 circles on each end with 105 sets of 3 circles in the middle. Raise S and T appropriately above the .9973967 . I. HESS the shape is È Ô ½ ´¾ · ¾ Ö · Ö. H10. A view of the pyramid looking down from S perpendicular to the plane ABCD will look like the ﬁgure below. Each has a radius ½.25 for the entire square or a circle inscribed in µ the square.

which leads to Ô ¼ ½ ¿¿ ¼ .8286 0.8163 0.3074 1. Clearly ST is not perpendicular to the plane.7892 0. 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.2270 1. Let È ´ µ be the probability of the server winning the game when the server has points and the receiver has points. and let Õ ½ Ô. · È ´ ¼ ¿¼µ Õ È ´ ¼ ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1. This leads to ¾ È ´ ¼ ¼µ Ô¾Ô Õ¾ . È ´ ¼ ¼µ ÔÈ ´ ¼ ¿¼µ · ÕÈ ´¿¼ ¼µ. È ´¿¼ ¼µ ÔÈ ´ ¼ ¼µ. ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ . Ô Õ At a score of Ò games to Ò (Ò ¼ to 5). Let Ô the probability of the server winning a point.3586 Pairing probability 0.7805 0 . ¾ ¾ Õ´ È ´¿¼ ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ .7849 .2801 1. ·¾ Ô ´½ ½ Õ µ È ´¼ ¼µ .2511 1. At Ò games to Ò.8027 0. È ´¼ ¼µ È ´ ¼ ¿¼µ when Ô¿ Ô¾ ¾Ô ½.PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit. È ´¼ ¼µ È ´¿¼ ½ µ when Ô ¾ Ô ¿. A computer program was written to calculate the probabilities shown in the table below.2069 1. H11. ¾ H12. È ´ ¼ ¿¼µ Ô · ÕÈ ´ ¼ ¼µ. which leads to Ô ¼ ½½ ¿ ¾ .

H14. we get (a) (b) H16. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. which must have been guessed by Smith. .042. H15.613. I. · ½. (b) 605 years. 7. 8. ´ prime. HESS H13. ¿ ½¾¼. (c) Unknown. We look for cases where (1) the oldest child is under 20. ´ ¾ · ½µ ¾ prime.321 is the ﬁrst of 5 in a row. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½.512. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. 15). ¼ ¼¼ ½ ¼.093. ´ ¾ ½µ ¾. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô . H17. ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . ¾ ½µ ¾.78 R. 12). (a) Length ½ ¿ ¾½ ½ ¼ as shown. (2) the younger two children have different ages.546. · ½. prime. 16). (d) Yes. The product and sum two years ago could be achieved by (2. which must have been guessed by Jones two years ago. There is only one set of ages that accomplishies this: (5. (b) Length ¾ ¼¾½½¾ ¿ as shown. The product and sum could be achieved by (4. Using a multiprecision continued fraction program. (a) 45 years. (c) 17. . (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ .641. ¼ ¼½¼ . Solutions occur for (b) ¾ ½.444 years. 6. . and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago. a=10.

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

I. HESS (2) Since ´ ÒµÞ is an entire function in the complex plane. (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 . 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 . 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 ¾.80 R. 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. 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 . Thus Þ ¾ Õ´Þ µ ´Ò·½µÞ is regular within . which still avoids enclosing the poles of Õ´Þ µ. we expand and deform to a contour the ﬁgure.

Since the equation in Ý is even.8505482 39. A pole of Õ´Þ µ occurs Þ . 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. (5) On Þ Ê for sufﬁciently large Ê.2916783 3.2238350 26. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to inﬁnity. which reduces to ´½ · Ü · Ý µ´Ü Ý µ ´ Ó×´Ò · ½µÝ × Ò´Ò · ½µÝ µ where ´Ü¾ · Ý¾ µ ´Ò·½µÜ . The table below gives numerical values of the ﬁrst few. ¼ for all . Ü 1 2 3 4 5 6 7 (6) The contribution to 2. where Þ at ½ · Þ Ü · Ý.4614892 13.PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ.5432385 32.1512074 45.8221528 Ý 7.6406814 3. Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ. Call the Ø such pole in the upper half-plane Þ Ü · Ý . Then Ý approximately equals ½·ÐÒ´ µ .8790560 20. × ÒÝ Ý we can look for poles in the upper half-plane only and reﬂect each one into the lower half-plane.0888430 2.6745053 3. ¾ × ´Ý · µ Ü Ý . The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ.0262969 3.5012690 3. not including its pole at Þ ¼. For larger deﬁne ´¾ ·½ ¾µ .4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ Þ µÕ´Þ µ ´½ · Þ µÒ·½.

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

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

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

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

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

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

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

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

and 1 legal positions . is about 50. note that O’Beirne’s ﬁrst solution (Figure 3) is of type Ô Â .90 R.000. which will be deposited in the Strens Collection in the Library at the University of Calgary. There are already more than 4200 in the collection. Now that you know how to classify solutions. for example. there are respectively 21. (Can you ﬁnd a larger class?) How many solutions are there? A wild guess. There are at least ten solutions of class Ô ¼ ¼ . 39. Figure 11. Coloring arguments don’t seem to lead to much restriction. GUY Figure 10. but perhaps some reader will be more perspicacious. With the Chevron in positions g. It isn’t very easy to give a good upper bound. although some of these turned out to be so. We have found 247 of these cases (no fewer than twenty-six bit the dust during the writing of this article). Continued: Coding the Butterﬂy’s position. and o. based on how rarely duplicates appear. There are 508 essentially different relative positions for the Chevron and Hexagon that have not been proved to be impossible. K. Remember that the code doesn’t specify the solution completely. Continued: Coding the Crown’s position. h.

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

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

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

Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to ﬁnd all the solutions. Figure 19. Six pieces meet at a point.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. for those who prefer their puzzles to have just one answer.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. Ô Ü (3-symmetry). Their numbers. president of Kadon Enterprises and discovered that she markets a set of “iamonds” . I wrote to Kate Jones. are 130 195 533 193 377 2214 111.94 R. by the Hexagon’s position Ø Ù Total 264 1094 124. GUY Figure 18. K. there are no fewer than 40 relative positions of the Chevron and Hexagon that determine such a unique solution: Ð × Ü ½ ¼ Ð Ø ¿ ¿ ¼ Ñ Ø Ü ¿ Ô Ô Ø ½ Ø Ü Ü ¾ ¾ ¼ Ô Õ ¼ Õ Ü Ö Ø ½ Ø Ø ¼ Ø Ü ½ ½ ¾ × Ø Ü Ü How long will it take you to ﬁnd them all without peeking at Marc’s database [5]? After writing a ﬁrst draft of this article. On the contrary.518 A 75.489 B 15.717 C 6675 D 7549 E 11. classiﬁed according to the position of the Chevron.

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

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

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

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!

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

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

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

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

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

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

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

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

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. parallelogram). The pieces can form diagonally doubled pentominoes and hexominoes with colormatching adjacency of tiles (Figure 9).112 K. including partitions into multiple shapes (Figure 10). JONES Figure 8. . Triangoes: orders 1 and 2 polytans permuted with two or more colors (square. Lemma: A matrix of multiple grids lends itself to polyomino packing with checkers in three colors. triangle.

THOSE PERIPATETIC PENTOMINOES 113 Figure 10. 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). 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). Figure 11. Gallop: On a ¢ ½¾ grid. double checkerboard hexomino shapes deﬁned by pawns are transformed with knight’s moves (borrowed from the Leap set). Figure 12. .

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

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

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

**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.

The numbers in parentheses are on the back side of the base. being black and white checkerboard patterns. 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. The four faces are continuous and identical. (Note: The trick fold for constructing the ﬂexagon can be done in a slightly different way that has the four faces not continuous. E. involving changing the ﬂat ﬂexagon into a three-dimensional ring and back again. and two slits in the center that intersect in the shape of a plus sign (see Figure 2). moves. The other cells come together to form two boxes open at the top. exactly as indicated in Figure 16. This base is manipulated in a third way. number the base from one to six. Note the two arrows in Figure 16. . Orient them just as indicated. They indicate that you are to make the base 3-D by pushing two opposite corners down and away from you. You will make two. When you pull them down and away from each other. There is a 6 cell at the bottom of each box (see Figure 16. the 3-D Tricky Way. the other two faces are quite difﬁcult to discover.122 R. and.) 3-D Tricky Fold In order to follow the instructions and understand the diagrams. very quick. and the second ﬂattens it again. This puzzle is interesting for two reasons: It is tricky to construct. they are turned over so you see the 5 cells. The ﬁrst renders the base 3-D. one corner in each hand. So reach underneath and hold the two free corners of the 1 cells. while four of the faces are easy to ﬁnd. One result was a puzzle made from a “cross” — a square (of 16 squares) with the four corner squares removed. The other two faces are solid black and solid white.

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

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

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

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

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

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

perhaps disguised in another form. as topological puzzles so often do? . and by that time I was rather tired of the whole thing. it is often the simplest things that prove to have the greatest appeal. 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. Or will it mischievously rise again.THE ODYSSEY OF THE FIGURE EIGHT PUZZLE 105 received numerous requests for the solution. When it comes to puzzles.

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

A couple of examples of ideas that have “popped” out of my mind at various times are explained and illustrated below. This puzzle idea came to me as I was driving to San Francisco to sell my puzzles at Fisherman’s Wharf. 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. I really can’t answer. 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. 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. in 1971 or 1972. IRBY and easily formed into whatever permanent shapes may be necessary to present a concept. The Bermuda Triangle Puzzle . sometimes it succeeds! If you devote your mind to something. Often it is easier to come up with and develop a new puzzle idea than to give it a good and catchy name.132 R. Thanks also to Tom Rodgers for his support of my work. I am frequently asked to explain the thought process involved in coming up with a new puzzle.” and for presenting me with the opportunity to meet Martin Gardner. I neither know or understand the process of any creativity than to say that it just happens. Unfortunately.

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

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

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

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

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

It really hurt. I must admit. there was a lull in my mind. not in the falling of the sand but in the ﬂowing of the liquid. and are indistinguishable in color and transparency. Hein realized. but what a clever ruse it is. with the exception that the sand was running downward in both hourglasses. When the cylinder is inverted. I admit. He knew that the sinking glass directly refuted his theory. it’s very simple. And on that tablet lay. “So do I. not the sand. Hein was awed by the brilliance of it all. The falling sand is just for misdirection. “somewhat worried and depressed on behalf of my beautiful theory. but of thinking.” Then.” he wrote. the heavier liquid is on top pushing down. but it would be easier to explain the symmetry of the phenomenon if it were falling upward in one of them!” Hein acknowledged Gardner’s appreciation of beauty in a theory.” he explained. MORRIS the ﬁrst time. I found myself in possession of the solution. “When the Alps in the lite of the (superelliptic) setting sun and the growing. exactly half moon (D for dynamics) were left behind us and all was dark. “I couldn’t think of anything else but the principle. He rationalized in jest: “My theory works.” he wrote on September 22.138 S. “This is not a question of fumbling one’s way to a solution.” The Solution Is in the Solution The key to the puzzle isn’t in the glass but around it. being false makes it less right. but he didn’t buy one to take home. and he found it painful. The hourglass puzzle is caused by an hourglass effect. Only when enough of it has seeped down does the glass begin to rise.” He later wrote about the “Aha” experience. My dear Watson. the solution of the hourglass mystery.” He couldn’t bear to abandon his pet idea completely. he was somber.” When Piet Hein left Milan on the afternoon of September 21.” Hein was forced to make a paradigm shift. But it does not make it any less beautiful. but let us be honest and not rate it any lower just because it is false. “I should like to meet the person who invented this effect and designed it so as to hide the solution so elegantly for us. a tabula rasa. . “I am glad you think my explanation is beautiful. Hein had a sudden crystallizing insight . but the relevant displacement is in the liquid.“At an altitude of 8500 meters. There must be two liquids of different speciﬁc weights that don’t mix completely. as the plane soared over Mont Blanc. like a small ﬁsh on the center of a huge platter.

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

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

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

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

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

Usually there is an “aha” reaction when students see that a cube can actually be constructed from the pieces. Three “longs” of size ¢ ¢ . For more advanced students it would be reasonable to ask in how many essentially different ways can the cube be constructed from the eight (uncolored) pieces. This will depend on just what is meant by the words “essentially 145 . a single cube).e. One “Biggie” of size ¢ ¢ (where ¾ ).Cube Puzzles Jeremiah Farrell Binomial Puzzle A new. 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. Now color each piece with six colors according to the scheme in Figure 1. has exactly two solutions — each with a different set of three colors. The problem then is to arrange the eight pieces into a cube so that opposite faces have the same color. if colored as above. 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 “ﬂats” of size ¢ ¢ . Before proceeding with this discussion the reader is urged to build a puzzle and attempt its solution. This puzzle. With proper insight it can be solved in a few minutes. rather amusing. Without this insight it typically takes several hours to arrive at a solution if one can be found at all..

additionally. try again to solve the puzzle. FARRELL Figure 1.½ distinct ways of constructing the cube. 0). and C will turn out to be a solution set.146 J. If. A. That is. If you cannot. B.½½ . 1.½ ¼. There are 93 solutions in this case. and the solver has the additional clue that a face is all of one color. each . 2. then there are ´ ¿µ´ µ´¿ µ¾ . but there still remain a great many cases that are nearly right. 2. Notice that the eight pieces must form the eight corners of the completed cube — one piece for each corner. B. These hints are to be regarded as progressive. 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. C. Therefore. A. after reading a hint. proceed to the next hint. ½º Call the six colors A. Trial and error is not a very fruitful way of trying to solve this puzzle. and 3. 1. 0. and 3 are any six distinctive colors. each of the 48 faces of the pieces is colored with a unique color. C. The combinatorial puzzle uses only six colors in its construction. B. Solution Hints.

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

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

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

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

the location of the third cube is determined.105 combinations are isomorphisms.105 of these combinations meet the three-cubes-of-each-color requirement. is that there are numerous solutions. It becomes necessary. as determined in Section 4. Since during the course of trying to solve this puzzle a conﬂict often occurs. The ﬁnal part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. Twelve of these can be selected in 1.152 S. The second observation is that for a cube to be assembled with the nine colors on each face. as is shown in Section 5. . it becomes apparent that most of the 133. The method of sorting the 10. none of the nine planes may contain two cubes of the same color. the results are shown in Table 1.700 (36!/12!24!) different ways of which.251. Fortunately. only 133. but not in the center of a face.677. A proof is given in Section 3. Since one person can assign the fourth and ﬁfth colors to orange and red and another person to red and orange. The question then arises: How many solutions are there? Most intriguing about this puzzle. then. The individual dicubes can be colored thirty-six different ways. this rule then tells the experimenter to backtrack. The previous question now becomes more interesting. some with the tricube on the edge and some with the tricube through the center of the cube. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. FARHI a difﬁcult exercise for a moderately proﬁcient programmer to verify the results obtained. The ﬁrst observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube.691 combinations need to be sorted into disjoint sets. The third observation. in which the number of isomorphisms using the set of thirty-six dicubes is also determined. to separate these cases into disjoint sets of isomorphisms. Thus.691 combinations into 148 disjoint sets is described in Section 6. A proof is given in Section 2. only thirty dicubes are required. This was achieved through the use of a computer program. and only 10. when two cubes of the same color are in position. obtained after some experimentation. and the most difﬁcult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1.

DZ[yz 1439 624 96 4 80 6 15 .:Zppp 10 10 0 0 0 0 0 . 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center .DOeyz 720 312 20 0 12 0 4 .THE NINE COLOR PUZZLE Table 1. 153 No.:Z[yz 60 60 0 0 0 0 0 .DZpez 720 312 198 12 83 10 16 .COeyz 360 180 8 0 16 16 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 .

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. FARHI 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 .THE NINE COLOR PUZZLE 155 No.

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

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

and 14. 3. Since the corner colors are displayed on three faces. cubes of the same color are never in the same plane.105 Possible Dicube Color Combinations The tricube colors are deﬁned as colors 1.158 S. accounting for two faces. 2.” The remaining six corners must have cubes of different colors. FARHI If an irregular tricube is used. and 11. then a third dicube with color 2 and any color from the second row is selected. If color 3 has been selected once. If color 2 is selected from the ﬁrst row. the center case ﬁlls spaces 5. 4. and 11. and on a centerface such as shown by the letter “B. then color 3 is combined with two colors from the third row. 11.” The remaining two colors must all be on edges. Determination of the 133. The two dicubes with color 1 may use any two colors in the ﬁrst row of the following table. 5. the edge case ﬁlls spaces 2. 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. then one color is selected from the third row and if color 3 has been selected twice. such as shown by the letter “C. If none of the previous selections include color 3. 2. then none are selected from the third row. The “No Two in the Same Plane” Rule Figure 3. and 3. 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. there are three possible locations. The corner case ﬁlls spaces 1. . the remaining two cubes must be located on an edge.” In each case. One cube is centrally located and thus not visible.

This results in requiring only thirty cubes and. 6. and 9 one can logically deduce the ﬁrst colors by the three . (6. 6.THE NINE COLOR PUZZLE 159 First Color 1 2 3 4 5 6 7 8 Second Selection 2 3 4 5 6 7 8 3 4 5 6 7 8 4 5 6 7 8 5 6 7 8 6 7 8 7 8 8 9 9 9 9 9 9 9 9 The process continues: color 4 is combined with colors in the fourth row such that there are three cubes with color 4. 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. (4. 6). 4. 8. or 5. 9. 7. (4. (1. For example. 9). (7. 8. 5). (2. 8). 9) and (8. 3). (3. 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. (7. 3. Identiﬁcation and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. the two new colors are identiﬁed as colors 6 and 7. using the same algorithm as before. 7). color 5 is combined with colors in the ﬁfth row such that there are three cubes with color 5. 9). given the second numbers 2. 2). 4.105 possible color combinations. 6). 9. 8). 5. only 10. 4). Using this algorithm. 5. Similarly if color 2 is not associated with colors 3. 6. a computer program established 133. Identiﬁcation of the second color of these ordered pairs is sufﬁcient to deﬁne all the colors. (5. (5.691 color combinations. etc.

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

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

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

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

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

Some months ago. O. . particularly inspired by Muroi’s “Four Sticks” (so closing the circle of mutual inspiration). Figure 2. Nine and One-Half Moves. Zigzag. V AN DEVENTER Figure 1.166 M. The three pieces of the puzzle form a square with internal voids.” This is is a true two-dimensional burr of only three pieces and it needs no less than nine and one-half moves to separate one piece from the other two. which I have called “Nine and One-Half Moves. I took up the challenge again and succeeded in ﬁnding a new design of a planar burr.

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

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

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

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

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

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

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.

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

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

ﬂexing. and Theoretical Puzzles. one or more pieces hold the rest together. They are solved by hinging. Jigsaw Puzzles (JIG) are made from cut or stamped-out pieces from a single complete object.e. provision is made for puzzles pending classiﬁcation.. Math. The mechanism of the puzzle is not usually apparent. or drinking therefrom. close it. the differences must be sufﬁciently minor so as not to obscure the similarity of the pieces. The pattern may be color. They usually comprise a single object or associated parts such as a box with its lid. etc.” 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. They may clip together but do not interlock in 3-D. Ambiguous Pictures and Puzzling Objects (AMB). Trick. or otherwise get it to work. Mystery. Other Types of Mechanical Puzzles and Objects (OTH). nor do they involve general assembly/disassembly of parts that interlock in 3-D. or a nut and bolt. magnetic poles. Sequential Movement Puzzles (SEQ) can be solved only by moves that can be seen to be dependent on previously made moves. The pattern required may be the matching of edges of squares. Pattern Puzzles (PAT) require the placing or arrangement of separate pieces of a similar nature to complete patterns according to deﬁned rules. Interlocking Puzzles (INT) interlock in three dimensions.CLASSIFICATION OF PUZZLES 177 ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ Opening Puzzles (OPN) are puzzles in which the principle object is to open it. Some have a container and are posed as packing problems. and the principle objective is to restore them to their unique original form. Measuring. i. Many clip-together puzzles are “non-interlocking. Logic. undo it. 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. Cutting. or the pieces are mutually selfsustaining. Folding and Hinged Puzzles (FOL) have parts that are joined together and usually do not come apart. Puzzles in which something appears impossible or ambiguous. Included in this group are Balancing. shape. Where the pattern is due to differences in shape. or folding. Also. etc. Jugs and Vessels (JUG). texture. pouring. Vessels having a mechanical puzzle or trick in their construction that affects the ﬁlling. a padlock and its hasp. remove something from it.

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

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

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

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. knives. berrocals. “Mayer’s Cube.” Cutler’s burr in a glass Cofﬁn’s “Saturn” JWIP and keychain animals. O’Beirne’s “Melting Block” . poison rings Nuts and bolts. Oskar’s keys. cast “ABC.” “Hazelgrove Box” Strijbos aluminium burr box Burrs.” “Nine of Swords.” 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. Soma. “Tak-it-Apart. checkerboards. pens.” Chinese rings. “T” puzzle Polycubes. cutlery. Pentominoes. Oscar’s “Dovetail.CLASSIFICATION OF PUZZLES RTF-2D3D RTF-3D Route mazes 2-D on 3-D surfaces (any path) Route mazes 3-D (any path) Maze on surface of cube Some hedge mazes.” “Margot Cube. “Hoffman” cube. gears. Dalgety’s “Devil’s Halo” Leather tanglement puzzles Boxes. purses Padlocks Chippendale tea chests. puzzle rings Hess wire puzzles.

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

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

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

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

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

if y is greater than x. if y is greater than x). if x is greater than y. I will now prove the following two obviously incompatible propositions. say. is equal to the excess of y over x. Then the excess of Ü over Ý. Proposition ¾. But look. 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. 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. one of which is twice as great as the other.e. And isn’t 100 surely greater than 50? 189 . if Ý is greater than Ü. The excess of x over y. is certainly 50 (since Ü is then 50).A Curious Paradox Raymond Smullyan Consider two positive integers Ü and Ý. is 100.. is greater than the excess of y over x. and the excess of Ý over Ü. the excess of x over y. suppose Ý. is certainly 100. The two amounts are really the same (i. Proposition ½. 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. if x is greater than y. if Ü is greater than Ý. Now.

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

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

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

I have answers only for N = 5.194 K. what’s the question?” Usually such a setup is too wide open to be interesting. 47. for me at least. Possible for 8. lies the intriguing question: Will we agree that one or another task statement lies within the spirit of the game? This is. ﬁnally clinched independently by Doris Rychener and A. and who then went on to prove that 77 is the largest integer that cannot be so partitioned. (Impossible for n = 2. as I think I have deﬁned it. even after developing a complete proof. Divide a rectangle into n disjoint subrectangles without creating a composite rectangle except for the whole (Frank Sinden). The trouble is that a misﬁre problem is . others quite esoteric: ´ µ ´ µ ´ µ ´ µ ´ µ ´ µ ´ µ Stated above.g. (Generally known. proved by Ron Graham. well known. and who did. KNOWLTON For a mathematician or logician. Design a polyhedron with Ò edges. 9. 3. 9. 7. The tetrahedron has 6 edges.” Meager as the current collection is. a sociological rather than mathematical question. of course. But a misﬁre problem. 10.. and 77. Some of these are well known. But “difﬁcult” may be the wrong word. 6. Cut a cube into n subcubes (called the “Hadwiger” problem. these are attrociously imprecise speciﬁcations. One of Martin Gardner’s books contains essentially this problem as a wire-identiﬁcation task. (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). But here. Zbinder who demonstrated a dissection into 54 subcubes). 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. 12 . its members do exhibit a special charm. my guess is that the matter lies somewhere on the non-crispness scale between agreement on “proof” and agreement on “elegance.) Cut a square into Ò acute triangles with clean topology: no triangle’s vertex may lie along the side of another triangle. 5. or 9. Partition n into distinct positive integers whose reciprocals sum to one.) What we have here is an inﬁnite number of problems of the form “Here’s the answer. e. has such a severely constrained answer that it is almost impossibly difﬁcult. 11. (Charles Cassidy and Graham Lord. with a rather obvious proof: Partition a square into n subsquares. 6 since 2 + 3 + 6 = 11 and ½ · ½ · ½ ½. tracked for years by Martin Gardner. the next have 8. such a partition for 11 is 2. edges. conjectured by Ken Knowlton.

I invite further contributions to the collection. say. Instead of searching directly for a task last misﬁring. for Ò ½¿.MISFIRING TASKS 195 a math problem with no implied search process whatever. . 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. Or arguments as to why this is too mushily stated a challenge.

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

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

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

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

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

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

204 B. GOSPER Times 100–105: .

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

he and Hartmut Holzwart have been producing oodles of orthogonal spaceships with speeds c/2. and one period 5. Also. inﬁnitely many period 3. Hartmut found another with speed 2c/5. (Mine runs only on an Apple II. We also have an argument. so we also get glider guns of those periods. speed c/3 orthogonal spaceships.) Lately. but 5. Due mostly to the work of Buckingham and 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. GOSPER resembling your initials. 6. The ﬁrst glider to hit them is reﬂected . of periods 5. c/3. For example. the 73 gen turn can emit a glider. 65. I built a p155 oscillator in which 20 Bs travel around a track of length ¿½¼¼´ ¾¼ ¢ ·¾ ¢ µ generations. by combining them we can build a closed loop whose length is any large multiple of 5 or 8. which reﬂects a symmetric pair of parallel gliders. it consists of a still life (“eater3”) and two spark producing oscillators. which implies that all sufﬁciently large periods are possible. based on construction universality. David Bell wrote a similar program that runs on faster machines. A period 150 gun ﬁres toward a distant pair of pentadecathlons. or 73 generations. we now have nontrivial examples (i.e. For the multiples of 8.206 B.) Construction of Oscillators. Another amusing oscillator uses the glider crystallization that Bill mentioned in his centinal letter a few years ago. speed c/4 orthogonal spaceship. (These are used in some of the large patterns with unusual growth rates mentioned later. and c/4. 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. we put several copies of the B in the track to reduce the period to one of those mentioned. JHC!).) 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). one period 4. a few oscillators with periods 5 (including some with useful sparks) and 6. speed c/2 orthogonal spaceships. and 47 are the only ones we know with the right sparks. in which a B heptomino is forced to turn a 90 Æ corner. (No doubt you ﬁgured that out for yourself long ago. or 47. one period 4. inﬁnitely many period 2. More recently. speed 2c/5 orthogonal spaceship.. (Any greater nonmultiple of 4 would also work.) Periods ½¿ · Ò and ½ ¼ · Ò use a mechanism developed by Buckingham. The · ¾ Ò ¾ · ¾ Ò ¼ · ¼Ò ¾¿¼ · ¼Ò ¾ ¾ · ¿ Ò and · ¿ Ò use a device discovered by Wainwright. 6. The turn can take either 64. speed c/4 diagonal spaceship (not the glider).

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

1 given. 33) (10. 50) (11.232 Version (10. 1 given. 40. 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. 36) (27. all given. 30. 30. 55) (78.6) # of Solutions (55. 33. 454. 36) (25. 40. 16) (16. SINGMASTER Solutions in Source 5 given. 17. 40) (10. 20) (10. 80) (10. At this point. 1 given. 26) (20. 140) . 2 given. 56. One can scale the set . 1 given. 136) (3. 1 given. 1 given. 60) (117. 81) (55. 30) (12. 20. 25. Bhaskara makes the following comment: “This. 36) (78. 1) (7. 50) (30. 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. 90) Date 1202? 1202? 1202? c.4) (70. 196) (45. 32) (12. 901) (10. 1 given. 37) (27. 16) D. 30) (18. 1 given. 1 given. 170) (305. 752. 1300 1300s 1489 1489 1489 c. 21) (225. 50) (10. 30. 1) (100. 6 given. 33) (20. 11. (20. makes this more speciﬁc by saying they ‘sold a part in lots and disposed of the remainder at one for ﬁve panas. Sridhara ﬁnds a one-parameter solution set. 35) (11. 1 given. 29.552. so we again have a three-dimensional solution set.400) (25. 119. 32. in some way to reduce the solution set to two dimensions. 36) (30. 20. 100) (45.9) (100. none given. 40. 4 given. 1 given with fractions. 12. 30) (8. 1 given. though it is not as easy to see how to do this as in the Western version. 60) . 1 given. (20. 48. . 40) (18. 49) (3. . 1 given.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 15) (31. 81) (10. 603. 60) (10. . 1 given. 55) (55. 82) (17. 20) (64. 68. one with fractions. 55) (16. none given. ¬ . 1 given.

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

18. pp. and the 13Ø edition. References [1] Alcuin. Johnson. [5] Bonnycastle. Claude-Gaspar. J. 1884. Mar 1992. reprinted by Blanchard. 1150. A.. A. [3] Bachet.” c. 1973. P. Prob. giving total numbers of 3295. Lyon. Lucas N. 242–244.. Gaz. 1817. Analysis similar to that done before shows that there are then ½ · ´ ¬ ½ Ò µ ¬ solutions.. Vieweg. Bhaskara solves his example by taking ¾. designed for the Use of Schools. 1923–24. 106–115.234 D. 3200 and 16. (There have been several reprints. Braunschweig. P. 10. [6] Bunt. 15. 13. annotated by David Singmaster and John Hadley. Bîjaganita. et al. London.. Paris. 3292. pp. Nunn et al. pp. Labosne. with Notes and Observations. pp.470 [4]. 4th ed.” [2] Ayyangar. Facsimile with transcription and commentary by John Wardroper.. Problemes plaisants & delectables qui se font par les nombres. 1879. Oxford University Press. [8] Diophantos. By the way. 178–186. 150 B. London.) Chap. The ﬁrst nine editions appeared “without any material alterations.. 2nd ed. 214–223. SINGMASTER For given and ¬ satisfying ´ ¼¼¼ µ and ´ ¼µ. Wynken de Worde. See No. 15.900 and of 3294. Book I. London..” J. Lyon. Rigaud. Rigaud. Arithmetica. John.(475). 1782. In: Henry Thomas Colebrooke. Problems to Sharpen the Young. nos. 1976. J... “Nine Chapters on the Mathematical Art. 1612. Indian Math. 5th ed.. H. Algebra. London. An Introduction to Algebra. there are only a ﬁnite number of solutions. Prob. pp. 3rd ed. I examined the 7 Ø edition. p. 102– 126. 1511. Translated by John Hadley. pp. Wiesbaden. Soc. 19. 1910. Krishnaswami. 122–126. L. 1824. [9] The Demaundes Joyous. 1st ed. aka Bhâskara.” which “may be considered as a concise abridgment” of his two-volume Treatise on Algebra. [7] Chiu Chang Suan Ching. trans. 1805. Gauthier-Villars. . 6.. Vogel. “A classical Indian puzzle-problem. 12. The Historical Roots of Elementary Mathematics. 1964. 24. Neun Bucher arithmetischer Technik.. Paris. Translated into German by K. In: T. [4] Bhaskara. Bijanganita. Math. 21. “an entire revision of the work. pp. 33. Nos. 1624. pp. which may be the same as the 1815 edition. 16: “De duobus hominibus boves ducentibus — Two men leading oxen. 24. 86–88. 5th ed.C. he produced a 10 Ø ed. Diophantos of Alexandria. 4392. 76. 2nd ed. John Murray. Prob. Heath.” In 1815. 134–136. and Other Places of Public Education. with Arithmetic and Mensuration from the Sanscrit of Brahmegupta and Bhascara. so . 1874. PrenticeHall. London. 1959. 1968. including Sandig. Revised by A. Gordon Fraser Gallery. 170. 1813. 54. reprinted by Dover.so and ¬ ½½¼. v.

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

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

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

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

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

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

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

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

We will have a solution to our problem if the Ò · ½ column has all entries 0. Ú ·½ . We can generate new columns moving from left to right or from right to left. let Ü¾ ¼.244 L.. Since there are only a ﬁnite number of adjacent pairs of the Ñ-tuple. Perhaps the next question might be For Ñ ﬁxed. · is odd.e. we can repeat the process for column 2. Ò is independent of Ú. 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. i. · · is odd. let Ü ¼. Now that every neighborhood for Ú½ is odd. either we get a column of 0’s or there must be an adjacent pair that repeats in our process.e. otherwise Ü ½. Of course. let Ü¿ ¼. and that label can be assigned so that the neighborhood is odd. Figure 5 shows an example where Ú ½ col´¼¼½½¼µ. Let Ú Ú ·½ be the ﬁrst pair that repeats. 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. Æ Ò is odd for square Ú in column Ò Ò. otherwise Ü ½. · is odd. let Ü½ ¼. and each new column will be dependent on just the preceding two.. Using the second occurrence of Ú of the ﬁrst occurrence of Ú Ú ·½ . SHADER will now have all but one entry labeled. · · is odd. otherwise Ü½ ½. is there an Ò such that for every starter Ú the Ñ ¢ Ò rectangle has all neighborhoods odd?. to show that Ò ¾¿ is a nice exercise for the reader. let Ü ¼. and Ò . . E. i. otherwise Ü¾ ½. Each neighborhood If If If If If This process can be repeated as many times as we like. For Ñ . · · is odd. Ñ is not limited to the value 5. Ú½ ¡ ¡ ¡ Ú Ú ·½ ¡ ¡ ¡ Ú Ú ·½ Figure 4. otherwise Ü¿ ½. Contradiction! Hence there is a column of 0’s and the proof is complete. and we generate a second Ú Ú ·½ somewhere to the right of the ﬁrst Ú Ú ·½ (see Figure 4).

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. If the ray would not return along the entry paths.Point Mirror Reﬂection M. 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. Solutions are shown in Figure 1. P = perpendicularly reﬂecting mirror). passing once again by entrance mirror E. half of the potential reﬂections would not be used. It is obvious that the strategy of 245 . Maximum number of reﬂections for two and three point mirrors (E = entrance mirror. 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. 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. if these are suitably placed and oriented.

At mirror E there can be one additional reﬂection for the entering ray. see Figure 2. Number of reﬂections for different mirror types. V AN DEVENTER Figure 2. Mirrors on the Corners of a Regular Polygon Can the theoretical maxima Ê ½ and Ê¾ be achieved for any Æ . from the other mirrors. so remains valid. . O. since the light paths are used in two directions. as shown for the regular pentagon in Figure 3. the angle between all the neighboring pairs of light paths is 180 Æ Æ . Only at mirror P can there be an odd number of reﬂections.246 M. the theoretical maximum number of reﬂections is Æ ´Æ ½µ · ½. which is an odd number. At all other mirrors there must be an even number of reﬂections. At each mirror a maximum of Æ ½ rays can be reﬂected. at the other Æ ¾ mirrors (other than the E and P mirrors) there can’t be Æ ½ reﬂections. going to every other mirror. At mirror P there can be at most Æ ½ reﬂections. for even Æ there is a parity complication. but only Æ ¾ to make it an even number. Therefore. At mirror E there can be at most Æ reﬂections. using a perpendicularly reﬂecting mirror P yields the maximum number of reﬂections. Æ ½ from the other mirrors and one from the entering ray. However. or ´½ µ Ê½ Æ¾ Æ · ½ 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. 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.

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

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

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

No general strategy was found. V AN DEVENTER Figure 8.250 M. 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 . Deleted paths for composite Æ up to 16. For composite Æ up to 16. which can be reached for prime Æ . solutions were found by a systematic trial and error search using a regular Æ -gon grid. O.

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

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

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

which concern extreme behaviors of ¿Ü · ½ iterates. for an ¾ explicit constant ½. [13]. Here we survey some recent results on stochastic models. [15]. Since a number Ò is multiplied by ½ if it is even and approximately ¿ if it is odd. one expects the iterates to decrease in size and eventually become periodic. The simplest of them assumes that this independent coin-ﬂip behavior persists until ½ is reached under iteration. which may be enough for a pseudorandom bit to be extracted. 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. Cloney. These models consist of a repeated random walk model for forward iterates by Ì . These mathematical models predict that the expected size of a “random” Ò after steps should be about ´ ¿ µ ¾Ò. cf. Lagarias [14] and Luby [17]. 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. obtained in joint work with Alan Weiss and David Applegate. Lagarias and Weiss [15] studied two different stochastic models for the behavior of ¿Ü · ½ function iterates. [20]. Several heuristic probabilistic models describing ¿Ü ·½ iterates are based on this idea. see [9]. [24]. et al. 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. There is excellent numerical agreement of this model’s prediction with ¿Ü · ½ function data. Garner [11] and Korec [12]) they also seem to possess some residual structurelessness. This model has the drawback that . In doing so we are in the atypical situation of modeling a purely deterministic process with a probabilistic model. 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. and a branching random walk model for backward iterates of Ì .3X + 1 FUNCTION ITERATES 255 One may summarize this by saying that the parity of successive iterates of Ì initially behaves like independent coin ﬂips. 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. see [20]. From this viewpoint it becomes interesting to determine how well the properties of ¿Ü · ½ iterates can be described by a stochastic model. Even though ¿Ü · ½ function iterates possess quite a bit of structure (cf. On the positive side. Since this is less than ½. one expects that ¾ ¾ on average it changes multiplicatively by their geometric mean. [24]. [5] proposed using the ¿Ü · ½ function as a pseudorandom number generator. which is ´ ¿ µ½ ¾ .

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

For each initial value Ò an entirely new random walk is used. The length of such trajectories approaches the limiting value ½ ´Òµ ÐÓ Ì ´ µ ´Òµ ÐÓ Ò ÐÓ Ò ¼ ½ ´Òµ . let Ø£ ´Òµ equal the maximal value taken during the random walk." 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. 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. For the extremal trajectories for Ø£ ´Òµ in (14. so that Ø£ ´Òµ ÐÓ Ò. Large deviation theory shows that. For the extremal trajectories for £ ´Òµ in (14. This random variable is ﬁnite with probability one.3X + 1 FUNCTION ITERATES 257 ¿ by starting at ÐÓ Ò and taking independent steps of size ÐÓ ¾ or ÐÓ ¾ ½ each. Large deviation theory also asserts that the graph of the logarithmically scaled trajectories ´½ º µ approaches a characteristic shape. £ and the expected size of ½ ´Òµ is given by (13. with probability 1. see Figure 1. this explains the name “repeated £ random walk model. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally.24).4). the limiting graph of these trajectories has a different appearance. The random variable ½ ´Òµ is the step number £ with probability ¾ at which 0 is crossed.1) it is a straight line segment with slope ¼ ¼¾¿½ starting from ´¼ ½µ and ending at ´ ¼µ.

starting at ´ ¾µ and ending at ´¾½ ¼µ. see Figure 2. LAGARIAS Figure 1.5) approximates the union of two line segments. and the graph (14. and ending at ´ ¾µ. Figure 2. Scaled trajectories of for ½ . solid for Ò ½¼). C. solid for ½¼). and the second with slope ¼ ½ ¿ for ÐÓ Ò ¾½ ÐÓ Ò. the ﬁrst with slope ¼ ½¿½ for ½ ÐÓ Ò. Ø´Òµµ ÐÓ Ò in ½¼ Ò . maximizing ´Òµ in ½¼ Ò ½¼ ·½ (dotted ¾½ ÐÓ Ò. Scaled trajectories of Ò maximizing ´ÐÓ ½¼ ·½ (dotted for ½ . starting from ´¼ ½µ.258 J.

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.6091 Table 2. Section 5] shows that in ½¼ ½½ trials one should only expect to ﬁnd a largest ´Òµ ¿¼.77 31.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6.456. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ´Òµ ¿ ¾ . The agreement with the stochastic model is quite good.884.4).5967 0.527 Random walks model ½ ´Òµ ´Òµ 70 108 165 214 329 429 592 593 706 755 21. .6020 0. ½ ´Òµ It gives the longest trajectory.68 24.799 6 8. A comparison of (14.400.5841 0. He found that Ò 12.48 18.6030 0.6026 0.24 16.890.5741 0.5927 0.371.91 19.328. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory.91 32.194. an analysis in [15.64 32.38 41.1) with numerical data up to ½¼ ½½ is given in Table 1.5818 0. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1.511 7 63.5857 0. Extensions of the data can be found in Leavens and Vermeulen [16].68 Ö´Òµ 0. ½ ½¼. 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 (14.527 has ½ ´Òµ ½¾ ½ and ´Òµ ¿ ¾ ½ . Maximal value of ´Òµ in intervals ½¼ Ò ½¼ ·½. 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.171 4 52.728.127 8 127. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale.13 26. where ´Òµ ÐÓ Ò .266.769.180.254 9 4. Vyssotsky [23] has found much larger numbers with high values of ´Òµ. and the large deviations extremal trajectory is indicated by a dotted line.6020 0. More recently V.815 10 13.527 5 837. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good.94 31. The large deviations extremal trajectory is indicated by a dotted line.

83 3. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ.24 16. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼.32 2.46 2.1).86 13.645 ´Òµ 21. see [15] for more explanation.65 5.06 4. The set of inverse iterates associated to any Ò has a tree structure. . if Ò ¾ or 8 ´ÑÓ µ.671 5 704.819 1.897 2.260 J. these trees have the minimal and maximal number of leaves possible for depth .909 1. (See Table 3 in Section 4. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿. In fact. or 7 ´ÑÓ µ.65 7. ½ ½¼. LAGARIAS Ò 1 27 2 703 3 9.10 20.05 19. respectively.976 1. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch.791 1.000 Ò 13.20 6.099 1. 15.86 5.903 2. Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2.84 19.559 Random walks model Table 3.528. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth . 5.02 21. In iterating the multivalued function Ì ½.817. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½.675 7 80.511 10 77.) empirically like . so do not have any signiﬁcant effect on the size of the tree. C.362.788 1.049.03 19. In this case. where ½ ´½ º½µ Ì ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½.831 9 8. 4.511 6 6.631.75 5. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ.790 1.391 8 319.663 4 77.566. Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk.86 7.14 11.59 23.48 12.804. In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed.560 1. there are duplications. using (15.

2). [15. È ÊÏ . cf. In [15] it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ . 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. satisﬁes (with probability one) ´½ º¾µ ÐÓ Ò£ ´ µ È³ ½ ÐÑ ½ ÐÓ This constant È ¬ ½ . . Theorem 4. 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 .e. and the value assigned to that vertex by the rules above is exactly Ù . The random walk at a given leaf node ends at a position Ù on the real line.1). 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.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. We assign labels ¼ or ½ to the edges. In order to prove the result (15. and where fying £ ´ µ ×ÙÔ ´½ º¿µ ÐÓ ´¾ · ½ ¾ ¿ ¿ ¼ This constant is provably the same constant as (14.4] show that the expected size of ÐÓ Ò£ ´ µ. and the root node is assigned the value 1. 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. i. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½.1]. Theorem 3. where ¬ is the unique positive number satis£ ´¬ µ ¼. it was necessary to determine the distribution of the number of leaves in a tree in the branching process. where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree. Lagarias and Weiss [15. Pruned ¿Ü · ½ trees of depth Ò ..

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

5 or 8 0 0 0 0 0 0 1 4 7 2 5 8 Figure 4. respectively largest number of leaves. and we consider the value of that vertex to be Ù . The paper [3] presents more evidence in favor of Conjecture ½These facts are derived using the probability density for Ï given in (15. The parent (bottom) always ¾Ò (edge label 0). (For actual ¿Ü · ½ trees. The number of leaves in actual ¿Ü · ½ trees empirically exhibits less variability than is predicted by this branching random walk model. 4 or 7 1 7 263 (prob. Transitions of the branching process [9].3X + 1 FUNCTION ITERATES (prob. and ¢ for a ﬁxed constant ¢ ½. 1/3 each) 2 8 5 4 1. Let Æ ´ µ and Æ · ´ µ denote the expected value of the smallest number of leaves.) We consider as random variables the smallest and largest number of leaves that occur among this set of trees. We make Ê´ µ independent draws of a tree of depth generated by this process. This value represents the initial value of the ¿Ü · ½ iteration starting from that vertex of the tree. The base vertex of the tree is located at the origin. choosing the root node uniformly ´ÑÓ µ. 1/3 each) 2. 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 9.4). ¾ takes a step ÐÓ ¾ along an edge labeled ¼ and takes a step ÐÓ ¿ along an edge labeled ½. choosing Ê´ µ Ê´ µ ¿ ·½ so ½ ¢ ¿. so that each vertex of the tree corresponds to a random walk ending at some position Ù on the real line. . 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.

750 0.83 1771.782 1.481 0.45 560.33 1.562 0.327 2.249 0. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a ﬁxed node Ò´ÑÓ ¿ ·½µ.32 17.09 56.481 0.342 0.375 0.898 1.238 1.268 0.486 0.62 1328.419 2.751 1.746 1.802 1.60 747.62 7.750 0.255 2.741 1.489 0.36 3149.742 1.300 0.534 0.489 0.507 0.422 0.802 1.481 0.481 0.482 0.289 0.491 1.778 1.67 0.500 1.750 1.83 99.762 1.669 1.490 0.490 0.738 1.232 2.501 0.869 1.633 0.78 2.491 0.738 1.259 0.534 0.500 1.492 0.911 1.38 236.791 1.307 0.375 2.360 2.281 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 4.50 315. 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 1.898 1.344 2.054 2.47 996.240 0.352 0.131 2.923 1.77 133.252 0.422 0.21 5.688 1.264 J.243 0.99 13.802 1. .302 0.292 2.451 0.246 0.475 0.81 4199.12 74.270 2.633 0.688 1.211 2.34 420.784 1.507 0.465 0.16 4.271 0.190 2.728 1.394 0.49 9. C.310 2.490 0.37 3.774 1.409 0.748 1.68 31.486 0.118 2.780 1.273 0.401 0.488 0.758 1.753 1.747 1.076 2.557 1.562 0.986 2.75 5599.265 0.475 0.758 1. Normalized extreme values for ¿Ü · ½ trees and for the branching process.475 0.77 2362.748 0.488 0.305 0.782 1.03 177.76 23.255 0.57 42.390 2.792 1.746 1.166 2.534 0.405 2.026 2.

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

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