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

Foreword Elwyn Berlekamp and Tom Rodgers

Ü ½ ¿ ½¿ ½ ¾½ ¾¿ ¾ ¾ ¾ ¿ ¿

I Personal Magic

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

II Puzzlers

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

v

vi

Contents

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

½ ¿

½¼¿ ½¼ ½½ ½¾ ½¿½ ½¿ ½ ½ ½ ½ ¿ ½

Contents

vii

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

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

III Mathemagics

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

Forward

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

ix

x

FORWARD

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

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

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

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

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

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

which was a big part of the puzzle page at Humpty.“There would be 50 magicians or so. “This kind of just happened. A friend showed him a novel way to fold a strip of paper into a series of hexagons. and Gardner was asked to do a monthly column. but.. If you write glibly. which led to an article on combinatorial geometry in Scientiﬁc American in December 1956.8 D. Gardner says. One of them intrigued him with a so-called hexaﬂexagon — a strip of paper folded into a hexagon. crammed with math journals in every language. Gardner drove to Princeton. When I ﬁrst met Asimov. a professional mathematician. I didn’t own a single math book. which turns inside out when two sides are pinched. but the mathematics in them has never gone much beyond second-year college level. I asked him if he was a professor at Boston University.” Gardner says he never does any original work. because that’s all the mathematics Gardner knows. as you know. gradually. because the two books are ﬁlled with all kinds of mathematical jokes. does tie in with math. Through the years they have grown far more sophisticated in logic. unless by accident. where graduate students had invented it.” [6] *** Every Saturday a group of conjureres would gather in a restaurant in lower Manhattan. “But I knew of some famous math books. because Lewis Carroll was. and I jumped at the chance. he watched his readers become more sophisticated. James R.“I’ve never made a discovery myself. So it wasn’t really too far aﬁeld from recreational math. that line one alcove in his study. demonstrating the appeal of math for the masses.” he recalls. RICHARDS inﬂuence of which can be seen in some of the activity pages which I am contributing to Humpty Dumpty. Newman’s The World of Mathematics had just been published. But it was a lucky idea because that’s been the best seller of all my books. he simply popularizes the work of others.He said no and . “At the time. of course. all doing magic tricks.” he explains with a shrug and a gesture toward the long rows of bookshelves.. asked me .”His ﬁrst columns were simple.” [14] *** At ﬁrst. you fool people.” Gardner reminisces. [23] *** He got into mathematics by way of paper folding. [16] *** “The Annotated Alice. as he studied the enormous literature on recreational math and learned more about it. “I’m really a journalist.Fascinated. 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.. the column was read mostly by high school students (he could tell by the mail).

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

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

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

“A Great Communicator of Mathematics and Other Games: A Conversation with Martin Gardner. 6.” Mathematical Intelligencer. NC]. 1997. March 11. 34–39. April 1978.” Two-Year College Mathematics Journal. RICHARDS [10] Hank Burchard. pp. 117–119. 36–40. no. vol. 47– 48. pp. December 9. The Flight of Peter Fromm. November 16.12 D. [24] Istvan Hargittai. no. 6E. 2. p. 101. “The Puckish High Priest of Puzzles. [12] John Braun. vol. 1981. 1. [22] Martin Gardner. [15] Rudy Rucker. 58. “The Annotated Gardner.” Time-News [Hendersonville. pp. “A Mind at Play. 56–61. p. [13] Anne Commire. 1–10. “The Magician of Math. “Martin Gardner. [19] Lee Dembart. 80–86. 1976. July/August 1981. “Martin Gardner: A Writer of Many Interests. December 12.” Newsweek. 3. 1976. [11] Sally Helgeson. pp. vol. no. 1B–2B.” The Charlotte [North Carolina] Observer. 5. pp. pp. no. 38. pp. 1979. January 1982. 1993. 10.” Skeptical Enquirer. December 6. March/April 1998. 89. pp. [17] Sara Lambert. 4. Dover. “A Conversation with Martin Gardner. “Every Day. Material taken from the Afterword. vol. 3. 8.” Omni. [21] Lawrence Toppman. pp. 1981. p.” Something About the Author. vol.” The Greenville [South Carolina] News. 10–21. 4. pp. 4. 232–244. “The Mathematical Gamester. 4. 66–69. 1E. pp. June 20.” Scientiﬁc American. [14] Anthony Barcellos. pp. 1994. no. “Martin Gardner. 2. [18] Lynne Lucas.” Science 81. “Mastermind. “[Martin Gardner]. vol. vol. . [26] Kendrick Frazier. 1981. vol. “The Math-e-magician of Hendersonville. “Magician of the Wonders of Numbers. 1979. “Interview: Martin Gardner.” Linking Ring. 40–41.” Bookletter. December 1995.” Los Angeles Times. [25] Michael Shermer. 1981. [16] Jerry Adler and John Carey. [23] Philip Yam. 32–37. 19. 16. pp.” Washington Post. December 5. no.” Skeptic. Impresario of Mathematical Games. [20] Scot Morris.

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

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

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

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

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

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

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

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

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

4 million mph Galaxy’s speed through the debris of the Big Bang .000 mph turning speed of Galaxy 1.roughaveragespeedofEarth 700.000mph.24 T. ROWETT Speed A child cycles ‘round the schoolyard Which lies on the Earth turning hard The Earth rounds the Sun As Sol does “the ton” And our Galaxy ﬂies — Gee! I’m tired 7 mph — child cyclist 700mphEarth’ssurface(NorthAfrica) 70.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

Michael W. Ecker

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

**The Sisyphus String: 123
**

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

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

42

M. W. ECKER

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

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

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

**What Is a Mathemagical Black Hole?
**

There are two key features that make our example interesting:

MATHEMAGICAL BLACK HOLES

43

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

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

44

M. W. ECKER

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

Words to Numbers: 4

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

**Narcissistic Numbers: 153
**

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

MATHEMAGICAL BLACK HOLES

45

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

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

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

ECKER The argument for why it works is similar to the case with the 123 example. which is to say. f (n) each value eventually is “pulled” into the black hole of 153. yet meets the two criteria for a black hole. But for those who enjoy programs. by cases or computer check. There are two ways to prove this. for large numbers n.e. for suitably small numbers. Repeat asking which column. for the black-hole attraction. laying out three across at a time). It’s a classic card trick. but be sure to sandwich the column that contains the chosen card between the other two columns. n. try this one from one of my readers. and then test to see whether other numbers are drawn to it. Card Tricks. Regroup the cards by picking up the cards by whole columns intact. card 11. Repeat one last time. you will need ﬁrst to discover a number that equals the sum of the fourth (or higher) power of its digits. 1¿ + 3¿ + 5¿ = 153. so 153 is indeed a ﬁxed point. I’ll omit the proof.46 M. First of all. the card chosen must be in the center of the array. there are some). Ask somebody to think of one of the cards without telling you which card he (or she) is thinking of. Arrange them in seven horizontal rows and three vertical columns. At the end. Now ask him (or her) which of the three columns contains the card. W. but the easier way is to draw a diagram that illustrates where a chosen card will end up next time. This is the card in the fourth row and second column. regrouping cards with the designated column being in the middle. Remove 21 cards from an ordinary deck. Now re-lay out the cards by laying out by rows (i. note that. ½¼ ¾¼ ½¼¼ ½½¼ ½¾¼ ½¿¼ ½ ¼ ½ ¼ ÄË ÈÊÁÆÌ Ì ¾½ Ö × ¹ ÈÖÓ ÈÊÁÆÌ ÔØ Ý Öº Åº Ïº ÁÆÌ ¹ ÁÅ Î´¾½µ¸ ´ ¸¿µ ÁÌ Ê ¼ ÇÊ Æ ½ ÌÇ ¾½ Î´Æµ Æ Æ ÈÊÁÆÌ ÈÊÁÆÌ È Ö ¸ ÔÐ Á ÁÌ Ê ¿ ÌÀ Æ ÈÊÁÆÌ ÈÊÁÆÌ Ö Ñ Ý Ë ÐÐÝ Ö ÞÞ Ö ÓÖ Ê Ì Æ × ººº ÈÊÁÆÌ ÓÙÖ Ö × Î´½½µ Æ . Second. and re-dealing out by rows. To ﬁnd an analagous black hole for larger powers (yes. Even Here’s an example that sounds a bit different.. Then.

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

If we multiply both sides by Ä we obtain a quadratic equation: ´£µ Ä¾ Ä ½ ¼ . Your own example may take more steps. That is. and 15. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. In fact.6 or so. 3. The ﬁrst two numbers numbers: 1. Divide both sides of the equation ´Òµ ½µ · ´Ò ¾µ by ´Ò ½µ. 1. This one may defy not only your ability to explain it. 2. This one is a bit more tedious. The ratios seem to be converging to a number around 1.). Iterate until a number repeats. ECKER Example: Start with 8028. and ´Òµ preceding two elements. but it is also that much more strange at the same time. 55. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. Repeat with 8532 to calculate 8532–2358 = 6174.48 M. The largest permutation is 8820.618 . but without using a program. Its value is roughly 1. 5. and these digits sum to 15. W. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. The black-hole number this time is 15. but your very equilibrium. including 1 and the number itself. we would have obtained the reciprocal of phi as the limit. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò call it Ä — as Ò increases. Its divisors are 1. 89. etc. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. 21. Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . Had we used the ratios ´Òµ ´Ò · ½µ instead. ½ . 34. phi. ´½µ ´Ò ½µ · ´Ò ¾µ for integers Ò ¾. 5. More formally. but you will reach 6174. 3. Now take the sum of the digits of these divisors. 8. 13. To get a quick and dirty feel for why phi arises. but fewer authors use that approach. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. the smallest is 0288. and the difference is 8532. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors.

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

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

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

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

Through transferring liquid among the containers. 0). g. and the other two start out empty: (15. 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. reverse the digits of the result. subtract 1 and you produce the original number. 3. Find the most efﬁcient solution. My regular racquetball opponent has a license plate whose three-digit part has the following property. 3. What is the number and what is the next greater number (possibly with more than three digits) having this property? . 4. Find a nine-digit number made up of 1. You are at a lake and have two empty containers capable of holding exactly ´ ¿ ½ ½ µ and ´ ¾ ½ ¾ ½ µ liters of liquid. 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. 63 9 38 33 32 12 88 25 16 18 15 Ü 5 E3. 0. 2 and 1. I. 5. Divide the number by 3. 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. and 6 pints. 25 and 9 combine to give 16) and use the rule to determine Ü. 6. 2. The 15pint container starts out full. Calculate the expansion of this 26 term expression: ´Ü µ´Ü µ´Ü µ´Ü µ ´Ü Ýµ´Ü Þ µ . E9. It is approximately 2244. Find the rule for combining numbers as shown below (e. How many transfers of liquid will it take you to get a volume of liquid in one container that is within one percent of exactly one liter? E7. E6. 8.5 nautical miles from Los Angeles to Honolulu. 7. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu.. How long does it take? E5. 4. 7. 6. 8. After going 10 miles in a straight line he ﬁnds himself back at camp. Where on earth could he be? E2.54 R. 5. HESS up and travels due North. 10 pints. You have containers that hold 15 pints.

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

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

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

I. between these points is greater than 3.58 R. An ant crawls along the surface of a 1 ¢ 1 ¢ 2 “dicube” shown below. It crawls at a rate of 1 foot per minute. 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. and (b) farthest from the far end of the band? M19. (a) If the ant starts at point A. M15. d. both of whom state that they cannot deduce the number on their own foreheads. . What number is written on your forehead? M18. The band is initially 100 feet long and is instantaneously and uniformly stretched an additional 100 feet at the end of each minute. A solution with eight triangles is known.) (b) What are the two points farthest apart from each other on the surface of the dicube? (The distance. Dissect a square into similar rectangles with sides in the ratio of 2 to 1 such that no two rectangles are the same size. A snail starts crawling from one end along a uniformly stretched elastic band. You are told that the three numbers are primes and that they form the sides of a triangle with prime perimeter.01). The snail maintains his grip on the band during the instant of each stretch. Divide an equilateral triangle into three contiguous regions of identical shape if (a) All three regions are the same size. You and two other people have numbers written on your foreheads. (c) two of the regions are the same size and the third region is a different size. A solution with nine rectangles is known. (b) all three regions are of different size. You see 5 and 7 on the other two people. M17. At what points in time is the snail (a) closest to the far end of the band. which point is the greatest distance away? (It is not B. HESS M14. M16.

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

Given that the width of the building is 125 ft and the rubber band must stretch an extra 10 cm to accommodate the building. What is the least number of irrational punches needed to eliminate all points of the plane? H4. “I grant you the land you can walk around in a day. “Take some of these stakes with you.851 ft for the radius of the earth. An irrational punch centered on point P in the plane removes all points from the plane that are an irrational distance from P. on the exact top is resting on the horizontal plane.60 R. 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. “Pound them into the ground along your way.) Assume that it takes Sir George 1 minute to pound a stake and that he walks at a constant speed between stakes. P.) H5.” (The Queen had read a little mathematics. I. HESS H2.” She pointed to a pile of wooden stakes. A billiard ball with a small black dot.” she continued.” the Queen said to Sir George. Where is the black dot when the ball returns to its initial resting place? . Imagine a rubber band stretched around the world and over a building as shown below. how tall is the building? (Use 20. How many stakes should he take with him to get as much land as possible? H3. and be back at your starting point in 24 hours. “To reward you for killing the dragon. All the land in the convex hull of your stakes will be yours.902.

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

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

(a) (b) (c) (d) When does one become a master? When does one become a sage? How long do the Lyrans live? Do ﬁve or six special years ever occur consecutively? H18. (A spandral is the Lyran unit of length. is unique. The inhabitants of Lyra III recognize special years when their age is of the form Ô¾Õ. On Lyra III one is a student until reaching a special year immediately following a special year. each of homogeneous composition. where Ô and Õ are different prime numbers. 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. The common axis of these layers is the planet’s rotational axis.331 to 1. The planet Alpha Lyra IV is an oblate spheroid. It (the Venerians are sexless) also reported that the straight line distance between it and its copod was 120 spandrals.PUZZLES FROM AROUND THE WORLD 63 H17. ﬁnally one becomes a sage until death. however. and the next inner layer is an anhydrous fromage. Cosmic. 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. just as one observes for Sol III.. which occurs in a special year that is the fourth in a row of consecutive special years. being made of a number of coaxial right circular cylindrical layers.) The reporting Venerian was mildly comforted because the distance to its copod was the minimum possible distance between the two. one then becomes a master until reaching a year that is the third in a row of consecutive special years. 18 and 20.) Unfortunately. Inc. (A copod corresponds roughly to something between a sibling and a rootshoot. It’s internal structure. The outermost of these layers is nearly pure krypton. Its axis of rotation coincides with the spheroid’s small axis. 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. 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 ﬁrst few such special years are 12. is contemplating mining the outer layer.

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

the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers.729. 6). 4. 2). 6. But this is not right because 18 is not the difference of 38 and 16. Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. . (9. ¾ gives Ø E5. 6). (1. 9). (1. 0. E2. (7. 4. 9. (3. and Ü ½ . ½ ¼½ or ¼ and ½ ¼½. The number is 381. 7. 0. 5). 10. 2).654. 0. 8. 0). so E8. 0). The license number is 741.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 6). ¼. 15 moves: (15. 10. (0. 2). . (11. 6. At ﬁrst glance. 6). E6. 0). 4. 0). (5. 8. Thus and Ò ¾½¼ transfers is the minimum. Ø¾ ¾¾ hours. (3.741. (1. so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. 0. 2. The trick here is that 1 knot = 1 nautical mile per hour. 6. E4. E3. (7. Anywhere within 10 miles of the north pole. E7. 10. E9. 6. 5. One of the terms is ´Ü Üµ. 2.425. In each case the numbers combine by summing the products of their digits. ¿ ½ ¼¼¼ are the smallest values satisfying the above equations. (5. (3. (3. The goal is to ﬁnd the smallest integer values of and so that Ò is a minimum and ¼ By numerical search we ﬁnd ¾´ · ½µ ½ ¼¼ ¼¾ . 6). (9. it appears that the rule might be subtraction. 6). Instead. 2. Thus Ü ½½. 0). 4. E10. 0). (11. the next greater number is 7.

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

0) (b) 40 moves. (23. 0). 124. (37. 0. (25. 0. 4. 6. 0). D will ﬁnish burning. (30. (20. (34. . 0. See the ﬁgure. 5. 4. 0). 4). ½¿ ½¿ ½ Ê 123. 6. 11. (31. 178. 6). 0). 10). (23. 6. 11. (31. 5). 6). When C is consumed. 158. 9. producing pieces C and D. 0). 11. (23. M2. (30. 0). (19. 0). (22. (25. 0. (22. 10). (29. (29. The probability of an acute triangle is ¿ ¾½ or ½ . (18. 6). 11. (17. 1. (22. 0). (26. 0. 4). 10). 11. 4). 0). 0). 0. (24. 128. 147. (35. (27. 9. 1). 2). 6. 6. 2. 3). If the hour hand is exactly on a second mark then the second hand will always be on the 12. (17. 0. 1. Cut B at that corresponding point. (24. 0. 6). the probability of a right triangle is ½ ¾½ or . 0. 6). 146. This is the shortest time interval that can be measured. 1. 11. 4. (24. (18. 6). (18. 0. (19. (19. 2. (23. 10). 11. (30. 10. 1). 4. 10). (26. 0). 10). 0). 0). 10). (35.75 minutes. 5). 0). 11. 0). 0. (30. start burning C from both ends and D from one end. 10). 6). 1). 11. 6). 0). 5. 127. 6. (33. 3. (26. (28. 1). (35. 10). (25. In 3. 11. (29. 0. 125. 0. 1). (23. 3. (21. At the same time. (25. 10). (20. Cut the lace in half. 0. 10). 11. (28. (29. (11. 0). 126. (30. 3. 148. (29. 0. 0). 1). (36. Burn A from both ends. 1. 0). 0). 0). producing pieces A and B. (25. 0). 0). 0. (25. 10). 0. 0. 11. 6). (36. (24. (31. 11. 6). 0).PUZZLES FROM AROUND THE WORLD 67 E16. 11. Solutions to Medium Problems M1. 5. 138. 1. 10). 10). (36. 1. 11. 168. (34. 0). 1. light the other end of D. 167. and the time is 8:24. 145. (25. (18. 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. 134. (19. 3). as seen from the list below. 6). 0). (29. 0. noting the point that burns last. E17. (30. (24. 0. 0). 1). 10. There are 7 ¢ 6/2 = 21 choices using vertex 1. (31. (21. 5. 11. 2). 156. (37. 0). 11. (31. (33. (23. 136. 1. 0. 3. 0). 4). (a) 32 moves. 1. 0. (24.

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

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

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

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

HESS (b) The two points are along the diagonal near P and its opposite Ô point on the ½ ¢ ½ face that includes A. Solutions to Hard Problems H1. 11. The points are ´ ¿ ½µ ¾ of the way along each diagonal for a distance of 3. 12. 14. 11. 8). The scores for the Good Eggs are (15. 13. (15.0119 . There are 7 Good Eggs. 7). (15. 10. 11. 11. 9). 6). 14. The number in each triangle indicates its area relative to the other triangles in the ﬁgure. I. 8). 13. 12. 14. (15. There are 5 subjects with 15 marks possible in each. (15. 14. Shown below. 14. M20.72 R. 9) and (14. 12. . 12. 12. 12. 13. 10). 13. Humpty Dumpty got a 10 in arithmetic. 10. (15.

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

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

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

(b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved.25 for the entire square or a circle inscribed in µ the square. There are 7 circles on each end with 105 sets of 3 circles in the middle. Note that the ratio È is 0. They deﬁne the four tangent lines drawn to produce ABCD as shown. I. Now consider two such circles of different sizes as shown. The ratio È takes on a maximum when Ö ¼¾ ¼ . The maximum ratio È Ö ¼¾ ¼ . The circles need to be packed as shown below. H10. Raise S and T appropriately above the .76 R. A view of the pyramid looking down from S perpendicular to the plane ABCD will look like the ﬁgure below. The smallest rectangle found to date containing 329 circles has 13 circles on each side of 101 sets of 3 and measures 2 ¢ 163. H9. The point S may be raised above the plane to give unit altitudes if the inscribing circle has radius ½. For Ò ½ there is just enough room for 329 circles.9973967 . HESS the shape is È Ô ½ ´¾ · ¾ Ö · Ö. Each has a radius ½.

7849 . ·¾ Ô ´½ ½ Õ µ È ´¼ ¼µ . and let Õ ½ Ô. · È ´ ¼ ¿¼µ Õ È ´ ¼ ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . At Ò games to Ò. which leads to Ô ¼ ½½ ¿ ¾ .3074 1. ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ .2801 1.8163 0. ¾ H12. A computer program was written to calculate the probabilities shown in the table below. Let Ô the probability of the server winning a point.7892 0. Ô Õ At a score of Ò games to Ò (Ò ¼ to 5).8027 0. H11.2511 1. ¾ ¾ Õ´ È ´¿¼ ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ . È ´ ¼ ¿¼µ Ô · ÕÈ ´ ¼ ¼µ. È ´ ¼ ¼µ ÔÈ ´ ¼ ¿¼µ · ÕÈ ´¿¼ ¼µ.3586 Pairing probability 0.7805 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. Let È ´ µ be the probability of the server winning the game when the server has points and the receiver has points.2069 1. È ´¼ ¼µ È ´ ¼ ¿¼µ when Ô¿ Ô¾ ¾Ô ½.PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit.2270 1. È ´¿¼ ¼µ ÔÈ ´ ¼ ¼µ. Clearly ST is not perpendicular to the plane. which leads to Ô ¼ ½ ¿¿ ¼ . Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1. This leads to ¾ È ´ ¼ ¼µ Ô¾Ô Õ¾ . È ´¼ ¼µ È ´¿¼ ½ µ when Ô ¾ Ô ¿.8286 0.

613. 12). ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . The product and sum two years ago could be achieved by (2. I. which must have been guessed by Smith.641. ´ ¾ ½µ ¾. . prime. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô . which must have been guessed by Jones two years ago. ¾ ½µ ¾. HESS H13.042. The product and sum could be achieved by (4. 7.093. There is only one set of ages that accomplishies this: (5. (d) Yes. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. . H15.321 is the ﬁrst of 5 in a row. ¼ ¼½¼ . 6. 8. (c) Unknown. (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ . Solutions occur for (b) ¾ ½. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. (2) the younger two children have different ages. (b) Length ¾ ¼¾½½¾ ¿ as shown.512. . Using a multiprecision continued fraction program. · ½. (b) 605 years.78 R. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½. we get (a) (b) H16. (a) 45 years. H14. ´ ¾ · ½µ ¾ prime. a=10. (c) 17. We look for cases where (1) the oldest child is under 20. (a) Length ½ ¿ ¾½ ½ ¼ as shown. ¿ ½¾¼. ¼ ¼¼ ½ ¼. 15). · ½. 16).546. ´ prime. H17. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago.444 years.

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

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

0888430 2.4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ Þ µÕ´Þ µ ´½ · Þ µÒ·½. The table below gives numerical values of the ﬁrst few. Ü 1 2 3 4 5 6 7 (6) The contribution to 2.1512074 45.2238350 26. Since the equation in Ý is even. ¼ for all . which reduces to ´½ · Ü · Ý µ´Ü Ý µ ´ Ó×´Ò · ½µÝ × Ò´Ò · ½µÝ µ where ´Ü¾ · Ý¾ µ ´Ò·½µÜ . ¾ × ´Ý · µ Ü Ý . Then Ý approximately equals ½·ÐÒ´ µ . Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ. The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ. Call the Ø such pole in the upper half-plane Þ Ü · Ý .5012690 3. A pole of Õ´Þ µ occurs Þ . (5) On Þ Ê for sufﬁciently large Ê.8790560 20. × ÒÝ Ý we can look for poles in the upper half-plane only and reﬂect each one into the lower half-plane. where Þ at ½ · Þ Ü · Ý. For larger deﬁne ´¾ ·½ ¾µ .6406814 3.2916783 3.0262969 3.5432385 32. Combining the poles in the upper and lower half-planes and some rearrangement ﬁnally produces ´Òµ where ½ ½ ÒÜ × Ò´ÒÝ µ (a) Since Ü Ü¾ · Ý ¾ Ü¾ · Ü · Ý¾ Ø Ò´Ý · µ Ý ÒÜ goes to zero as Ò goes to inﬁnity.8505482 39.PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ.6745053 3.4614892 13. not including its pole at Þ ¼.8221528 Ý 7. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to inﬁnity.

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

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

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

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

and subtracting each of the labels 11 to 19 from 20 we arrive at the labeling in Figure 1. Crown. a parallelogram drawn in the opposite way from the usual textbook fashion. The Butterﬂy has the same symmetries as a rectangle: Its position is described by that of its body. The nineteen Hexiamond pieces. . with its pointer on the left. GUY Figure 1. and each is positioned by its apex. sailing to the left. with its handle on the left. . 5. and 9. the multiples of 4 have rotational symmetry through 180 degrees. to the Signpost. . which have no symmetry. as do pieces 1. 7. K. with its hook on the left. ½. and the other even numbers have reﬂective symmetries. . whose head is on the left. We don’t know how O’Beirne decided whether to give a piece a number less than or greater than 10. Pieces 4 and 8 have only rotational symmetry and exist in enantiomorphous pairs. to the Crook. together with those that have rotational symmetry only. and Lobster each have a single symmetry of reﬂection. to the Snake. 3. but our mnemonics are as follows. The Chevron. O’Beirne had already suggested a numbering of the pieces from 1 to 19.86 R. Pieces with odd-numbered labels have no symmetry and. . which appears to be turning to the left. Negative (bar sinister?) labels are given to the Bar. exist in enantiomorphous pairs. ¿. to the Yacht. . Notice that 0 has the greatest symmetry. to the Hook. The symmetries of the Hexagon are those of the dihedral group D6. By subtracting 10 from the Hexagon. and the Sphinx.

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

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

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

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

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

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

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

On the contrary. Ô Ü (3-symmetry). by the Hexagon’s position Ø Ù Total 264 1094 124. for those who prefer their puzzles to have just one answer.518 A 75. Their numbers. 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.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. Six pieces meet at a point.94 R. K. are 130 195 533 193 377 2214 111.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to ﬁnd all the solutions. I wrote to Kate Jones.717 C 6675 D 7549 E 11. GUY Figure 18. Figure 19. president of Kadon Enterprises and discovered that she markets a set of “iamonds” .489 B 15. classiﬁed according to the position of the Chevron.

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

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

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

98

S. TAKAGI

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

A square with a center hole

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

Some patterns formed from the Sei Shonagon pieces

½1988, W.H. Freeman and Co., NY

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

Bernhard Wiezorke

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

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

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

99

100

B. WIEZORKE

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

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

TANGRAM CAT HAPPILY TURNS INTO PINK PANTHER

101

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

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

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

102

B. WIEZORKE

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

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

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

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

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

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

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

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

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

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

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

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

Tiny Tans: Four triangle-based pieces can make some pentomino shapes. JONES Figure 13. 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.114 K. Rhombiominoes: A skewed embodiment of pentominoes. This is a limited-edition set. the six pieces comprising 20 balls can make most of the pentominoes in double size. 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. The T and U puzzles are actually dissected pentominoes (Figure 13). The 20 distinct pieces form a ½¼ ¢ ½¼ rhombus (Figure 14).

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

JONES Figure 16.116 K. if there is a 90-degree angle or parallelogram to be found. . Martin Gardner. Lewis Carroll Chess Wordgame. the pentominoes will announce themselves in yet another manifestation. Thank you. beautiful playthings for the mind. And assuredly.

**Self-Designing Tetraﬂexagons
**

Robert E. Neale

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

Figure 1. Square window.

Figure 2. Cross plus-slit.

117

118

R. E. NEALE

Figure 3. Square plus-slit.

Figure 4. Square slash-slit.

Figure 5. Square cross-slit.

Square Window

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

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

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

SELF-DESIGNING TETRAFLEXAGONS

119

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

Figure 6.

Figure 7.

Figure 8.

Figure 9.

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

120

R. E. NEALE

Figure 10.

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

Figure 11.

Figure 12.

SELF-DESIGNING TETRAFLEXAGONS

121

Figure 13.

Figure 14.

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

Figure 15.

So reach underneath and hold the two free corners of the 1 cells. the 3-D Tricky Way. one corner in each hand. they are turned over so you see the 5 cells. while four of the faces are easy to ﬁnd. E. The numbers in parentheses are on the back side of the base. There is a 6 cell at the bottom of each box (see Figure 16. and the second ﬂattens it again. When you pull them down and away from each other. very quick. The ﬁrst renders the base 3-D. The other two faces are solid black and solid white. being black and white checkerboard patterns. exactly as indicated in Figure 16. and. and two slits in the center that intersect in the shape of a plus sign (see Figure 2). 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. involving changing the ﬂat ﬂexagon into a three-dimensional ring and back again. Note the two arrows in Figure 16. The other cells come together to form two boxes open at the top. This base is manipulated in a third way. This puzzle is interesting for two reasons: It is tricky to construct. Orient them just as indicated. number the base from one to six. moves. They indicate that you are to make the base 3-D by pushing two opposite corners down and away from you. . 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.122 R. the other two faces are quite difﬁcult to discover. One result was a puzzle made from a “cross” — a square (of 16 squares) with the four corner squares removed.) 3-D Tricky Fold In order to follow the instructions and understand the diagrams. You will make two.

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

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

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

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

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

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

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

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

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

The triangle can be moved over the entirety of the larger conﬁguration. The Bermuda Triangle rates about a medium level of difﬁculty. which is a triangular piece. We didn’t have any nightmares about it.METAGROBOLIZERS OF WIRE 133 The object of the Bermuda Triangle is to save the UFO that is trapped in the puzzle. 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. The solutions to many puzzles can be elusive until the puzzle has been manipulated many times. 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. but the name stuck. the UFO being a ring with an abstract shape mated to it. The Nightmare . as I sat playing with it. my wife joined me in critiquing my latest design. this one came to me fully formed and complete. taking the UFO with it as it moves. 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. There are several places where the UFO may be separated from the triangle but only one place where the separation will allow the solution to be executed. The puzzle is generally set up with the ring around the Bermuda Triangle. although moderately difﬁcult for the average puzzler to solve initially. As with most of my puzzle ideas. this one is relatively easy to remember once the solution has been seen. After making the prototype.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

none of the nine planes may contain two cubes of the same color. obtained after some experimentation. The method of sorting the 10. Since one person can assign the fourth and ﬁfth colors to orange and red and another person to red and orange.251.677. as determined in Section 4. . The ﬁrst observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube. This was achieved through the use of a computer program. The previous question now becomes more interesting. as is shown in Section 5. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions.152 S. Thus. Twelve of these can be selected in 1. it becomes apparent that most of the 133.105 combinations are isomorphisms. The ﬁnal part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. when two cubes of the same color are in position. The third observation. is that there are numerous solutions.691 combinations need to be sorted into disjoint sets. Fortunately. It becomes necessary. Since during the course of trying to solve this puzzle a conﬂict often occurs. the location of the third cube is determined. but not in the center of a face. A proof is given in Section 2. only 133.700 (36!/12!24!) different ways of which. The second observation is that for a cube to be assembled with the nine colors on each face. to separate these cases into disjoint sets of isomorphisms. The question then arises: How many solutions are there? Most intriguing about this puzzle. in which the number of isomorphisms using the set of thirty-six dicubes is also determined. and only 10. FARHI a difﬁcult exercise for a moderately proﬁcient programmer to verify the results obtained. only thirty dicubes are required. and the most difﬁcult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1. then. this rule then tells the experimenter to backtrack. the results are shown in Table 1. The individual dicubes can be colored thirty-six different ways.691 combinations into 148 disjoint sets is described in Section 6.105 of these combinations meet the three-cubes-of-each-color requirement. A proof is given in Section 3. some with the tricube on the edge and some with the tricube through the center of the cube.

:Zppp 10 10 0 0 0 0 0 .:Z[yz 60 60 0 0 0 0 0 .DZ[yz 1439 624 96 4 80 6 15 . 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.DZpez 720 312 198 12 83 10 16 .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 . 153 No.COeyz 360 180 8 0 16 16 0 .

FARHI No.154 S. 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 .

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.

FARHI 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.

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

there are three possible locations. The “No Two in the Same Plane” Rule Figure 3. FARHI If an irregular tricube is used.105 Possible Dicube Color Combinations The tricube colors are deﬁned as colors 1.” The remaining six corners must have cubes of different colors. the edge case ﬁlls spaces 2. If color 2 is not selected from the ﬁrst row then color 2 is combined with any two colors from the second row for the third and fourth dicubes. The only way the remaining two cubes of that color can be displayed on all six faces is for them to be located on opposite corners as shown in Figure 3 by the letter “A. 11. If color 2 is selected from the ﬁrst row. If none of the previous selections include color 3. The two dicubes with color 1 may use any two colors in the ﬁrst row of the following table. Determination of the 133. the remaining two cubes must be located on an edge.” The remaining two colors must all be on edges. and 14. such as shown by the letter “C. and 11. and on a centerface such as shown by the letter “B.158 S. then one color is selected from the third row and if color 3 has been selected twice. 5. The corner case ﬁlls spaces 1. accounting for two faces. 2. 2. 3. One cube is centrally located and thus not visible. Since the corner colors are displayed on three faces. and 11. 4. then color 3 is combined with two colors from the third row. then a third dicube with color 2 and any color from the second row is selected. the center case ﬁlls spaces 5. cubes of the same color are never in the same plane. and 3. then none are selected from the third row. If color 3 has been selected once.” In each case. .

9. 6. using the same algorithm as before.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. 5. 4. 4).105 possible color combinations. (7. 8. 3). 3.691 color combinations. 6). 6. 5). (7. (5. the two new colors are identiﬁed as colors 6 and 7. (4. 6). (4. 9). 8). 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. Similarly if color 2 is not associated with colors 3. For example. 9) and (8. 7). (6. 5. 8. 4. Identiﬁcation and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. Identiﬁcation of the second color of these ordered pairs is sufﬁcient to deﬁne all the colors. and 9 one can logically deduce the ﬁrst colors by the three . a computer program established 133. or 5. given the second numbers 2. 6. 7. (2. 2). (1. 8). etc. color 5 is combined with colors in the ﬁfth row such that there are three cubes with color 5. 9. This results in requiring only thirty cubes and. only 10. Using this algorithm. 9). (3. (5. 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.

3). 67. The ASCII codes for these characters are: 46. Note: The smallest possible number is 23 and the largest is 99. 8). The ASCII code for 23 is not a printable character. 2).:Z[yz. The 10. 9). (35). (5. three at a time. The solutions are identiﬁed by a three-digit number indicating the color of the buried cube and the two “edge” colors. (67). An inspection of the ASCII table will explain why it was decided to add 23. including pairs of two at a time. which is in the center of the tricube. and (99). Colors 1 and 3. (7. 98. (4.691 cases into 148 disjoint sets of isomorphisms. but not with any of the other colors. and 122. Any permissible interchange of colors is an isomorphism. (4. 9). 8). FARHI cubes of each color requirement and the restriction that the second number be greater than the ﬁrst. 68. Shown in Figure 4 are an edge solution and a center solution for the color combination of Figure 1. (1. The process was then continued until the list was exhausted. 91. If the solutions are to be catalogued.. 90. 6).160 S. cannot be interchanged with other colors. 9). ﬁve at a time including three at a time with two at a time. by adding twenty-three to each and using the ASCII character set codes (ASCII stands for American Standard Code for Information Interchange). four at a time. . has sixty isomorphisms. (68). its isomorphisms are determined and removed from the list.:Zppp. and six at a time including a triplet of two at a time. then additional criteria for recording solutions are recommended. (98). 7). and 99. can be swapped with each other. Not all of these mappings produce a new isomorphism. Subtracting twenty-three results in: 23.691 color combinations obtained by the algorithm described in Section 5 were listed in order according to their ASCII characters.:Z[yz. These were removed from the list. Different solutions often have the same three-digit identiﬁcation. Identiﬁcation was further simpliﬁed by placing these numbers in pairs (see Note below). the ﬁrst entry in Table 1 is . pairs of three at a time and four at a time with two at a time. (6. (7. Decoding a name is quite simple. The use of this six-character identiﬁcation greatly simpliﬁed the process of sorting the 10. 121. on the ends of the tricube. (24). i. and (8. (5. It is then a simple process to recognize that this represents the dicubes colored (1.e. 3). (2. . For example. Color 2. 58. This process then identiﬁed the 148 disjoint sets of isomorphisms. and then assigning a printable character to each. The remaining six colors may be interchanged in numerous ways: two at a time. The ﬁrst entry. 5). 35. 6). (4. The head of the list then became . Thus this color combination is identiﬁed as /:Z[yz.

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

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

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

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

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. O. particularly inspired by Muroi’s “Four Sticks” (so closing the circle of mutual inspiration). Nine and One-Half Moves. V AN DEVENTER Figure 1.” 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. Some months ago. The three pieces of the puzzle form a square with internal voids.166 M. Zigzag. Figure 2.

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

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

but must be used efﬁciently to solve the puzzle. Box 2: ¢ ¢ ½¼ Pieces: ½ ¢ ¿ ¢ . No Holes. rigorous analysis of the puzzle can be done by hand in about 15 minutes. ½ ¢ ¾ ¢ ¿ ¢ . although not exactly alike. A complete. The pieces of the second type are smaller and easier to use. 9. ¾ ¢ ¿ ¢ . not counting rotations and/or reﬂections. but they must not be wasted. The large box is moderately difﬁcult. but it does make a nice set of puzzles. ¾ ¢ ¿¢¿¢ ¢ . No Holes. The seven pieces can be placed in the box in ten different ways. ¿ ¢ ¿ ¢ ¿. ½ ¢ ¿ ¢ . The 18 pieces are all different. ¾ ¢ ¢ . Limited Number of Piece Types The puzzles I know of in this category follow a common principal: There are basically two types of pieces — a large supply of one type and a limited supply of another.¿¢¿¢ ¢ . (13) ½ ¢ ¾ ¢ In the ﬁrst of these. Quadron does not use any special tricks that I am aware of. ¾ ¢ ¿ ¢ . The rest of the pieces.170 B. In the second design. CUTLER 8. By analyzing “checkerboard” colorings of the layers in the box. The small box is very easy. ¾¢¿¢ . and has many solutions. Naef): Box 1: ¢ ¢ Pieces: ¾ ¢ ¿ ¢ ¿. . function similarly to the ½ ¢ ¾ ¢ ¾ pieces in the ﬁrst puzzle. The middle-size box is difﬁcult — there is only one solution. See [2] or [5] for more information. the three ½ ¢ ½ ¢ ¿ pieces must be used sparingly. and then the rest is easy. ¿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. ¢ ¢ Box 3: ¢ ¢ ½¼ Pieces: all 18 pieces from ﬁrst two boxes . The solver must determine exactly where the second set of pieces must be placed. ½ ¢ ¿ ¢ . ½ ¢ ¿ ¢ ½¼. (6) ½ ¢ ¾ ¢ ¾ unnamed (18) (John Conway): Box: ¢ ¢ Pieces: (3) ½ ¢ ½ ¢ ¿. the three individual cubes are obviously easy to place.¿¢¿¢ . (1) ½ ¢ ¾ ¢ ¾. and the three boxes offer a wide range of difﬁculty. Pieces Mostly Different º Quadron (18) (Jost Hanny. (1) ¾ ¢ ¾ ¢ ¾.¾¢ ¢ .

about once a second a piece is added or removed from the box! Using compiled BASIC. ½¼ ¢ ½½. ¢ ½¼. ½½ ¢ ½½ and two each of ¢ . but there are more efﬁcient algorithms that can be used for block-packing puzzles. There are 3. This leaves 4 pieces that might be placed sideways. º Parcel Post Puzzle (18) (designer unknown. the rate increases to 40 pieces/second. the widths and lengths are ¢ . The programmer can use algorithms that are used for pentominoe problems. ½ ¢ ¿ ¢ . The solution involves use of the following obvious trick (is that an oxymoron?): Some piece(s) are placed sideways in the box. 10 are too wide to ﬁt into the box sideways and 4 are of width 5. all very similar. ¢ ¾½. ¢ ½ . Since all the pieces are of the same thickness and the box depth equals three thicknesses. The analysis was done on about 20 powerful IBM workstations. a Commodore 64.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. Bill Cutler Puzzles): Box: ½ ¢ ½ ¢ ½ Pieces: ½¿ ¢ ½½¾ ¢ ½ ½. the machines had constructed 2 1/2 trillion different partially ﬁlled boxes. When running these programs on more powerful computers. ½ ¢ ¢ ¼.480 solutions. The second box was analyzed in about a minute of mainframe computer time. Of the 18 pieces. ¾¼ ¢ ¼ ¢ ¾. ¢ ½¿. ½ ¢ ½ ¢ ¾ ¢ . 1978. copied from a model in the collection of Abel Garcia): Box: ¢ ½ ¢ ¾ Pieces: all pieces are of thickness 2 units. ¢ ½ . ¢ . ½ ¢ ¼ ¢ . The Commodore 64 is such a wonderously slow machine — when running the program in interpreter BASIC. the difference between the three boxes is stunning: The ﬁrst box can be completely analyzed in a small fraction of a second. ¢ ½½. One or two individual layers can be constructed. º Boxed Box (23) (Cutler.450. The program displayed the status of the box at any instant using color graphics. and they all have three of these four pieces placed sideways. but the process cannot be completed. which is no good for this purpose. ¢ . ¢ ½ . I did a complete analysis of the third box. and ¢ ½¿. I wrote such a program on my ﬁrst computer. In early 1996. The total CPU time used was about 8500 hours. By the end of the runs. it is tempting to solve the puzzle by constructing three layers of pieces. or the equivalent of one year on one machine. . ¾¢ . There are four solutions to the puzzle. ¾½ ¢ ¢ ½ ¼. ¢ ½¿. ½ ¢ ¾ ¢ ¾. not counting rotations and reﬂections. I painted pieces of an actual model to match the display. The result was a fascinating demonstration of how a computer can be used to solve such a puzzle.

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

. 6(2). I won’t say which one (it should be easy to ﬁgure out). Spring 1973. 20–23. The Melting Block is more of a paradox than a puzzle. Mathematical Games column in Scientiﬁc American. being sure that the unﬁlled space is concealed at the bottom and is stable. pp. this is not another oxymoron.” J.” J. readers. This ﬁts into the box with a little room all around.” in The Mathematical Gardner. except one. but seems to the casual observer to ﬁll up the box completely. but rather a tautology to the 99% of the world that would never even have started to read this article. Hoffman. This should keep him busy for quite some time! References [1] W. 122–127. Recreational Math. Gardner. You are now prepared for your next encounter with a boring puzzle-nut. ½ ¢ ¾ ¢ ½ ¢ ¢ . “Packing Problems and Inequalities. one of the puzzles listed above is impossible. 104–111. Gardner.BLOCK-PACKING JAMBALAYA 173 12. pp. By the way. Place the box on your puzzle shelf with the remaining piece hidden behind the box. Klarner (Wadsworth International.) Pick up the box and last piece with both hands. ¿ ¢¾ ¢ .¿ ¢¾ ¢ .) This is a great puzzle to show to “non-puzzle people” and is one of my favorites. [2] M. February 1976. the pieces can be rearranged to make a ¢ ¢ ½¿¿ rectangular solid. 212– 225. pp. “Brick-Packing Puzzles. . [5] D. Pack all the pieces. [4] D.¿ ¢ ¢ ¢ ¢ . (No. 1979–80.½ . pp. and then dump the pieces onto the ﬂoor. It is a valuable weapon in every puzzle collector’s arsenal. being careful to keep the renegade piece hidden from view. Klarner. including the one in your hand.¿ ¢ ¢ plus a second copy of ½ ¢ ¾ ¢ . 12(2). 1981).. “Subdividing a Box into Completely Incongruent Boxes. [3] M. Cutler. into the box. February 1979. When the ninth piece is added to the group. Recreational Math. Miscellaneous ½½º Melting Block (8-9) (Tom O’Beirne) Box: ¢ ¢ ½¿¿ Pieces: ½ ¢ ¾ ¢ . 112–117. pp. The eight pieces ﬁt together easily to form a rectangular block ¢ ¢ ½¿¾. edited by D. Mathematical Games column in Scientiﬁc American. (This second construction is a little more difﬁcult. Show off the solved box to your victim.

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

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

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

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

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

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

“Le Pendu.” “Tandem Maze” (complex). jumping beans. 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. visiting places en route. bridges. etc.” Tomy’s “Crazy Maze.” puzzles using tops Pentangle “Roly-Poly” puzzles Ramps. HORDERN Detailed Puzzle Classiﬁcation Class DEXDEX-UNCA Description Dexterity Puzzles Dexterity or other physical skills in their solution Example Cup and Ball.” “Theo der Turnier. avoiding objects.” colour mazes. plain balls into holes Dexterity with sundry objects and/or obstacles Liquid objects Dexterity in liquid Indirect viewing Mechanized Using tools and magnetic tools Route following dexterity Objects concealed from view Electrica and electronic dexterities Pinball-related dexterities Other dexterities.180 J. DALGETY AND E. “Pike’s Peak” Icosian Game. number totalling mazes. Most hedge mazes RTF-2D Route path) mazes 2-D (any . 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. Konigsburg Bridges “Worried Woodworm. “Bootlegger” (complex) Ring and hole mazes.” “Yankee.” Engel’s “Black Box” Bagatelle “Frying Pan.

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

nine-piece ivory cube Pack the Plums. DALGETY AND E.” Loyd’s Donkeys. Josephus.182 ASS-POLY ASS-SHAP 3-D Assembly Polyhedra and Spheres 3D Assembly Other Shapes J. HORDERN Ball pyramids. “Managon. BlackBox. Squashed Soma. and weaving 2-D patterns Patterns with 2-D parts on 3-D surface “Shmuzzles” tesselations JIG-PART JIG-SLOP JIG-LAYR JIG-2D3D JIG-BLOX PATPAT-2DPG Bilhourd’s jigsaws “Sculpture Puzzles” “Toyznet” Queens on chess board.” “Phoney Baloney. or pieces. Apple and Worms. number puzzles Heads and tails “Testa” PAT-STIX PAT-NUM PAT-2DEG PAT-2D PAT-STAK PAT-2D3D Stacking transparent layers. weaving puzzles “Dodeca” . overlapping.” “Even Steven. pegs. Waddington’s “Black Box” Match puzzles. “Lapin. 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. Jensen’s “Tricky Laberint” Magic squares.” 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.

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

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

soot on unglazed part of ashtray Landscape turned to make a portrait.” Penrose triangle “Vanishing Leprechauns. impossible dovetails. Topsey Turveys.” “Find the 5th Pig” needing coloured overlays.” random dot stereograms “Naughty Butterﬂies. Oskar’s “Escher Puzzle. “Spot the Difference. 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 . 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.” Hooper’s paradox Anamorphic pictures Pictures and objects of one subject made from completely unrelated objects Devinettes (obscure outlines). courtship/matrimony Optical illusions.CLASSIFICATION OF PUZZLES OTH-PEND AMBAMB-POBJ Pending Classiﬁcation !!! Ambiguous Paradoxical objects (objects that apparently cannot be made) Vanishing images 185 Puzzles awaiting classiﬁcation Arrow through bottle. OHOs.

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

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

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

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

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

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

Or arguments as to why this is too mushily stated a challenge. .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. Instead of searching directly for a task last misﬁring. except to review all the math and geometry that you already know. for Ò ½¿. I invite further contributions to the collection. say.

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

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

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

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

The ﬁrst player has a winning strategy in n-spot Sprouts if and only if n is 3. Sleator believes in the Sprouts Conjecture and the Misère Sprouts Conjecture. a “2” means the second player has a winning strategy.edu/~sleator . Misère sprouts conjecture.cmu. and an asterisk indicates a new result obtained by our program. 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 ﬁrst player has a winning strategy in n-spot misère Sprouts if and only if n is 0 or 1 modulo 5.½ ½For a paper describing these results in more detail. This problem seems to lie just beyond our program. To reduce the space still further we used a data compression technique. The n-spot Sprouts positions evaluated so far fall into a remarkably simple pattern.cs.COMPUTER ANALYSIS OF SPROUTS 201 ¯ of previous searches in a hash table. characterized by the following conjecture: Sprouts conjecture. The data for misère Sprouts ﬁt a similar pattern. Applegate doesn’t believe in these conjectures. go to http://www. and our ingenuity. and searching the successors of a position in order from lowest degree to highest. We are still left with the nagging problem of resolving a bet between two of the authors. 4. The size of the hash table turned out to be a major limitation of the program. or 5 modulo 6. We discovered that saving only the losing positions reduced the space requirement by a large factor. our computational resources. and we devised and implemented two methods to save space without losing too much time efﬁciency. Our Results Here’s what our program found: Number of Spots normal play misère play 1 2 1 2 2 2 3 1 2 4 1 2 ¾ £ ¾£ ½£ ½£ ½£ ½ £ ½£ ¾£ ¾£ ¾£ 5 1 6 2 7 8 9 10 11 A “1” means the ﬁrst player to move has a winning strategy. The only remaining case required to resolve the bet is the 10-spot misère game.

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

GOSPER Times 100–105: .204 B.

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

) Construction of Oscillators. The turn can take either 64. one period 4. speed c/4 diagonal spaceship (not the glider). speed c/4 orthogonal spaceship. which implies that all sufﬁciently large periods are possible.. 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.) Periods ½¿ · Ò and ½ ¼ · Ò use a mechanism developed by Buckingham. and 47 are the only ones we know with the right sparks. For the multiples of 8. For example. or 47. Also. (These are used in some of the large patterns with unusual growth rates mentioned later. More recently. A period 150 gun ﬁres toward a distant pair of pentadecathlons. of periods 5.206 B. one period 4. but 5. Hartmut found another with speed 2c/5. We also have an argument. by combining them we can build a closed loop whose length is any large multiple of 5 or 8. we put several copies of the B in the track to reduce the period to one of those mentioned.) Lately. we now have nontrivial examples (i. I built a p155 oscillator in which 20 Bs travel around a track of length ¿½¼¼´ ¾¼ ¢ ·¾ ¢ µ generations. it consists of a still life (“eater3”) and two spark producing oscillators. based on construction universality. so we also get glider guns of those periods. The ﬁrst glider to hit them is reﬂected . Due mostly to the work of Buckingham and Wainwright. in which a B heptomino is forced to turn a 90 Æ corner. inﬁnitely many period 3.) 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). Another amusing oscillator uses the glider crystallization that Bill mentioned in his centinal letter a few years ago. the 73 gen turn can emit a glider. he and Hartmut Holzwart have been producing oodles of orthogonal spaceships with speeds c/2. 6. a few oscillators with periods 5 (including some with useful sparks) and 6. 65. inﬁnitely many period 2. speed c/3 orthogonal spaceships. (Any greater nonmultiple of 4 would also work. The · ¾ Ò ¾ · ¾ Ò ¼ · ¼Ò ¾¿¼ · ¼Ò ¾ ¾ · ¿ Ò and · ¿ Ò use a device discovered by Wainwright. and one period 5. speed c/2 orthogonal spaceships. 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. and c/4.e. or 73 generations. JHC!). c/3. 6. which reﬂects a symmetric pair of parallel gliders. (Mine runs only on an Apple II. (No doubt you ﬁgured that out for yourself long ago. speed 2c/5 orthogonal spaceship. David Bell wrote a similar program that runs on faster machines.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

850 c. 300 B. 3. 11. 2. and to the second thirty apples. c. 1) (30. 3. 3. 2. I quote a version that appears as No. 7) I-(7. 2. (1. 1. although the material is believed to be based on older Arabic sources. 7) II-(1. 5. 3. 510 c. and brought home in like such money. 10. 5. 2. Sridhara [18]. 10.” printed by Wynken de Worde in 1511 [9]. 10. 2. 3. 0. 11. 6) (1. 3. 11. 22. certain apples. 2. 23. 6) I-(7. Now how many sold each of them for a penny? . with the problem statement giving the 6 as a 7 Fibonacci Fibonacci. 10. 11. 9. 50. 5. which is rather a mouthful. 5) (2. The problem is treated in one form by Mahavira [15]. 3. 8. 3) (10. and to the youngest ten apples. and Bhaskara [4]. 6) I-(7. 5. 10) (7. with the problem statement giving the 6 as a 7 Fibonacci. 5) c. 9. 9. 250 c. A man had three daughters of three ages. which daughters he delivered. noting that this system is inconsistent Fibonacci.C. 1. there seems to be no common name for it. 5.SOME DIOPHANTINE RECREATIONS 225 from the early fourteenth century and hence after Fibonacci. 8. 2. 1. 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. 5) (100. to sell. 6) I-(7. 850 c. 4. 4. 9. 2) I-(9. And he took to the eldest daughter ﬁfty apples. 4) (2.5) I-(25. 4. 2. and all these three sold in like many for a penny. 4. 5. so I will ﬁrst describe the later form which is ﬁrst found in Fibonacci [11]. hence the above title. 50 in the ﬁrst printed English riddle collection “The Demaundes Joyous. 7) I-(7. 6. 2. 4. 11. 3. 1. 510 9Ø century c. 4. 9. 5. 11. but this form is less clearly expressed and has inﬁnitely many solutions. done two ways Selling Different Amounts at the Same Price Yielding the Same Although this problem is quite old.

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

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

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

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

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

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

11. 1 given. 56. 1 given. 40) (18. 1 given. 33. 32) (12. . 1 given. 15) (31. 50) (30. 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.232 Version (10. (20. 32. 6 given. 1 given. 55) (55. (20. 49) (3. 25. 17. 1 given. 80) (10. none given. 36) (27. 100) (45. 68. 4 given. 48. 37) (27.400) (25. 136) (3. 1 given. 1 given with fractions. 1 given. 901) (10. 81) (55. 40) (10. 33) (10. 36) (78. At this point. 140) . . 30. 30) (18. 20) (10. 170) (305. 36) (30.552. 1) (100. 40. 1 given. 30. 21) (225. 119. 1 given. 30. SINGMASTER Solutions in Source 5 given. 60) (117. one with fractions. ¬ .9) (100. 55) (78. 81) (10. 20) (64. 12. 1300 1300s 1489 1489 1489 c. Sridhara ﬁnds a one-parameter solution set. 20. 90) Date 1202? 1202? 1202? c. makes this more speciﬁc by saying they ‘sold a part in lots and disposed of the remainder at one for ﬁve panas. 454. 1 given. 603. 20. 1 given. 50) (10. 82) (17. 1 given. 1) (7. 196) (45. 35) (11. 50) (11.6) # of Solutions (55. 40. 55) (16. 60) (10.4) (70. 30) (12. . all given. 26) (20.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 1 given. 16) D. 29. 30) (8. 16) (16. though it is not as easy to see how to do this as in the Western version. 60) . so we again have a three-dimensional solution set. 752. none given. 40. One can scale the set . 33) (20. 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. 36) (25. 2 given. 1 given. in some way to reduce the solution set to two dimensions. Bhaskara makes the following comment: “This.

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

Wynken de Worde. 5th ed. Wiesbaden. Problems to Sharpen the Young. 1805. “A classical Indian puzzle-problem.so and ¬ ½½¼. pp.234 D. and the 13Ø edition. 2nd ed. Paris. 102– 126. 13. Prob. 1824. 12. 3200 and 16. H. reprinted by Dover. 1150. Algebra.(475). [5] Bonnycastle. 54.. 1884. 1782. 3292. he produced a 10 Ø ed.” which “may be considered as a concise abridgment” of his two-volume Treatise on Algebra.” [2] Ayyangar. 1511. By the way. London. Problemes plaisants & delectables qui se font par les nombres. Prob. 1968. Indian Math.. Book I. I examined the 7 Ø edition. 150 B. Labosne. 1959. Johnson. Revised by A. Paris. 1624. annotated by David Singmaster and John Hadley. 4392. 10. An Introduction to Algebra. 1813. In: T. 24. [8] Diophantos. Math. J. L. 106–115. London. [6] Bunt. Rigaud.C. 16: “De duobus hominibus boves ducentibus — Two men leading oxen.. Heath. 1879. John Murray.” c. so . pp. 1910. P. 214–223. Lucas N. Krishnaswami. with Notes and Observations. and Other Places of Public Education. 1874. giving total numbers of 3295. Rigaud. 1817. London. Analysis similar to that done before shows that there are then ½ · ´ ¬ ½ Ò µ ¬ solutions.. reprinted by Blanchard. 1973.470 [4]. [9] The Demaundes Joyous. Vieweg. London. Braunschweig. “an entire revision of the work.. et al. SINGMASTER For given and ¬ satisfying ´ ¼¼¼ µ and ´ ¼µ. Bîjaganita. 1923–24. A. designed for the Use of Schools. 86–88. Bhaskara solves his example by taking ¾. In: Henry Thomas Colebrooke. Vogel. 15. there are only a ﬁnite number of solutions. Lyon. 18... Neun Bucher arithmetischer Technik.. Bijanganita. pp. J.. The Historical Roots of Elementary Mathematics. v. The ﬁrst nine editions appeared “without any material alterations. Facsimile with transcription and commentary by John Wardroper.” In 1815. pp. 76. Mar 1992. Prob. “Nine Chapters on the Mathematical Art. pp. Oxford University Press. [7] Chiu Chang Suan Ching. 4th ed. Gordon Fraser Gallery. 242–244. Nos. John. with Arithmetic and Mensuration from the Sanscrit of Brahmegupta and Bhascara. 134–136. nos. 6. 21. pp..” J. Soc. 3rd ed. Gaz. 33.. 122–126. . [3] Bachet. which may be the same as the 1815 edition. Translated by John Hadley. P. Lyon. 1964. 1976.) Chap. 5th ed. trans. Claude-Gaspar. PrenticeHall. 2nd ed. p. 178–186. aka Bhâskara. London.900 and of 3294. 1st ed. pp. 24. including Sandig. Diophantos of Alexandria. Nunn et al. Arithmetica. See No. A.. 19. 15. 170. 1612. (There have been several reprints. [4] Bhaskara. pp. Translated into German by K. Gauthier-Villars. References [1] Alcuin.

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

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

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

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

Yamasaki. 1979. 818–827. Simon and Schuster. then by erasing the stone in the encircled cell. Bleicher. A winning Opening in Reverse Hex. The position in the real game is obtained from the position in the imaginary game by putting É’s stone in the encircled cell. . The American Mathematical Monthly. Theory of Division Games. 1959. Inst. Res. and transfers the circle from its current location to the new cell. M. SLEATOR The imaginary game and the real game differ in up to two places. pp. Excursions into Mathematics.. J. [2] R. Recreational Mathematics. Publ. pp. Worth. Beck. This contradicts our assumption that È has a winning strategy. Sci. Evans. Player É eventually wins the imaginary game. The position in the imaginary game is a winning position for É. It is easy to see that the relationship between the real game and the imaginary game is maintained. and changing the contents of Ô¼ from a É-stone to a È -stone. and J. 73–83. 337–358.240 J. Gardner. The imaginary game is obtained from the real game by ﬁrst changing Ô¼ from È ’s stone to É’s stone. Kyoto Univ. References [1] A. [5] Y. Crow. pp. 1978. 327–339. The Game of Hex and the Brouwer Fixed-Point Theorem. [4] M. 1969. 86 (10). Gale. 14. then É arbitrarily chooses a different empty cell in which to move. 7(3). pp. The fact that Ñ´Ä Ê µ ½ ensures that both players can continue to move. as follows. Math. and the monotonicity property ensures that the corresponding position in the real game is also a win for É. pp 189–192. If the strategy ÄÊ requires a move into the encircled cell. [3] D. New York. New York. LAGARIAS AND D. Summer 1974. The Scientiﬁc American Book of Mathematical Puzzles and Diversions.

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

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

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

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

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

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

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

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

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

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

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

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

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

Since a number Ò is multiplied by ½ if it is even and approximately ¿ if it is odd. which concern extreme behaviors of ¿Ü · ½ iterates. [24]. [5] proposed using the ¿Ü · ½ function as a pseudorandom number generator. These models consist of a repeated random walk model for forward iterates by Ì . There is excellent numerical agreement of this model’s prediction with ¿Ü · ½ function data. Lagarias [14] and Luby [17]. From this viewpoint it becomes interesting to determine how well the properties of ¿Ü · ½ iterates can be described by a stochastic model. Garner [11] and Korec [12]) they also seem to possess some residual structurelessness. In doing so we are in the atypical situation of modeling a purely deterministic process with a probabilistic model. [24]. et al. On the positive side. Lagarias and Weiss [15] studied two different stochastic models for the behavior of ¿Ü · ½ function iterates. one expects that ¾ ¾ on average it changes multiplicatively by their geometric mean. for an ¾ explicit constant ½. 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. obtained in joint work with Alan Weiss and David Applegate. which is ´ ¿ µ½ ¾ . and a branching random walk model for backward iterates of Ì . [20]. which may be enough for a pseudorandom bit to be extracted. one expects the iterates to decrease in size and eventually become periodic. The pseudorandom character of ¿Ü ·½ function iterates can be viewed as the source of the difﬁculty of obtaining a rigorous proof of the ¿Ü · ½ Conjecture. the 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. These mathematical models predict that the expected size of a “random” Ò after steps should be about ´ ¿ µ ¾Ò. cf. Here we survey some recent results on stochastic models. so that the average number of steps to reach 1 should be ´½¿º¾ µ ½ ½ ÐÓ ÐÓ Ò ¾ ¿ ¾½¾ ÐÓ Ò Furthermore.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. [15]. see [20]. Even though ¿Ü · ½ function iterates possess quite a bit of structure (cf. Since this is less than ½. 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. This model has the drawback that . Several heuristic probabilistic models describing ¿Ü ·½ iterates are based on this idea. The simplest of them assumes that this independent coin-ﬂip behavior persists until ½ is reached under iteration. [13]. Cloney. see [9].

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

£ and the expected size of ½ ´Òµ is given by (13." 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.24). we consider the random variable Ê ¾ ÐÓ ½ ¾ · ¿ ¾ Ö£ ´Òµ ½ ½ ´Òµ Ò Ì ´ µ ´Òµ Ì ´ ½µ´Òµ with ½ ½ ´Òµ Ó Large deviation theory predicts that with probability 1 it satisﬁes Ð Ñ ×ÙÔ Ö£ ´Òµ Ò ½ ³¼ ¼ ¼ ¼ The quantity corresponds to an estimate of the maximum fraction of elements in a ¿Ü ·½ trajectory that are odd. For the extremal trajectories for £ ´Òµ in (14.4). this explains the name “repeated £ random walk model. with probability 1. The length of such trajectories approaches the limiting value ½ ´Òµ ÐÓ Ì ´ µ ´Òµ ÐÓ Ò ÐÓ Ò ¼ ½ ´Òµ .1) it is a straight line segment with slope ¼ ¼¾¿½ starting from ´¼ ½µ and ending at ´ ¼µ. Large deviation theory also asserts that the graph of the logarithmically scaled trajectories ´½ º µ approaches a characteristic shape. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally. For the extremal trajectories for Ø£ ´Òµ in (14. Large deviation theory shows that. The random variable ½ ´Òµ is the step number £ with probability ¾ at which 0 is crossed. This random variable is ﬁnite with probability one. so that Ø£ ´Òµ ÐÓ Ò. let Ø£ ´Òµ equal the maximal value taken during the random walk.3X + 1 FUNCTION ITERATES 257 ¿ by starting at ÐÓ Ò and taking independent steps of size ÐÓ ¾ or ÐÓ ¾ ½ each. For each initial value Ò an entirely new random walk is used. see Figure 1. the limiting graph of these trajectories has a different appearance.

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

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.884.769.48 18.6026 0. an analysis in [15.194. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale.799 6 8.5818 0.815 10 13. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1.180.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6.266.527 5 837.5841 0.13 26.1) with numerical data up to ½¼ ½½ is given in Table 1.728.890. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ´Òµ ¿ ¾ .6091 Table 2. He found that Ò 12.38 41.24 16.5927 0.511 7 63.4).5967 0.171 4 52.64 32.91 19.127 8 127. Section 5] shows that in ½¼ ½½ trials one should only expect to ﬁnd a largest ´Òµ ¿¼. The agreement with the stochastic model is quite good.456.6020 0. where ´Òµ ÐÓ Ò .68 24.5741 0.6030 0.68 Ö´Òµ 0. ½ ½¼. More recently V. The large deviations extremal trajectory is indicated by a dotted line.328. and the large deviations extremal trajectory is indicated by a dotted line. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good.371.254 9 4. Maximal value of ´Òµ in intervals ½¼ Ò ½¼ ·½. ½ ´Òµ It gives the longest trajectory. A comparison of (14.77 31.400. He also found a number around ½¼½½¼ with ´Òµ exceeding ¿ ¼ Ø´Òµ in intervals ½¼ Ò ½¼ ·½ up to ½¼½½ with the prediction of (14. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory.5857 0. .527 has ½ ´Òµ ½¾ ½ and ´Òµ ¿ ¾ 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. Extensions of the data can be found in Leavens and Vermeulen [16].6020 0.91 32.527 Random walks model ½ ´Òµ ´Òµ 70 108 165 214 329 429 592 593 706 755 21. Vyssotsky [23] has found much larger numbers with high values of ´Òµ.94 31.

819 1.560 1.817.528. respectively. Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2. LAGARIAS Ò 1 27 2 703 3 9.645 ´Òµ 21.02 21. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼. see [15] for more explanation.566. ½ ½¼.831 9 8. (See Table 3 in Section 4.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½.05 19. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth . In iterating the multivalued function Ì ½.24 16. using (15.65 7. there are duplications.790 1.46 2. these trees have the minimal and maximal number of leaves possible for depth . 15.32 2.03 19.976 1.362.83 3. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿.631.14 11.59 23.791 1.86 7.20 6. or 7 ´ÑÓ µ. where ½ ´½ º½µ Ì ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½.511 10 77. 4.675 7 80.06 4.10 20.099 1.65 5. The set of inverse iterates associated to any Ò has a tree structure.511 6 6. Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk.) empirically like .48 12.903 2.559 Random walks model Table 3.1).788 1. 5. C. In fact. so do not have any signiﬁcant effect on the size of the tree. In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed.391 8 319. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ.909 1. .86 5. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ.663 4 77.75 5. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ. if Ò ¾ or 8 ´ÑÓ µ. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree.86 13. In this case. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch.049.897 2.84 19.000 Ò 13.260 J.804.671 5 704.

where ¬ is the unique positive number satis£ ´¬ µ ¼. We assign labels ¼ or ½ to the edges. The random walk at a given leaf node ends at a position Ù on the real line. In [15] it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ . È ÊÏ .2). and the root node is assigned the value 1. [15.3X + 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.. 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. In order to prove the result (15. cf.e. i. satisﬁes (with probability one) ´½ º¾µ ÐÓ Ò£ ´ µ È³ ½ ÐÑ ½ ÐÓ This constant È ¬ ½ . Theorem 4.4] show that the expected size of ÐÓ Ò£ ´ µ. and where fying £ ´ µ ×ÙÔ ´½ º¿µ ÐÓ ´¾ · ½ ¾ ¿ ¿ ¼ This constant is provably the same constant as (14. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½. and the value assigned to that vertex by the rules above is exactly Ù . it was necessary to determine the distribution of the number of leaves in a tree in the branching process. 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. 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 . where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree.1].1). Pruned ¿Ü · ½ trees of depth Ò . . Theorem 3. Lagarias and Weiss [15. 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.

they proposed that the normalized extreme quantities remain bounded. 4.262 J. 16. 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. 8 pictured in Figure 4. The branching process is a multi-type Galton–Watson process with six types labeled 1. Since the number of leaves is expected to grow like ´ ¿ µ .) The random walk arises from assigning the labels ¼ and ½ to the edges of the resulting tree. 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. Based on this data. Let Æ ´ µ and Æ · ´ µ denote the minimal and maximal number of leaves. the corresponding random walk . LAGARIAS and as ½ the leaf probability distribution is of the form Ï ´ ¿ µ . 2. C. with all nodes that have a label congruent to ¼´ÑÓ ¿µ removed. Here the pruned 3x+1 tree Ì ´ µ is the tree of inverse iterates of the ¿Ü · ½ function grown backward to depth . (See Athreya and Ney [4] for a general discussion of such processes. respectively. 7. Distribution of 3x + 1 Trees Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees. from a given root node . This can be formalized as Conjecture C . 5. Theorem 3. In this table. Ê´ µ counts the number of distinct edge-labeled ¿Ü · ½ trees of depth . where ¼ represents the entering vertex being even and ½ represents the entering vertex being odd. they studied the normalized extreme values ´½ º½µ ·´ µ ¿ Æ ´ µ Ò ´ µ ¿ Æ ·´ µ Table 3 gives data up to ¿¼.2]. see [15. 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].

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

746 1.486 0.750 0.492 0. .264 J.562 0.633 0.310 2.67 0.419 2.688 1.307 0.778 1.774 1.342 0.99 13.45 560.481 0.026 2.118 2.742 1.268 0.669 1.327 2.49 9. C.475 0.869 1.923 1.34 420.232 2.054 2.748 0.12 74.741 1.488 0.390 2. 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 1.75 5599.475 0.422 0.782 1.728 1.489 0.375 2.688 1.16 4.482 0.238 1.758 1.252 0.747 1.76 23.481 0.791 1.33 1.21 5.36 3149.166 2.746 1.489 0.273 0.255 2.394 0.507 0.748 1.534 0.249 0.076 2.465 0.753 1.281 0.190 2.77 133.758 1.68 31.633 0.488 0.802 1.792 1.534 0.422 0.211 2. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a ﬁxed node Ò´ÑÓ ¿ ·½µ.481 0.911 1.562 0.898 1.81 4199.60 747.47 996.490 0.500 1.433 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 4.50 315.780 1.557 1. Normalized extreme values for ¿Ü · ½ trees and for the branching process.09 56.37 3.270 2.131 2.77 2362.782 1.243 0.475 0.802 1.57 42.62 7.62 1328.78 2.255 0.802 1.344 2.738 1.409 0.375 0.289 0.750 0.451 0.534 0.405 2.491 0.265 0.491 1.83 99.500 1.271 0.305 0.750 1.401 0.784 1.03 177.751 1.486 0.762 1.292 2.490 0.302 0.300 0.38 236.898 1.986 2.738 1.352 0.240 0.481 0.246 0.360 2.32 17.501 0.507 0.83 1771.259 0.490 0.

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

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

Sign up to vote on this title

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

Are you sure?

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

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

Get the full title to continue

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