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

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

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

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

His breakcame when he sold a humorous short story called “The Horse on the Escalator” to Esquire magazine. too.” “I amused myself on nightwatch by thinking up crazy plots. but it has taken world chaos to make his father say that his oldest son is perhaps the sanest of his family.. 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... accept 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.Would Mr.. then based in Chicago. [18] *** His career as a professional writer started in 1946 shortly after he returned from four years on a destroyer escort in World War II. straight magic and card tricks. the University of Chicago. Those sales marked a turning point in Gardner’s career. but he doesn’t like being branded.. Gardner was hanging around his alma mater.He called up Mr.” said the soft-spoken Gardner. 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. Christ. Christ and asked if he might come out to Christ’s home. “pitchmen’s” novelties. The editor invited the starving writer for lunch at a good restaurant.. His personal philosophy has been described as a loose Platonism.Still ﬂush with musteringout pay. who had the largest collection he had ever heard of. If he were to rest his thoughts upon one quotation it would be Lord Dunsany’s: “Man is a small thing. To Gardner’s family his way of life has at last become understandable. .” [5] *** Martin Gardner ’36 is a professional [sic] magician. 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. but where to park them while he was peregrinating over the globe was a problem.6 D. and he thinks Plato. might object with sound reason. He arrived with a large suitcase full of puzzles! Puzzles had been a hobby with him.During the past few months a determined outpouring of ideas for booklets on paper-cutting and other tricks. Those mental plots evolved into imaginative short stories that he sold to Esquire magazine. and the night is large and full of wonder. He tours the world pulling rabbits out of hats. writing and taking an occasional GI Bill philosophy course.

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

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

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. And then every once in a while I get a sort of scoop. ‘you just read books by the professors and rewrite them?’ That’s really what I do. Vladimir Nabokov pays a twinkling tribute by introducing one Martin Gardiner.’ he said. For example.Let me say that I think not knowing too much about a subject is an asset for a journalist. Auden constantly quoted from Gardner.” [14] *** “I’m very ill at ease in front of an audience. It’s a very intriguing paradox and I’m not sure that it’s even resolved. I have also managed to provoke some outspoken enemies. ‘You mean you’re in the same racket I am. as the mathemagician admits.” Nevertheless. Matrix that the millionth digit of . Professors at Stanford University have just programmed a computer to carry to the millionth digit.. The last scoop that I got was when I heard about a public-key cryptography system at MIT.D. poet W. In his novel Ada. P. I did a column a few years ago on a marvelous paradox called Newcomb’s paradox. His wife interjected: “The fact is he doesn’t want to do . whom he calls “an invented philospher.” Gardner said.MARTIN GARDNER: A “DOCUMENTARY” 9 where I got my Ph. and “proof” by a ﬁctional Dr. not a liability. The late Jacob Bronowski was a devotee. 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.. “not all my readers are fans. Snow’s famous “gulf of mutual incomprehesion” that lies between the technical and literary cultures. in decision theory. if it were ever computed. and that stumped him for a moment. would be the number 5. [8] *** “I particularly enjoy writing columns that overlap with philosophical issues. and that was the ﬁrst publication the general public had on it.” [20] *** His “Mathematical Games” column in Scientiﬁc American is one of the few bridges over C. I said I didn’t have one and he looked startled. 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. 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.” [11] *** “I can’t think of any deﬁnition of ‘mathematician’ or ‘scientist’ that would apply to me. I realized what a big breakthrough this was and based a column on it. H.

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

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

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

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

his disturbance went much deeper. Byrd: All tight. As I said.14 R. Doyle insisted that Houdini was lying! If that’s not psychotic paranoia. 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. I have no idea who wrote it. he was surely one of the most interesting writers of the last century. whamo! Ambrose (After a pause): That’s quite a cute idea! But frankly. when the public was convinced of his rationality. how could someone of Gardner’s caliber ever entertain the silly notion that Doyle never wrote the Holmes stories? Ambrose: To me the solution is obvious: Martin Gardner never wrote that chapter! The chapter is a complete forgery. Ambrose: I’m glad you realize that. Byrd: But now something else puzzles me: Martin Gardner was no fool. then. I never had the slightest doubt that Doyle wrote the Holmes stories. he absolutely refused to believe Houdini when he said that there was a perfectly naturalistic explanation for the escapes. 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. 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. but now your explanation of the apparent contradiction between Doyle the rationalist and Doyle the crank makes some sense. Then. SMULLYAN also pointed out that Doyle’s interest in spiritualism started much earlier in life than is generally realized. what is? Byrd: I guess you’re right. Now. Don’t you know that he believed that the famous Harry Houdini escaped from locked trunks by dematerializing and going out through the keyhole? What’s even worse. . it’s just as implausible as Gardner’s idea that Doyle never wrote the Holmes stories at all.

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

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

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

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

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

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

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

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.roughaveragespeedofEarth 700.000mph.000 mph turning speed of Galaxy 1.4 million mph Galaxy’s speed through the debris of the Big Bang .

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

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

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

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

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

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

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

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

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

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

) These curious combinations occur 660 times a day. (3) Find the longest time span without a palindromic time. (2) Find the two palindromes whose difference is closest to 12 hours. 44. Problem 5: Calculate the expansion of the following 26 terms. 13:22:31. 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. 30. 6. 42. 54. x. 23:55:32. 50. 4. 34. y Problem 7: The ﬁgure shown here is the solution to the problem of dividing the ﬁgure into four identical shapes. Problem 6: Which two numbers come at the end of this sequence? 2. 36. (1) Find the closest such times. 64. 40. 1:01:01. 62. (Ignore the colons. 3:59:53. 46. .38 N. 66. 2:41:42. 56. 32. etc. YOSHIGAHARA Problem 4: Arrange the following ﬁve pieces to make the shape of a star. For example. 52. 60.

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.CREATIVE PUZZLE THINKING 39 Problem 9: A right triangle with sides 3. . 4.

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

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

Michael W. Ecker

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

**The Sisyphus String: 123
**

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

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

42

M. W. ECKER

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

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

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

**What Is a Mathemagical Black Hole?
**

There are two key features that make our example interesting:

MATHEMAGICAL BLACK HOLES

43

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

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

44

M. W. ECKER

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

Words to Numbers: 4

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

**Narcissistic Numbers: 153
**

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

MATHEMAGICAL BLACK HOLES

45

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

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

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

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

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

6 or so. but you will reach 6174. 5. ECKER Example: Start with 8028. but your very equilibrium. Its divisors are 1. This one may defy not only your ability to explain it. and ´Òµ preceding two elements. we would have obtained the reciprocal of phi as the limit. This one is a bit more tedious. Your own example may take more steps. The ratios seem to be converging to a number around 1. The ﬁrst two numbers numbers: 1.). Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . and 15.618 . 8. 13. including 1 and the number itself. W. ½ . the smallest is 0288. 3. To get a quick and dirty feel for why phi arises. Had we used the ratios ´Òµ ´Ò · ½µ instead. Now take the sum of the digits of these divisors. Iterate until a number repeats. ´½µ ´Ò ½µ · ´Ò ¾µ for integers Ò ¾. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. but fewer authors use that approach.48 M. Divide both sides of the equation ´Òµ ½µ · ´Ò ¾µ by ´Ò ½µ. but it is also that much more strange at the same time. 21. The black-hole number this time is 15. 89. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. In fact. Its value is roughly 1. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. 55. and these digits sum to 15. etc. 5. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. The largest permutation is 8820. Repeat with 8532 to calculate 8532–2358 = 6174. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò call it Ä — as Ò increases. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. 3. More formally. That is. 34. 2. phi. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors. 1. and the difference is 8532. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ Ä¾ Ä ½ ¼ . but without using a program.

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

Unsolved Problems as Black Holes Even unsolved problems sometimes fall into this black-hole scenario. Because this is so easy to program. Must you always reach 1? If you start with 5. iteration of matrix powers draws one to a result that represents long-term stability independent of an input . As things are set up. then 2. Once you hit 4. take half. Keep iterating. I will omit a program here. the numbers ´Ò ½µ ´Òµ. a cycle a cycle of length three. ECKER obtain our two golden numbers. and conversely. All real inputs except zero. you stay at 4. If odd. then 4. Ulam. this problem has the paradoxical property that. the number ½ ½ is an attractor with the black-hole property.) Now triple the largest odd factor and then add 1. W. although one of the hardest to settle deﬁnitively. 84 is ¾ ¢ ¾ ¢ ¿ ¢ . while ¼ ½ is a repeller. dating back to the 1930s and still an open question (though sometimes also identiﬁed with the names of Hailstone. Under certain conditions. Now try some examples. then 1 again. Close Other examples of mathemagical black holes arise in the study of stochastic processes. triple and add one. is among the easiest to program (a few lines). then 1. then 2. Hmm now only in black holes. let’s be creative and ﬁx this up by modifying the problem. The only non-odd prime is 2. If you do reach 1 — and nobody has either proved you must. We’re interested that repeats ad inﬁnitum. (Just multiply all the odd prime factors together. Anybody who proves the Collatz Conjecture will prove that my variation is a mathemagical black hole as well. and the second golden number lead to the attractor. This answer is the next iterate. and ¿¢½·½ . Pick the largest odd factor. So. nor shown any example that doesn’t — then you next get 4. Deﬁne the process instead by taking the starting number and breaking it down completely into factors that are odd and even. you get 16.50 M. Success! In fact. our black hole called phi. then 8. which really are cycles of length one. and Syracuse). Consider the Collatz Conjecture. In the example that would be ¿¢ ¾½. Start with a natural number. because the largest odd factor in 4 is 1. You should ﬁnd that you keep getting 4. For instance. If even.

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

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

Find the most efﬁcient solution. 7. 8. 5. How long does it take? E5. 4. 3. 10 pints. 25 and 9 combine to give 16) and use the rule to determine Ü. After going 10 miles in a straight line he ﬁnds himself back at camp. The 15pint container starts out full. reverse the digits of the result. Find a nine-digit number made up of 1. E9. 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. Through transferring liquid among the containers. and the other two start out empty: (15.. My regular racquetball opponent has a license plate whose three-digit part has the following property.5 nautical miles from Los Angeles to Honolulu. 63 9 38 33 32 12 88 25 16 18 15 Ü 5 E3. I. E6. 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. 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. 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. 2 and 1. 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. Divide the number by 3. Calculate the expansion of this 26 term expression: ´Ü µ´Ü µ´Ü µ´Ü µ ´Ü Ýµ´Ü Þ µ . What is the number and what is the next greater number (possibly with more than three digits) having this property? . subtract 1 and you produce the original number. 4. 7. 0. 6. 5. 0). 8. 6. Find the rule for combining numbers as shown below (e. It is approximately 2244.54 R. g. 2. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu. HESS up and travels due North. Where on earth could he be? E2. You have containers that hold 15 pints. and 6 pints.

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

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

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

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

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

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

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

(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. where he (1) picks out three socks at random. I’ll let you off the hook.0107 . (a) The ﬁgure shows two Pythagorean triangles with a common side where three of the ﬁve side-lengths are prime numbers. I. and proceeds to explain: “I normally ask guests to determine the ages of my three children. Deﬁne the goodness of ﬁt of a /b to x as g (a.0107 0. six black. given the sum and product of their ages.a . e ) 2.” Your response is “No need to tell me more.62 R. x) = a bx . . Minimum Cutting Length. where e = H16. (b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes? . into ﬁve parts of equal area? (c) an equilateral triangle into ﬁve parts of equal area? H15. their ages are ” H14. My uncle’s ritual for dressing each morning except Sunday includes a trip to the sock drawer. ) 0. Since Smith missed the problem tonight and Jones missed it at the party two years ago.0107.71828 . The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem. The drawer starts with 16 socks each Monday morning (eight blue. b. b) Find the minimum c and d so that g (c. d. ) = 0. two brown) and ends up with four socks each Saturday evening. HESS H12. An approximation to is 355/113. (a) Find the minimum a and b so that g (a. b. with g (355.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. Find other such examples. 113. A number may be approximated by a /b where a and b are integers. 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).

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

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

6).741. 0). 8. 5). ¾ gives Ø E5. ¿ ½ ¼¼¼ are the smallest values satisfying the above equations. In each case the numbers combine by summing the products of their digits. (7. E4. 4. Thus Ü ½½. (3. (9.654. 0. 6). Instead. E10. (9. The trick here is that 1 knot = 1 nautical mile per hour. But this is not right because 18 is not the difference of 38 and 16. and Ü ½ . 6). (5. 6). 0). 9. it appears that the rule might be subtraction. ½ ¼½ or ¼ and ½ ¼½. 0). 6). so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. One of the terms is ´Ü Üµ.425. 7. 2. E9. (1. The license number is 741. 8. 4. (3. 2). 10. (11. E7. the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. 4. At ﬁrst glance. 2). 6). .729. 10. 10. (0. 6. 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 . Ø¾ ¾¾ hours. E6. . 6. Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. 9). Anywhere within 10 miles of the north pole. E2. 0. 2. (1. Thus and Ò ¾½¼ transfers is the minimum. 0. 6. 0). ¼. The number is 381. 5.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 15 moves: (15. (3. (1. 0. 4. 2. (3. 2). so E8. (7. 6. 0). 0). the next greater number is 7. (5. E3. (11.

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

(18. (19. 0. (36. 167. 6). 10). (25. 11. 0. (25. producing pieces C and D. (18. (30. 126. (33. 0). 6). 0). 9. 6). (23. (19. 6. (28. 1). 0). Burn A from both ends. 6). (20. (29. 1.PUZZLES FROM AROUND THE WORLD 67 E16. 148. (26. (21. 0. 5. 11. start burning C from both ends and D from one end. 1). (23. 2. 1). 0). 1. 0). (25. (23. 125. 0). 1. 6. Cut the lace in half. 1. (a) 32 moves. 124. (25. 134. 0). 10). 1. (30. (21. 10. 10). 11. 10). 0). (34. 6). (30. 1). 11. 1. 11. 6. 147. 3). (25. producing pieces A and B. 3. 0). 11. 11. 0). the probability of a right triangle is ½ ¾½ or . D will ﬁnish burning. 10). (11. (37. 11. ½¿ ½¿ ½ Ê 123. 127. 6. (26. 0. (24. (22. 6. 6). (35. 0). (31. 1. The probability of an acute triangle is ¿ ¾½ or ½ . 0. (29. 1. 10). (29. (36. noting the point that burns last. 138. 0. 0. (24. 0). (37. See the ﬁgure. 5). (23. 2). (31. 168. 0). (30.75 minutes. 0). (24. 11. (22. 0. 10). 0. (35. 0. 4. 146. (29. 0). 10). 0. 0. (29. (17. 2). 128. (20. 0). light the other end of D. 0). (27. 0). . Cut B at that corresponding point. 6). and the time is 8:24. 9. (19. (29. 0. 4). 5. 0. 156. 0). This is the shortest time interval that can be measured. 158. 10). 11. 4). 11. 136. 3. 0). (24. 2. 10. (17. In 3. 11. 0). 0). 6. 4. 0). 0. 11. 10). 5). 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. (18. (34. 6). 0. 4. 1). 11. 10). 3. (18. 0. (31. (22. 4). as seen from the list below. 4). 1). 0). 0. (31. 0. 1. 11. E17. (26. 0). (31. (33. 0. 10). 10). (23. M2. 6). 11. (28. 0). (19. 0). 4. (25. (24. There are 7 ¢ 6/2 = 21 choices using vertex 1. If the hour hand is exactly on a second mark then the second hand will always be on the 12. 0). 178. When C is consumed. (25. Solutions to Medium Problems M1. 6). 3). At the same time. (35. (23. 5. 145. 0) (b) 40 moves. 3. (30. 11. 0). (24. (36. 10). 1). 5. 0. 0. (30. 6).

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

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

P. 2PS.465. 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. Each number is the length of the shorter side of the rectangle. 6S. 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. PS.789. 4S. 5S. . Shown below.123.70 R. and 3P. 2S. M14. P4S. 2P. P3S. I. 2P2S. M15. A nine-rectangle solution is shown below. P2S. 3S.

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

(15. Solutions to Hard Problems H1. 14. (15. There are 7 Good Eggs. The number in each triangle indicates its area relative to the other triangles in the ﬁgure. 11. 14. 9). 14. 7). There are 5 subjects with 15 marks possible in each. The points are ´ ¿ ½µ ¾ of the way along each diagonal for a distance of 3. 12. 10). 13. 12. 9) and (14. 8). (15. 10. 8). Humpty Dumpty got a 10 in arithmetic. 13. 6). Shown below. 12. 14. 12.72 R. 13. The scores for the Good Eggs are (15. (15.0119 . 12. 12. HESS (b) The two points are along the diagonal near P and its opposite Ô point on the ½ ¢ ½ face that includes A. 13. (15. 11. 11. . 10. 14. M20. 11. I.

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

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

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

Each has a radius ½. For Ò ½ there is just enough room for 329 circles. H10. The smallest rectangle found to date containing 329 circles has 13 circles on each side of 101 sets of 3 and measures 2 ¢ 163. The ratio È takes on a maximum when Ö ¼¾ ¼ . A view of the pyramid looking down from S perpendicular to the plane ABCD will look like the ﬁgure below. They deﬁne the four tangent lines drawn to produce ABCD as shown. Note that the ratio È is 0.9973967 . There are 7 circles on each end with 105 sets of 3 circles in the middle. The maximum ratio È Ö ¼¾ ¼ . HESS the shape is È Ô ½ ´¾ · ¾ Ö · Ö. H9. The circles need to be packed as shown below.76 R. Raise S and T appropriately above the . Now consider two such circles of different sizes as shown. The point S may be raised above the plane to give unit altitudes if the inscribing circle has radius ½.25 for the entire square or a circle inscribed in µ the square. (b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved. I.

È ´¿¼ ¼µ ÔÈ ´ ¼ ¼µ. which leads to Ô ¼ ½ ¿¿ ¼ . Let È ´ µ be the probability of the server winning the game when the server has points and the receiver has points. ¾ ¾ Õ´ È ´¿¼ ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ . which leads to Ô ¼ ½½ ¿ ¾ .8286 0. È ´¼ ¼µ È ´ ¼ ¿¼µ when Ô¿ Ô¾ ¾Ô ½. This leads to ¾ È ´ ¼ ¼µ Ô¾Ô Õ¾ . ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ . È ´¼ ¼µ È ´¿¼ ½ µ when Ô ¾ Ô ¿.2801 1.7805 0 . ·¾ Ô ´½ ½ Õ µ È ´¼ ¼µ . 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. A computer program was written to calculate the probabilities shown in the table below.8027 0.3586 Pairing probability 0.8163 0.PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit.3074 1.7892 0. H11. · È ´ ¼ ¿¼µ Õ È ´ ¼ ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . È ´ ¼ ¼µ ÔÈ ´ ¼ ¿¼µ · ÕÈ ´¿¼ ¼µ.2069 1. and let Õ ½ Ô. Let Ô the probability of the server winning a point.2270 1. Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1.7849 . Ô Õ At a score of Ò games to Ò (Ò ¼ to 5).2511 1. ¾ H12. At Ò games to Ò. È ´ ¼ ¿¼µ Ô · ÕÈ ´ ¼ ¼µ. Clearly ST is not perpendicular to the plane.

7. (c) 17. prime.321 is the ﬁrst of 5 in a row. There is only one set of ages that accomplishies this: (5. Solutions occur for (b) ¾ ½. 16).444 years. (c) Unknown. a=10. (d) Yes. (a) Length ½ ¿ ¾½ ½ ¼ as shown. H14. (a) 45 years. ´ prime. ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . ¿ ½¾¼. ¼ ¼½¼ . · ½.641. we get (a) (b) H16. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago. HESS H13. (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ . Using a multiprecision continued fraction program. The product and sum could be achieved by (4. We look for cases where (1) the oldest child is under 20.78 R. H17. I. . 12). .512.613. (b) Length ¾ ¼¾½½¾ ¿ as shown. ¾ ½µ ¾. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. ´ ¾ · ½µ ¾ prime. 6. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô . The product and sum two years ago could be achieved by (2. 8. . ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½. (b) 605 years. ¼ ¼¼ ½ ¼. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. · ½.042.546. (2) the younger two children have different ages. which must have been guessed by Jones two years ago.093. ´ ¾ ½µ ¾. 15). which must have been guessed by Smith. H15.

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

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. The only pole of Õ´Þ µ ´Ò·½µÞ within is a double pole at Þ ¼. and by Cauchy’s Theorem we have ½ ¾ Thus ÞÕ´Þ µ ´Ò·½µÞ Þ Þ ¾ Õ´Þ µ ´Ò·½µÞ Þ ¼ ¾ ´¾Ò · ¿ µ ´Òµ ¾½ ÞÕ´Þ µ ´½ · Þ µÒ·½ To evaluate ´Òµ. which still avoids enclosing the poles of Õ´Þ µ. Thus Þ ¾ Õ´Þ µ ´Ò·½µÞ is regular within .80 R. we expand and deform to a contour the ﬁgure. (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 . HESS (2) Since ´ ÒµÞ is an entire function in the complex plane. As trarily large it still encloses the same poles as . I.

Then Ý approximately equals ½·ÐÒ´ µ . The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ.0888430 2.6406814 3.8790560 20.6745053 3. where Þ at ½ · Þ Ü · Ý.1512074 45.5012690 3. 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. ¼ for all . Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to inﬁnity. A pole of Õ´Þ µ occurs Þ .PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ. ¾ × ´Ý · µ Ü Ý .2916783 3. not including its pole at Þ ¼.5432385 32. For larger deﬁne ´¾ ·½ ¾µ .8505482 39. Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ.0262969 3.2238350 26. Ü 1 2 3 4 5 6 7 (6) The contribution to 2.4614892 13. Since the equation in Ý is even.8221528 Ý 7. (5) On Þ Ê for sufﬁciently large Ê. The table below gives numerical values of the ﬁrst few.4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ Þ µÕ´Þ µ ´½ · Þ µÒ·½. Call the Ø such pole in the upper half-plane Þ Ü · Ý . which reduces to ´½ · Ü · Ý µ´Ü Ý µ ´ Ó×´Ò · ½µÝ × Ò´Ò · ½µÝ µ where ´Ü¾ · Ý¾ µ ´Ò·½µÜ . × ÒÝ Ý we can look for poles in the upper half-plane only and reﬂect each one into the lower half-plane.

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

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

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

and 8 from Figure 3 and replace them in a different way.O’Beirne’s Hexiamond Richard K. O’Beirne thought that the result would be more pleasing if the Hexagon were in the center. and several copies had to be manufactured. If we count reﬂections as different. ﬁve had reﬂexive symmetry. Maybe the only other places where it has appeared in print are Berlekamp et al. 5. but the one that grabbed us the most was the Hexiamond. 6. We became adept at ﬁnding new solutions based on the old. while seven did not. among the shapes that can be formed by adjoining six equilateral triangles. but I haven’t found one yet. He showed us many remarkable puzzles. No one went to bed for about 48 hours. In the interim he had found solutions with the Hexagon in two other of its seven possible positions (Figures 4 and 5). [1] and the not very accessible reference Guy [2]. ¿. 5. Or try using 0. Remove pieces ½. It soon became necessary to devise a classiﬁcation scheme. and then by twelve more. Martin [3] uses O’Beirne’s names for the shapes but does not distinguish between reﬂections so his problems only involve 12 shapes. and ¿. particularly on this side of the Atlantic. . But it does appear in his column in the New Scientist. 6. I was sure that I would ﬁnd references to the Hexiamond in Martin Gardner’s column. 85 . 3. then there are 19 shapes (Figure 1). which can be surrounded by six more hexagons. and in Figure 7. It must have been in 1959 that O’Beirne noticed that. and in Figure 5. ½. since it was not easy to decide whether a solution was new or not. 1. When I began to write this article. The next solution in my collection is Figure 7 by Mike Guy (March 1960). Nor is it mentioned in O’Beirne’s own book [4]. since everyone wanted to try it at once. giving a ﬁgure (Figure 2) having the same total area as the 19 shapes. It was soon after this that O’Beirne visited the Guy family in London. One of these is the regular hexagon. Guy Tom O’Beirne is not as celebrated a puzzler as he deserves to be. 3. Rearrange pieces 1. and . 7. and in January 1960 he found the solution shown in Figure 6. 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 subtracting each of the labels 11 to 19 from 20 we arrive at the labeling in Figure 1. We don’t know how O’Beirne decided whether to give a piece a number less than or greater than 10. and each is positioned by its apex. sailing to the left. Crown. GUY Figure 1. ½. and the other even numbers have reﬂective symmetries. . a parallelogram drawn in the opposite way from the usual textbook fashion. together with those that have rotational symmetry only. . to the Snake. The Chevron. The nineteen Hexiamond pieces. as do pieces 1.86 R. to the Crook. the multiples of 4 have rotational symmetry through 180 degrees. 7. to the Signpost. 5. . Pieces 4 and 8 have only rotational symmetry and exist in enantiomorphous pairs. . 3. The symmetries of the Hexagon are those of the dihedral group D6. Pieces with odd-numbered labels have no symmetry and. The Butterﬂy has the same symmetries as a rectangle: Its position is described by that of its body. . to the Yacht. By subtracting 10 from the Hexagon. whose head is on the left. Negative (bar sinister?) labels are given to the Bar. O’Beirne had already suggested a numbering of the pieces from 1 to 19. and 9. and the Sphinx. with its hook on the left. to the Hook. which appears to be turning to the left. which have no symmetry. exist in enantiomorphous pairs. but our mnemonics are as follows. with its pointer on the left. . Notice that 0 has the greatest symmetry. with its handle on the left. K. and Lobster each have a single symmetry of reﬂection.

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

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

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

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

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

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

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

Their numbers.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. GUY Figure 18. are 130 195 533 193 377 2214 111.489 B 15.94 R. Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to ﬁnd all the solutions. by the Hexagon’s position Ø Ù Total 264 1094 124. for those who prefer their puzzles to have just one answer.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and.518 A 75.717 C 6675 D 7549 E 11. I wrote to Kate Jones. Six pieces meet at a point. classiﬁed according to the position of the Chevron. 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. Figure 19. president of Kadon Enterprises and discovered that she markets a set of “iamonds” . K. Ô Ü (3-symmetry). On the contrary.

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

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

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

98

S. TAKAGI

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

A square with a center hole

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

Some patterns formed from the Sei Shonagon pieces

½1988, W.H. Freeman and Co., NY

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

Bernhard Wiezorke

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

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

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

99

100

B. WIEZORKE

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

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

TANGRAM CAT HAPPILY TURNS INTO PINK PANTHER

101

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

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

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

102

B. WIEZORKE

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

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

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

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

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

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

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

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

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

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

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

Gallop: On a ¢ ½¾ grid. 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). double checkerboard hexomino shapes deﬁned by pawns are transformed with knight’s moves (borrowed from the Leap set). Another puzzle is to change a simple hexomino made of six pawns into as many different other hexominoes as possible by moving the pawns with chinese-checker jumps (Figure 12). Figure 12. Figure 11.THOSE PERIPATETIC PENTOMINOES 113 Figure 10. .

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

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

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

**Self-Designing Tetraﬂexagons
**

Robert E. Neale

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

Figure 1. Square window.

Figure 2. Cross plus-slit.

117

118

R. E. NEALE

Figure 3. Square plus-slit.

Figure 4. Square slash-slit.

Figure 5. Square cross-slit.

Square Window

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

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

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

SELF-DESIGNING TETRAFLEXAGONS

119

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

Figure 6.

Figure 7.

Figure 8.

Figure 9.

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

120

R. E. NEALE

Figure 10.

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

Figure 11.

Figure 12.

SELF-DESIGNING TETRAFLEXAGONS

121

Figure 13.

Figure 14.

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

Figure 15.

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

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

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

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

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

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

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

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

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

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

METAGROBOLIZERS OF WIRE 133 The object of the Bermuda Triangle is to save the UFO that is trapped in the puzzle. as I sat playing with it. I keep tools and wire handy for just such events having learned that three-dimensional ideas are difﬁcult to decipher and duplicate from two-dimensional scribbling on a piece of paper. The solutions to many puzzles can be elusive until the puzzle has been manipulated many times. this one is relatively easy to remember once the solution has been seen. As with most of my puzzle ideas. although moderately difﬁcult for the average puzzler to solve initially. taking the UFO with it as it moves. 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 UFO being a ring with an abstract shape mated to it. The Nightmare . The Bermuda Triangle rates about a medium level of difﬁculty. The triangle can be moved over the entirety of the larger conﬁguration. which is a triangular piece. 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. this one came to me fully formed and complete. 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. my wife joined me in critiquing my latest design. We didn’t have any nightmares about it. After making the prototype. The puzzle is generally set up with the ring around the Bermuda Triangle.

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

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

I immediately ordered some. Tim Rowett had brought one from England. Gardner knew how it must work and wrote about it in his August and September columns without even having . My September 1992 column introduced the Floating Hourglass 25 years after I ﬁrst heard about it. but his description was clear enough. I found a thick folder on the Hourglass. Hein told him about a toy he had seen in the Paris airport. He didn’t bring one back. a treasure trove of letters and drawings. I saw my second Floating Hourglass and knew I could ﬁnally write about it. made by Ray Bathke of London. but no magazine article could possibly contain it. since I only published puzzles in Omni that I knew my readers could ﬁnd. I asked my clever readers to submit theories to explain it. When Martin allowed me to look through his ﬁles. the Danish sculptor/inventor (the Soma Cube. the Super Egg) and poet/artist (Grooks). The results appeared in the January 1993 issue. MORRIS At the 1991 Puzzle Collectors Party in Los Angeles. This book ﬁnally gives me the chance to tell the history of the Floating Hourglass Puzzle. had visited Martin in early 1966.136 S. The First Theories Piet Hein. For years I have itched to tell this story.

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

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

” The Hourglass Letters While Hein was having his epiphany in Europe. also from the U. Gardner was beginning to wonder if his friction explanation told the whole story. In short. and it didn’t last long. or perhaps still another one. having relied (unfortunately) on an account given to me by a friend who examined the toy in Paris. It is possible that one version of the toy works on the principle you mention. Well. Reid later adapted his letter for publication and his short article “Weight of an Hourglass” appeared in American .” 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. and to suggest that you not publish a correction. 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 . 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. “I am writing to put your mind at ease (on the impact theory posed by Altman). Albert Altman wrote from the U.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. at this point I am hopelessly confused. Naval Ordnance Laboratory at Silver Spring. Reid. The principle is simply too clever to be true. One man wrote that he owned one of the Paris cylinders. You are the inventor.S. perhaps depending on the temperature. I have not yet seen the toy. Naval Ordnance Lab.S. A letter dated September 29 came from Walter P. Gardner was getting letters in response to the September issue. Could temperature and sand impact also be factors? Would he have to print a correction? He replied to Altman on September 12th: I am embarrassed to admit that your explanation may be right. and the height through which the sand has fallen. On September 6th. where is the rate of the ﬂow of the sand. and the other on the principle I suggested. the acceleration of gravity. Hopelessly confused? Martin Gardner?! That is something that surely doesn’t happen very often. I am sure that your explanation was correct. but did not bring one back with him. but reported that his glass sometimes ﬂoated and sometimes sank. all you have to do is shake your own hand.

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

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

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

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

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

each of the 48 faces of the pieces is colored with a unique color. Trial and error is not a very fruitful way of trying to solve this puzzle. Therefore. FARRELL Figure 1. each . after reading a hint. If. additionally. These hints are to be regarded as progressive. A. C. ½º Call the six colors A. 0). That is. The combinatorial puzzle uses only six colors in its construction. 1.146 J. Solution Hints. and 3 are any six distinctive colors. Notice that the eight pieces must form the eight corners of the completed cube — one piece for each corner. but there still remain a great many cases that are nearly right. If you cannot. C. proceed to the next hint.½ distinct ways of constructing the cube. 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. 0. B. and 3. There are 93 solutions in this case. B. and the solver has the additional clue that a face is all of one color. 2. 1.½ ¼. A. and C will turn out to be a solution set. 2. then there are ´ ¿µ´ µ´¿ µ¾ .½½ . B. try again to solve the puzzle.

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

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

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

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

in which the number of isomorphisms using the set of thirty-six dicubes is also determined.691 combinations into 148 disjoint sets is described in Section 6. the results are shown in Table 1. The question then arises: How many solutions are there? Most intriguing about this puzzle. The second observation is that for a cube to be assembled with the nine colors on each face.251. is that there are numerous solutions. when two cubes of the same color are in position. The ﬁrst observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube. The third observation. and the most difﬁcult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1. and only 10. this rule then tells the experimenter to backtrack.691 combinations need to be sorted into disjoint sets. The ﬁnal part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. A proof is given in Section 2. none of the nine planes may contain two cubes of the same color. it becomes apparent that most of the 133. Since one person can assign the fourth and ﬁfth colors to orange and red and another person to red and orange. A proof is given in Section 3. . only thirty dicubes are required. only 133. as is shown in Section 5. to separate these cases into disjoint sets of isomorphisms. then. This was achieved through the use of a computer program. Thus. It becomes necessary.677.700 (36!/12!24!) different ways of which. as determined in Section 4. Twelve of these can be selected in 1. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. Fortunately. The individual dicubes can be colored thirty-six different ways. FARHI a difﬁcult exercise for a moderately proﬁcient programmer to verify the results obtained. The previous question now becomes more interesting. obtained after some experimentation. some with the tricube on the edge and some with the tricube through the center of the cube.152 S.105 of these combinations meet the three-cubes-of-each-color requirement.105 combinations are isomorphisms. but not in the center of a face. the location of the third cube is determined. Since during the course of trying to solve this puzzle a conﬂict often occurs. The method of sorting the 10.

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

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 . FARHI No.154 S.

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.

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

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

the remaining two cubes must be located on an edge. and 3. cubes of the same color are never in the same plane. there are three possible locations. then none are selected from the third row. and 11. Determination of the 133. If none of the previous selections include color 3.” In each case. 2. and 14. 3. such as shown by the letter “C. 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. If color 3 has been selected once. then color 3 is combined with two colors from the third row. One cube is centrally located and thus not visible. then a third dicube with color 2 and any color from the second row is selected. The “No Two in the Same Plane” Rule Figure 3. FARHI If an irregular tricube is used. The two dicubes with color 1 may use any two colors in the ﬁrst row of the following table. 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.105 Possible Dicube Color Combinations The tricube colors are deﬁned as colors 1. accounting for two faces. . and on a centerface such as shown by the letter “B. 4. Since the corner colors are displayed on three faces. and 11. the center case ﬁlls spaces 5.” The remaining two colors must all be on edges. If color 2 is selected from the ﬁrst row. the edge case ﬁlls spaces 2. then one color is selected from the third row and if color 3 has been selected twice. 5. 11.158 S.” The remaining six corners must have cubes of different colors. The corner case ﬁlls spaces 1. 2.

6). Identiﬁcation and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. 6). color 5 is combined with colors in the ﬁfth row such that there are three cubes with color 5. 8). (3. (5. 9). (4. 6. 2). the two new colors are identiﬁed as colors 6 and 7. For example. 4). or 5. This results in requiring only thirty cubes and. 7).THE NINE COLOR PUZZLE 159 First Color 1 2 3 4 5 6 7 8 Second Selection 2 3 4 5 6 7 8 3 4 5 6 7 8 4 5 6 7 8 5 6 7 8 6 7 8 7 8 8 9 9 9 9 9 9 9 9 The process continues: color 4 is combined with colors in the fourth row such that there are three cubes with color 4. 9. 5). 8). 9. 3. 6. 9). (4. using the same algorithm as before. 5. 7. a computer program established 133. 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 . Using this algorithm. (6. 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 5.691 color combinations. 6. (5. Similarly if color 2 is not associated with colors 3.105 possible color combinations. (7. 4. 4. 8. 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. etc. (2. given the second numbers 2. 3). only 10. 9) and (8. (7. (1. 5. 8.

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

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

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

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

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

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

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

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

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

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

¢ ¢ . See [1] or [3] for more information. Holes. Simpliﬁed (15) (Cutler. ½½ ¢ ½¾ ¢ ¾. ¾ ¢ ¢ ½¾¿. ½¾ ¢ ½ ¢ ½ . ¢ ¢ ½½. The extra space makes available a whole new realm of possibilities. basic. Holes. 1981. The net result was a puzzle that is extremely difﬁcult. ½¼ ¢ ½ ¢ ¿½. version by cutting some of the pieces into two or three smaller pieces. ½ ¢ ½ ¢ ¾½. none having any symmetry or pattern. ½½ ¢ ½½ ¢ ¾ . ¾ ¢ ¿ ¢ . but it is not. ½¼ ¢ ½½ ¢ ¾. ¢ . Bill Cutler Puzzles) Box: ¼ ¢ ¾ ¢ ¾ Pieces: ¢ ½ ¢ ¾ . ¢ ¾¼ ¢ ¾¼. ¢ ¢ The dimensions of the pieces are all different numbers. ½¼ ¢ ½ ¢ ¾ . ½½ ¢ ½ ¢ ¾ . The pieces ﬁt into the box with no extra space. ¾ ¢ ¿¼ ¢ ¢ ½¿ . ½½ ¢ ½ ¢ ½ . The smallest number for which this can be done is 23. Almost 15 years later. See [4]. Pieces the Same or Similar º Hoffman’s Blocks (27) (Dean Hoffman. CUTLER ¾¾ ¢ ½¼ ¢ ½¿½. ¿¿ ¢ ¿ ¢ ¿ ¢ ½¾½. 1976) Box: ½ ¢ ½ ¢ ½ Pieces: (27) ¢ ¢ This sounds like a simple puzzle. The box is a cube with side equal to the sum. There are many other 23-piece solutions that are combinatorially different from the above design. ¿ ¢ ½½¼ ¢ ½¿ . 1986. ¢ ½¼ ¢ ½½. They can be any three different numbers. ½½ ¢ ½ ¢ ¿¼. There are 21 solutions. where the smallest is greater than one-quarter of the sum. this puzzle still fascinates me. ¢ ½¼ ¢ ½½ 11. ½¼ ¢ ½¾ ¢ ¾ . I will not say anything more about this design except that the trick involved is different from any of those used by the other designs in this paper. 10. ½¼ ¢ ½ ¢ ¾ . ¿ ¢ ¾ ¢ ½¾ . º Hoffman Junior (8) (NOB Yoshigahara. ¿½ ¢ ¢ . Pieces Different ½¼º Cutler’s Dilemma. ¢ ¼. Hikimi Puzzland) Box: ½ ¢ ½ ¢ ½ Pieces: Two each of ¢ ¢ ½¼.172 B. The dimensions of the pieces can be modiﬁed. . ½ ¢ ½ ¢ ¾½ The original design of Cutler’s Dilemma had 23 pieces and was constructed from the above. I like the dimensions above because it tempts the solver to stack the pieces three deep in the middle dimension. ¾¿ ¢ ½ ¢ ¿. ¿ ¢ ¾ ¢ ¼.

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

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.

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

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

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

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

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

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

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

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

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

Chinese winepots Royale’s patent Luminations “Columbus Egg” Jugs and liquids.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. DALGETY AND E. “Jail Nixon” Strip polyhedra Standard “block holes and suck” solution “Jolly Jugs” Cadogan teapots. or those that can only be represented on a computer . crosswords on bathroom tissue OTH-MATH OTH-LOGI OTH-TRIK OTH-MAGI OTH-MYST OTH-SET OTH-THEO Mathematical puzzles (excluding number pattern arrangements) Logic puzzles Trick or catch puzzles (solution needs subterfuge) Conjuring tricks presented as puzzles Objects whose function or material is a mystery Sets of puzzles of mixed type Puzzles whose existence is only theoretically possible Cartoon pictures to arrange in order “Infernal Bottle” Disappearing coin slide box Creteco spacers such as 4-D puzzles. rebus plates and prints. HORDERN Map folding. Archimedes gold Cork for three holes. 12 Golf Balls. ﬁve square puzzle 19th century riddle prints Anagrams.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

204 B. GOSPER Times 100–105: .

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

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

Buckingham mercifully outmoded it with a comparatively tiny. When the crystal reaches the guns. it uses various mechanisms to interleave larger period streams. New Glider Guns. The period ½¿ · Ò guns based on Buckingham’s B heptomino turns are also smaller. Subsequent gliders grow a crystal upstream. its bounding box is 13. 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. the cycle repeats every ¢ ¾ generations. two gliders delete one pair of beehives. The period 94 gun is 143 by 607. forming gliders that stabilize the ends of a row. based on a period 72 device discovered by Bob Wainwright. Using normal kickbacks. If you delete the trafﬁc light. I’ll just add that some of Buckingham’s syntheses of still lifes and billiard table oscillators are aweinspiring. 11 gliders add a pair of beehives to the crystal. an eater stops its growth. (Sadly. there are other ways to get any multiple of 30. two such rows. centinal-based ¼¼ · ¾¼¼Ò. the resulting period must be a multiple of 8. of size 149 by 149. the process begins again. Fortunately. they don’t work with period 30 streams. In the period 94 gun. Rich has already mentioned this. based on the period 100 centinal.) It’s also possible to build a gun that produces a glider stream of any period ½ . emit gliders that can crash to form MWSSs. there’s now a fairly small period 44 gun that Buckingham built recently.584 by 12.112.STRANGE NEW LIFE FORMS: UPDATE 207 180Æ and collides with the second to form a honey farm. The other guns are all pretty big. so we can build a long row of them that is unstable at both ends. you (JHC) described how to thin a glider stream by kicking a glider back and forth between two streams. A honey farm starts to form but is modiﬁed by an eater and a block. forms a trafﬁc light. which hit eaters. In addition to the long-familiar period 30 and period 46 guns. and it begins to decay. and then starts forming another honey farm in a different location. Here’s how to turn 2 MWSSs into a glider: In Winning Ways. When they’re all gone. with a total of 36 AK47s. I’ve found some eaterassisted kickbacks that give any even period. Buckingham has also found extremely weird and elegant guns of periods 144 and 216. For example. It’s fairly easy to get any period down . The largest by far that’s actually been built is Bill’s period 1100 gun (extensible to ¼¼ · ¾¼¼Ò). it’s based on the “AK47” reaction discovered independently by Dave Buckingham and Rich Schroeppel. It emits a glider. The gun itself has a larger period. A close pair of AK47s can delete each other’s trafﬁc lights.

GOSPER to 18 this way (period 23 is especially simple). (At least they seem unusual. This will make it possible to produce glider streams of any period. 16.” (Less than 14 is physically impossible. if the glider is delayed by 60 (or. Ò · ) generations. 10. (I built a pseudo-period 15 gun based on his reactions. and 17. this reaction leads to a period 30 LWSS gun: Large Constructions. 150. I have a working construction to build a P14 glider stream by inserting a glider between two gliders 28 gens apart! . and 360. middle-. . 120. more generally. . and Buckingham has found clever ways to get 15. 300. and heavyweight spaceships from gliders. it escapes: There are also combinations of two period 30 guns that give fairly small guns with periods 90. For example.) New Year’s Eve Newsﬂash: Buckingham has just announced “. Here it’s shown deleting a glider. so we have guns that produce these as well. based on the . The smallest known uses a period 8 oscillator (“blocker”) found by Wainwright to delete half the gliders from a period 60 stream.) Returning to Hickerson’s account: Many of the large constructions mentioned later require a period 120 gun. 210.208 B. We can also synthesize light-. . . Universality implies the existence of Life patterns with various unusual properties. its bounding box is 373 by 372.

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

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

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

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

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

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

A viable hollow maze that cannot be constructed by the “cut and connect” method. ¯ Recursion. Regular control mazes. Replace each point by a hollow maze (see Figure 11). I keep receiving letters from readers of Scientiﬁc American. Construction of a hollow maze by recursion. There is still much research to be done on multiple silhouettes and hollow mazes. V AN DEVENTER Figure 9.218 M. Figure 10. O. . and perhaps some of the problems mentioned above will be solved in the future. Figure 11.

D.. and the material goes back several centuries). complicated versions like “How Old Is Ann?” appeared. but there is an additional diophantine requirement that the data and/or the answers should be integral. Ü · Ü or Ü or Ü . but the solution requires some ingenuity. The second is the Ass and Mule Problem. The ﬁrst is a simple age 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.e. compiled around A. 500 (although the date of this is uncertain.Some Diophantine Recreations David Singmaster There are a number of recreational problems that lead to equations to be solved. Other times the equations are straightforward. Sometimes the difﬁculty is in setting up the equations. the problem of Diophantus’ age gives us Ü · Ü Ü Ü · · · · ½¾ ¾ Ü i. leading to one equation in one unknown. I discuss three examples of such diophantine recreations. where the problem was in understanding the phrasing of the question. In the late nineteenth century. The simplest are just problems of the ‘aha’ or ‘heap’ type. with the solution then being easy. A Simple Age Problem Age problems have been popular since at least The Greek Anthology [16]. which contained statements such as “Mary is twice as old as Ann was when Mary was as old as Ann is now. The third is the problem of Selling Different Amounts at the Same Price Yielding the Same. such as exploiting the symmetry of the problem.” (See Loyd [14]) 219 . In a third class of problems. for which I have just found a reasonably simple condition for integer data to produce an integer solution. the equations and their solution are relatively straightforward. In this case. for which I give an algorithm for ﬁnding all solutions and a new simple formula forthe number of solutions. In The Greek Anthology. See Tropfke [20] for this and related problems.

at the time he was married. after years. what were their ages on their wedding day? We can state the general form of this problem as follows. The solution of these is easily ´¿µ ´ ½µ ´ ½µ assuming . more comprehensible. . version. Her age to mine exactly bears. is times as Thus the problem in The American Tutor’s Assistant and Bonnycastle is Problem (3. 15.220 D. . when are and integers? For example. he was only twice as old. SINGMASTER Between these are a number of types of problem. we can give a pretty simple answer: is integral if and only if ½I haven’t actually seen the book. Since is integral if is. A person. 1824 [5]. The type that I want to examine is illustrated by the following. which is the earliest I have found. reportedly from The American Tutor’s Assistant. My age was to that of my bride As three times three to three But now when ten and half ten years. and are integers. 2). 3. I pray. 2) does not have an integral solution. Problem ´ old as . Now we give a diophantine aspect to our problem by asking the question: If . Now tell. but with the same numbers. et al. As eight is to sixteen. µ. It is easily seen that the problem leads to the equations ´¾µ where and found to be denote the ages of · and ´ · µ .½ When ﬁrst the marriage knot was ty’d Between my wife and me. is now times as old as . but after they had lived together 15 years. Problem (4. What were our ages when we wed? A more typical. 1791. appears at about the same time in Bonnycastle. albeit less poetic. [6]. was 3 times as old as his wife. We man and wife have been. the problem is quoted in Bunt. from what I’ve said.

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

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

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

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

2. 6) I-(7. 1. 3. 11. to sell. And he took to the eldest daughter ﬁfty apples. 5) (2. 5. 6) (1. and to the youngest ten apples. 1) (30. done two ways Selling Different Amounts at the Same Price Yielding the Same Although this problem is quite old. 2. 11. 5. 9. and all these three sold in like many for a penny. 6. 5. 7) I-(7. 8. 10) (7. 9. 2. so I will ﬁrst describe the later form which is ﬁrst found in Fibonacci [11]. The problem is treated in one form by Mahavira [15]. with the problem statement giving the 6 as a 7 Fibonacci Fibonacci. 8. 3. 4. 3. 0. A man had three daughters of three ages. 3. 2. 1. 4. 9. but this form is less clearly expressed and has inﬁnitely many solutions. 5. which is rather a mouthful. Now how many sold each of them for a penny? . 7) II-(1. 2. 11. 7) I-(7. 3. 510 9Ø century c. 10. hence the above title. 9.” printed by Wynken de Worde in 1511 [9]. 10. 50 in the ﬁrst printed English riddle collection “The Demaundes Joyous. 11. 50. certain apples. 300 B. 23. 3. 10.C. and brought home in like such money. 4. 11. 5. c. 2) I-(9. which daughters he delivered. 4. 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. 3) (10.5) I-(25. 2. 1. 3. 6) I-(7. 10. although the material is believed to be based on older Arabic sources. 4.SOME DIOPHANTINE RECREATIONS 225 from the early fourteenth century and hence after Fibonacci. and to the second thirty apples. 5) c. 2. 850 c. 4. with the problem statement giving the 6 as a 7 Fibonacci. 3. Sridhara [18]. (1. 9. 2. and Bhaskara [4]. 510 c. 5. 11. I quote a version that appears as No. 850 c. there seems to be no common name for it. 250 c. 22. 5) (100. 1. 5. noting that this system is inconsistent Fibonacci. 4) (2. 6) I-(7. 2.

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

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

Solutions in the literature. We also get ´«¾ · ¾ µ . we have ´ µ« ¾ ¾ . Let ´«¾ ¾ µ. Now use equation (1) to eliminate from (2). ´«¾ · ¾ µ «¾ ½ · ¾ «¾. divides . as solutions that differ only in price scale are really the same. and they are considered to give the same sets of solutions. Glaisher considers ¿.. are often given with fractional values. and we note that ½ «¾ ½ . From just before (6). Hence ¾ .. and I claim that any such «¾ generates an integer solution of our problem. these prices are considered the same as ¾½ ½. In the present approach. and (4) gives us ½ µ ´ divides . giving us ´¿µ ´ µ ´ ´ µ Subtracting each of these from the case µ´ ½ µ ´ ½ µ Since ´ µ ½ implies ´ µ ½. we must have ¾ dividing «¾. This is quite reasonable. 30. this implies that ¾ divides . In some versions of the problem. i. . Since ´«¾ ¾ µ ½. where ´ µ ´«¾ ¾ µ ½. in particular ½ ¬ for some integer ¬ — e.e.g. From (6). so that µ ´ µ « is a multiple of Now we will take « ¾ as the ﬁrst basic parameter of our solution. This is also equivalent to scaling the prices. which restricts the solutions to a subset of those for the case ¾½ ½. we have «¾ ¾ . SINGMASTER solution set to two. This tells us that . so «¾ ¾ . «¾ . so the previous equation becomes ´ µ« ¾ ¾ . and the solution set is again twodimensional. unless one imposes some further condition such as being the reciprocal of an integer. the most common solution of the (10.228 D. 50) problem has ¿ ½ — and the usual solutions have each person selling as many groups of ¬ as possible. From the case ¾ of (4). the yield is speciﬁed. 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 . ½ with the condition that be integral. which is the condition (5) for all the equations (4) to have integral solutions. however.

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

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

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

119. 1 given. 49) (3. 60) (10. 81) (10. none given. 1 given. 35) (11. 1) (7. 30) (18. 68. 20.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 1 given.552. 30) (12. 50) (11. 40. 50) (10. 32) (12. 1 given. 140) . 6 given. 36) (27. though it is not as easy to see how to do this as in the Western version. SINGMASTER Solutions in Source 5 given. 33) (20. 1 given. 4 given. so we again have a three-dimensional solution set. 1 given. 20) (64.232 Version (10. Sridhara ﬁnds a one-parameter solution set. 32. 60) (117. 100) (45. 1 given. 196) (45. . 40. (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. Bhaskara makes the following comment: “This. in some way to reduce the solution set to two dimensions. 1 given. 1 given. 1300 1300s 1489 1489 1489 c. 2 given. one with fractions. 170) (305. At this point. 40. 1 given. 30) (8. 901) (10. 55) (78. 11. 36) (78. 81) (55. ¬ . 1 given with fractions. 56. none given. 55) (16. 36) (30. 37) (27. 603. 50) (30. . makes this more speciﬁc by saying they ‘sold a part in lots and disposed of the remainder at one for ﬁve panas. 30. 12. 136) (3. 33) (10. 1) (100. 30. 90) Date 1202? 1202? 1202? c. 26) (20. . 454. 16) (16. (20. I must point out that some Indian versions give and/or as fractions! The Indian authors do not get very general solutions of these problems. 1 given. 40) (18. 33. 36) (25.4) (70. 17. 16) D. 1 given. 1 given.400) (25. 1 given. 20) (10. 60) . 48.6) # of Solutions (55. 82) (17. 80) (10. One can scale the set . 20. 40) (10. 29. 25. 30. 21) (225. all given.9) (100. 15) (31. 1 given. 752. 55) (55.

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

Neun Bucher arithmetischer Technik.900 and of 3294. Wynken de Worde.. aka Bhâskara. 2nd ed. Vogel. Labosne. trans. 15. London. 178–186.470 [4]. and Other Places of Public Education. 1624. which may be the same as the 1815 edition. designed for the Use of Schools. Gauthier-Villars. [8] Diophantos. 76. Lucas N..” J. 24. 6. 1817. 1813. Rigaud. 1150. pp. 86–88. 3rd ed. In: T. he produced a 10 Ø ed. Mar 1992. Wiesbaden. 1511. there are only a ﬁnite number of solutions. See No.so and ¬ ½½¼. reprinted by Dover.. Problems to Sharpen the Young. v. Facsimile with transcription and commentary by John Wardroper.(475). [4] Bhaskara. 5th ed. 1968. References [1] Alcuin. [5] Bonnycastle. 3200 and 16. 1879. 170. John. et al. [6] Bunt. Gordon Fraser Gallery.” which “may be considered as a concise abridgment” of his two-volume Treatise on Algebra. Indian Math.. 1964. Arithmetica. Math.” c. Algebra.. 4th ed. p. 15. The Historical Roots of Elementary Mathematics. (There have been several reprints.. 33. Oxford University Press. 21. Heath. “an entire revision of the work. Bhaskara solves his example by taking ¾. The ﬁrst nine editions appeared “without any material alterations. reprinted by Blanchard.” In 1815. 2nd ed. [9] The Demaundes Joyous. 24. 1824. 122–126.. Analysis similar to that done before shows that there are then ½ · ´ ¬ ½ Ò µ ¬ solutions. Rigaud. [7] Chiu Chang Suan Ching. 106–115. . 10. 1874. Lyon. 5th ed. Bijanganita. 3292. nos. Nos. In: Henry Thomas Colebrooke. J. “A classical Indian puzzle-problem. pp. Gaz. and the 13Ø edition. pp. 214–223. By the way. [3] Bachet. Vieweg.C. pp.) Chap. London. Paris. 19. P. 150 B. 54. 1976. London. J. H. Claude-Gaspar. P. 1959. pp. Revised by A. John Murray. 1973. pp. Translated by John Hadley. PrenticeHall. I examined the 7 Ø edition. pp. Paris. 1612. 1923–24. with Notes and Observations. Diophantos of Alexandria. A. Krishnaswami. pp. 13. Johnson.. 4392. 102– 126. 1910. SINGMASTER For given and ¬ satisfying ´ ¼¼¼ µ and ´ ¼µ.” [2] Ayyangar. London. An Introduction to Algebra. 242–244. 1884. 1st ed.. giving total numbers of 3295. Translated into German by K.. Book I. Lyon. Bîjaganita. 134–136. annotated by David Singmaster and John Hadley. Prob. 1805. L. with Arithmetic and Mensuration from the Sanscrit of Brahmegupta and Bhascara. so . Nunn et al. Prob. 1782. Problemes plaisants & delectables qui se font par les nombres. including Sandig. 12.234 D. Soc. 18.. Braunschweig. Prob. “Nine Chapters on the Mathematical Art. London. A. 16: “De duobus hominibus boves ducentibus — Two men leading oxen..

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

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

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

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

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

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

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

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

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

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

Æ ½ from the other mirrors and one from the entering ray. at the other Æ ¾ mirrors (other than the E and P mirrors) there can’t be Æ ½ reﬂections. or ´½ µ Ê½ Æ¾ Æ · ½ However. At mirror E there can be one additional reﬂection for the entering ray. Number of reﬂections for different mirror types. for even Æ there is a parity complication. . 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. going to every other mirror. O. At mirror E there can be at most Æ reﬂections. Mirrors on the Corners of a Regular Polygon Can the theoretical maxima Ê ½ and Ê¾ be achieved for any Æ . as shown for the regular pentagon in Figure 3. and what should the positions and the orientations of the mirrors be? A well-known geometrical property may help us: When the Æ mirrors are put on the corners of a regular Æ -gon. At all other mirrors there must be an even number of reﬂections. At mirror P there can be at most Æ ½ reﬂections. so remains valid. Therefore. However. At each mirror a maximum of Æ ½ rays can be reﬂected. V AN DEVENTER Figure 2. which is an odd number. Only at mirror P can there be an odd number of reﬂections. since the light paths are used in two directions. the angle between all the neighboring pairs of light paths is 180 Æ Æ . using a perpendicularly reﬂecting mirror P yields the maximum number of reﬂections. see Figure 2.246 M. but only Æ ¾ to make it an even number. the theoretical maximum number of reﬂections is Æ ´Æ ½µ · ½. from the other mirrors.

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

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

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

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

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

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

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

These mathematical models predict that the expected size of a “random” Ò after steps should be about ´ ¿ µ ¾Ò. [24]. Cloney. Even though ¿Ü · ½ function iterates possess quite a bit of structure (cf. 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. Since a number Ò is multiplied by ½ if it is even and approximately ¿ if it is odd. 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. [13]. which is ´ ¿ µ½ ¾ . which may be enough for a pseudorandom bit to be extracted. and a branching random walk model for backward iterates of Ì . see [9]. On the positive side. Since this is less than ½. which concern extreme behaviors of ¿Ü · ½ iterates. see [20]. The repeated random walk model results of [15] show almost perfect agreement between empirical data for ¿Ü · ½ function iterates for Ò up to ½¼ ½½. From this viewpoint it becomes interesting to determine how well the properties of ¿Ü · ½ iterates can be described by a stochastic model. [20]. [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 . Here we survey some recent results on stochastic models. [24]. 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. [15]. There is excellent numerical agreement of this model’s prediction with ¿Ü · ½ function data.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. In doing so we are in the atypical situation of modeling a purely deterministic process with a probabilistic model. one expects the iterates to decrease in size and eventually become periodic. This model has the drawback that . obtained in joint work with Alan Weiss and David Applegate. for an ¾ explicit constant ½. et al. cf. Several heuristic probabilistic models describing ¿Ü ·½ iterates are based on this idea. Lagarias [14] and Luby [17]. The simplest of them assumes that this independent coin-ﬂip behavior persists until ½ is reached under iteration. Garner [11] and Korec [12]) they also seem to possess some residual structurelessness. Lagarias and Weiss [15] studied two different stochastic models for the behavior of ¿Ü · ½ function iterates. A well-developed theory of pseudorandom number generators shows that even a single bit of pseudorandomness can be inﬂated into an arbitrarily efﬁcient pseudorandom number generator. one expects that ¾ ¾ on average it changes multiplicatively by their geometric mean.

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

For each initial value Ò an entirely new random walk is used. this explains the name “repeated £ random walk model. Large deviation theory also asserts that the graph of the logarithmically scaled trajectories ´½ º µ approaches a characteristic shape.24). with probability 1. so that Ø£ ´Òµ ÐÓ Ò. The random variable ½ ´Òµ is the step number £ with probability ¾ at which 0 is crossed. For the extremal trajectories for Ø£ ´Òµ in (14. Large deviation theory shows that. 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. £ and the expected size of ½ ´Òµ is given by (13. see Figure 1. For the extremal trajectories for £ ´Òµ in (14.4). the limiting graph of these trajectories has a different appearance." 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. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally. This random variable is ﬁnite with probability one. 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. 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.

and the graph (14. maximizing ´Òµ in ½¼ Ò ½¼ ·½ (dotted ¾½ ÐÓ Ò. solid for ½¼). and the second with slope ¼ ½ ¿ for ÐÓ Ò ¾½ ÐÓ Ò. see Figure 2. Figure 2.5) approximates the union of two line segments. the ﬁrst with slope ¼ ½¿½ for ½ ÐÓ Ò. solid for Ò ½¼). starting at ´ ¾µ and ending at ´¾½ ¼µ. and ending at ´ ¾µ. Scaled trajectories of Ò maximizing ´ÐÓ ½¼ ·½ (dotted for ½ . Ø´Òµµ ÐÓ Ò in ½¼ Ò . LAGARIAS Figure 1. C. starting from ´¼ ½µ. Scaled trajectories of for ½ .258 J.

6091 Table 2.91 19.5818 0. He found that Ò 12.400.68 24.6026 0.728.5741 0.77 31. Vyssotsky [23] has found much larger numbers with high values of ´Òµ.94 31.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6.371.4). More recently V.511 7 63. where ´Òµ ÐÓ Ò .91 32.254 9 4. .527 has ½ ´Òµ ½¾ ½ and ´Òµ ¿ ¾ ½ . and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory.180. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale.171 4 52.456.769.5927 0.38 41. A comparison of (14.68 Ö´Òµ 0. Section 5] shows that in ½¼ ½½ trials one should only expect to ﬁnd a largest ´Òµ ¿¼.527 5 837. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good. Maximal value of ´Òµ in intervals ½¼ Ò ½¼ ·½.24 16. 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 (14.48 18. 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. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ´Òµ ¿ ¾ .6030 0. an analysis in [15. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1. ½ ´Òµ It gives the longest trajectory.13 26.328.6020 0.266.815 10 13. The agreement with the stochastic model is quite good. The large deviations extremal trajectory is indicated by a dotted line.64 32.5967 0.1) with numerical data up to ½¼ ½½ is given in Table 1.884.127 8 127.799 6 8.6020 0.527 Random walks model ½ ´Òµ ´Òµ 70 108 165 214 329 429 592 593 706 755 21. 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. ½ ½¼. Extensions of the data can be found in Leavens and Vermeulen [16]. and the large deviations extremal trajectory is indicated by a dotted line.194.890.5857 0.5841 0.

respectively.03 19.84 19. ½ ½¼. see [15] for more explanation. In fact.528. 4.46 2. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch. Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree. where ½ ´½ º½µ Ì ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½. or 7 ´ÑÓ µ.260 J.86 13.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½.675 7 80.10 20.05 19.831 9 8. there are duplications.976 1.560 1.663 4 77. so do not have any signiﬁcant effect on the size of the tree.559 Random walks model Table 3.1). .099 1.59 23. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿. using (15. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth .14 11.000 Ò 13.791 1.903 2.20 6. C.645 ´Òµ 21.391 8 319. Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2. In this case. In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed. In iterating the multivalued function Ì ½.02 21. LAGARIAS Ò 1 27 2 703 3 9.75 5. The set of inverse iterates associated to any Ò has a tree structure. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼.788 1.65 7.06 4.511 10 77.671 5 704. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ. 5. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ.24 16.32 2.790 1.819 1.362.804.909 1.897 2.83 3.049.48 12.86 7. if Ò ¾ or 8 ´ÑÓ µ.86 5.) empirically like . (See Table 3 in Section 4.566. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ. 15. these trees have the minimal and maximal number of leaves possible for depth .511 6 6.631.817.65 5.

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

the corresponding random walk . Let Æ ´ µ and Æ · ´ µ denote the minimal and maximal number of leaves. 2. 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 . from a given root node . C. In this table. where ¼ represents the entering vertex being even and ½ represents the entering vertex being odd. Ê´ µ counts the number of distinct edge-labeled ¿Ü · ½ trees of depth . Let ´½ º¾µ ÐÑ Ò ½ ¼ ´ µ Ò ½ · · Ð Ñ ×ÙÔ · ´ µ ½ Then these quantities satisfy the inequalities ½ Applegate and Lagarias [3] compared this conjecture with predictions derived from one of the branching random walk models of [15]. 5. Here the pruned 3x+1 tree Ì ´ µ is the tree of inverse iterates of the ¿Ü · ½ function grown backward to depth . Based on this data.) The random walk arises from assigning the labels ¼ and ½ to the edges of the resulting tree. see [15. 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. 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. with all nodes that have a label congruent to ¼´ÑÓ ¿µ removed.2]. 16. 4. 7. 8 pictured in Figure 4. This can be formalized as Conjecture C . respectively. they proposed that the normalized extreme quantities remain bounded. Theorem 3.262 J. they studied the normalized extreme values ´½ º½µ ·´ µ ¿ Æ ´ µ Ò ´ µ ¿ Æ ·´ µ Table 3 gives data up to ¿¼. Distribution of 3x + 1 Trees Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees. LAGARIAS and as ½ the leaf probability distribution is of the form Ï ´ ¿ µ . (See Athreya and Ney [4] for a general discussion of such processes.

(For actual ¿Ü · ½ trees. and it yields a second child by the yields a child by the map Ò ½ multivalued map Ò ¿ ´¾Ò ½µ (edge label 1) if Ò ¾ or 8 mod 9. This value represents the initial value of the ¿Ü · ½ iteration starting from that vertex of the tree. Let Æ ´ µ and Æ · ´ µ denote the expected value of the smallest number of leaves. The parent (bottom) always ¾Ò (edge label 0). 1/3 each) 2. ¾ takes a step ÐÓ ¾ along an edge labeled ¼ and takes a step ÐÓ ¿ along an edge labeled ½. 4 or 7 1 7 263 (prob. The paper [3] presents more evidence in favor of Conjecture ½These facts are derived using the probability density for Ï given in (15. 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. 1/3 each) 2 8 5 4 1.3X + 1 FUNCTION ITERATES (prob. and we consider the value of that vertex to be Ù . so that each vertex of the tree corresponds to a random walk ending at some position Ù on the real line.4).) We consider as random variables the smallest and largest number of leaves that occur among this set of trees. 5 or 8 0 0 0 0 0 0 1 4 7 2 5 8 Figure 4. 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. and ¢ for a ﬁxed constant ¢ ½. respectively largest number of leaves. . choosing Ê´ µ Ê´ µ ¿ ·½ so ½ ¢ ¿. Transitions of the branching process [9]. The number of leaves in actual ¿Ü · ½ trees empirically exhibits less variability than is predicted by this branching random walk model.

131 2.802 1.869 1.47 996.782 1.249 0.405 2.289 0.475 0.33 1.501 0.490 0.492 0.534 0.76 23.264 J.486 0.489 0.344 2.451 0.026 2.688 1.50 315.475 0.422 0.342 0.252 0.898 1.500 1.738 1.507 0.401 0.490 0.62 1328.762 1.37 3.375 2.750 0.302 0. .490 0. C.78 2.49 9.36 3149.259 0.190 2.481 0.753 1.758 1.802 1.77 2362.255 2.81 4199.911 1.482 0.898 1.34 420.390 2.75 5599.238 1.923 1.394 0.750 0. Normalized extreme values for ¿Ü · ½ trees and for the branching process.246 0.741 1.03 177.273 0.780 1.751 1.747 1.32 17.791 1.305 0.465 0.728 1.375 0.742 1.986 2.271 0.507 0.562 0.292 2.45 560.270 2.076 2.486 0.62 7.255 0.16 4.774 1.488 0.750 1.489 0.265 0.433 10850 20436 38550 72806 137670 260612 493824 936690 1778360 3379372 6427190 12232928 23300652 44414366 84713872 161686324 308780220 100 134 178 237 311 413 548 736 988 1314 1744 2309 3084 4130 5500 7336 9788 114 154 206 274 363 484 649 868 1159 1549 2052 2747 Table 4.491 1.491 0.778 1.268 0.669 1.307 0.409 0.83 99.77 133.327 2.232 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 1.352 0.746 1.557 1.38 236.243 0.782 1.784 1.481 0.118 2.633 0.738 1.240 0.419 2.60 747.57 42.633 0.12 74.300 0.67 0.68 31.83 1771. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a ﬁxed node Ò´ÑÓ ¿ ·½µ.758 1.562 0.534 0.488 0.688 1.422 0.360 2.99 13.748 1.748 0.746 1.21 5.534 0.166 2.211 2.475 0.281 0.09 56.481 0.310 2.500 1.054 2.792 1.802 1.481 0.

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

[22] B. [25] G. 21 (1985). private communication. Thwaites. J. 1681. Lecture Notes in Math. pp. Terras. Wirsching. pp. [24] S. Vyssotsky. Wagon. The Collatz Problem. LAGARIAS [21] R. Springer-Verlag. The Dynamical System Generated by the ¿Ò · ½ Function. 241–252. Math. Appl. My Conjecture. [23] V. . No. 1998. Inst. Math. Intelligencer 7 (1985). C. 35–41.266 J. 72–76. pp. Bull. New York. Acta Arithmetica 30 (1976). A Stopping Time Problem on the Positive Integers.

Sign up to vote on this title

UsefulNot useful- Classic Brainteasers - Martin Gardener
- Mathematical Magic Tricks
- Martin Gardner- Wheels,Life and Other Mathematical Amusements
- Math Puzzles Sacred Geometry Mysteries Etc
- Martin Gardner - The Colossal Book of Mathematics
- GARDNER04
- The Penguine Book of Curious and Interesting Puzzles
- Martin Gardner the Last Recreations
- Math Puzzles and Games
- Everybody's Illustrated Book of Puzzles
- Puzzlers' Tribute - A Feast for the Mind - David Wolfe & Tom Rodgers, Eds.
- Martin Gardner - A Die of Another Color
- The Big Book of Mind-Bending Puzzles
- Magic Tricks (Misc.)
- Test Your Math IQ
- 54601090 Maths Puzzles
- Chinese Maths Olympiad
- Modern Conjurer Magic Trick Ebook
- Colin Mcleod - Deceit
- Math Tricks, Brain Twisters and Puzzles by Joseph Degrazia PDF Tam
- IQ Puzzles
- Mathematics
- A Friendly Mathematics Competition - 35 Years of Teamwork in Indiana
- Math Olympiad Problems All Countries (1989 2009)
- 40 Fun-Tabulous Math Puzzles
- MATHEMATICS
- Mathematical Puzzles
- Math Puzzles and Games
- Mathemagic - Advanced Math Puzzles