Contents

Foreword Elwyn Berlekamp and Tom Rodgers

Ü ½ ¿ ½¿ ½ ¾½ ¾¿ ¾ ¾ ¾ ¿ ¿

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

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

vi

Contents

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

½ ¿

½¼¿ ½¼ ½½ ½¾ ½¿½ ½¿ ½ ½ ½ ½ ¿ ½

Contents

vii

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

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

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

Forward

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

x

FORWARD

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

000mph.4 million mph Galaxy’s speed through the debris of the Big Bang . ROWETT Speed A child cycles ‘round the schoolyard Which lies on the Earth turning hard The Earth rounds the Sun As Sol does “the ton” And our Galaxy flies — Gee! I’m tired 7 mph — child cyclist 700mphEarth’ssurface(NorthAfrica) 70.24 T.000 mph turning speed of Galaxy 1.roughaveragespeedofEarth 700.

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

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

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

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

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

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

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

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]. ÏÇÅ Æ ÇÄÁÎ ÅÁÌ Ë Î ÆÌ Æ ËÌË ÄÄË ÉÍ Ä ÄÍ Ê Ä Ê ËÄ È References Martin Gardner. E. KNUTH Puzzle #7.34 D. Chapter 6. Puzzles. and Word Plays (New York: Copernicus. . The Universe in a Handkerchief: Lewis Carroll’s Mathematical Recreations. 1996). Games.

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

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

(1) Find the closest such times. 36. 32. 2:41:42. Problem 6: Which two numbers come at the end of this sequence? 2. 6. For example. 44. 3:59:53. Can you divide the figure into three identical shapes? ´Ü   µ´Ü   µ´Ü   µ ´Ü   ݵ´Ü   Þ µ Problem 8: A 24-hour digital watch has many times that are palindromic.) These curious combinations occur 660 times a day. 23:55:32. . 64. 30. 50. Problem 5: Calculate the expansion of the following 26 terms. (Ignore the colons. 56. 1:01:01. 4. 60. 46. 40. etc. 66. (3) Find the longest time span without a palindromic time. 13:22:31. 42. 34. y Problem 7: The figure shown here is the solution to the problem of dividing the figure into four identical shapes.38 N. 54. x. 52. (2) Find the two palindromes whose difference is closest to 12 hours. YOSHIGAHARA Problem 4: Arrange the following five pieces to make the shape of a star. 62.

4. Can you divide this triangle into two parts with equal area using only 2 matchsticks? Problem 10: Cover the 7 ¢ 7 square on the left with the 12 L-shaped pieces on the right.CREATIVE PUZZLE THINKING 39 Problem 9: A right triangle with sides 3. 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. but you may rotate them in the plane. .

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

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 ź Ϻ 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 Ðݸ 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 Á̸½µ 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°´ ΠƵ · 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׺ºº ÈÊÁÆÌ ÈÊÁÆÌ ÓÔÝÖ Ø ½ ¿¸ Öº ź Ϻ 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 Á̵ Î Ä´ÅÁ °´Æ°¸ Á Á̸½µµ ³ Ø Ú ÐÙ Ó Ø Í´ Á Á̵ δ Á Á̵¶Î´ Á Á̵¶Î´ Á Á̵ ³ Ù Ø ËÍÅ Í Ë ËÍÅ Í Ë · Í´ Á Á̵ 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.

e. Even Here’s an example that sounds a bit different. At the end. the card chosen must be in the center of the array. Ask somebody to think of one of the cards without telling you which card he (or she) is thinking of.. Now ask him (or her) which of the three columns contains the card. f (n) each value eventually is “pulled” into the black hole of 153. and then test to see whether other numbers are drawn to it. 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. and re-dealing out by rows. but be sure to sandwich the column that contains the chosen card between the other two columns. Repeat asking which column. There are two ways to prove this. First of all. Then. so 153 is indeed a fixed point. laying out three across at a time). Now re-lay out the cards by laying out by rows (i. card 11. But for those who enjoy programs. Repeat one last time. It’s a classic card trick. To find an analagous black hole for larger powers (yes. Second. W. which is to say. Card Tricks. Regroup the cards by picking up the cards by whole columns intact. there are some). for large numbers n. you will need first to discover a number that equals the sum of the fourth (or higher) power of its digits.46 M. n. Arrange them in seven horizontal rows and three vertical columns. ECKER The argument for why it works is similar to the case with the 123 example. This is the card in the fourth row and second column. yet meets the two criteria for a black hole. for suitably small numbers. ½¼ ¾¼ ½¼¼ ½½¼ ½¾¼ ½¿¼ ½ ¼ ½ ¼ ÄË ÈÊÁÆÌ Ì ¾½ Ö × ¹ ÈÖÓ ÈÊÁÆÌ ÔØ Ý Öº ź Ϻ ÁÆÌ ¹ ÁŠδ¾½µ¸ ´ ¸¿µ ÁÌ Ê ¼ ÇÊ Æ ½ ÌÇ ¾½ δƵ 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 . 1¿ + 3¿ + 5¿ = 153. Remove 21 cards from an ordinary deck. regrouping cards with the designated column being in the middle. I’ll omit the proof. try this one from one of my readers. note that. for the black-hole attraction.

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 δƵ 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 ¿  Ǵõ Æ Æ · ½ δƵ ´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 Ǵõ Ã Æ Ì Ã ËÏ È Ç´ µ¸ Ç´¾µ ÇÊ Á ½ ÌÇ Perhaps it is not surprising.e. Fall 1992. as with the sisyphian strings. No. generalizes somewhat. further iteration with 6174 is pointless: 7641–1467 = 6174. don’t take one of the nine numbers with four identical digits). write down the largest permutation of the number. Within the total of seven steps. 48. nonetheless. That is. and subtract. the smallest permutation (allowing initial zeros as digits). involve numbers. but this trick.. you always reach 6174. . ¾ Kaprekar’s Constant: What a Difference 6174 Makes! Most black holes. At that point. Apply this same process to the difference just obtained. ¾REC. 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.

and ´Òµ preceding two elements. and these digits sum to 15. and 15. 55. Now take the sum of the digits of these divisors. 3. This one is a bit more tedious. The black-hole number this time is 15. but you will reach 6174. That is. 3. the smallest is 0288. 5. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. ECKER Example: Start with 8028. To get a quick and dirty feel for why phi arises. ´½µ ´Ò   ½µ · ´Ò   ¾µ for integers Ò ¾. 89.). Iterate until a number repeats. ½ . The first two numbers numbers: 1. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ ľ   Ä   ½ ¼ . 21. In fact. 1. The ratios seem to be converging to a number around 1. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. This one may defy not only your ability to explain it. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. we would have obtained the reciprocal of phi as the limit. The largest permutation is 8820. 5. Its divisors are 1. More formally. Its value is roughly 1. etc. but without using a program. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò   call it Ä — as Ò increases. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. 8.618 . Your own example may take more steps. 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. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors.48 M. including 1 and the number itself. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. and the difference is 8532. 13. but fewer authors use that approach. Repeat with 8532 to calculate 8532–2358 = 6174. but your very equilibrium. phi. 34. 2. Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . Had we used the ratios ´Òµ ´Ò · ½µ instead.6 or so. W.

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

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

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

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

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

6. and they must cross in 25 minutes. . The Bridge.. B takes 2 minutes. Divide the figure below into (a) 4 congruent pieces and (b) 3 congruent pieces. 8. (a) There are six men with crossing times of 1. Find nine single-digit numbers other than (1. 6. 2. 2. 8. and D takes 10 minutes. 9. E14.PUZZLES FROM AROUND THE WORLD 55 E10. If it cannot move to a square previously visited. Can all four men cross the bridge? If so. 3. Four men must cross a bridge. and the bridge will hold up to three men at a time. Place the numbers from 1 to 15 in the 3 ¢ 5 array so that each column has the same sum and each row has the same sum. A pair must walk together at the rate of the slower man’s pace. 4. They all start on the same side and have 17 minutes to get across. and 9) with a sum of 45 and a product of 9 362.880. 3. E11. and 9 minutes and they must cross in 31 minutes. A knight is placed on an infinite checkerboard.. It is night. Each man walks at a different speed: A takes 1 minute to cross. . and they need their one flashlight to guide them on any crossing. 7. E13.. (b) There are seven men with crossing times of 1. and 10 minutes. how? Try these other problems. A maximum of two people can cross at one time. how can you make it unable to move in as few moves as possible? E12. C takes 5 minutes.

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

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

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

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

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

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

A number may be approximated by a /b where a and b are integers.0107 0.” Your response is “No need to tell me more. Define the goodness of fit of a /b to x as g (a. e ) 2. d. The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem. 113. I’ll let you off the hook. with g (355. b) Find the minimum c and d so that g (c. I. Minimum Cutting Length. Since Smith missed the problem tonight and Jones missed it at the party two years ago. their ages are ” H14. The drawer starts with 16 socks each Monday morning (eight blue. six black. My uncle’s ritual for dressing each morning except Sunday includes a trip to the sock drawer. Find other such examples. HESS H12.0107 . then (2) wears any matching pair and returns the odd sock to the drawer or (3) returns the three socks to the drawer if he has no matching pair and repeats steps (1) and (3) until he completes step (2). b. ) = 0. An approximation to is 355/113. (a) Find the minimum a and b so that g (a. x) = a bx .71828 . . into five parts of equal area? (c) an equilateral triangle into five parts of equal area? H15. ) 0.0107. two brown) and ends up with four socks each Saturday evening. and proceeds to explain: “I normally ask guests to determine the ages of my three children. (b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes? . (a) The figure shows two Pythagorean triangles with a common side where three of the five side-lengths are prime numbers. given the sum and product of their ages. (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 e = H16.62 R.a . b.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. where he (1) picks out three socks at random.

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

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

10. (11. ¾ gives Ø E5. Anywhere within 10 miles of the north pole. 6). (7. 4. 10. and Ü ½ . 0). ؾ ¾¾ hours. E2. 6). (5. In each case the numbers combine by summing the products of their digits. E6. 4. 2. 5). But this is not right because 18 is not the difference of 38 and 16. 6). 0. so E8. 5. 15 moves: (15. 2). (0. The number is 381. so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. At first glance. 6. (1.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 9). 8. 2). (3. One of the terms is ´Ü   ܵ. 6). (1. it appears that the rule might be subtraction. the next greater number is 7. 6). 6. E3. 2. E4. 2). (5. ¼. 4. (11. (7. . Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. 7. 0). 0). E10. (3.729. (9. (3. The goal is to find the smallest integer values of and so that Ò is a minimum and ¼   By numerical search we find ¾´ ·   ½µ   ½ ¼¼ ¼¾ . the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. 8. 6. 4. . 0. The trick here is that 1 knot = 1 nautical mile per hour. Instead. The license number is 741.741. 0). (3. E9.425. (1. 6). 0. E7. 10. ½ ¼½ or ¼ and ½ ¼½. 0). 0.654. 2. ¿   ½ ¼¼¼   are the smallest values satisfying the above equations. 6. (9. Thus Ü ½½. 0). 9. Thus and Ò ¾½¼ transfers is the minimum.

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

11. 1). 178. (35. 11. 5. (23. D will finish burning. (30. 10). (23. (17. 6). 10). 10). In 3. 0). 1). Cut B at that corresponding point. (35. 146. 0. producing pieces A and B. (37. 3. (31. 3). 0). 134. 5. When C is consumed.PUZZLES FROM AROUND THE WORLD 67 E16. (30. 3. 0. 0). Solutions to Medium Problems M1. (31. 168. 0. 6). (19. 10). 5. 11. 2). 0. 0). 145. (24. 1). 127. 0). (27. (23. 0. . 11. 10. 0. 6). (24. 3. 2). 11. (33. (31. 9. (20. (31. 6. 0). (25. 11. 0. 0). 0). (28. 0. (34. 0. 148. 4). 0. 1. 136. (29. (22. 11. 5. 4. 0). 1. (22. 3). (24. 9. E17. 125. (23. 6). ½¿ ½¿ ½ Ê 123. (22. 0. the probability of a right triangle is ½ ¾½ or . 10). (29. 1). 10. 3. See the figure. 6). (19. 2. 0). 4). 0. (20. (18. 11. 4. 0). 0. (17. (30. 0. noting the point that burns last. 0). 1. This is the shortest time interval that can be measured. 2. (23. 0. 0. 0). 6). (29. (25. 0). 11. (31. At the same time. 0. 0). and the time is 8:24. 6). 0). 10). 138. 0. 1. 1. 5). (25. M2. (a) 32 moves. (21. (29. 0). If the hour hand is exactly on a second mark then the second hand will always be on the 12. 11. 124. 0). (35. (25. 11. 0). (18. 158. 1. light the other end of D. 11. (25. producing pieces C and D. 0. (28. (26. 0). (29. 0. 10). Cut the lace in half. 167. (37. (24. 1). 1. (36. 11. (18. (25. 6. 4. (34. 1. (24. Burn A from both ends. (30. 11. 6). 0). (26. (25. 0. (33. 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. 11. (36. (30. 10). 11. 0). 1. (19. (21. 11. (19. 4). (36. 10). 0) (b) 40 moves. 1). 0). There are 7 ¢ 6/2 = 21 choices using vertex 1. (23. 128. (26. 6. (30. 156.75 minutes. 126. 147. 0). 0. 1). 6). 6. 0). 6). (18. 4. 10). 10). 10). 0). 10). 10). 0). 5). 6). (29. 6. (11. The probability of an acute triangle is ¿ ¾½ or ½ . 6. (24. 0). 0). start burning C from both ends and D from one end. 4). as seen from the list below.

The series can be expressed by a simple continuous function: ´Òµ To get Ü. Ë ¹ À ½¼ Â Ë À ¿¾ ¿¾ ¿¾ ¿¾ Ë Ãɽ¼ À ¹ ¹ ÃÉ Ë ¹ À Ãɽ¼ ÃÉ ½¼ M4. North wins the sixth trick with the 10 of diamonds. North and East discard all their diamonds on these tricks. HESS M3. East playing the 9. we take the limit of ½¾¼´¾Ò   ½µ Ò ÓÖ Ò ´Òµ as Ò goes to 0. North next leads a heart which West trumps with the queen. M6. ¿½ ½¾¼ ÐÒ ¾ ¼ Ü M5. For multiple of 3 for Ô   ½. The final two tricks are won by South’s long diamonds. I. with East underruffing North each time. The only real root is Ü ¿ we get the prime 17. For ¾. There are six ways as shown below. ¿. Ü¿   ¾Ü   ¼. The first five tricks are alternate spade and heart ruffs by North and West. The next four tricks are won by South with the 5432 of clubs. so that ¾Ô · Ô¾ is always a Ô Ô ¿ we have Ô · ½ or . Ô M7.68 R. Then ¾Ô ¿Ò   ½ and Ô¾ ¿Ñ · ½. The hand below does the job if the play goes as follows.

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

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

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

10).0119 . Solutions to Hard Problems H1. 8). 11. 9) and (14. I. 12. Shown below. 14. 14. 11.72 R. The scores for the Good Eggs are (15. (15. 11. 13. 12. 11. The number in each triangle indicates its area relative to the other triangles in the figure. 7). (15. 14. . 14. (15. (15. 12. M20. 13. The points are ´ ¿   ½µ ¾ of the way along each diagonal for a distance of 3. 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. There are 5 subjects with 15 marks possible in each. 8). 12. There are 7 Good Eggs. 9). Humpty Dumpty got a 10 in arithmetic. 10. 12. 12. 13. 14. (15. 10. 6).

and stretch in the band due to the building. 0. Ö earths radius. where . If Sir George travels at speed Ú then × Ú ´½ ¼   Òµ Ò where 1440 is the number of minutes in a day. 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. 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. Its area is Ò× ¾.PUZZLES FROM AROUND THE WORLD 73 H2. The distance from B to the distance BC. Relative to the center point of the ball. H4. For a full rotation of the ball on the circle the point P will execute ¾ rotations about the axis. Let Û width of the building. Three equally spaced punches in a straight line do the job if the square of the spacing is irrational. 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. Imagine a regular polygon of Ò sides after Sir George’s trip. H5. Point P will return to the top again as Ô well. For Ö ½È ´ ¼ ¿ ¾  ¼ ¾ ½¼ ½ ¼ ¼ ¼ ½ µ its initial position was (0.300. Consider the ball rotating about the axis AC without slipping. with 159. as shown below. height of the building. the point P will have coordinates È Ô Ô  Ö × Ò ¾ Ö´ Ó×   ½µ ¾ Ö´ Ó× · ½µ ¾. BCB rolls along the plane. . 1). TheÔ cone.1968 Ú ¾ H3.

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

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

The ratio È takes on a maximum when Ö ¼¾ ¼ . Now consider two such circles of different sizes as shown. The maximum ratio È Ö ¼¾ ¼ . A view of the pyramid looking down from S perpendicular to the plane ABCD will look like the figure below. There are 7 circles on each end with 105 sets of 3 circles in the middle. 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. For Ò ½ there is just enough room for 329 circles. HESS the shape is È Ô ½ ´¾ · ¾ Ö ·   Ö. H9. They define the four tangent lines drawn to produce ABCD as shown. I. The circles need to be packed as shown below.9973967 . Raise S and T appropriately above the . Note that the ratio È is 0.76 R. The point S may be raised above the plane to give unit altitudes if the inscribing circle has radius ½. (b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved. Each has a radius ½.25 for the entire square or a circle inscribed in µ the square.

Let Ô the probability of the server winning a point. È ´ ¼   ¿¼µ Ô · ÕÈ ´ ¼ ¼µ.PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit. 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. È ´¼   ¼µ È ´¿¼   ½ µ when Ô ¾   Ô ¿. and let Õ ½   Ô.3586 Pairing probability 0.8027 0.3074 1.2511 1. È ´¼   ¼µ È ´ ¼   ¿¼µ when Ô¿   Ô¾   ¾Ô ½. · È ´ ¼   ¿¼µ Õ È ´ ¼   ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . which leads to Ô ¼ ½½ ¿ ¾ . ¾ H12. È ´¿¼   ¼µ ÔÈ ´ ¼   ¼µ.7849 . Clearly ST is not perpendicular to the plane.2801 1. This leads to ¾ È ´ ¼   ¼µ Ô¾Ô Õ¾ . ·¾ Ô ´½ ½ Õ µ È ´¼   ¼µ . Let È ´   µ be the probability of the server winning the game when the server has points and the receiver has points. H11.2270 1. At Ò games to Ò. which leads to Ô ¼ ½ ¿¿ ¼ . È ´ ¼   ¼µ ÔÈ ´ ¼   ¿¼µ · ÕÈ ´¿¼   ¼µ. ¾ ¾ Õ´ È ´¿¼   ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ . Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1. Ô  Õ At a score of Ò games to Ò (Ò ¼ to 5). ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ .2069 1. A computer program was written to calculate the probabilities shown in the table below.7805 0 .8286 0.8163 0.7892 0.

512. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô . 12). ¼ ¼¼ ½ ¼. ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . (b) 605 years. . (a) Length ½ ¿ ¾½ ½ ¼ as shown. 8.042. prime. I. we get (a) (b) H16. We look for cases where (1) the oldest child is under 20. (a) 45 years. which must have been guessed by Smith. Solutions occur for (b) ¾ ½. The product and sum two years ago could be achieved by (2. · ½.546.444 years. ´ ¾   ½µ ¾. H17.613. ¿ ½¾¼. · ½. 16). 15).78 R. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½. (c) 17. ¼ ¼½¼ . ´ ¾ · ½µ ¾ prime.093. Using a multiprecision continued fraction program. H14. a=10. . ´ prime. HESS H13. . which must have been guessed by Jones two years ago. (d) Yes. 7. H15. ¾   ½µ ¾. (b) Length ¾ ¼¾½½¾ ¿ as shown. The product and sum could be achieved by (4.641. 6. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. (c) Unknown.321 is the first of 5 in a row. (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ . (2) the younger two children have different ages. There is only one set of ages that accomplishies this: (5. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago.

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

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

6406814 3. Call the Ø such pole in the upper half-plane Þ Ü · Ý . Since the equation in Ý is even. which reduces to ´½ · Ü · Ý µ´Ü   Ý µ ´ Ó×´Ò · ½µÝ   × Ò´Ò · ½µÝ µ where ´Ü¾ · ݾ µ ´Ò·½µÜ .5432385 32. ¼ for all . not including its pole at Þ ¼. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to infinity.2238350 26.8505482 39. Then Ý approximately equals   ½·ÐÒ´ µ . The table below gives numerical values of the first few. For larger define ´¾ ·½ ¾µ . Ü 1 2 3 4 5 6 7 (6) The contribution to 2.PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ.5012690 3. A pole of Õ´Þ µ occurs Þ .  ¾ × ´Ý · µ  Ü Ý .4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ   Þ µÕ´Þ µ ´½ · Þ µÒ·½. (5) On Þ Ê for sufficiently large Ê.6745053 3. where Þ at ½ · Þ Ü · Ý.1512074 45.8221528 Ý 7. The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ.4614892 13.0888430 2.0262969 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 ܾ · Ü · ݾ Ø Ò´Ý · µ Ý  ÒÜ goes to zero as Ò goes to infinity.8790560 20. Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ.2916783 3. × ÒÝ Ý we can look for poles in the upper half-plane only and reflect each one into the lower half-plane.

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

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

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

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

Pieces with odd-numbered labels have no symmetry and. but our mnemonics are as follows. By subtracting 10 from the Hexagon. . 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. . to the Crook. Pieces 4 and 8 have only rotational symmetry and exist in enantiomorphous pairs. whose head is on the left. with its handle on the left. Notice that 0 has the greatest symmetry. Crown. with its hook on the left. and the other even numbers have reflective symmetries. a parallelogram drawn in the opposite way from the usual textbook fashion. Negative (bar sinister?) labels are given to the Bar. K. which appears to be turning to the left. to the Hook. exist in enantiomorphous pairs. The symmetries of the Hexagon are those of the dihedral group D6. and 9. . to the Yacht. sailing to the left. The Chevron. The nineteen Hexiamond pieces. 5. GUY Figure 1. which have no symmetry. to the Signpost. and the Sphinx.86 R. O’Beirne had already suggested a numbering of the pieces from 1 to 19. the multiples of 4 have rotational symmetry through 180 degrees. and subtracting each of the labels 11 to 19 from 20 we arrive at the labeling in Figure 1. and Lobster each have a single symmetry of reflection. 7. . . to the Snake. 3. . with its pointer on the left. as do pieces 1. ¿. ½. The Butterfly has the same symmetries as a rectangle: Its position is described by that of its body. together with those that have rotational symmetry only.

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

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

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

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

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

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

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

by the Hexagon’s position Ø Ù Total 264 1094 124.94 R. K. are 130 195 533 193 377 2214 111. I wrote to Kate Jones.518 A 75. 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. for those who prefer their puzzles to have just one answer. Their numbers. GUY Figure 18. classified according to the position of the Chevron.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. Six pieces meet at a point.447 F 5727 G 1914 I don’t think that this need take the fun out of one of the best two-dimensional puzzles ever invented. Figure 19. Ô Ü (3-symmetry).489 B 15. president of Kadon Enterprises and discovered that she markets a set of “iamonds” . Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to find all the solutions.717 C 6675 D 7549 E 11. On the contrary.

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

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

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

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

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

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

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

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

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

double polyominoes can be formed with ¾ ¢ ¾ quadrants of each color and then dispersed through a maze-like sequence of moves to the desired final position (Figure 6). with checkerboarded arrangements (Figure 5). formed with checkers (Figure 4). Figure 5. The pieces can form duplicate and triplicate models of themselves. . 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. Quintachex: the pentominoes plus a ¾ ¢ ¾ square checkerboarded on both sides (different on the two sides).110 K. or diagonal. 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. In another use. column.

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

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

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

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

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

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

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.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

Therefore. C. That is. then there are ´ ¿µ´ µ´¿ µ¾ . C. If you cannot. additionally. 2. 1. There are 93 solutions in this case. 0. Trial and error is not a very fruitful way of trying to solve this puzzle. B. try again to solve the puzzle. 0). These hints are to be regarded as progressive. 1. but there still remain a great many cases that are nearly right. and 3. The combinatorial puzzle uses only six colors in its construction.146 J.½½ . proceed to the next hint. each . and 3 are any six distinctive colors. B. If. Notice that the eight pieces must form the eight corners of the completed cube — one piece for each corner. A. Solution Hints.½ distinct ways of constructing the cube. B. each of the 48 faces of the pieces is colored with a unique color. FARRELL Figure 1. A. ½º Call the six colors A. different” but one interpretation could be to orient the cube to sit in the positive octant in space with Biggie always occupying the corner (0.½ ¼. and the solver has the additional clue that a face is all of one color. after reading a hint. and C will turn out to be a solution set. 2.

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

and 3). or off-diagonal around all four lateral faces is always 42. The solver will notice that the completed cube has a “fault-free” property. Then place the longs and. B. Figure 3. This will assure that only one solution is attainable with the colors A–B–C (the proof is left to the reader). That is. 2. The Magic Die has the amazing property that the sum of any row. If only one solution is desired. Smally. one that gives only one solution. FARRELL must go by looking at its A–B–C corner. We like to have three Smallys prepared: one that yields two solutions. This will change the parity on one corner of the cube so that a solution is impossible using that set of three colors. take any piece and interchange two of the colors A. and C (or two of 1. 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. . column. There is another solution to this puzzle using the colors 1– 2–3.148 J. no seam runs completely through the cube in any direction and therefore no rotation of any part of the cube is possible. finally. main diagonal (upper left to lower right).

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

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

when two cubes of the same color are in position. The second observation is that for a cube to be assembled with the nine colors on each face. then. Since during the course of trying to solve this puzzle a conflict often occurs. the location of the third cube is determined. and only 10. FARHI a difficult exercise for a moderately proficient programmer to verify the results obtained.152 S. It becomes necessary. as determined in Section 4. as is shown in Section 5.700 (36!/12!24!) different ways of which. to separate these cases into disjoint sets of isomorphisms. Fortunately. and the most difficult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1.105 of these combinations meet the three-cubes-of-each-color requirement. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. is that there are numerous solutions.251. The method of sorting the 10. only 133. The third observation.691 combinations into 148 disjoint sets is described in Section 6. This was achieved through the use of a computer program. in which the number of isomorphisms using the set of thirty-six dicubes is also determined. The individual dicubes can be colored thirty-six different ways. this rule then tells the experimenter to backtrack. 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 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 3. some with the tricube on the edge and some with the tricube through the center of the cube. only thirty dicubes are required. .691 combinations need to be sorted into disjoint sets.677. A proof is given in Section 2. Twelve of these can be selected in 1. Since one person can assign the fourth and fifth colors to orange and red and another person to red and orange. the results are shown in Table 1. The question then arises: How many solutions are there? Most intriguing about this puzzle. but not in the center of a face. The previous question now becomes more interesting.105 combinations are isomorphisms. it becomes apparent that most of the 133. Thus. obtained after some experimentation. none of the nine planes may contain two cubes of the same color.

:Zppp 10 10 0 0 0 0 0 .DZppp 360 156 428 24 208 8 36 /90eyz 180 60 16 0 38 32 0 /:0eyz 360 120 80 4 24 4 4 /:Z[yz 720 240 182 8 50 16 24 /:Zpez 360 120 124 4 38 44 16 /:Zppp 180 60 408 0 186 32 64 /COeyz 360 72 0 0 6 4 0 /DXeyz 720 132 18 0 4 2 0 /DZ[yz 720 132 72 0 32 12 0 /DZepz 1439 264 41 0 40 9 0 /DZppp 360 66 80 0 117 28 0 /MNeyz 60 8 0 0 0 0 0 /MOeyz 720 96 8 0 34 8 16 /MXeyz 720 84 32 0 30 12 12 /MYeyz 720 84 26 0 18 6 4 /MZ[yz 1439 168 62 2 70 21 23 /MZepz 1439 168 41 3 35 15 16 /MZpez 1439 168 51 4 57 17 20 /MZppp 720 84 68 4 73 26 14 /NXeyz 720 96 8 0 36 12 12 /NYeyz 1439 192 40 2 64 17 24 /NZmyz 720 96 8 0 10 2 0 /NZnpz 1439 192 53 2 53 11 17 /NZpez 720 96 36 4 48 24 32 /NZpfp 1439 192 10 0 49 34 23 /Ob[yz 1439 168 25 1 81 24 16 /Obmyz 1439 168 98 5 65 18 19 /Obofz 720 84 6 0 18 6 8 /Obpez 1439 168 124 3 53 8 17 .:Z[yz 60 60 0 0 0 0 0 . 153 No.THE NINE COLOR PUZZLE Table 1.DZ[yz 1439 624 96 4 80 6 15 .DZpez 720 312 198 12 83 10 16 .DOeyz 720 312 20 0 12 0 4 . 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center .COeyz 360 180 8 0 16 16 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 .154 S. FARHI No.

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

FARHI 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 .

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

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. and 3.105 Possible Dicube Color Combinations The tricube colors are defined as colors 1. 4. If none of the previous selections include color 3. FARHI If an irregular tricube is used. . there are three possible locations.” The remaining two colors must all be on edges. then one color is selected from the third row and if color 3 has been selected twice. cubes of the same color are never in the same plane. If color 3 has been selected once. the remaining two cubes must be located on an edge. and 14. such as shown by the letter “C.158 S. 5. The only way the remaining two cubes of that color can be displayed on all six faces is for them to be located on opposite corners as shown in Figure 3 by the letter “A. then color 3 is combined with two colors from the third row. Since the corner colors are displayed on three faces. the edge case fills spaces 2. 2. The two dicubes with color 1 may use any two colors in the first row of the following table. then a third dicube with color 2 and any color from the second row is selected. then none are selected from the third row. The “No Two in the Same Plane” Rule Figure 3. 11. If color 2 is selected from the first row. 2. and 11.” The remaining six corners must have cubes of different colors. One cube is centrally located and thus not visible.” In each case. and on a centerface such as shown by the letter “B. Determination of the 133. accounting for two faces. and 11. the center case fills spaces 5. The corner case fills spaces 1. 3.

or 5. 6. 3. (7. color 5 is combined with colors in the fifth row such that there are three cubes with color 5. 9). 5). Identification of the second color of these ordered pairs is sufficient to define all the colors.105 possible color combinations. using the same algorithm as before. 8). 9. 4).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. a computer program established 133. (5. the two new colors are identified as colors 6 and 7. (4. 6. etc. 9). 2). Similarly if color 2 is not associated with colors 3. 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. Using this algorithm. 7. 8. 5. given the second numbers 2. 4. For example.691 color combinations. 6). 8). (2. and 9 one can logically deduce the first colors by the three . (1. 3). 8. (3. (7. (4. 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. 6). only 10. (5. 9) and (8. 7). 6. 9. Identification and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. This results in requiring only thirty cubes and. 4. (6. 5.

9). (68). Shown in Figure 4 are an edge solution and a center solution for the color combination of Figure 1. and then assigning a printable character to each. the first entry in Table 1 is . three at a time. and 99. Colors 1 and 3. 8). then additional criteria for recording solutions are recommended. 8). (5. has sixty isomorphisms.:Z[yz.e. . pairs of three at a time and four at a time with two at a time. 91. by adding twenty-three to each and using the ASCII character set codes (ASCII stands for American Standard Code for Information Interchange). 9). The use of this six-character identification greatly simplified the process of sorting the 10.691 cases into 148 disjoint sets of isomorphisms.:Z[yz. For example. four at a time. (4. Different solutions often have the same three-digit identification. 121. The ASCII code for 23 is not a printable character. (67). 90. (98). (4. . 6). its isomorphisms are determined and removed from the list. If the solutions are to be catalogued. (5. The first entry. The head of the list then became .691 color combinations obtained by the algorithm described in Section 5 were listed in order according to their ASCII characters. Any permissible interchange of colors is an isomorphism.160 S. (1. The 10. FARHI cubes of each color requirement and the restriction that the second number be greater than the first. 9). (35). The solutions are identified by a three-digit number indicating the color of the buried cube and the two “edge” colors. including pairs of two at a time. cannot be interchanged with other colors. 58. 7). The ASCII codes for these characters are: 46. 3). Not all of these mappings produce a new isomorphism..:Zppp. on the ends of the tricube. can be swapped with each other. Decoding a name is quite simple. This process then identified the 148 disjoint sets of isomorphisms. (24). Thus this color combination is identified as /:Z[yz. and six at a time including a triplet of two at a time. 6). (2. and 122. 2). five at a time including three at a time with two at a time. which is in the center of the tricube. The process was then continued until the list was exhausted. (7. It is then a simple process to recognize that this represents the dicubes colored (1. 5). 67. 3). (6. The remaining six colors may be interchanged in numerous ways: two at a time. An inspection of the ASCII table will explain why it was decided to add 23. and (99). Color 2. but not with any of the other colors. Subtracting twenty-three results in: 23. 98. (4. and (8. 68. Note: The smallest possible number is 23 and the largest is 99. These were removed from the list. 35. i. (7. Identification was further simplified by placing these numbers in pairs (see Note below).

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

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

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

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

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

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

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

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

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

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

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

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. .

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

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

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

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

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

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

” “Hazelgrove Box” Strijbos aluminium burr box Burrs. Oscar’s “Dovetail. “Mayer’s Cube. purses Padlocks Chippendale tea chests. cutlery. “Hoffman” cube. puzzle rings Hess wire puzzles.” “Nine of Swords.” “Margot Cube. knives.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” .” Cutler’s burr in a glass Coffin’s “Saturn” JWIP and keychain animals. poison rings Nuts and bolts. Soma. “T” puzzle Polycubes.” 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. checkerboards. “Tak-it-Apart. pens. Dalgety’s “Devil’s Halo” Leather tanglement puzzles Boxes.” Chinese rings. berrocals. gears. cast “ABC. Pentominoes. ball in cube of cubelets 181 ¢ TNGTNG-RIGI TNG-R&F TNG-FLEX OPNOPN-BOX OPN-LOCK OPN-HID Tanglement Puzzles Tanglement of rigid and semirigid parts Tanglement of rigid and flexible parts All flexible Opening Puzzles Opening containers Opening locks Opening/finding hidden compartments not originally designed as puzzles Opening other objects Wire puzzles. Oskar’s keys.

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

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

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. “Jail Nixon” Strip polyhedra Standard “block holes and suck” solution “Jolly Jugs” Cadogan teapots. five square puzzle 19th century riddle prints Anagrams. HORDERN Map folding. or those that can only be represented on a computer . 12 Golf Balls. rebus plates and prints. DALGETY AND E. Archimedes gold Cork for three holes.184 FOL-SH2D FOL-SH3D JUGJUG-STD JUG-CPLX JUG-BASE JUG-NLID JUG-OTHR OTHOTH-ELEC OTH-BAL OTH-MEAS OTH-CUT OTH-RIDD OTH-WORD Folding sheets and strips into 2-D solution Folding sheets and strips into 3-D shapes Puzzle Jugs Puzzle mugs standard Complex mugs requiring special manipulation Pour from base Lidless wine jugs Self-pouring and other patents Other Types of Puzzle Electrical and electronic (non-dexterous) Balancing (non-dexterity) Measuring and weighing puzzles Cutting puzzles Riddles Word puzzles J. Chinese winepots Royale’s patent Luminations “Columbus Egg” Jugs and liquids.

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

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

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

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

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

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

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

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

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

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

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

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

characterized by the following conjecture: Sprouts conjecture. Misère sprouts conjecture. and our ingenuity. The size of the hash table turned out to be a major limitation of the program. The n-spot Sprouts positions evaluated so far fall into a remarkably simple pattern. and searching the successors of a position in order from lowest degree to highest. The only remaining case required to resolve the bet is the 10-spot misère game. This problem seems to lie just beyond our program. 4. To reduce the space still further we used a data compression technique.cmu. 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. Applegate doesn’t believe in these conjectures. We are still left with the nagging problem of resolving a bet between two of the authors. and we devised and implemented two methods to save space without losing too much time efficiency.COMPUTER ANALYSIS OF SPROUTS 201 ¯ of previous searches in a hash table. our computational resources. The data for misère Sprouts fit a similar pattern.cs. go to http://www. Sleator believes in the Sprouts Conjecture and the Misère Sprouts Conjecture. The first player has a winning strategy in n-spot misère Sprouts if and only if n is 0 or 1 modulo 5. or 5 modulo 6. The first player has a winning strategy in n-spot Sprouts if and only if n is 3. a “2” means the second player has a winning strategy. and an asterisk indicates a new result obtained by our program.edu/~sleator .½ ½For a paper describing these results in more detail. We discovered that saving only the losing positions reduced the space requirement by a large factor. 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 toll in human productivity probably exceeded the loss in computer time. found by Hickerson (at U. Davis) has period 5! Stretching between is a beam of darts that seems to move rightward at the speed of light. With the great reduction in computation costs. alien Life forms. with human programmers as the vector. it would be nice to report that. the beam shortens in both directions at the speed of light. While the Life programs distributed with personal computers are harmless toys whose infective power is 100% cancelled by TV. of which the weirdest. by singlehandedly popularizing Conway’s game of Life during the 1970s. Martin Gardner. sabotaged the Free World’s computer industry beyond the wildest dreams of the KGB. like smallpox. and then both ends explode. The biggest shock has been Dean Hickerson’s transformation of the computer from simulation vehicle to automated seeker and explorer. Sadly. as of late 1992. leading to hundreds of unnatural. new developments in the game are still spreading through computer networks. has to be Time 0: Note that it extends leftward with velocity ½ . and the solution of most of the questions initially posed by the game. But the (stationary) right end. with the left end (found by Hartmut Holzwart at the University of Stuttgart) repeating with period 4.C. this is not the case. But if you interrupt it.Strange New Life Forms: Update Bill Gosper The Gathering for Gardner Atlanta International Museum of Art and Design Atlanta. the Life bug no longer poses a threat. Back when only the large corporations could afford computers because they cost hundreds of dollars an hour. Georgia Dear Gathering. 203 . infecting some of the world’s best brains and machinery. clandestine Life programs spread like a virus.

204 B. GOSPER Times 100–105: .

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

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

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

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

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

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

STRANGE NEW LIFE FORMS: UPDATE

211

ºººººººººººººººººººººÓººººººººººººººº ºººººººººººººººººººººÓºÓººººººººººººº ºººººººººººººººººººººÓÓºººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººÓºººººººººººººººººººººººººººººº ººººÓºÓººººººººººººººººººººººÓºÓººººº ºººººÓÓººººººººººººººººººººººÓÓºººººº ººººººººººººººººººººººººººººººÓºººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººÓººººººººººººººººººÓººººººº ºººººººººººÓÓººººººººººººººººÓºÓººººº ººººººººººÓÓºººººººººººººººººÓÓºººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººÓÓÓºººººººººººº ººººººººººººººººººººººÓºººººººººººººº ºººººººººººÓºººººººººººÓººººººººººººº ºººººººººººÓÓºººººººººººººÓÓÓººººÓÓºº ººººººººººÓºÓºººººººººººººÓººººººÓºÓº ºººººººººººººººººººººººººººÓºººººÓººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ÓÓººººººººººººººººººººººººººººººººººº ºÓÓºººººººººººººººººººººººººººººººººº ÓºººººººÓÓººººººººººººººººººººººººÓÓº ºººººººººÓÓºººººººººººººººººººººººÓºÓ ººººººººÓºººººººººººººººººººººººººÓºº ººººººººººººººººººººººººººººººººººººº ºººººººººººººººººººººººººººÓÓºººººººº ºººººººººººººººººººººººººººÓºÓººººººº ºººººººººººººººººººººººººººÓººººººººº ººººººººººººººººººººººººººººººººººººº ºººººººººÓÓºººººººººººººººººººººººººº ººººººººÓºÓºººººººººººººººººººººººººº ººººººººººÓºººººººººººººººººººººººººº
Conway called such oscillators billiard tables, and along with the rest of us, never imagined they could be made by colliding gliders.

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

212

B. GOSPER

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

ØØÔ »»ÛÛÛºÑ Ò ×ÔÖ Ò º ÓÑ» Ð Ò »Ð

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

» Ò Üº ØÑÐ

STRANGE NEW LIFE FORMS: UPDATE

213

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

Hollow Mazes
M. Oskar van Deventer

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

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

Figure 1. Wedge of Wallis.

A silhouette defines a cylindrical region in space, perpendicular to the silhouette surface. With this meaning the word no longer refers to an object, but only to a region of a flat surface (Figure 2). Two or more silhouettes (not necessarily perpendicular to each other) define unambiguously a region or object in space that is the cross-section of two or more cylinders. Dewdney introduced the term projective cast for an object defined by a number of silhouettes. Not every object can be defined as the projective
213

214

M. O. V AN DEVENTER

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

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

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

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

Each side is a two-dimensional control maze: a surface into which slots have been cut. Three viable silhouettes yielding a projective cast with two parts. since a spar cannot jump from one part to another. . Figure 5. There can be no cycles since the center of a cycle would fall out. Each spar passes from one side of the box to the other. Control mazes are always viable because of their construction. A cursor consisting of three mutually perpendicular spars registers one’s position in the hollow maze. Hollow Mazes A hollow maze is a rectangular box with six sides (Figure 6). The resemblance between multiple silhouettes and hollow mazes is evident. sliding along the slots of the control maze on each side. 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.HOLLOW MAZES 215 Figure 4. a single three-dimensional maze is produced from three pairs of two-dimensional mazes. Three viable silhouettes yielding a projective cast with one cycle. and how can a projective cast be given certain properties by its silhouettes? I do not know whether there is a general answer. In this way. Control mazes cannot consist of multiple parts.

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

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

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

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

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

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

I’d have times you. this doesn’t hold. the age problem. Ü and Ý are integers if and only if the second terms in (2) are integers. and it is obvious from (1). . unfortunately. but somewhat more complex than. Parameswaran for the letter that inspired this investigation.” and the second responds: “And if I had from you. assuming ½. Looking for a more symmetric solution. so we only need to check one of the second terms in (2). say Ü Ý. which is similar to. and I thought that the integrality of this was equivalent to integrality of Ü and Ý. assuming ½. this false trail led me to the following simple result. but it is traditional for the parameters and the solutions. This still doesn’t give us a very simple or satisfactory criterion. which is if and only if   ½ divides GCD ´ · ½µ´ · µ. 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. Luckily. One can see from (2). This also divides ´ · ½µ´ · ½µ   ´ · ½µ   ´ · ½µ   ½. The criterion allows us to pick arbitrary and . that Ü is an integer if and only if Ý is an integer. where ´ · ½ · ½µ denotes the GCD of · ½ and · ½.222 D. ´ · ½µ´ · µ ´ · µ´ · ½ · ½µ." Many versions of this problem occur. I first noted that Ü · Ý has a symmetric expression. Consider ´ · ½ · ½µ. as before. though. Hence a natural question is to consider the diophantine question Which integer problems have integer solutions? In the early 1990s [1]. 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 am also intrigued to see that any and any can be used. however. 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. The general solution of the problem gives somewhat complicated expressions for Ü and Ý. Here I present a reasonably simple criterion that allows one to generate all the integer problems with integer solutions. to be integers. which I would not have predicted. The values of Ü and Ý are integers if and only if   ½ divides both ´ · ½µ´ · µ and ´ · ½µ´ · µ. I am indebted to S. I find it particularly striking that and only enter via the sum · . SINGMASTER times you.

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

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

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

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

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

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

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

then ½ we get à ´ ½   ½µ ¾ Æ ´ ¾   ½µ Æ· ´ ½   ¾µ¾   ½ . If ½ is odd. . We then have à ½ Æ ½´ ½ · ½µ ¾ Æ· ´ ½   ¾µ´ ½   ½µ ¾.. The majority of examples are APs with Ò ¿. ´½ µ Again. These occur when ½ ¼. Subtracting ¾Ã or ¾Ã   ½ from Æ yields Æ· . we find that this occurs when   ½ ­Ò .. we have à ½ Æ ¿ Æ· ½.230 D. ½ ½ . Then we have ­Ò ´Ò   ½µ .. so we can write ½ ·´   ½µ . when the upper bound for given in (13) is an exact integer. unless it happens that ½ «¾­Ò ­¾ .e. we have à ­ ´ ½   Ò · ½µ   ½ Æ Ã ´ ½ · ½µ   ­Ò à ´Ã · ½µ ¾  ½¼ ¾ and ¾ ¢ ½½   ¾¼ ¢ ¾ ¢ ¿ ¾ ¢ Most of the earlier writers excluded zero values and only considered positive solutions.e. we have ¾Ã   ½ solutions with zero values. Looking again at our example. Most of the examples in the literature have the in arithmetic progression (AP). Rearranging. which is trivially an AP. Tartaglia’s problem 139 is the one with two sizes of pearls and is (10. ½ «¾­Ò ­¾ by (7). A little thought shows that then   is the common difference of the progression. i. 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 . Æ· ½¼   ¾ ¢ ¾ . 20. i. Substituting (11) into (8) and summing gives us the number of solutions. ½ A few longer APs occur — I have seen examples with Ò . so we let Ã Æ  ½ ­  Ò à where [ ] is the greatest integer function.e. We can count the number Æ · of positive solutions by finding the number of solutions with a zero value. and when Ò ¼. when two zero values occur in the same solution. 90). From (15). All but one of these have ½ even. A few cases occur that are not APs (see below). i. in our example. In that case. ­Ò   Ò   ½: ´½ µ à A number of early cases have Ò ¾. Hence we normally get ¾Ã situations with a zero value. SINGMASTER Then (10) becomes ´½¾µ ´½¿µ ½  ½ ­Ò The upper bound may not be an integer. and then à ½ ¾ Æ ¾ Æ· ´ ½   ¾µ¾ . perhaps as intended by De Lagny.

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

25. 11. 40. one with fractions. though it is not as easy to see how to do this as in the Western version.6) # of Solutions (55. 20) (64. (20. 50) (11. 1300 1300s 1489 1489 1489 c.4) (70. 30. 4 given. 55) (55. 81) (55. 454. 56. 29. 6 given. 100) (45. 30. Sridhara finds a one-parameter solution set. 49) (3. At this point. SINGMASTER Solutions in Source 5 given. Bhaskara makes the following comment: “This. 752. 36) (78. 60) (117. 20) (10. 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. makes this more specific by saying they ‘sold a part in lots and disposed of the remainder at one for five panas. 901) (10. 1 given. 50) (30. 1 given. 90) Date 1202? 1202? 1202? c. 30) (18. 48. 1 given. 136) (3. 55) (16. 1 given. 1 given. 30) (12. all given. 1 given. 15) (31. 1 given. 2 given. 17. 32) (12. 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. 40. 40. 36) (30. 60) (10. in some way to reduce the solution set to two dimensions. 33. . 40) (10. 36) (27. 20. 119. 20. 1 given. 40) (18. 1) (7. 1 given. 30) (8. 1 given. 37) (27. 33) (20. 36) (25.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 1 given. 1 given. (20. 30. 35) (11.400) (25. 33) (10. ¬ .9) (100. 81) (10.552. 140) . 1 given. 170) (305. so we again have a three-dimensional solution set. 1 given. 68. 60) . 80) (10. 1 given. 603. 32. 1 given with fractions. 12. One can scale the set . .232 Version (10. 1) (100. 16) (16. 21) (225. 26) (20. 50) (10. . none given. 82) (17. 16) D. 196) (45. 55) (78. none given.

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

Math. Bîjaganita. 1805. pp. reprinted by Blanchard. 10. 214–223. with Notes and Observations. v. John. 4th ed. 1884. Paris. [5] Bonnycastle. p. 12. Lucas N. Paris.” [2] Ayyangar. 1910. and Other Places of Public Education. pp. including Sandig. Gordon Fraser Gallery. 122–126. 2nd ed. pp. 18.) Chap. A. Wiesbaden.. P. aka Bhâskara. 16: “De duobus hominibus boves ducentibus — Two men leading oxen. pp. designed for the Use of Schools. Bhaskara solves his example by taking ¾. 19. 76. . The first nine editions appeared “without any material alterations. Revised by A. Gaz.. Labosne. London. References [1] Alcuin. Braunschweig. “an entire revision of the work.. I examined the 7 Ø edition. 15. The Historical Roots of Elementary Mathematics. reprinted by Dover. 1824. 54. 15. A. 3rd ed.so   and ¬ ½½¼. Book I.. London. Translated by John Hadley.. 4392. Vogel. An Introduction to Algebra. 3200 and 16. 13.. 24.” J. London. 1923–24. Heath. “Nine Chapters on the Mathematical Art.” which “may be considered as a concise abridgment” of his two-volume Treatise on Algebra. Diophantos of Alexandria. Bijanganita. London.234 D. 1968. “A classical Indian puzzle-problem. Analysis similar to that done before shows that there are then ½ · ´ ¬ ½   Ò µ  ¬ solutions. 1813.. Johnson. 178–186. By the way.” c. 6. there are only a finite number of solutions. 106–115. 86–88. [4] Bhaskara. et al. Prob. [6] Bunt. Problems to Sharpen the Young. In: Henry Thomas Colebrooke. Gauthier-Villars.. Mar 1992. with Arithmetic and Mensuration from the Sanscrit of Brahmegupta and Bhascara. 1624. Krishnaswami. Soc. 2nd ed. Claude-Gaspar. Oxford University Press.. pp. annotated by David Singmaster and John Hadley. and the 13Ø edition. 170.. pp. (There have been several reprints. pp. 5th ed. giving total numbers of 3295.470 [4]. 1874. trans. Facsimile with transcription and commentary by John Wardroper. H. London. 3292. so . P. Vieweg. 1959.” In 1815. Lyon. 1879.C. See No. Rigaud.900 and of 3294. nos. 1612. Indian Math. Nunn et al. [3] Bachet. 1150. 1973.. Arithmetica. Problemes plaisants & delectables qui se font par les nombres. 1511. 1817. Prob. [9] The Demaundes Joyous. Translated into German by K. Algebra. Lyon. 150 B. 1964. 33. 242–244.(475). SINGMASTER For given and ¬ satisfying ´ ¼¼¼ µ and ´ ¼µ. L. Nos. Rigaud. 24. J. he produced a 10 Ø ed. PrenticeHall. John Murray. 134–136. [8] Diophantos. Wynken de Worde. Neun Bucher arithmetischer Technik. 1782. 102– 126. 5th ed. In: T. J. Prob. which may be the same as the 1815 edition.. 21. 1976. 1st ed. pp. [7] Chiu Chang Suan Ching.

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

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

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

. Now.WHO WINS Misère HEX? 239 define Ѵĵ to be the minimum (over all possible games of Misère Hex in which È plays strategy Ä) of the number of cells left uncovered at the end of the game. supposing that Ѵĵ ½. Eventually É will win the imaginary game because Ä Ê is a winning strategy. and draws a circle around the cell containing this move. 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. and draws a new circle around the move just played. Thus the real game can continue. This contradicts our assumption that È has a winning strategy. The position is still a win for É. Therefore it is possible for È to move. she plays instead into another empty cell (chosen arbitrarily). We are now ready to prove the theorem. because of the monotonicity property. however. Let Ô ¼ be È ’s first move. Player É begins by encircling Ô¼. Suppose that É is the first player. such a position contains a È -path. The contradiction will be to show that under this assumption É has a winning strategy. while in the real game. playing out ÄÊ in an imaginary game.) The proof is complicated. When this happens she has also won the real game. This relationship will be maintained throughout the game. By definition. She makes an arbitrary first move. 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. except that in the imaginary game the encircled cell is empty. when it is É’s turn to move there must be at least three empty cells in the imaginary game. and further modified by changing any subset of É’s stones into È ’s stones. and there must be at least one empty cell in the real game. (Note that Ѵĵ Ñ´Ä Ê µ. We shall argue by contradiction. suppose that É is the second player. that cell contains a É-stone. because none of these changes would interfere with the È -path. so there are at least two empty cells in the real game. Because Ñ´ÄÊ µ ½. When the strategy Ä Ê requires É to play in the encircled cell. Player É applies the following strategy. 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. The state of this game is exactly like that of the real game. From now on she applies strategy Ä Ê in what we shall call the imaginary game. We must show that Ѵĵ ¼. Consider a terminal position of a game that is a win for É. Suppose the position is modified by filling in any subset of the empty cells with É’s stones. erases the circle.) Similarly. We shall make use of the following monotonicity property of the game. when it is È ’s turn to move there must be at least two empty cells in the imaginary game. The basic idea resembles Nash’s proof that the first player has a winning strategy for Hex.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

4).6026 0.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6.91 32. ½ ´Òµ It gives the longest trajectory.171 4 52.48 18. Section 5] shows that in ½¼ ½½ trials one should only expect to find a largest ­ ´Òµ ¿¼.456.5927 0.6030 0. Maximal value of ­ ´Òµ in intervals ½¼ Ò ½¼ ·½.884.6020 0. More recently V. A comparison of (14.799 6 8.5857 0. Extensions of the data can be found in Leavens and Vermeulen [16].371.527 Random walks model ½ ´Òµ ­ ´Òµ 70 108 165 214 329 429 592 593 706 755 21.127 8 127.328.5967 0.1) with numerical data up to ½¼ ½½ is given in Table 1. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good.254 9 4.400.527 5 837.890.511 7 63.13 26.94 31.5818 0.815 10 13. where ­ ´Òµ ÐÓ Ò .728.68 Ö´Òµ 0. ½ ½¼. .6020 0.91 19.6091 Table 2.77 31.5741 0.266. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory.24 16. The large deviations extremal trajectory is indicated by a dotted line. and the large deviations extremal trajectory is indicated by a dotted line.38 41. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ­ ´Òµ ¿ ¾ .180.64 32.5841 0.769. Vyssotsky [23] has found much larger numbers with high values of ­ ´Òµ. an analysis in [15. The agreement with the stochastic model is quite good. 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. He found that Ò 12.194. He also found a number around ½¼½½¼ with ­ ´Òµ exceeding ¿ ¼ Ø´Òµ in intervals ½¼ Ò ½¼ ·½ up to ½¼½½ with the prediction of (14.527 has ½ ´Òµ ½¾ ½ and ­ ´Òµ ¿ ¾ ½ .68 24. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1. 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.

In this case. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch.000 Ò 13.65 5. so do not have any significant effect on the size of the tree.049.897 2.817.02 21.675 7 80.831 9 8. 4.790 1.362. see [15] for more explanation.819 1.631. Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk.528.46 2. if Ò ¾ or 8 ´ÑÓ µ.645 ­ ´Òµ 21. 5.06 4. C.84 19.) empirically like . In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ.59 23. (See Table 3 in Section 4. In fact. The set of inverse iterates associated to any Ò has a tree structure. using (15.65 7. LAGARIAS Ò 1 27 2 703 3 9.86 13. there are duplications.788 1. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼. these trees have the minimal and maximal number of leaves possible for depth .86 7. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree.671 5 704. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ.24 16.791 1.75 5. respectively. 15.804.560 1.663 4 77.511 6 6.391 8 319. ½ ½¼. .1). Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½.32 2.260 J. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿.099 1. In iterating the multivalued function Ì  ½. where ½ ´½ º½µ Ì  ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½.14 11.86 5.976 1. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ. or 7 ´ÑÓ µ.83 3.48 12.903 2.559 Random walks model Table 3.03 19.20 6.511 10 77.909 1. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth .10 20.566.05 19.

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

16. Distribution of 3x + 1 Trees Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees. 7. Ê´ µ counts the number of distinct edge-labeled ¿Ü · ½ trees of depth . Since the number of leaves is expected to grow like ´ ¿ µ . Theorem 3. Based on this data. respectively. 2. Here the pruned 3x+1 tree Ì ´ µ is the tree of inverse iterates of the ¿Ü · ½ function grown backward to depth . The branching process is a multi-type Galton–Watson process with six types labeled 1. 5.2]. This can be formalized as Conjecture C . they proposed that the normalized extreme quantities remain bounded. 4. the corresponding random walk . where ¼ represents the entering vertex being even and ½ represents the entering vertex being odd. (See Athreya and Ney [4] for a general discussion of such processes.262 J. Let Æ   ´ µ and Æ · ´ µ denote the minimal and maximal number of leaves. In this table. they studied the normalized extreme values ´½ º½µ ·´ µ ¿ Æ  ´ µ Ò  ´ µ ¿ Æ ·´ µ Table 3 gives data up to ¿¼. C.) The random walk arises from assigning the labels ¼ and ½ to the edges of the resulting tree. 8 pictured in Figure 4. LAGARIAS and as ½ the leaf probability distribution is of the form Ï ´ ¿ µ . where Ï is a random variable satisfying ´½ º µ and Û prob Ï Û ´Üµ Ü for ¼ ½ ´Üµ is strictly positive on ´¼ ½µ. 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. 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]. from a given root node . see [15.

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

748 0.986 2.249 0.054 2.405 2.751 1.33 1.77 2362.475 0.327 2.342 0.633 0.83 1771.310 2.488 0.246 0.750 0.390 2.232 2.255 2.898 1.32 17.47 996.557 1. .268 0.792 1.747 1.475 0.240 0.491 0.738 1.211 2.422 0.489 0.750 1.62 1328.78 2.57 42.501 0.03 177.255 0.300 0.911 1.131 2.67 0.38 236.482 0.305 0.758 1.344 2.409 0.500 1.490 0.270 2.748 1.401 0. C. 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 1.688 1.352 0.419 2.742 1.09 56.481 0.49 9.75 5599.76 23.490 0.302 0.21 5.026 2.782 1.741 1.782 1.68 31.633 0.762 1.252 0.507 0.273 0.292 2.534 0.898 1.451 0.486 0.076 2.869 1.746 1.562 0.307 0.50 315.688 1. Normalized extreme values for ¿Ü · ½ trees and for the branching process.12 74.802 1. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a fixed node Ò´ÑÓ ¿ ·½µ.562 0.375 0.738 1.118 2.778 1.360 2.289 0.99 13.784 1.490 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 4.238 1.923 1.83 99.481 0.481 0.802 1.36 3149.488 0.500 1.60 747.422 0.271 0.77 133.507 0.746 1.780 1.16 4.264 J.481 0.492 0.534 0.475 0.37 3.394 0.758 1.486 0.190 2.491 1.281 0.243 0.62 7.753 1.774 1.45 560.791 1.534 0.728 1.750 0.34 420.166 2.465 0.802 1.81 4199.375 2.669 1.259 0.489 0.

Comp. 1996. D. 85–89. R. Generalized Parcal Triangles. 1990. Density Bounds for the ¿Ü · ½ Problem I. 1972. New York. of QuFu Normal University. On Heights in the Collatz ¿Ò · ½ Problem. C. The ´¿Ü · ½µ ¾ Problem: A Statistical Approach. Lagarias. The ¿Ü · ½ Problem: a Quasi-Cellular Automaton. Comp. Branching Processes. Collatz. pp. Weiss. Providence. pp. Math. B. 12. and G. TreeSearch Method. 2 (1992). 42. Monthly 82 (1985). Korec. pp. Pseudorandomness and Cryptographic Applications. Unpredictable Iterations. C. 25 (1977). No. Proc.. Density Bounds for the ¿Ü · ½ Problem II. Luby.I. Annals Appl. on Stochastic Processes. Math. Comp. 11 (1992). 55 (1985). Prob. Natural Science Edition. Amer. Rawsthorne. [17] M. pp. Iteration of the Number Theoretic Function ´¾Òµ ½µ ¿Ò · ¾. Springer-Verlag. Conway. 349–360. pp. 9–11. (1976). C. pp. pp. Discrete Math. On the Distribution of ¿Ü · ½ Trees. pp. E. The ¿Ü · ½ Problem and Its Generalizations. Lagarias and A. (1995). Leavens and M. [6] L. Tartani. [9] M. Muller. Lagarias. Proc.. . 64 (1995). Mitteilungen der Math. Univ. Proc. Lagarias. E. Applegate and J. J. Hamburg 12 (1991). Appl. 44 (1994). 3rd Intl. 58 (1985). Maximum Excursion and Stopping Time Record Holders for the ¿Ü · ½ Problem: Computational Results. in: Cryptology and Computational Number Theory. 427–438. [16] G. C. AMS. Math. 3–23. Pomerance. [12] I. pp. Mathematical Games. and Cellular Automata. and R.. (June). 64 (1995). [20] D. Oliveira e Silva. Math. Math. Math. pp. Pseudorandom Number Generators in Cryptography and Number Theory. Advances in Math. [15] J. Complex Systems 1 (1987). On the origin of the ´¿Ò · ½µ Problem (in Chinese). Math. Applegate and J. pp. Conf. NJ. Colorado. H. Merlini. Princeton University Press: Princeton. Appl. Slovaca. [11] L. Athreya and P. 4. [18] H. 24. Ò¸ ´¾Ò · [8] C. Ney. Garner. 1972 Number Theory Conference.. [19] T. [10] Martin Gardner. ed. [3] D. of Colorado. Goles. Vichniac. pp. 57–64. pp. 3. 1972. Ges. 42–45. 49–52. C. 79–99.3X + 1 FUNCTION ITERATES 265 References [1] D. 115–143. Computers Math. A. Applegate and J. Cloney. Symp. [13] J. 231–251. Lagarias. 1990. [14] J. Vol. pp. 172–176. E. The ¿Ü · ½ Problem: Two Stochastic Models. [7] J. Imitation of an Iteration. Physics and Geometry-Locarno. C. (1999). pp. Y. [5] T. [2] D. 229–261. 114–118. C. Krasikov Inequalities. R. pp. Boulder. Experimental Math. 411–426. Muriel. C. J. Lagarias. The ¿Ü · ½ Problem. 68. [4] K. Everett. Vermeulen. Feix. Mag. 371– 384. ¿Ü · ½ Search Programs. 101–117. Scientific American 226 (1972). Das ¿Ò · ½ Problem. No. pp.

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

Sign up to vote on this title
UsefulNot useful

Master Your Semester with Scribd & The New York Times

Special offer for students: Only $4.99/month.

Master Your Semester with a Special Offer from Scribd & The New York Times

Cancel anytime.