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



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



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


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



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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

seen as an Army Sergeant Major. You. Light Sun and Earth. Bang!. the Earth Sun’s System. Friday night At a minute to twelve Eve spin. right? A week = 7 days corresponds to 14 Billion Years 1daycorresponds to2BillionYears(USA) 1 minute corresponds to 2 Million Years 1 Millisecond is 23 Years 23 . and a bit more ½¼  meters = Nucleus of a human cell ½¼ ½ meters = Atom’s nucleus Time The Creator. Adam delve In the last millisecond. First thing on Monday. form up. Cell’s girth Our bodies.Three Limericks — On Space. Time. and Speed Tim Rowett Space Seven steps each ten million to one Describe the whole space dimension The Atom. 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. barks out his orders for the week.

ROWETT Speed A child cycles ‘round the schoolyard Which lies on the Earth turning hard The Earth rounds the Sun As Sol does “the ton” And our Galaxy flies — Gee! I’m tired 7 mph — child cyclist 700mphEarth’ssurface(NorthAfrica) 70.24 T.roughaveragespeedofEarth 700.4 million mph Galaxy’s speed through the debris of the Big Bang .000mph.000 mph turning speed of Galaxy 1.

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

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

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

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

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

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

. Genesis 32 : 15. . Ì Ê Ë among the . . . . charity shall 33 . Ê of good men. shoulder Aaron . they were . in Revelation 3 : 11. . . . . . ÌÁÊ Ë shall be . . . . . Ï Î Ë. Daniel 3 : 21. . . . . . I have . . . . À Ï Agag in pieces . . John 18 : 18. use intermediate words found in Matthew 24 : 35. violence of the . Ë Î Ê the wicked . . . Gather up thy . . . . . . . Matthew 6 : 28. hazarded their . For example.) Puzzle #4. . from the . for example. ( 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. again. . . forth. Ezra 34 : 25. Ï Ê Ë out of . remnant shall be . as a . . Ruth 3 : 16. . hospitality. . . . which is . Psalm 84 : 6. . . and . . . . . . and to the . . . . . . Ezra 7 : 24. and which . . At his .) Puzzle #6. . . . . . . . (But ‘writ’ is not a Biblical word. till seven Ë Î Æ. And your . and . hewed stones. . Ecclesiastes 12 : 11. Ë of the sea. . Psalm 145 : 15. a . they married . . . . Therefore . . . . Suitable intermediate words can be found. . And the soldiers’ . Ë Ï sackcloth . . . 2 Corinthians 11 : 19. seventh hour the . Acts 2 : 45. times . . and they . ( Matthew 18 : 22 ) ( Matthew 13 : 49 ) ( John 4 : 52 ) ( Numbers 33 : 54 ) ( Deuteronomy 29 : 11 ) ( 1 Samuel 15 : 33 ) ( Job 16 : 15 ) ( 1 Kings 7 : 9 ) ( Romans 9 : 27 ) ( Leviticus 9 : 21 ) ( Acts 27 : 41 ) ( Jeremiah 10 : 17 ) ( Matthew 13 : 25 ) ( Ezekiel 24 : 23 ) ( Matthew 18 : 21 ) . . . . Ë Î : For he will . . but none are strictly increasing or decreasing. Ï Î for a wave . . . . . Ï Ê ye shall give . . . . . . . . . Ë. Job 40 : 17. me. Î Ê left him.BIBLICAL LADDERS . ÌÁÅ Ë? Jesus saith . . And Samuel . . À Ï Ê of thy wood . . looked in the . . . Numbers 1 : 2. Micah 1 : 8. .) Puzzle #5. . and sowed . . . Ê the multitude . Ê. Ë for the name . they were . . . . the mighty . Matthew 25 : 33. (See also Genesis 27 : 44. . . . Ë Ï with saws. .

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

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

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

66. 1:01:01. 34. 4. (3) Find the longest time span without a palindromic time. 23:55:32. 36. (1) Find the closest such times. 32. 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. 6. y Problem 7: The figure shown here is the solution to the problem of dividing the figure into four identical shapes. 30. 54.) These curious combinations occur 660 times a day. 50. (Ignore the colons. 64. etc. 46. 42. For example. 62. 44. 52. . YOSHIGAHARA Problem 4: Arrange the following five pieces to make the shape of a star. 40. 13:22:31.38 N. Problem 5: Calculate the expansion of the following 26 terms. Problem 6: Which two numbers come at the end of this sequence? 2. 2:41:42. 3:59:53. (2) Find the two palindromes whose difference is closest to 12 hours. 56. 60. x.

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

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

Number Play, Calculators, and Card Tricks: Mathemagical Black Holes
Michael W. Ecker

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

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

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



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:



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.



(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



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.

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

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

55. That is. 5. 1. ECKER Example: Start with 8028. including 1 and the number itself. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò   call it Ä — as Ò increases. Repeat with 8532 to calculate 8532–2358 = 6174. and 15.618 .48 M. 34. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. The ratios seem to be converging to a number around 1. More formally. In fact. ½ . but it is also that much more strange at the same time. Divide both sides of the equation ´Òµ ½µ · ´Ò   ¾µ by ´Ò   ½µ. Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . 3.6 or so. the smallest is 0288. 3. 5. but your very equilibrium. but you will reach 6174. Your own example may take more steps. To get a quick and dirty feel for why phi arises. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. and ´Òµ preceding two elements. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ ľ   Ä   ½ ¼ . and these digits sum to 15. This one is a bit more tedious. Iterate until a number repeats. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors. but without using a program. Now take the sum of the digits of these divisors. Its divisors are 1. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. 2. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. The largest permutation is 8820. phi. but fewer authors use that approach. Its value is roughly 1. ´½µ ´Ò   ½µ · ´Ò   ¾µ for integers Ò ¾.). and the difference is 8532. 8. The first two numbers numbers: 1. 13. etc. This one may defy not only your ability to explain it. 89. The black-hole number this time is 15. W. 21. Had we used the ratios ´Òµ ´Ò · ½µ instead. we would have obtained the reciprocal of phi as the limit.

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

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

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

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

Divide the number by 3. 3.5 nautical miles from Los Angeles to Honolulu. How long does it take? E5. 2 and 1. HESS up and travels due North. 8. My regular racquetball opponent has a license plate whose three-digit part has the following property. 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. 63 9 38 33 32 12 88 25 16 18 15 Ü 5 E3. 0). 4. Find the rule for combining numbers as shown below (e. 25 and 9 combine to give 16) and use the rule to determine Ü. 4. 8. 2.54 R. 0. You have containers that hold 15 pints. The 15pint container starts out full. 5. reverse the digits of the result. What is the number and what is the next greater number (possibly with more than three digits) having this property? . and 6 pints. 5. Calculate the expansion of this 26 term expression: ´Ü   µ´Ü   µ´Ü   µ´Ü   µ ´Ü   ݵ´Ü   Þ µ . 10 pints. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu. 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. 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. E6. After going 10 miles in a straight line he finds himself back at camp. 7. and the other two start out empty: (15. E9. 6. g. subtract 1 and you produce the original number. Find a nine-digit number made up of 1. Where on earth could he be? E2. 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. I. It is approximately 2244. 6. 7. 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. Find the most efficient solution..

. The Bridge. 8. Find nine single-digit numbers other than (1. 2. . Can all four men cross the bridge? If so. and 10 minutes. 7. and the bridge will hold up to three men at a time. E11. A pair must walk together at the rate of the slower man’s pace. 9. 4. and 9 minutes and they must cross in 31 minutes.. They all start on the same side and have 17 minutes to get across. A knight is placed on an infinite checkerboard. Place the numbers from 1 to 15 in the 3 ¢ 5 array so that each column has the same sum and each row has the same sum. how? Try these other problems.PUZZLES FROM AROUND THE WORLD 55 E10.880. (b) There are seven men with crossing times of 1. 3. 8. A maximum of two people can cross at one time. (a) There are six men with crossing times of 1. B takes 2 minutes. Four men must cross a bridge.. C takes 5 minutes. and they must cross in 25 minutes. and 9) with a sum of 45 and a product of 9 362. 6. 6. and D takes 10 minutes. E13. . If it cannot move to a square previously visited. 2. 3. and they need their one flashlight to guide them on any crossing. E14. Each man walks at a different speed: A takes 1 minute to cross. Divide the figure below into (a) 4 congruent pieces and (b) 3 congruent pieces. It is night. how can you make it unable to move in as few moves as possible? E12.

You have a 37-pint container full of a refreshing drink. 1260. x. M4. What is the probability that they form (a) an acute triangle. 450. N thirsty customers arrive. Suppose a clock’s second hand is exactly on a second mark and exactly 18 second marks ahead of the hour hand. Despite your lack of power you took 6 tricks. Find a simple continuous function to generate the series and compute the surprise answer for Ü. The lace burns irregularly like a fuse and takes 60 minutes to burn from end to end. . HESS E15. What is the minimum time interval you can measure? Medium Problems M1. You are given some matches. You are allowed to draw a closed path on the surface of each of the potatoes shown below. one having an 11-pint container and another having a 2Npint container. hearts.56 R. a shoelace. 280. Potato Curves. (b) N = 5? M2. 180. 60. Produce the other hands and a line of play that allows this to occur. Three points have been chosen randomly from the vertices of a cube. How will you most efficiently measure out 1 pint of drink for each customer to drink in turn and end up with N pints in the 11-pint container and 37-2N pints in the 37-pint container if (a) N = 3. Shoelace Clock. What time is it? E17. 744. 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. From England comes the series 35. 45. In clubs you held 5432. (b) a right triangle? M3. and diamonds. Can you draw the two paths so that they are identical to each other in three-dimensional space? E16. making a 4-club contract. I. 120. You were playing bridge as South and held 432 in spades. and a pair of scissors.

and [(13). You wish to cross from the south to the north side and have only two thin planks of length L and width 1 ft to help you get across. How many ways can four points be arranged in the plane so that the six distances between pairs of points take on only two different values? M6. (34)]. 26 packages labeled A to Z are known to each weigh whole numbers of pounds in the range 1 to 26. From the USSR we are asked to simplify Ü M7. (b) Now do it with three weights. (4)] are legal combinations while [(1). (34)] = [(21). Ô ¾ · ¿½¼¿ · Õ ¿ Ô ¾   ¿½¼¿ . Show that no more than two of its vertices lie on grid points. Music on the planet Alpha Lyra IV consists of only the notes A and B. Legal combinations allow the buttons to be pushed in specific order either singly or in pairs without pushing any button more than once. [(1). The river shown below is 10 feet wide and has a jog in it. M11. Õ ¿ Ô such that ¾Ô + Ô¾ is also prime. [(134)]. Find the smallest prime number that contains each digit from 1 to 9 at least once. A regular pentagon is drawn on ordinary graph paper. Prove there are no M8. (14)] are not. Also. it never includes three repetitions of any sequence nor does the repetition BB ever occur. (a) Determine the weight of each package with a two-pan balance and four weights of your own design. (14)]. Many crypto doorknob locks use doors with five buttons numbered from 1 to 5. Find all primes others. M10. how many legal combinations would there be? M13. and [(2).PUZZLES FROM AROUND THE WORLD 57 M5. (13). What is the longest Lyran musical composition? M12. What is the least value for L that allows a successful plan for crossing the river? M9. Thus [(12). (3)]. (a) How many legal combinations are there? (b) If a sixth button were added. .

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

Alice was almost in tears. “Don’t be a fool. Isn’t it differential equations. or something I’ve never learned?” Humpty Dumpty opened one eye. that was — I got a better mark in arithmetic than any of the others who qualified. I’ll tell you now. “Anyone ought to be able to do it who is able to count on five fingers. It was a number equal to four times the maximum obtainable in one subject. child. 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. I’ll put it another way. when I tell you that no two of us obtained the same assortment of marks — a thing which was only just possible — you’ll be well on the way to the answer. “Good at sums. Well. child? I don’t very much.” What was Humpty Dumpty’s mark in Arithmetic? Hard Problems H1. many years ago.” “How many subjects were there?” said Alice. “I’m coming to that. “So I am.” “But you haven’t told me .” he said crossly. “How many of us were there? Well.” said Alice.” said Alice. oh certainly. “do any of it.” began Alice. But what has that got to do with liking them? When I qualified as a Good Egg — many. How many ways can you find to do this? (Nineteen solutions are known.) . multiplied by the number of subjects (I’ve told you about that already).” said Humpty Dumpty. “You don’t like arithmetic. Similar Triangles. gives a product equal to half the number of marks obtained by each Good Egg. Not that that’s saying a lot. And I ought to mention that in no two subjects did I get the same mark. Humpty Dumpty.” He composed himself for a nap. “Ah!” said Humpty Dumpty.PUZZLES FROM AROUND THE WORLD 59 M20. “But I thought you were good at sums. The figure following this problem shows a 30–60–90 triangle divided into four triangles of the same shape.” she said. And we all just managed to qualify. and that is also true of the other Good Eggs who qualified. The number of other Good Eggs who qualified when I did. interested. “I must think. “I can’t.” “But how many .” said Humpty Dumpty.” said Humpty Dumpty. But to make it as easy as I can for you. The number of subjects was one third of the number of marks obtainable in any one subject. And now you can find out all you want to know.

how tall is the building? (Use 20.) Assume that it takes Sir George 1 minute to pound a stake and that he walks at a constant speed between stakes.902. and be back at your starting point in 24 hours. HESS H2. “I grant you the land you can walk around in a day. on the exact top is resting on the horizontal plane. “To reward you for killing the dragon. P.” she continued. Where is the black dot when the ball returns to its initial resting place? . A billiard ball with a small black dot.60 R. How many stakes should he take with him to get as much land as possible? H3.” the Queen said to Sir George. Given that the width of the building is 125 ft and the rubber band must stretch an extra 10 cm to accommodate the building. It rolls without slipping or twisting so that its contact point with the plane follows a horizontal circle of radius equal to that of the ball. “Take some of these stakes with you.) H5. Imagine a rubber band stretched around the world and over a building as shown below.” She pointed to a pile of wooden stakes.” (The Queen had read a little mathematics.851 ft for the radius of the earth. An irrational punch centered on point P in the plane removes all points from the plane that are an irrational distance from P. What is the least number of irrational punches needed to eliminate all points of the plane? H4. “Pound them into the ground along your way. I. All the land in the convex hull of your stakes will be yours.

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

The drawer starts with 16 socks each Monday morning (eight blue. . An approximation to is 355/113.0107 . A number may be approximated by a /b where a and b are integers. ) = 0. (b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes? . six black.62 R. Define the goodness of fit of a /b to x as g (a. where he (1) picks out three socks at random. (a) The figure shows two Pythagorean triangles with a common side where three of the five side-lengths are prime numbers. Find other such examples. HESS H12.0107 0. Minimum Cutting Length. and proceeds to explain: “I normally ask guests to determine the ages of my three children. My uncle’s ritual for dressing each morning except Sunday includes a trip to the sock drawer.” Your response is “No need to tell me more. Since Smith missed the problem tonight and Jones missed it at the party two years ago.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. their ages are ” H14. where e = H16. (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. then (2) wears any matching pair and returns the odd sock to the drawer or (3) returns the three socks to the drawer if he has no matching pair and repeats steps (1) and (3) until he completes step (2). I.0107. (a) Find the minimum a and b so that g (a. with g (355. b) Find the minimum c and d so that g (c. b. e ) 2.71828 . d. x) = a bx . b. into five parts of equal area? (c) an equilateral triangle into five parts of equal area? H15.a . given the sum and product of their ages. two brown) and ends up with four socks each Saturday evening. The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem. I’ll let you off the hook. 113. ) 0.

The common axis of these layers is the planet’s rotational axis. The outermost of these layers is nearly pure krypton. 18 and 20. Inc. the little that is known about the krypton layer was received via sub-etherial communication from a Venerian pioneer immediately prior to its demise at the hands (some would say wattles) of a frumious snatcherband. (A spandral is the Lyran unit of length. The planet Alpha Lyra IV is an oblate spheroid. The inhabitants of Lyra III recognize special years when their age is of the form Ô¾Õ. finally one becomes a sage until death. 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. (A copod corresponds roughly to something between a sibling and a rootshoot. On Lyra III one is a student until reaching a special year immediately following a special year. Cosmic. and the next inner layer is an anhydrous fromage. The first few such special years are 12.331 to 1. It (the Venerians are sexless) also reported that the straight line distance between it and its copod was 120 spandrals. where Ô and Õ are different prime numbers.PUZZLES FROM AROUND THE WORLD 63 H17. Its axis of rotation coincides with the spheroid’s small axis.) The reporting Venerian was mildly comforted because the distance to its copod was the minimum possible distance between the two. It’s internal structure. being made of a number of coaxial right circular cylindrical layers. which occurs in a special year that is the fourth in a row of consecutive special years. just as one observes for Sol III. each of homogeneous composition. 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. one then becomes a master until reaching a year that is the third in a row of consecutive special years.. By nature the Venerians can only survive at the krypton/fromage boundary and by tragic mistake the two copods had landed on disconnected branches of . (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. is contemplating mining the outer layer. is unique.) Unfortunately. however.

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

0). But this is not right because 18 is not the difference of 38 and 16. 10. E4. the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. 4. 6. ¼. E9. (11. E10. 0). Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. Instead. 8. At first glance.729. 10. The license number is 741. and Ü ½ . 2. 6). (1. Anywhere within 10 miles of the north pole. 6. 5.425. 2). (7. E7. . 0. the next greater number is 7. 0). 4. so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. 0). ½ ¼½ or ¼ and ½ ¼½. 6).PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 0. 0). ¿   ½ ¼¼¼   are the smallest values satisfying the above equations. 9). (5. 6). The trick here is that 1 knot = 1 nautical mile per hour. One of the terms is ´Ü   ܵ.654. 2). 6. 6. (3. 5). (7.741. The number is 381. it appears that the rule might be subtraction. (1. E3. (1. (0. E6. 6). ؾ ¾¾ hours. 0). 2. so E8. 7. 2). (3. 15 moves: (15. 0. (3. 8. In each case the numbers combine by summing the products of their digits. 4. 10. . 6). 4. (9. E2. 9. 0. (11. 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 . 2. (9. ¾ gives Ø E5. Thus Ü ½½. 6). Thus and Ò ¾½¼ transfers is the minimum. (5. (3.

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

0). 1). 11. 5). (37. E17. (25. (30. (33. In 3. (11. 0). 6. (21. (31. (18. 5. 6. light the other end of D. 6). 0) (b) 40 moves. (30. (29. (24. 10). (30. 11. 134. 2. 5. 0. 4). 6). 0). (29. 10). When C is consumed. 4). 11. 0. 3. 0). (36. 4. 3. (29. 0. 0. (25. 6. 0. 11. 0. producing pieces A and B. 0). 1). This is the shortest time interval that can be measured. 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. (33. (23. (28. 6. 10). 0). 0). 11. (27. (30. 10). 10). (23. 0). (17. 125. and the time is 8:24. as seen from the list below. 1). 4). (19. 0). (26. (23.75 minutes. (31. (25. 5. 1. 0). producing pieces C and D. 1). 124. 10). (36. M2. 1. 3). Solutions to Medium Problems M1. 6). 0). 0. 0. 11. 11. 0. Cut B at that corresponding point. 10). 10. 148. 0). See the figure. (22. 0. 11. 1). 127. 0. 6). 128. 0. 11. (a) 32 moves. noting the point that burns last.PUZZLES FROM AROUND THE WORLD 67 E16. 0). 10). (34. 0. 0). 1. 0). 6). 11. (31. 4. (36. 0. 4. 9. (23. (26. 0). 3. 0. 6). 2. Cut the lace in half. 9. 11. 10). 0. 126. 0). 0). (28. 0. (29. (29. (35. 0. (29. 0. the probability of a right triangle is ½ ¾½ or . ½¿ ½¿ ½ Ê 123. (20. 1). There are 7 ¢ 6/2 = 21 choices using vertex 1. If the hour hand is exactly on a second mark then the second hand will always be on the 12. 1). 3. 0). 10). 10). 167. 11. 1. Burn A from both ends. (25. 0). (24. 6). (37. 10). 0). (31. (19. (23. 10). 6). 0. 0. 0). 6). 11. 0). 2). 156. 0. 178. start burning C from both ends and D from one end. 11. (34. 5. 1. 158. (20. 4. 6. (30. 168. 0). 3). (25. 1. 0). 138. 1. 6. (23. 6). 11. (30. (19. 10). (26. . (35. (25. 136. (24. 146. (19. 145. 0). 1. 2). 11. 0). 4). D will finish burning. (18. 6). (31. The probability of an acute triangle is ¿ ¾½ or ½ . At the same time. (22. 0). (22. (35. (24. 0). 147. (21. 11. (24. (24. (17. 10. (25. (18. 5). 1. (18.

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

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

P4S. 2P2S. 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. Shown below. 2S. M15. and 3P. 2PS. 2P. 4S. PS. Each number is the length of the shorter side of the rectangle. . 3S. M14.70 R. P. 1. 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 M13.789.465. A nine-rectangle solution is shown below.123. I. P3S. 6S. P2S. 5S.

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

Solutions to Hard Problems H1. 13. 10. The scores for the Good Eggs are (15. M20. 6). Shown below. . 11. 9). There are 7 Good Eggs. 14. (15. The number in each triangle indicates its area relative to the other triangles in the figure. 10). 8). HESS (b) The two points are along the diagonal near P and its opposite Ô point on the ½ ¢ ½ face that includes A. (15. 7). 11. 13. 12.0119 . (15. 14. 14. 14. 12. Humpty Dumpty got a 10 in arithmetic. 13. 13. (15. I. 9) and (14.72 R. 12. 11. 8). 12. 14. 10. 12. The points are ´ ¿   ½µ ¾ of the way along each diagonal for a distance of 3. There are 5 subjects with 15 marks possible in each. 12. (15. 11.

the point P will have coordinates È Ô Ô  Ö × Ò ¾ Ö´ Ó×   ½µ ¾ Ö´ Ó× · ½µ ¾.300. Three equally spaced punches in a straight line do the job if the square of the spacing is irrational. and stretch in the band due to the building. Then the ¾Ø Ò Û Ö· Ö ÖØ Ò Û ¾Ö · ¾ · ¾ ¿ ½ ¾Ö Ø Ò · Û   ¾Ö´ · µ These equations can be iterated to produce ft. 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 . For a full rotation of the ball on the circle the point P will execute ¾ rotations about the axis. H5. Ö earths radius. 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.1968 Ú ¾ H3. as shown below. 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. height of the building. 1). Consider the ball rotating about the axis AC without slipping. If Sir George travels at speed Ú then × Ú ´½ ¼   Òµ Ò where 1440 is the number of minutes in a day. TheÔ cone. Point P will return to the top again as Ô well. Its area is Ò× ¾. 0. H4. The distance from B to the distance BC. BCB rolls along the plane. Imagine a regular polygon of Ò sides after Sir George’s trip. with 159. Relative to the center point of the ball. For Ö ½È ´ ¼ ¿ ¾  ¼ ¾ ½¼ ½ ¼ ¼ ¼ ½ µ its initial position was (0. Let Û width of the building. . where .PUZZLES FROM AROUND THE WORLD 73 H2.

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

Its form will take on the shape of four quarter circles of radius Ö at the corners as shown in the figure. There are an infinite number of choices of Ñ and Ò followed by the appropriate to make an infinite density of ´ µ’s in the ¼ ¾   ¼ interval. (g) H7. The perimeter of (c) Define ¡ ¡ ¡ Ô ¡ ¡ ¡ where there are Ò square roots. The area of such a shape is ½·´   µÖ¾ . (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. H8. or ½¾¼Æ. (a) Imagine an expanding soap bubble in the unit square. For the right . Pick and as any digits other than 0. Clearly . which contains ½ . ´ µ can be put in the interval ¼ ¾   ¼ .PUZZLES FROM AROUND THE WORLD 75 (e) (f) · ¼Æ ¼Æ. .

Now consider two such circles of different sizes as shown. 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. The circles need to be packed as shown below. Raise S and T appropriately above the . The smallest rectangle found to date containing 329 circles has 13 circles on each side of 101 sets of 3 and measures 2 ¢ 163. HESS the shape is È Ô ½ ´¾ · ¾ Ö ·   Ö. H9.9973967 . Note that the ratio È is 0. H10. Each has a radius ½.25 for the entire square or a circle inscribed in µ the square. The ratio È takes on a maximum when Ö ¼¾ ¼ . The point S may be raised above the plane to give unit altitudes if the inscribing circle has radius ½. The maximum ratio È Ö ¼¾ ¼ . (b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved. For Ò ½ there is just enough room for 329 circles. I. They define the four tangent lines drawn to produce ABCD as shown.76 R.

This leads to ¾ È ´ ¼   ¼µ Ô¾Ô Õ¾ . which leads to Ô ¼ ½ ¿¿ ¼ .2801 1.3586 Pairing probability 0. ¾ H12. ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ . 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. ·¾ Ô ´½ ½ Õ µ È ´¼   ¼µ .7849 . A computer program was written to calculate the probabilities shown in the table below. Clearly ST is not perpendicular to the plane. and let Õ ½   Ô.2270 1. ¾ ¾ Õ´ È ´¿¼   ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ .2069 1.3074 1. which leads to Ô ¼ ½½ ¿ ¾ . È ´¼   ¼µ È ´ ¼   ¿¼µ when Ô¿   Ô¾   ¾Ô ½. Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1.PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit. H11. È ´ ¼   ¼µ ÔÈ ´ ¼   ¿¼µ · ÕÈ ´¿¼   ¼µ.8286 0. È ´¼   ¼µ È ´¿¼   ½ µ when Ô ¾   Ô ¿. È ´ ¼   ¿¼µ Ô · ÕÈ ´ ¼ ¼µ.7892 0.2511 1. · È ´ ¼   ¿¼µ Õ È ´ ¼   ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . Let È ´   µ be the probability of the server winning the game when the server has points and the receiver has points. È ´¿¼   ¼µ ÔÈ ´ ¼   ¼µ. At Ò games to Ò.7805 0 .8163 0. Let Ô the probability of the server winning a point.8027 0. Ô  Õ At a score of Ò games to Ò (Ò ¼ to 5).

093. HESS H13. (a) Length ½ ¿ ¾½ ½ ¼ as shown.613. . a=10. . Solutions occur for (b) ¾ ½. (b) 605 years. H14. · ½.444 years. ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . ´ prime.78 R. (b) Length ¾ ¼¾½½¾ ¿ as shown.042. There is only one set of ages that accomplishies this: (5. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½.512. 16). 7. we get (a) (b) H16. ¾   ½µ ¾. 8. ¿ ½¾¼. . 12). ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. (d) Yes.641. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. (a) 45 years. 15). (2) the younger two children have different ages. H15.546. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô . H17. (c) Unknown. which must have been guessed by Jones two years ago. which must have been guessed by Smith. 6. ¼ ¼½¼ .321 is the first of 5 in a row. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago. (c) 17. ¼ ¼¼ ½ ¼. prime. The product and sum two years ago could be achieved by (2. ´ ¾   ½µ ¾. I. The product and sum could be achieved by (4. Using a multiprecision continued fraction program. · ½. ´ ¾ · ½µ ¾ prime. (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ . We look for cases where (1) the oldest child is under 20.

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

we expand and deform to a contour the figure. As trarily large it still encloses the same poles as .80 R. The only pole of Õ´Þ µ  ´Ò·½µÞ within is a double pole at Þ ¼. and by Cauchy’s Theorem we have ½ ¾ Thus ÞÕ´Þ µ  ´Ò·½µÞ Þ Þ ¾ Õ´Þ µ  ´Ò·½µÞ Þ ¼ ¾  ´¾Ò · ¿ µ ´Òµ   ¾½ ÞÕ´Þ µ ´½ · Þ µÒ·½ To evaluate ´Òµ. HESS (2) Since ´  ÒµÞ is an entire function in the complex plane. (3) The summation over can be carried out explicitly to give ´Òµ ¾ ½ ¾Ò· · ¿ ¾ Þ Þ ½   ´½ · Þ µ  Þ ´½ · Þ µÒ·½ (4) Define Õ´Þ µ ½ ½   ´½ · Þ µ  Þ . Thus Þ ¾ Õ´Þ µ  ´Ò·½µÞ is regular within . which still avoids enclosing the poles of Õ´Þ µ. 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 . 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. I.

(5) On Þ Ê for sufficiently large Ê.8790560 20.6406814 3.4614892 13.  ¾ × ´Ý · µ  Ü Ý .0888430 2. 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. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to infinity. The table below gives numerical values of the first few. Call the Ø such pole in the upper half-plane Þ Ü · Ý . ¼ for all . The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ.4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ   Þ µÕ´Þ µ ´½ · Þ µÒ·½. Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ.5012690 3. For larger define ´¾ ·½ ¾µ . A pole of Õ´Þ µ occurs Þ .1512074 45.0262969 3. Ü 1 2 3 4 5 6 7 (6) The contribution to 2.2916783 3.8221528 Ý 7. where Þ at ½ · Þ Ü · Ý.2238350 26. which reduces to ´½ · Ü · Ý µ´Ü   Ý µ ´ Ó×´Ò · ½µÝ   × Ò´Ò · ½µÝ µ where ´Ü¾ · ݾ µ ´Ò·½µÜ . × ÒÝ Ý we can look for poles in the upper half-plane only and reflect each one into the lower half-plane.5432385 32.PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ.8505482 39. Since the equation in Ý is even. Then Ý approximately equals   ½·ÐÒ´ µ . not including its pole at Þ ¼.6745053 3.

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

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

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

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

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

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

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

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

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

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

All edges of the pieces lie in one of three different directions. Solutions found by machine by Bert Buckley. found by John Conway in 1964. We have always drawn the hexiamond with one of these directions vertical. pieces placed symmetrically with respect to the two kinds of axis. For example. in Figure 14. Figure 15 shows eight solutions. or ½¿ . respectively 13.92 R. at angles of 60 degrees to one another. Conway showed that you can’t have more than 15. K. You can . It seems unlikely that anyone can beat 11. ¼Ô Á Notice that there are two different kinds of axes of symmetry. To deduce the others from the single diagram shown. but you may prefer to put one of them horizontal. rearrange pieces ¾¾ or ¿¿ and note that the hexagon formed by ¼½½ has the symmetries of a rectangle. each with 11 symmetrically placed pieces. GUY ½ ÒÜ ¼ Ó¿ 66-05-30 ½ÕÜ Ó¿ 66-08-15 Figure 13. Æ Figure 14. after which you can swap ¼¾ with ¿ . Some solutions come in large groups. rearrange pieces ¾¿¿.

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

94 R.489 B 15. GUY Figure 18.717 C 6675 D 7549 E 11. president of Kadon Enterprises and discovered that she markets a set of “iamonds” . are 130 195 533 193 377 2214 111. Figure 19. Their numbers. Ô Ü (3-symmetry). classified according to the position of the Chevron. for those who prefer their puzzles to have just one answer. 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.518 A 75. K. there are no fewer than 40 relative positions of the Chevron and Hexagon that determine such a unique solution: Ð × Ü ½ ¼ Ð Ø ¿ ¿ ¼ Ñ Ø Ü ¿ Ô Ô Ø ½ Ø Ü Ü ¾ ¾ ¼ Ô Õ ¼ Õ Ü Ö Ø ½ Ø Ø ¼ Ø Ü ½ ½ ¾ × Ø Ü Ü How long will it take you to find them all without peeking at Marc’s database [5]? After writing a first draft of this article. I wrote to Kate Jones.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. by the Hexagon’s position Ø Ù Total 264 1094 124. On the contrary. Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to find all the solutions.

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

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

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



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,



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.



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.



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!

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

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

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

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

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

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

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

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

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

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

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

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

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

Self-Designing Tetraflexagons
Robert E. Neale

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

Figure 1. Square window.

Figure 2. Cross plus-slit.




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.



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.



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.



Figure 13.

Figure 14.

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

Figure 15.

the other two faces are quite difficult to discover. being black and white checkerboard patterns. involving changing the flat flexagon into a three-dimensional ring and back again. and two slits in the center that intersect in the shape of a plus sign (see Figure 2). exactly as indicated in Figure 16. while four of the faces are easy to find. The other two faces are solid black and solid white. very quick. There is a 6 cell at the bottom of each box (see Figure 16. and the second flattens it again. This puzzle is interesting for two reasons: It is tricky to construct.122 R. one corner in each hand. So reach underneath and hold the two free corners of the 1 cells. Orient them just as indicated. . The other cells come together to form two boxes open at the top. moves. When you pull them down and away from each other. They indicate that you are to make the base 3-D by pushing two opposite corners down and away from you. and. You will make two. (Note: The trick fold for constructing the flexagon can be done in a slightly different way that has the four faces not continuous.) 3-D Tricky Fold In order to follow the instructions and understand the diagrams. they are turned over so you see the 5 cells. E. The numbers in parentheses are on the back side of the base. Note the two arrows in Figure 16. This base is manipulated in a third way. The four faces are continuous and identical. NEALE Cross Plus-Slit The window model inspired me to think of other ways to form a flexagon without having to attach the ends to each other. the 3-D Tricky Way. number the base from one to six. The first renders the base 3-D. One result was a puzzle made from a “cross” — a square (of 16 squares) with the four corner squares removed.

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

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

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

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

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

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

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

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

Often it is easier to come up with and develop a new puzzle idea than to give it a good and catchy name. I am frequently asked to explain the thought process involved in coming up with a new puzzle. Unfortunately. 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. Thanks also to Tom Rodgers for his support of my work. Many thanks must go to Martin Gardner as an inspiration to the millions enlightened by his myriad works. It is my suspicion that the subconscious mind is constantly at work attempting to fit pieces of countless puzzles together. A couple of examples of ideas that have “popped” out of my mind at various times are explained and illustrated below.” and for presenting me with the opportunity to meet Martin Gardner. I really can’t answer.132 R. sometimes it succeeds! If you devote your mind to something. 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. either you become good at it or you are devoting your mind to the wrong thing. This puzzle idea came to me as I was driving to San Francisco to sell my puzzles at Fisherman’s Wharf. in 1971 or 1972. IRBY and easily formed into whatever permanent shapes may be necessary to present a concept. The Bermuda Triangle Puzzle .

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

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

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

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

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

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

perhaps depending on the temperature. and the height through which the sand has fallen. Hopelessly confused? Martin Gardner?! That is something that surely doesn’t happen very often. Naval Ordnance Laboratory at Silver Spring. the acceleration of gravity. I am sure that your explanation was correct. On September 6th.S. 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.S. all you have to do is shake your own hand. I have not yet seen the toy. or perhaps still another one. Maryland: Another solution to the hourglass science teaser is that the momentum carried by the falling sand causes the hourglass plus sand to weigh more than its static weight by an amount Ô¾ . The principle is simply too clever to be true. and it didn’t last long. Naval Ordnance Lab. having relied (unfortunately) on an account given to me by a friend who examined the toy in Paris. and the other on the principle I suggested. Albert Altman wrote from the U. You are the inventor. at this point I am hopelessly confused. Well. It is possible that one version of the toy works on the principle you mention.” The Hourglass Letters While Hein was having his epiphany in Europe. and to suggest that you not publish a correction. Reid. but reported that his glass sometimes floated and sometimes sank. A letter dated September 29 came from Walter P. Gardner was getting letters in response to the September issue. 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. In short.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. 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. where is the rate of the flow of the sand. Gardner was beginning to wonder if his friction explanation told the whole story. but did not bring one back with him. also from the U. “I am writing to put your mind at ease (on the impact theory posed by Altman). One man wrote that he owned one of the Paris cylinders.

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

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

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

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

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

If. A.½ ¼. The combinatorial puzzle uses only six colors in its construction. C. 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. and 3 are any six distinctive colors. Trial and error is not a very fruitful way of trying to solve this puzzle. each of the 48 faces of the pieces is colored with a unique color. try again to solve the puzzle. and 3. FARRELL Figure 1. Therefore. B.½½ . and the solver has the additional clue that a face is all of one color. 1. These hints are to be regarded as progressive.½ distinct ways of constructing the cube. 1. additionally. but there still remain a great many cases that are nearly right. 2.146 J. Notice that the eight pieces must form the eight corners of the completed cube — one piece for each corner. proceed to the next hint. after reading a hint. C. There are 93 solutions in this case. each . 0). A. then there are ´ ¿µ´ µ´¿ µ¾ . If you cannot. B. 0. ½º Call the six colors A. That is. B. 2. Solution Hints. and C will turn out to be a solution set.

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

Smally. This will change the parity on one corner of the cube so that a solution is impossible using that set of three colors. There is another solution to this puzzle using the colors 1– 2–3. Then place the longs and. That is. or off-diagonal around all four lateral faces is always 42. Figure 3. . and C (or two of 1.148 J. If only one solution is desired. column. 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. one that gives only one solution. and 3). take any piece and interchange two of the colors A. main diagonal (upper left to lower right). This will assure that only one solution is attainable with the colors A–B–C (the proof is left to the reader). The Magic Die has the amazing property that the sum of any row. FARRELL must go by looking at its A–B–C corner. no seam runs completely through the cube in any direction and therefore no rotation of any part of the cube is possible. 2. finally. B. The solver will notice that the completed cube has a “fault-free” property. We like to have three Smallys prepared: one that yields two solutions.

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

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

The question then arises: How many solutions are there? Most intriguing about this puzzle. The previous question now becomes more interesting. when two cubes of the same color are in position.152 S. A proof is given in Section 2. only thirty dicubes are required.677. It becomes necessary. is that there are numerous solutions. FARHI a difficult exercise for a moderately proficient programmer to verify the results obtained. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. the results are shown in Table 1. obtained after some experimentation.691 combinations into 148 disjoint sets is described in Section 6. only 133. this rule then tells the experimenter to backtrack. . as determined in Section 4. and the most difficult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1. The second observation is that for a cube to be assembled with the nine colors on each face. Thus. Since during the course of trying to solve this puzzle a conflict often occurs. as is shown in Section 5.700 (36!/12!24!) different ways of which. The method of sorting the 10. A proof is given in Section 3. The first observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube. some with the tricube on the edge and some with the tricube through the center of the cube. Fortunately.691 combinations need to be sorted into disjoint sets. The third observation. The final part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. Twelve of these can be selected in 1. it becomes apparent that most of the 133.105 of these combinations meet the three-cubes-of-each-color requirement. to separate these cases into disjoint sets of isomorphisms.251. then. and only 10. but not in the center of a face. none of the nine planes may contain two cubes of the same color. Since one person can assign the fourth and fifth colors to orange and red and another person to red and orange. The individual dicubes can be colored thirty-six different ways. This was achieved through the use of a computer program. the location of the third cube is determined. in which the number of isomorphisms using the set of thirty-six dicubes is also determined.105 combinations are isomorphisms.

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

FARHI No. 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center /Obppp 720 84 38 2 104 46 20 /Oceyz 720 84 18 0 46 6 8 /Ocmyz 1439 168 34 3 27 12 2 /Ocnpz 1439 168 126 4 117 16 19 /Ocnyz 1439 168 40 1 36 11 11 /Ocnzp 1439 168 49 3 45 24 9 /Ocoxz 1439 168 112 5 14 3 2 /Ocozp 1439 168 18 1 35 28 9 /Oewez 720 84 32 2 38 48 4 /Oewnz 1439 168 52 3 20 15 1 /Oewpp 1439 168 17 6 38 12 4 /Oeyxp 720 84 36 0 12 6 4 9DEeyz 180 36 0 0 16 0 0 9DNeyz 30 6 0 0 0 0 0 9DOeyz 360 72 0 0 32 12 4 9DXeyz 720 108 10 0 0 0 0 9DYeyz 720 108 22 2 38 18 6 9DZ[yz 1439 216 69 3 32 17 2 9DZepz 1439 216 71 7 32 16 7 9DZpez 1439 216 55 6 52 13 5 9DZppp 720 108 54 12 72 18 20 9ODeyz 360 36 4 0 4 0 0 9OF[yz 360 36 76 0 72 8 4 9OFepz 720 72 82 0 45 0 6 9OFppp 180 18 208 0 134 8 0 9Ob[yz 720 72 54 4 41 12 2 9Obepz 1439 144 145 7 48 13 5 9Obmyz 1439 144 87 2 31 8 0 9Obnpz 720 72 2 0 18 4 0 9Obofz 720 72 68 4 28 2 0 9Obpez 720 72 22 2 25 4 0 9Obpfp 720 72 174 6 71 10 0 9Obppp 1439 144 30 3 26 5 4 9Odmpz 720 72 186 6 54 10 2 9Odmzp 360 36 16 0 6 12 0 9Odoyz 360 36 76 8 66 16 4 9Oewpp 360 36 40 0 16 4 0 DDDeyz 15 1 0 0 0 0 0 .154 S.

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

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

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

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

4. (2. (4. given the second numbers 2. 5. (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). 9. (3. 3). 5. This results in requiring only thirty cubes and. Using this algorithm. (5. 5). (7. 9). etc. only 10. 9). 8. (7. a computer program established 133. 3. and 9 one can logically deduce the first colors by the three . the two new colors are identified as colors 6 and 7. 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. Identification of the second color of these ordered pairs is sufficient to define all the colors. (6.105 possible color combinations. (1. 8. 2). 7). 8). 4. or 5. Identification and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1.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. using the same algorithm as before.691 color combinations. 7. For example. color 5 is combined with colors in the fifth row such that there are three cubes with color 5. Similarly if color 2 is not associated with colors 3. 6. 9) and (8. 8). 6). 9. (5. 6. 6. 4).

9). (4. It is then a simple process to recognize that this represents the dicubes colored (1. (6.:Zppp. This process then identified the 148 disjoint sets of isomorphisms. The ASCII codes for these characters are: 46. (4. including pairs of two at a time. and (8. .:Z[yz. 6). (2. 58. pairs of three at a time and four at a time with two at a time. can be swapped with each other.:Z[yz. The use of this six-character identification greatly simplified the process of sorting the 10. (1. (24). and 99. on the ends of the tricube. 6). Subtracting twenty-three results in: 23. 9).691 color combinations obtained by the algorithm described in Section 5 were listed in order according to their ASCII characters. Different solutions often have the same three-digit identification. (98). and then assigning a printable character to each. cannot be interchanged with other colors. which is in the center of the tricube. then additional criteria for recording solutions are recommended. The process was then continued until the list was exhausted. 3). four at a time. 8). (5. three at a time. Color 2. . 67. For example. 2). 9). (35). its isomorphisms are determined and removed from the list. and 122. 35. An inspection of the ASCII table will explain why it was decided to add 23. 91. 68. i. (4. and (99). FARHI cubes of each color requirement and the restriction that the second number be greater than the first. 90.691 cases into 148 disjoint sets of isomorphisms. but not with any of the other colors. The ASCII code for 23 is not a printable character. Note: The smallest possible number is 23 and the largest is 99. Shown in Figure 4 are an edge solution and a center solution for the color combination of Figure 1. 7). the first entry in Table 1 is . (67)..e.160 S. The 10. The first entry. (7. has sixty isomorphisms. The remaining six colors may be interchanged in numerous ways: two at a time. by adding twenty-three to each and using the ASCII character set codes (ASCII stands for American Standard Code for Information Interchange). Colors 1 and 3. The solutions are identified by a three-digit number indicating the color of the buried cube and the two “edge” colors. Identification was further simplified by placing these numbers in pairs (see Note below). Not all of these mappings produce a new isomorphism. (5. The head of the list then became . 121. five at a time including three at a time with two at a time. 8). and six at a time including a triplet of two at a time. These were removed from the list. Any permissible interchange of colors is an isomorphism. 98. 3). (7. Decoding a name is quite simple. (68). If the solutions are to be catalogued. 5). Thus this color combination is identified as /:Z[yz.

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

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

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

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

which I have called “Nine and One-Half Moves. V AN DEVENTER Figure 1. O. Nine and One-Half Moves. Zigzag.” 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). Figure 2. I took up the challenge again and succeeded in finding a new design of a planar burr.166 M. . The three pieces of the puzzle form a square with internal voids. Some months ago.

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

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

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

¢ ½½. ½ ¢ ½ ¢ ¾ ¢ . I wrote such a program on my first computer. ¢ ½ . ¢ ½¿. ¢ ½ . and ¢ ½¿. ¢ ¾½.480 solutions. º Boxed Box (23) (Cutler.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. There are four solutions to the puzzle. The Commodore 64 is such a wonderously slow machine — when running the program in interpreter BASIC. 10 are too wide to fit into the box sideways and 4 are of width 5. ¢ ½ . The solution involves use of the following obvious trick (is that an oxymoron?): Some piece(s) are placed sideways in the box. In early 1996. Since all the pieces are of the same thickness and the box depth equals three thicknesses. Bill Cutler Puzzles): Box: ½ ¢ ½ ¢ ½ Pieces: ½¿ ¢ ½½¾ ¢ ½ ½. The program displayed the status of the box at any instant using color graphics. ¢ ½¿. the widths and lengths are ¢ . ½½ ¢ ½½ and two each of ¢ . the machines had constructed 2 1/2 trillion different partially filled boxes. about once a second a piece is added or removed from the box! Using compiled BASIC. the difference between the three boxes is stunning: The first box can be completely analyzed in a small fraction of a second. When running these programs on more powerful computers. ½ ¢ ¿ ¢ . º Parcel Post Puzzle (18) (designer unknown. ½ ¢ ¾ ¢ ¾. ½ ¢ ¢ ¼.450. ¾¼ ¢ ¼ ¢ ¾. ½¼ ¢ ½½. which is no good for this purpose. 1978. but the process cannot be completed. By the end of the runs. ½ ¢ ¼ ¢ . ¢ . The analysis was done on about 20 powerful IBM workstations. ¾¢ . or the equivalent of one year on one machine. ¢ ½¼. There are 3. copied from a model in the collection of Abel Garcia): Box: ¢ ½ ¢ ¾ Pieces: all pieces are of thickness 2 units. Of the 18 pieces. The second box was analyzed in about a minute of mainframe computer time. and they all have three of these four pieces placed sideways. not counting rotations and reflections. all very similar. ¾½ ¢ ¢ ½ ¼. The total CPU time used was about 8500 hours. This leaves 4 pieces that might be placed sideways. the rate increases to 40 pieces/second. it is tempting to solve the puzzle by constructing three layers of pieces. I painted pieces of an actual model to match the display. but there are more efficient algorithms that can be used for block-packing puzzles. I did a complete analysis of the third box. a Commodore 64. . ¢ . The programmer can use algorithms that are used for pentominoe problems. The result was a fascinating demonstration of how a computer can be used to solve such a puzzle. One or two individual layers can be constructed.

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

pp. “Subdividing a Box into Completely Incongruent Boxes. pp. Klarner (Wadsworth International. being sure that the unfilled space is concealed at the bottom and is stable. [4] D. Recreational Math. readers. Place the box on your puzzle shelf with the remaining piece hidden behind the box. Gardner. 1981).. [5] D. This fits into the box with a little room all around. the pieces can be rearranged to make a ¢ ¢ ½¿¿ rectangular solid. 1979–80. pp. (No. 12(2). (This second construction is a little more difficult. Show off the solved box to your victim. Spring 1973. February 1976. When the ninth piece is added to the group. including the one in your hand.. Pack all the pieces.¿ ¢¾ ¢ . 104–111.½ . You are now prepared for your next encounter with a boring puzzle-nut. This should keep him busy for quite some time! References [1] W. By the way. “Brick-Packing Puzzles.¿ ¢ ¢ plus a second copy of ½ ¢ ¾ ¢ . ¿ ¢¾ ¢ . being careful to keep the renegade piece hidden from view. Cutler.” J. Miscellaneous ½½º Melting Block (8-9) (Tom O’Beirne) Box: ¢ ¢ ½¿¿ Pieces: ½ ¢ ¾ ¢ .BLOCK-PACKING JAMBALAYA 173 12.¿ ¢ ¢ ¢ ¢ . one of the puzzles listed above is impossible. Gardner. 20–23. but rather a tautology to the 99% of the world that would never even have started to read this article. It is a valuable weapon in every puzzle collector’s arsenal.) Pick up the box and last piece with both hands. Mathematical Games column in Scientific American. . ½ ¢ ¾ ¢ ½ ¢ ¢ . I won’t say which one (it should be easy to figure out). 122–127. pp. and then dump the pieces onto the floor. [3] M. Recreational Math. 6(2). except one. The Melting Block is more of a paradox than a puzzle.” J. February 1979. 212– 225. The eight pieces fit together easily to form a rectangular block ¢ ¢ ½¿¾. into the box. Mathematical Games column in Scientific American. [2] M. 112–117.” in The Mathematical Gardner. this is not another oxymoron. edited by D. pp. “Packing Problems and Inequalities. but seems to the casual observer to fill up the box completely.) This is a great puzzle to show to “non-puzzle people” and is one of my favorites. Hoffman. Klarner.

.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.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

is left as an exercise for the reader. GOLOMB it is a well-established principle that a property P is true for all members of a set S if it can be shown to be true for an arbitrary member of the set S. it is sufficient to prove it for an arbitrary government. W. To prove the assertion for all governments. it is obviously unjust. And since this is true for an arbitrary government. and extending the method to other models of mathematical proof. Finding further theorems of this type.192 S. Proof. We exploit this to obtain the following important result. All governments are unjust. If a government is arbitrary. it is true for all governments. . Theorem 3.

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

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

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

Out from the node labeled Ü the two directed edges go to the nodes labeled ¾Ü and ¾Ü ·½. When the number of nodes is ¾ Ò we can label them with the integers from ¼ to ¾ Ò   ½ and notice that they correspond to all the numbers expressible with Ò bits (Ò binary digits). Figure 1.Drawing de Bruijn Graphs Herbert Taylor The simplest kind of de Bruijn graph. If ¾Ü or ¾Ü ·½ is bigger than ¾Ò   ½. in which the number of nodes is a power of two. . two edges directed out and two edges directed in. at each node. Ò 197 . has. just subtract ¾ Ò from it to bring the number back into the range from ¼ to ¾ Ò   ½.

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

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

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

We are still left with the nagging problem of resolving a bet between two of the authors.COMPUTER ANALYSIS OF SPROUTS 201 ¯ of previous searches in a hash table. The first player has a winning strategy in n-spot misère Sprouts if and only if n is 0 or 1 modulo 5. This problem seems to lie just beyond our program. The first player has a winning strategy in n-spot Sprouts if and only if n is 3.½ ½For a paper describing these results in more . go to http://www. To reduce the space still further we used a data compression technique. our computational resources. The size of the hash table turned out to be a major limitation of the program.cs. and we devised and implemented two methods to save space without losing too much time efficiency. and searching the successors of a position in order from lowest degree to highest. Sleator believes in the Sprouts Conjecture and the Misère Sprouts Conjecture. and he bet Sleator a six-pack of beer of the winner’s choice that one of them would fail on some game up to 10 spots. The data for misère Sprouts fit a similar pattern. We discovered that saving only the losing positions reduced the space requirement by a large factor. and an asterisk indicates a new result obtained by our program. The only remaining case required to resolve the bet is the 10-spot misère game. a “2” means the second player has a winning strategy. Misère sprouts conjecture. or 5 modulo 6. characterized by the following conjecture: Sprouts conjecture. 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 n-spot Sprouts positions evaluated so far fall into a remarkably simple pattern. Applegate doesn’t believe in these conjectures.cmu. and our ingenuity. 4.

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

204 B. GOSPER Times 100–105: .

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

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

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

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

But it’s fairly easy to build a pattern for which this isn’t true: Take two glider puffers. headed east and west. although almost all large patterns grow to infinity. Even infinite growth is rare for small patterns. In Winning Ways you (JHC) describe how to pull a block three units using two gliders and push it three units using 30. These use a glider salvo to push a block (or blinker). it eventually becomes periodic. I’ve also built some aperiodic patterns with bounded populations. so some of us have spent many hours putting together guns and puffers to produce various results. (Figuring out how to achieve a particular behavior makes a pleasant puzzle.) Most of my large constructions are designed to achieve unusual population growth rates.STRANGE NEW LIFE FORMS: UPDATE 209 small things we normally look at. using a kickback reaction each time it hits one of the glider waves. I found a way to pull it one unit with two gliders and push it one unit with three gliders. similar to the one you described. Both of these patterns have populations tending to infinity. This is even true for guns.) The first such pattern (other than glider guns) was Bill’s breeder. In addition. Using this I built a sliding block memory register. and breeders.) 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. Smaller breeders are now known. firing gliders southwest and northeast. Add a glider that travels northwest and southeast. actually building the thing is mostly tedious.” versions of which were built independently by Bill and myself. respectively. a signal is produced whenever a decrement operation reduces the value to 0. the reaction also sends back a glider (or two) which triggers the release of the next salvo. puffers. (It has one small difference: the “test for zero” is not a separate operation. Instead. and probably others. whose population grows like Ø ¾. such as Ø ½ ¾ Ø¿ ¾ ÐÓ ´Øµ ÐÓ ´Øµ¾ Ø ÐÓ ´Øµ and Ø ÐÓ ´Øµ¾ . Probably the second such pattern is the “exponential aperiodic. depending on which cells you’re looking at. and linear growth with an irrational growth rate. Bill also built an “arithmetic aperiodic” in which the gap between occupations increases in an arithmetic progression. there are several . If you look at a finite region in a typical small pattern.

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



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



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



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



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

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

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

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

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

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

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

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

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

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

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

with the problem statement giving the 6 as a 7 Fibonacci. 8. 0. 250 c. 50 in the first printed English riddle collection “The Demaundes Joyous. 2. 1. 9. 2. 5. 4. 510 9Ø century c. 9. 3. 9.C. 1010 1150 1202? 1202? 1202? 1202? 1202? 1202? 1202? Euclid Diophantus The Greek Anthology The Greek Anthology Alcuin Mahavira Mahavira al-Karkhi Bhaskara Fibonacci Fibonacci Fibonacci.5) I-(25. 10. 7) I-(7. 4) (2.” printed by Wynken de Worde in 1511 [9]. 2. to sell. which daughters he delivered. done two ways Selling Different Amounts at the Same Price Yielding the Same Although this problem is quite old.SOME DIOPHANTINE RECREATIONS 225 from the early fourteenth century and hence after Fibonacci. 11. 2. 6. and to the second thirty apples. Now how many sold each of them for a penny? . 10. Sridhara [18]. 3. with the problem statement giving the 6 as a 7 Fibonacci Fibonacci. 3. 2. 850 c. 5. hence the above title. 7) II-(1. although the material is believed to be based on older Arabic sources. 22. 3. and Bhaskara [4]. 1. The problem is treated in one form by Mahavira [15]. 50. 2. 850 c. 9. 3) (10. 4. 4. 3. 11. 10) (7. 3. 1. c. 6) I-(7. 6) (1. 1. 10. 3. there seems to be no common name for it. 23. 5. 11. 6) I-(7. 9. 5. (1. which is rather a mouthful. certain apples. 10. 2. 5) c. noting that this system is inconsistent Fibonacci. 5) (100. 8. 4. 4. 5. 2. 11. 4. 6) I-(7. 11. and all these three sold in like many for a penny. 2. A man had three daughters of three ages. and to the youngest ten apples. but this form is less clearly expressed and has infinitely many solutions. 2) I-(9. 3. 11. so I will first describe the later form which is first found in Fibonacci [11]. I quote a version that appears as No. 510 c. 5. 5) (2. 5. 1) (30. 300 B. and brought home in like such money. And he took to the eldest daughter fifty apples. 7) I-(7.

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

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

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

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

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. we have ¾Ã   ½ solutions with zero values. From (15).e.e. Æ· ½¼   ¾ ¢ ¾ . Then we have ­Ò ´Ò   ½µ . Subtracting ¾Ã or ¾Ã   ½ from Æ yields Æ· . A little thought shows that then   is the common difference of the progression. ­Ò   Ò   ½: ´½ µ à A number of early cases have Ò ¾. Most of the examples in the literature have the in arithmetic progression (AP).. 90). These occur when ½ ¼.230 D. Looking again at our example. and then à ½ ¾ Æ ¾ Æ· ´ ½   ¾µ¾ . and when Ò ¼. Tartaglia’s problem 139 is the one with two sizes of pearls and is (10. All but one of these have ½ even. ½ ½ . unless it happens that ½ «¾­Ò ­¾ . so we let Ã Æ  ½ ­  Ò à where [ ] is the greatest integer function. . We then have à ½ Æ ½´ ½ · ½µ ¾ Æ· ´ ½   ¾µ´ ½   ½µ ¾.e. so we can write ½ ·´   ½µ . ´½ µ Again. Substituting (11) into (8) and summing gives us the number of solutions. SINGMASTER Then (10) becomes ´½¾µ ´½¿µ ½  ½ ­Ò The upper bound may not be an integer. perhaps as intended by De Lagny. The majority of examples are APs with Ò ¿. Rearranging. when the upper bound for given in (13) is an exact integer. we find that this occurs when   ½ ­Ò . ½ A few longer APs occur — I have seen examples with Ò . we have à ½ Æ ¿ Æ· ½. then ½ we get à ´ ½   ½µ ¾ Æ ´ ¾   ½µ Æ· ´ ½   ¾µ¾   ½ . i. A few cases occur that are not APs (see below). We can count the number Æ · of positive solutions by finding the number of solutions with a zero value.. ½ «¾­Ò ­¾ by (7). 20. If ½ is odd.. when two zero values occur in the same solution. In that case. Hence we normally get ¾Ã situations with a zero value. which is trivially an AP. 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 . in our example. i. i.

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

29. 1 given. 140) . 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. 40. 1 given.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 30) (8. 68. 82) (17. 16) (16. so we again have a three-dimensional solution set. One can scale the set . 1 given. 81) (55. 30. 40. 32) (12. 55) (55. 15) (31. 32.400) (25. 36) (30. none given. 1 given. 1 given. 30) (12. 752.6) # of Solutions (55. 20) (10. 901) (10. At this point. 11. 36) (27. 454. 33. 40) (10. 1 given. 60) (117. 12. 30. Sridhara finds a one-parameter solution set. 603. 40. Bhaskara makes the following comment: “This. 50) (10. 1 given. 60) . 36) (78. 33) (20. 136) (3.232 Version (10. 50) (30. 50) (11. 55) (16. ¬ . 20. 1) (7. 26) (20. (20. 100) (45. 40) (18. 1 given. 81) (10. 16) D. . 37) (27. 20) (64. 48. 1 given. (20. 90) Date 1202? 1202? 1202? c.552. 1) (100. 196) (45. 30) (18. SINGMASTER Solutions in Source 5 given. none given. 30. . 49) (3. 1 given. 25. 1 given. 1300 1300s 1489 1489 1489 c. 36) (25. 60) (10. 55) (78. 1 given. all given. 2 given. makes this more specific by saying they ‘sold a part in lots and disposed of the remainder at one for five panas. 1 given with fractions. 33) (10. 1 given. 170) (305. 1 given. in some way to reduce the solution set to two dimensions.9) (100. 4 given. . 6 given. 1 given. 20. one with fractions. 21) (225. 56.4) (70. though it is not as easy to see how to do this as in the Western version. 17. 119. 80) (10. 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. 35) (11.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

V AN DEVENTER Figure 8.250 M. which can be reached for prime Æ . and many questions remain open: ¯ ¯ ¯ ¯ ¯ Is there a better search strategy for composite Æ ? What are the properties of solutions for Æ ½ and higher? Are there more Æ with multiple solutions? Can a sharper theoretical maximum for odd and composite expressed in a closed form? Are there solutions that use the back side of the mirrors? Æ be . O. solutions were found by a systematic trial and error search using a regular Æ -gon grid. No general strategy was found. For composite Æ up to 16. Deleted paths for composite Æ up to 16.

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

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

We let Ø´Òµ denote the largest value reached in the trajectory of Ò. . It is usually credited to Lothar Collatz.e.. each possible binary pattern occurs exactly once in ½ Ò ¾ . C. i. 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.. Finally we let Ö´Òµ count the fraction of odd iterates in a trajectory up to the point that ½ is reached. Prior to that it circulated for many years by word of mouth. The first mathematical papers on it appeared around 1976 ([8]. Ò Ó ´½¿º¾½µ Ì ´ µ´Òµ ½ ½ ´Òµ Ñ Ò ´½¿º¾¼µ Ç´Òµ where ½ ´Òµ ½ if no iterate equals ½.254 J. then the distribution of the -vector ½ ½ ´Òµ Ò Ì ´ µ ´Òµ ½ ´ÑÓ ¾µ for ¼ ½ ´Òµ Ó Ú ´Òµ Ò Ì ´Òµ Ì ´  ½µ´Òµ ´ÑÓ ¾µ is exactly uniform. Furthermore the resulting pattern is periodic in Ò with period ¾ .. so that ´½¿º¾¾µ Ø´Òµ ×ÙÔ Ì ´ µ ´Òµ Ò ¼ Ó where Ø´Òµ ½ if the sequence of iterates is unbounded. Müller [18]. Surveys of the known results on the ¿Ü · ½ problem can be found in Lagarias [13]. 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. i. who studied problems similar to it in the 1930s. if one takes the uniform distribution on ½ ¾ . For example. [21]). see [22]. The total stopping time ½ ´Òµ of Ò is the minimal number of iterates needed to reach ½. LAGARIAS The ¿Ü·½ problem appeared in Martin Gardner’s “Mathematical Games” column in June 1972 [10]. This basic property was independently discovered by Everett [8] and Terras [21]. i. Thwaites in 1952. and now more than one hundred papers have been written about the problem.e. ´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. It was also independently discovered by B.e. and Wirsching [25]. and who has stated that he circulated the problem in the early 1950s [6]. To describe iterates of the ¿Ü·½ function we introduce some terminology.

In doing so we are in the atypical situation of modeling a purely deterministic process with a probabilistic model.3X + 1 FUNCTION ITERATES 255 One may summarize this by saying that the parity of successive iterates of Ì initially behaves like independent coin flips. The simplest of them assumes that this independent coin-flip behavior persists until ½ is reached under iteration. These mathematical models predict that the expected size of a “random” Ò after steps should be about ´ ¿ µ ¾Ò. This model has the drawback that . The pseudorandom character of ¿Ü ·½ function iterates can be viewed as the source of the difficulty of obtaining a rigorous proof of the ¿Ü · ½ Conjecture. Garner [11] and Korec [12]) they also seem to possess some residual structurelessness. [24]. On the positive side. for an ¾ explicit constant ½. Here we survey some recent results on stochastic models. and a branching random walk model for backward iterates of Ì . 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. see [20]. The repeated random walk model results of [15] show almost perfect agreement between empirical data for ¿Ü · ½ function iterates for Ò up to ½¼ ½½. obtained in joint work with Alan Weiss and David Applegate. From this viewpoint it becomes interesting to determine how well the properties of ¿Ü · ½ iterates can be described by a stochastic model. Several heuristic probabilistic models describing ¿Ü ·½ iterates are based on this idea. Since a number Ò is multiplied by ½ if it is even and approximately ¿ if it is odd. cf. [15]. which concern extreme behaviors of ¿Ü · ½ iterates. one expects that ¾ ¾ on average it changes multiplicatively by their geometric mean. [20]. which may be enough for a pseudorandom bit to be extracted. one expects the iterates to decrease in size and eventually become periodic. These models consist of a repeated random walk model for forward iterates by Ì . Lagarias and Weiss [15] studied two different stochastic models for the behavior of ¿Ü · ½ function iterates. [13]. Cloney. which is ´ ¿ µ½ ¾ . Even though ¿Ü · ½ function iterates possess quite a bit of structure (cf. [5] proposed using the ¿Ü · ½ function as a pseudorandom number generator. et al. the number of steps ½ ´Òµ for Ò to iterate to 1 should be nor  ¡ ½ Ô mally distributed with mean ½ ÐÓ ¿ ÐÓ Ò and variance ½ ÐÓ Ò. see [9]. Since this is less than ½. Lagarias [14] and Luby [17]. There is excellent numerical agreement of this model’s prediction with ¿Ü · ½ function data. 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. [24].

for these probabilistic models the deviation from independence exhibited by ¿Ü ·½ trajectories does not affect the asymptotic maximal length of extremal trajectories. C. Various authors have obtained rigorous results in the direction of the ¿Ü ·½ Conjecture. The ¿Ü · ½ Conjecture remains unsolved and is viewed as intractable. ´Üµ ܼ ½ 14. which are defined in Section 3. They observed that the distribution of the largest and smallest number of leaves possible in such trees of depth appears to be narrower than what would be predicted by the model of Lagarias and Weiss. indeed actual ¿Ü ·½ trajectories coalesce to form a tree structure. Repeated Random Walk Model Lagarias and Weiss [15] studied a repeated random walk model for forward iterates by Ì . This leads to two conjectures stated in Section 4. Lagarias and Weiss introduced a family of branching random walk models to describe the ensemble of such trees. the successive iterates ÐÓ Ì ´ µ´Òµ are modeled . More recently Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees. That is. In this model. and found small systematic deviations of the distribution of ¿Ü · ½ inverse iterates from that predicted by the branching random walk model above. They compared empirical data with predictions made from stochastic models. see [2]. LAGARIAS it does not model the fact that ¿Ü·½ trajectories are not independent. 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. consisting of lower bounds for the number of integers Ò below ½. It remains a challenge to exploit such regularities to obtain new rigorous results on the ¿Ü · ½ problem. The local structure of backward ¿Ü · ½ iterates can be explicitly described using ¿Ü · ½ trees.256 J. More generally. one may a value Ü that have some iterate Ì ´ µ ´Òµ estimate for a positive integer the quantity Ü with some Ì ´ µ´Òµ It has been conjectured that for each positive ½ or ¾ ´ÑÓ ¿µ there is a ´Üµ card Ò Ò ½ Ò Ó positive constant such that ´Üµ Ü for all Ü see Applegate and Lagarias [2] and Wirsching [25]. 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.

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. the limiting graph of these trajectories has a different appearance." What is the extremal size of ½ ´Òµ? Large deviation theory shows that with probability one ´½ º½µ The constant ­ÊÏ is the unique solution with ­ tional equation £ ´Òµ Ð Ñ ×ÙÔ ½ Ò ½ ÐÓ Ò ­ÊÏ ³ ½ ´ ½ ÐÓ ¿ µ ½ of the func¾ ´½ º¾µ where ­ ´ µ ×ÙÔ ¾ ½ ­ ½ ´½ º¿µ Next. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally. this explains the name “repeated £ random walk model. For each initial value Ò an entirely new random walk is used.4). This random variable is finite with probability one. The length of such trajectories approaches the limiting value ½ ´Òµ ÐÓ Ì ´ µ ´Òµ ÐÓ Ò ÐÓ Ò ¼ ½ ´Òµ . £ and the expected size of ½ ´Òµ is given by (13.3X + 1 FUNCTION ITERATES 257 ¿ by starting at ÐÓ Ò and taking independent steps of size   ÐÓ ¾ or ÐÓ ¾ ½ each. so that Ø£ ´Òµ ÐÓ Ò. Large deviation theory also asserts that the graph of the logarithmically scaled trajectories ´½ º µ approaches a characteristic shape. The random variable ½ ´Òµ is the step number £ with probability ¾ at which 0 is crossed. For the extremal trajectories for Ø£ ´Òµ in (14. Large deviation theory shows that.24). with probability 1. see Figure 1. For the extremal trajectories for ­ £ ´Òµ in (14.1) it is a straight line segment with slope  ¼ ¼¾¿½ starting from ´¼ ½µ and ending at ´­ ¼µ. let Ø£ ´Òµ equal the maximal value taken during the random walk.

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

½ ´Òµ It gives the longest trajectory.94 31. A comparison of (14.64 32.5841 0. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory.171 4 52.5857 0.1) with numerical data up to ½¼ ½½ is given in Table 1. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale.77 31.371. The large deviations extremal trajectory is indicated by a dotted line. Maximal value of ­ ´Òµ in intervals ½¼ Ò ½¼ ·½. 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. More recently V. and the large deviations extremal trajectory is indicated by a dotted line.5741 0.6026 0.527 Random walks model ½ ´Òµ ­ ´Òµ 70 108 165 214 329 429 592 593 706 755 21.728.91 32.127 8 127. The deviations from the stochastic model are well within the range that would be expected from the size of the expected standard deviation for such models. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ­ ´Òµ ¿ ¾ .884.180. Vyssotsky [23] has found much larger numbers with high values of ­ ´Òµ.68 24. where ­ ´Òµ ÐÓ Ò . an analysis in [15.91 19. ½ ½¼. 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. Section 5] shows that in ½¼ ½½ trials one should only expect to find a largest ­ ´Òµ ¿¼.254 9 4. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good.527 has ½ ´Òµ ½¾ ½ and ­ ´Òµ ¿ ¾ ½ .6020 0.24 16.13 26.511 7 63.5927 0. .799 6 8.5967 0.890. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1.68 Ö´Òµ 0. The agreement with the stochastic model is quite good.456.6020 0. He found that Ò 12.5818 0.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6.4).38 41.48 18. Extensions of the data can be found in Leavens and Vermeulen [16].400.328.527 5 837.6030 0.6091 Table 2.194.815 10 13.769.266.

or 7 ´ÑÓ µ. In iterating the multivalued function Ì  ½.528.675 7 80. using (15. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿.86 7.631.788 1.511 6 6. Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk.976 1. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ.511 10 77. C. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth .05 19. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼. LAGARIAS Ò 1 27 2 703 3 9.) empirically like . respectively. In this case. so do not have any significant effect on the size of the tree. In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed.59 23.819 1. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch.75 5.362.65 5.566.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½.83 3. see [15] for more explanation. . Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2.48 12.14 11. The set of inverse iterates associated to any Ò has a tree structure. 15. In fact.000 Ò 13.65 7. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ. ½ ½¼. where ½ ´½ º½µ Ì  ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½.02 21. these trees have the minimal and maximal number of leaves possible for depth .46 2.903 2.791 1.06 4.671 5 704. if Ò ¾ or 8 ´ÑÓ µ. there are duplications.32 2.817. 5. (See Table 3 in Section 4.20 6.663 4 77.260 J.897 2.049.909 1.560 1.24 16.790 1.86 13.645 ­ ´Òµ 21.1).84 19. 4.10 20.03 19. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree.831 9 8.804. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ.391 8 319.099 1.559 Random walks model Table 3.86 5.

Theorem 3.1]. We assign labels ¼ or ½ to the edges. ­ È ­ÊÏ . . satisfies (with probability one) ´½ º¾µ ÐÓ Ò£ ´ µ ­ ȳ ½ ÐÑ ½ ÐÓ This constant ­ È ¬  ½ . it was necessary to determine the distribution of the number of leaves in a tree in the branching process. where the random walk starts at the origin on the real axis at the root node and takes a step ÐÓ ¾ along an edge labeled ¼ and a step ÐÓ ¾ along an edge ¿ labeled ½. [15. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½. In order to prove the result (15. Lagarias and Weiss [15. and the root node is assigned the value 1. where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree. cf. and where fying £ ´ µ ×ÙÔ ´½ º¿µ   ÐÓ ´¾ · ½ ¾ ¿ ¿ ¼ This constant is provably the same constant as (14. The random walk at a given leaf node ends at a position Ù on the real line.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. Pruned ¿Ü · ½ trees of depth Ò . 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.2). 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 . i.4] show that the expected size of ÐÓ Ò£ ´ µ. Theorem 4.1). and the value assigned to that vertex by the rules above is exactly Ù .. starting from root nodes Ò and The problem of determining the largest trajectory to get to 1 under iteration in a ¿Ü · ½ tree is equivalent to studying the leaf in a ¿Ü · ½ tree having the smallest value.e. In [15] it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ . where ¬ is the unique positive number satis£ ´¬ µ ¼.

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

so that each vertex of the tree corresponds to a random walk ending at some position Ù on the real line. 5 or 8 0 0 0 0 0 0 1 4 7 2 5 8 Figure 4. This value represents the initial value of the ¿Ü · ½ iteration starting from that vertex of the tree. and ¢ for a fixed constant ¢ ½. and we consider the value of that vertex to be Ù .4). respectively largest number of leaves. 1/3 each) 2.3X + 1 FUNCTION ITERATES (prob. The paper [3] presents more evidence in favor of Conjecture ½These facts are derived using the probability density for Ï given in (15. and it yields a second child by the yields a child by the map Ò ½ multivalued map Ò ¿ ´¾Ò ½µ (edge label 1) if Ò ¾ or 8 mod 9. The base vertex of the tree is located at the origin. Transitions of the branching process [9].) We consider as random variables the smallest and largest number of leaves that occur among this set of trees. choosing the root node uniformly ´ÑÓ µ.   ¾ takes a step ÐÓ ¾ along an edge labeled ¼ and takes a step ÐÓ ¿ along an edge labeled ½. The parent (bottom) always ¾Ò (edge label 0). We make Ê´ µ independent draws of a tree of depth generated by this process. . 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. 4 or 7 1 7 263 (prob. Let Æ  ´ µ and Æ · ´ µ denote the expected value of the smallest number of leaves. (For actual ¿Ü · ½ trees. The number of leaves in actual ¿Ü · ½ trees empirically exhibits less variability than is predicted by this branching random walk model. choosing Ê´ µ Ê´ µ ¿ ·½ so ½ ¢ ¿. 1/3 each) 2 8 5 4 1.

911 1.481 0.37 3.327 2.746 1.750 0.076 2.270 2.748 0.300 0.489 0.486 0.211 2.390 2.750 0.898 1.534 0.264 J.118 2.475 0.60 747.305 0.688 1.255 2.738 1.249 0.753 1. .488 0.49 9.792 1.465 0.747 1.481 0.758 1.401 0.405 2.633 0.375 2.16 4.342 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 4.780 1.238 1.352 0.746 1.83 99.491 1.265 0.289 0.50 315.481 0.255 0.45 560.67 0.302 0.791 1.34 420.409 0.09 56.751 1.802 1.482 0.762 1.748 1.12 74.232 2.475 0.21 5.375 0.57 42.492 0.62 7.451 0.898 1. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a fixed node Ò´ÑÓ ¿ ·½µ.281 0.778 1.490 0.534 0.32 17.562 0.246 0.738 1.669 1.166 2.252 0.802 1.77 2362.758 1.03 177.62 1328.742 1.507 0.36 3149.259 0.99 13.292 2.422 0.422 0.501 0.923 1.47 996.782 1.344 2. C.488 0.728 1.500 1.481 0.243 0.190 2.240 0.75 5599.475 0.68 31.054 2.688 1.419 2.78 2.307 0.534 0.750 1.268 0. Normalized extreme values for ¿Ü · ½ trees and for the branching process.38 236.557 1.741 1.310 2.491 0.633 0.273 0.360 2.83 1771.76 23.500 1.394 0.489 0.81 4199.774 1.562 0.486 0.784 1.131 2.869 1.77 133.782 1.507 0. LAGARIAS Ê´ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 µ 4 8 14 24 42 76 138 254 470 876 1638 3070 5766 Æ  ´ µ 1 1 1 2 2 3 4 5 6 9 11 16 20 27 36 48 64 87 Æ ·´ µ 2 3 4 6 8 10 14 18 24 32 42 55 74   ¡ ¿ ¿Ü · ½ Function Trees  ´ Branching Process µ · ´ µ  ´ µ · ´ µ 1.802 1.490 0.026 2.271 0.490 0.986 2.33 1.

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

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

Sign up to vote on this title
UsefulNot useful