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

That is. I sold my collections and chucked the whole business. there are various interviews and articles about Gardner. without narrative but arranged to tell a story. but merely bleach it. *** “I have recently read an article on handwriting and forgeries in which it is stated that ink eradicators do not remove ink. [1932 yearbook caption.. and 3 . while a sixteen-year-old student at Tulsa Central High.” [2] *** An able cartoonist with an adept mind for science. magic. and I’ve always talked clearly when people interview me. having been a fan for some time. and writing were to stay with him. It’s lived mainly inside my brain. 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.] *** [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 . The early interest in science. Can you give me a description of this process. Instead of a true biography. a month later. The first two times Gardner appeared in print were in 1930. what chemicals are used. we present here a portrait in the style of a documentary. and how it is performed?” [1] *** “Enclosed find a dollar bill for a year’s subscription to The Cryptogram. was a query to “The Oracle” in Gernsback’s magazine Science and Invention. I don’t think my life is too interesting.. quoted below. I am deeply interested in the success of the organization. [21] While there is no biography of Martin Gardner. we give a collection of quotes and excerpts. puzzles. and that ink so bleached can be easily brought out by a process of ‘fuming’ known to all handwriting experts. The first.Also below are two quotes showing a strong childhood interest in puzzles.

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

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

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

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

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

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

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

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

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

rational. I’ve heard of it. by Martin Gardner? Professor Byrd: No. Bad. He was a very famous science writer of the last century. 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. He actually advances the thesis that Conan Doyle never wrote the Sherlock Holmes stories—that these stories are forgeries. Byrd: To tell you the truth. Gardner. and Bogus.The year is 2050 A. and Doyle Raymond Smullyan SCENE I .Ambrose.D. 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. He 13 . 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. I guess the answer is that Doyle went senile in his last years? Ambrose: No. Byrd: Oh? Ambrose: The chapter is titled “The Irrelevance of Conan Doyle”. Professor Ambrose: Have you ever read the book Science: Good. Byrd: That is weird! Especially from Gardner! On what does he base it? Ambrose: On absolutely nothing! His whole argument is that no one with the brilliant. and of course I’ve heard of Martin Gardner.

Then. how could someone of Gardner’s caliber ever entertain the silly notion that Doyle never wrote the Holmes stories? Ambrose: To me the solution is obvious: Martin Gardner never wrote that chapter! The chapter is a complete forgery. 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. I never had the slightest doubt that Doyle wrote the Holmes stories. it’s just as implausible as Gardner’s idea that Doyle never wrote the Holmes stories at all.14 R. 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. when the public was convinced of his rationality. 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. whamo! Ambrose (After a pause): That’s quite a cute idea! But frankly. his disturbance went much deeper. he absolutely refused to believe Houdini when he said that there was a perfectly naturalistic explanation for the escapes. I have no idea who wrote it. So senility is not the explanation. then. 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. 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. SMULLYAN also pointed out that Doyle’s interest in spiritualism started much earlier in life than is generally realized. Ambrose: I’m glad you realize that. Doyle insisted that Houdini was lying! If that’s not psychotic paranoia. As I said. . Byrd: But now something else puzzles me: Martin Gardner was no fool. Byrd: All tight. he was surely one of the most interesting writers of the last century. Now.

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

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

is one of Gardner’s pseudonyms. The review ends with the sentence “George Groth. until Martin kindly informed me that the whole thing was simply a hoax! Martin is really great on hoaxes—for example. The Flight of Peter Fromm. The Ancient Mariner.” . his annotated editions of Alice in Wonderland. and Casey at the Bat—not to mention his religious novel. GARDNER. The Hunting of the Snark. In Martin’s book. an opening move in chess (pawn to Queen’s rook four) that guaranteed a certain win for white. by the way. Whys and Wherefores (University of Chicago Press. he reported the discovery of a map that required five colors. a discovery of a fatal flaw in the theory of relativity. his numerous puzzle books. so well known and highly respected for the mathematical games column he wrote for years for Scientific American. is reprinted a scathing review of his The Whys of a Philosophical Scrivener by a writer named George Groth. 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. 1989).A Hundred Years Later (To be supplied by the reader) Discussion: How come this same Martin Gardner. How could he ever believe anything that bizarre? Broad: Ah. and his Whys of a Philosophical Scrivener—how come he wrote such a crazy chapter as “The Irrelevance of Conan Doyle”? This troubled me for a long time. AND DOYLE 17 Cranby: But what puzzles me is how such a sober and reliable historian as Ambrose could ever believe that Gardner didn’t write that chapter.AMBROSE. in his April 1975 column in Scientific American.

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

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

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

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

roughaveragespeedofEarth 700.000mph.24 T.000 mph turning speed of Galaxy 1.4 million mph Galaxy’s speed through the debris of the Big Bang . ROWETT Speed A child cycles ‘round the schoolyard Which lies on the Earth turning hard The Earth rounds the Sun As Sol does “the ton” And our Galaxy flies — Gee! I’m tired 7 mph — child cyclist 700mphEarth’ssurface(NorthAfrica) 70.

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

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

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

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

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

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

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

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

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

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

46. 32. Problem 5: Calculate the expansion of the following 26 terms. 23:55:32. y Problem 7: The figure shown here is the solution to the problem of dividing the figure into four identical shapes. 64. 6. 13:22:31.) These curious combinations occur 660 times a day. Problem 6: Which two numbers come at the end of this sequence? 2. 36. For example. 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. (1) Find the closest such times. (Ignore the colons. 42. 44.38 N. 30. 52. 4. 34. 56. . (2) Find the two palindromes whose difference is closest to 12 hours. 60. YOSHIGAHARA Problem 4: Arrange the following five pieces to make the shape of a star. etc. x. 2:41:42. 1:01:01. 54. 3:59:53. (3) Find the longest time span without a palindromic time. 40. 50. 66. 62.

and 5 matchsticks long is divided into two parts with equal area using 3 matchsticks. . but you may rotate them in the plane. Can you divide this triangle into two parts with equal area using only 2 matchsticks? Problem 10: Cover the 7 ¢ 7 square on the left with the 12 L-shaped pieces on the right.CREATIVE PUZZLE THINKING 39 Problem 9: A right triangle with sides 3. You are not allowed to turn over any of the pieces. 4.

How many pieces do you need? .40 N. 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.

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

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

13. That is. 21. but without using a program.). The black-hole number this time is 15. 1. but your very equilibrium. Divide both sides of the equation ´Òµ ½µ · ´Ò   ¾µ by ´Ò   ½µ. Its divisors are 1. we would have obtained the reciprocal of phi as the limit. phi.618 . One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. Now take the sum of the digits of these divisors. 5. 34. Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . The ratios seem to be converging to a number around 1. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors. the smallest is 0288. 89. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò   call it Ä — as Ò increases. ECKER Example: Start with 8028. The first two numbers numbers: 1. Iterate until a number repeats. More formally. and ´Òµ preceding two elements. Its value is roughly 1. The largest permutation is 8820. This one may defy not only your ability to explain it. ´½µ ´Ò   ½µ · ´Ò   ¾µ for integers Ò ¾. Repeat with 8532 to calculate 8532–2358 = 6174. etc. 55. and the difference is 8532. but it is also that much more strange at the same time. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. This one is a bit more tedious. Had we used the ratios ´Òµ ´Ò · ½µ instead. In fact.48 M. 5. ½ . Your own example may take more steps. 8.6 or so. To get a quick and dirty feel for why phi arises. including 1 and the number itself. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. 2. 3. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. 3. but fewer authors use that approach. and 15. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. and these digits sum to 15. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ ľ   Ä   ½ ¼ . W. but you will reach 6174.

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

All real inputs except zero. Under certain conditions. then 2. and conversely. dating back to the 1930s and still an open question (though sometimes also identified with the names of Hailstone. you get 16. which really are cycles of length one. iteration of matrix powers draws one to a result that represents long-term stability independent of an input . is among the easiest to program (a few lines). For instance. You should find that you keep getting 4. and Syracuse). then 2. let’s be creative and fix this up by modifying the problem. (Just multiply all the odd prime factors together. Unsolved Problems as Black Holes Even unsolved problems sometimes fall into this black-hole scenario. then 1. and the second golden number lead to the attractor. Anybody who proves the Collatz Conjecture will prove that my variation is a mathemagical black hole as well. Ulam. Define the process instead by taking the starting number and breaking it down completely into factors that are odd and even. then 8. I will omit a program here. you stay at 4. the number ½ ½ is an attractor with the black-hole property. a cycle a cycle of length three. Now try some examples. We’re interested that repeats ad infinitum. nor shown any example that doesn’t — then you next get 4. Once you hit 4. this problem has the paradoxical property that. This answer is the next iterate. If even. triple and add one. As things are set up. our black hole called phi. In the example that would be ¿¢ ¾½. take half. Hmm now only in black holes. Success! In fact. 84 is ¾ ¢ ¾ ¢ ¿ ¢ . ECKER obtain our two golden numbers. Close Other examples of mathemagical black holes arise in the study of stochastic processes. If you do reach 1 — and nobody has either proved you must. The only non-odd prime is 2. Must you always reach 1? If you start with 5.50 M. then 4. because the largest odd factor in 4 is 1. then 1 again. If odd. and ¿¢½·½ . So. Consider the Collatz Conjecture.) Now triple the largest odd factor and then add 1. Start with a natural number. Because this is so easy to program. the numbers   ´Ò   ½µ ´Òµ. Pick the largest odd factor. Keep iterating. although one of the hardest to settle definitively. W. while  ¼ ½ is a repeller.

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

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

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. 4. reverse the digits of the result. 0). 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? E7.54 R. 10 pints. g. 6. 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. After going 10 miles in a straight line he finds himself back at camp. 8. 3. 0. Find the most efficient solution. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu. You have containers that hold 15 pints. 8. Calculate the expansion of this 26 term expression: ´Ü   µ´Ü   µ´Ü   µ´Ü   µ ´Ü   ݵ´Ü   Þ µ . Divide the number by 3. How long does it take? E5. 4. Find a nine-digit number made up of 1. Find the rule for combining numbers as shown below (e. subtract 1 and you produce the original number. 7. It is approximately 2244. 5. Through transferring liquid among the containers. 63 9 38 33 32 12 88 25 16 18 15 Ü 5 E3. and 6 pints. The 15pint container starts out full. 5. What is the number and what is the next greater number (possibly with more than three digits) having this property? . 3. Where on earth could he be? E2. HESS up and travels due North. 6. and the other two start out empty: (15. E6. My regular racquetball opponent has a license plate whose three-digit part has the following property. 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. 25 and 9 combine to give 16) and use the rule to determine Ü. I. 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. 7. 2.5 nautical miles from Los Angeles to Honolulu. 2 and 1. E9..

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

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

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

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

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

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

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

six black. e ) 2. I’ll let you off the hook. Find other such examples. . into five parts of equal area? (c) an equilateral triangle into five parts of equal area? H15. their ages are ” H14. d. b.What is the minimum cut-length needed to divide (a) a unit-sided equilateral triangle into four parts of equal area? (b) a unit square. (b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes? . 113.71828 . I.62 R. where he (1) picks out three socks at random. ) = 0. two brown) and ends up with four socks each Saturday evening. An approximation to is 355/113. My uncle’s ritual for dressing each morning except Sunday includes a trip to the sock drawer. A number may be approximated by a /b where a and b are integers. Since Smith missed the problem tonight and Jones missed it at the party two years ago.0107 . (a) Find the minimum a and b so that g (a. ) 0. The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem. (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. given the sum and product of their ages. HESS H12.” Your response is “No need to tell me more. Define the goodness of fit of a /b to x as g (a. b) Find the minimum c and d so that g (c.0107. where e = H16.a . and proceeds to explain: “I normally ask guests to determine the ages of my three children. The drawer starts with 16 socks each Monday morning (eight blue. Minimum Cutting Length. (a) The figure shows two Pythagorean triangles with a common side where three of the five side-lengths are prime numbers. with g (355. b.0107 0. x) = a bx . 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).

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

a rather boastful sort. swearing at the difficulties he had in crossing the channel of the haggard dagger digger. Before they were wiped out by a permeous accretion of Pfister-gas. Define ´Òµ where Ò ¾Ò · ¾   ¿ Ò Ò  ½ ´ ¼   Òµ   ½¾¿ and ¾ ½ ¾ ½ (a) Prove that ´Òµ goes to zero as Ò goes to infinity. Should Cosmic. Taurus. lying along the intersection of a plane with the satellite’s surface. HESS the curve of intersection of the krypton/fromage boundary and the planetary surface. they dug a series of channels in the satellite surface. was occupied by a race of knife-makers eons ago. I. (c) Find the smallest Ñ such that the magnitude of ´Ñµ is less than the magnitude of ´Ñ · ½µ. He had dug the entire length of a longer channel but never crossed the path of the first dagger digger.64 R. How long was the channel which the braggart haggard dagger digger dug? H20. He found nothing but the fact that the length of the channel was 30 spandrals. An even more curious fact is that Taurus is a torus (doughnut shape). The third student had explored a channel 50 spandrals in length. 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. A curious feature of these channels is that each is a complete and perfect circle.. (b) Find ´½¼¼¼µ to three significant figures. was also tired because he had thoroughly dug the entire length of the largest possible channel. a rather lazy fellow (a laggard dagger digger?). The fourth student. had merely walked the 60 spandral length of another channel. The second student was very tired from his work. . undertake the mining venture? H19. The fifth student. crossing the channel of the haggard dagger digger. a moon of a-Lyra IV (hieronymous). Inc.

(0. (7. 0). The license number is 741. ¼. ¾ gives Ø E5. 2. 0. Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. (5. E7. E3. 0). 0. (3. 0. ½ ¼½ or ¼ and ½ ¼½.654. 15 moves: (15. 6). 6). 5. it appears that the rule might be subtraction. But this is not right because 18 is not the difference of 38 and 16. so E8. (9. Instead. 4.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 0). (3. (11. 2). E2. 4. (5. (11. ¿   ½ ¼¼¼   are the smallest values satisfying the above equations.729. 8. Thus Ü ½½. 9. 0. 6. E4. 10. (3. 8. 0). Thus and Ò ¾½¼ transfers is the minimum.741. 2. E6. ؾ ¾¾ hours. (1. One of the terms is ´Ü   ܵ. 4. 10. E9. the next greater number is 7. (1. E10. 6). 10. 6). 6. (7. 4. (1.425. so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. 0). 2). 6. In each case the numbers combine by summing the products of their digits. 6). At first glance. The trick here is that 1 knot = 1 nautical mile per hour. 5). (3. 2. 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 . . 9). Anywhere within 10 miles of the north pole. 6). and Ü ½ . 6. 0). 7. The number is 381. . the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. 2). (9.

8 and 9 go across. 1 comes back. . 1 comes back. 1 and 3 go across. I. The figure below shows how to trap the knight after 15 moves. 1 and 6 go across. (a) 1 and 3 go across. 9. 1 and 4 go across. B comes back. A and B go across. (b) 1 and 2 go across. 1 comes back. 1 and 2 go across. A comes back. 1 comes back. Yes. 1. 3 comes back. E12.and 7 go across. and 10 go across. See the array below. E15. E13. HESS E11. 1 comes back. The path of their intersecting surfaces is a desired path. A and B go across. 8.66 R. E14. C and D go across. 6. 2 comes back. The solutions are shown in the following figures. Imagine intersecting the potatoes with each other.

The probability of an acute triangle is ¿ ¾½ or ½ . 0). the probability of a right triangle is ½ ¾½ or . 5). 11. 11. 156. 1. and the time is 8:24. ½¿ ½¿ ½ Ê 123. 0. (35. (36. 0. (24. (22. 11. 11. 0). 3). 6). 1). Cut B at that corresponding point. 6). 6). 6. 0). (25. 0). 0). 4). 5. (29. 126. (21. 10). 6. 10). 11. 0. (25. 168. 0. (24. 2). (31. 0. 128. Solutions to Medium Problems M1. 0. 9. 6). 3). 10). 0). 0). (22. 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. (23. 0. (25. 0). (a) 32 moves. 6).75 minutes. 0). 1). 5. 148. 0). (29. 6). (37. 147. 0. (19. (30. See the figure. 2. as seen from the list below. 11. 11. 136. (26. 11. 0. 0. 0). (26. 1. (20. 11. 146. (29. 10. 6. 6. 4). 9. (25. When C is consumed. 11. (25. (29. (25. (19. (18. (36. (24. 4. (31. (33. 11. 1).PUZZLES FROM AROUND THE WORLD 67 E16. (36. producing pieces C and D. (37. 10). 11. (24. (23. 0. 0. This is the shortest time interval that can be measured. (30. (26. 1). start burning C from both ends and D from one end. (29. (19. 6). 10). 178. 0. 127. (31. 167. 2. (19. (23. 1. (24. 0. M2. 0). 10). D will finish burning. 1. 0. 134. (34. 0) (b) 40 moves. 1. 1. (30. 0). (20. (21. (34. 6). 6). (35. 10). 11. 0). 4). 11. (18. . 10. (23. (30. 0). 0). 4. (30. 0). 1. (11. At the same time. 158. (25. 6. 10). 4. producing pieces A and B. 0). (35. 0). 125. (27. (18. Cut the lace in half. (24. 11. 0). 0. 10). Burn A from both ends. In 3. 1). (17. 5). There are 7 ¢ 6/2 = 21 choices using vertex 1. 0. 10). 1). 10). 2). (23. 0). 0. (31. 3. 0). 1. 0. (31. (23. light the other end of D. 0). (28. 4). (22. (33. 10). 124. 3. 6. 0). 0). E17. 1. If the hour hand is exactly on a second mark then the second hand will always be on the 12. 11. noting the point that burns last. 1). 10). (28. 3. 6). 0). (17. 5. (30. 138. (18. 10). (29. 0). 0. 4. 145. 0). 3. 5. 0). 0. 6). 0. 11.

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

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

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

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

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

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

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

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

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

which leads to Ô ¼ ½ ¿¿ ¼ .3586 Pairing probability 0.3074 1.2511 1. È ´ ¼   ¿¼µ Ô · ÕÈ ´ ¼ ¼µ. ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ . Let Ô the probability of the server winning a point. È ´¿¼   ¼µ ÔÈ ´ ¼   ¼µ.8286 0. ¾ H12. My uncle is expected to take the longest time to dress on Saturday and on Friday he is least likely to get a pair from the first three socks chosen. È ´ ¼   ¼µ ÔÈ ´ ¼   ¿¼µ · ÕÈ ´¿¼   ¼µ.2801 1. ·¾ Ô ´½ ½ Õ µ È ´¼   ¼µ .8027 0.7849 . At Ò games to Ò. which leads to Ô ¼ ½½ ¿ ¾ . Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1.2270 1. This leads to ¾ È ´ ¼   ¼µ Ô¾Ô Õ¾ .8163 0.PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit. and let Õ ½   Ô. È ´¼   ¼µ È ´ ¼   ¿¼µ when Ô¿   Ô¾   ¾Ô ½. H11.7805 0 . Ô  Õ At a score of Ò games to Ò (Ò ¼ to 5).7892 0. ¾ ¾ Õ´ È ´¿¼   ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ . · È ´ ¼   ¿¼µ Õ È ´ ¼   ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . A computer program was written to calculate the probabilities shown in the table below. Clearly ST is not perpendicular to the plane. È ´¼   ¼µ È ´¿¼   ½ µ when Ô ¾   Ô ¿. Let È ´   µ be the probability of the server winning the game when the server has points and the receiver has points.2069 1.

. 16). which must have been guessed by Smith. The product and sum could be achieved by (4. (b) Length ¾ ¼¾½½¾ ¿ as shown. we get (a) (b) H16. ¾   ½µ ¾. The product and sum two years ago could be achieved by (2. (a) Length ½ ¿ ¾½ ½ ¼ as shown. prime. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago. ´ ¾   ½µ ¾. Solutions occur for (b) ¾ ½. · ½. 7.093. (d) Yes. a=10. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½. ¼ ¼½¼ . There is only one set of ages that accomplishies this: (5. (b) 605 years. 15). . ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼.042. (a) 45 years. ´ prime. (c) Unknown. · ½.78 R.321 is the first of 5 in a row. ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ .546. ¿ ½¾¼. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. H14. Using a multiprecision continued fraction program.512. ´ ¾ · ½µ ¾ prime. HESS H13. H15.641. We look for cases where (1) the oldest child is under 20. H17. (c) 17. 6.444 years. . Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô .613. 12). I. which must have been guessed by Jones two years ago. (2) the younger two children have different ages. ¼ ¼¼ ½ ¼. 8.

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

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. As trarily large it still encloses the same poles as . 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 . I. HESS (2) Since ´  ÒµÞ is an entire function in the complex plane.80 R. Thus Þ ¾ Õ´Þ µ  ´Ò·½µÞ is regular within . it follows from Cauchy’s Theorem that ´Òµ where must enclose Þ ¾Ò · ¾   ¿ Ò  ½ and will be taken to be Þ  ´Ò·½µÞ  ½ ½   ´½ · Þ µ  Þ ¾ ½ ¾ ¼ Þ ´  ÒµÞ ´½ · Þ µ ·½ ¾. The only pole of Õ´Þ µ  ´Ò·½µÞ within is a double pole at Þ ¼. we expand and deform to a contour the figure. which still avoids enclosing the poles of Õ´Þ µ. (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 ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ. × ÒÝ Ý we can look for poles in the upper half-plane only and reflect each one into the lower half-plane. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to infinity. where Þ at ½ · Þ Ü · Ý. A pole of Õ´Þ µ occurs Þ . which reduces to ´½ · Ü · Ý µ´Ü   Ý µ ´ Ó×´Ò · ½µÝ   × Ò´Ò · ½µÝ µ where ´Ü¾ · ݾ µ ´Ò·½µÜ .8221528 Ý 7. Then Ý approximately equals   ½·ÐÒ´ µ .2238350 26.1512074 45.8790560 20.  ¾ × ´Ý · µ  Ü Ý .0888430 2. not including its pole at Þ ¼.5432385 32. For larger define ´¾ ·½ ¾µ .PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ.8505482 39. Call the Ø such pole in the upper half-plane Þ Ü · Ý .5012690 3. Ü 1 2 3 4 5 6 7 (6) The contribution to 2.4614892 13. The table below gives numerical values of the first few. (5) On Þ Ê for sufficiently large Ê.6406814 3.6745053 3. Since the equation in Ý is even. ¼ for all .4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ   Þ µÕ´Þ µ ´½ · Þ µÒ·½.2916783 3. Combining the poles in the upper and lower half-planes and some rearrangement finally produces ´Òµ where ½ ½  ÒÜ × Ò´ÒÝ   µ (a) Since Ü Ü¾ · Ý ¾ ܾ · Ü · ݾ Ø Ò´Ý · µ Ý  ÒÜ goes to zero as Ò goes to infinity.0262969 3. The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ.

E1. E4. Unknown. Vol. MIT 59. Problem 3a. 3. Reprinted with permission. edited by Alan Gottlieb. Unknown. p. 9. Unknown. M3. Schoenberg. Reprinted with permission. 302. From Nobuyuki Yoshigahara (private communication). p. ­ ­ ­ ­ ­ . 8. From Nobuyuki Yoshigahara (private communication). E3. All rights reserved. J. No. “Puzzle Corner. ´ ¼¼µ Sources ¼ ½ ¢ ½¼  ¾ ´ ¼½µ   ½¾¾ ¢ ½¼  ¾ All problems written by the author unless otherwise indicated. Submitted by S. Vol. Unknown. Reprinted with permission. Reprinted with permission.” Crux Mathematicorum. E10. October 1992. p. From Dieter Gebhardt (private communication). E7. Problem 7 from the 1980 Leningrad High School Olympiad. It is at Ñ ¼¼. October 1987. “Olympiad Corner. From Nobuyuki Yoshigahara (private communication).” in MIT’s Technology Review. 1983. E2. 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.” in MIT’s Technology Review. From Nobuyuki Yoshigahara (private communication). From Yoshiyuki Kotani (private communication). M5. MIT 55. E5. E12. No. J. page 178. Problem 1. M2. edited by Alan Gottlieb. M4. E14. HESS  ½ ½ (b) For Ò ½¼. All rights reserved. All rights reserved. From David Singmaster (private communication). Thus ´½¼¼¼µ is very nearly ½  ½¼¼¼Ü½ × Ò´½¼¼¼Ý½   ½ µ. E16. ´Òµ is dominated by the first pole and its reflected pole. private communication). E11. I. M6. edited by Allan Gottlieb. All rights reserved. “Puzzle Section. Einhorn and I. (a) and (b) original problems.” Pi Mu Epsilon Journal. Submitted by Lawrence Kells. “Puzzle Corner. “Puzzle Corner. E8. Unknown. 1985. Reprinted with permission of the Canadian Mathematical Society. From David Singmaster (attributed to Roger Penrose. p. Problem 3. E15. 1990.82 R. 10. MIT 55. E13.” in MIT’s Technology Review. From Nobuyuki Yoshigahara (private communication). 1987. 1992. May–June 1990. All rights reserved. From David Singmaster (private communication). Modification of Problem 1.

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

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

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

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

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

4. shown in Figure 10. as in the first solution in Figure 14. so first rotate the board so that the Chevron. 2 or 4 for a subscript. the center. If both Chevron and Hexagon are symmetrically placed. Then. 2. Figure 9. Coding the Chevron. 8. reflect the board left to right in order to bring it onto the right half. depending on how far it is from the center. f and m are equidistant from opposite sides of the board. is pointing upward. K. reflect the board if necessary to bring the Hexagon into the right half.5%. Read off the lowercase letter at the apex. D. or Ü the side of the board it is nearest to.) Next use Figure 9 to describe the position of the Hexagon. s. If the Chevron is central (positions b to g in Figure 8). and t are omitted. piece number 0.88 R. piece number ¾. since placing the Butterfly there does not allow a solution to be completed. A. but it’s clear that they can’t occur in a solution. piece number 2. Coding the Hexagon. This is indicated by a lowercase letter. The subscripts on C are about ½ more than their ¾ clock hour. A to G. is unique and requires no subscript. Figure 8. and G. 6. GUY to be made quickly. B. This will be a capital letter. F only even ones. if the Chevron is on the left-hand side of the board. and are given only 0. l. (The exact figure is 89. A and E have only odd subscripts. it is possible to place the Chevron in such positions. is that of its “body. The subscripts a. Note that the letters a and n are missing. together with an even subscript. of the Chevron from Figure 8. 0. Then append a subscript ¼ ½ Ü indicating its position on the clock. We won’t count a solution as different if it is just a rotation or a reflection. More than three-quarters of the positions found so far have the Chevron in position h. If all three are symmetrically placed. reflect if . reflect the board if necessary to bring the Butterfly into the right half. e. The position of the Butterfly.” the edge that bisects it.

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

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

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

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

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

by the Hexagon’s position Ø Ù Total 264 1094 124. GUY Figure 18. are 130 195 533 193 377 2214 111. president of Kadon Enterprises and discovered that she markets a set of “iamonds” . On the contrary.717 C 6675 D 7549 E 11. Six pieces meet at a point. for those who prefer their puzzles to have just one answer. Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to find all the solutions.489 B 15. classified according to the position of the Chevron. Their numbers. K. Ô Ü (3-symmetry).518 A 75. Figure 19. 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.94 R.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. I wrote to Kate Jones.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.

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

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

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



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!

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

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

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

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

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

joy and reverence that we sent one of the first sets to the inspiration of our enterprise. these are products made by Kadon Enterprises. Martin Gardner. Fives quints quintillions! It was with great pride. We had much to learn about marketing. keeping the checkerboard pattern (Figure 3). and the product names are trademarks of Kadon. not just a single product. And it was encouraging to us neophyte entrepreneurs when Games magazine reviewed Quintillions and included it on the “Games 100” list in 1980.½ Super Quintillions: 17 non-planar pentacubes (plus one duplicate piece to help fill the box). 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. 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. Figure 3. checkerboarded format. The puzzle challenge: Change any one into another in the minimum number of chess knight’s moves.. so we’d have to think of another name.THOSE PERIPATETIC PENTOMINOES 109 a registered trademark of Solomon Golomb. Here (chronologically) are the many guises and offspring of the dozen shapes that entered the culture through the doorway Martin Gardner opened in 1957. And so Quintillions begat a large number of kindred puzzle sets. ½All . and most of them sneak in some form of pentomino or polyomino entity among their other activities.

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

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

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

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

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

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

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

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

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

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

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

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

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

Over the intervening years. I soon became convinced that this was impossible. I was at a loss for any sort of formal proof.104 S. I had continued to . Alaska. knotted a cord around it. More recently. Next it appeared in a 1976 issue of a British magazine on puzzles and games. I published this simply as a curiosity in a 1974 newsletter (later reprinted in Puzzle Craft). he begged me for help. full of diagrams and such. somewhat to my surprise. purporting to prove that the puzzle was unsolvable. which then left them utterly baffled as to finding it. (In a recent survey of my puzzle customers. during an idle moment one day I formed some wire into the figure-eight shape shown below. Some readers misread that purposely vague write-up and assumed that it must have a solution. The proof appeared to be a rather complicated. and I did not spend a lot of time trying to digest it. When some of his customers started asking for the solution. The puzzle editor made the surprising observation that it was topologically equivalent to the Double-Treble-Clef Puzzle made by Pentangle and therefore solvable. Unlike the simple example shown above. I received a seven-page document from someone in Japan. and tried to remove it. 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). For a slight variation of this. but being a novice in the field of topology. Alas. To add to the confusion. since the Pentangle puzzle was. But a careful check showed that the two were not quite equivalent. the puzzle appeared in Creative Puzzles of the World by van Delft and Botermans (1978) with a farce of a convoluted “solution” thrown in for added amusement.) The story does not end there. they reported that topological puzzles in general were their least favorite of any puzzles I had produced. however. Royce Lowe of Juneau. decided to add the Figure Eight to the line of puzzles that he made and sold in his spare time. COFFIN Finally. none of these ideas enjoyed any success. some of these puzzles can be quite baffling to solve.

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

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

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

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

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

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

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

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

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

You are the inventor. all you have to do is shake your own hand. It is possible that one version of the toy works on the principle you mention. and the height through which the sand has fallen. “I am writing to put your mind at ease (on the impact theory posed by Altman). the acceleration of gravity. Could temperature and sand impact also be factors? Would he have to print a correction? He replied to Altman on September 12th: I am embarrassed to admit that your explanation may be right. having relied (unfortunately) on an account given to me by a friend who examined the toy in Paris. Reid.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. Maryland: Another solution to the hourglass science teaser is that the momentum carried by the falling sand causes the hourglass plus sand to weigh more than its static weight by an amount Ô¾ . and to suggest that you not publish a correction. A letter dated September 29 came from Walter P. perhaps depending on the temperature. where is the rate of the flow of the sand. Albert Altman wrote from the U. Hopelessly confused? Martin Gardner?! That is something that surely doesn’t happen very often. or perhaps still another one.S. One man wrote that he owned one of the Paris cylinders. and it didn’t last long. The principle is simply too clever to be true.” The Hourglass Letters While Hein was having his epiphany in Europe. but reported that his glass sometimes floated and sometimes sank. I have not yet seen the toy. Well. The height decreases due to the buildup of sand on the bottom of the hourglass and at a critical value the net force on the hourglass acts upward and rises. I am sure that your explanation was correct. also from the U. Gardner was getting letters in response to the September issue. and the other on the principle I suggested. at this point I am hopelessly confused. In short. Gardner was beginning to wonder if his friction explanation told the whole story. 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. but did not bring one back with him. Naval Ordnance Laboratory at Silver Spring.S. Naval Ordnance Lab. On September 6th.

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

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

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

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

Usually there is an “aha” reaction when students see that a cube can actually be constructed from the pieces. This puzzle. Three “flats” of size ¢ ¢ . One “Biggie” of size ¢ ¢ (where ¾ ). combinatorial puzzle can be constructed by acquiring twenty-seven uniform cubes of size ¢ ¢ and gluing them together to form the following eight pieces: One “Smally” of size ¢ ¢ (i. Three “longs” of size ¢ ¢ . Most Martin Gardner aficionados will recognize that the eight pieces model the situation represented by the binomial expansion ´½µ ´ · µ¿ ¿·¿ ¾ ·¿ ¾· ¿ Gardner calls such models of mathematical theorems “look-see” proofs.Cube Puzzles Jeremiah Farrell Binomial Puzzle A new. Before proceeding with this discussion the reader is urged to build a puzzle and attempt its solution. 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. rather amusing. if colored as above. a single cube).. 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. With proper insight it can be solved in a few minutes. This will depend on just what is meant by the words “essentially 145 . The problem then is to arrange the eight pieces into a cube so that opposite faces have the same color.e. In fact he has recommended that every teacher of algebra construct a set of eight pieces for classroom use. has exactly two solutions — each with a different set of three colors.

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

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

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

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

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

the results are shown in Table 1. Thus. A proof is given in Section 2. it becomes apparent that most of the 133. This was achieved through the use of a computer program. Since one person can assign the fourth and fifth colors to orange and red and another person to red and orange. FARHI a difficult exercise for a moderately proficient programmer to verify the results obtained. The first observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube. when two cubes of the same color are in position.691 combinations need to be sorted into disjoint sets. . The individual dicubes can be colored thirty-six different ways. then.152 S.105 of these combinations meet the three-cubes-of-each-color requirement. The third observation.691 combinations into 148 disjoint sets is described in Section 6. the location of the third cube is determined. The second observation is that for a cube to be assembled with the nine colors on each face. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. is that there are numerous solutions. Since during the course of trying to solve this puzzle a conflict often occurs.105 combinations are isomorphisms. Twelve of these can be selected in 1. to separate these cases into disjoint sets of isomorphisms. this rule then tells the experimenter to backtrack. The previous question now becomes more interesting. but not in the center of a face. only 133. only thirty dicubes are required. in which the number of isomorphisms using the set of thirty-six dicubes is also determined.700 (36!/12!24!) different ways of which. obtained after some experimentation.251. as determined in Section 4.677. and the most difficult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1. It becomes necessary. The question then arises: How many solutions are there? Most intriguing about this puzzle. some with the tricube on the edge and some with the tricube through the center of the cube. and only 10. as is shown in Section 5. The method of sorting the 10. Fortunately. none of the nine planes may contain two cubes of the same color. A proof is given in Section 3. The final part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations.

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

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

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

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

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

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

and 9 one can logically deduce the first colors by the three . 8). For example. 8). 4). 3. 6. Identification and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. 6).691 color combinations. (5. (2. 6). a computer program established 133. (1. 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. Identification of the second color of these ordered pairs is sufficient to define all the colors.105 possible color combinations. 2). 6. only 10. 9. This results in requiring only thirty cubes and. (5. using the same algorithm as before. 4. 9). 9. (7. 9). 3). 8. (4. Reduction of the Number of Isomorphisms The table in Section 4 is simplified by restricting the selection in the first two rows as shown below. 6.THE NINE COLOR PUZZLE 159 First Color 1 2 3 4 5 6 7 8 Second Selection 2 3 4 5 6 7 8 3 4 5 6 7 8 4 5 6 7 8 5 6 7 8 6 7 8 7 8 8 9 9 9 9 9 9 9 9 The process continues: color 4 is combined with colors in the fourth row such that there are three cubes with color 4. 5. the two new colors are identified as colors 6 and 7. 7. given the second numbers 2. Similarly if color 2 is not associated with colors 3. (4. (7. color 5 is combined with colors in the fifth row such that there are three cubes with color 5. Using this algorithm. etc. 5). 5. 8. or 5. 7). 4. (3. (6. 9) and (8.

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

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

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

For the expert. The object of the puzzle in each case is to move block A to the bottom right corner. The puzzles are both rated as difficult. however. . and the second puzzle uses block 2(b). Once this has been mastered. Blocks 1 and 5 have both channels (in the form of a cross) and can go anywhere. only to get hopelessly stuck…. Figure 3.164 E. they must only move in a cross-shaped area. The delight of the first puzzle is that it is very easy to move block A to just above the bottom right corner. These blocks. a plan can be made as to which blocks have to be moved so as to allow block A to pass. the shortest known solutions are 50 moves for the first puzzle and a staggering 70 moves for the second. and yet so far! In puzzle 2 it is quite a task to move block A more than a square or two. Both solutions are believed (but not proved) to be minimum-move solutions. which has to be overcome…. 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. It is quite an achievement just to solve them. Blocks 4 and 2(a) have a single vertical channel and can only approach the corners from a vertical direction. as well as block A. This causes something of a traffic jam. Since they can’t go into the corners. The real “problem” blocks are block 7 in the first puzzle and blocks 7 and 2(b) in the second. can only move freely up and down the center of the puzzle. The first puzzle uses block 2(a). or to get anywhere at all! Hints for Solving In both puzzles the “nuisance” blocks are 3 and 6. Blocks 3 and 6 have no channels and cannot go into any corner. So near. The Puzzle Figure 3 shows the start position. HORDERN direction. the second being the harder of the two.

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

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

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

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

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

¢ ½ . and ¢ ½¿. ¢ . ¢ ½¼. There are 3. This leaves 4 pieces that might be placed sideways.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. One or two individual layers can be constructed. all very similar. ¢ ¾½. I painted pieces of an actual model to match the display. The result was a fascinating demonstration of how a computer can be used to solve such a puzzle. When running these programs on more powerful computers. Bill Cutler Puzzles): Box: ½ ¢ ½ ¢ ½ Pieces: ½¿ ¢ ½½¾ ¢ ½ ½. ½ ¢ ½ ¢ ¾ ¢ . the difference between the three boxes is stunning: The first box can be completely analyzed in a small fraction of a second. I wrote such a program on my first computer. º Boxed Box (23) (Cutler. ½¼ ¢ ½½. The analysis was done on about 20 powerful IBM workstations. and they all have three of these four pieces placed sideways. Since all the pieces are of the same thickness and the box depth equals three thicknesses. ½ ¢ ¾ ¢ ¾. about once a second a piece is added or removed from the box! Using compiled BASIC. ¾¼ ¢ ¼ ¢ ¾. The Commodore 64 is such a wonderously slow machine — when running the program in interpreter BASIC. ¢ ½¿. The second box was analyzed in about a minute of mainframe computer time. There are four solutions to the puzzle. . I did a complete analysis of the third box. Of the 18 pieces. 10 are too wide to fit into the box sideways and 4 are of width 5. The solution involves use of the following obvious trick (is that an oxymoron?): Some piece(s) are placed sideways in the box. not counting rotations and reflections. or the equivalent of one year on one machine. ½ ¢ ¢ ¼.450. By the end of the runs. the rate increases to 40 pieces/second. ½ ¢ ¿ ¢ . The programmer can use algorithms that are used for pentominoe problems. ½½ ¢ ½½ and two each of ¢ . the widths and lengths are ¢ . ¢ . º Parcel Post Puzzle (18) (designer unknown. ¢ ½ . In early 1996. the machines had constructed 2 1/2 trillion different partially filled boxes. copied from a model in the collection of Abel Garcia): Box: ¢ ½ ¢ ¾ Pieces: all pieces are of thickness 2 units. ¾½ ¢ ¢ ½ ¼. but the process cannot be completed. ¢ ½½. The program displayed the status of the box at any instant using color graphics.480 solutions. The total CPU time used was about 8500 hours. ¾¢ . which is no good for this purpose. it is tempting to solve the puzzle by constructing three layers of pieces. but there are more efficient algorithms that can be used for block-packing puzzles. ½ ¢ ¼ ¢ . ¢ ½ . 1978. ¢ ½¿. a Commodore 64.

¾¿ ¢ ½ ¢ ¿. 10. º Hoffman Junior (8) (NOB Yoshigahara. ¿¿ ¢ ¿ ¢ ¿ ¢ ½¾½.172 B. I like the dimensions above because it tempts the solver to stack the pieces three deep in the middle dimension. Bill Cutler Puzzles) Box: ¼ ¢ ¾ ¢ ¾ Pieces: ¢ ½ ¢ ¾ . ¿½ ¢ ¢ . ½¼ ¢ ½½ ¢ ¾. ½½ ¢ ½¾ ¢ ¾. Simplified (15) (Cutler. version by cutting some of the pieces into two or three smaller pieces. basic. Hikimi Puzzland) Box: ½ ¢ ½ ¢ ½ Pieces: Two each of ¢ ¢ ½¼. ½¾ ¢ ½ ¢ ½ . Holes. ¢ ¢ The dimensions of the pieces are all different numbers. ½½ ¢ ½ ¢ ¿¼. ¾ ¢ ¿¼ ¢ ¢ ½¿ . 1986. ¢ ½¼ ¢ ½½ 11. The pieces fit into the box with no extra space. 1981. this puzzle still fascinates me. See [1] or [3] for more information. 1976) Box: ½ ¢ ½ ¢ ½ Pieces: (27) ¢ ¢ This sounds like a simple puzzle. Holes. ¢ ¢ . 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. Pieces the Same or Similar º Hoffman’s Blocks (27) (Dean Hoffman. There are 21 solutions. ¢ ¾¼ ¢ ¾¼. Almost 15 years later. The dimensions of the pieces can be modified. none having any symmetry or pattern. ½½ ¢ ½ ¢ ¾ . ¿ ¢ ¾ ¢ ¼. The net result was a puzzle that is extremely difficult. ¢ . ½¼ ¢ ½ ¢ ¿½. ½½ ¢ ½½ ¢ ¾ . ½¼ ¢ ½¾ ¢ ¾ . ½ ¢ ½ ¢ ¾½. but it is not. They can be any three different numbers. ½¼ ¢ ½ ¢ ¾ . ¢ ¼. ¾ ¢ ¢ ½¾¿. The smallest number for which this can be done is 23. where the smallest is greater than one-quarter of the sum. The box is a cube with side equal to the sum. Pieces Different ½¼º Cutler’s Dilemma. The extra space makes available a whole new realm of possibilities. ¢ ½¼ ¢ ½½. CUTLER ¾¾ ¢ ½¼ ¢ ½¿½. There are many other 23-piece solutions that are combinatorially different from the above design. ½½ ¢ ½ ¢ ½ . ½¼ ¢ ½ ¢ ¾ . ½ ¢ ½ ¢ ¾½ The original design of Cutler’s Dilemma had 23 pieces and was constructed from the above. ¾ ¢ ¿ ¢ . . ¢ ¢ ½½. ¿ ¢ ¾ ¢ ½¾ . See [4]. ¿ ¢ ½½¼ ¢ ½¿ .

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

.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.

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

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

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

Required for INT ASS PAT. 2-D on 3-D. This has resulted in an unwieldy list. Required for INT JIG ASS PAT RTF SEQ FOL. Examples of approved structure adjectives that may be used include ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ Identical: All pieces identical Cartesian: Three mutually perpendicular axes Diagonal: Pieces rotated 45 degrees along their axis Skewed: Like a squashed puzzle Polyhedral Geometrical: Non-Cartesian but geometrical structures Organic: Amorphously shaped pieces Ball: Pieces made from joined spheres Rod: Square. Required for SEQ. the type of structure. where XXX is the main class and YYYY is the sub-class. Examples of puzzles in each class are given in the right-hand column. Whether needed. 3-D. while not strictly puzzles. Number of Dimensions. DALGETY AND E. and the standard Stellated Rhombic Dodecahedron has a six-piece Diagonal Cartesian structure. not needed. Linked: Pieces joined together by hinges. Thus a standard six-piece burr uses square rods in a Cartesian structure. etc. 2-D. 3-D. ¯ Proposed Developments Prior to 1998 the subclasses attempted to incorporate the number of dimensions. and any group/non-group moves. Required for INT JIG ASS PAT. Number of Pieces. We hope to reduce the number of sub-classes substantially in the near future. which. triangular. etc. HORDERN Ephemera (EPH). This category has been included because most puzzle collections include related ephemera. 2-D. 2-D to 3-D. and 4-D. Type of Piece Structure. need to be classified as part of the collection. hexagonal. A detailed classification with second-level classes is given in the separate table.178 J. Group Moves. . or partly used in solution. Puzzle Class Abbreviations (PZCODE) are standardized to a maximum of eight characters: XXX-YYYY. strings. ribbons. We are now working on defining what should be included automatically in the subclasses where they are relevant.

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

” “Theo der Turnier.” Engel’s “Black Box” Bagatelle “Frying Pan. HORDERN Detailed Puzzle Classification Class DEXDEX-UNCA Description Dexterity Puzzles Dexterity or other physical skills in their solution Example Cup and Ball. bridges. jumping beans. plain balls into holes Dexterity with sundry objects and/or obstacles Liquid objects Dexterity in liquid Indirect viewing Mechanized Using tools and magnetic tools Route following dexterity Objects concealed from view Electrica and electronic dexterities Pinball-related dexterities Other dexterities. avoiding objects. pneumatic operation Route-Finding Puzzles Route-finding with changing path and/or Complex Traveller Route-finding step mazes Unicursal route-finding Shortest route Complex route mazes with special objectives Four Generations “Ball in Block. number totalling mazes. “Le Pendu.” “Tandem Maze” (complex).” “Yankee.” puzzles using tops Pentangle “Roly-Poly” puzzles Ramps. etc. visiting places en route. DALGETY AND E.” colour mazes. “Pike’s Peak” Icosian Game. Konigsburg Bridges “Worried Woodworm.180 J.” Tomy’s “Crazy Maze. Mercury manipulation Water-filled puzzles View by mirror Tomy’s “Pocketeers” DEX-BALL DEX-SDRY DEX-LQOB DEX-INLQ DEX-MIRR DEX-MECH DEX-TOOL DEX-RTFL DEX-HIDD DEX-ELEC DEX-PINB DEX-OTHR RTFRTF-CHNG RTF-STEP RTF-UNIC RTF-SHOR RTF-CPLX Dexterity. Most hedge mazes RTF-2D Route path) mazes 2-D (any . “Bootlegger” (complex) Ring and hole mazes.

knives.” “Nine of Swords. 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. “T” puzzle Polycubes. cast “ABC. puzzle rings Hess wire puzzles. Soma. purses Padlocks Chippendale tea chests. “Tak-it-Apart. Oscar’s “Dovetail. “Mayer’s Cube.” Cutler’s burr in a glass Coffin’s “Saturn” JWIP and keychain animals. gears.” “Margot Cube. Oskar’s keys. poison rings Nuts and bolts. checkerboards. “Hoffman” cube. cutlery.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.” Chinese rings.” keychain cars “Plato’s Plight” OPN-OTHR INTINT-BOX INT-CART INT-POLY INT-SHAP Interlocking Puzzles Boxes that disassemble Cartesian (has parts along three mutually perpendicular axes) Interlocking polyhedral Other rigid shapes INT-TENS ASSASS-MAT ASS-2D ASS-CART Tensegrity structures in which compression and tension elements separate Assembly Puzzles Match stick puzzles and tricks 2-D Assembly 3-D Cartesian Assemblies Tangram. Dalgety’s “Devil’s Halo” Leather tanglement puzzles Boxes. pens. O’Beirne’s “Melting Block” . berrocals.” “Hazelgrove Box” Strijbos aluminium burr box Burrs. Pentominoes.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

GOSPER Times 100–105: .204 B.

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

(Mine runs only on an Apple II. or 47. I built a p155 oscillator in which 20 Bs travel around a track of length ¿½¼¼´ ¾¼ ¢ ·¾ ¢ µ generations. a few oscillators with periods 5 (including some with useful sparks) and 6. speed c/4 orthogonal spaceship. More recently. Hartmut found another with speed 2c/5. we put several copies of the B in the track to reduce the period to one of those mentioned. so we also get glider guns of those periods. based on construction universality. in which a B heptomino is forced to turn a 90 Æ corner. Due mostly to the work of Buckingham and Wainwright. but 5. and one period 5. For the multiples of 8. speed c/2 orthogonal spaceships.) The multiples of ¿¼ and ½¼¼ and the · ½¾¼Ò and ½¿ · ½¾¼Ò are based on gliders shuttling back and forth between either oscillators (or periods ½ ¿¼ and ½¼¼) or output streams of glider guns (of periods ¿¼ and ¼¼ · ¾¼¼Ò).. one period 4. A period 150 gun fires toward a distant pair of pentadecathlons. he and Hartmut Holzwart have been producing oodles of orthogonal spaceships with speeds c/2. 6. we now have nontrivial examples (i. infinitely many period 3. speed 2c/5 orthogonal spaceship.e. it consists of a still life (“eater3”) and two spark producing oscillators. 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. 6. the 73 gen turn can emit a glider. We also have an argument. and 47 are the only ones we know with the right sparks. and c/4. speed c/4 diagonal spaceship (not the glider). GOSPER resembling your initials. David Bell wrote a similar program that runs on faster machines. The · ¾ Ò ¾ · ¾ Ò ¼ · ¼Ò ¾¿¼ · ¼Ò ¾ ¾ · ¿ Ò and · ¿ Ò use a device discovered by Wainwright. (These are used in some of the large patterns with unusual growth rates mentioned later. one period 4. The turn can take either 64. For example. by combining them we can build a closed loop whose length is any large multiple of 5 or 8. which reflects a symmetric pair of parallel gliders.206 B. Another amusing oscillator uses the glider crystallization that Bill mentioned in his centinal letter a few years ago. which implies that all sufficiently large periods are possible. or 73 generations.) Periods ½¿ · Ò and ½ ¼ · Ò use a mechanism developed by Buckingham. 65. (No doubt you figured that out for yourself long ago. 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. Also. (Any greater nonmultiple of 4 would also work. The first glider to hit them is reflected . of periods 5. JHC!). speed c/3 orthogonal spaceships. infinitely many period 2.) Lately. c/3.) Construction of Oscillators.

The other guns are all pretty big. centinal-based ¼¼ · ¾¼¼Ò. its bounding box is 13. there’s now a fairly small period 44 gun that Buckingham built recently. The period 94 gun is 143 by 607. the cycle repeats every ¢ ¾ generations. The period ½¿ · Ò guns based on Buckingham’s B heptomino turns are also smaller.584 by 12. the process begins again. New Glider Guns. with a total of 36 AK47s. It emits a glider. A honey farm starts to form but is modified by an eater and a block. The gun itself has a larger period. they don’t work with period 30 streams. and it begins to decay. which hit eaters.STRANGE NEW LIFE FORMS: UPDATE 207 180Æ and collides with the second to form a honey farm. Fortunately. Rich has already mentioned this. I’ll just add that some of Buckingham’s syntheses of still lifes and billiard table oscillators are aweinspiring.112. based on a period 72 device discovered by Bob Wainwright. the resulting period must be a multiple of 8. emit gliders that can crash to form MWSSs. Buckingham has also found extremely weird and elegant guns of periods 144 and 216. Here’s how to turn 2 MWSSs into a glider: In Winning Ways. forming gliders that stabilize the ends of a row. It’s fairly easy to get any period down . In addition to the long-familiar period 30 and period 46 guns. A close pair of AK47s can delete each other’s traffic lights. there’s a period ½ ´ ¢ ¿ · ¢ µ gun that’s 119 by 119 and a period ½¿ ´ ¼ ¢ ¿ · ½ ¢ µ ¾ gun that’s 309 by 277.) It’s also possible to build a gun that produces a glider stream of any period ½ . an eater stops its growth. In the period 94 gun. there are other ways to get any multiple of 30. 11 gliders add a pair of beehives to the crystal. forms a traffic light. Using normal kickbacks. When they’re all gone. I’ve found some eaterassisted kickbacks that give any even period. of size 149 by 149. it uses various mechanisms to interleave larger period streams. so we can build a long row of them that is unstable at both ends. it’s based on the “AK47” reaction discovered independently by Dave Buckingham and Rich Schroeppel. Subsequent gliders grow a crystal upstream. If you delete the traffic light. (Sadly. based on the period 100 centinal. you (JHC) described how to thin a glider stream by kicking a glider back and forth between two streams. Buckingham mercifully outmoded it with a comparatively tiny. two such rows. two gliders delete one pair of beehives. The largest by far that’s actually been built is Bill’s period 1100 gun (extensible to ¼¼ · ¾¼¼Ò). Glider Syntheses. When the crystal reaches the guns. For example. and then starts forming another honey farm in a different location.

Here it’s shown deleting a glider. (I built a pseudo-period 15 gun based on his reactions. so we have guns that produce these as well. and 17. For example. .” (Less than 14 is physically impossible. The smallest known uses a period 8 oscillator (“blocker”) found by Wainwright to delete half the gliders from a period 60 stream.) New Year’s Eve Newsflash: Buckingham has just announced “. and heavyweight spaceships from gliders. . 300. We can also synthesize light-. 16. 210. 10. it escapes: There are also combinations of two period 30 guns that give fairly small guns with periods 90. . Universality implies the existence of Life patterns with various unusual properties. middle-. 120. and Buckingham has found clever ways to get 15. if the glider is delayed by 60 (or. its bounding box is 373 by 372.208 B. more generally. . 150. I have a working construction to build a P14 glider stream by inserting a glider between two gliders 28 gens apart! . (At least they seem unusual. This will make it possible to produce glider streams of any period. based on the . Ò · ) generations.) Returning to Hickerson’s account: Many of the large constructions mentioned later require a period 120 gun. and 360. this reaction leads to a period 30 LWSS gun: Large Constructions. GOSPER to 18 this way (period 23 is especially simple).

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

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



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

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

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

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

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

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

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

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

but it is traditional for the parameters and the solutions. The general problem ´ µ leads to the system of equations ´½µ ´¾µ Ü· Ü · ´Ý   µ Ý· Ý · ´Ü   µ The solutions are readily computed to be Thus. The criterion allows us to pick arbitrary and . this doesn’t hold. ´ · ½µ´ · µ ´ · µ´ · ½ · ½µ.222 D. . which is if and only if   ½ divides GCD ´ · ½µ´ · µ." Many versions of this problem occur. though. I am indebted to S. and then determines which values of and give integral solutions. assuming ½. Hence a natural question is to consider the diophantine question Which integer problems have integer solutions? In the early 1990s [1]. I find it particularly striking that and only enter via the sum · . where ´ · ½ · ½µ denotes the GCD of · ½ and · ½. I’d have times you. and I thought that the integrality of this was equivalent to integrality of Ü and Ý. the age problem. This also divides ´ · ½µ´ · ½µ   ´ · ½µ   ´ · ½µ   ½. Here I present a reasonably simple criterion that allows one to generate all the integer problems with integer solutions. which I would not have predicted. as before. I said I knew of no way to decide this that was simpler than actually seeing if the solutions were integers. Looking for a more symmetric solution. This still doesn’t give us a very simple or satisfactory criterion. The general solution of the problem gives somewhat complicated expressions for Ü and Ý. I am also intrigued to see that any and any can be used. SINGMASTER times you. The values of Ü and Ý are integers if and only if   ½ divides both ´ · ½µ´ · µ and ´ · ½µ´ · µ. Parameswaran for the letter that inspired this investigation. but somewhat more complex than. I first noted that Ü · Ý has a symmetric expression. which is similar to.” and the second responds: “And if I had from you. unfortunately. so we only need to check one of the second terms in (2). to be integers. this false trail led me to the following simple result. say Ü Ý. and it is obvious from (1). Ü and Ý are integers if and only if the second terms in (2) are integers. assuming ½. that Ü is an integer if and only if Ý is an integer. One can see from (2). however. Luckily. Consider ´ · ½ · ½µ. 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.

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

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

3. Sridhara [18]. 11. I quote a version that appears as No. 2) I-(9. 4) (2. 2. 10. 2. hence the above title. 3) (10. 4. 250 c. 2. 5. 10. 6) I-(7. with the problem statement giving the 6 as a 7 Fibonacci. and to the youngest ten apples. 6) I-(7. 850 c. 4. 7) I-(7. 2. 3. certain apples. 2. 11. 2.C. The problem is treated in one form by Mahavira [15]. 50 in the first printed English riddle collection “The Demaundes Joyous. 6) I-(7. 3. but this form is less clearly expressed and has infinitely many solutions. 9. 11. 1) (30. 3. although the material is believed to be based on older Arabic sources. 5. 6) (1. 2. 1. to sell. 6. 5. 11. 4. 5) (100. 7) I-(7.” printed by Wynken de Worde in 1511 [9]. 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. which daughters he delivered. 11. noting that this system is inconsistent Fibonacci. 5) (2. 50. 1. so I will first describe the later form which is first found in Fibonacci [11]. which is rather a mouthful. 3. 5. 23. c. (1. 3. and to the second thirty apples. 8. 11. 5) c. 10) (7. 10. and all these three sold in like many for a penny. Now how many sold each of them for a penny? . 4. 2.SOME DIOPHANTINE RECREATIONS 225 from the early fourteenth century and hence after Fibonacci. 2. with the problem statement giving the 6 as a 7 Fibonacci Fibonacci. 1. A man had three daughters of three ages. 9. 9. 510 9Ø century c. 5. 510 c. And he took to the eldest daughter fifty apples. 9.5) I-(25. 9. 4. 7) II-(1. 850 c. 8. 10. 4. there seems to be no common name for it. 5. and Bhaskara [4]. 3. 22. and brought home in like such money. 300 B. 1. 0. 3. done two ways Selling Different Amounts at the Same Price Yielding the Same Although this problem is quite old.

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

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

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

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

Tartaglia’s problem 139 is the one with two sizes of pearls and is (10. ½ ½ . These occur when ½ ¼.. we have ¾Ã   ½ solutions with zero values. Most of the examples in the literature have the in arithmetic progression (AP). we have à ½ Æ ¿ Æ· ½. 20.230 D. If ½ is odd. Substituting (11) into (8) and summing gives us the number of solutions. we find that this occurs when   ½ ­Ò . We can count the number Æ · of positive solutions by finding the number of solutions with a zero value. so we can write ½ ·´   ½µ . 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 . and when Ò ¼. SINGMASTER Then (10) becomes ´½¾µ ´½¿µ ½  ½ ­Ò The upper bound may not be an integer. Rearranging. when the upper bound for given in (13) is an exact integer.e. We then have à ½ Æ ½´ ½ · ½µ ¾ Æ· ´ ½   ¾µ´ ½   ½µ ¾. ­Ò   Ò   ½: ´½ µ à A number of early cases have Ò ¾. In that case. Looking again at our example... Subtracting ¾Ã or ¾Ã   ½ from Æ yields Æ· . unless it happens that ½ «¾­Ò ­¾ . which is trivially an AP. ½ A few longer APs occur — I have seen examples with Ò . then ½ we get à ´ ½   ½µ ¾ Æ ´ ¾   ½µ Æ· ´ ½   ¾µ¾   ½ . ´½ µ Again.e. i. i. A little thought shows that then   is the common difference of the progression. The majority of examples are APs with Ò ¿. A few cases occur that are not APs (see below). so we let Ã Æ  ½ ­  Ò à where [ ] is the greatest integer function. Then we have ­Ò ´Ò   ½µ . From (15). and then à ½ ¾ Æ ¾ Æ· ´ ½   ¾µ¾ . perhaps as intended by De Lagny. i. Æ· ½¼   ¾ ¢ ¾ . when two zero values occur in the same solution. in our example. 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. Hence we normally get ¾Ã situations with a zero value. 90). . ½ «¾­Ò ­¾ by (7).e. All but one of these have ½ even.

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

32. 1500 1513 1515 1521? 1556 1556 1556 1612 1725? 1874 1874 1874 1874 1893 1905 1924 Source Fibonacci Fibonacci Fibonacci “Abraham” Munich codex14684 Widman Widman Widman Pacioli Blasius Tagliente Ghaligai Tartaglia Tartaglia Tartaglia Bachet Ozanam from De Lagny Labosne Labosne Labosne Labosne Hoffmann Dudeney Glaisher (171. 90) Date 1202? 1202? 1202? c. 16) (16. 170) (305. 26) (20. 20. 1 given. 2 given. 49) (3. 33) (10. 196) (45. 81) (55. none given. . 1 given. makes this more specific by saying they ‘sold a part in lots and disposed of the remainder at one for five panas. One can scale the set . 454.232 Version (10. 60) (117. 40) (10.400) (25. 80) (10. 4 given. one with fractions. 50) (30. 20. SINGMASTER Solutions in Source 5 given. 36) (25. 1) (7. 40. At this point. 119. 40. Sridhara finds a one-parameter solution set. 60) (10. ¬ . 1 given. Bhaskara makes the following comment: “This. 1 given. 33. 1 given. . though it is not as easy to see how to do this as in the Western version. 68. 36) (78. 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. 901) (10. in some way to reduce the solution set to two dimensions. 6 given. so we again have a three-dimensional solution set. 1 given. 48. 1 given. 40. 55) (55. 56. 30. none given. 55) (78. all given. (20. 140) . 752. 1) (100. 136) (3. 1 given. 30) (18. 29.9) (100. 1 given. (20. 30. 100) (45. 36) (30. 12. 603. 25. 11. 36) (27.6) # of Solutions (55. 20) (64. 60) . 50) (11.4) (70.552. 15) (31. 82) (17.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 81) (10. 1 given. 50) (10. 37) (27. 30) (8. 16) D. 30) (12. 1 given. 20) (10. 33) (20. . 1 given. 1 given. 21) (225. 55) (16. 40) (18. 17. 32) (12. 1 given with fractions. 30. 35) (11. 1 given. 1300 1300s 1489 1489 1489 c. 1 given.

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

A. reprinted by Blanchard. 21.” [2] Ayyangar. 242–244. which may be the same as the 1815 edition. Nunn et al. 24. (There have been several reprints. pp. Gordon Fraser Gallery. London. 18. reprinted by Dover. p. Gaz. London. [7] Chiu Chang Suan Ching. Heath. “A classical Indian puzzle-problem. 3rd ed. 76. John Murray.900 and of 3294. See No. 1813. 86–88. . 150 B. Bijanganita. 1964. Indian Math. Lyon.. Wynken de Worde. 2nd ed. Lucas N. [8] Diophantos. 12. nos. Lyon. London.” In 1815. pp. In: T. 1805. By the way. pp. and the 13Ø edition. Krishnaswami. PrenticeHall. Gauthier-Villars. 1824. 1879. Book I. London.. 16: “De duobus hominibus boves ducentibus — Two men leading oxen. annotated by David Singmaster and John Hadley. Revised by A.. Bîjaganita. 170. [5] Bonnycastle. v. 5th ed. “Nine Chapters on the Mathematical Art.. 1884. Vogel. The first nine editions appeared “without any material alterations. 1976.234 D. 1959. 1968. 102– 126. Prob. 1874. including Sandig. Algebra.. Bhaskara solves his example by taking ¾. 1910. 178–186. Analysis similar to that done before shows that there are then ½ · ´ ¬ ½   Ò µ  ¬ solutions. 1973. Diophantos of Alexandria. 3292.” c. I examined the 7 Ø edition. et al. Translated into German by K. SINGMASTER For given and ¬ satisfying ´ ¼¼¼ µ and ´ ¼µ.. Labosne. Mar 1992. Johnson. he produced a 10 Ø ed. The Historical Roots of Elementary Mathematics. [9] The Demaundes Joyous.. Prob. 33. P. References [1] Alcuin. 1923–24. Paris. 13. 6.. Facsimile with transcription and commentary by John Wardroper. 2nd ed. P. 3200 and 16. Prob. Translated by John Hadley. London.) Chap. Oxford University Press. L. 19. 122–126. Claude-Gaspar. and Other Places of Public Education. 134–136. 5th ed. pp. Vieweg. Problemes plaisants & delectables qui se font par les nombres. Problems to Sharpen the Young. giving total numbers of 3295. trans. so ..” J. there are only a finite number of solutions. [4] Bhaskara. 214–223.. [6] Bunt. 54. pp. In: Henry Thomas Colebrooke. 1511. with Arithmetic and Mensuration from the Sanscrit of Brahmegupta and Bhascara. Math.(475). Paris.. Nos. with Notes and Observations. 10. Soc. pp. 15. A..C. 1624. 1782. 1st ed. Rigaud. “an entire revision of the work. 15. Neun Bucher arithmetischer Technik. designed for the Use of Schools. Wiesbaden. 1817. Braunschweig. [3] Bachet. 4th ed.” which “may be considered as a concise abridgment” of his two-volume Treatise on Algebra. 106–115. pp. 24. aka Bhâskara. J. pp. 1150.   and ¬ ½½¼. An Introduction to Algebra. Rigaud. John. 4392.470 [4]. J. 1612. H.

Lucknow University. 1010. Hildesheim. In: Scritti di Leonardo Pisano. London. 136–139. (1202). Supp. Translated by M. Boncompagni. 4 Ø edition. Berlin. 1–131. Mahavira. Jacques. Tartaglia. Cambridge. p. Bonwick. Paton.D.SOME DIOPHANTINE RECREATIONS 235 [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] 1971. 900. 1725.000 Puzzles. Abo^ Beqr Mohammed Ben Alhacen. . The Greek Anthology. 1228. 50. 298–302. Prob. R. Jombert. 44–49 and 94. 1916–18. Reprinted by Georg Olms Verlag.” L’Imprimerie Imperiale. 850.” c. edited and published by B. Transcribed and translated by K. Ragacarya. book 16. Nicolo. W. New edition. 79–80. pp. 53. prob. 90. pp. aka Leonardo Pisano. English version: Recreations Mathematical and Physical. prob. “On certain puzzle-questions occurring in early arithmetical writings and the general partition problems with which they are connected. Paris. prob. reprinted 1976. Reprint. pp. edited by Grandin. VIII. 60-62. pp. Paris. Ozanam. 1556. and Heinemann. V. 26–27 of the transcription. Amsterdam.. aka Mahâvîrâ(cârya). 1914. Tropfke. Translated by W. 1853. Government Press. Alkarkh. pp. prob. 1916. pp. 6 of the facsimile. Lucknow. Part 1. Sam Loyd’s Cyclopedia of 5. R. This is the oldest riddle collection printed in England. 1976. and Helmuth Gericke. Rome. 5. Revised by Kurt Vogel. 1912.” Messenger of Mathematics. Glaisher. Vol. J. 1923–24.. Lamb Publishing. et al. 24. aka Srîdharâcârya. Loeb Classical Library. Liber Abbaci. ex. pp. 24. Venice. 1708. Madras. surviving in a single example in the Cambridge University Library. Edited into French by Franz Woepcke as “Extrait du Fakhrî. Curtio Troiano. 1694 (not seen). Fibonacci. Karin Reich.. 28. Menge. Tricks and Conundrums. Edited by J. 286–287. Pâtîganita. 1959. London. Mass. Harvard University Press. L. Geschichte der Elementarmathematik. Paris. General Trattato di Numeri et Misure. Sridhara. vol. “Ganita-sâra-sangraha. Paris. Shukla. pp. p. 1857. Pinnacle or Corwin. 256r–256v. I. Edited by Sam Loyd II. C. pp. Sam. 1980. 68–70. S. Teubner. Sect. 76–77.” A. Johannes. no. Vol. 53 and 346. A. Vol. 5. 3. four vols. De Gruyter. Metrodorus (compiler). pp. 1982. L. 1: Arithmetik und Algebra. Leipzig. c. Heiberg and H. Euclid.. 201–210. Opera. MS 952. Arabe de la Bibliotheque Imperiale. Loyd. 1696. Recreations Mathematiques et Physiques. Untitled manu uscript called “Alfakhrî. 2 Ò edition. aka al-Karagi.

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

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

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

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

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

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

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

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

passing once again by entrance mirror E. 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. 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. Figure 1. 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. P = perpendicularly reflecting mirror). If the ray would not return along the entry paths. if these are suitably placed and oriented. Oskar van Deventer The Problem It is well known that a ray of light can reflect many times between two ordinary line mirrors. half of the potential reflections would not be used. Solutions are shown in Figure 1.Point Mirror Reflection M. It is obvious that the strategy of 245 .

However. as shown for the regular pentagon in Figure 3. the theoretical maximum number of reflections is Æ ´Æ   ½µ · ½. going to every other mirror. 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. for even Æ there is a parity complication. since the light paths are used in two directions. At mirror E there can be one additional reflection for the entering ray. Therefore. using a perpendicularly reflecting mirror P yields the maximum number of reflections. the angle between all the neighboring pairs of light paths is 180 Æ Æ . V AN DEVENTER Figure 2. at the other Æ   ¾ mirrors (other than the E and P mirrors) there can’t be Æ   ½ reflections. At mirror P there can be at most Æ  ½ reflections. At mirror E there can be at most Æ reflections. so remains valid. but only Æ   ¾ to make it an even number. which is an odd number. see Figure 2.246 M. At each mirror a maximum of Æ   ½ rays can be reflected. 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. O. 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. or ´½ µ ʽ ƾ   Æ · ½ However. . At all other mirrors there must be an even number of reflections. Æ   ½ from the other mirrors and one from the entering ray. from the other mirrors. Number of reflections for different mirror types.

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

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

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

V AN DEVENTER Figure 8. O. No general strategy was found. and many questions remain open: ¯ ¯ ¯ ¯ ¯ Is there a better search strategy for composite Æ ? What are the properties of solutions for Æ ½ and higher? Are there more Æ with multiple solutions? Can a sharper theoretical maximum for odd and composite expressed in a closed form? Are there solutions that use the back side of the mirrors? Æ be . Deleted paths for composite Æ up to 16. For composite Æ up to 16.250 M. which can be reached for prime Æ . solutions were found by a systematic trial and error search using a regular Æ -gon grid.

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

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

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

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

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

3X + 1 FUNCTION ITERATES 257 ¿ by starting at ÐÓ Ò and taking independent steps of size   ÐÓ ¾ or ÐÓ ¾ ½ each. let Ø£ ´Òµ equal the maximal value taken during the random walk.4). Large deviation theory shows that. Large deviation theory also asserts that the graph of the logarithmically scaled trajectories ´½ º µ approaches a characteristic shape. For the extremal trajectories for ­ £ ´Òµ in (14. 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. see Figure 1. This random variable is finite with probability one." What is the extremal size of ½ ´Òµ? Large deviation theory shows that with probability one ´½ º½µ The constant ­ÊÏ is the unique solution with ­ tional equation £ ´Òµ Ð Ñ ×ÙÔ ½ Ò ½ ÐÓ Ò ­ÊÏ ³ ½ ´ ½ ÐÓ ¿ µ ½ of the func¾ ´½ º¾µ where ­ ´ µ ×ÙÔ ¾ ½ ­ ½ ´½ º¿µ Next. £ and the expected size of ½ ´Òµ is given by (13. For the extremal trajectories for Ø£ ´Òµ in (14. For each initial value Ò an entirely new random walk is used. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally. the limiting graph of these trajectories has a different appearance. so that Ø£ ´Òµ ÐÓ Ò. this explains the name “repeated £ random walk model.24). The length of such trajectories approaches the limiting value ½ ´Òµ ÐÓ Ì ´ µ ´Òµ ÐÓ Ò ÐÓ Ò ¼ ½ ´Òµ . with probability 1. The random variable ½ ´Òµ is the step number £ with probability ¾ at which 0 is crossed.1) it is a straight line segment with slope  ¼ ¼¾¿½ starting from ´¼ ½µ and ending at ´­ ¼µ.

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

3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6. 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. Maximal value of ­ ´Òµ in intervals ½¼ Ò ½¼ ·½.127 8 127. ½ ½¼.527 Random walks model ½ ´Òµ ­ ´Òµ 70 108 165 214 329 429 592 593 706 755 21.38 41. .6020 0.5927 0. The agreement with the stochastic model is quite good. an analysis in [15. 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. Vyssotsky [23] has found much larger numbers with high values of ­ ´Òµ.94 31.91 32.890.5857 0.1) with numerical data up to ½¼ ½½ is given in Table 1.769. More recently V.48 18.328.5967 0.511 7 63.180.815 10 13.24 16.6020 0.5841 0. Section 5] shows that in ½¼ ½½ trials one should only expect to find a largest ­ ´Òµ ¿¼.91 19. 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. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1.728.799 6 8.68 Ö´Òµ 0. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale.194.884.6026 0.5818 0. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ­ ´Òµ ¿ ¾ .400.6030 0.77 31.527 has ½ ´Òµ ½¾ ½ and ­ ´Òµ ¿ ¾ ½ .527 5 837. ½ ´Òµ It gives the longest trajectory. and the large deviations extremal trajectory is indicated by a dotted line.6091 Table 2.64 32.171 4 52. A comparison of (14.371.13 26.5741 0. The large deviations extremal trajectory is indicated by a dotted line.266. Extensions of the data can be found in Leavens and Vermeulen [16].68 24.4). where ­ ´Òµ ÐÓ Ò .254 9 4.456. He found that Ò 12.

790 1.099 1.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½.05 19.511 10 77.48 12.817. In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed.46 2.84 19.391 8 319. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼.000 Ò 13.20 6.06 4.260 J.566. (See Table 3 in Section 4.560 1.049. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ.791 1. C. 4. so do not have any significant effect on the size of the tree.10 20.559 Random walks model Table 3.14 11. In this case. In fact.03 19. Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch.65 7.788 1.645 ­ ´Òµ 21.903 2. Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2. 5.59 23.631.86 7. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth . In iterating the multivalued function Ì  ½.804.819 1.83 3. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ. 15.511 6 6. there are duplications.32 2.02 21. if Ò ¾ or 8 ´ÑÓ µ. using (15.831 9 8. The set of inverse iterates associated to any Ò has a tree structure.663 4 77.528. or 7 ´ÑÓ µ. respectively.75 5. LAGARIAS Ò 1 27 2 703 3 9.897 2.) empirically like .86 13. where ½ ´½ º½µ Ì  ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½.1). these trees have the minimal and maximal number of leaves possible for depth .976 1. see [15] for more explanation. ½ ½¼.909 1.675 7 80. . where the branching at a node in the tree is determined by the node label ´ ÑÓ µ.24 16.65 5. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿.671 5 704.86 5.362.

cf. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½. The random walk at a given leaf node ends at a position Ù on the real line. . and the root node is assigned the value 1.2).1]. Pruned ¿Ü · ½ trees of depth Ò . In order to prove the result (15. it was necessary to determine the distribution of the number of leaves in a tree in the branching process. satisfies (with probability one) ´½ º¾µ ÐÓ Ò£ ´ µ ­ ȳ ½ ÐÑ ½ ÐÓ This constant ­ È ¬  ½ . Theorem 4. 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. ­ È ­ÊÏ . We assign labels ¼ or ½ to the edges. Lagarias and Weiss [15.e. and where fying £ ´ µ ×ÙÔ ´½ º¿µ   ÐÓ ´¾ · ½ ¾ ¿ ¿ ¼ This constant is provably the same constant as (14.3X + 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. Theorem 3.4] show that the expected size of ÐÓ Ò£ ´ µ. and the value assigned to that vertex by the rules above is exactly Ù . i. In [15] it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ .. where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree.1). where the values of the inverse image nodes of a node Ò of value Ò are ¾Ò or ¾Ò ¾¿ depending on whether the tree has one or two branches. [15. where ¬ is the unique positive number satis£ ´¬ µ ¼. The branching random walk arises from associating a walk to each path in the tree from a root node to a leaf node at depth . where 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.

(See Athreya and Ney [4] for a general discussion of such processes. LAGARIAS and as ½ the leaf probability distribution is of the form Ï ´ ¿ µ . they studied the normalized extreme values ´½ º½µ ·´ µ ¿ Æ  ´ µ Ò  ´ µ ¿ Æ ·´ µ Table 3 gives data up to ¿¼. 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. 2.) The random walk arises from assigning the labels ¼ and ½ to the edges of the resulting tree. 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. from a given root node . 4.2]. Here the pruned 3x+1 tree Ì ´ µ is the tree of inverse iterates of the ¿Ü · ½ function grown backward to depth . see [15. C. 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]. Since the number of leaves is expected to grow like ´ ¿ µ . Ê´ µ counts the number of distinct edge-labeled ¿Ü · ½ trees of depth . Distribution of 3x + 1 Trees Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees.262 J. Based on this data. 5. 7. This can be formalized as Conjecture C . with all nodes that have a label congruent to ¼´ÑÓ ¿µ removed. respectively. the corresponding random walk . Theorem 3. they proposed that the normalized extreme quantities remain bounded. 16. where ¼ represents the entering vertex being even and ½ represents the entering vertex being odd. The branching process is a multi-type Galton–Watson process with six types labeled 1. Let Æ   ´ µ and Æ · ´ µ denote the minimal and maximal number of leaves. 8 pictured in Figure 4. In this table.

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

302 0.327 2.307 0.240 0.09 56.33 1.802 1.923 1.394 0.488 0.500 1.78 2.751 1.21 5.802 1.12 74.255 2.259 0. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a fixed node Ò´ÑÓ ¿ ·½µ.99 13.289 0.488 0.490 0.305 0.782 1.534 0.344 2.481 0. C.75 5599.232 2.118 2.501 0.688 1.49 9.77 2362.741 1.486 0.986 2.62 1328.300 0.375 2.83 99.780 1.238 1.419 2.352 0.252 0.271 0.633 0.243 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 1.747 1.742 1.360 2.669 1.81 4199.562 0.475 0.281 0.68 31.451 0.465 0.026 2.405 2.67 0.489 0.782 1.57 42.750 1.746 1.738 1.491 1.38 236.47 996.791 1.076 2.310 2.688 1.534 0.753 1.50 315.292 2.750 0.34 420.481 0.375 0.16 4.60 747.45 560.802 1.475 0.491 0.32 17.562 0.738 1.409 0.270 2.482 0.534 0.507 0.83 1771.481 0.054 2.490 0.898 1.36 3149.898 1.211 2.746 1.273 0.507 0.264 J.869 1.422 0.190 2.784 1.748 0.748 1.489 0.390 2.265 0.246 0.557 1.792 1. Normalized extreme values for ¿Ü · ½ trees and for the branching process.268 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 4.911 1.401 0.758 1.774 1.500 1.255 0.481 0.475 0.166 2.76 23.37 3.249 0.762 1.633 0.131 2.342 0.77 133.62 7.492 0.03 177.750 0.490 0. .728 1.778 1.758 1.486 0.422 0.

349–360. pp. Math. Lagarias. C. Symp.. Univ. Comp. (June). pp. pp. 24. The ¿Ü · ½ Problem: Two Stochastic Models. Leavens and M. Mathematical Games. Comp. Maximum Excursion and Stopping Time Record Holders for the ¿Ü · ½ Problem: Computational Results. Oliveira e Silva. 1972 Number Theory Conference. Lagarias. Weiss. Cloney. Prob. E. Rawsthorne. Das ¿Ò · ½ Problem. Collatz. Slovaca. 42. B. pp. New York. [12] I. [15] J. Applegate and J. Natural Science Edition. Applegate and J. C. Math. Appl. pp. Luby. Pseudorandom Number Generators in Cryptography and Number Theory. [14] J. . [5] T. 85–89. A. and G. Math. Vermeulen. Comp. Advances in Math. [7] J. 231–251. Conf. 101–117. 25 (1977). pp. 57–64. Imitation of an Iteration. Ò¸ ´¾Ò · [8] C. [6] L. 411–426. J. Monthly 82 (1985). 49–52. of Colorado. (1995). TreeSearch Method. Princeton University Press: Princeton. Proc. C. 1990. pp. 3–23. Vol.. On the origin of the ´¿Ò · ½µ Problem (in Chinese). AMS. Springer-Verlag. Physics and Geometry-Locarno. pp. J. C. Proc. The ¿Ü · ½ Problem. 68. [3] D. Branching Processes. Math. Ney. Density Bounds for the ¿Ü · ½ Problem I. Feix. 1972. 3. (1976). Iteration of the Number Theoretic Function ´¾Òµ ½µ ¿Ò · ¾. 44 (1994). Merlini. on Stochastic Processes. Vichniac. On Heights in the Collatz ¿Ò · ½ Problem. 11 (1992).. Muller. pp. [11] L. Mag. 1996.3X + 1 FUNCTION ITERATES 265 References [1] D. [9] M. Everett. 371– 384. Boulder. Math. 4. pp. 9–11. pp. Athreya and P. [18] H. 12. [13] J. 1990. No. 115–143. Computers Math. pp. Unpredictable Iterations. Annals Appl. [16] G. 64 (1995). Mitteilungen der Math. NJ. D. 1972. 2 (1992). 3rd Intl. Krasikov Inequalities. Garner. H.. Generalized Parcal Triangles. 64 (1995). pp. Math. Amer. E. Experimental Math. [10] Martin Gardner. pp. E. On the Distribution of ¿Ü · ½ Trees. Providence. [20] D. (1999). Lagarias and A. The ´¿Ü · ½µ ¾ Problem: A Statistical Approach. 58 (1985). Goles. Ges. Proc. R. Pomerance. Korec. pp. The ¿Ü · ½ Problem and Its Generalizations. Lagarias. Muriel. and R. Density Bounds for the ¿Ü · ½ Problem II. C. 55 (1985). 114–118. [19] T. Discrete Math. C. 172–176. No. [2] D. ¿Ü · ½ Search Programs. Scientific American 226 (1972). Lagarias. and Cellular Automata. C. Appl. The ¿Ü · ½ Problem: a Quasi-Cellular Automaton. Pseudorandomness and Cryptographic Applications. 42–45. R. 79–99. pp. 229–261. Lagarias. Applegate and J. ed. pp. Colorado. Y.I. Complex Systems 1 (1987). in: Cryptology and Computational Number Theory. Hamburg 12 (1991). Math. of QuFu Normal University. Conway. [17] M. Tartani. C. [4] K. 427–438.

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

Sign up to vote on this title
UsefulNot useful