Click on your interest section for more information :
Acne q Advertising q Aerobics & Cardio q Affiliate Revenue q Alternative Medicine q Attraction q Online Auction q Streaming Audio & Online Music q Aviation & Flying q Babies & Toddler q Beauty q Blogging, RSS & Feeds q Book Marketing q Book Reviews q Branding q Breast Cancer q Broadband Internet q Muscle Building & Bodybuilding q Careers, Jobs & Employment q Casino & Gambling q Coaching q Coffee q College & University q Cooking Tips q Copywriting q Crafts & Hobbies q Creativity q Credit q Cruising & Sailing q Currency Trading q Customer Service q Data Recovery & Computer Backup q Dating q Debt Consolidation q Debt Relief q Depression q Diabetes q Divorce q Domain Name q E-Book q E-commerce q Elder Care q Email Marketing q Entrepreneur q Ethics q Exercise & Fitness q Ezine Marketing q Ezine Publishing q Fashion & Style q Fishing

Fitness Equipment q Forums q Game q Goal Setting q Golf q Dealing with Grief & Loss q Hair Loss q Finding Happiness q Computer Hardware q Holiday q Home Improvement q Home Security q Humanities q Humor & Entertainment q Innovation q Inspirational q Insurance q Interior Design & Decorating q Internet Marketing q Investing q Landscaping & Gardening q Language q Leadership q Leases & Leasing q Loan q Mesothelioma & Asbestos Cancer q Business Management q Marketing q Marriage & Wedding q Martial Arts q Medicine q Meditation q Mobile & Cell Phone q Mortgage Refinance q Motivation q Motorcycle q Music & MP3 q Negotiation q Network Marketing q Networking q Nutrition q Get Organized - Organization q Outdoors q Parenting q Personal Finance q Personal Technology q Pet q Philosophy q Photography q Poetry

Political q Positive Attitude Tips q Pay-Per-Click Advertising q Public Relations q Pregnancy q Presentation q Psychology q Public Speaking q Real Estate q Recipes & Food and Drink q Relationship q Religion q Sales q Sales Management q Sales Telemarketing q Sales Training q Satellite TV q Science Articles q Internet Security q Search Engine Optimization (SEO) q Sexuality q Web Site Promotion q Small Business q Software q Spam Blocking q Spirituality q Stocks & Mutual Fund q Strategic Planning q Stress Management q Structured Settlements q Success q Nutritional Supplements q Tax q Team Building q Time Management q Top Quick Tips q Traffic Building q Vacation Rental q Video Conferencing q Video Streaming q VOIP q Wealth Building q Web Design q Web Development q Web Hosting q Weight Loss q Wine & Spirits q Writing q Article Writing q Yoga


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

G4G1 was so successful that a second gathering was held in January 1995 and a third in January 1998. Included here are the papers from G4G1 and a few that didn’t make it into the initial collection.x FORWARD Scott Kim. The success of these gatherings has depended on the generous donations of time and talents of many people. 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. so many people have expressed interest in the papers presented at prior gatherings that A K Peters.. Georgia . Tyler Barrett has played a key role in scheduling the talks. California Tom Rodgers Atlanta. 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. distributed to the conference participants. Charlotte. has agreed to publish this archival record. Elwyn Berlekamp Berkeley. Ltd. who has made his extraordinary career possible. As the gatherings have expanded. and presented to Gardner at the meeting.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

Bang!. right? A week = 7 days corresponds to 14 Billion Years 1daycorresponds to2BillionYears(USA) 1 minute corresponds to 2 Million Years 1 Millisecond is 23 Years 23 . Light Sun and Earth. You. Time. 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. and a bit more ½¼  meters = Nucleus of a human cell ½¼ ½ meters = Atom’s nucleus Time The Creator.Three Limericks — On Space. 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. form up. Friday night At a minute to twelve Eve spin. barks out his orders for the week. First thing on Monday. seen as an Army Sergeant Major.

roughaveragespeedofEarth 700. ROWETT Speed A child cycles ‘round the schoolyard Which lies on the Earth turning hard The Earth rounds the Sun As Sol does “the ton” And our Galaxy flies — Gee! I’m tired 7 mph — child cyclist 700mphEarth’ssurface(NorthAfrica) 70.24 T.4 million mph Galaxy’s speed through the debris of the Big Bang .000 mph turning speed of Galaxy 1.000mph.

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

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

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

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

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

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

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

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

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

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

2:41:42. 6. 64. 36. For example. 66. Can you divide the figure into three identical shapes? ´Ü   µ´Ü   µ´Ü   µ ´Ü   ݵ´Ü   Þ µ Problem 8: A 24-hour digital watch has many times that are palindromic.) These curious combinations occur 660 times a day. Problem 6: Which two numbers come at the end of this sequence? 2. (2) Find the two palindromes whose difference is closest to 12 hours. 23:55:32. . 62. 1:01:01. 40. y Problem 7: The figure shown here is the solution to the problem of dividing the figure into four identical shapes. (1) Find the closest such times. 60. 46. x. 44.38 N. 30. 54. 3:59:53. 52. 56. YOSHIGAHARA Problem 4: Arrange the following five pieces to make the shape of a star. (3) Find the longest time span without a palindromic time. 4. 13:22:31. 42. 32. 50. 34. etc. (Ignore the colons. Problem 5: Calculate the expansion of the following 26 terms.

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

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.

I call such individual instances mathemagical black holes. Ecker The legend of Sisyphus is a lesson in inevitability.Number Play. 2. and 15 have in common with each other. 41 . These are 3 (three evens). 6174. and the total number of digits. All numbers in this universe are drawn to 123 by this process. The number 123 with respect to this process and the universe of numbers is a mathemagical black hole. and Card Tricks: Mathemagical Black Holes Michael W. Based on articles appearing in (REC) Recreational and Educational Computing. as well as with various card tricks? These are mathematical delights interesting in their own right.759. never to escape. the physical universe has strange entities called black holes that pull everything toward them. 4. 3. 347. No matter how Sisyphus tried. Like the legend. and 7 (seven is the total number of digits). total number. after all. never to escape. 4 (four odds). Calculators. use these digits to form the next string or number. If we repeat with 123. pulled inexorably by gravity. but much more so collectively because of the common theme linking them all. What. So. 153. such as 9. these bodies may be even more difficult to identify in the world of number play than their more famous brethren in physics. counting evens. to get 1.288. The Sisyphus String: 123 Suppose we start with any natural number. respectively. so write down 123. odds. the small boulder he rolled up the hill would always come down at the last minute. could numbers such as 123. regarded as a string. Count the number of even digits. But did you know that we have comparable bodies in recreational mathematics? At first glance. we get 123 again. the number of odd digits. Now repeat with 347.

Iterating for 202. and a final iteration from 303 produces 123.42 M. odds. 25. At this point. So. Or revise the program to automate the testing for all natural numbers in some interval. and total are 20. The numbers of evens. since 123 yields 123 again. If you wish. odds. One more iteration using 426 produces 303. and 45. 2. the number obtained from 20.545 we find 4. 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. 45. so we have 426 now. 25. any further iteration is futile in trying to get away from the black hole of 123. W. respectively. ECKER But will every number really be sent to 123? Try a really big number now. say 122333444455555666666777777788888888999999999 (or pick one of your own). our next iterate is 202. What Is a Mathemagical Black Hole? There are two key features that make our example interesting: . you can test a lot more numbers more quickly with a computer program in BASIC or other high-level programming language. total. and 6 for evens. modify line 110 to allow the program to start again.545.

. I’ve personally checked the iterates of f (n) for n = 1 to 999 by a computer program such as the one above. any number to which other elements (usually numbers) are drawn by some stated process. In other words. The direct proof is actually faster and easier. Every element subject to the force of the black hole (the process applied to the chosen universe) is eventually pulled into it. 2. f). Why does this example work. 48.MATHEMAGICAL BLACK HOLES 43 1. f (b) = b . Fall 1992. and why do most mathemagical black holes occur? My argument is to show that large inputs have smaller outputs. At that point. suffices. In mathematical terms. # odd digits. once drawn in per point 2. starting at 1000 or above eventually pulls one down to under 1000.) Thus. but try induction if you would like rigor. we can argue as follows: If n f (n) n. No. A mathemagical black hole is. where b is an element of a set U and f: U satisfying: 1. # odd digits. b = 123. total # digits. loosely. total # digits): ½For generalized sisyphian strings. there exists a natural number k satisfying f (x) = b. the real trick is in finding interesting processes. an element never escapes. all (b. as point 1 ensures. and the superscript indicates k-fold (repeated) composition of functions. thus reducing an infinite universe to a manageable finite one. an argument by cases. U. the new number that counts the digits is smaller than the original number. you never get out. see REC. a black hole is a triple U is a function. then In the case of the 123 hole. b plays the role of the black-hole element. or a computer check of the finitely many cases. (It’s intuitively obvious. For each x in U. sufficient iteration of the process applied to any starting number must eventually result in reaching 123. just as reaching a black hole of physics implies no escape. For the Sisyphus String½ . For n 1000. Here. Once you hit 123. Formalized Definition. U = natural numbers ¸ and f (number) = the number obtained by writing down the string counting # even digits of number. Of course. 2. as a three-digit string for a number must have one of these possibilities for (# even digits. In this case. 999. Though the number itself is the star of the show.

In turn. 370. 1. As another instance. is. Add up the cubes to form a new . other natural languages may have a comparable property. In this case. Count the number of characters in the spelling. and finally 4. To create a black hole. this gives 12. For instance. 3) (2. such as FIVE for the usual 5. Since we are going to be doing some arithmetic. 3) So. Narcissistic Numbers: 153 It is well known that. One at a time. 2. 6.111 (six ones) is a multiple of 3 because the sum of the digits. it is 4 — or FOUR. 111. Though this result is clearly language-dependent. take the cube of each digit. Of these. Write down your multiple of 3. Repeat with 4 to get 4 again. 0. 3) (1. Now apply the rule to each of these four and you’ll see that you always produce (1. 163 appears as ONE HUNDRED SIXTY-THREE for a total count of 23. just one has a black-hole property. 3) (3. We start with any positive whole number that is a multiple of 3. I’ll arbitrarily say that we will include spaces and hyphens in our count. and 407. Recall that there is a special shortcut to test whether you have a multiple of 3. we need to define a universe (set U ) and a process (function f ). work now with the 4 or FOUR. Just add up the digits and see whether that sum is a multiple of 3. the only natural numbers that equal the sum of the cubes of their digits are 153. then 6. To avoid ambiguity. if n 1000. Words to Numbers: 4 Here is one that master recreationist Martin Gardner wrote to tell me about several years ago. 1. Then. 2. you may wish to take out a hand calculator and/or some paper. other than the trivial examples of 0 and 1. W. ECKER (0. 371. within one iteration you must get one of these four triples. Take any whole number and write out its numeral in English. then 5. So. 3) — thus resulting in the claimed number of 123.111 (seven ones) is not.111.44 M. then 3. try 163. but not necessarily with 4 as the black hole. However. 3.

Then b = 153 is the unique black-hole element. at which point future iterations keep producing 153. which leads to 1458. Note also that this operation or process preserves divisibility by 3 in the successive numbers. (It is well known that none exists for the sum of the squares of the digits. it is necessarily unique. if a black hole exists. And once you reach 153. 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. one more iteration just gets you 153 again. with a paper or problem proposal in one of the smaller math journals every few years. this particular result. (For a given universe. You must reach 153. 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. Let’s test just one initial instance. Now repeat the process. finally leading to 153. then 702. . without the “black hole” terminology or perspective. which yields 351. Using the sum of the cubes of the digits. so break out after you’ve grown weary. gets discovered and re-discovered annually.) In more formal language.MATHEMAGICAL BLACK HOLES 45 number. if we start with 432 — a multiple of 3 — we get 99. as one gets cycles. One nice thing is that it is easy to edit this program to test for black holes using larger powers.) Not incidentally.

Regroup the cards by picking up the cards by whole columns intact. Repeat one last time. Now ask him (or her) which of the three columns contains the card. ECKER The argument for why it works is similar to the case with the 123 example. for large numbers n. Card Tricks. card 11. It’s a classic card trick.. This is the card in the fourth row and second column. note that. laying out three across at a time). try this one from one of my readers. I’ll omit the proof. W. To find an analagous black hole for larger powers (yes. Then. At the end. regrouping cards with the designated column being in the middle. Remove 21 cards from an ordinary deck. Even Here’s an example that sounds a bit different. Ask somebody to think of one of the cards without telling you which card he (or she) is thinking of. which is to say. n. by cases or computer check. Repeat asking which column. but the easier way is to draw a diagram that illustrates where a chosen card will end up next time. There are two ways to prove this. and then test to see whether other numbers are drawn to it. 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 . Arrange them in seven horizontal rows and three vertical columns. First of all. f (n) each value eventually is “pulled” into the black hole of 153. and re-dealing out by rows. Now re-lay out the cards by laying out by rows (i. you will need first to discover a number that equals the sum of the fourth (or higher) power of its digits. so 153 is indeed a fixed point. there are some). but be sure to sandwich the column that contains the chosen card between the other two columns. yet meets the two criteria for a black hole. for the black-hole attraction. the card chosen must be in the center of the array. 1¿ + 3¿ + 5¿ = 153.46 M.e. But for those who enjoy programs. Second. for suitably small numbers.

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

This one is a bit more tedious. Divide both sides of the equation ´Òµ ½µ · ´Ò   ¾µ by ´Ò   ½µ. Your own example may take more steps. That is. ECKER Example: Start with 8028. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. 21. In fact. ½ . The first two numbers numbers: 1. 2. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. 5. Its divisors are 1. 34. 89. Had we used the ratios ´Òµ ´Ò · ½µ instead. The ratios seem to be converging to a number around 1. 8. Its value is roughly 1. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors. 5. but your very equilibrium. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ ľ   Ä   ½ ¼ . Now take the sum of the digits of these divisors.). etc. ´½µ ´Ò   ½µ · ´Ò   ¾µ for integers Ò ¾. and 15. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. 1. More formally. Iterate until a number repeats. but without using a program. phi. 3. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò   call it Ä — as Ò increases. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. including 1 and the number itself.618 . Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . Repeat with 8532 to calculate 8532–2358 = 6174. 3. and ´Òµ preceding two elements.48 M. the smallest is 0288. and these digits sum to 15. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. but it is also that much more strange at the same time. The black-hole number this time is 15. but fewer authors use that approach. but you will reach 6174. we would have obtained the reciprocal of phi as the limit. 55. 13. The largest permutation is 8820. W.6 or so. To get a quick and dirty feel for why phi arises. and the difference is 8532. This one may defy not only your ability to explain it.

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

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

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

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

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

and they must cross in 25 minutes. A knight is placed on an infinite checkerboard. . and D takes 10 minutes. . 6.880. 4. (a) There are six men with crossing times of 1. 7. 9. It is night. 2. A pair must walk together at the rate of the slower man’s pace. Four men must cross a bridge. 2. E13. 8. and 9 minutes and they must cross in 31 minutes. Can all four men cross the bridge? If so. E14. and 10 minutes. Find nine single-digit numbers other than (1. A maximum of two people can cross at one time. B takes 2 minutes. C takes 5 minutes.. how? Try these other problems. 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. 3.PUZZLES FROM AROUND THE WORLD 55 E10. how can you make it unable to move in as few moves as possible? E12... 3. Each man walks at a different speed: A takes 1 minute to cross. E11. 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. 8. They all start on the same side and have 17 minutes to get across. and the bridge will hold up to three men at a time. (b) There are seven men with crossing times of 1. and 9) with a sum of 45 and a product of 9 362. The Bridge. If it cannot move to a square previously visited. 6.

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

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

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

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

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

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

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

and the next inner layer is an anhydrous fromage. the little that is known about the krypton layer was received via sub-etherial communication from a Venerian pioneer immediately prior to its demise at the hands (some would say wattles) of a frumious snatcherband. The inhabitants of Lyra III recognize special years when their age is of the form Ô¾Õ. By nature the Venerians can only survive at the krypton/fromage boundary and by tragic mistake the two copods had landed on disconnected branches of . one then becomes a master until reaching a year that is the third in a row of consecutive special years. (a) (b) (c) (d) When does one become a master? When does one become a sage? How long do the Lyrans live? Do five or six special years ever occur consecutively? H18. It (the Venerians are sexless) also reported that the straight line distance between it and its copod was 120 spandrals. however. It’s internal structure. The outermost of these layers is nearly pure krypton.PUZZLES FROM AROUND THE WORLD 63 H17. The common axis of these layers is the planet’s rotational axis. 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. which occurs in a special year that is the fourth in a row of consecutive special years. being made of a number of coaxial right circular cylindrical layers. 18 and 20.331 to 1. 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. Inc. just as one observes for Sol III. (A spandral is the Lyran unit of length. finally one becomes a sage until death. is contemplating mining the outer layer. 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. Its axis of rotation coincides with the spheroid’s small axis.. is unique. Cosmic. (A copod corresponds roughly to something between a sibling and a rootshoot. each of homogeneous composition. The planet Alpha Lyra IV is an oblate spheroid. The first few such special years are 12.) Unfortunately. where Ô and Õ are different prime numbers.

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

9. . it appears that the rule might be subtraction. (0. (3. 6. The trick here is that 1 knot = 1 nautical mile per hour. (7. E4. Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. 4. 2. E2. 15 moves: (15. 0). (7. Thus and Ò ¾½¼ transfers is the minimum. 9). 6). 2). (1.425. Instead. ؾ ¾¾ hours. At first glance. Thus Ü ½½. E7. In each case the numbers combine by summing the products of their digits. E9. The number is 381. (3. and Ü ½ . 5). (9. the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. 0.741. 10. (1. E10. E6. 5. ¼. E3. the next greater number is 7. (9. 10. ¾ gives Ø E5. (5. so E8. (11. 8. 6. But this is not right because 18 is not the difference of 38 and 16. 6. 0). 6). One of the terms is ´Ü   ܵ. 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 . (3. ¿   ½ ¼¼¼   are the smallest values satisfying the above equations. 6). 2. 2). 8. (3. 10. 0). 6. The license number is 741. (11. 0. 7. 2. . 6). 0. 4. 6). (1. 4.654. 0). ½ ¼½ or ¼ and ½ ¼½.729. Anywhere within 10 miles of the north pole. 0.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. 4. (5. 2). so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. 0). 0). 6).

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

0). 134. (30. 0). 11. 3. (18. 3. 128. (21. 0. (24.PUZZLES FROM AROUND THE WORLD 67 E16. In 3. (30. 2. 147. (20. 0. 1. 5). 11. (29. (25. 0). (31. 125. 11. 6). (31. 11. M2. 4. 0). There are 7 ¢ 6/2 = 21 choices using vertex 1. producing pieces A and B. 1. 11. 0. (34. 6). (a) 32 moves. D will finish burning. (26. 0). (23. ½¿ ½¿ ½ Ê 123. 136. . (35. (23. The probability of an acute triangle is ¿ ¾½ or ½ . 158. 0. 11. 11. When C is consumed. (20. 1. (19. (29. 0). 10). 0. (24. Solutions to Medium Problems M1. (11. as seen from the list below. 6). 6). 0). (31. 6. 10). 2). 6. 3). 4. 3. 11. (28. (18. 10). 0). (26. 167. start burning C from both ends and D from one end. 0. 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. 0). 0). 9. 1. 0. (25. 4). (24. (28. light the other end of D. 1). 168. (24. 3. (30. 148. 11. 1. 10). (22. 11. and the time is 8:24. (19. 6). (23. 4). If the hour hand is exactly on a second mark then the second hand will always be on the 12. 10). 1). (33. 0. See the figure. 0). 11. This is the shortest time interval that can be measured. (25. 6). (37. 127. (35. the probability of a right triangle is ½ ¾½ or . 126. 6. 0). (19. 0. (35. (24. E17. 6). 0). (33. 6). (30. 1). 0). 11. 0). (23. (34. 6). (29. (22. 11. (25. (30. 2). (36. 10). 1). 0). 145. (18. (17. 11. producing pieces C and D. (29. 0). 6. 5. 0. 9. (25. 11. 0. 0). (18. (36. Burn A from both ends. 0). (25. 10). 0). 1. (31.75 minutes. 10). 6. 5. 0). 10). (30. (26. 6). 1). 5. 1). 178. (17. 1). (24. 124. (23. Cut B at that corresponding point. 0). 10). 0). 0. Cut the lace in half. 146. 10. 10. 0. 0). 1. 10). 1. 6). 10). 0). (23. (29. noting the point that burns last. 4. 10). 0. 0. (25. 0. 6. (22. 0. 4). 0) (b) 40 moves. 0). 4. (29. 11. (21. 11. 2. 0. 156. 0). 5). 3). 4). 138. 0. 0). (37. (31. 0). (27. 0. 5. 10). (19. At the same time. 0. 0. 1. (36.

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

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

Each number is the length of the shorter side of the rectangle. 6S. 2S. P. . 2P2S. PS. P3S.465.70 R.789. 4S. M15. A nine-rectangle solution is shown below. 2PS. 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 M13.123. I. P4S. 5S. and 3P. 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. 2P. 1. Shown below. M14. P2S. 3S.

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

I.0119 . 13. (15. 13. 10. Humpty Dumpty got a 10 in arithmetic. Solutions to Hard Problems H1. 9). 14. 8). 6). The points are ´ ¿   ½µ ¾ of the way along each diagonal for a distance of 3. Shown below. 8). (15. 9) and (14. (15. There are 5 subjects with 15 marks possible in each. 11. 14. 12. 14. 12. 14. 13. There are 7 Good Eggs. 14.72 R. 12. The number in each triangle indicates its area relative to the other triangles in the figure. 7). (15. (15. 11. . 12. 13. 12. 10. HESS (b) The two points are along the diagonal near P and its opposite Ô point on the ½ ¢ ½ face that includes A. 12. 11. M20. 11. The scores for the Good Eggs are (15. 10).

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

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

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

(b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved. I. Raise S and T appropriately above the .25 for the entire square or a circle inscribed in µ the square. The point S may be raised above the plane to give unit altitudes if the inscribing circle has radius ½. For Ò ½ there is just enough room for 329 circles. The maximum ratio È Ö ¼¾ ¼ . H10. H9. Note that the ratio È is 0.9973967 . Now consider two such circles of different sizes as shown. The ratio È takes on a maximum when Ö ¼¾ ¼ . A view of the pyramid looking down from S perpendicular to the plane ABCD will look like the figure below. The smallest rectangle found to date containing 329 circles has 13 circles on each side of 101 sets of 3 and measures 2 ¢ 163. Each has a radius ½. HESS the shape is È Ô ½ ´¾ · ¾ Ö ·   Ö.76 R. The circles need to be packed as shown below. They define the four tangent lines drawn to produce ABCD as shown. There are 7 circles on each end with 105 sets of 3 circles in the middle.

A computer program was written to calculate the probabilities shown in the table below.2270 1.8027 0. H11. È ´¿¼   ¼µ ÔÈ ´ ¼   ¼µ.3586 Pairing probability 0. which leads to Ô ¼ ½ ¿¿ ¼ .8163 0. ¾ ¾ Õ´ È ´¿¼   ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ .2069 1. È ´ ¼   ¼µ ÔÈ ´ ¼   ¿¼µ · ÕÈ ´¿¼   ¼µ. which leads to Ô ¼ ½½ ¿ ¾ . Ô  Õ At a score of Ò games to Ò (Ò ¼ to 5). At Ò games to Ò. Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1. ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ .PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit. ¾ H12. Clearly ST is not perpendicular to the plane.7805 0 . Let È ´   µ be the probability of the server winning the game when the server has points and the receiver has points.7892 0.3074 1. ·¾ Ô ´½ ½ Õ µ È ´¼   ¼µ .2801 1. È ´¼   ¼µ È ´¿¼   ½ µ when Ô ¾   Ô ¿. · È ´ ¼   ¿¼µ Õ È ´ ¼   ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . Let Ô the probability of the server winning a point.7849 . This leads to ¾ È ´ ¼   ¼µ Ô¾Ô Õ¾ . and let Õ ½   Ô. My uncle is expected to take the longest time to dress on Saturday and on Friday he is least likely to get a pair from the first three socks chosen. È ´¼   ¼µ È ´ ¼   ¿¼µ when Ô¿   Ô¾   ¾Ô ½. È ´ ¼   ¿¼µ Ô · ÕÈ ´ ¼ ¼µ.8286 0.2511 1.

¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ . which must have been guessed by Jones two years ago. Solutions occur for (b) ¾ ½. (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ .042. I. ¼ ¼½¼ . and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago.093. we get (a) (b) H16. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½. prime. ¿ ½¾¼. (a) 45 years. (a) Length ½ ¿ ¾½ ½ ¼ as shown. a=10. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. H17. . · ½. The product and sum two years ago could be achieved by (2. 8.613. (2) the younger two children have different ages. Using a multiprecision continued fraction program. ¼ ¼¼ ½ ¼.444 years. (c) Unknown. (c) 17. We look for cases where (1) the oldest child is under 20. which must have been guessed by Smith. 15).546. (b) 605 years. ¾   ½µ ¾. . (d) Yes. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. · ½. ´ ¾ · ½µ ¾ prime. HESS H13.641. There is only one set of ages that accomplishies this: (5. ´ prime. H15.78 R. 7.512. H14. 12). ´ ¾   ½µ ¾. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô .321 is the first of 5 in a row. 6. 16). . The product and sum could be achieved by (4. (b) Length ¾ ¼¾½½¾ ¿ as shown.

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

HESS (2) Since ´  ÒµÞ is an entire function in the complex plane. 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. Thus Þ ¾ Õ´Þ µ  ´Ò·½µÞ is regular within . which still avoids enclosing the poles of Õ´Þ µ.80 R. 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. we expand and deform to a contour the figure. The only pole of Õ´Þ µ  ´Ò·½µÞ within is a double pole at Þ ¼. (3) The summation over can be carried out explicitly to give ´Òµ ¾ ½ ¾Ò· · ¿ ¾ Þ Þ ½   ´½ · Þ µ  Þ ´½ · Þ µÒ·½ (4) Define Õ´Þ µ ½ ½   ´½ · Þ µ  Þ . 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.

 ¾ × ´Ý · µ  Ü Ý . Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to infinity. not including its pole at Þ ¼.0888430 2. Then Ý approximately equals   ½·ÐÒ´ µ . Since the equation in Ý is even. × ÒÝ Ý we can look for poles in the upper half-plane only and reflect each one into the lower half-plane. which reduces to ´½ · Ü · Ý µ´Ü   Ý µ ´ Ó×´Ò · ½µÝ   × Ò´Ò · ½µÝ µ where ´Ü¾ · ݾ µ ´Ò·½µÜ .1512074 45. The table below gives numerical values of the first few. Call the Ø such pole in the upper half-plane Þ Ü · Ý .5432385 32. A pole of Õ´Þ µ occurs Þ . (5) On Þ Ê for sufficiently large Ê. ¼ for all . where Þ at ½ · Þ Ü · Ý. 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.8221528 Ý 7.0262969 3.8505482 39.6406814 3.5012690 3. Ü 1 2 3 4 5 6 7 (6) The contribution to 2.2238350 26. Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ.PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ.6745053 3. The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ.8790560 20.4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ   Þ µÕ´Þ µ ´½ · Þ µÒ·½.4614892 13. For larger define ´¾ ·½ ¾µ .2916783 3.

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

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

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

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

By subtracting 10 from the Hexagon.86 R. with its pointer on the left. 3. 7. with its hook on the left. which have no symmetry. to the Signpost. GUY Figure 1. sailing to the left. to the Crook. which appears to be turning to the left. Notice that 0 has the greatest symmetry. and subtracting each of the labels 11 to 19 from 20 we arrive at the labeling in Figure 1. . . O’Beirne had already suggested a numbering of the pieces from 1 to 19. the multiples of 4 have rotational symmetry through 180 degrees. Negative (bar sinister?) labels are given to the Bar. We don’t know how O’Beirne decided whether to give a piece a number less than or greater than 10. . Crown. and Lobster each have a single symmetry of reflection. to the Snake. The Chevron. ¿. but our mnemonics are as follows. 5. and each is positioned by its apex. . and the Sphinx. with its handle on the left. . to the Yacht. and 9. and the other even numbers have reflective symmetries. . The Butterfly has the same symmetries as a rectangle: Its position is described by that of its body. K. to the Hook. exist in enantiomorphous pairs. as do pieces 1. The nineteen Hexiamond pieces. together with those that have rotational symmetry only. whose head is on the left. a parallelogram drawn in the opposite way from the usual textbook fashion. Pieces 4 and 8 have only rotational symmetry and exist in enantiomorphous pairs. Pieces with odd-numbered labels have no symmetry and. ½. The symmetries of the Hexagon are those of the dihedral group D6.

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

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

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

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

the second by the present writer. the first found by John Conway and Mike Guy in 1963. I didn’t think he would be successful. Coding the position of the Lobster. each with as many as 11 of the 19 pieces symmetrically placed. For example. Figure 13 shows two of Bert Buckley’s machinemade solutions. and solutions are known in all of these cases. In 1966 Bert Buckley. the discerning solver will soon wish to specialize. As we go to press. then a graduate student at the University of Calgary.O’BEIRNE’S HEXIAMOND 91 Figure 12. With such a plethora of solutions. but after a few months of intermittent CPU time on an IBM 1620 he found half a dozen or so solutions from which it was possible to deduce another fifty by hand. how symmetrical a solution can you get? Figure 14 shows two solutions. for the Hexagon. suggested looking for solutions on a machine. Marc Paulhus has established that there are just ½¿ · ½ · ½ · ½ · ½ · ¾½ · ¿ · ¾ · ¾ · ¿¼ · ¾ · ¿¼ · ½ · ¾¿ · ¿¿ · · ¿¿ · ¾¾ · ¿¾ ¿¿ different relative positions for the Chevron and Hexagon that yield solutions. .

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

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

Ô Ü (3-symmetry). GUY Figure 18.447 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. classified according to the position of the Chevron.717 C 6675 D 7549 E 11. president of Kadon Enterprises and discovered that she markets a set of “iamonds” . are 130 195 533 193 377 2214 111.489 B 15. 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. Figure 19.518 A 75. K. Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to find all the solutions.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. by the Hexagon’s position Ø Ù Total 264 1094 124. Six pieces meet at a point. for those who prefer their puzzles to have just one answer. I wrote to Kate Jones. Their numbers. On the contrary.

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

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

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

“Shigeo Takagi. a Tokyo magician.” A square with a center hole I know of two other books that are collections of patterns of the Sei Shonagon pieces. TAKAGI shapes are inaccurate.” and Edo Chie-kata (Ingenious Patterns in Edo) was published in 1837. W. In about 1780. Freeman and Co. The puzzle was introduced to the world by Martin Gardner in his book Time Travel ½ . I possess a sheet with wood-block printing on which we can see patterns of the Sei Shonagon pieces. was kind enough to send me a photocopy of this rare book. With the Chinese tans it is not possible to put a square hole anywhere inside a large square. and he elaborated as follows. Unlike the Chinese tans. the Shonagon pieces will form a square in two different ways.98 S.. Can you find the second pattern? The pieces also will make a square with a central square hole in the same orientation. In addition. Some patterns formed from the Sei Shonagon pieces ½1988.H. NY . Takahiro Nakada wrote a manuscript entitled “Narabemono 110 (110 Patterns of an Arrangement Pattern). but its author and date of publication are unknown.

a linear extension (Figure 1). one of an infinite number. or better generate. Tangram properties preserved in Trigo-Tangram are the linearity of the sides. Depending on the complexity of the transformation. That’s why I gave it the name TrigoTangram. Not preserved. 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. 99 . a Tangram derivative called Trigo-Tangram. others not. certain properties of the original puzzle are preserved in the derivative. the convexity of the pieces. Tangram submitted to a linear extension. The result is a set of pieces for a new puzzle. for example. fit into a grid of equilateral triangles. As an example. While the tangram pieces can be arranged in an orthogonal grid. a trigonal grid. due to ¿ as extension factor. The result will be a new puzzle. and — if you do it the right way — a nice one.How a Tangram Cat Happily Turns into the Pink Panther Bernhard Wiezorke Do you want to create. and the side length and area ratios. Figure 1. let us take the good old Tangram and try a very simple transformation. Puzzles generated by geometrical transformations I call derivatives . our new puzzle is a Tangram derivative. the Ô pieces of our new puzzle.

copy and cut out the pieces. For Trigo-Tangram. WIEZORKE Figure 2. you get funny results. As you can see.100 B. and enjoy this new puzzle. are the angles and the congruence of both the two big and the two small triangles. Here is another way to have fun with a puzzle and its derivative: Take the original Tangram. For our derivatives. Figure 4 shows how a Tangram bird stretches its neck and how a Tangram cat turns into the Pink Panther. The transformation of the solution can be a solution for the derivative (right image) or not (upper image). we do not only need the pieces. and try to make a corresponding figure with the pieces of Trigo-Tangram. . take a piece of cardboard. two problems can be generated by extending this original Tangram problem in two different directions. Figure 3 shows some problems for Trigo-Tangram. we need problems as well. So. which gives us two different problems. 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. In Figure 2 this is done for a Tangram problem and its solution. And you can see another important fact: The transformation of the solution does not always render a solution for the derivative. The images of the figures will be suitable problems for the derivative. lay out a figure. the corresponding linear extension can be performed in two different directions. Suitable problems for a puzzle derivative can be found by transforming problems of the original puzzle. In many cases.

Transforming two-dimensional puzzles provides endless fascination. the two puzzles Figure 4. You obtain better results by applying linear extensions diagonally (Figure 5). Just look what happens if you rotate the Tangram in Figure 1 by ¼ Æ and then perform the transformation. and a Tangram cat turns into the Pink Panther. Some problems for the Trigo-Tangram puzzle.TANGRAM CAT HAPPILY TURNS INTO PINK PANTHER 101 Figure 3. The square and parallelogram of the original are transformed into congruent rhombi. A Tangram bird stretches its neck. . which makes the derivative less convenient for a puzzle. Since in this case the congruence between the two large and two small Tangram triangles is preserved.

I hope I have stimulated your imagination about what can be done by transforming puzzles so you may start your own work.102 B. WIEZORKE Figure 5. and have fun generating individual derivatives for yourself and your friends! . generated are even more Tangram-like than the Trigo-Tangram. Take your favorite puzzle. Transforming the Tangram by diagonal extensions leads to two derivatives which are more Tangram-like than the one shown in Figure 1. One set of problems will serve for both of them. To create problems for these derivatives is left to the reader.

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

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

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

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

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

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



Void: a ¢ grid on which a “switching of the knights” puzzle is applied to pairs of polyomino shapes from domino to heptomino in size, formed with checkers (Figure 4). What is the minimum number of moves to exchange the congruent groups of black and white knights?

Figure 4.

Quintachex: the pentominoes plus a ¾ ¢ ¾ square checkerboarded on both sides (different on the two sides). The pieces can form duplicate and triplicate models of themselves, with checkerboarded arrangements (Figure 5).

Figure 5.

Colormaze: Square tiles in six colors form double polyomino shapes with no duplicate color in any row, column, or diagonal. In another use, 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).



Figure 6.

Poly-5: two-dimensional polyominoes of orders 1 to 5, in a four-way symmetrical tray of 89 unit squares (Figure 7).

Figure 7.

Sextillions: the hexomino shapes can form double through sextuple copies of themselves and various enlargements of the smaller polyominoes. Sizecompatible with Poly-5. Snowflake Super Square: the 24 tiles are all the permutations of three contours—straight, convex, concave—on the four sides of a square. They can form various single- and double-size polyomino shapes (Figure 8).



Figure 8.

Triangoes: orders 1 and 2 polytans permuted with two or more colors (square, triangle, parallelogram). The pieces can form diagonally doubled pentominoes and hexominoes with colormatching adjacency of tiles (Figure 9).

Figure 9.

Lemma: A matrix of multiple grids lends itself to polyomino packing with checkers in three colors. 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, including partitions into multiple shapes (Figure 10).



Figure 10.

Multimatch II: The 24 tri-color squares with vertex coloring (each tile is a ¾ ¢ ¾ of smaller squares) can form polyominoes both as shapes with colormatching and as color patches on top of the tiles (Figure 11).

Figure 11.

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

Figure 12.



Figure 13.

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

Figure 14.



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

Figure 15. Game of Solomon.

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

Cross plus-slit. Sometimes the opening is in a different direction (mountain or valley fold). with a different color on each side (as in standard origami paper). flexagons seem more elegant when no glue or paste is required. There are five different starting bases (paper shapes. There are three different ways of folding the bases. The four-sided ones discussed here are “flexed” by folding them in half along an axis. the base for construction can be any rectangle. Warning: Not all the results are interesting. Neale Flexagons are paper structures that can be manipulated to bring different surfaces into view.” What follows is the result of my explorations. sometimes along the other axis. as black and white. Although the text refers only to a square. the first version of the square slash-slit (for the minimalist base and puzzle subtlety). Square window. 117 . But some are. Note that the text assumes you are using two-colored paper. and the third version of the square cross-slit (for its self-design). for simplicity. and then opening them up a different way to reveal a new face. Although usually constructed from a strip of paper with attachment of the ends. Interesting designs of the faces can be found when the material is paper.Self-Designing Tetraflexagons Robert E. Figure 2. I call these “self-designing. or cardboard. referred to. see Figures 1–5). my favorites being three: the first version of the cross plus-slit (for its puzzle difficulty). and sometimes both. Figure 1.

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. It has six faces that are easy to find. that there are two ways of constructing this flexagon.O. The result is shown in Figure 7. Booklet No. so is a separate manuscript.” This involves decorating the paper with markers. Square cross-slit. It should be noted. . both of which are discussed below. Square plus-slit. It is a hexa-tetraflexagon made from a “window” — a square (of 16 squares) with the center (of 4 squares) removed (Figure 1). Figure 5. Square Window The first flexagon I saw that did not need glue was shown to me by Giuseppi Baggi years ago. 1978. As the dotted line indicates in Figure 7. B. however. Same Way Fold.118 R. valley fold the left edge to the center.S. E. As the dotted line indicates in Figure 6. NEALE Figure 3. Square slash-slit. The result is ½Directions for constructing the hexa-tetraflexagon are found on pages 18–19 of Paul Jackson’s Flexagons. Figure 4. 11. valley fold the upper edge to the center. England.



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.



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

Figure 16.



Figure 17.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

691 combinations into 148 disjoint sets is described in Section 6.105 combinations are isomorphisms.251. obtained after some experimentation. It becomes necessary. The individual dicubes can be colored thirty-six different ways. and only 10. This was achieved through the use of a computer program.700 (36!/12!24!) different ways of which. then. The second observation is that for a cube to be assembled with the nine colors on each face. only 133. this rule then tells the experimenter to backtrack. The method of sorting the 10. Since one person can assign the fourth and fifth colors to orange and red and another person to red and orange. The third observation.691 combinations need to be sorted into disjoint sets. the results are shown in Table 1. as determined in Section 4. some with the tricube on the edge and some with the tricube through the center of the cube. when two cubes of the same color are in position. the location of the third cube is determined. A proof is given in Section 3. The first observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube. Twelve of these can be selected in 1. only thirty dicubes are required. but not in the center of a face. . none of the nine planes may contain two cubes of the same color.152 S. FARHI a difficult exercise for a moderately proficient programmer to verify the results obtained.677. Since during the course of trying to solve this puzzle a conflict often occurs. The final part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. it becomes apparent that most of the 133. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. as is shown in Section 5. in which the number of isomorphisms using the set of thirty-six dicubes is also determined. to separate these cases into disjoint sets of isomorphisms. A proof is given in Section 2. Thus. The previous question now becomes more interesting. is that there are numerous solutions. Fortunately.105 of these combinations meet the three-cubes-of-each-color requirement. and the most difficult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1. The question then arises: How many solutions are there? Most intriguing about this puzzle.

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

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

THE NINE COLOR PUZZLE 155 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 .

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.156 S.

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

such as shown by the letter “C. Since the corner colors are displayed on three faces. then a third dicube with color 2 and any color from the second row is selected. then color 3 is combined with two colors from the third row. . 5. and 14. 2. If color 2 is not selected from the first row then color 2 is combined with any two colors from the second row for the third and fourth dicubes. the edge case fills spaces 2.158 S.” The remaining six corners must have cubes of different colors. The “No Two in the Same Plane” Rule Figure 3. The corner case fills spaces 1. then none are selected from the third row. One cube is centrally located and thus not visible. 4. the center case fills spaces 5. Determination of the 133. 3. If color 3 has been selected once. If none of the previous selections include color 3. and on a centerface such as shown by the letter “B. FARHI If an irregular tricube is used. then one color is selected from the third row and if color 3 has been selected twice. 2. there are three possible locations. cubes of the same color are never in the same plane. the remaining two cubes must be located on an edge. The two dicubes with color 1 may use any two colors in the first row of the following table. and 11. 11. 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. If color 2 is selected from the first row. and 3.” In each case.” The remaining two colors must all be on edges.105 Possible Dicube Color Combinations The tricube colors are defined as colors 1. accounting for two faces. and 11.

8. 6. 3). 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. 8). 2). Identification of the second color of these ordered pairs is sufficient to define all the colors. 9). 5. 6).691 color combinations. This results in requiring only thirty cubes and. Identification and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. a computer program established 133. 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. 8). using the same algorithm as before. (2. etc. For example. (4. (3. (7. Similarly if color 2 is not associated with colors 3.105 possible color combinations. 6). 9. 9). (6. (7. 5). 8. 9) and (8. given the second numbers 2. (1. 4. 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. and 9 one can logically deduce the first colors by the three . 4. or 5. 9. (5. 7). 7. 3. 4). Using this algorithm. color 5 is combined with colors in the fifth row such that there are three cubes with color 5. (4. 6. 5. (5. the two new colors are identified as colors 6 and 7. only 10.

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

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

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

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

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

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

so I count it as a move and a half. . 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. we can glue one piece of the puzzle between two square plates as indicated in the top left corner of Figure 2. By using opaque plates the design is hidden as well. The ninth move is a slide-plus-rotate move.PLANAR BURRS 167 The pieces.

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. The number of such puzzles that I am aware of is quite small, but the “tricks,” or unique features that the puzzles employ are many and varied. I know of no other small group of puzzles that encompasses such a rich diversity of ideas. Presented here are 11 such “block-packing” puzzles. The tricks to most of the puzzles are discussed here, but complete solutions are not given. The puzzles are grouped according to whether there are holes in the assembled puzzle and whether the pieces are all the same or different. For each puzzle, the total number of pieces is in parentheses. If known, the inventor of the puzzle, date of design, and manufacturer are given.

7. No Holes, All Pieces the Same
“Aren’t these puzzles trivial?,” you ask. Well, you are not far from being completely correct, but there are some interesting problems. David Klarner gives a thorough discussion of this case in [5]. 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 ¢



8. No Holes, 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 pieces of the second type are smaller and easier to use, but must be used efficiently to solve the puzzle. The solver must determine exactly where the second set of pieces must be placed, and then the rest is easy.

¿º º

unnamed (9) (Slothouber–Graatsma): Box: ¿ ¢ ¿ ¢ ¿ Pieces: (3) ½ ¢ ½ ¢ ½, (6) ½ ¢ ¾ ¢ ¾ unnamed (18) (John Conway): Box: ¢ ¢ Pieces: (3) ½ ¢ ½ ¢ ¿, (1) ½ ¢ ¾ ¢ ¾, (1) ¾ ¢ ¾ ¢ ¾, (13) ½ ¢ ¾ ¢

In the first of these, the three individual cubes are obviously easy to place, but they must not be wasted. By analyzing “checkerboard” colorings of the layers in the box, it is easy to see that the cubes must be placed on a main diagonal. In the second design, the three ½ ¢ ½ ¢ ¿ pieces must be used sparingly. The rest of the pieces, although not exactly alike, function similarly to the ½ ¢ ¾ ¢ ¾ pieces in the first puzzle. See [2] or [5] for more information.

9. No Holes, Pieces Mostly Different


Quadron (18) (Jost Hanny, Naef): Box 1: ¢ ¢ Pieces: ¾ ¢ ¿ ¢ ¿, ¾ ¢ ¿ ¢ , ¾ ¢


¢ ,¾¢ ¢ ,¿¢¿¢ ,¿¢¿¢ ¢ , ¾¢¿¢


Box 2: ¢ ¢ ½¼ Pieces: ½ ¢ ¿ ¢ , ½ ¢ ¿ ¢ , ½ ¢ ¿ ¢ , ½ ¢ ¿ ¢ ½¼, ½ ¢ ¾ ¢ ¿ ¢ , ¾ ¢ ¿ ¢ , ¾ ¢ ¢ , ¿ ¢ ¿ ¢ ¿, ¢ ¢ Box 3: ¢ ¢ ½¼ Pieces: all 18 pieces from first two boxes


Quadron does not use any special tricks that I am aware of, but it does make a nice set of puzzles. The 18 pieces are all different, and the three boxes offer a wide range of difficulty. The small box is very easy. The seven pieces can be placed in the box in ten different ways, not counting rotations and/or reflections. A complete, rigorous analysis of the puzzle can be done by hand in about 15 minutes. The middle-size box is difficult — there is only one solution. The large box is moderately difficult, and has many solutions.



Quadron also makes for a nice entrance into the realm of computer analysis of puzzles and the limitations of such programs. The programmer can use algorithms that are used for pentominoe problems, but there are more efficient algorithms that can be used for block-packing puzzles. I wrote such a program on my first computer, a Commodore 64. The program displayed the status of the box at any instant using color graphics. 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. The Commodore 64 is such a wonderously slow machine — when running the program in interpreter BASIC, about once a second a piece is added or removed from the box! Using compiled BASIC, the rate increases to 40 pieces/second. When running these programs on more powerful computers, the difference between the three boxes is stunning: The first box can be completely analyzed in a small fraction of a second. The second box was analyzed in about a minute of mainframe computer time. In early 1996, I did a complete analysis of the third box. There are 3,450,480 solutions, not counting rotations and reflections. The analysis was done on about 20 powerful IBM workstations. The total CPU time used was about 8500 hours, or the equivalent of one year on one machine. By the end of the runs, the machines had constructed 2 1/2 trillion different partially filled boxes.


Parcel Post Puzzle (18) (designer unknown; copied from a model in the collection of Abel Garcia): Box: ¢ ½ ¢ ¾ Pieces: all pieces are of thickness 2 units; the widths and lengths are ¢ , ¢ ½ , ¢ ¾½, ¢ , ¢ ½¼, ¢ ½¿, ¢ ½ , ¢ ½ , ¢ ½½, ¢ ½¿, ½¼ ¢ ½½, ½½ ¢ ½½ and two each of ¢ , ¢ , and ¢ ½¿.

Since all the pieces are of the same thickness and the box depth equals three thicknesses, it is tempting to solve the puzzle by constructing three layers of pieces. One or two individual layers can be constructed, but the process cannot be completed. The solution involves use of the following obvious trick (is that an oxymoron?): Some piece(s) are placed sideways in the box. Of the 18 pieces, 10 are too wide to fit into the box sideways and 4 are of width 5, which is no good for this purpose. This leaves 4 pieces that might be placed sideways. There are four solutions to the puzzle, all very similar, and they all have three of these four pieces placed sideways.


Boxed Box (23) (Cutler, 1978, Bill Cutler Puzzles): Box: ½ ¢ ½ ¢ ½ Pieces: ½¿ ¢ ½½¾ ¢ ½ ½, ½ ¢ ¼ ¢ , ½ ¢ ¢ ¼, ½ ¢ ½ ¢ ¾ ¢ , ½ ¢ ¾ ¢ ¾, ½ ¢ ¿ ¢ , ¾¼ ¢ ¼ ¢ ¾, ¾½ ¢

¢ ½ ¼, ¾¢ ,



¾¾ ¢ ½¼ ¢ ½¿½, ¾¿ ¢ ½ ¢ ¿, ¾ ¢ ¿¼ ¢ ¢ ½¿ , ¿½ ¢ ¢ , ¿¿ ¢ ¿ ¢ ¿ ¢ ½¾½, ¿ ¢ ¾ ¢ ¼, ¢

¢ , ¾ ¢ ¿ ¢ , ¾ ¢ ¢ ½¾¿, ¢ ¼, ¿ ¢ ½½¼ ¢ ½¿ , ¿ ¢ ¾ ¢ ½¾ , ¢ , ¢ ¢

The dimensions of the pieces are all different numbers. The pieces fit into the box with no extra space. The smallest number for which this can be done is 23. There are many other 23-piece solutions that are combinatorially different from the above design. Almost 15 years later, this puzzle still fascinates me. See [1] or [3] for more information.

10. Holes, Pieces the Same or Similar


Hoffman’s Blocks (27) (Dean Hoffman, 1976) Box: ½ ¢ ½ ¢ ½ Pieces: (27) ¢ ¢

This sounds like a simple puzzle, but it is not. The extra space makes available a whole new realm of possibilities. There are 21 solutions, none having any symmetry or pattern. The dimensions of the pieces can be modified. They can be any three different numbers, where the smallest is greater than one-quarter of the sum. The box is a cube with side equal to the sum. I like the dimensions above because it tempts the solver to stack the pieces three deep in the middle dimension. See [4].


Hoffman Junior (8) (NOB Yoshigahara, 1986, Hikimi Puzzland) Box: ½ ¢ ½ ¢ ½ Pieces: Two each of ¢ ¢ ½¼, ¢ ¢ ½½, ¢ ½¼ ¢ ½½, ¢ ½¼ ¢ ½½

11. Holes, Pieces Different


Cutler’s Dilemma, Simplified (15) (Cutler, 1981, Bill Cutler Puzzles) Box: ¼ ¢ ¾ ¢ ¾ Pieces: ¢ ½ ¢ ¾ , ¢ ¾¼ ¢ ¾¼, ½¼ ¢ ½½ ¢ ¾, ½¼ ¢ ½¾ ¢ ¾ , ½¼ ¢ ½ ¢ ¿½, ½¼ ¢ ½ ¢ ¾ , ½¼ ¢ ½ ¢ ¾ , ½½ ¢ ½½ ¢ ¾ , ½½ ¢ ½¾ ¢ ¾, ½½ ¢ ½ ¢ ½ , ½½ ¢ ½ ¢ ¿¼, ½½ ¢ ½ ¢ ¾ , ½¾ ¢ ½ ¢ ½ , ½ ¢ ½ ¢ ¾½, ½ ¢ ½ ¢ ¾½

The original design of Cutler’s Dilemma had 23 pieces and was constructed from the above, basic, version by cutting some of the pieces into two or three smaller pieces. The net result was a puzzle that is extremely difficult. 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.



12. Miscellaneous


Melting Block (8-9) (Tom O’Beirne) Box: ¢ ¢ ½¿¿ Pieces: ½ ¢ ¾ ¢ , ½ ¢ ¾ ¢ ½ ¢ ¢ ,¿ ¢¾ ¢ ,¿ ¢ ¢ plus a second copy of ½ ¢ ¾ ¢

,½ ,¿

¢ ¢ ¢ ¢



¢¾ ¢


The Melting Block is more of a paradox than a puzzle. The eight pieces fit together easily to form a rectangular block ¢ ¢ ½¿¾. This fits into the box with a little room all around, but seems to the casual observer to fill up the box completely. When the ninth piece is added to the group, the pieces can be rearranged to make a ¢ ¢ ½¿¿ rectangular solid. (This second construction is a little more difficult.) This is a great puzzle to show to “non-puzzle people” and is one of my favorites. By the way, one of the puzzles listed above is impossible. I won’t say which one (it should be easy to figure out). It is a valuable weapon in every puzzle collector’s arsenal. Pack all the pieces, except one, into the box, being sure that the unfilled space is concealed at the bottom and is stable. Place the box on your puzzle shelf with the remaining piece hidden behind the box. You are now prepared for your next encounter with a boring puzzle-nut. (No, readers, this is not another oxymoron, but rather a tautology to the 99% of the world that would never even have started to read this article.) Pick up the box and last piece with both hands, being careful to keep the renegade piece hidden from view. Show off the solved box to your victim, and then dump the pieces onto the floor, including the one in your hand. This should keep him busy for quite some time!

[1] W. Cutler, “Subdividing a Box into Completely Incongruent Boxes,” J. Recreational Math., 12(2), 1979–80, pp. 104–111. [2] M. Gardner, Mathematical Games column in Scientific American, February 1976, pp. 122–127. [3] M. Gardner, Mathematical Games column in Scientific American, February 1979, pp. 20–23. [4] D. Hoffman, “Packing Problems and Inequalities,” in The Mathematical Gardner, edited by D. Klarner (Wadsworth International, 1981), pp. 212– 225. [5] D. Klarner, “Brick-Packing Puzzles,” J. Recreational Math., 6(2), Spring 1973, pp. 112–117.

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

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

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

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

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

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

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

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

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

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

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

” “Find the 5th Pig” needing coloured overlays.” Penrose triangle “Vanishing Leprechauns. “Spot the Difference. OHOs. impossible dovetails.” random dot stereograms “Naughty Butterflies. hold to light advertisements Obscure non-rebus heiroglyphic prints Requiring red and green glasses for either 3-D or movement effects Framed strip prints showing different views from different directions . weight illusions AMB-VANI AMB-DIST AMB-ARCH Distortions Archimboldesque objects AMB-HIDD Hidden image pictures (no manipulation required) Hidden image pictures (manipulation required) AMB-HMAN AMB-TURN Pictures that require turning to show different images Perception illusions Ephemera related to puzzles Wiggle Woggle HTL Micro Printing Moire effect Hold to light Heiroglyphs (non-rebus) Anaglyphs AMB-ILLU EPHEPH-WIWO EPH-MICR EPH-MOIR EPH-HTL EPH-HEIR EPH-ANAG EPH-STRP Strip prints (three views in one frame) Hold to light cards producing movements by shadows Images concealed by extreme smallness Puzzles/effects produced by moire/fringe effects Protean views. Oskar’s “Escher Puzzle.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. courtship/matrimony Optical illusions. Topsey Turveys.” 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.

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

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

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

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

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

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

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

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

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

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

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

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

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

204 B. GOSPER Times 100–105: .

In 1989 I wrote a program to search for oscillators and spaceships. It found many oscillators of periods 3 and 4 (including some smaller than any previously known.STRANGE NEW LIFE FORMS: UPDATE 205 So. 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. For a few months I ran it almost constantly.” 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. David Bell. and I’ve run it occasionally since then. and myself Oscillator and Spaceship Searches. 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. Bill Gosper. and myself Large constructions mostly by Buckingham. Bill Gosper. and a period 4 .

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

Glider Syntheses. A honey farm starts to form but is modified by an eater and a block. 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. based on a period 72 device discovered by Bob Wainwright. it uses various mechanisms to interleave larger period streams. If you delete the traffic light. you (JHC) described how to thin a glider stream by kicking a glider back and forth between two streams. of size 149 by 149. Subsequent gliders grow a crystal upstream. Fortunately. Buckingham has also found extremely weird and elegant guns of periods 144 and 216. In addition to the long-familiar period 30 and period 46 guns. forms a traffic light. two gliders delete one pair of beehives. the cycle repeats every ¢ ¾ generations. centinal-based ¼¼ · ¾¼¼Ò. forming gliders that stabilize the ends of a row.) It’s also possible to build a gun that produces a glider stream of any period ½ . the resulting period must be a multiple of 8. New Glider Guns. The gun itself has a larger period. which hit eaters. based on the period 100 centinal. The largest by far that’s actually been built is Bill’s period 1100 gun (extensible to ¼¼ · ¾¼¼Ò). emit gliders that can crash to form MWSSs. Buckingham mercifully outmoded it with a comparatively tiny. A close pair of AK47s can delete each other’s traffic lights. the process begins again. there are other ways to get any multiple of 30.112. with a total of 36 AK47s. there’s now a fairly small period 44 gun that Buckingham built recently. its bounding box is 13. It’s fairly easy to get any period down . 11 gliders add a pair of beehives to the crystal. an eater stops its growth. Rich has already mentioned this. In the period 94 gun. they don’t work with period 30 streams.STRANGE NEW LIFE FORMS: UPDATE 207 180Æ and collides with the second to form a honey farm. it’s based on the “AK47” reaction discovered independently by Dave Buckingham and Rich Schroeppel. When they’re all gone. Using normal kickbacks. so we can build a long row of them that is unstable at both ends. two such rows. The other guns are all pretty big. It emits a glider. and it begins to decay. I’ll just add that some of Buckingham’s syntheses of still lifes and billiard table oscillators are aweinspiring. The period 94 gun is 143 by 607.584 by 12. (Sadly. The period ½¿ · Ò guns based on Buckingham’s B heptomino turns are also smaller. For example. When the crystal reaches the guns. and then starts forming another honey farm in a different location. Here’s how to turn 2 MWSSs into a glider: In Winning Ways. I’ve found some eaterassisted kickbacks that give any even period.

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

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

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

you can waste computer time all night long and nobody says anything. never imagined they could be made by colliding gliders. perhaps the ideas wouldn’t have prospered so well.) “Martin Gardner is a skilled presenter of ideas. “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. on just that level of computer. Not to mention that the time was ideal.STRANGE NEW LIFE FORMS: UPDATE 211 ºººººººººººººººººººººÓººººººººººººººº ºººººººººººººººººººººÓºÓººººººººººººº ºººººººººººººººººººººÓÓºººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººÓºººººººººººººººººººººººººººººº ººººÓºÓººººººººººººººººººººººÓºÓººººº ºººººÓÓººººººººººººººººººººººÓÓºººººº ººººººººººººººººººººººººººººººÓºººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººÓººººººººººººººººººÓººººººº ºººººººººººÓÓººººººººººººººººÓºÓººººº ººººººººººÓÓºººººººººººººººººÓÓºººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººÓÓÓºººººººººººº ººººººººººººººººººººººÓºººººººººººººº ºººººººººººÓºººººººººººÓººººººººººººº ºººººººººººÓÓºººººººººººººÓÓÓººººÓÓºº ººººººººººÓºÓºººººººººººººÓººººººÓºÓº ºººººººººººººººººººººººººººÓºººººÓººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ÓÓººººººººººººººººººººººººººººººººººº ºÓÓºººººººººººººººººººººººººººººººººº ÓºººººººÓÓººººººººººººººººººººººººÓÓº ºººººººººÓÓºººººººººººººººººººººººÓºÓ ººººººººÓºººººººººººººººººººººººººÓºº ººººººººººººººººººººººººººººººººººººº ºººººººººººººººººººººººººººÓÓºººººººº ºººººººººººººººººººººººººººÓºÓººººººº ºººººººººººººººººººººººººººÓººººººººº ººººººººººººººººººººººººººººººººººººº ºººººººººÓÓºººººººººººººººººººººººººº ººººººººÓºÓºººººººººººººººººººººººººº ººººººººººÓºººººººººººººººººººººººººº Conway called such oscillators billiard tables. and along with the rest of us. if all those computer hours hadn’t been around to waste. ººÓÓºÓÓººººº ººÓºººÓººººº ºººÓÓÓºººººº ºººººººººººº ºººººÓÓÓºººº ººººÓºººÓºÓÓ ºÓººÓºººÓºÓÓ ÓºÓºÓºÓºÓººº ºÓÓºÓºººÓººº ºººººÓÓÓºººº ºººººººººººº ºººººººÓÓÓºº ººººººÓºººÓº ººººººÓÓºÓÓº Finally. and Life was an excellent idea for him to have the opportunity to present. nobody has yet . That raises the question. 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. ‘Have other hours of computer time been spent more profitably?’ (Nowadays. after a trial run of the foregoing around the Life net.

) “Nor should it be overlooked that there is a more serious mathematical theory of automata. “It might be worth mentioning the intellectual quality of Gardner’s presentation of Life. —Bill Gosper ØØÔ »»ÛÛÛºÑ Ò ×ÔÖ Ò º ÓÑ» Ð Ò »Ð Many remarkable discoveries followed this writing. of all the games. Nor that there are things about Life. So there is really something there which is worth studying. and other automata. See. at least). which certainly owes something to the work which has been performed on Life.212 B. » Ò Üº ØÑÐ . puzzles and tricks that made their appearance in his columns over the years. that can be foreseen by the use of the theories that were stimulated by all the playing around that was done (some of it. for example. GOSPER come up with another automaton with anything like the logical intricacy of Life. did anything excite nearly as much curiosity? (Well. there were flexagons.” The Life you save may not fit on your disk.

37. 43. 31. These are representative oscillators of the (small) periods (indicated by the still-Life numerals) known by the end of 1992. 23. . 49. As of the end of 1998. and 53. 41. 38.STRANGE NEW LIFE FORMS: UPDATE 213 An early “stamp collection” by Dean Hickerson. 27. all periods have been found except 19.

but only to a region of a flat surface (Figure 2). The first part is about multiple silhouettes in general. In this article I shall give a more detailed description of the subject.Hollow Mazes M. Figure 1. A silhouette defines a cylindrical region in space.. e. K. Oskar van Deventer In October 1983 I discovered hollow mazes. With this meaning the word no longer refers to an object.g. Multiple Silhouettes Silhouettes are often used in. A. a triangular. mechanical engineering (working drawings) and robotics (pattern recognition). Wedge of Wallis. Dewdney introduced the term projective cast for an object defined by a number of silhouettes. the second part is about their application to hollow mazes. Not every object can be defined as the projective 213 . and a square silhouette?” The solution is the well-known Wedge of Wallis (Figure 1). One particular mathematical problem reads: “Which solid object has a circular. Dewdney presented the concept in Scientific American. September 1988. 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. perpendicular to the silhouette surface.

cast of a (limited) number of silhouettes: One cannot cast a hollow cube or a solid sphere.214 M. Dewdney used the term viable for a silhouette or a maze that consists of one part and has no cycles. (b) The cylindrical region in space defined by it. no cycles. (a) Negative of a silhouette. Figure 3. O. Topological properties of mazes and graphs: (a) multiple parts with cycles. (b) one part with cycles. no cycles. a graph may consist of either one part or multiple parts (Figure 3). to projective casts. a graph may have either cycles or no cycles. and to mazes and graphs in general. Second. (d) one part. There are two interesting topological properties that apply to silhouettes. V AN DEVENTER Figure 2. First. Even when silhouettes are viable. The basic problem is the analysis . 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. (c) multiple parts.

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

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

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

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

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

when are and integers? For example. A person. Her age to mine exactly bears. more comprehensible. As eight is to sixteen. what were their ages on their wedding day? We can state the general form of this problem as follows. reportedly from The American Tutor’s Assistant. version. 1824 [5]. The solution of these is easily ´¿µ ´   ½µ   ´   ½µ   assuming . 15. Since is integral if is. It is easily seen that the problem leads to the equations ´¾µ where and found to be denote the ages of · and ´ · µ . at the time he was married. is times as Thus the problem in The American Tutor’s Assistant and Bonnycastle is Problem (3. . et al. the problem is quoted in Bunt. Problem (4. but with the same numbers. µ. was 3 times as old as his wife. The type that I want to examine is illustrated by the following. after years. Problem ´ old as . Now tell. [6]. he was only twice as old. Now we give a diophantine aspect to our problem by asking the question: If . 3.½ When first the marriage knot was ty’d Between my wife and me. appears at about the same time in Bonnycastle. we can give a pretty simple answer: is integral if and only if ½I haven’t actually seen the book. which is the earliest I have found. but after they had lived together 15 years. from what I’ve said. . 1791. is now times as old as . albeit less poetic. My age was to that of my bride As three times three to three But now when ten and half ten years. 2) does not have an integral solution.220 D. SINGMASTER Between these are a number of types of problem. What were our ages when we wed? A more typical. I pray. 2). and are integers. We man and wife have been.

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

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

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

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

there seems to be no common name for it. 7) I-(7. 4. 10) (7. 4. 2. although the material is believed to be based on older Arabic sources. (1. and Bhaskara [4]. and to the youngest ten apples. c. 2. 2. 9. 2. 1. 5. with the problem statement giving the 6 as a 7 Fibonacci Fibonacci. 50 in the first printed English riddle collection “The Demaundes Joyous. 2. 3) (10. 5. 4. and brought home in like such money. 10. noting that this system is inconsistent Fibonacci. 7) II-(1. 4) (2. 3.” printed by Wynken de Worde in 1511 [9]. and to the second thirty apples. 4. certain apples. 4. 5. A man had three daughters of three ages. 850 c. and all these three sold in like many for a penny. 4. 50. 510 c. 510 9Ø century c. hence the above title. to sell. 6) I-(7. 3. I quote a version that appears as No. 5. 10. 5) c. Now how many sold each of them for a penny? . 6. 11. 3. 2. Sridhara [18]. 6) I-(7. 9. 250 c. 9. 11.5) I-(25. 7) I-(7. which daughters he delivered. 2. 10. which is rather a mouthful. 5) (100. And he took to the eldest daughter fifty apples. 6) I-(7. with the problem statement giving the 6 as a 7 Fibonacci.SOME DIOPHANTINE RECREATIONS 225 from the early fourteenth century and hence after Fibonacci. 5) (2. 9. 1) (30. 3. 22. 3. done two ways Selling Different Amounts at the Same Price Yielding the Same Although this problem is quite old. 3. 850 c. 300 B. 2. 3. 1.C. 5. 8. 3. 2. 1. 11. 0. 1. 6) (1. 8. 5. but this form is less clearly expressed and has infinitely many solutions. 23. 10. 9. 5. 11. 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. 11. so I will first describe the later form which is first found in Fibonacci [11]. The problem is treated in one form by Mahavira [15]. 2) I-(9. 11.

The 1708 English edition of Ozanam [17]. Although the objects are usually eggs. 40). Let us denote the problem with numbers ½ ¾ by ´ ½ ¾ µ. for which he often gives fractional answers. but we both conjecture that he was counting the positive solutions. (12. The extensively revised and enlarged edition of 1725 (or 1723) outlines a fairly general method and correctly says there are ten solutions. 25. and he gives five solutions for his first version — but there are 55. or raising them late in the day when few items are left and new buyers arrive. Even with this trick exposed. he added several more versions as well. He gives a fairly general rule for generating solutions [12]. Fibonacci has no truck with trick questions and gives the necessary information — the sellers sell part of their stock at one price and then the remainders at another price. 30. or sometimes fruit. such as the first receives twice the second. During the 1980s. find solutions where the amounts received are in some ratio. which would produce the positive solutions with the smaller price equal to 1. 33). it can take a fair amount of time to work out a solution. which normally are all of the same value. SINGMASTER ‘In like many’ means ‘the same number’ and ‘in like much money’ means ‘the same sum of money. is there a solution. dropping Bachet’s rule and replacing it with some rather vague algebra. Fibonacci goes on to consider solutions with certain properties — e. 32). Fibonacci only gives two-person versions: (10. Tartaglia [19] justifies the different prices by having large and small pearls. I tried several times to find an easy way to generate the solutions. contrary to De Lagny’s statement that there are six solutions.. or lowering prices late in the day in an attempt to sell their remainders. of which 36 are positive. In 1874. distinctly against the spirit of the problem. which is essentially the same as the first edition of 1694. This is historically by far the most common version. 30). so that the above is (10. considers (10. of which there are indeed six. 50). or simply changing prices after lunch. The question of finding all solutions for a three-person problem seems to have first been tackled in the 1600s. (12.’ Anyone meeting this problem for the first time soon realises there must be some trick to it. the 1612 and 1624 editions of Bachet [3] ¾ give a fairly general rule for this case and apply it to (20. 30) and gives two solutions. I have now arrived at a reasonably simple approach that turns ¾I have not seen these editions. Labosne revised the material.226 D. Neither Glaisher nor I have seen the relevant work of De Lagny. . and it is not immediately obvious how to find or count all the solutions. 30. According to Glaisher [12]. Fibonacci is one of the few to give more than a single solution.g. This may happen by the sellers going to different markets. if the prices are also given. find all solutions with one price fixed.

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

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

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

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

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

40. 1300 1300s 1489 1489 1489 c. 36) (25. 90) Date 1202? 1202? 1202? c. 29. 1 given. 49) (3. 1 given. 55) (78. 1 given. 55) (55. 4 given. 1 given. 33) (10. 1 given with fractions. 1 given. 48. .6) # of Solutions (55. 33) (20.232 Version (10. 752.400) (25.552. so we again have a three-dimensional solution set. 32. 196) (45. all given. 170) (305. 1) (7. 82) (17. 20. 56. 1 given. 36) (30. 1 given. 1 given. 40) (18. 20) (64. though it is not as easy to see how to do this as in the Western version. 454. 30) (18. 32) (12. 603. 16) D. 60) . (20. 50) (30. 30) (8. 36) (78. 140) . 21) (225. 25. 50) (11. 40. (20. 2 given. 1 given. 901) (10. 80) (10. 55) (16. 40) (10. ¬ . in some way to reduce the solution set to two dimensions. none given. 30. . 30. SINGMASTER Solutions in Source 5 given. 60) (117. 119. 60) (10. 26) (20. one with fractions.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. . 81) (55. 1 given. 1 given. 1 given.9) (100. I must point out that some Indian versions give and/or as fractions! The Indian authors do not get very general solutions of these problems. 1 given. 37) (27. 12. 1500 1513 1515 1521? 1556 1556 1556 1612 1725? 1874 1874 1874 1874 1893 1905 1924 Source Fibonacci Fibonacci Fibonacci “Abraham” Munich codex14684 Widman Widman Widman Pacioli Blasius Tagliente Ghaligai Tartaglia Tartaglia Tartaglia Bachet Ozanam from De Lagny Labosne Labosne Labosne Labosne Hoffmann Dudeney Glaisher (171. Sridhara finds a one-parameter solution set. 100) (45. 40. 16) (16. makes this more specific by saying they ‘sold a part in lots and disposed of the remainder at one for five panas. 36) (27. 20) (10. 17. 1 given. 20. At this point. 50) (10. 68. 35) (11.4) (70. none given. 15) (31. 30. 1) (100. 33. 81) (10. 1 given. 136) (3. 6 given. 11. 30) (12. One can scale the set . Bhaskara makes the following comment: “This.

i. we get ¬ ¬  ½   ´ ¬   ½µ « divides and the condition for all the equations to have integral solutions is   If we set ´ ¼¼ µ then ´ ¼µ gives ´ ¬   ½µ ­   We need to determine which and ¬ will make ´ ¼¼ µ hold. (Ayyangar notes only that ´  µ must divide . owing to restriction. then subtracting to eliminate and using the same abbreviations as before. the answer must by the intelligent be elicited by the exercise of ingenuity. we let ¬ ¬ .. and such a supposition introduced. but in order for «½ ­Ò   ½ . so our equations become · ¬¬ ·¬ ´¾¼¼ µ ´ ¼µ ´ ¼µ Ayyangar gives a reasonable solution process for this problem. which is if and only if Ö divides ­ ´  µ.) Both and ¬ can take on an infinite range of values. as has brought out a result in an unrestricted case as in a restricted one.e. ´ ¼¼¼µ ´ ¬   ½µ ¼ ´mod  µ Let ­ ´  µ. disappoints. which is if and only if Ö divides . Then ´ ¼¼¼ µ is equivalent to ¬   ½ ¼ ´mod   ­ µ and this is solvable if and only if ´   ­ µ ½. has been by some means reduced to equation. Then   ­ is relatively prime to if and only if   ­ divides ×.SOME DIOPHANTINE RECREATIONS 233 ´½¼¼ µ which is instanced by ancient writers as an example of a solution resting on unconfirmed ground. and ´  µ can be a proper divisor of Ö. we must have Ò ´ ¼µ ¬ ½ ´ ¼µ « . Eliminating ¬ . pointing out an improvement. when the operation. but since his notation is different. where Ö comprises all the prime powers in   whose primes also occur in . Write   Ö×.” Amen! For ease of expression. In the like suppositions. I will sketch the solution in the present notation. This gives us a rather peculiar condition.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

For composite Æ up to 16. solutions were found by a systematic trial and error search using a regular Æ -gon grid.250 M. and many questions remain open: ¯ ¯ ¯ ¯ ¯ Is there a better search strategy for composite Æ ? What are the properties of solutions for Æ ½ and higher? Are there more Æ with multiple solutions? Can a sharper theoretical maximum for odd and composite expressed in a closed form? Are there solutions that use the back side of the mirrors? Æ be . V AN DEVENTER Figure 8. Deleted paths for composite Æ up to 16. O. No general strategy was found. which can be reached for prime Æ .

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

The ¿Ü · ½ problem concerns the behavior under iteration of the ¿Ü · ½ function Ì . 253 ¾ . Huge numbers of computer cycles have been expended studying it. the largest integer occurring . The ¿Ü · ½ Conjecture combines simplicity of statement with apparent intractability. Simple probabilistic models appear to model the average behavior of the initial iterates of a “random” positive integer Ò. Lagarias Abstract. The predictions of these models are consistent with empirical data for the ¿Ü · ½ function. where Ì ´¾µ ´Òµ Ì ´Ì ´Òµµ. in a trajectory starting from an integer Ò should be of size Ò ¾´½·Ó´½µµ as Ò and trajectories should be of length at most ½ ÐÓ Ò before reaching ½. see Oliveira e Silva [19]. For example. ¾ ½ 13. Introduction The well-known ¿Ü · ½ problem concerns the behavior under iteration of the ¿Ü · ½ function Ì . This paper surveys results concerning more sophisticated probability models. it has now been verified for all Ò ¾ ¼¾ ¢ ½¼½ . which predict the behavior of “extreme” trajectories of ¿Ü · ½ iterates.How Random Are 3x + 1 Function Iterates? Jeffrey C. etc. defined by ´½¿º½ µ Ì ´Òµ ¿Ò·½ ¾ Ò if Ò is odd if Ò is even The 3x+1 Conjecture asserts that for each positive integer Ò there is some iterate with Ì ´ µ´Òµ ½. In particular. 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 .

This basic property was independently discovered by Everett [8] and Terras [21]. and now more than one hundred papers have been written about the problem. ´½¿º¾¿µ Ö´Òµ The ¿Ü · ½ Conjecture is made plausible by the observation that ¿Ü · ½ iterates behave “randomly” in some average sense. Thwaites in 1952. The total stopping time ½ ´Òµ of Ò is the minimal number of iterates needed to reach ½.. Müller [18]. 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.. Furthermore the resulting pattern is periodic in Ò with period ¾ .e.e. 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. each possible binary pattern occurs exactly once in ½ Ò ¾ . C. Finally we let Ö´Òµ count the fraction of odd iterates in a trajectory up to the point that ½ is reached. To describe iterates of the ¿Ü·½ function we introduce some terminology. [21]). . The first mathematical papers on it appeared around 1976 ([8].e. This randomness property can be formalized by taking an input probability distribution on the integers and looking at the probability distribution resulting from iterating the ¿Ü · ½ function some number of times. i. It was also independently discovered by B. see [22]. LAGARIAS The ¿Ü·½ problem appeared in Martin Gardner’s “Mathematical Games” column in June 1972 [10]. It is usually credited to Lothar Collatz. i. Ò Ó ´½¿º¾½µ Ì ´ µ´Òµ ½ ½ ´Òµ Ñ Ò ´½¿º¾¼µ Ç´Òµ where ½ ´Òµ ½ if no iterate equals ½. and Wirsching [25]. i. We let Ø´Òµ denote the largest value reached in the trajectory of Ò. Surveys of the known results on the ¿Ü · ½ problem can be found in Lagarias [13]. Prior to that it circulated for many years by word of mouth. if one takes the uniform distribution on ½ ¾ . who studied problems similar to it in the 1930s. so that ´½¿º¾¾µ Ø´Òµ ×ÙÔ Ì ´ µ ´Òµ Ò ¼ Ó where Ø´Òµ ½ if the sequence of iterates is unbounded.254 J. and who has stated that he circulated the problem in the early 1950s [6].. For example.

which may be enough for a pseudorandom bit to be extracted. [24]. et al. This model has the drawback that . Cloney. [24]. In doing so we are in the atypical situation of modeling a purely deterministic process with a probabilistic model. Since a number Ò is multiplied by ½ if it is even and approximately ¿ if it is odd. From this viewpoint it becomes interesting to determine how well the properties of ¿Ü · ½ iterates can be described by a stochastic model.3X + 1 FUNCTION ITERATES 255 One may summarize this by saying that the parity of successive iterates of Ì initially behaves like independent coin flips. obtained in joint work with Alan Weiss and David Applegate. [13]. These mathematical models predict that the expected size of a “random” Ò after steps should be about ´ ¿ µ ¾Ò. and a branching random walk model for backward iterates of Ì . 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. see [9]. Even though ¿Ü · ½ function iterates possess quite a bit of structure (cf. Garner [11] and Korec [12]) they also seem to possess some residual structurelessness. The simplest of them assumes that this independent coin-flip behavior persists until ½ is reached under iteration. There is excellent numerical agreement of this model’s prediction with ¿Ü · ½ function data. [5] proposed using the ¿Ü · ½ function as a pseudorandom number generator. Here we survey some recent results on stochastic models. Lagarias [14] and Luby [17]. one expects the iterates to decrease in size and eventually become periodic. cf. On the positive side. 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. 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. [20]. 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. Lagarias and Weiss [15] studied two different stochastic models for the behavior of ¿Ü · ½ function iterates. Several heuristic probabilistic models describing ¿Ü ·½ iterates are based on this idea. one expects that ¾ ¾ on average it changes multiplicatively by their geometric mean. which is ´ ¿ µ½ ¾ . [15]. which concern extreme behaviors of ¿Ü · ½ iterates. These models consist of a repeated random walk model for forward iterates by Ì . 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. Since this is less than ½. see [20]. for an ¾ explicit constant ½.

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

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

Figure 2. and ending at ´ ¾µ. Scaled trajectories of Ò maximizing ´ÐÓ ½¼ ·½ (dotted for ½ . maximizing ­ ´Òµ in ½¼ Ò ½¼ ·½ (dotted ¾½ ÐÓ Ò. solid for ½¼). see Figure 2. and the second with slope  ¼ ½ ¿ for ÐÓ Ò ¾½ ÐÓ Ò. LAGARIAS Figure 1. Scaled trajectories of for ½ . starting from ´¼ ½µ. solid for Ò ½¼). and the graph (14.5) approximates the union of two line segments. Ø´Òµµ ÐÓ Ò in ½¼ Ò .258 J. starting at ´ ¾µ and ending at ´¾½ ¼µ. the first with slope ¼ ½¿½ for ½ ÐÓ Ò. C.

Extensions of the data can be found in Leavens and Vermeulen [16].94 31.769.371.5967 0.456.13 26.24 16.6020 0.171 4 52.5818 0.400. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory. .254 9 4. He found that Ò 12.68 24.5857 0.1) with numerical data up to ½¼ ½½ is given in Table 1.266.890. Section 5] shows that in ½¼ ½½ trials one should only expect to find a largest ­ ´Òµ ¿¼.6020 0. More recently V.815 10 13. ½ ´Òµ It gives the longest trajectory.527 has ½ ´Òµ ½¾ ½ and ­ ´Òµ ¿ ¾ ½ . The deviations from the stochastic model are well within the range that would be expected from the size of the expected standard deviation for such models.527 Random walks model ½ ´Òµ ­ ´Òµ 70 108 165 214 329 429 592 593 706 755 21.6030 0.77 31.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale. ½ ½¼. an analysis in [15.91 19.799 6 8. 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. Vyssotsky [23] has found much larger numbers with high values of ­ ´Òµ.68 Ö´Òµ 0. The agreement with the stochastic model is quite good.728.511 7 63.127 8 127. 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. A comparison of (14.6091 Table 2.64 32.91 32. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ­ ´Òµ ¿ ¾ .5927 0.4).48 18. where ­ ´Òµ ÐÓ Ò .180.527 5 837.194.38 41. and the large deviations extremal trajectory is indicated by a dotted line. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1. Maximal value of ­ ´Òµ in intervals ½¼ Ò ½¼ ·½.328.6026 0. The large deviations extremal trajectory is indicated by a dotted line.5741 0.5841 0. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good.884.

671 5 704.788 1. In fact.566.03 19. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ.75 5.83 3.791 1.897 2. respectively.560 1. where ½ ´½ º½µ Ì  ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree.362.05 19. 4.903 2.391 8 319. In iterating the multivalued function Ì  ½. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ. if Ò ¾ or 8 ´ÑÓ µ. these trees have the minimal and maximal number of leaves possible for depth . ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch.10 20.65 5.48 12. In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed. In this case. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ.) empirically like .528.84 19.14 11.06 4. 15.909 1.817. Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk. see [15] for more explanation.819 1.831 9 8.59 23.20 6.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½.86 13. .46 2.86 7. or 7 ´ÑÓ µ. LAGARIAS Ò 1 27 2 703 3 9.049. C.675 7 80.645 ­ ´Òµ 21.631. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿. using (15. (See Table 3 in Section 4. The set of inverse iterates associated to any Ò has a tree structure. Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2.663 4 77.559 Random walks model Table 3.24 16.099 1.260 J. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth .02 21. 5.511 6 6.790 1.32 2.65 7. there are duplications. ½ ½¼.511 10 77.976 1.804.1). so do not have any significant effect on the size of the tree.000 Ò 13.86 5.

4] show that the expected size of ÐÓ Ò£ ´ µ. The random walk at a given leaf node ends at a position Ù on the real line. . ­ È ­ÊÏ .. Theorem 4. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½. [15. where ¬ is the unique positive number satis£ ´¬ µ ¼.1). i. cf. In order to prove the result (15.3X + 1 FUNCTION ITERATES 128 64 32 16 8 4 40 20 10 5 13 224 112 56 28 14 7 74 37 261 Figure 3.2). and where fying £ ´ µ ×ÙÔ ´½ º¿µ   ÐÓ ´¾ · ½ ¾ ¿ ¿ ¼ This constant is provably the same constant as (14. Theorem 3. where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree. We assign labels ¼ or ½ to the edges. it was necessary to determine the distribution of the number of leaves in a tree in the branching process.e. 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 . satisfies (with probability one) ´½ º¾µ ÐÓ Ò£ ´ µ ­ ȳ ½ ÐÑ ½ ÐÓ This constant ­ È ¬  ½ . Pruned ¿Ü · ½ trees of depth Ò . Lagarias and Weiss [15. 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. and the value assigned to that vertex by the rules above is exactly Ù . In [15] it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ . and the root node is assigned the value 1. starting from root nodes Ò and The problem of determining the largest trajectory to get to 1 under iteration in a ¿Ü · ½ tree is equivalent to studying the leaf in a ¿Ü · ½ tree having the smallest value.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.

262 J. 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. see [15. (See Athreya and Ney [4] for a general discussion of such processes. Based on this data. Distribution of 3x + 1 Trees Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees. 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. C. 7. they proposed that the normalized extreme quantities remain bounded. Let ´½ º¾µ   ÐÑ Ò ½ ¼  ´ µ Ò   ½ · · Ð Ñ ×ÙÔ · ´ µ ½ Then these quantities satisfy the inequalities ½ Applegate and Lagarias [3] compared this conjecture with predictions derived from one of the branching random walk models of [15]. 5.2]. respectively. Ê´ µ counts the number of distinct edge-labeled ¿Ü · ½ trees of depth . Here the pruned 3x+1 tree Ì ´ µ is the tree of inverse iterates of the ¿Ü · ½ function grown backward to depth . 16. where ¼ represents the entering vertex being even and ½ represents the entering vertex being odd. In this table. Let Æ   ´ µ and Æ · ´ µ denote the minimal and maximal number of leaves. Theorem 3. with all nodes that have a label congruent to ¼´ÑÓ ¿µ removed. This can be formalized as Conjecture C . from a given root node . 4. 8 pictured in Figure 4. the corresponding random walk . The branching process is a multi-type Galton–Watson process with six types labeled 1. Applegate and Lagarias [1] empirically studied the extremal distribution of the number of leaves in a ¿Ü ·½ tree of depth as the root node is varied. Since the number of leaves is expected to grow like ´ ¿ µ .) The random walk arises from assigning the labels ¼ and ½ to the edges of the resulting tree. 2.

This value represents the initial value of the ¿Ü · ½ iteration starting from that vertex of the tree. respectively largest number of leaves. 1/3 each) 2 8 5 4 1. and ¢ for a fixed constant ¢ ½. Transitions of the branching process [9]. The number of leaves in actual ¿Ü · ½ trees empirically exhibits less variability than is predicted by this branching random walk model.3X + 1 FUNCTION ITERATES (prob.) We consider as random variables the smallest and largest number of leaves that occur among this set of trees. . 5 or 8 0 0 0 0 0 0 1 4 7 2 5 8 Figure 4.   ¾ takes a step ÐÓ ¾ along an edge labeled ¼ and takes a step ÐÓ ¿ along an edge labeled ½. and we consider the value of that vertex to be Ù . so that each vertex of the tree corresponds to a random walk ending at some position Ù on the real line. Let Æ  ´ µ and Æ · ´ µ denote the expected value of the smallest number of leaves. (For actual ¿Ü · ½ trees. The analogues of normalized extreme values in the models are ´½ º¿µ  ´ µ ¿ Æ  ´ µ Ò ·´ µ ¿ Æ ·´ µ In [3] it is proved½ that and Ð Ñ ×ÙÔ · ´ µ ÐÑ Ò ½  ´ µ ½ ¼ These results for the branching random walk model do not agree with Conjecture C above.4). The parent (bottom) always ¾Ò (edge label 0). and it yields a second child by the yields a child by the map Ò ½ multivalued map Ò ¿ ´¾Ò ½µ (edge label 1) if Ò ¾ or 8 mod 9. We make Ê´ µ independent draws of a tree of depth generated by this process. 4 or 7 1 7 263 (prob. 1/3 each) 2. The base vertex of the tree is located at the origin. choosing the root node uniformly ´ÑÓ µ. choosing Ê´ µ Ê´ µ ¿ ·½ so ½ ¢ ¿. The paper [3] presents more evidence in favor of Conjecture ½These facts are derived using the probability density for Ï given in (15.

77 133.026 2.758 1.166 2.557 1.243 0.491 0.490 0.633 0.77 2362.238 1.780 1.375 0.986 2.62 1328.49 9.451 0.422 0.898 1.482 0.302 0.246 0.307 0.792 1.669 1.750 0.738 1.688 1.292 2.03 177.211 2.401 0.310 2.405 2.76 23.746 1. Normalized extreme values for ¿Ü · ½ trees and for the branching process.738 1.190 2.751 1.923 1.259 0.118 2.489 0.47 996.500 1.16 4.688 1.12 74.507 0.232 2.255 0.344 2.500 1.375 2.492 0.490 0. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a fixed node Ò´ÑÓ ¿ ·½µ.481 0.633 0.481 0.747 1.09 56.742 1.99 13.898 1.281 0.240 0.352 0.68 31.390 2.300 0.83 1771.81 4199.289 0.784 1.360 2.475 0.748 0.762 1.489 0.67 0.054 2.50 315.076 2.741 1.268 0.327 2.75 5599.305 0.778 1. C.252 0.486 0.758 1.486 0.791 1.491 1.782 1.45 560.490 0.481 0.62 7.534 0. .534 0.750 1. 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 1.21 5.249 0.802 1.869 1.465 0.36 3149.419 2.802 1.433 10850 20436 38550 72806 137670 260612 493824 936690 1778360 3379372 6427190 12232928 23300652 44414366 84713872 161686324 308780220 100 134 178 237 311 413 548 736 988 1314 1744 2309 3084 4130 5500 7336 9788 114 154 206 274 363 484 649 868 1159 1549 2052 2747 Table 4.255 2.57 42.750 0.394 0.271 0.33 1.753 1.264 J.60 747.507 0.728 1.38 236.488 0.534 0.562 0.83 99.37 3.748 1.273 0.475 0.501 0.131 2.782 1.774 1.409 0.34 420.481 0.911 1.488 0.265 0.475 0.562 0.802 1.342 0.78 2.422 0.746 1.270 2.32 17.

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

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

Sign up to vote on this title
UsefulNot useful