P. 1
(eBook) Ma the Magic (Magic Tricks)

(eBook) Ma the Magic (Magic Tricks)

|Views: 10|Likes:
Published by siriuskar

More info:

Published by: siriuskar on Sep 05, 2011
Copyright:Attribution Non-commercial


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less







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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

Of course. I’m familiar with much of Ambrose’s excellent work. 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. He seriously maintained that Conan Doyle never wrote the Sherlock Holmes stories. Gardner. Broad: Do you believe that Doyle wrote the Holmes stories? Cranby: Of course! Why should I doubt it for one minute? Broad: Then how do you answer Gardner’s objection that no one with a mind so rational as to write the Holmes stories could possibly be so irrational as to get involved with spiritualism in the peculiarly anti-rational way that he did? Cranby: Oh. Broad: And do you recall the chapter. Cranby: Oh.16 R. where did you send it? Broad: To your Connecticut address. then I won’t get it for a couple of days. 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. Bad. Why do you ask? Broad: Well.One Hundred Years Later Professor Broad: Did you get my paper. What is it about? Broad: You know the twentieth century writer. and Doyle”? Professor Cranby: No. Martin Gardner? Cranby: Of course! I’m quite a fan of his. are you familiar with the Ambrose paper. ”Gardner and Doyle”? Cranby: No. Cranby: Good God! That’s ridiculous! That’s just as crazy as Gardner’s idea that Doyle didn’t write Holmes. certainly. come on now! That’s no objection! It’s obvious that Doyle. you remember his book. I think I have just about everything he ever wrote. SMULLYAN SCENE II . ”Ambrose. Science: Good. and Bogus? Cranby: Oh. with all his brilliance. What is it about? Broad: Well. but not this one. had an insane streak that simply got worse through the years. “The Irrelevance of Conan Doyle”? Cranby: Oh yes! As a matter of fact that is the strangest chapter of the book and is quite unlike anything else Gardner ever wrote. Of course Gardner wrote that chapter! Broad: Of course he did! .

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

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

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

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

Friday night At a minute to twelve Eve spin. our Galaxy — done! ½¼ ¼ meter = 1 meter = a Human Body ½¼ meters = Earth’s diameter ½¼½ meters = outer Solar System ½¼¾½ meters = Galaxy’s diameter ½¼ ¾ meters = Universe. barks out his orders for the week. form up. Cell’s girth Our bodies. and Speed Tim Rowett Space Seven steps each ten million to one Describe the whole space dimension The Atom. right? A week = 7 days corresponds to 14 Billion Years 1daycorresponds to2BillionYears(USA) 1 minute corresponds to 2 Million Years 1 Millisecond is 23 Years 23 . You. Adam delve In the last millisecond. and a bit more ½¼  meters = Nucleus of a human cell ½¼ ½ meters = Atom’s nucleus Time The Creator. Light Sun and Earth. Time. First thing on Monday.Three Limericks — On Space. the Earth Sun’s System. seen as an Army Sergeant Major. Bang!.

roughaveragespeedofEarth 700.4 million mph Galaxy’s speed through the debris of the Big Bang .24 T.000 mph turning speed of Galaxy 1.000mph. 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.

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

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

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

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

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

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

. . . . . . Genesis 32 : 15. and they . . Psalm 84 : 6. seventh hour the . Ë. I have . . . . Psalm 145 : 15. till seven Ë Î Æ. . . for example. . . . but none are strictly increasing or decreasing. Ê of good men. which is . . . . . from the . . . . . shoulder Aaron . 2 Corinthians 11 : 19. À Ï Agag in pieces . . again. . . violence of the . . Acts 2 : 45. . . . . Ï Î for a wave . (See also Genesis 27 : 44. remnant shall be . Suitable intermediate words can be found. . . Ecclesiastes 12 : 11. ÌÁÅ Ë? Jesus saith . Matthew 25 : 33. . . . . . (But ‘writ’ is not a Biblical word. in Revelation 3 : 11. times . . . . looked in the . they married . . a . . Ë Î Ê the wicked . . For example. . John 18 : 18. À Ï Ê of thy wood . . . Î Ê left him. . me. . . . . Ë Î : For he will . . Ë Ï with saws. .BIBLICAL LADDERS . and to the . . ÌÁÊ Ë shall be . . . and which . . Ì Ê Ë among the . Ê. they were . and . Ë for the name . . Micah 1 : 8. . . the mighty . .) Puzzle #5. . . . . Ezra 34 : 25. . . charity shall 33 . Ë Ï sackcloth . . . Job 40 : 17. . . . forth. Ï Î Ë. as a . Gather up thy . they were . . Ï Ê ye shall give . Daniel 3 : 21. Therefore . . . And Samuel . Matthew 6 : 28. Ezra 7 : 24. . . . hewed stones. . ( 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 ) . hospitality. Ê the multitude . . . And the soldiers’ . . use intermediate words found in Matthew 24 : 35. . . . . . hazarded their . ( Genesis 3 : 7 ) ( Zechariah 4 : 1 ) ( Exodus 29 : 27 ) ( Psalm 93 : 4 ) ( Luke 17 : 27 ) ( Acts 15 : 26 ) ( Ezekiel 21 : 21 ) ( Titus 1 : 8 ) ( 1 Peter 4 : 8 ) Æ Ã Ï Ã Ï Î Ï Î ÏÁÎ ÄÁÎ ÄÁÎ ÄÇÎ ÇÎ (Many other solutions are possible. .) Puzzle #6. . . . . . And your . Ruth 3 : 16. Numbers 1 : 2. . and sowed . . Ë of the sea. . . Ï Ê Ë out of . At his . .) Puzzle #4. and . . . . .

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

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

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

1:01:01. 64. For example. 46. 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. (Ignore the colons. 4. (1) Find the closest such times. 3:59:53. 60. x. 32. Problem 6: Which two numbers come at the end of this sequence? 2. 44. Problem 5: Calculate the expansion of the following 26 terms. y Problem 7: The figure shown here is the solution to the problem of dividing the figure into four identical shapes. 2:41:42. 66. 40. 36. etc. 52. 30. 23:55:32. .) These curious combinations occur 660 times a day. 42. 13:22:31. 50. (2) Find the two palindromes whose difference is closest to 12 hours. 34. 56. (3) Find the longest time span without a palindromic time. 54. 6. 62.38 N. YOSHIGAHARA Problem 4: Arrange the following five pieces to make the shape of a star.

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

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

after all. use these digits to form the next string or number. 4. But did you know that we have comparable bodies in recreational mathematics? At first glance. so write down 123. 3. Now repeat with 347. These are 3 (three evens). and 7 (seven is the total number of digits). counting evens. The Sisyphus String: 123 Suppose we start with any natural number. these bodies may be even more difficult to identify in the world of number play than their more famous brethren in physics. 347. Count the number of even digits. could numbers such as 123. to get 1. the small boulder he rolled up the hill would always come down at the last minute. The number 123 with respect to this process and the universe of numbers is a mathemagical black hole. 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. So.Number Play. never to escape. total number. such as 9. 4 (four odds). 41 . 2. What. the number of odd digits. never to escape. the physical universe has strange entities called black holes that pull everything toward them. No matter how Sisyphus tried. Ecker The legend of Sisyphus is a lesson in inevitability.759. and Card Tricks: Mathemagical Black Holes Michael W. pulled inexorably by gravity. If we repeat with 123. and 15 have in common with each other. regarded as a string. respectively. All numbers in this universe are drawn to 123 by this process. but much more so collectively because of the common theme linking them all. I call such individual instances mathemagical black holes. Like the legend. 6174. and the total number of digits. 153. Calculators. we get 123 again. odds.288.

odds. So. At this point. since 123 yields 123 again. and a final iteration from 303 produces 123.545 we find 4. and total are 20. respectively. 25. say 122333444455555666666777777788888888999999999 (or pick one of your own). One more iteration using 426 produces 303. and 6 for evens. W. our next iterate is 202. The numbers of evens. 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. 25. and 45. ECKER But will every number really be sent to 123? Try a really big number now. total. modify line 110 to allow the program to start again. If you wish. so we have 426 now. odds.545. the number obtained from 20.42 M. you can test a lot more numbers more quickly with a computer program in BASIC or other high-level programming language. 45. Iterating for 202. What Is a Mathemagical Black Hole? There are two key features that make our example interesting: . 2. any further iteration is futile in trying to get away from the black hole of 123. Or revise the program to automate the testing for all natural numbers in some interval.

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

then 6. However. try 163. but not necessarily with 4 as the black hole. Though this result is clearly language-dependent. Words to Numbers: 4 Here is one that master recreationist Martin Gardner wrote to tell me about several years ago. if n 1000. Add up the cubes to form a new . 3) — thus resulting in the claimed number of 123. Take any whole number and write out its numeral in English. We start with any positive whole number that is a multiple of 3. within one iteration you must get one of these four triples. Repeat with 4 to get 4 again. 2. Now apply the rule to each of these four and you’ll see that you always produce (1. I’ll arbitrarily say that we will include spaces and hyphens in our count. Just add up the digits and see whether that sum is a multiple of 3. ECKER (0.111 (six ones) is a multiple of 3 because the sum of the digits. and 407. then 5. Of these. Narcissistic Numbers: 153 It is well known that. just one has a black-hole property. So. 6. other natural languages may have a comparable property. 3) So.111. 3) (1. this gives 12. As another instance. it is 4 — or FOUR. In this case.44 M. other than the trivial examples of 0 and 1. is. 370. To create a black hole. Count the number of characters in the spelling. 3) (2. 1. 2. 111. 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 ). Since we are going to be doing some arithmetic. 3) (3. 1. Then. take the cube of each digit.111 (seven ones) is not. To avoid ambiguity. 3. One at a time. W. For instance. 0. the only natural numbers that equal the sum of the cubes of their digits are 153. work now with the 4 or FOUR. In turn. then 3. and finally 4. 371. 163 appears as ONE HUNDRED SIXTY-THREE for a total count of 23. such as FIVE for the usual 5. Write down your multiple of 3. you may wish to take out a hand calculator and/or some paper.

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

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

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

For large Ò this equation is approximately the same as Ä ½ · ½ Ä. In fact. 89. The first two numbers numbers: 1. Its divisors are 1.618 . but your very equilibrium. 21. but it is also that much more strange at the same time. 8. The largest permutation is 8820. This one is a bit more tedious. 5. and the difference is 8532. Its value is roughly 1. 55. 1. Had we used the ratios ´Òµ ´Ò · ½µ instead. That is. ECKER Example: Start with 8028. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ¾ ¾ ¾ ½ ¿ ½ ½ ½ (approx. Your own example may take more steps. 3. phi. ´½µ ´Ò   ½µ · ´Ò   ¾µ for integers Ò ¾. Fibonacci Numbers: Classic Results as Black Holes Many an endearing problem has charmed mathophiles with the Fibonacci . Divide both sides of the equation ´Òµ ½µ · ´Ò   ¾µ by ´Ò   ½µ.). including 1 and the number itself. etc. The Divisive Number 15 Take any natural number larger than 1 and write down its divisors.6 or so. and these digits sum to 15.48 M. The ratios seem to be converging to a number around 1. 5. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò   call it Ä — as Ò increases. but fewer authors use that approach. the smallest is 0288. 13. 34. and 15. but without using a program. The black-hole number this time is 15. 2. This one may defy not only your ability to explain it. Repeat with 8532 to calculate 8532–2358 = 6174. it is well known that the sequence converges to Ô ´½ · µ ¾ or the golden number. If we multiply both sides by Ä we obtain a quadratic equation: ´£µ ľ   Ä   ½ ¼ . More formally. To get a quick and dirty feel for why phi arises. but you will reach 6174. Now take the sum of the digits of these divisors. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ. Iterate until a number repeats. One of the classic results is that the ratio of consecutive terms has a lim¿ iting value. ½ . and ´Òµ preceding two elements. W. 3. we would have obtained the reciprocal of phi as the limit.

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

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

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

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

Calculate the expansion of this 26 term expression: ´Ü   µ´Ü   µ´Ü   µ´Ü   µ ´Ü   ݵ´Ü   Þ µ .5 nautical miles from Los Angeles to Honolulu. 4. 8. 7. 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. How long does it take? E5. 6. 5. subtract 1 and you produce the original number. 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. 9 in some perE mutation such that when digits are removed one at a time from the right the number remaining is divisible in turn by 8. E6. 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. g. My regular racquetball opponent has a license plate whose three-digit part has the following property. reverse the digits of the result. and the other two start out empty: (15. You have containers that hold 15 pints. HESS up and travels due North. 7. Find a nine-digit number made up of 1. 4. How many transfers of liquid will it take you to get a volume of liquid in one container that is within one percent of exactly one liter? E7. 63 9 38 33 32 12 88 25 16 18 15 Ü 5 E3. and 6 pints. Find the rule for combining numbers as shown below (e. 2. 10 pints. What is the number and what is the next greater number (possibly with more than three digits) having this property? . 3.54 R. 5. I. 0. 3. Through transferring liquid among the containers. 6. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu. E9. After going 10 miles in a straight line he finds himself back at camp.. It is approximately 2244. 25 and 9 combine to give 16) and use the rule to determine Ü. 2 and 1. 0). Where on earth could he be? E2. The 15pint container starts out full. Find the most efficient solution. Divide the number by 3. 8.

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

x. You are given some matches. What is the minimum time interval you can measure? Medium Problems M1. 744. From England comes the series 35. Despite your lack of power you took 6 tricks. You were playing bridge as South and held 432 in spades. Find a simple continuous function to generate the series and compute the surprise answer for Ü. 45. What is the probability that they form (a) an acute triangle. You are allowed to draw a closed path on the surface of each of the potatoes shown below. 280. one having an 11-pint container and another having a 2Npint container. Suppose a clock’s second hand is exactly on a second mark and exactly 18 second marks ahead of the hour hand. M4. a shoelace. . making a 4-club contract. Produce the other hands and a line of play that allows this to occur. 120. Three points have been chosen randomly from the vertices of a cube. 180. In clubs you held 5432. HESS E15. It has a symmetry property in that the burn rate a distance Ü from the left end is the same as the burn rate the same distance Ü from the right end. Potato Curves. 1260. 450. 60. and diamonds. You have a 37-pint container full of a refreshing drink. Shoelace Clock. What time is it? E17. (b) N = 5? M2. The lace burns irregularly like a fuse and takes 60 minutes to burn from end to end. 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 = 3.56 R. I. and a pair of scissors. Can you draw the two paths so that they are identical to each other in three-dimensional space? E16. hearts. N thirsty customers arrive. (b) a right triangle? M3.

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

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

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

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

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

A number may be approximated by a /b where a and b are integers. The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem. b) Find the minimum c and d so that g (c. 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. where he (1) picks out three socks at random. . b. 113. ) = 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. e ) 2.a . (a) The figure shows two Pythagorean triangles with a common side where three of the five side-lengths are prime numbers. their ages are ” H14. and proceeds to explain: “I normally ask guests to determine the ages of my three children. ) 0.0107 . x) = a bx .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.62 R. b. I. Find other such examples.” Your response is “No need to tell me more. The drawer starts with 16 socks each Monday morning (eight blue. d. (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. HESS H12. six black. (b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes? .0107. Since Smith missed the problem tonight and Jones missed it at the party two years ago. I’ll let you off the hook. 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). with g (355. Minimum Cutting Length. two brown) and ends up with four socks each Saturday evening. An approximation to is 355/113.71828 . Define the goodness of fit of a /b to x as g (a. where e = H16.0107 0.

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

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

0). Instead. (11. 6. (3. (3. 6. 5). 0. (0. 8. Thus and Ò ¾½¼ transfers is the minimum. .654. the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. E7. 2. 2). The license number is 741. 4. 10. The trick here is that 1 knot = 1 nautical mile per hour. At first glance.PUZZLES FROM AROUND THE WORLD 65 Solutions to Easy Problems E1. ؾ ¾¾ hours. 7. 0. 0). (1. The number is 381. E4.729. (9. (3. (7. so E8. 4. 6). E6. 2. 4. 10. 2). ¼. the next greater number is 7. 0). 0). In each case the numbers combine by summing the products of their digits. But this is not right because 18 is not the difference of 38 and 16. 6). Thus Ü ½½. (7. 6). 6). 0). 9). so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. 0. 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 .425. it appears that the rule might be subtraction. ¿   ½ ¼¼¼   are the smallest values satisfying the above equations. E9. E3. 2. and Ü ½ . 10. 6. (1. ½ ¼½ or ¼ and ½ ¼½. ¾ gives Ø E5. (11. 0. 2). 6. E2. 9. (3. (1. (5. 5. Thus x = 4 ¢ 6 + 2 ¢ 2 = 28. 8. 4. One of the terms is ´Ü   ܵ. 6). 6). E10. 0). (9. (5. . 15 moves: (15. Anywhere within 10 miles of the north pole.741.

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

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

For ¾. The series can be expressed by a simple continuous function: ´Òµ To get Ü. The next four tricks are won by South with the 5432 of clubs. East playing the 9.68 R. Then ¾Ô ¿Ò   ½ and Ô¾ ¿Ñ · ½. with East underruffing North each time. Ü¿   ¾Ü   ¼. ¿. North wins the sixth trick with the 10 of diamonds. Ô M7. Ë ¹ À ½¼ Â Ë À ¿¾ ¿¾ ¿¾ ¿¾ Ë Ãɽ¼ À ¹ ¹ ÃÉ Ë ¹ À Ãɽ¼ ÃÉ ½¼ M4. The only real root is Ü ¿ we get the prime 17. The hand below does the job if the play goes as follows. The final two tricks are won by South’s long diamonds. M6. North and East discard all their diamonds on these tricks. HESS M3. There are six ways as shown below. ¿½ ½¾¼ ÐÒ ¾ ¼ Ü M5. I. For multiple of 3 for Ô   ½. we take the limit of ½¾¼´¾Ò   ½µ Ò ÓÖ Ò ´Òµ as Ò goes to 0. so that ¾Ô · Ô¾ is always a Ô Ô ¿ we have Ô · ½ or . North next leads a heart which West trumps with the queen. The first five tricks are alternate spade and heart ruffs by North and West.

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

P3S. P2S. 2S. P.789. P4S. 3S. M15. 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. 5S. 6S. 2PS. M14. I. and 3P. 4S. 1. PS. 2P. Each number is the length of the shorter side of the rectangle. 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. Shown below. .70 R. 2P2S.123.465. A nine-rectangle solution is shown below.

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

12. HESS (b) The two points are along the diagonal near P and its opposite Ô point on the ½ ¢ ½ face that includes A. 14. 14. 10). 8). (15. 14. 12. The points are ´ ¿   ½µ ¾ of the way along each diagonal for a distance of 3. 11. 11. There are 7 Good Eggs. 9) and (14. There are 5 subjects with 15 marks possible in each. The number in each triangle indicates its area relative to the other triangles in the figure. 11. (15. 13. 10. 11. Humpty Dumpty got a 10 in arithmetic. (15. 9). 6). 13. . M20. 12. 12. 13. 13. 12. (15. The scores for the Good Eggs are (15. (15. 7). 14. Shown below. 10.0119 . I. 8). 12. Solutions to Hard Problems H1.72 R. 14.

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

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

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

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

È ´¼   ¼µ È ´ ¼   ¿¼µ when Ô¿   Ô¾   ¾Ô ½. ·¾ Ô ´½ ½ Õ µ È ´¼   ¼µ .8286 0.7849 . H11. At Ò games to Ò. È ´ ¼   ¿¼µ Ô · ÕÈ ´ ¼ ¼µ.7892 0.2069 1.3586 Pairing probability 0. È ´ ¼   ¼µ ÔÈ ´ ¼   ¿¼µ · ÕÈ ´¿¼   ¼µ.2511 1. Clearly ST is not perpendicular to the plane.PUZZLES FROM AROUND THE WORLD 77 plane so that all altitudes are unit. and let Õ ½   Ô.8163 0.2270 1. Day Monday Tuesday Wednesday Thursday Friday Saturday Average Selections Required 1. Let È ´   µ be the probability of the server winning the game when the server has points and the receiver has points. A computer program was written to calculate the probabilities shown in the table below. È ´¿¼   ¼µ ÔÈ ´ ¼   ¼µ. ¾ ¾ Õ´ È ´¿¼   ½ µ Ô¾´½ · Õµ · Ô Ô¾ ÔÕÕ·½µ . This leads to ¾ È ´ ¼   ¼µ Ô¾Ô Õ¾ . which leads to Ô ¼ ½ ¿¿ ¼ .2801 1. which leads to Ô ¼ ½½ ¿ ¾ . Let Ô the probability of the server winning a point.3074 1. Ô  Õ At a score of Ò games to Ò (Ò ¼ to 5). ¾ ¾ Similarly one derives Ô · Ô¾Ô·ÕÕ¾ . My uncle is expected to take the longest time to dress on Saturday and on Friday he is least likely to get a pair from the first three socks chosen. · È ´ ¼   ¿¼µ Õ È ´ ¼   ½ µ Ô · ÔÕ · ÔÔ ·Õ¾ . È ´¼   ¼µ È ´¿¼   ½ µ when Ô ¾   Ô ¿.7805 0 . ¾ H12.8027 0.

444 years. ¿ ½¾¼. We look for cases where (1) the oldest child is under 20. (2) the younger two children have different ages. . 15). ¼ ¼¼ ½ ¼. ¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½ .613. H17.321 is the first of 5 in a row. HESS H13. ´ ¾   ½µ ¾. 12). which must have been guessed by Jones two years ago. (b) Length ¾ ¼¾½½¾ ¿ as shown.546. There is only one set of ages that accomplishies this: (5. ¼ ¼½¼ . (a) Length ½ ¿ ¾½ ½ ¼ as shown. (d) Yes. Ô½ · ½ ¾£Ô£Ô ·¾ £ Ô¾ · ¿ £ Ô¿ · ¾ £Ô . we get (a) (b) H16. (a) 45 years. which must have been guessed by Smith. 16). Using a multiprecision continued fraction program.78 R. (b) 605 years. (c) Unknown.093. 7.042. a=10. Solutions occur for (b) ¾ ½. 8. The product and sum could be achieved by (4. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago. · ½. . The product and sum two years ago could be achieved by (2. H14. I. ´ ¾ · ½µ ¾ prime. ¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼. prime. (c) 17. ¾   ½µ ¾. . 6. ¿ ´ ¾ · ½µ ¾ ¾½¿ ¾½. ¾¾ ¾ ¿ ¼ ¿ ¿ ½½ ½ ¾¼. · ½. (a) ½ ¾ ¢ ½¼¾ ¼¾ ¢ ½¼½ . ´ prime.641.512. H15.

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

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

(5) On Þ Ê for sufficiently large Ê.8221528 Ý 7.5432385 32. where Þ at ½ · Þ Ü · Ý.8505482 39.6745053 3.4473849 of ´Òµ from pole is the limit as Þ approaches Þ ´Þ   Þ µÕ´Þ µ ´½ · Þ µÒ·½.6406814 3. The resulting equations in Ü and Ý are ´½ Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ. × ÒÝ Ý we can look for poles in the upper half-plane only and reflect each one into the lower half-plane. The table below gives numerical values of the first few. Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ.8790560 20.  ¾ × ´Ý · µ  Ü Ý .4614892 13. For larger define ´¾ ·½ ¾µ . ¼ for all .2916783 3.0262969 3. Combining the poles in the upper and lower half-planes and some rearrangement finally produces ´Òµ where ½ ½  ÒÜ × Ò´ÒÝ   µ (a) Since Ü Ü¾ · Ý ¾ ܾ · Ü · ݾ Ø Ò´Ý · µ Ý  ÒÜ goes to zero as Ò goes to infinity.5012690 3.0888430 2. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to infinity. Then Ý approximately equals   ½·ÐÒ´ µ . Ü 1 2 3 4 5 6 7 (6) The contribution to 2.PUZZLES FROM AROUND THE WORLD 81 where is the counterclockwise path about the Ø pole of Õ´Þ µ.2238350 26. A pole of Õ´Þ µ occurs Þ . not including its pole at Þ ¼. which reduces to ´½ · Ü · Ý µ´Ü   Ý µ ´ Ó×´Ò · ½µÝ   × Ò´Ò · ½µÝ µ where ´Ü¾ · ݾ µ ´Ò·½µÜ . Since the equation in Ý is even. Call the Ø such pole in the upper half-plane Þ Ü · Ý .1512074 45.

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

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

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

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

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

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

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

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

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

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

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

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

I wrote to Kate Jones.489 B 15.717 C 6675 D 7549 E 11. K. Ô Ü (3-symmetry).94 R. On the contrary.460 584 985 885 Ð Ñ Ó Ô Õ Ö 31 × 1537 637 914 749 498 1238 and. Their numbers. by the Hexagon’s position Ø Ù Total 264 1094 124. are 130 195 533 193 377 2214 111. 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.518 A 75. classified according to the position of the Chevron. GUY Figure 18. president of Kadon Enterprises and discovered that she markets a set of “iamonds” .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. for those who prefer their puzzles to have just one answer. Õ ¼ Ø Paulhus wrote a program that used only a few days of computer time to find all the solutions. Six pieces meet at a point. Figure 19.

London. are given in the following table. where the last line counts reflections as different: Order 1 2 3 4 5 6 7 # of iamonds 1 1 1 3 4 12 24 with reflections 1 1 1 4 6 19 44 The Iamond Ring is a beautiful arrangement.O’BEIRNE’S HEXIAMOND 95 under the trademark “Iamond Ring.” For this purpose the iamonds are not counted as different if they are reflections of one another. 787–788. diamonds. discovered by Kate herself. J. Academic Press. doesn’t do justice Figure 20. and R. Guy. Conway. if you want to use them to make O’Beirne’s hexiamond puzzle. Berlekamp. so. you must get two sets. . References [1] E. Kate Jones’ Iamond Ring. triamonds. R. The numbers of iamonds. Figure 20. . 1982. K. in which each x-iamond is labeled Ü ½ Ü to the very appealing colored pieces of the professionally produced puzzle. H. Winning Ways for Your Mathematical Plays. pp.

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

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

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

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

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

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

I hope I have stimulated your imagination about what can be done by transforming puzzles so you may start your own work. and have fun generating individual derivatives for yourself and your friends! . generated are even more Tangram-like than the Trigo-Tangram. WIEZORKE Figure 5. Transforming the Tangram by diagonal extensions leads to two derivatives which are more Tangram-like than the one shown in Figure 1.102 B. To create problems for these derivatives is left to the reader. Take your favorite puzzle. One set of problems will serve for both of them.

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

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

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

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

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

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



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.

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

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

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



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.

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

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

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

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

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

When it comes to puzzles. and by that time I was rather tired of the whole thing. perhaps disguised in another form. it is often the simplest things that prove to have the greatest appeal. as topological puzzles so often do? . 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. probably not even realized at the start. Or will it mischievously rise again.

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

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

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

There are two outer and two inner loops. . not promptly corrected. with the wire ends making small rings that are wrapped around the wire in such a manner as to eliminate any usable ends. and the object is to remove the cord completely from the puzzle. The difficulty may be increased by. Any wrong moves. 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. after the cord is removed. On a scale of 1 to 10.134 R. The Nightmare puzzle is made from one continuous strand of wire. I rate the Nightmare an 8. adding a ring to the cord that will not pass through either of the small end rings then attempting to replace the cord. In addition to the difficulty in conceptualizing the convoluted shape of the puzzle. A cord encircles the two inner loops of the puzzle. IRBY has more than lived up to its name with a convoluted three-dimensional shape that exacts extreme effort and concentration from all who attempt it.

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

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

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

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

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

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

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

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

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

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

B. Notice that the eight pieces must form the eight corners of the completed cube — one piece for each corner. proceed to the next hint. Solution Hints. C. B.½ distinct ways of constructing the cube. A. If. FARRELL Figure 1. 1. but there still remain a great many cases that are nearly right. and C will turn out to be a solution set. then there are ´ ¿µ´ µ´¿ µ¾ . 2.½½ . The combinatorial puzzle uses only six colors in its construction. A. That is. There are 93 solutions in this case. try again to solve the puzzle. 2. and the solver has the additional clue that a face is all of one color. Trial and error is not a very fruitful way of trying to solve this puzzle. after reading a hint. ½º Call the six colors A. 1. 0). If you cannot. different” but one interpretation could be to orient the cube to sit in the positive octant in space with Biggie always occupying the corner (0. and 3 are any six distinctive colors. each of the 48 faces of the pieces is colored with a unique color. These hints are to be regarded as progressive. 0. each . additionally. Therefore. and 3.146 J. B.½ ¼. C.

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

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

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

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

The final part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. and only 10.105 of these combinations meet the three-cubes-of-each-color requirement.691 combinations into 148 disjoint sets is described in Section 6. none of the nine planes may contain two cubes of the same color.152 S. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. as determined in Section 4. it becomes apparent that most of the 133. and the most difficult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1. A proof is given in Section 3.251. when two cubes of the same color are in position. then. but not in the center of a face. only thirty dicubes are required. Twelve of these can be selected in 1. It becomes necessary. to separate these cases into disjoint sets of isomorphisms. FARHI a difficult exercise for a moderately proficient programmer to verify the results obtained. This was achieved through the use of a computer program. The individual dicubes can be colored thirty-six different ways. some with the tricube on the edge and some with the tricube through the center of the cube.677. . The method of sorting the 10. only 133. as is shown in Section 5. this rule then tells the experimenter to backtrack. The second observation is that for a cube to be assembled with the nine colors on each face.691 combinations need to be sorted into disjoint sets. obtained after some experimentation.105 combinations are isomorphisms. the location of the third cube is determined. The first observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube. Since one person can assign the fourth and fifth colors to orange and red and another person to red and orange. Since during the course of trying to solve this puzzle a conflict often occurs. in which the number of isomorphisms using the set of thirty-six dicubes is also determined. the results are shown in Table 1. The previous question now becomes more interesting.700 (36!/12!24!) different ways of which. is that there are numerous solutions. The third observation. The question then arises: How many solutions are there? Most intriguing about this puzzle. A proof is given in Section 2. Fortunately. Thus.

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

FARHI No.154 S. 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 .

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.

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

then color 3 is combined with two colors from the third row. The two dicubes with color 1 may use any two colors in the first row of the following table.” The remaining two colors must all be on edges. the edge case fills spaces 2. The only way the remaining two cubes of that color can be displayed on all six faces is for them to be located on opposite corners as shown in Figure 3 by the letter “A. 11. and 3. and on a centerface such as shown by the letter “B. If color 3 has been selected once. then a third dicube with color 2 and any color from the second row is selected. . The “No Two in the Same Plane” Rule Figure 3. cubes of the same color are never in the same plane. 3. 5. If color 2 is selected from the first row. there are three possible locations. Determination of the 133. and 14. 2.” In each case. such as shown by the letter “C. One cube is centrally located and thus not visible. 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. accounting for two faces. 4. 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. the center case fills spaces 5.” The remaining six corners must have cubes of different colors. Since the corner colors are displayed on three faces. and 11. the remaining two cubes must be located on an edge.158 S. The corner case fills spaces 1. If none of the previous selections include color 3. and 11.105 Possible Dicube Color Combinations The tricube colors are defined as colors 1. then none are selected from the third row.

6. (1. 9) and (8. 5). 5. (7. 8. Reduction of the Number of Isomorphisms The table in Section 4 is simplified by restricting the selection in the first two rows as shown below. using the same algorithm as before. 6. Identification of the second color of these ordered pairs is sufficient to define all the colors. 8. 7). Identification and Separation into Disjoint Sets The color combinations in Figure 1 use dicubes with color (1. 9. 3). (2. 8). etc.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. For example. (4. given the second numbers 2. 9). and 9 one can logically deduce the first colors by the three . color 5 is combined with colors in the fifth row such that there are three cubes with color 5. the two new colors are identified as colors 6 and 7. (5. Using this algorithm. 4).105 possible color combinations. 6. Similarly if color 2 is not associated with colors 3.691 color combinations. 6). only 10. (6. 9). or 5. 3. (4. (7. 9. 2). (5. 5. 4. 6). a computer program established 133. 4. (3. This results in requiring only thirty cubes and. 7. 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).

has sixty isomorphisms. four at a time. cannot be interchanged with other colors. 98. FARHI cubes of each color requirement and the restriction that the second number be greater than the first. (7.:Zppp. by adding twenty-three to each and using the ASCII character set codes (ASCII stands for American Standard Code for Information Interchange). 9).:Z[yz. Color 2. ..691 color combinations obtained by the algorithm described in Section 5 were listed in order according to their ASCII characters. and 99. . (6. including pairs of two at a time. (24). (4. (68). This process then identified the 148 disjoint sets of isomorphisms. Different solutions often have the same three-digit identification. 58. can be swapped with each other. and then assigning a printable character to each. The 10. 6). 2). Thus this color combination is identified as /:Z[yz. its isomorphisms are determined and removed from the list. 35. If the solutions are to be catalogued. The solutions are identified by a three-digit number indicating the color of the buried cube and the two “edge” colors. which is in the center of the tricube. then additional criteria for recording solutions are recommended. (35). (7. 7). Not all of these mappings produce a new isomorphism. It is then a simple process to recognize that this represents the dicubes colored (1. Decoding a name is quite simple. 3). (5. The ASCII code for 23 is not a printable character. on the ends of the tricube. but not with any of the other colors. The ASCII codes for these characters are: 46. The process was then continued until the list was exhausted. (4. pairs of three at a time and four at a time with two at a time. five at a time including three at a time with two at a time. and 122. These were removed from the list. 5). 8).e. 6). 90.160 S. and (99). i. Any permissible interchange of colors is an isomorphism. The head of the list then became . For example. 9). Shown in Figure 4 are an edge solution and a center solution for the color combination of Figure 1.691 cases into 148 disjoint sets of isomorphisms. (67). 9). The use of this six-character identification greatly simplified the process of sorting the 10. 3). An inspection of the ASCII table will explain why it was decided to add 23. The remaining six colors may be interchanged in numerous ways: two at a time. 67. 8). (5. Subtracting twenty-three results in: 23. (2. and six at a time including a triplet of two at a time. Identification was further simplified by placing these numbers in pairs (see Note below). and (8. 68. (98). Colors 1 and 3. 91. 121. three at a time. (4. The first entry. Note: The smallest possible number is 23 and the largest is 99. the first entry in Table 1 is . (1.:Z[yz.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

“Lapin. and weaving 2-D patterns Patterns with 2-D parts on 3-D surface “Shmuzzles” tesselations JIG-PART JIG-SLOP JIG-LAYR JIG-2D3D JIG-BLOX PATPAT-2DPG Bilhourd’s jigsaws “Sculpture Puzzles” “Toyznet” Queens on chess board. Waddington’s “Black Box” Match puzzles. number puzzles Heads and tails “Testa” PAT-STIX PAT-NUM PAT-2DEG PAT-2D PAT-STAK PAT-2D3D Stacking transparent layers. Apple and Worms. “Managon. nine-piece ivory cube Pack the Plums. HORDERN Ball pyramids. DALGETY AND E.” “Even Steven.” 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. BlackBox. Josephus. overlapping. pegs. Squashed Soma.” Loyd’s Donkeys. or pieces. Jensen’s “Tricky Laberint” Magic squares.” “Phoney Baloney.182 ASS-POLY ASS-SHAP 3-D Assembly Polyhedra and Spheres 3D Assembly Other Shapes J. 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. weaving puzzles “Dodeca” .

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

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

weight illusions AMB-VANI AMB-DIST AMB-ARCH Distortions Archimboldesque objects AMB-HIDD Hidden image pictures (no manipulation required) Hidden image pictures (manipulation required) AMB-HMAN AMB-TURN Pictures that require turning to show different images Perception illusions Ephemera related to puzzles Wiggle Woggle HTL Micro Printing Moire effect Hold to light Heiroglyphs (non-rebus) Anaglyphs AMB-ILLU EPHEPH-WIWO EPH-MICR EPH-MOIR EPH-HTL EPH-HEIR EPH-ANAG EPH-STRP Strip prints (three views in one frame) Hold to light cards producing movements by shadows Images concealed by extreme smallness Puzzles/effects produced by moire/fringe effects Protean views. “Spot the Difference. 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).” random dot stereograms “Naughty Butterflies.CLASSIFICATION OF PUZZLES OTH-PEND AMBAMB-POBJ Pending Classification !!! Ambiguous Paradoxical objects (objects that apparently cannot be made) Vanishing images 185 Puzzles awaiting classification Arrow through bottle. soot on unglazed part of ashtray Landscape turned to make a portrait. Oskar’s “Escher Puzzle. impossible dovetails. OHOs. hold to light advertisements Obscure non-rebus heiroglyphic prints Requiring red and green glasses for either 3-D or movement effects Framed strip prints showing different views from different directions .” “Find the 5th Pig” needing coloured overlays.” Penrose triangle “Vanishing Leprechauns.

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

We are not told whether it is Ü or Ý that is the greater of the two.e. if y is greater than x. is 100. if Ý is greater than Ü. and the excess of Ý over Ü. Proposition ½. say. one of which is twice as great as the other. the excess of x over y. Then the excess of Ü over Ý. Ä Ø Ø «ÖÒ ØÛ Ò Ü Ò Ý ÓÖ Û Ø × Ø × Ñ Ø Ò ¸ Ø Ð ×× Ö Ó Ø ØÛÓº Ì Ò Ó Ú ÓÙ×ÐÝ Ø Ü ×× Ó Ü ÓÚ Ö Ý¸ Ü × Ö Ø Ö Ø Ò Ý¸ × ¸ Ò Ø Ü ×× Ó Ý ÓÚ Ö Ü¸ Ý × Ö Ø Ö Ø Ò Ü¸ × Ò º ËÒ ¸ ÈÖÓÔÓ× Ø ÓÒ ¾ × ×Ø Ð × Proof of Proposition 1. The excess of x over y. is certainly 100. 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 equal to the excess of y over x. I will now prove the following two obviously incompatible propositions. ËÙÔÔÓ× Ü × Ö Ø Ö Ø Ò Ýº Ì Ò Ü ¾Ý¸ Ò Ø Ü ×× Ó Ü ÓÚ Ö Ý × Ø Ò Ýº Ì Ù× Ø Ü ×× Ó Ü ÓÚ Ö Ý¸ Ü × Ö Ø Ö Ø Ò Ý¸ × Ýº ÆÓÛ¸ ×ÙÔÔÓ× Ý × Ö Ø Ö Ø Ò Üº Ì Ò Ü ½ ݸ Ò Ø Ü ×× ¾ Ó Ý ÓÚ Ö Ü × Ø Ò Ý   ½ Ý ½ ݺ Ì Ù× Ø Ü ×× Ó Ý ÓÚ Ö Ü¸ Ý × Ö Ø Ö ¾ ¾ ½ ½ Ø Ò Ü¸ × ¾ ݺ Ë Ò Ý × Ö Ø Ö Ø Ò ¾ ݸ Ø × ÔÖÓÚ × Ø Ø Ø Ü ×× Ó Ü ÓÚ Ö Ý¸ Ü × Ö Ø Ö Ø Ò Ý¸ × Ö Ø Ö Ø Ò Ø Ü ×× Ó Ý ÓÚ Ö Ü¸ Ý × Ö Ø Ö Ø Ò Üº Ì Ù× ÈÖÓÔÓ× Ø ÓÒ ½ × ×Ø Ð × º Proof of Proposition 2. if Ü is greater than Ý. But look.A Curious Paradox Raymond Smullyan Consider two positive integers Ü and Ý. Proposition ¾. The two amounts are really the same (i. And isn’t 100 surely greater than 50? 189 . is greater than the excess of y over x. Now. if x is greater than y. if x is greater than y.. if y is greater than x). suppose Ý. is certainly 50 (since Ü is then 50).

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

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

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

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

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

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

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

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

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

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

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

204 B. GOSPER Times 100–105: .

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

speed c/3 orthogonal spaceships. the 73 gen turn can emit a glider.) Construction of Oscillators. The · ¾ Ò ¾ · ¾ Ò ¼ · ¼Ò ¾¿¼ · ¼Ò ¾ ¾ · ¿ Ò and · ¿ Ò use a device discovered by Wainwright. The turn can take either 64.) 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).. or 73 generations. or 47. We also have an argument. 6. I built a p155 oscillator in which 20 Bs travel around a track of length ¿½¼¼´ ¾¼ ¢ ·¾ ¢ µ generations. one period 4.206 B.e. The first glider to hit them is reflected . Some of the c/3s and c/4s have sparks at the back which can do various things to c/2 spaceships that catch up with them. and 47 are the only ones we know with the right sparks. More recently. and c/4.) Lately.) Periods ½¿ · Ò and ½ ¼ · Ò use a mechanism developed by Buckingham. c/3. infinitely many period 2. speed c/2 orthogonal spaceships. (Mine runs only on an Apple II. For the multiples of 8. Hartmut found another with speed 2c/5. he and Hartmut Holzwart have been producing oodles of orthogonal spaceships with speeds c/2. based on construction universality. 6. which implies that all sufficiently large periods are possible. GOSPER resembling your initials. one period 4. we put several copies of the B in the track to reduce the period to one of those mentioned. which reflects a symmetric pair of parallel gliders. and one period 5. A period 150 gun fires toward a distant pair of pentadecathlons. in which a B heptomino is forced to turn a 90 Æ corner. but 5. (No doubt you figured that out for yourself long ago. (Any greater nonmultiple of 4 would also work. Also. of periods 5. JHC!). a few oscillators with periods 5 (including some with useful sparks) and 6. speed 2c/5 orthogonal spaceship. by combining them we can build a closed loop whose length is any large multiple of 5 or 8. speed c/4 orthogonal spaceship. Due mostly to the work of Buckingham and Wainwright. 65. Another amusing oscillator uses the glider crystallization that Bill mentioned in his centinal letter a few years ago. David Bell wrote a similar program that runs on faster machines. it consists of a still life (“eater3”) and two spark producing oscillators. For example. so we also get glider guns of those periods. infinitely many period 3. not just lcms of smaller period oscillators acting almost independently) of oscillators of periods ½ ½ ½ ¾ ¾ ¾ ¿¼ ¿¾ ¿ ¼ ¾ ¼ ¾ ½¼¼ ½¼ ½¾ · ½¾¼Ò ½¿ · ½¾¼Ò · ¾ Ò ¾ · ¾ Ò ¼ · ¾ Ò ¾¿¼ · ¼Ò ¾ ¾ · ¿ Ò ·¿ Ò ½¿ · Ò ½ ¼· Ò and all multiples of ¿¼ and ½¼¼. speed c/4 diagonal spaceship (not the glider). (These are used in some of the large patterns with unusual growth rates mentioned later. we now have nontrivial examples (i.

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

more generally. and 360. this reaction leads to a period 30 LWSS gun: Large Constructions. Universality implies the existence of Life patterns with various unusual properties.) New Year’s Eve Newsflash: Buckingham has just announced “. . . (At least they seem unusual. 300. I have a working construction to build a P14 glider stream by inserting a glider between two gliders 28 gens apart! . Here it’s shown deleting a glider. its bounding box is 373 by 372. We can also synthesize light-. . if the glider is delayed by 60 (or. it escapes: There are also combinations of two period 30 guns that give fairly small guns with periods 90. 210. and 17.208 B. 16. 10. and Buckingham has found clever ways to get 15.” (Less than 14 is physically impossible. 120. This will make it possible to produce glider streams of any period. The smallest known uses a period 8 oscillator (“blocker”) found by Wainwright to delete half the gliders from a period 60 stream. based on the .) Returning to Hickerson’s account: Many of the large constructions mentioned later require a period 120 gun. For example. . GOSPER to 18 this way (period 23 is especially simple). and heavyweight spaceships from gliders. (I built a pseudo-period 15 gun based on his reactions. middle-. 150. so we have guns that produce these as well. Ò · ) generations.

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

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

That raises the question. never imagined they could be made by colliding gliders. after a trial run of the foregoing around the Life net. and Life was an excellent idea for him to have the opportunity to present. you can waste computer time all night long and nobody says anything. ‘Have other hours of computer time been spent more profitably?’ (Nowadays. “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. if all those computer hours hadn’t been around to waste. on just that level of computer. perhaps the ideas wouldn’t have prospered so well. ººÓÓºÓÓººººº ººÓºººÓººººº ºººÓÓÓºººººº ºººººººººººº ºººººÓÓÓºººº ººººÓºººÓºÓÓ ºÓººÓºººÓºÓÓ ÓºÓºÓºÓºÓººº ºÓÓºÓºººÓººº ºººººÓÓÓºººº ºººººººººººº ºººººººÓÓÓºº ººººººÓºººÓº ººººººÓÓºÓÓº Finally. nobody has yet . and along with the rest of us.) “Martin Gardner is a skilled presenter of ideas.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. Not to mention that the time was ideal. 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.

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

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

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

214 M. a graph may have either cycles or no cycles. no cycles. no cycles. There are two interesting topological properties that apply to silhouettes. First. to projective casts. The basic problem is the analysis . (d) one part. V AN DEVENTER Figure 2. Even when silhouettes are viable. Dewdney used the term viable for a silhouette or a maze that consists of one part and has no cycles. (a) Negative of a silhouette. cast of a (limited) number of silhouettes: One cannot cast a hollow cube or a solid sphere. Figure 3. and to mazes and graphs in general. There is no simple connection between the properties of the silhouettes and the properties of their projective cast. (c) multiple parts. Topological properties of mazes and graphs: (a) multiple parts with cycles. a graph may consist of either one part or multiple parts (Figure 3). (b) one part with cycles. their projective cast can still have cycles (Figure 4) or multiple parts (Figure 5). (b) The cylindrical region in space defined by it. Second. O.

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

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

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

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

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

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

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

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

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

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

1. 11. 2. noting that this system is inconsistent Fibonacci. 6) (1. 10. 4. 3) (10. although the material is believed to be based on older Arabic sources. 5) c. which is rather a mouthful. 11. 4. hence the above title. 3. 5) (100. And he took to the eldest daughter fifty apples. 1010 1150 1202? 1202? 1202? 1202? 1202? 1202? 1202? Euclid Diophantus The Greek Anthology The Greek Anthology Alcuin Mahavira Mahavira al-Karkhi Bhaskara Fibonacci Fibonacci Fibonacci. 2. 850 c. 250 c. 4) (2. 1. 9. 6) I-(7. 5.5) I-(25. Now how many sold each of them for a penny? . I quote a version that appears as No. 6) I-(7. 4. 22. (1. 23. 8. 10) (7. 2. 2. 8. 3. 2. and to the youngest ten apples. 5. 3. 5. 1. 2. 510 9Ø century c. so I will first describe the later form which is first found in Fibonacci [11]. 10. 11. 3. 9. 11. 2.” printed by Wynken de Worde in 1511 [9]. and Bhaskara [4]. 5. but this form is less clearly expressed and has infinitely many solutions. 4. certain apples. 4. and all these three sold in like many for a penny. 5) (2. 0. 7) II-(1. there seems to be no common name for it. The problem is treated in one form by Mahavira [15]. 5. 3.SOME DIOPHANTINE RECREATIONS 225 from the early fourteenth century and hence after Fibonacci. 5. 6. 510 c. with the problem statement giving the 6 as a 7 Fibonacci. 10. 3. 7) I-(7. 50 in the first printed English riddle collection “The Demaundes Joyous. 300 B. 11. 4. Sridhara [18]. with the problem statement giving the 6 as a 7 Fibonacci Fibonacci. 7) I-(7. 3. 9. 1. 5. which daughters he delivered. 850 c. 50. to sell.C. 6) I-(7. done two ways Selling Different Amounts at the Same Price Yielding the Same Although this problem is quite old. 2. 9. 3. 10. 2. 9. and brought home in like such money. 11. and to the second thirty apples. 1) (30. c. 2) I-(9. A man had three daughters of three ages.

g. and it is not immediately obvious how to find or count all the solutions. During the 1980s. such as the first receives twice the second. or sometimes fruit. 30. In 1874. Fibonacci only gives two-person versions: (10. (12. which is essentially the same as the first edition of 1694. the 1612 and 1624 editions of Bachet [3] ¾ give a fairly general rule for this case and apply it to (20. if the prices are also given. This may happen by the sellers going to different markets. (12. which normally are all of the same value. 33). 40). but we both conjecture that he was counting the positive solutions. Fibonacci is one of the few to give more than a single solution. of which 36 are positive. 30. considers (10. which would produce the positive solutions with the smaller price equal to 1. 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. or lowering prices late in the day in an attempt to sell their remainders. so that the above is (10. I tried several times to find an easy way to generate the solutions. Even with this trick exposed. Fibonacci goes on to consider solutions with certain properties — e. The question of finding all solutions for a three-person problem seems to have first been tackled in the 1600s.. Labosne revised the material. .’ Anyone meeting this problem for the first time soon realises there must be some trick to it. 30). contrary to De Lagny’s statement that there are six solutions. According to Glaisher [12]. He gives a fairly general rule for generating solutions [12]. 30) and gives two solutions. SINGMASTER ‘In like many’ means ‘the same number’ and ‘in like much money’ means ‘the same sum of money. he added several more versions as well. 50). Tartaglia [19] justifies the different prices by having large and small pearls. or simply changing prices after lunch. is there a solution. of which there are indeed six. and he gives five solutions for his first version — but there are 55. The extensively revised and enlarged edition of 1725 (or 1723) outlines a fairly general method and correctly says there are ten solutions. Although the objects are usually eggs. or raising them late in the day when few items are left and new buyers arrive. I have now arrived at a reasonably simple approach that turns ¾I have not seen these editions. Neither Glaisher nor I have seen the relevant work of De Lagny. 25.226 D. dropping Bachet’s rule and replacing it with some rather vague algebra. Let us denote the problem with numbers ½ ¾ by ´ ½ ¾ µ. The 1708 English edition of Ozanam [17]. 32). it can take a fair amount of time to work out a solution. find all solutions with one price fixed. find solutions where the amounts received are in some ratio. for which he often gives fractional answers. This is historically by far the most common version. distinctly against the spirit of the problem.

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

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

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

in our example. i. i. we have ¾Ã   ½ solutions with zero values. In that case. unless it happens that ½ «¾­Ò ­¾ .. Hence we normally get ¾Ã situations with a zero value.. and when Ò ¼. We then have à ½ Æ ½´ ½ · ½µ ¾ Æ· ´ ½   ¾µ´ ½   ½µ ¾. Tartaglia’s problem 139 is the one with two sizes of pearls and is (10. ½ ½ . i. ´½ µ Again. . when the upper bound for given in (13) is an exact integer. If ½ is odd. From (15). 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 . These occur when ½ ¼. Looking again at our example. we have à ½ Æ ¿ Æ· ½. so we let Ã Æ  ½ ­  Ò à where [ ] is the greatest integer function. A few cases occur that are not APs (see below). 90). ­Ò   Ò   ½: ´½ µ à A number of early cases have Ò ¾. The majority of examples are APs with Ò ¿.e. so we can write ½ ·´   ½µ . Æ· ½¼   ¾ ¢ ¾ .e. then ½ we get à ´ ½   ½µ ¾ Æ ´ ¾   ½µ Æ· ´ ½   ¾µ¾   ½ . SINGMASTER Then (10) becomes ´½¾µ ´½¿µ ½  ½ ­Ò The upper bound may not be an integer. perhaps as intended by De Lagny. We can count the number Æ · of positive solutions by finding the number of solutions with a zero value.. 20. Substituting (11) into (8) and summing gives us the number of solutions. which is trivially an AP. ½ «¾­Ò ­¾ by (7). ½ A few longer APs occur — I have seen examples with Ò . A little thought shows that then   is the common difference of the progression. we have à ­ ´ ½   Ò · ½µ   ½ Æ Ã ´ ½ · ½µ   ­Ò à ´Ã · ½µ ¾  ½¼ ¾ and ¾ ¢ ½½   ¾¼ ¢ ¾ ¢ ¿ ¾ ¢ Most of the earlier writers excluded zero values and only considered positive solutions. we find that this occurs when   ½ ­Ò . Most of the examples in the literature have the in arithmetic progression (AP). All but one of these have ½ even. Then we have ­Ò ´Ò   ½µ .e.230 D. Rearranging. when two zero values occur in the same solution. Subtracting ¾Ã or ¾Ã   ½ from Æ yields Æ· . and then à ½ ¾ Æ ¾ Æ· ´ ½   ¾µ¾ .

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

55) (78. 1 given. 40. ¬ . 1 given. 55) (16. 136) (3. all given.4) (70. 4 given. 30) (12. 1 given. (20. 21) (225. At this point. 81) (10. none given. 82) (17. 90) Date 1202? 1202? 1202? c. 32. 36) (27. 36) (30. 30) (8. 37) (27. 11. 33) (10. 26) (20. 33) (20.6) # of Solutions (55. 33. 1 given. 60) (10. 60) . 30. 20. . 1 given. 32) (12. 119. 1 given.400) (25. . 1 given. 170) (305. Sridhara finds a one-parameter solution set. 50) (11. though it is not as easy to see how to do this as in the Western version. 40) (10.9) (100. 16) (16. 1300 1300s 1489 1489 1489 c. in some way to reduce the solution set to two dimensions. 1 given. makes this more specific by saying they ‘sold a part in lots and disposed of the remainder at one for five panas. 100) (45. 60) (117. 25. 50) (10.232 Version (10. 12. One can scale the set . 1 given. 29. 17. 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. 56. 196) (45. 30. Bhaskara makes the following comment: “This. 6 given. 20) (64. 2 given. 30) (18.’ Thus (2) becomes ´¾¼µ · We now have ¾Ò · ¿ unknowns and ¾Ò equations. 1 given. 15) (31. 40. 454. 20. (20. 50) (30. one with fractions. 55) (55. 901) (10. 20) (10. 603. 40. 80) (10. I must point out that some Indian versions give and/or as fractions! The Indian authors do not get very general solutions of these problems. 40) (18. 36) (78. 1 given. 1 given. 1) (7. 1 given. SINGMASTER Solutions in Source 5 given. 1 given. none given. 35) (11. 140) . 30. 16) D. 1) (100. 68. 49) (3. so we again have a three-dimensional solution set. 1 given with fractions. . 81) (55. 752. 36) (25. 1 given.552. 48.

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

13. 2nd ed. [6] Bunt.so   and ¬ ½½¼. Indian Math. Labosne. Krishnaswami. PrenticeHall. Nos. Translated into German by K. Rigaud. Prob. References [1] Alcuin.. 16: “De duobus hominibus boves ducentibus — Two men leading oxen. 3292. 33. 106–115. Wiesbaden. 24. Soc. designed for the Use of Schools.C. pp. Bîjaganita. 5th ed. Neun Bucher arithmetischer Technik. 5th ed. 1923–24. [9] The Demaundes Joyous. giving total numbers of 3295.. aka Bhâskara. A. John. 1964. 54. pp. Braunschweig. John Murray.” In 1815. Math. 15. Algebra. 12. Lyon. (There have been several reprints. Paris. H. Claude-Gaspar. “Nine Chapters on the Mathematical Art. See No. 1624. 4392. 15. 18. London. reprinted by Dover. Johnson. Bhaskara solves his example by taking ¾.. 3200 and 16. with Arithmetic and Mensuration from the Sanscrit of Brahmegupta and Bhascara. Revised by A. I examined the 7 Ø edition. [3] Bachet. 242–244. Prob. The Historical Roots of Elementary Mathematics. annotated by David Singmaster and John Hadley. Heath..” J.” [2] Ayyangar.. Rigaud.234 D. 19. London. [4] Bhaskara. “A classical Indian puzzle-problem. Lucas N. p. pp. 102– 126. and Other Places of Public Education. 1973. 1150. P. 1817. 122–126.470 [4]. he produced a 10 Ø ed. 1824. Mar 1992. 1st ed. 1976.. [8] Diophantos. and the 13Ø edition. pp. there are only a finite number of solutions. London. 24. London. Book I. pp.. L.. London.” c. 76. Prob. 1805. 21. 1813. 1884. 86–88. 178–186. pp. [5] Bonnycastle. The first nine editions appeared “without any material alterations. trans. J.. pp. 150 B. 1879. 3rd ed.. 1612. 10. Vieweg. Gauthier-Villars. 170. In: Henry Thomas Colebrooke. Nunn et al. including Sandig. Problemes plaisants & delectables qui se font par les nombres..) Chap.. 1874. “an entire revision of the work. pp. 4th ed. Translated by John Hadley. 134–136. Arithmetica. Oxford University Press. In: T. et al. [7] Chiu Chang Suan Ching. Diophantos of Alexandria. 214–223. Lyon.(475). reprinted by Blanchard. Gaz.” which “may be considered as a concise abridgment” of his two-volume Treatise on Algebra. nos. 1959. Bijanganita. Gordon Fraser Gallery. which may be the same as the 1815 edition. Problems to Sharpen the Young. Paris. 1910. 1511. .900 and of 3294. with Notes and Observations. 1968. An Introduction to Algebra. By the way. so . v. Facsimile with transcription and commentary by John Wardroper. 1782. SINGMASTER For given and ¬ satisfying ´ ¼¼¼ µ and ´ ¼µ. A. Vogel. 6. 2nd ed. J. Analysis similar to that done before shows that there are then ½ · ´ ¬ ½   Ò µ  ¬ solutions. Wynken de Worde. P.

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

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

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

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

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

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

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

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

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

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

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

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

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

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

V AN DEVENTER Figure 8. For composite Æ up to 16.250 M. No general strategy was found. which can be reached for prime Æ . Deleted paths for composite Æ up to 16. solutions were found by a systematic trial and error search using a regular Æ -gon grid. O. 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 .

Reflection diagrams for composite Æ up to 9. lasers or burglar alarms? . for example.POINT MIRROR REFLECTION 251 Figure 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.

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

[21]). 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. each possible binary pattern occurs exactly once in ½ Ò ¾ . This basic property was independently discovered by Everett [8] and Terras [21]. C. To describe iterates of the ¿Ü·½ function we introduce some terminology. For example. Furthermore the resulting pattern is periodic in Ò with period ¾ . i. i.. LAGARIAS The ¿Ü·½ problem appeared in Martin Gardner’s “Mathematical Games” column in June 1972 [10].. Prior to that it circulated for many years by word of mouth. see [22].. 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. who studied problems similar to it in the 1930s.e. We let Ø´Òµ denote the largest value reached in the trajectory of Ò. and who has stated that he circulated the problem in the early 1950s [6]. if one takes the uniform distribution on ½ ¾ . then the distribution of the -vector ½ ½ ´Òµ Ò Ì ´ µ ´Òµ ½ ´ÑÓ ¾µ for ¼ ½ ´Òµ Ó Ú ´Òµ Ò Ì ´Òµ Ì ´  ½µ´Òµ ´ÑÓ ¾µ is exactly uniform. 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]. The first mathematical papers on it appeared around 1976 ([8]. i.e.e. Thwaites in 1952. ´½¿º¾¿µ Ö´Òµ The ¿Ü · ½ Conjecture is made plausible by the observation that ¿Ü · ½ iterates behave “randomly” in some average sense. so that ´½¿º¾¾µ Ø´Òµ ×ÙÔ Ì ´ µ ´Òµ Ò ¼ Ó where Ø´Òµ ½ if the sequence of iterates is unbounded.254 J. It was also independently discovered by B. Surveys of the known results on the ¿Ü · ½ problem can be found in Lagarias [13]. and now more than one hundred papers have been written about the problem. Finally we let Ö´Òµ count the fraction of odd iterates in a trajectory up to the point that ½ is reached. The total stopping time ½ ´Òµ of Ò is the minimal number of iterates needed to reach ½. It is usually credited to Lothar Collatz.

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

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

this explains the name “repeated £ random walk model. The random variable ½ ´Òµ is the step number £ with probability ¾ at which 0 is crossed." 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.24). Large deviation theory shows that.1) it is a straight line segment with slope  ¼ ¼¾¿½ starting from ´¼ ½µ and ending at ´­ ¼µ. This random variable is finite with probability one. the limiting graph of these trajectories has a different appearance. with probability 1. Large deviation theory also asserts that the graph of the logarithmically scaled trajectories ´½ º µ approaches a characteristic shape. £ and the expected size of ½ ´Òµ is given by (13. For each initial value Ò an entirely new random walk is used. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally.4). see Figure 1. The length of such trajectories approaches the limiting value ½ ´Òµ ÐÓ Ì ´ µ ´Òµ ÐÓ Ò ÐÓ Ò ¼ ½ ´Òµ .3X + 1 FUNCTION ITERATES 257 ¿ by starting at ÐÓ Ò and taking independent steps of size   ÐÓ ¾ or ÐÓ ¾ ½ each. so that Ø£ ´Òµ ÐÓ Ò. For the extremal trajectories for Ø£ ´Òµ in (14. let Ø£ ´Òµ equal the maximal value taken during the random walk. For the extremal trajectories for ­ £ ´Òµ in (14. we consider the random variable Ê ¾   ÐÓ ½ ¾ · ¿ ¾ Ö£ ´Òµ ½ ½ ´Òµ Ò Ì ´ µ ´Òµ Ì ´  ½µ´Òµ with ½ ½ ´Òµ Ó Large deviation theory predicts that with probability 1 it satisfies Ð Ñ ×ÙÔ Ö£ ´Òµ Ò ½ ³¼ ¼ ¼ ¼ The quantity corresponds to an estimate of the maximum fraction of elements in a ¿Ü ·½ trajectory that are odd.

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

815 10 13. where ­ ´Òµ ÐÓ Ò .5818 0.266.5857 0. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good.4).5741 0.48 18. 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 (14.328.6091 Table 2.180.728.171 4 52.6026 0.77 31.527 has ½ ´Òµ ½¾ ½ and ­ ´Òµ ¿ ¾ ½ . The agreement with the stochastic model is quite good.6020 0.13 26. 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.456.5927 0. Vyssotsky [23] has found much larger numbers with high values of ­ ´Òµ. More recently V.511 7 63. Maximal value of ­ ´Òµ in intervals ½¼ Ò ½¼ ·½. A comparison of (14.127 8 127.91 19. and the large deviations extremal trajectory is indicated by a dotted line. ½ ½¼.527 5 837.38 41.64 32.799 6 8.884. fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1. He found that Ò 12.68 24. an analysis in [15.68 Ö´Òµ 0.194.1) with numerical data up to ½¼ ½½ is given in Table 1.91 32.5841 0.400. Extensions of the data can be found in Leavens and Vermeulen [16].24 16. .769.6030 0.5967 0.3X + 1 FUNCTION ITERATES 259 Ò 1 27 2 703 3 6.371. 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.94 31. and that Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ­ ´Òµ ¿ ¾ .254 9 4. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale. Section 5] shows that in ½¼ ½½ trials one should only expect to find a largest ­ ´Òµ ¿¼.6020 0.527 Random walks model ½ ´Òµ ­ ´Òµ 70 108 165 214 329 429 592 593 706 755 21. The large deviations extremal trajectory is indicated by a dotted line.890. ½ ´Òµ It gives the longest trajectory.

.817.831 9 8.000 Ò 13.06 4. Largest value of ÐÓ Ø´Òµ ÐÓ Ò Ñ Ü ´Òµ ÐÓ 2.663 4 77.65 5.55 ÐÓ Ø´Òµ for ½¼ ÐÓ Ò Ò ½¼ ·½. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch. ½ ½¼.909 1.976 1. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ. In fact.675 7 80. LAGARIAS Ò 1 27 2 703 3 9. or 7 ´ÑÓ µ.86 7.897 2.02 21.645 ­ ´Òµ 21. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿.559 Random walks model Table 3. so do not have any significant effect on the size of the tree.10 20. In [15] these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed.788 1.099 1. where ½ ´½ º½µ Ì  ½´Òµ ´ ¨ ¾Ò ¾Ò ½ © if Ò ½. In iterating the multivalued function Ì  ½.790 1.362. if Ò ¾ or 8 ´ÑÓ µ.903 2. using (15.1).671 5 704.32 2.83 3. (See Table 3 in Section 4. C. 15.65 7. The set of inverse iterates associated to any Ò has a tree structure.03 19. 4. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ. 5.86 13.) empirically like . Branching Random Walk Model Lagarias and Weiss [15] modeled backward iteration of the ¿Ü · ½ function by a branching random walk.48 12. In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree.24 16.260 J.75 5.560 1. respectively.566.391 8 319.05 19.84 19.511 10 77.14 11. In this case. these trees have the minimal and maximal number of leaves possible for depth . Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼.59 23. there are duplications. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ.049.528.631.819 1.86 5. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth .20 6.511 6 6.804. see [15] for more explanation.46 2.791 1.

. and the root node is assigned the value 1. satisfies (with probability one) ´½ º¾µ ÐÓ Ò£ ´ µ ­ ȳ ½ ÐÑ ½ ÐÓ This constant ­ È ¬  ½ . [15. Pruned ¿Ü · ½ trees of depth Ò . In [15] it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ . where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree. In order to prove the result (15. Lagarias and Weiss [15. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½. and where fying £ ´ µ ×ÙÔ ´½ º¿µ   ÐÓ ´¾ · ½ ¾ ¿ ¿ ¼ This constant is provably the same constant as (14. The random walk at a given leaf node ends at a position Ù on the real line. We assign labels ¼ or ½ to the edges.e. where ¬ is the unique positive number satis£ ´¬ µ ¼. and the value assigned to that vertex by the rules above is exactly Ù .1). Theorem 4. cf. 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]. i. 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 .2). ­ È ­ÊÏ . it was necessary to determine the distribution of the number of leaves in a tree in the branching process. where the random walk starts at the origin on the real axis at the root node and takes a step ÐÓ ¾ along an edge labeled ¼ and a step ÐÓ ¾ along an edge ¿ labeled ½. 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.4] show that the expected size of ÐÓ Ò£ ´ µ. Theorem 3.3X + 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.

Distribution of 3x + 1 Trees Applegate and Lagarias ([1] and [3]) studied properties of the ensemble of all ¿Ü · ½ trees. 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 . This can be formalized as Conjecture C . (See Athreya and Ney [4] for a general discussion of such processes. 8 pictured in Figure 4. 4. Theorem 3. 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].) The random walk arises from assigning the labels ¼ and ½ to the edges of the resulting tree. the corresponding random walk . 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. LAGARIAS and as ½ the leaf probability distribution is of the form Ï ´ ¿ µ . see [15. with all nodes that have a label congruent to ¼´ÑÓ ¿µ removed. 2. where Ï is a random variable satisfying ´½ º µ and Û prob Ï Û ´Üµ Ü for ¼ ½ ´Üµ is strictly positive on ´¼ ½µ. 5. Since the number of leaves is expected to grow like ´ ¿ µ . In this table. Let Æ   ´ µ and Æ · ´ µ denote the minimal and maximal number of leaves. they studied the normalized extreme values ´½ º½µ ·´ µ ¿ Æ  ´ µ Ò  ´ µ ¿ Æ ·´ µ Table 3 gives data up to ¿¼. 16.262 J. C. 7. where ¼ represents the entering vertex being even and ½ represents the entering vertex being odd. from a given root node . Based on this data.2]. The branching process is a multi-type Galton–Watson process with six types labeled 1. respectively. they proposed that the normalized extreme quantities remain bounded.

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

77 133.252 0.488 0.492 0.422 0.507 0.490 0.633 0.342 0.534 0.238 1.375 2.344 2.738 1.750 0. Normalized extreme values for ¿Ü · ½ trees and for the branching process.780 1.409 0.36 3149.16 4.491 1.249 0.268 0.488 0.68 31.633 0.118 2.475 0.50 315.742 1.76 23.802 1.534 0.38 236.62 7.747 1.792 1.360 2.246 0.37 3.746 1.451 0.32 17.270 2.898 1.898 1.986 2.500 1.45 560.748 1.475 0.728 1.791 1.33 1.076 2.486 0. C.500 1.405 2.401 0.131 2.255 0.12 74.281 0.486 0.911 1.83 1771.305 0.62 1328.501 0.255 2.774 1.923 1.482 0.60 747.57 42. .489 0.475 0.419 2.534 0.481 0.300 0.166 2.273 0.77 2362.491 0. C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a fixed node Ò´ÑÓ ¿ ·½µ.782 1.211 2.67 0.232 2.746 1.748 0.741 1.758 1.327 2.802 1.259 0.243 0.352 0.99 13.302 0.190 2.03 177.422 0.669 1.47 996.307 0.09 56.433 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 4.054 2.738 1.562 0.762 1.784 1.34 420.264 J.490 0.507 0.465 0.489 0.481 0.782 1.802 1.78 2.751 1.026 2.289 0.688 1.265 0.869 1.375 0.83 99.750 1.481 0.310 2.778 1.753 1.758 1.75 5599.562 0.688 1.81 4199.481 0.49 9.750 0.240 0.292 2.557 1.390 2.394 0.490 0. LAGARIAS Ê´ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 µ 4 8 14 24 42 76 138 254 470 876 1638 3070 5766 Æ  ´ µ 1 1 1 2 2 3 4 5 6 9 11 16 20 27 36 48 64 87 Æ ·´ µ 2 3 4 6 8 10 14 18 24 32 42 55 74   ¡ ¿ ¿Ü · ½ Function Trees  ´ Branching Process µ · ´ µ  ´ µ · ´ µ 1.271 0.21 5.

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

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

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->