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
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
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 ¿
A Curious Paradox Raymond Smullyan A Powerful Procedure for Proving Practical Propositions Solomon W. Golomb Misﬁring Tasks Ken Knowlton Drawing de Bruijn Graphs Herbert Taylor Computer Analysis of Sprouts David Applegate, Guy Jacobson, and Daniel Sleator Strange New Life Forms: Update Bill Gosper Hollow Mazes M. Oskar van Deventer Some Diophantine Recreations David Singmaster Who Wins Misère Hex? Jeffrey Lagarias and Daniel Sleator An Update on Odd Neighbors and Odd Neighborhoods Leslie E. Shader Point Mirror Reﬂection M. Oskar van Deventer How Random Are 3x + 1 Function Iterates? Jeffrey C. Lagarias
Martin Gardner has had no formal education in mathematics, but he has had an enormous inﬂuence on the subject. His writings exhibit an extraordinary ability to convey the essence of many mathematically sophisticated topics to a very wide audience. In the words ﬁrst uttered by mathematician John Conway, Gardner has brought “more mathematics, to more millions, than anyone else." In January 1957, Martin Gardner began writing a monthly column called “Mathematical Game” in Scientiﬁc American. He soon became the inﬂuential center of a large network of research mathematicians with whom he corresponded frequently. On browsing through Gardner’s old columns, one is struck by the large number of now-prominent names that appear therein. Some of these people wrote Gardner to suggest topics for future articles; others wrote to suggest novel twists on his previous articles. Gardner personally answered all of their correspondence. Gardner’s interests extend well beyond the traditional realm of mathematics. His writings have featured mechanical puzzles as well as mathematical ones, Lewis Carroll, and Sherlock Holmes. He has had a life-long interest in magic, including tricks based on mathematics, on sleight of hand, and on ingenious props. He has played an important role in exposing charlatans who have tried to use their skills not for entertainment but to assert supernatural claims. Although he nominally retired as a regular columnist at Scientiﬁc American in 1982, Gardner’s proliﬁc output has continued. Martin Gardner’s inﬂuence has been so broad that a large percentage of his fans have only infrequent contacts with each other. Tom Rodgers conceived the idea of hosting a weekend gathering in honor of Gardner to bring some of these people together. The ﬁrst “Gathering for Gardner” (G4G1) was held in January 1993. Elwyn Berlekamp helped publicize the idea to mathematicians. Mark Setteducati took the lead in reaching the magicians. Tom Rodgers contacted the puzzle community. The site chosen was Atlanta, partly because it is within driving distance of Gardner’s home. The unprecedented gathering of the world’s foremost magicians, puzzlists, and mathematicians produced a collection of papers assembled by
Scott Kim, distributed to the conference participants, and presented to Gardner at the meeting. G4G1 was so successful that a second gathering was held in January 1995 and a third in January 1998. As the gatherings have expanded, so many people have expressed interest in the papers presented at prior gatherings that A K Peters, Ltd., has agreed to publish this archival record. Included here are the papers from G4G1 and a few that didn’t make it into the initial collection. The success of these gatherings has depended on the generous donations of time and talents of many people. Tyler Barrett has played a key role in scheduling the talks. We would also like to acknowledge the tireless effort of Carolyn Artin and Will Klump in editing and formatting the ﬁnal version of the manuscript. All of us felt honored by this opportunity to join together in this tribute to the man in whose name we gathered and to his wife, Charlotte, who has made his extraordinary career possible. Elwyn Berlekamp Berkeley, California Tom Rodgers Atlanta, Georgia
while a sixteen-year-old student at Tulsa Central High. and
. what chemicals are used. I sold my collections and chucked the whole business. quoted below.Martin Gardner: A “Documentary”
I’ve never consciously tried to keep myself out of anything I write. I don’t think my life is too interesting. That is. and that ink so bleached can be easily brought out by a process of ‘fuming’ known to all handwriting experts. was a query to “The Oracle” in Gernsback’s magazine Science and Invention. and I’ve always talked clearly when people interview me.. 
While there is no biography of Martin Gardner. The early interest in science. but merely bleach it. Can you give me a description of this process. and how it is performed?”  *** “Enclosed ﬁnd a dollar bill for a year’s subscription to The Cryptogram. a month later. The ﬁrst. magic. puzzles.] ***  “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 . we give a collection of quotes and excerpts. It’s lived mainly inside my brain.Also below are two quotes showing a strong childhood interest in puzzles. and writing were to stay with him. having been a fan for some time.. I am deeply interested in the success of the organization. The ﬁrst two times Gardner appeared in print were in 1930. Instead of a true biography. [1932 yearbook caption. *** “I have recently read an article on handwriting and forgeries in which it is stated that ink eradicators do not remove ink. there are various interviews and articles about Gardner.”  *** An able cartoonist with an adept mind for science. we present here a portrait in the style of a documentary. The second was the “New Color Divination” in the magic periodical The Sphinx. without narrative but arranged to tell a story.
. That institution in the 1930s was under the inﬂuence of Robert Maynard Hutchins. The erosion of my beliefs was even slower than my conversion... I can recall my satisfaction in keeping my head upright during assemblies when we were asked to lower our head in prayer. [and] Seventh-Day Adventist Carlyle B.”  *** “My mother was a dedicated Methodist who treasured her Bible and. Thus it was several years ago I decided to make a collection of mechanical puzzles. There was no particular day or even year during which I decided to stop calling myself a Christian... a lapsed Baptist who became a wellknown humanist.. Haynes... so Gardner went to the University of Chicago for what he thought would be his ﬁrst two years.. 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. “The ﬁrst and only puzzle collector I ever met was a ﬁctitious character. I learned later.” 
.. through the ﬁrst three years at the University of Chicago.  *** “My fundamentalism lasted. however. RICHARDS
began to look for something new to collect. never missed a Sunday service unless she was ill. Gardner. He was the chief detective in a series of short stories that ran many years ago in one of the popular mystery magazines. as far as I know. I became convinced that evolution was a satanic myth. For about a year I actually attended an Adventist church... A major inﬂuence on me at the time was a course on comparative religions taught by Albert Eustace Haydon.. took courses in the philosophy of science and then in philosophy. It wasn’t long until I discovered Dwight L. At that time. My father.”  *** Gardner was intrigued by geometry in high school and wanted to go to Caltech to become a physicist.... incredibly. Caltech accepted undergraduates only after they had completed two years of college. who had decreed that everyone should have a broad liberal education with no specialization at ﬁrst. I was one of the organizers of the Chicago Christian Fellowship.. but at any rate I have found the hobby equally as fascinating. Personally I can’t say that I have reaped from my collection the professional beneﬁt which this man did... Throughout my ﬁrst year in high school I considered myself an atheist. thus prevented from pursuing math and science. was a pantheist. Knowing little then about geology. then as now a citadel of secular humanism.. which wound up displacing his interest in physics and Caltech.. Moody .4
with ﬁfty or a hundred dollars at the eleventh hour — the result of having sold an idea
. Martin disposed of it all. are now his most precious. Oklahoma. and triumphantly turning up. and somewhat rareﬁed library. The rest of the year ﬁnds him periodically down to his last ﬁve dollars.. I decided I didn’t want to teach.. I wanted to write. and almost only possession. his simian stride and posture is contrasted by the gentilityand ﬂuent deftness of his hands.  *** 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. Desperate Desmond fashion. diversiﬁed. These clippings he mounted. The card entries run from prostitutes to Plautus — which is not too far — and from Plato to police museums. jetted eyebrows and spading chin.He tired of visiting oil companies every day. facing eviction from the Homestead Hotel. Gardner broke from established routine to launch himself upon his self-chosen method of traveling light through life. Those hands can at any time be his passport to fame and fortune.” Gardner said of his reporting stint. and he has spent the last halfdozen years of his life eliminating both from his consideration. In a civilization of property rights and personal belongings. for competent magicians consider him one of the ﬁnest intimate illusionists in this country today. But to fame Gardner is as indifferent as he is to fortune. Possessor a few years ago of a large. ﬁlling some twenty-ﬁve shoe boxes. 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.  ***  A slim.“Real dull stuff. Those cards. together with the summarized total of his knowledge. family that is the essence of upper middle-class substantiality. Martin Gardner is a Robinson Crusoe by choice.”  *** Gardner returned to his home state after college to work as assistant oil editor for the Tulsa Tribune. and took a job . divesting himself of all material things to which he might be forced to give some consideration. in Chicago. upon a series of thousands of ﬁling cards. He is doing his stint again this season. after having ﬁrst cut out from the important books the salient passages he felt worth saving or remembering. The son of a wellto-do Tulsa.MARTIN GARDNER: A “DOCUMENTARY”
“After I had graduated and spent another year at graduate work.
 *** His career as a professional writer started in 1946 shortly after he returned from four years on a destroyer escort in World War II. When Professor Jay Christ (Business Law) was exhibiting his series of puzzles at the Club late last Fall Gardner chanced to be in town and saw one of the exhibits. Gardner was hanging around his alma mater. too. Those mental plots evolved into imaginative short stories that he sold to Esquire magazine. and he thinks Plato.Still ﬂush with musteringout pay. accept Mr.6
D.”  *** Martin Gardner ’36 is a professional [sic] magician. RICHARDS
for a magic trick or a sales-promotion angle to any one of a half-dozen companies who look to him for specialties. the University of Chicago. but it has taken world chaos to make his father say that his oldest son is perhaps the sanest of his family..Would Mr. His personal philosophy has been described as a loose Platonism. and the night is large and full of wonder.He called up Mr. He tours the world pulling rabbits out of hats. then based in Chicago.” said the soft-spoken Gardner. Gardner’s four or ﬁve hundred?  *** He was appointed yeoman of the destroyer escort in the North Atlantic “when they found out I could type. Christ.
. The editor invited the starving writer for lunch at a good restaurant. 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. but where to park them while he was peregrinating over the globe was a problem. writing and taking an occasional GI Bill philosophy course.During the past few months a determined outpouring of ideas for booklets on paper-cutting and other tricks. who had the largest collection he had ever heard of.” “I amused myself on nightwatch by thinking up crazy plots. might object with sound reason.. He arrived with a large suitcase full of puzzles! Puzzles had been a hobby with him. 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. If he were to rest his thoughts upon one quotation it would be Lord Dunsany’s: “Man is a small thing. “pitchmen’s” novelties.. His breakcame when he sold a humorous short story called “The Horse on the Escalator” to Esquire magazine.... straight magic and card tricks. but he doesn’t like being branded. Those sales marked a turning point in Gardner’s career.
. early ESP research. He designed features and wrote stories for Humpty. It was Simon who introduced Martin and Charlotte (Mrs.. 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. another magic friend. I suppose I really didn’t get into it seriously until I wrote my ﬁrst book. Fads and Fallacies in the Name of Science. I’ve always enjoyed keeping up with those ideas.And the Immanuel Velikovsky business had just started. “was an old Navy pea jacket that smelled of diesel oil. I’ve been fascinated by crazy science and such things as perpetual motion machines and logical paradoxes. now called Scientology. I remember the hatcheck girl looking askance when I handed her the ﬁlthy rag.”  About 1947. I wrote about those three things in an article for the Antioch Review. and one of the greatest writers of children’s fantasy in the history of world literature. then expanded that article into a book by adding chapters on dowsing. Frank Baum “Oz” books.  *** “Ever since I was a boy. Children’s Digest. and Bill Simon.”  *** Although Gardner is a brand-new children’s writer.” He adds. which was then promoted by John Campbell in Astounding Science Fiction.” Gardner recalls. For about a year. Paul Curry. Dai Vernon. pyramidology.”  *** Their ﬁrst son was born in 1955 and their second three years later. Clayton Rawson. “I was brought up on John Martin’s magazine. having read all of them as a child. I was inﬂuenced by the Dianetics movement. He says that he is a great admirer of the L. Atlantis. and Polly Pigtails.“Those were good years at Humpty. but it really took off when they started attacking it on the Long John Nebel Show. performed the ceremony. Judge George Starke. the hollow-earth theories. and so on. I was astonished at how rapidly the thing had become a cult.. It took a long time for the book to start selling. Gardner) and served as best man at their wedding. Piggity’s. and regards Baum as “the greatest writer of children’s ﬁction yet to be produced by America. almost every night. I had friends who were sitting in Wilhelm Reich’s orgone energy accumulators. the book would be mentioned on the show by some guest who was attacking it. ﬂying saucers. the
. he has a good background for the task. too.MARTIN GARDNER: A “DOCUMENTARY”
“The only coat I had. he moved to New York where he soon became friends with such well-known magic devotees as the late Bruce Elliot. Persi Diaconis.
I asked him if he was a professor at Boston University. crammed with math journals in every language. all doing magic tricks. that line one alcove in his study. where graduate students had invented it.” he explains with a shrug and a gesture toward the long rows of bookshelves.  *** He got into mathematics by way of paper folding.” he recalls.”  *** Every Saturday a group of conjureres would gather in a restaurant in lower Manhattan. a professional mathematician.” Gardner reminisces. Newman’s The World of Mathematics had just been published. One of them intrigued him with a so-called hexaﬂexagon — a strip of paper folded into a hexagon. gradually. demonstrating the appeal of math for the masses. he watched his readers become more sophisticated. James R.  *** “The Annotated Alice. because that’s all the mathematics Gardner knows.”  *** At ﬁrst.Fascinated.. because Lewis Carroll was. I was lucky there in that I really didn’t have anything new to say in The Annotated Alice because I just looked over the literature and pulled together everything in the form of footnotes. “This kind of just happened. because the two books are ﬁlled with all kinds of mathematical jokes. When I ﬁrst met Asimov.”His ﬁrst columns were simple. RICHARDS
inﬂuence of which can be seen in some of the activity pages which I am contributing to Humpty Dumpty. “But I knew of some famous math books. A friend showed him a novel way to fold a strip of paper into a series of hexagons.8
D. you fool people. which turns inside out when two sides are pinched. as you know.“I’ve never made a discovery myself. does tie in with math. which led to an article on combinatorial geometry in Scientiﬁc American in December 1956.” Gardner says he never does any original work.. the column was read mostly by high school students (he could tell by the mail). but.“There would be 50 magicians or so. Gardner drove to Princeton. and Gardner was asked to do a monthly column. unless by accident. But it was a lucky idea because that’s been the best seller of all my books. of course. and I jumped at the chance. If you write glibly.. asked me
. “I’m really a journalist. he simply popularizes the work of others. as he studied the enormous literature on recreational math and learned more about it. which was a big part of the puzzle page at Humpty. but the mathematics in them has never gone much beyond second-year college level. Gardner says. So it wasn’t really too far aﬁeld from recreational math.He said no and . “At the time. I didn’t own a single math book. Through the years they have grown far more sophisticated in logic.
D. ‘you just read books by the professors and rewrite them?’ That’s really what I do. I did a column a few years ago on a marvelous paradox called Newcomb’s paradox. in decision theory.  *** “I particularly enjoy writing columns that overlap with philosophical issues. The last scoop that I got was when I heard about a public-key cryptography system at MIT. if it were ever computed. and “proof” by a ﬁctional Dr. I have also managed to provoke some outspoken enemies. H. I said I didn’t have one and he looked startled. 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.Let me say that I think not knowing too much about a subject is an asset for a journalist. would be the number 5. as the mathemagician admits. Snow’s famous “gulf of mutual incomprehesion” that lies between the technical and literary cultures.MARTIN GARDNER: A “DOCUMENTARY”
where I got my Ph. And then every once in a while I get a sort of scoop. Vladimir Nabokov pays a twinkling tribute by introducing one Martin Gardiner. For example. P. Matrix that the millionth digit of . poet W. whom he calls “an invented philospher..”  *** His “Mathematical Games” column in Scientiﬁc American is one of the few bridges over C. In his novel Ada. I realized what a big breakthrough this was and based a column on it.”  *** “I can’t think of any deﬁnition of ‘mathematician’ or ‘scientist’ that would apply to me.”  *** “I’m very ill at ease in front of an audience. and that stumped him for a moment. His wife interjected: “The fact is he doesn’t want to do
. It’s a very intriguing paradox and I’m not sure that it’s even resolved. Auden constantly quoted from Gardner. He was asked how he knew he was ill at ease if he had never done it. “not all my readers are fans. not a liability.’ he said. 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. Maybe I can explain things more clearly than a professional mathematician can.” Gardner said. and that was the ﬁrst publication the general public had on it..” Nevertheless. To everyone’s surprise — especially the hoaxer’s — the number turned out to be 5. Professors at Stanford University have just programmed a computer to carry to the millionth digit. The late Jacob Bronowski was a devotee..” 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. ‘You mean you’re in the same racket I am.
and I continue to contribute original tricks to magic periodicals. for that matter. K. and other books about math and science. Charles S. My second major hobby as a child was chess. A hobby I acquired late in life is playing the musical saw. From James he derived his notion that belief in God is a matter of faith only. Peirce.”  *** Gardner himself does not own a computer (or. during which he gave his computer to one of his two sons. William James. and most enjoyed playing was tennis. “I don’t believe God interrupts natural laws or tinkers with the universe.” he expounds. I have written a number of small books that are sold only in magic shops.  *** “Certain authors have been a big inﬂuence on me. and I have retained an interest in it ever since. and enumerates them. G.” .” 
. To my knowledge he’ll shop only for books.”  *** “My earliest hobby was magic. “I’m a scissors-and-rubber-cement man. A lot of it is a waste of time. or The Ambidextrous Universe. “From Chesterton I got a sense of mystery in the universe. From each Gardner has culled some wisdom.” he remarks. He once did — and got hooked playing chess on it.10
D. Rudolf Carnap. Miguel de Unamuno. Although I have written no general trade books on conjuring. and I looked down and saw the pattern of the chessboard on the surface on the water. Chesterton. Besides Plato and Kant. The sport I most enjoyed watching as a boy was baseball..”  *** Gardner takes refuge in magic. the wise visitor will keep his money in his pocket. a fax or answering machine). I much more enjoyed writing the book with Carnap. “Then one day I was doing dishes with my wife. The retinal retention lasted about a week. “I don’t think there’s any way to prove the existence of God logically. But when Gardner brings out his green baize gaming board. at which he probably is good enough to earn yet another living. why anything exists. “From Wells I took his tremendous interest in and respect for science. RICHARDS
it the same way he doesn’t want to shop for clothes. Gardner peers at the world with such wide-eyed wonder as to inspire trust in all who meet him. and H..” he recalls. Wells.” Gardner says.”  *** “In a way I regret spending so much time debunking bad science. I would have had little time for anything else. there are G. but I stopped playing after my college days for the simple reason that had I not done so.
1. “The Mathemagician..” Hobbies. April 21. 28. p. 7. these are the most widespread and the most damaging. “Escape to Bohemia.” 
 Martin Gardner. He knew what he wanted to do. It really is wonderful that he achieved what he achieved. destructive pseudo-scientiﬁc notions linked to race and religion. 1975. Sharpless Hickman. 63.. pp. We don’t hold meetings. no. no. for as Voltaire once said: ‘Men will commit atrocities as long as they believe absurdities.. I’m on the board of a foundation exposing this problem.”  *** “There are.  Martin Gardner. vol.. pp. pp. p. 1939. Fall 1968.  Tower Topics [University of Chicago]. vol. p. p. 27–29.’ ”  *** “In the medical ﬁeld [scientiﬁc ignorance] could lead to horrendous results. 8.” Tulsa World (Sunday Magazine).”  *** “Martin never sold out.” Science and Invention. September 1934. April 1930. No. The Cryptogram. p. August 22. One of the most damaging examples of pseudoscience is false memory syndrome.  C. “Exclusive Interview with Martin Gardner.. “Now It Is Now It Isn’t.” Pulse [University of Chicago]. and always have been. April 28. 2. The mysterians are not an organized group or anything.  Bernard Sussman.
. 3. 1119.” Time. [Letter]. 7–11. “He would never do anything that he wasn’t really interested in. 16–17. and he starved. He was poor for a very long time until he ﬁt into something.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. October 1940.  LaVere Anderson.  Betsy Bliss. 1975. “A Puzzling Collection.  Martin Gardner.” Southwind [Miami-Dade Junior College].. Mysterians believe that at this point in our evolutionary history there are mysteries that cannot be resolved.” Diaconis said.” Chicago Daily News. 1957. 2. Hopefully. “Martin Gardner’s Tongue-in-Cheek Science. p.  [Stefan Kanfer]. “Under the Reading Lamp.MARTIN GARDNER: A “DOCUMENTARY”
*** “As a member of a group called the mysterians I believe that we have no idea whether free will exists or how it works. 4. April 1932. 1. educated people can succeed in shedding light into these areas of prejudice and ignorance.
 Sara Lambert. pp. “Magician of the Wonders of Numbers. 1976. 1981. “A Great Communicator of Mathematics and Other Games: A Conversation with Martin Gardner. pp. 2. p. 38. 1994.  Anne Commire. Dover. 6E. pp.  Lawrence Toppman. “Mastermind. 10–21. “The Annotated Gardner.” Scientiﬁc American.” Skeptical Enquirer.  Philip Yam.” Newsweek. pp. 47– 48. 19. 8. “Martin Gardner. “Martin Gardner. no.” Washington Post. 10.  Lynne Lucas. 1B–2B. 89.” Omni.12
D. vol. December 9.” Skeptic. June 20. vol.” Los Angeles Times. 1–10.” Time-News [Hendersonville. vol. pp. pp. “The Puckish High Priest of Puzzles. 1979. 2.  Jerry Adler and John Carey.” Mathematical Intelligencer. RICHARDS
 Hank Burchard. March/April 1998. January 1982. 80–86.” Something About the Author.  Rudy Rucker. December 12.” Science 81. pp.  Scot Morris. 66–69. “[Martin Gardner]. “A Mind at Play. pp. no. 1981. 1993.  Kendrick Frazier.  Istvan Hargittai. 3. Material taken from the Afterword. December 5. pp.” Bookletter. 1. NC]. “The Mathematical Gamester.  Lee Dembart.  Sally Helgeson. 4. 6. p. 40–41.  John Braun. vol. vol. 36–40. November 16. 1979. December 1995. pp. 16. no. “The Math-e-magician of Hendersonville. 1981. 117–119. March 11. “Interview: Martin Gardner. pp. no. vol. p.  Martin Gardner. 4. 58.” The Charlotte [North Carolina] Observer. pp. The Flight of Peter Fromm. 1997.
. Impresario of Mathematical Games. April 1978. no. no. 4. 232–244. “Every Day. 34–39.  Michael Shermer. vol.” Linking Ring. pp. 1981. 5. July/August 1981. 1E. 32–37. 4. “Martin Gardner: A Writer of Many Interests.” The Greenville [South Carolina] News. 3. December 6. vol.  Anthony Barcellos.” Two-Year College Mathematics Journal. “A Conversation with Martin Gardner. “The Magician of Math. 56–61. 1976. 101.
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. He was a very famous science writer of the last century. I’ve heard of it.The year is 2050 A.D.
Professor Ambrose: Have you ever read the book Science: Good. Byrd: Oh? Ambrose: The chapter is titled “The Irrelevance of Conan Doyle”. Bad. and Bogus. and Doyle
SCENE I . Gardner. Byrd: To tell you the truth.Ambrose. Why do you ask? Ambrose: Because the book contains one weird chapter — It is totally unlike anything else that Gardner ever wrote. this fact has often puzzled me! How could anyone with the brilliance to write the Sherlock Holmes stories ever get involved with spiritualism—and in such a crazy way? Ambrose: You mean that you have doubts that Doyle wrote the Holmes stories? Byrd: Of course not! That thought has never crossed my mind! All I said was that I ﬁnd the situation puzzling. no! Gardner correctly pointed out that all the available evidence shows that Doyle remained quite keen and active to the end. He
. by Martin Gardner? Professor Byrd: No. scientiﬁc mind to write the Sherlock Holmes stories could possibly have spent his last twelve years in a tireless crusade against all rationality—I’m talking about his crazy involvement with spiritualism. He actually advances the thesis that Conan Doyle never wrote the Sherlock Holmes stories—that these stories are forgeries. and of course I’ve heard of Martin Gardner. I guess the answer is that Doyle went senile in his last years? Ambrose: No. rational.
Then. Now. As I said. Byrd: I just thought of another idea! Perhaps Doyle was planning all along to foist his spiritualism on the public and started out writing his rational Holmes stories to gain everybody’s conﬁdence. 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. then. Byrd: All tight. 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. his disturbance went much deeper. 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. he was surely one of the most interesting writers of the last century. So senility is not the explanation. SMULLYAN
also pointed out that Doyle’s interest in spiritualism started much earlier in life than is generally realized. how could someone of Gardner’s caliber ever entertain the silly notion that Doyle never wrote the Holmes stories? Ambrose: To me the solution is obvious: Martin Gardner never wrote that chapter! The chapter is a complete forgery. it’s just as implausible as Gardner’s idea that Doyle never wrote the Holmes stories at all. I never had the slightest doubt that Doyle wrote the Holmes stories.
. 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.14
R. Byrd: But now something else puzzles me: Martin Gardner was no fool. Doyle insisted that Houdini was lying! If that’s not psychotic paranoia. but now your explanation of the apparent contradiction between Doyle the rationalist and Doyle the crank makes some sense. Ambrose: I’m glad you realize that. whamo! Ambrose (After a pause): That’s quite a cute idea! But frankly. when the public was convinced of his rationality. I have no idea who wrote it.
But the alternative that Gardner actually wrote that chapter is not just improbable. but he certainly never wrote that chapter. GARDNER. And as Holmes wisely said: Whenever we have eliminated the impossible. they are perfectly consistent in spirit with all the sensible things that Gardner ever wrote. Byrd (After a long pause): I guess you’re right. the theory is most improbable. I don’t see how there can be the slightest doubt that this chapter is a complete forgery! Byrd: But that raises serious problems! All right. I don’t go as far as some historians who believe that Martin Gardner never existed. But that one chapter sticks out like a sore thumb—not just with respect to the other chapters. In fact. AND DOYLE
but it was certainly not Martin Gardner. whatever remains. you must be right! It is certainly not conceivable that anyone as rational as Gardner could entertain such a stange notion. The title is ”Gardner and Doyle”. nobody in his right mind could believe that Gardner actually wrote that chapter. however improbable must be the truth. 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 completely out of the question. We can only hope that future research will answer the question of how that strange chapter ever got into the book.AMBROSE. the more I think about it. why would Gardner have ever allowed the chapter to be included? Or could it possibly have gotten there without his knowledge? That also seems implausible. But surely. All the other chapters are obviously genuine. your theory strikes me as most improbable! Ambrose: I agree with you wholeheartedly. I’ll send you a copy. A person of Gardner’s caliber could never have written anything like that! Byrd: Now just a minute. No. And so I am forced to the conclusion that Martin Gardner never wrote that chapter. I believe that he did exist. he couldn’t possibly have written such a chapter. Will you please explain one thing: How did the chapter ever get there? No. are you talking about the whole book or just that one chapter? Ambrose: Just that one chapter. But now I think you’ve made a very important historical discovery! Why don’t you publish it? Ambrose: I am publishing it. 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. I can see how an entire book by an alleged author might be a forgery. Now.
where did you send it? Broad: To your Connecticut address. SMULLYAN
SCENE II . Broad: Do you believe that Doyle wrote the Holmes stories? Cranby: Of course! Why should I doubt it for one minute? Broad: Then how do you answer Gardner’s objection that no one with a mind so rational as to write the Holmes stories could possibly be so irrational as to get involved with spiritualism in the peculiarly anti-rational way that he did? Cranby: Oh. I’m familiar with much of Ambrose’s excellent work. Cranby: Good God! That’s ridiculous! That’s just as crazy as Gardner’s idea that Doyle didn’t write Holmes. had an insane streak that simply got worse through the years. and Bogus? Cranby: Oh. “The Irrelevance of Conan Doyle”? Cranby: Oh yes! As a matter of fact that is the strangest chapter of the book and is quite unlike anything else Gardner ever wrote. certainly.16
R. Science: Good. and Doyle”? Professor Cranby: No. Why do you ask? Broad: Well. are you familiar with the Ambrose paper. ”Gardner and Doyle”? Cranby: No. 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.One Hundred Years Later
Professor Broad: Did you get my paper. Bad. then I won’t get it for a couple of days. Martin Gardner? Cranby: Of course! I’m quite a fan of his. but not this one. Of course Gardner wrote that chapter! Broad: Of course he did!
. What is it about? Broad: Well. with all his brilliance. 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. you remember his book. ”Ambrose. Cranby: Oh. Of course. I think I have just about everything he ever wrote. Gardner. What is it about? Broad: You know the twentieth century writer. Broad: And do you recall the chapter. come on now! That’s no objection! It’s obvious that Doyle.
he reported the discovery of a map that required ﬁve colors. The Ancient Mariner. GARDNER. The Hunting of the Snark. In Martin’s book. so well known and highly respected for the mathematical games column he wrote for years for Scientiﬁc American. that’s where my paper comes in! I maintain that Ambrose never wrote that paper—it must be a complete forgery!
SCENE III . 1989). by the way. his numerous puzzle books. The Flight of Peter Fromm.”
. The review ends with the sentence “George Groth. Whys and Wherefores (University of Chicago Press. How could he ever believe anything that bizarre? Broad: Ah. is one of Gardner’s pseudonyms.A Hundred Years Later
(To be supplied by the reader) Discussion: How come this same Martin Gardner. an opening move in chess (pawn to Queen’s rook four) that guaranteed a certain win for white.AMBROSE. a discovery of a fatal ﬂaw in the theory of relativity. until Martin kindly informed me that the whole thing was simply a hoax! Martin is really great on hoaxes—for example. in his April 1975 column in Scientiﬁc American. is reprinted a scathing review of his The Whys of a Philosophical Scrivener by a writer named George Groth. and Casey at the Bat—not to mention his religious novel. AND DOYLE
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. 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. his annotated editions of Alice in Wonderland.
The competitions led me to believe I had a talent. while the latter was ﬁlled with ﬂights of fancy and wonderment. Welcome aboard. but now I see another side of the story. I would be remiss if I didn’t cover the topic. The contrast with my teachers in school was striking. I did get to the fun stuff. led me to the more important lesson that. the techniques of integration for a later course. there seemed to be two completely different kinds of mathematics: the kind you learned in school and the kind you learned from Martin Gardner. I learned sneaky ways of doing difﬁcult computations (a round bullet shot through the center of a sphere comes to mind). both groups being beer lovers. I learned of hexaﬂexagons (I still have somewhere in my cluttered ofﬁce a model of a rotating ring I made while in high school). and I learned of islands populated only by truth tellers and liars. with the national mood for education reform. Do I duplicate the school experiences I had with my students? Well.A Truth Learned Early
It was in high school that I decided to be a mathematician. Today. I surely try not to. But Martin Gardner. sooner or later I would get to the fun stuff. say. through his books and columns. this was no small thing.
. The former was ﬁlled with one dreary numerical problem after another. This colorful world stood in stark contrast to school mathematics. I ﬁgured that if I could just stick it out long enough. perhaps. above all else. and when my students need to know. From Martin Gardner I learned of logical and language paradoxes. The credit (or. In fact. and for an adolescent unsure of himself and his place in the world. A good part of my job now is being a teacher. It was true. blame!) for this can be laid squarely on mathematical competitions and Martin Gardner. such as the condemned prisoner who wasn’t supposed to know the day of his execution (I don’t think I understand this even now!). it seems the rest of the country is ﬁnally learning this truth too. This fundamental truth was learned from Martin Gardner when I was young and impressionable. mathematics is fun. Technical proﬁciency is a worthy goal. and it is a truth I carry in my heart. 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.
) ´¾µ I had remarked in an issue of Word Ways (May. (It can be no coincidence that the even number SIX. Matrix — two faces of the same coin. Dr. while seven is the only odd prime that on removal of one letter becomes EVEN.” Not to be outdone (as usual). His brilliant future as the world’s premiere mathematical wordsmith had been fated since the day he was christened. Gardner who introduced me to the wiley doctor over ten years ago. Matrix notes that 13 is an emirp since its reversal is also prime. an enigma. 1981. Matrix has informed me that a better statement is “Martin Gardner: a man and rarer enigma. The ﬁrst and last names have six and seven letters. upon subtraction of the same letter becomes the odd number IX.” (He also found that the square had contained a prediction for the 1980 presidential election: “Reagan ran. page 88) that the ¿ ¢ ¿ word square in Figure 1 spells out with chess king moves the laudatory phrase “Martin Gardner. In!”)
Figure 1.Martin Gardner = Mint! Grand! Rare!
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.
. For instance: ´½µ There are 13 letters in MARTIN GARDNER. It was Mr. Since then Matrix and I have marveled over the inevitability of Gardner’s career choice. Six is the ﬁrst perfect number.
Dr. Martin Gardner is indeed Mint! Grand! Rare! He has my love.22
J. Matrix claims this to be an unusual property.
. exactly one from each row and one from each column. This tribute could be continued. A common English word will be the result. Choose three letters.
A different arrangement of the nine letters produces the square shown in Figure 2. not often found in squares composed from names. but the point is clear. FARRELL
and a bit more ½¼ meters = Nucleus of a human cell ½¼ ½ meters = Atom’s nucleus
The Creator.Three Limericks — On Space. barks out his orders for the week. the Earth Sun’s System. Light Sun and Earth. Friday night At a minute to twelve Eve spin. You. Adam delve In the last millisecond. 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. Cell’s girth Our bodies. 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
. Bang!. Time. First thing on Monday. and Speed
Seven steps each ten million to one Describe the whole space dimension The Atom. form up. seen as an Army Sergeant Major.
000mph.000 mph turning speed of Galaxy 1.4 million mph Galaxy’s speed through the debris of the Big Bang
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.
I won’t give the solution. or 6 is on top. Around 1980 I started creating more of these things (which I now think could best be described as “mazes with rules”). The next page shows one of the mazes from my book. These columns presented rolling cube puzzles by Roland Sprague and John Harris. The puzzles involved tipping cubes from one square to another on a grid. At the time I thought these were puzzles. then you can tip it onto a square with an Ä. 5.) I chose this particular maze because it illustrates the cross-fertilization that Martin’s columns created. I got the original idea for this maze from remembering columns that Martin wrote in December 1963. . In my maze. What you have to do is tip the die off the starting square. If a 4. You can tip the die from one square to the next. Ó . If (and only if ) a 1. I wrote half the dopey stories and I sort of like some of them. but it takes 66 moves. 2. As Martin’s columns said. If a 1. 3. and you can only tip it onto squares that contain letters. you can tip the die onto a square with an . Mad Mazes. and March 1975. and in 1990 I had a book of them published. Position the die so that the 2 is on top and the 6 is facing you (that is. before my publisher added art work and dopey stories. He used another of my puzzles in the November 1963 column — this one involved traveling in three dimensions through a ¢ ¢ grid.
. and Ú Ò.A Maze with Rules
In his October 1962 column. you can tip the die onto a square with an Ç. or 6 is on top. then ﬁnd a way to get it back onto that square. November 1965. you can tip it onto a square with an À. or 3 is on top of the die. This is my manuscript version of the maze. The letters stand for ÐÓÛ. (Actually. or 5 is on top. but later I realized they were more like mazes. 4. Martin Gardner presented a puzzle of mine that involved traveling through a city that had various arrows at the intersections. you should think of a cube as a large carton that is too heavy to slide but that can be tipped over on an edge. the 6 faces the bottom edge of the page). If a 2. place a die on the square marked START.
but you can also download it off my website. While you’re there. You might try enlarging it on a copier. check out the rest of the site. December 1998." The ﬁrst of these walk-through mazes appeared at the Gathering for Gardner in January 1993. I have a long write-up (with pictures) of something called “walk-through mazes-with-rules. several of the mazes were built as adjuncts to large cornﬁeld mazes. Since then the concept has grown.28
R. Go to ØØÔ »» ÓÑ º ØØºÒ Ø» ÖÓ Ø ÓØØ»ÖÓÐÐº ØÑÐ. 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. ABBOTT
Maze Addendum. In the summer of 1998.
. . because your . is a six-step sequence that we might call Jacob’s Ladder. And Jacob
Â Å Æ Å Æ Î Ï Î ÏÁÎ ÄÁÎ ÄÁÎ
Ë. the words should all be present in the Bible that was used in Dodgson’s day. . and their . When the . Dodgson also was quite familiar with the Bible. . for example. seen of . . not the words. The following tableau shows. Ë
then of all .)
. that there’s a Biblical word ladder corresponding to such a transition. . the King James translation. . on the other hand. in fact. not a newfangled translation! Incidentally. bitter with . . . the sequence of verses in this ladder is strictly increasing through the Old Testament. But the tableau lists only the verse numbers. We can go from ÌÀÁË to ÌÀ Ì in three such steps: ÌÀÁË. made their . . of death . Here. . Ë Ë. ÌÀ Ì. in which one word metamorphoses into another by changing a letter at a time.Biblical Ladders
Donald E. . . . . . . ÌÀ Æ. So let’s play a game that combines both activities: Let’s construct word ladders in which all words are Biblical. . Jacob’s Ladder. . And his . was strictly decreasing. More precisely. . in the land . are written . had many . . .
( 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. . Many people consider the Bible to be a story of transition from ÏÊ ÌÀ to ÁÌÀ. . . As an ordained deacon of the Church of England. Ë Ë. what are the missing words? (Remember to use a King James Bible for reference. never backtracking in Biblical order. Knuth
Charles Lutwidge Dodgson. ÌÀÁÆ. invented a popular pastime now called word ladders. . . and their . aka Lewis Carroll. . because ‘James’ is a form of ‘Jacob’:
ËÀ Ê . ..
ÇÎ Ê the multitude . .... that his . . . . .... ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ººº ÁÌÀ. Go from ËÏÇÊ to (plow) ËÀ Ê in four steps:
.. . .. .. . every man his
against ..30 . . if you have a good concordance or a King James computer ﬁle. . . charity shall
. . .. .. . . . .. ... . ...
( Genesis 39 : 19 ) ( Genesis 40 : 2 ) ( Exodus 24 : 4 ) ( Exodus 34 : 1 ) ( Leviticus 13 : 3 ) ( Leviticus 14 : 46 ) ( Leviticus 25 : 29 ) ( Deuteronomy 22 : 21 ) ( Joshua 11 : 4 ) ( 1 Samuel 13 : 20 ) ( 1 Samuel 15 : 3 ) ( 1 Samuel 26 : 13 ) ( 1 Kings 10 : 15 ) ( Psalm 10 : 14 ) ( Isaiah 3 : 17 ) ( Isaiah 54 : 16 ) ( Isaiah 54 : 17 ) ( Habakkuk 2 : 4 )
Puzzle #2.. they sware. .......... they were .. . How about going from Æ Ã to ÇÎ Ê... . and they ... . . .. . . ... .. people never swore.... .) Puzzle #3... . .. . KNUTH
ÏÊ ÌÀ was kindled. not lift up a . E. .
. .. .. . Yea also.. . . . in eight steps? A suitable middle verse is provided as a clue.. . Of course #1 was too easy. ..
ººº ººº ººº ººº ººº ººº ººº
( Genesis 3 : 7 ) ( ) ( ) ( ) ( Luke 17 : 27 ) ( ) ( ) ( ) ( 1 Peter 4 : 8 )
... . . and his ... . . .
ººº ººº ººº
( Micah 4 : 3 ) ( ) ( ) ( ) ( 1 Samuel 13 : 20 )
(Hint: In the time of King James. Of course #2 was also pretty easy. by his
D. So neither words nor verse numbers will be given this time.
. . ....) Construct a 12-step Biblical ladder from ÇÊ Ë (Judges 3 : 28) to ÊÇÄÄË (Ezra 6 : 1)... seventy times ... . Of course #5 was too hard..... a change of pace: Construct ¢ word squares. .. .. .... (For worshippers of automobiles. . ..)
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
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. . ... Puzzle #5.
Ë Î Æ.
. which “comes back on itself” in an unexpected way... . Here’s one that anybody can do.. with only a Bible in hand.. Puzzle #6... Therefore .BIBLICAL LADDERS
Puzzle #4. Complete the following Biblical ladder. unless you have special resources. using only words from the King James Bible verses shown. . Finally. Find a Biblical word ladder from ÀÇÄ to ÏÊÁÌ.. .. . (The words will read the same down as they do across. . ..
Puzzle #1. . . . . . on the . 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 . play the . . to requite . . . the LORD. . ËÈ Ê ËÈ ËÈÁ ËÈÁÌ . . . . . the sea . . . . . mischief and . have. . within a . . . sheep that are even . is of me. man his .. Pharaoh was . . . . . Thus the middle word must be ÏÁÎ Ë. Here’s a strictly decreasing solution:
. . . . Other words can now be ﬁlled in. . . So it must be ÏÁÎ Ë or ÁÎ Æ. not lift up a . is turned . . being between . . . . . . in multitude. . still avoiding forward steps. E. . The Lord GOD hath . . . beside Eloth. . Lord will . and 1 Kings 4 : 29 could also be used for ËÀÇÊ . . . . and I will . . . . ÏÀÁÄ ÏÀÇÄ ÏÀÇÊ ËÀÇÊ ËÀ Ê . .
( Genesis 39 : 19 ) ( Genesis 40 : 2 ) ( Exodus 24 : 4 ) ( Exodus 34 : 1 ) ( Leviticus 13 : 3 ) ( Leviticus 14 : 46 ) ( Leviticus 25 : 29 ) ( Deuteronomy 22 : 21 ) ( Joshua 11 : 4 ) ( 1 Samuel 13 : 20 ) ( 1 Samuel 15 : 3 ) ( 1 Samuel 26 : 13 ) ( 1 Kings 10 : 15 ) ( Psalm 10 : 14 ) ( Isaiah 3 : 17 ) ( Isaiah 54 : 16 ) ( Isaiah 54 : 17 ) ( Habakkuk 2 : 4 )
Puzzle #2. . and ÁÎ Æ doesn’t work. . .
. a great . that it is . ..
( Micah 4 : 3 ) ( Amos 6 : 8 ) ( Song of Solomon 4 : 2 ) ( 1 Kings 9 : 26 ) ( 1 Samuel 13 : 20 )
There are 13 other possible citations for ËÏÇÊÆ. . . and the . .
was kindled. . merchants. . . . . . and . . . . . . . . ËÀÇÊ ËÀ Ê . . . . since neither Ã nor ÆÁÃ nor Æ Î nor Æ Î Æ is a word. . . .32
D. them not. and his . . in her . . by himself. . . . Yea also. with a scab . year after . . and the step after Æ Ã must be Ï Ã . all the . every man his
ËÏÇÊ ËÏÇÊÆ ËÀÇÊÆ. . . and his . . . . all the . which came . . . against two . and . Puzzle #3. upon these . . of the Red . . . that bloweth . . of the . ËÅÁÌ ËÅÁÌÀ Ë ÁÌÀ ÁÌÀ. created the . . . .
against . by his
ÏÊ ÌÀ ÏÊÇÌÀ ÏÊÇÌ ÏÊÁÌ ÏÀÁÌ . . . . .
Micah 1 : 8. . Ezra 34 : 25. . . Matthew 6 : 28. . Ë Î Ê the wicked . . . . the mighty . . . Genesis 32 : 15. . . . .) Puzzle #4. . and . Ecclesiastes 12 : 11. . . . . and sowed . . . Ï Î for a wave . . . . Ì Ê Ë among the . a . . John 18 : 18. ( 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. . . 2 Corinthians 11 : 19. looked in the . Ë. . . . and which . . . hazarded their . . (But ‘writ’ is not a Biblical word. . . . . . . . Job 40 : 17. Ê. . . Ï Ê Ë out of . Suitable intermediate words can be found. seventh hour the . And the soldiers’ . . ÌÁÅ Ë? Jesus saith . Ë of the sea. . . .
. . Psalm 145 : 15. À Ï Agag in pieces .) Puzzle #6. Ï Î Ë. . . . . Ë for the name . . Ruth 3 : 16. . . from the . . . . . For example. Ê of good men. . again. . till seven
Ë Î Æ.
( 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 )
. hewed stones. . . . . me. . Numbers 1 : 2. in Revelation 3 : 11. . Ë Î : For he will . . À Ï Ê of thy wood . Ê the multitude . . Therefore . they married . use intermediate words found in Matthew 24 : 35. . And Samuel . . . Ezra 7 : 24. . . forth. . . they were . At his . which is . Ï Ê ye shall give . as a . Matthew 25 : 33. but none are strictly increasing or decreasing. . and to the . and they . hospitality. times . Gather up thy . charity shall
33 . remnant shall be . . . . . . (See also Genesis 27 : 44. . and . Acts 2 : 45. . . . . . . . . violence of the .BIBLICAL LADDERS . . Daniel 3 : 21. . . ÌÁÊ Ë shall be . for example. Î Ê left him. . . Ë Ï with saws. . . they were . . . . shoulder Aaron . .) Puzzle #5. I have . Psalm 84 : 6. . Ë Ï sackcloth . . And your .
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]. KNUTH
Puzzle #7. Chapter 6.34
D. Games. Puzzles. The Universe in a Handkerchief: Lewis Carroll’s Mathematical Recreations. E.
ÏÇÅ Æ ÇÄÁÎ ÅÁÌ Ë Î ÆÌ Æ ËÌË
ÄÄË ÉÍ Ä ÄÍ Ê Ä Ê ËÄ È
Martin Gardner. and Word Plays (New York: Copernicus. 1996).
“Finally. They were soon banned by the Church. and BettoÌ Å . Diamonds. the Church (Spades. are all twosided card games. Clubs. Peasants.
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!
. like other forms of entertainment and gambling. Aristocracy.Card Game Trivia
14th Century: Decks of one-sided Tarot playing cards ﬁrst appeared in Europe.
20th Century: After 600 years of playing with one-sided cards. MaxxÌ Å .
18th Century: Symmetric backs and fronts were designed to prevent cheating by signaling to other players. Hearts). two-sided playing cards and games were invented by Stewart Lamle. uniﬁed (feudal) society as promoted by Joan of Arc: Nobility.) Card-playing spread like wildﬁre. you can play with a full deck!”—Zeus Ì Å .
19th Century: The Joker was devised by a Mississippi riverboat gambler to increase the odds of getting good Poker hands. competed with Holy services. (Cards.
and an even number plus an odd number is an odd number. OK?” “OK.” “No! It is an odd number! I can prove it!” How? Problem 2: Move two matches so that no triangle remains. OK?” “Yes.” “Then an even number times an even number is an odd number. and an odd number times an even number is an even number.” “An even number plus an even number is an even number. OK?” “No! It is an even number.Creative Puzzle Thinking
Problem 1: “An odd number plus an odd number is an even number.” “An odd number times an odd number is an odd number.
Problem 3: What number belongs at ? in this sequence? 13 24 28 33 34 32 11 17 16 18 14 36 35 46 52 53 22 ? 33 19 24
. OK?” “Of course.
N. (3) Find the longest time span without a palindromic time. 23:55:32. 6. 13:22:31. 50. Can you divide the ﬁgure into three identical shapes?
´Ü µ´Ü µ´Ü µ ´Ü Ýµ´Ü Þ µ
Problem 8: A 24-hour digital watch has many times that are palindromic. 3:59:53. Problem 6: Which two numbers come at the end of this sequence? 2. 30. etc. 1:01:01. 44. x. 62. 2:41:42. 4. 42.
. (Ignore the colons. (1) Find the closest such times. 56. 46. YOSHIGAHARA
Problem 4: Arrange the following ﬁve pieces to make the shape of a star. y Problem 7: The ﬁgure shown here is the solution to the problem of dividing the ﬁgure into four identical shapes. (2) Find the two palindromes whose difference is closest to 12 hours. 32. 40. 60. 52. 54. 34.) These curious combinations occur 660 times a day. 64. For example.
Problem 5: Calculate the expansion of the following 26 terms.
. You are not allowed to turn over any of the pieces. and 5 matchsticks long is divided into two parts with equal area using 3 matchsticks. but you may rotate them in the plane. 4.
Can you divide this triangle into two parts with equal area using only 2 matchsticks?
Problem 10: Cover the 7 ¢ 7 square on the left with the 12 L-shaped pieces on the right.CREATIVE PUZZLE THINKING
Problem 9: A right triangle with sides 3.
N. How many pieces do you need?
. divide the left-hand shape and rearrange the pieces to make the right-hand shape. YOSHIGAHARA
Problem 11: Using as few cuts as possible.
Number Play, Calculators, and Card Tricks: Mathemagical Black Holes
Michael W. Ecker
The legend of Sisyphus is a lesson in inevitability. No matter how Sisyphus tried, the small boulder he rolled up the hill would always come down at the last minute, pulled inexorably by gravity. Like the legend, the physical universe has strange entities called black holes that pull everything toward them, never to escape. But did you know that we have comparable bodies in recreational mathematics? At ﬁrst glance, these bodies may be even more difﬁcult to identify in the world of number play than their more famous brethren in physics. What, after all, could numbers such as 123, 153, 6174, 4, and 15 have in common with each other, as well as with various card tricks? These are mathematical delights interesting in their own right, but much more so collectively because of the common theme linking them all. I call such individual instances mathemagical black holes.
The Sisyphus String: 123
Suppose we start with any natural number, regarded as a string, such as 9,288,759. Count the number of even digits, the number of odd digits, and the total number of digits. These are 3 (three evens), 4 (four odds), and 7 (seven is the total number of digits), respectively. So, use these digits to form the next string or number, 347. Now repeat with 347, counting evens, odds, total number, to get 1, 2, 3, so write down 123. If we repeat with 123, we get 123 again. The number 123 with respect to this process and the universe of numbers is a mathemagical black hole. All numbers in this universe are drawn to 123 by this process, never to escape.
Based on articles appearing in (REC) Recreational and Educational Computing. 41
M. W. ECKER
But will every number really be sent to 123? Try a really big number now, say 122333444455555666666777777788888888999999999 (or pick one of your own). The numbers of evens, odds, and total are 20, 25, and 45, respectively. So, our next iterate is 202,545, the number obtained from 20, 25, 45. Iterating for 202,545 we ﬁnd 4, 2, and 6 for evens, odds, total, so we have 426 now. One more iteration using 426 produces 303, and a ﬁnal iteration from 303 produces 123. At this point, any further iteration is futile in trying to get away from the black hole of 123, since 123 yields 123 again. If you wish, you can test a lot more numbers more quickly with a computer program in BASIC or other high-level programming language. Here’s a fairly generic one (Microsoft BASIC):
½ ÄË ¾ ÈÊÁÆÌ Ì ½¾¿ Å Ø Ñ Ð Ð ÀÓÐ » ´ µ ½ ¿¸ Öº Åº Ïº 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 ÐÝ¸ 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 ÁÌ¸½µ 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°´ Î Æµ · ËÌÊ°´Ç µ · ËÌÊ°´ Î Æ · Ç µ ½¼¼ ÈÊÁÆÌ Î Æ Ç Î Æ · Ç ¹¹¹ Æ Û ÒÙÑ Ö × Î Ä´ÆÍ°µ ½½¼ ÈÊÁÆÌ Á Î Ä´ÆÍ°µ Î Ä´Æ°µ ÌÀ Æ ÈÊÁÆÌ ÓÒ º Æ ½¾¼ Æ° ÆÍ° Î Æ ¼ Ç ¼ ÇÌÇ ¿¼
If you wish, modify line 110 to allow the program to start again. Or revise the program to automate the testing for all natural numbers in some interval.
What Is a Mathemagical Black Hole?
There are two key features that make our example interesting:
MATHEMAGICAL BLACK HOLES
1. Once you hit 123, you never get out, just as reaching a black hole of physics implies no escape. 2. Every element subject to the force of the black hole (the process applied to the chosen universe) is eventually pulled into it. In this case, sufﬁcient iteration of the process applied to any starting number must eventually result in reaching 123. Of course, once drawn in per point 2, an element never escapes, as point 1 ensures. A mathemagical black hole is, loosely, any number to which other elements (usually numbers) are drawn by some stated process. Though the number itself is the star of the show, the real trick is in ﬁnding interesting processes. Formalized Deﬁnition. In mathematical terms, a black hole is a triple U is a function, all (b, U, f), where b is an element of a set U and f: U satisfying: 1. f (b) = b . 2. For each x in U, there exists a natural number k satisfying f (x) = b. Here, b plays the role of the black-hole element, and the superscript indicates k-fold (repeated) composition of functions. For the Sisyphus String½ , b = 123, U = natural numbers ¸ and f (number) = the number obtained by writing down the string counting # even digits of number, # odd digits, total # digits. Why does this example work, and why do most mathemagical black holes occur? My argument is to show that large inputs have smaller outputs, thus reducing an inﬁnite universe to a manageable ﬁnite one. At that point, an argument by cases, or a computer check of the ﬁnitely many cases, sufﬁces. 999, then In the case of the 123 hole, we can argue as follows: If n f (n) n. In other words, the new number that counts the digits is smaller than the original number. (It’s intuitively obvious, but try induction if you would like rigor.) Thus, starting at 1000 or above eventually pulls one down to under 1000. For n 1000, I’ve personally checked the iterates of f (n) for n = 1 to 999 by a computer program such as the one above. The direct proof is actually faster and easier, as a three-digit string for a number must have one of these possibilities for (# even digits, # odd digits, total # digits):
½For generalized sisyphian strings, see REC, No. 48, Fall 1992.
M. W. ECKER
(0, 3, 3) (1, 2, 3) (2, 1, 3) (3, 0, 3) So, if n 1000, within one iteration you must get one of these four triples. Now apply the rule to each of these four and you’ll see that you always produce (1, 2, 3) — thus resulting in the claimed number of 123.
Words to Numbers: 4
Here is one that master recreationist Martin Gardner wrote to tell me about several years ago. Take any whole number and write out its numeral in English, such as FIVE for the usual 5. Count the number of characters in the spelling. In this case, it is 4 — or FOUR. So, work now with the 4 or FOUR. Repeat with 4 to get 4 again. As another instance, try 163. To avoid ambiguity, I’ll arbitrarily say that we will include spaces and hyphens in our count. Then, 163 appears as ONE HUNDRED SIXTY-THREE for a total count of 23. In turn, this gives 12, then 6, then 3, then 5, and ﬁnally 4. Though this result is clearly language-dependent, other natural languages may have a comparable property, but not necessarily with 4 as the black hole.
Narcissistic Numbers: 153
It is well known that, other than the trivial examples of 0 and 1, the only natural numbers that equal the sum of the cubes of their digits are 153, 370, 371, and 407. Of these, just one has a black-hole property. To create a black hole, we need to deﬁne a universe (set U ) and a process (function f ). We start with any positive whole number that is a multiple of 3. Recall that there is a special shortcut to test whether you have a multiple of 3. Just add up the digits and see whether that sum is a multiple of 3. For instance, 111,111 (six ones) is a multiple of 3 because the sum of the digits, 6, is. However, 1,111,111 (seven ones) is not. Since we are going to be doing some arithmetic, you may wish to take out a hand calculator and/or some paper. Write down your multiple of 3. One at a time, take the cube of each digit. Add up the cubes to form a new
MATHEMAGICAL BLACK HOLES
number. Now repeat the process. You must reach 153. And once you reach 153, one more iteration just gets you 153 again. Let’s test just one initial instance. Using the sum of the cubes of the digits, if we start with 432 — a multiple of 3 — we get 99, which leads to 1458, then 702, which yields 351, ﬁnally leading to 153, at which point future iterations keep producing 153. Note also that this operation or process preserves divisibility by 3 in the successive numbers.
½¼ ¾¼ ¿¼ ¼ ¼ ¼ ¼ ¼ ¼ ½¼¼ ½½¼ ½¾¼ ½¿¼ ½ ¼ ½ ¼ ½ ¼ ½ ¼ ½ ¼ ÄË ÈÊÁÆÌ Ì Ñ Ø Ñ Ð Ð ÓÐ ½ ¿ººº ÈÊÁÆÌ ÈÊÁÆÌ Ë ÑÔÐ Ø ×Ø ÔÖÓ Ö Ñ ÓÖ ×ÙÑ Ó Ù × Ó Ø×ººº ÈÊÁÆÌ ÈÊÁÆÌ ÓÔÝÖ Ø ½ ¿¸ Öº Åº Ïº 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 ÁÌµ Î Ä´ÅÁ °´Æ°¸ Á ÁÌ¸½µµ ³ Ø Ú ÐÙ Ó Ø Í´ Á ÁÌµ Î´ Á ÁÌµ¶Î´ Á ÁÌµ¶Î´ Á ÁÌµ ³ Ù Ø ËÍÅ Í Ë ËÍÅ Í Ë · Í´ Á ÁÌµ ³Ã Ô ÖÙÒÒ Ò ØÓØ Ð Ó ×ÙÑ Ó Ù × Æ Ì Á ÁÌ ÈÊÁÆÌ Ì ×ÙÑ Ó Ø Ù × Ó Ø Ø× Ó Æ × ËÍÅ Í Ë Á ËÍÅ Í Ë Æ ÌÀ Æ ÈÊÁÆÌ ËÙ ×× ÓÙÒ Ð ÓÐ ÈÊÁÆÌ ÇÌÇ ¼ Æ ËÍÅ Í Ë ÇÌÇ ½½¼
This program continues forever, so break out after you’ve grown weary. One nice thing is that it is easy to edit this program to test for black holes using larger powers. (It is well known that none exists for the sum of the squares of the digits, as one gets cycles.) In more formal language, we obtain the 153 mathemagical black hole by letting U = 3 · = all positive integral multiples of 3 and f (n) = the sum of the cubes of the digits of n. Then b = 153 is the unique black-hole element. (For a given universe, if a black hole exists, it is necessarily unique.) Not incidentally, this particular result, without the “black hole” terminology or perspective, gets discovered and re-discovered annually, with a paper or problem proposal in one of the smaller math journals every few years.
and re-dealing out by rows. To ﬁnd an analagous black hole for larger powers (yes. First of all. by cases or computer check. which is to say.46
M. card 11. so 153 is indeed a ﬁxed point. Remove 21 cards from an ordinary deck. There are two ways to prove this. for suitably small numbers. Now re-lay out the cards by laying out by rows (i. for large numbers n. W.e. Now ask him (or her) which of the three columns contains the card. but be sure to sandwich the column that contains the chosen card between the other two columns. ECKER
The argument for why it works is similar to the case with the 123 example. It’s a classic card trick. the card chosen must be in the center of the array. Ask somebody to think of one of the cards without telling you which card he (or she) is thinking of. Repeat asking which column. 1¿ + 3¿ + 5¿ = 153. but the easier way is to draw a diagram that illustrates where a chosen card will end up next time.
½¼ ¾¼ ½¼¼ ½½¼ ½¾¼ ½¿¼ ½ ¼ ½ ¼ ÄË ÈÊÁÆÌ Ì ¾½ Ö × ¹ ÈÖÓ ÈÊÁÆÌ ÔØ Ý Öº Åº Ïº ÁÆÌ ¹ ÁÅ Î´¾½µ¸ ´ ¸¿µ ÁÌ Ê ¼ ÇÊ Æ ½ ÌÇ ¾½ Î´Æµ 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
. for the black-hole attraction. yet meets the two criteria for a black hole. Second. laying out three across at a time). But for those who enjoy programs. and then test to see whether other numbers are drawn to it. n. Regroup the cards by picking up the cards by whole columns intact. try this one from one of my readers. f (n) each value eventually is “pulled” into the black hole of 153. Then. I’ll omit the proof. Even
Here’s an example that sounds a bit different. regrouping cards with the designated column being in the middle. note that. you will need ﬁrst to discover a number that equals the sum of the fourth (or higher) power of its digits.. At the end. Repeat one last time. there are some). Arrange them in seven horizontal rows and three vertical columns. This is the card in the fourth row and second column.
. Fall 1992. That is. the smallest permutation (allowing initial zeros as digits). Rearrange the digits of your number to form the largest and smallest strings possible. as with the sisyphian strings. but this trick. nonetheless.
¾REC. No. further iteration with 6174 is pointless: 7641–1467 = 6174. Within the total of seven steps. and subtract. don’t take one of the nine numbers with four identical digits). generalizes somewhat. Take any four-digit number except an integral multiple of 1111 (i. Apply this same process to the difference just obtained. write down the largest permutation of the number. involve numbers.MATHEMAGICAL BLACK HOLES
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 Î´Æµ 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 ¿ Â Ç´Ãµ Æ Æ · ½ Î´Æµ ´Á¸Âµ Æ Ì Á Æ Ì Ã ÇÌÇ ½ ¼
½ ÌÇ ¿
Ö ´½¸ ¾¸ ÓÖ ¿µ ¿ ÌÀ Æ ¾ ¼ Ç´Ãµ Ã Æ Ì Ã ËÏ È Ç´ µ¸ Ç´¾µ ÇÊ Á ½ ÌÇ
Perhaps it is not surprising. you always reach 6174. 48.
. At that point. ¾
Kaprekar’s Constant: What a Difference 6174 Makes!
Most black holes.e.
One of the classic results is that the ratio of consecutive terms has a lim¾ ¿ iting value. phi. In fact. ECKER
Example: Start with 8028. Its value is roughly 1. 55. the smallest is 0288. 5. This one is a bit more tedious. 5. The black-hole number this time is 15. but without using a program. To get a quick and dirty feel for why phi arises. Now take the sum of the digits of these divisors. 3. For large Ò this equation is approximately the same as Ä ½ · ½ Ä. ½ . we would have obtained the reciprocal of phi as the limit. but fewer authors use that approach. Divide both sides of the equation ´Òµ ½µ · ´Ò ¾µ by ´Ò ½µ.48
M. 13. That is. assume that the numbers ´Ò · ½µ ´Òµ approach some limit — ´Ò call it Ä — as Ò increases. it is well known that the sequence converges to ´½ · µ ¾ or the golden number. 3. Had we used the ratios ´Òµ ´Ò · ½µ instead. 34. Iterate until a number repeats. and the difference is 8532. W. 8.618 . and ´Òµ preceding two elements. form the ratios ´Ò · ½µ ´Òµ: ½ ½ ½½ ¾¾ ½ ¿ ½ (approx. More formally.). but it is also that much more strange at the same time. 21. are each 1 and successive terms are obtained by adding the immediately ½ ´¾µ.6 or so. and 15. If we multiply both sides by Ä we obtain a quadratic equation:
Ä¾ Ä ½
. Repeat with 8532 to calculate 8532–2358 = 6174. but you will reach 6174. ´½µ ´Ò ½µ · ´Ò ¾µ for integers Ò ¾. 2. Your own example may take more steps. The ﬁrst two numbers numbers: 1. etc. but your very equilibrium. and these digits sum to 15. The ratios seem to be converging to a number around Ô 1.
The Divisive Number 15
Take any natural number larger than 1 and write down its divisors. Its divisors are 1.
Fibonacci Numbers: Classic Results as Black Holes
Many an endearing problem has charmed mathophiles with the Fibonacci . including 1 and the number itself. 89. The largest permutation is 8820. 1. This one may defy not only your ability to explain it.
Notice that the above plausibility argument did not use the values F (1) and F (2). We have not ﬁnished. One obtains convergence to phi — but still no appearance of the second number that is the solution to the quadratic equation above. Each is also one minus the other. one of which is phi. then ¿ . The ﬁrst two numbers need not even be whole numbers or positive. Ü may not equal ½ either. one gets the same result: convergence to phi. Thus. This. too. is easy to test in a program that you can write yourself. Is there a connection between the two solutions of the quadratic equation? First. Second. . We must now avoid ´Üµ ½ (otherwise. in succession. could not use an input Third. Thus. Consider the function ´Üµ ½·½ Ü for nonzero real numbers Ü. For ´Üµ ½·½ Ü. if we extend the deﬁnition of black hole to require only that the iterates get closer and closer to one number. ﬁnd. But there is still another black hole one can derive from this. more generally.MATHEMAGICAL BLACK HOLES
Solving by the quadratic formula yields two solutions. we
. Solving ½ · ½ Ü ½ gives Ü ½ ¾. If we continue now working backward with function ’s preimages we ¾ . I selected (or rather stumbled across) this function because of the simple argument above that gave phi as a limit. . ´Üµ ½·½ Ü. In closing out this brief connection of Fibonacci numbers to the topic. Notice that these fractions are precisely the negatives of the ratios of consecutive Fibonacci numbers in the reverse order that we considered. along with 0. The solution to the equation ´Üµ ½ · ½ Ü ¼ is just Ü ½. must also be eliminated from the universe. ´ ´Üµµ ¼ and then ´ ´ ´Üµµµ is undeﬁned). had we formed the ratios ´Òµ ´Ò · ½µ. Start with a seed number Ü and iterate function values ´Üµ ´ ´Üµµ. 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. each number is the negative reciprocal of the other. if you take any additive sequence (any sequence — no matter what the ﬁrst two terms — in which the third term and beyond are obtained by adding the preceding two). Indeed. All of these must be eliminated from the universe for . that we must similarly rule out ¿ ½¿ etc. 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. etc. More about the second solution in a moment. the second golden ratio. we have a black hole once again. The limiting ratio.
(Just multiply all the odd prime factors together. If even. Start with a natural number. the numbers ´Ò ½µ ´Òµ. The only non-odd prime is 2. Success! In fact. Now try some examples. this problem has the paradoxical property that. If odd. For instance. Must you always reach 1? If you start with 5. then 2. is among the easiest to program (a few lines). our black hole called phi. W. while ¼ ½ is a repeller. This answer is the next iterate. ECKER
obtain our two golden numbers. triple and add one. In the example that would be ¿¢ ¾½. iteration of matrix powers draws one to a result that represents long-term stability independent of an input
. dating back to the 1930s and still an open question (though sometimes also identiﬁed with the names of Hailstone. As things are set up. Because this is so easy to program. You should ﬁnd that you keep getting 4. which really are cycles of length one. Pick the largest odd factor.) Now triple the largest odd factor and then add 1. then 1 again. Once you hit 4. If you do reach 1 — and nobody has either proved you must. because the largest odd factor in 4 is 1. then 8. then 2. nor shown any example that doesn’t — then you next get 4. and conversely. So. take half. although one of the hardest to settle deﬁnitively. Keep iterating. then 1.
Unsolved Problems as Black Holes
Even unsolved problems sometimes fall into this black-hole scenario. you get 16. Hmm now only in black holes. then 4. 84 is ¾ ¢ ¾ ¢ ¿ ¢ . I will omit a program here. and the second golden number lead to the attractor. Anybody who proves the Collatz Conjecture will prove that my variation is a mathemagical black hole as well. Under certain conditions. We’re interested that repeats ad inﬁnitum. All real inputs except zero. a cycle a cycle of length three. Deﬁne the process instead by taking the starting number and breaking it down completely into factors that are odd and even. let’s be creative and ﬁx this up by modifying the problem. Consider the Collatz Conjecture.50
Other examples of mathemagical black holes arise in the study of stochastic processes. and Syracuse). Ulam. and ¿¢½·½ . you stay at 4. the number ½ ½ is an attractor with the black-hole property.
MATHEMAGICAL BLACK HOLES
. Ecker. the teasers and problems here are intriguing in their own right. There are striking resemblances to convergence results in such disciplines as differential equations. I invite readers to send me other examples or correspondence: Dr. Pennsylvania 18411-9206 USA Happy Black-Hole hunting during your salute to Martin Gardner. whom we are proud to make a tiny claim to as our Senior Contributing Mathematical Editor. Recreational and Eductional Computing. Michael W. Editor Recreational and Educational Computing 909 Violet Terrace Clarks Summit. the real value for me is in seeing the black-hole idea as the unifying theme of seemingly disparate recreations. in which we’ve had additional ones. too. Quite apart from any utility. This is an ongoing pursuit in my own newsletter. Moreover. however.
Dieter Gebhardt. add information on the source of the problem. a company newsletter published at Logicon. Harry Nelson. Clayton Dodge. A sources section provides information on the source of each problem as far as I know. After lunch he gets
. Karl Scherer. James Dalgety. David Singmaster. the solution section gives an approach to answering each of the problems. Yoshiyuki Kotani. Source information begins on page 82. Naoaki Takashima. medium. After going 10 miles in a straight line he stops for lunch. Inc. Laurie Brokenshire. He gets up and travels due North. Martin Gardner. other solvers through word of mouth. I would be delighted to hear from anyone who can improve the solution approach.Puzzles from Around the World
Richard I. Dario Uri. and hard . Hess
Most of the puzzles in this collection were presented in the Logigram. especially. Nick Baxter. These problems are for the enjoyment of the solver and should be passed on to others for their enjoyment as well. A man has breakfast at his camp. or entirely new problems of my own creation. it consisted of problems from a number of sources: other problem columns. Brian Barwell. where I worked for 27 years.. Allan Gottlieb. The regular problem column appeared from 1984 to 1994 and was called “Puzzles from Around the World”. Nob Yoshigahara. embellishments or adaptations of such problems. and. Bob Wainwright. I thank the following people for providing problems and ideas that have contributed to the richness of the problems involved: Leon Bankoff.
E1. or offer more interesting problems for future enjoyment. The problems are arranged in three general categories of easy.
Through transferring liquid among the containers. After going 10 miles in a straight line he ﬁnds himself back at camp. A boat starts from being at rest in Los Angeles Harbor and proceeds at 1 knot per hour to Honolulu. 25 and 9 combine to give 16) and use the rule to determine Ü. 9 in some permutation such that when digits are removed one at a time from the right the number remaining is divisible in turn by 8. 7. 4. 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. Find a nine-digit number made up of 1. Find the rule for combining numbers as shown below (e. reverse the digits of the result. 6. The 15pint container starts out full. E6. 2. g. E9. subtract 1 and you produce the original number. 10 pints. I.54
R. How long does it take? E5.. 3. 63 9 38 33 32 12 88 25 16 18 15
E3. 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. 2 and 1. 6. Divide the number by 3. 3. and the other two start out empty: (15. Where on earth could he be? E2. 0. It is approximately 2244. You have containers that hold 15 pints. 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. HESS
up and travels due North. E8. 5. What number belongs at Ü in this sequence? 34 32 36 46 64 75 50 35 34 16 18 14 22
40 35 15 20 12
E4.5 nautical miles from Los Angeles to Honolulu. What is the number and what is the next greater number (possibly with more than three digits) having this property?
. Find the most efﬁcient solution. My regular racquetball opponent has a license plate whose three-digit part has the following property. Calculate the expansion of this 26 term expression:
´Ü µ´Ü µ´Ü µ´Ü µ ´Ü Ýµ´Ü Þ µ
. 0). 7. 4. 8. 5. and 6 pints.
how? Try these other problems. (b) There are seven men with crossing times of 1. 9. It is night. and they must cross in 25 minutes. 4. Can all four men cross the bridge? If so. and they need their one ﬂashlight to guide them on any crossing. Find nine single-digit numbers other than (1..
E13. and the bridge will hold up to three men at a time. Each man walks at a different speed: A takes 1 minute to cross. C takes 5 minutes.880.. 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. E11. and 9) with a sum of 45 and a product of 9 362. 7. They all start on the same side and have 17 minutes to get across. If it cannot move to a square previously visited. E14. 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.
. Divide the ﬁgure below into (a) 4 congruent pieces and (b) 3 congruent pieces. The Bridge. 3. 3. B takes 2 minutes. 6. A knight is placed on an inﬁnite checkerboard.PUZZLES FROM AROUND THE WORLD
E10. (a) There are six men with crossing times of 1. Four men must cross a bridge. . 8. and 10 minutes. A maximum of two people can cross at one time. 2. and D takes 10 minutes. 2. 6. and 9 minutes and they must cross in 31 minutes. 8.
a shoelace. Despite your lack of power you took 6 tricks. What is the minimum time interval you can measure?
M1. You are given some matches. Potato Curves. N thirsty customers arrive. It has a symmetry property in that the burn rate a distance Ü from the left end is the same as the burn rate the same distance Ü from the right end. From England comes the series 35. (b) a right triangle? M3. In clubs you held 5432.56
R. 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. What is the probability that they form (a) an acute triangle. You were playing bridge as South and held 432 in spades. and diamonds. Shoelace Clock. making a 4-club contract. and a pair of scissors. 45. Three points have been chosen randomly from the vertices of a cube. hearts. Suppose a clock’s second hand is exactly on a second mark and exactly 18 second marks ahead of the hour hand. What time is it? E17. 60. You are allowed to draw a closed path on the surface of each of the potatoes shown below. x. 120. 450. HESS
. You have a 37-pint container full of a refreshing drink. 744. Find a simple continuous function to generate the series and compute the surprise answer for Ü. one having an 11-pint container and another having a 2Npint container. 1260. (b) N = 5? M2. Can you draw the two paths so that they are identical to each other in three-dimensional space?
E16. 180. The lace burns irregularly like a fuse and takes 60 minutes to burn from end to end. I. Produce the other hands and a line of play that allows this to occur. 280. M4.
(a) How many legal combinations are there? (b) If a sixth button were added. Show that no more than two of its vertices lie on grid points. Also. M11. What is the longest Lyran musical composition? M12. What is the least value for L that allows a successful plan for crossing the river?
M9. (14)] are not. (4)] are legal combinations while [(1). Thus [(12). (14)].
Ô¿ · ¾ · ¿½¼
Ô¿ . 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). Find the smallest prime number that contains each digit from 1 to 9 at least once. (a) Determine the weight of each package with a two-pan balance and four weights of your own design. Music on the planet Alpha Lyra IV consists of only the notes A and B. The river shown below is 10 feet wide and has a jog in it. M10. (b) Now do it with three weights. Many crypto doorknob locks use doors with ﬁve buttons numbered from 1 to 5. (34)] = [(21). A regular pentagon is drawn on ordinary graph paper. (34)]. how many legal combinations would there be? M13. and [(13). (13). ¾ ¿½¼
Prove there are no
M8. 26 packages labeled A to Z are known to each weigh whole numbers of pounds in the range 1 to 26.
Ô such that ¾Ô + Ô¾ is also prime.PUZZLES FROM AROUND THE WORLD
M5. (3)]. Find all primes others. Legal combinations allow the buttons to be pushed in speciﬁc order either singly or in pairs without pushing any button more than once. 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. From the USSR we are asked to simplify Ü M7. [(134)]. it never includes three repetitions of any sequence nor does the repetition BB ever occur.
Dissect a square into similar rectangles with sides in the ratio of 2 to 1 such that no two rectangles are the same size. You and two other people have numbers written on your foreheads. M16. A solution with nine rectangles is known.01). A solution with eight triangles is known. and (b) farthest from the far end of the band? M19. A snail starts crawling from one end along a uniformly stretched elastic band. You are told that the three numbers are primes and that they form the sides of a triangle with prime perimeter. What number is written on your forehead? M18. d. You see 5 and 7 on the other two people. M15. (c) two of the regions are the same size and the third region is a different size. which point is the greatest distance away? (It is not B.) (b) What are the two points farthest apart from each other on the surface of the dicube? (The distance. The snail maintains his grip on the band during the instant of each stretch.58
M14. Divide an equilateral triangle into three contiguous regions of identical shape if (a) All three regions are the same size. The band is initially 100 feet long and is instantaneously and uniformly stretched an additional 100 feet at the end of each minute. It crawls at a rate of 1 foot per minute. 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. both of whom state that they cannot deduce the number on their own foreheads. (b) all three regions are of different size. M17. (a) If the ant starts at point A.
. I. At what points in time is the snail (a) closest to the far end of the band. An ant crawls along the surface of a 1 ¢ 1 ¢ 2 “dicube” shown below. between these points is greater than 3.
” she said. And we all just managed 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. The ﬁgure following this problem shows a 30–60–90 triangle divided into four triangles of the same shape. Well. Not that that’s saying a lot. Isn’t it differential equations.” said Humpty Dumpty. many years ago. or something I’ve never learned?” Humpty Dumpty opened one eye. “But I thought you were good at sums. Humpty Dumpty.” What was Humpty Dumpty’s mark in Arithmetic?
H1.PUZZLES FROM AROUND THE WORLD
M20. I’ll tell you now.” “But you haven’t told me . “Don’t be a fool. interested. “Good at sums. and that is also true of the other Good Eggs who qualiﬁed.” he said crossly. “You don’t like arithmetic. “I know I haven’t told you how many marks in all one had to obtain to qualify. child? I don’t very much. And now you can ﬁnd out all you want to know. How many ways can you ﬁnd to do this? (Nineteen solutions are known. “I can’t. The number of subjects was one third of the number of marks obtainable in any one subject. that was — I got a better mark in arithmetic than any of the others who qualiﬁed.” “How many subjects were there?” said Alice. “I must think. “Ah!” said Humpty Dumpty.” He composed himself for a nap.” said Alice. multiplied by the number of subjects (I’ve told you about that already). “I’m coming to that. But what has that got to do with liking them? When I qualiﬁed as a Good Egg — many. Alice was almost in tears. And I ought to mention that in no two subjects did I get the same mark.” said Humpty Dumpty. But to make it as easy as I can for you. child. “So I am.” said Alice. 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. “do any of it.” “But how many .” said Humpty Dumpty. None of us did as well in arithmetic as in any other subject. The number of other Good Eggs who qualiﬁed when I did. oh certainly. Similar Triangles.” began Alice. “How many of us were there? Well. I’ll put it another way. It was a number equal to four times the maximum obtainable in one subject.)
) Assume that it takes Sir George 1 minute to pound a stake and that he walks at a constant speed between stakes. All the land in the convex hull of your stakes will be yours. HESS
H2. An irrational punch centered on point P in the plane removes all points from the plane that are an irrational distance from P. P. how tall is the building? (Use 20. “I grant you the land you can walk around in a day. Given that the width of the building is 125 ft and the rubber band must stretch an extra 10 cm to accommodate the building. and be back at your starting point in 24 hours.902. “Take some of these stakes with you. 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. “To reward you for killing the dragon. on the exact top is resting on the horizontal plane.” she continued.)
H5. I. Imagine a rubber band stretched around the world and over a building as shown below. A billiard ball with a small black dot.” (The Queen had read a little mathematics. Where is the black dot when the ball returns to its initial resting place?
.” the Queen said to Sir George. “Pound them into the ground along your way. How many stakes should he take with him to get as much land as possible? H3.851 ft for the radius of the earth. What is the least number of irrational punches needed to eliminate all points of the plane? H4.” She pointed to a pile of wooden stakes.60
(11. A good approximation of
¿ ½. What is the smallest value of n for which you can ﬁt 2n + 1 such circles into a 2 ¢ n rectangle?
H10. decimal points.
(a) Find the best approximation using two digits of your choice. 4). H11. (8. You may use · ¢ . Two evenly matched tennis players are playing a tiebreak set. 2). (c) Show that ½ can be approximated with arbitrary accuracy using any two digits if the rules of part (b) are followed. As shown below it is easy to place 2n unit diameter circles in a 2 ¢ n rectangle. Prove or disprove that line ST is perpendicular to plane ABCD. (9. No square roots or other functions are allowed.PUZZLES FROM AROUND THE WORLD
H6. exponents. (12. 4). taken from vertices S and T. (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. 6). 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. H8. (12. Tennis Paradox. The server in any game wins each point with a ﬁxed probability p. and parentheses. 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. (b) The same as part (a) except square roots are allowed. 5). 4). where 0 p 1.n) = (a) (b) (c) (d) (e) (f) (g) (3. The altitudes of their eight triangular faces. 3). You are given two pyramids SABCD and TABCD. (7. are all equal to 1.
their ages are ” H14. The drawer starts with 16 socks each Monday morning (eight blue.0107 0. ) 0. Find other such examples. x) = a bx .a . 113. d. six black.62
R. . given the sum and product of their ages. An approximation to is 355/113.0107. and proceeds to explain: “I normally ask guests to determine the ages of my three children.” Your response is “No need to tell me more. where e =
H16. The hostess at her 20th wedding anniversary party tells you that the youngest of her three children likes her to pose this problem. My uncle’s ritual for dressing each morning except Sunday includes a trip to the sock drawer. ) = 0. I’ll let you off the hook. b) Find the minimum c and d so that g (c. b. (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. HESS
H12. b. I. e ) 2. Minimum Cutting Length. into ﬁve parts of equal area? (c) an equilateral triangle into ﬁve parts of equal area? H15.
(b) Can a third Pythagorean triangle be abutted such that 4 of the 7 lengths are primes?
. Since Smith missed the problem tonight and Jones missed it at the party two years ago. (a) Find the minimum a and b so that g (a. 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). Deﬁne the goodness of ﬁt of a /b to x as g (a. where he (1) picks out three socks at random.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. (a) The ﬁgure shows two Pythagorean triangles with a common side where three of the ﬁve side-lengths are prime numbers.0107 .71828 . A number may be approximated by a /b where a and b are integers. with g (355. two brown) and ends up with four socks each Saturday evening.
) Unfortunately. Its axis of rotation coincides with the spheroid’s small axis. 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. On Lyra III one is a student until reaching a special year immediately following a special year. (A spandral is the Lyran unit of length. It’s internal structure. is unique. each of homogeneous composition. The outermost of these layers is nearly pure krypton. one then becomes a master until reaching a year that is the third in a row of consecutive special years. By nature the Venerians can only survive at the krypton/fromage boundary and by tragic mistake the two copods had landed on disconnected branches of
. 18 and 20. Inc. where Ô and Õ are different prime numbers. It (the Venerians are sexless) also reported that the straight line distance between it and its copod was 120 spandrals. ﬁnally one becomes a sage until death.331 to 1. and the next inner layer is an anhydrous fromage. being made of a number of coaxial right circular cylindrical layers. (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.PUZZLES FROM AROUND THE WORLD
H17. is contemplating mining the outer layer. which occurs in a special year that is the fourth in a row of consecutive special years. The planet Alpha Lyra IV is an oblate spheroid. and the company’s ﬁnancial planners have found that the venture will be sound if there are more than one million cubic spandrals of krypton in the outer layer. The inhabitants of Lyra III recognize special years when their age is of the form Ô¾Õ. (A copod corresponds roughly to something between a sibling and a rootshoot. however. The common axis of these layers is the planet’s rotational axis. The ﬁrst few such special years are 12. just as one observes for Sol III..) The reporting Venerian was mildly comforted because the distance to its copod was the minimum possible distance between the two. 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. Cosmic.
swearing at the difﬁculties he had in crossing the channel of the haggard dagger digger. (b) Find ´½¼¼¼µ to three signiﬁcant ﬁgures. 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.
How long was the channel which the braggart haggard dagger digger dug? H20. An even more curious fact is that Taurus is a torus (doughnut shape). they dug a series of channels in the satellite surface. a rather boastful sort. Before they were wiped out by a permeous accretion of Pﬁster-gas. Should Cosmic. undertake the mining venture? H19. had merely walked the 60 spandral length of another channel. Taurus.64
R. The second student was very tired from his work. The third student had explored a channel 50 spandrals in length. The ﬁfth student. (c) Find the smallest Ñ such that the magnitude of ´Ñµ is less than the magnitude of ´Ñ · ½µ. HESS
the curve of intersection of the krypton/fromage boundary and the planetary surface. He found nothing but the fact that the length of the channel was 30 spandrals. The fourth student. lying along the intersection of a plane with the satellite’s surface. Inc. was occupied by a race of knife-makers eons ago. crossing the channel of the haggard dagger digger. a rather lazy fellow (a laggard dagger digger?). a moon of a-Lyra IV (hieronymous). Deﬁne
½ ´ ¼
¾ ½ ¾ ½
(a) Prove that ´Òµ goes to zero as Ò goes to inﬁnity. A curious feature of these channels is that each is a complete and perfect circle.. I.
. He had dug the entire length of a longer channel but never crossed the path of the ﬁrst dagger digger. was also tired because he had thoroughly dug the entire length of the largest possible channel.
(0. 0. 6. 2). (9. 0. Anywhere within 10 miles of the north pole. 6).
E9. 2). The goal is to ﬁnd the smallest integer values of and so that
is a minimum and ¼ By numerical search we ﬁnd
½ ¼¼ ¼¾
. 10. 0. 0). Thus Ü ½½. 0). Thus x = 4 ¢ 6 + 2 ¢ 2 = 28.425. 5).654.
½ ¼½ or ¼
½ ¼½. 2). 4. E7. 0). (11. In each case the numbers combine by summing the products of their digits. 2. 9. E6. 0). E3. (7.729. 4. so E8. But this is not right because 18 is not the difference of 38 and 16. (1. E10. 0). At ﬁrst glance. the next greater number is 7. (1. and Ü ½ .
¼. the rule is that two numbers combine to give a number that is the sum of the digits of the two numbers. (11. 6). 7. One of the terms is ´Ü Üµ. 9). 8. 2. 0). 10. 10. Thus and Ò ¾½¼ transfers is the minimum.
.741. 5. it appears that the rule might be subtraction. (3. E4. 6).PUZZLES FROM AROUND THE WORLD
Solutions to Easy Problems
E1. 8. . 2. The trick here is that 1 knot = 1 nautical mile per hour. (7. (9. 15 moves: (15.
are the smallest values satisfying the above equations. 6). 6. 0. so 1 knot per hour is a constant acceleration of 1 nmi per hour per hour. (5. 6. 4. (5. ¾ gives Ø E5. 6. Instead. (3. (3. 6). (1. The number is 381. Ø¾ ¾¾ hours. 4. E2. The license number is 741. (3. 6).
A comes back. I. and 10 go across. 2 comes back. 6. 9. 1 and 2 go across. A and B go across. The solutions are shown in the following ﬁgures. 1 comes back. C and D go across. 1 comes back. E14. HESS
E11. A and B go across. Imagine intersecting the potatoes with each other. 8. 1 comes back. B comes back. 8 and 9 go across.
E15. Yes. (b) 1 and 2 go across. 3 comes back. The path of their intersecting surfaces is a desired path. (a) 1 and 3 go across. 1 comes back. The ﬁgure below shows how to trap the knight after 15 moves.66
R. See the array below. 1 and 3 go across. 1. 1 and 6 go across.
E12. 1 comes back.
.and 7 go across.
E13. 1 and 4 go across.
11. (23. (25.
Solutions to Medium Problems
M1. (35. noting the point that burns last. (35. 0). 0). (24. 0. 6. 11. (26. 0). (30. 10.PUZZLES FROM AROUND THE WORLD
E16. (27. (22. 11. 10. (a) 32 moves. (11. 2. (31.75 minutes. 6). (30. 1). (23. 6). 11. 6). (21. 5. 0. 4. start burning C from both ends and D from one end. 10). 0. 11. 1). 6). D will ﬁnish burning. 0. (24. 0. 1). 147. (25. (19. (24. (29. 10). 11. 10). There are 7 ¢ 6/2 = 21 choices using vertex 1. 0. 0). 10). 5. 6. (18. 0). (25. 0. 6. 0). (19. (18. 2. 3. (34. 1. 1. 4). 168. 1. 138. 0. 9. (30. 0. 11. (23. 5. 4. (29. (36. (31. (23. 0. (23. (25. 10). 6). 6). 4). 3). 9. When C is consumed. 0). 0. 3). (31. (35. and the time is 8:24. 11. 0). 10). (19. 1). 0). The probability of an acute triangle is ¿ ¾½ or ½ . 5). 3. (28. (25. 167. 1. 4). 11. (17. 0). 11. 0). 146. 11. 0). 1). 4). (28. (18. 145. 0). 0). 11. 0. (37. See the ﬁgure. 0. Cut the lace in half. (31. (29. (18. 10). 4. 4. (31. (24. 0). 6. 6). 0. 10).
. 11. 125. 0). (24. (29. 11. 0. At the same time. 1. (33. 148. In 3. 0). (37. 1). 127. E17. (26. 1. producing pieces A and B. 0). 156. M2. 0. 0. 3. 0. 136. 0). as seen from the list below. (25. ½¿ ½¿ ½ Ê 123. 11. (24. If the hour hand is exactly on a second mark then the second hand will always be on the 12. (33. 2). Burn A from both ends. 6). (20. 6). 10). (26. 1. producing pieces C and D. 0). 0). 1. (30. (29. 2). 0). 0). 128. 10). 10). 0). (36. (22. 0) (b) 40 moves. 10). 158. 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. (20. (23. (21. (29. 1). 0. 6). (30. 126. 5. 3. (19. 10). (36. 11. (17. 0). 10). the probability of a right triangle is ½ ¾½ or . light the other end of D. 0. 11. Cut B at that corresponding point. 124. 0). This is the shortest time interval that can be measured. 0). 178. 6. 1. 0. 5). (30. 0). 0). 134. 0. 6. (22. (25. (34. 6).
The series can be expressed by a simple continuous function:
To get Ü. There are six ways as shown below. East playing the 9. I. Then ¾Ô ¿Ò ½ and Ô¾ ¿Ñ · ½. HESS
¾. so that ¾Ô · Ô¾ is always a
¿ we have Ô
· ½ or
M6. Ü¿ ¾Ü
¼. North wins the sixth trick with the 10 of diamonds. The next four tricks are won by South with the 5432 of clubs.
Ë ¹ À ½¼ Â Ë À ¿¾ ¿¾ ¿¾ ¿¾
Ë ÃÉÂ½¼ À ¹ ¹ ÃÉ
Ë ¹ À ÃÉÂ½¼ ÃÉÂ ½¼
M4. ¿½ ½¾¼ ÐÒ ¾
M5. North and East discard all their diamonds on these tricks. The ﬁrst ﬁve tricks are alternate spade and heart ruffs by North and West.68
R. we take the limit of
½¾¼´¾Ò ½µ Ò ÓÖ Ò ´Òµ as Ò goes to 0. For
multiple of 3 for Ô
½. The ﬁnal two tricks are won by South’s long diamonds. The only real root is Ü ¿ we get the prime 17. ¿. The hand below does the job if the play goes as follows. North next leads a heart which West trumps with the queen. with East underrufﬁng North each time.
P3S. The triangle they form will always include a vertex with an angle of 36 degrees as shown below. the planks can be marginally longer than Ä
M9.PUZZLES FROM AROUND THE WORLD
M8. P2S. (b) Weights of 2. 2P.
Place that vertex at the origin and suppose the other two vertices of the triangle are at grid points ´ µ and ´ µ. If these lines are interpreted as theÔ diagonals of the planks then ½ ¿ ¿ ¼ feet. 3. Assume that three vertices do fall on grid points. Corresponding to each of these types the
· ´ ¾ · ¾µ´ ¾ · ¾µ
´ · µ¾ ¾ ´ · ¾ µ´ ¾ · ¾ µ
. 4S. P. The longest musical composition without three consecutive repetitions of any sequence are AABABAABABAABAAB and its reverse. 6. M11. 2PS. 3S. 9. PS. M10. and 27 do the job. 5S. In the ﬁgure below the oblique lines each have a length of ´¾¼ ¾µ ¿ ¾ ¼ feet. This contradiction proves that three vertices of a regular pentagon cannot lie on grid points. Note that a 1-pound package is the only one that wont balance or lift a 2-pound weight. Then
Ó×´¿ Æ µ
But is irrational and therefore cannot be the ratio of two integers. M12. 2S. (a) Weights of 1. 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.
R. 2S. PS. 4S. 5S. P4S.789. 2P2S.123. 6S. Each number is the length of the shorter side of the rectangle.
. A nine-rectangle solution is shown below. P2S. 1. I. 2PS. The number of combinations is
¾¡ ¡ · ¡ · ¿¡ ¡ · · · ¾¡½ ¾¡¼ ¾¡¾¡¾ ¾¡¾ ¾¡¾ ¾¡¾¡¾ ¿ ¾ ½ ¼
¾¡ ¿¡ · · ¾¡¿ ¾¡¾ ½
M13. 2P. M14. 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. and 3P. P. P3S. Shown below.465. 3S.
But if he has 3 then he reasons that person B sees (5. This 1 foot is overcome by the snail’s normal progress during that minute.
M17. Since A doesn’t know his number. 7 or 11 on your forehead.
. That happens for Ø ½ ¼ ¾ ¢ ½¼ minutes. M18. Since B doesn’t know his number. 3) would immediately know he had 5. ¿ Ù M19.PUZZLES FROM AROUND THE WORLD
M16. 5) and concludes that he must have 3 or 7. An eight-triangle solution is shown below. 7) and concludes he has 3 or 5. Since B doesn’t draw this conclusion you know you don’t have a 7. (b) When the snail has traversed 99 % of the band. The snail eventually arrives at the ½ ½ ¿ far end when ½· ½ ¾ · ¿ · · Ø ½¼¼. or Ù ¾¾ ¢ ½¼ ¾ Ø . 3) and would know his number is 7. B can eliminate 3 because anyone seeing (3. The farthest is at point P one quarter of the way along the diagonal on the 1 ¢ 1 face from B. (a) During the Òth minute the band is ½¼¼Ò long and the snail travels a fraction of the band equal to ½ ½¼¼Ò. (a) Unfold the box and examine shortest path distances from A to various points. 3) and would know his number is 5 or 3. A would conclude that he has a 7. This happens ½ · ½ · ¡¡¡· ½ when ½ · ¾ . then B would conclude he has a 5. (3) If you have a 7 then person B with a 5 sees (7. Since A doesn’t draw this conclusion you know you don’t have a 5. But if he has 3 then he reasons that person A with 7 sees (7. Therefore you have 11 on your forehead. (2) If you have a 5. (1) The three possibilities are that you have 5. the 100-foot stretch causes the far end to move 1 foot farther from the snail. then person A with a 7 sees (5.
8). 12. 13. (15. 11. 9). 14. 12. (15. 13. 7). 12. 12. 13.0119 . (15. There are 7 Good Eggs. I. Shown below. (15. The scores for the Good Eggs are (15. 12. 12. The points are ´ ¿ ½µ ¾ of the way along each diagonal for a distance of 3. 11. 11. 6). HESS
(b) The two points are along the diagonal near ÔP and its opposite point on the ½ ¢ ½ face that includes A. 9) and (14. 13.
Solutions to Hard Problems
H1. 14. 14. There are 5 subjects with 15 marks possible in each.72
R. M20. 10. The number in each triangle indicates its area relative to the other triangles in the ﬁgure.
. 10. 14. Humpty Dumpty got a 10 in arithmetic. 8). 14. (15. 11. 10).
Relative to the Ô center point of the ball. Three equally spaced punches in a straight line do the job if the square of the spacing is irrational. H4. To maximize the area is to ﬁnd the Ò that maximizes
The area is largest for Ò
½ ¼ µ´½ ¼ Òµ¾ ´ Òµ Ò ½ . 0. where . with 159.PUZZLES FROM AROUND THE WORLD
H2.300.1968 Ú ¾
H3. BCB rolls along the plane. the point P will have Ô coordinates È Ö × Ò ¾ Ö´ Ó× ½µ ¾ Ö´ Ó× · ½µ ¾. For a full rotation of the ball on the circle the point P will execute ¾ rotations about the axis. 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. Its area is Ò× ¾. Imagine a regular polygon of Ò sides after Sir George’s trip.
. as shown below. Consider the ball rotating about the axis AC without slipping. 1). Then the
Û ¾Ö · ¾ ·
¾Ö Ø Ò · Û ¾Ö´ · µ
These equations can be iterated to produce
ft. The distance from B to the distance BC. height of the building. For Ö ½È ´ ¼ ¿ ¾ ¼ ¾ ½¼ ½ ¼ ¼ ¼ ½ µ its initial position was (0. 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. Let Û width of the building. Ö earths radius. TheÔ cone. If Sir George travels at speed Ú then × Ú ´½ ¼ Òµ Ò where 1440 is the number of minutes in a day. and stretch in the band due to the building.
H5. Point P will return to the top again as Ô well.
The solutions are shown below. (a)
¼Æ ¼Æ ½ ¼Æ ¾ ¼Æ ÓÖ ¿¼¼Æ.
¼Æ ¼Æ ÓÖ½ ¼Æ
· · ¼Æ
¼Æ ¼Æ ½ ¼Æ ¾ ¼Æ. HESS
H6. ÓÖ ¿¼¼Æ
which contains ½ .
. Pick and as any digits other than 0. The perimeter of
¡ ¡ ¡ Ô ¡ ¡ ¡ where there are Ò square roots. The area of such a shape is ½·´ µÖ¾ . or ½¾¼Æ. (a) (b)
¿ ½ ¾¾ ¿ ½ ½ ¾¾ ´Òµ
´Òµ ´Ñµ can be made arbitrarily close to 0.
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. Clearly
H7.PUZZLES FROM AROUND THE WORLD
· ¼Æ ¼Æ. Its form will take on the shape of four quarter circles of radius Ö at the corners as shown in the ﬁgure. Ó×´¾ µ · Ó×´¾ µ ¾ Ó× · ¾ Ó×´ µ Ó×´
µ. ´ µ can be put in the interval ¼ ¾ ¼ . For the right . (a) Imagine an expanding soap bubble in the unit square.
9973967 . I. A view of the pyramid looking down from S perpendicular to the plane ABCD will look like the ﬁgure below.
The point S may be raised above the plane to give unit altitudes if the inscribing circle has radius ½. Note that the ratio È is 0.76
R. There are 7 circles on each end with 105 sets of 3 circles in the middle. H9. The smallest rectangle found to date containing 329 circles has 13 circles on each side of 101 sets of 3 and measures 2 ¢ 163. Raise S and T appropriately above the
. They deﬁne the four tangent lines drawn to produce ABCD as shown. The ratio È takes on a maximum when Ö ¼¾ ¼ . HESS
the shape Ô is È
½ ´¾ ·
¾ Ö · Ö.25 for the entire square or a circle inscribed in µ
(b) The corresponding problem of maximizing the ratio of volume to surface area within the unit cube remains unsolved. The circles need to be packed as shown below. The maximum ratio È Ö ¼¾ ¼ .
H10. For Ò ½ there is just enough room for 329 circles. Each has a radius ½. Now consider two such circles of different sizes as shown.
2801 1. and let Õ ½ Ô. My uncle is expected to take the longest time to dress on Saturday and on Friday he is least likely to get a pair from the ﬁrst three socks chosen.7805 0 . which leads to Ô ¼ ½½ ¿ ¾ . Ô ´½ ½ Õ µ È ´¼ ¼µ . È ´¼ ¼µ È ´ ¼ ¿¼µ when Ô¿ Ô¾ ¾Ô ½.
H11.2069 1.3074 1.2270 1.8286 0. Let Ô the probability of the server winning a point. È ´ ¼ ¿¼µ Ô · ÕÈ ´ ¼ ¼µ.
Monday Tuesday Wednesday Thursday Friday Saturday
Average Selections Required 1.
Similarly one derives
Ô · Ô¾Ô·ÕÕ¾ .2511 1. At Ò games to Ò. This leads to ¾ È ´ ¼ ¼µ Ô¾Ô ·Õ¾ . È ´¼ ¼µ È ´¿¼ ½ µ when Ô ¾ Ô ¿.7849
. È ´¿¼ ¼µ ÔÈ ´ ¼ ¼µ. A computer program was written to calculate the probabilities shown in the table below.8163 0. Ô Õ At a score of Ò games to Ò (Ò ¼ to 5). Clearly ST is not perpendicular to the plane.
Pairing probability 0.8027 0. È ´ ¼ ¼µ ÔÈ ´ ¼ ¿¼µ · ÕÈ ´¿¼ ¼µ. Let È ´ µ be the probability of the server winning the game when the server has points and the receiver has points.PUZZLES FROM AROUND THE WORLD
plane so that all altitudes are unit.
È ´ ¼ ¿¼µ
Õ È ´ ¼ ½ µ Ô · ÔÕ · ÔÔ ¾ ·Õ ¾ . ¾ Õ ´ÔÕ ·½µ È ´¿¼ ½ µ Ô¾´½ · Õµ · Ô Ô ¾ ·Õ ¾ . which leads to Ô ¼ ½ ¿¿ ¼ .7892 0.
641. Using a multiprecision continued fraction program. · ½.
prime. 8. We look for cases where (1) the oldest child is under 20. (a) Length ½ ¿ ¾½ ½ ¼ as shown. 16). we get (a) (b) H16.093.042. I.546. (b) Length ¾ ¼¾½½¾ ¿ as shown. (c) 17. HESS
H13. 6.512. ¿ ½¾¼.321 is the ﬁrst of 5 in a row.
¾½ ¾ ½¼¿½ ½¼ ½ ½½ ½ ¾½ ¾¼.
´ ¾ · ½µ ¾
prime. (a) 45 years. which must have been guessed by Jones two years ago. (d) Yes.
´ ¾ ½µ ¾.
´ ¾ · ½µ ¾ ¾½¿ ¾½. which must have been guessed by Smith.444 years.
¾ ½µ ¾. There is only one set of ages that accomplishies this: (5. 15).613. a=10. and (3) there is a product and sum that give rise to ambiguity for the ages both this year and two years ago.78
R. ¼ ¼¼ ½ ¼. H14. The product and sum could be achieved by (4. 12).
¼ ¼½¼ .
¾¾ ¾ ¿
¿ ½½ ½ ¾¼.
Solutions occur for (b) ¾ ½. . (2) the younger two children have different ages. (a)
¾ ¢ ½¼¾ ¼¾ ¢ ½¼½
. ´ prime.
Ô½ · ½
£ Ô¾ · ¿
£ Ô¿ ·
¾ ¾ ¾ ¢ ½¼¾ ½ ½¼¿ ¿¿ ¢ ½¼½
. 7. (b) 605 years. (c) Unknown. · ½. The product and sum two years ago could be achieved by (2. .
The ﬁrst case is not possible because Ê ¾Ö ¼. H20. (1) Deﬁne
Á ´Ò Üµ
Òµ ´ ÒµÜ
¾ Á ´Ò ½µ ¾Ò · ¿
. (b) Channels with radii between Ê ¾Ö and Ê are possible.
H19. The ﬁgure below shows a cross-section of the torus. From the descriptions of the ﬁrst two students they must have dug channels of type (b).
. which is the intersection of the plane shown by the slanting line and the torus.
Let Ö and Ê dug on its surface.PUZZLES FROM AROUND THE WORLD
Six in a row is impossible because three consecutive even special years cannot occur. the other case gives Ê Ö ¿¼ and Ö ¾ . One case gives Ê Ö ¾ and Ö ¿¼.
H18. (c) A less well-known third type of circular channel with radius Ê Ö is possible. There are three classes of channels that can be
(a) It is clear that many channels with radius Ö are possible. The venture should be undertaken since the volume in cubic spandrals can be determined using calculus to be
´½ ¿¿½µ¾ ¿´ ¼µ¿
It is noteworthy that the volume doesn’t depend on the polar or equilateral radii of Alpha Lyra IV. The third and fourth students must have explored channels of type (a) and (c) in some order. Thus the second case applies and ¾ Ê ½½¼ .
As trarily large it still encloses the same poles as . and by Cauchy’s Theorem we have
ÞÕ´Þ µ ´Ò·½µÞ
Þ ¾ Õ´Þ µ ´Ò·½µÞ
¾ ´¾Ò · ¿ µ
ÞÕ´Þ µ ´½ · Þ µÒ·½
To evaluate ´Òµ. Thus
¼ like that in ¼ gets arbi-
ÞÕ´Þ µ ´½ · Þ µÒ·½ ÞÕ´Þ µ ½ · Ò·½ ´½ · Þ µ ¾ Ê
ÞÕ´Þ µ ´½ · Þ µÒ·½
. which still avoids enclosing the poles of Õ´Þ µ.
(3) The summation over
can be carried out explicitly to give
¾ ½ ¾Ò· · ¿ ¾
Þ ½ ´½ · Þ µ Þ ´½ · Þ µÒ·½
(4) Deﬁne Õ´Þ µ ½ ½ ´½ · Þ µ Þ . I. HESS
(2) Since ´ ÒµÞ is an entire function in the complex plane. The only pole of Õ´Þ µ ´Ò·½µÞ within is a double pole at Þ ¼. it follows from Cauchy’s Theorem that
where must enclose Þ
½ and will be taken to be Þ
´Ò·½µÞ ½ ½ ´½ · Þ µ Þ ¾
½ ¾ ¼
´ ÒµÞ ´½ · Þ µ ·½
R. Thus Þ ¾ Õ´Þ µ ´Ò·½µÞ is regular within . we expand and deform to a contour the ﬁgure.
6406814 3. Then Ý approximately equals ½·ÐÒ´ µ .8505482 39.2916783 3. not including its pole at Þ ¼.0888430 2. (5) On Þ Ê for sufﬁciently large Ê. Õ´Þ µ is bounded and the integral on the path tends to zero as Ê tends to inﬁnity.4614892 13.8221528
7. The table below gives numerical values of the ﬁrst few.0262969 3.5012690 3. The resulting equations in Ü and Ý are ´½ × Ò Ý Ý Ý ÓØ Ýµ and Ü ÐÒ´Ý × Ò Ýµ. which reduces to ´½ · Ü · Ý µ´Ü Ý µ ´ Ó×´Ò · ½µÝ × Ò´Ò · ½µÝ µ
where ´Ü¾ · Ý¾ µ ´Ò·½µÜ .5432385 32. goes to zero as Ò goes to inﬁnity.
¾ × ´Ý · µ Ü Ý
. where Þ at ½ · Þ Ü · Ý.6745053 3. For larger deﬁne ´¾ ·½ ¾µ . Since the equation in Ý is even. we can look for poles in the upper half-plane only and reﬂect each one into the lower half-plane. Thus ´Òµ is the sum of integrals on small circle paths about the poles of Õ´Þ µ. A pole of Õ´Þ µ occurs Þ . Combining the poles in the upper and lower half-planes and some rearrangement ﬁnally produces
ÒÜ × Ò´ÒÝ
(a) Since Ü
Ü¾ · Ý ¾ Ü¾ · Ü · Ý¾ Ø Ò´Ý · µ Ý ÒÜ ¼ for all .PUZZLES FROM AROUND THE WORLD
where is the counterclockwise path about the Ø pole of Õ´Þ µ.4473849 of
´Òµ from pole is the limit as Þ approaches Þ ´Þ Þ µÕ´Þ µ ´½ · Þ µÒ·½.8790560 20.2238350 26.
1 2 3 4 5 6 7 (6) The contribution to 2.1512074 45. Call the Ø such pole in the upper half-plane Þ Ü · Ý .
All rights reserved. Reprinted with permission. May–June 1990. Schoenberg.” in MIT’s Technology Review. Vol. page 178. From Nobuyuki Yoshigahara (private communication). From David Singmaster (private communication). Reprinted with permission of the Canadian Mathematical Society. All rights reserved. “Olympiad Corner. 1983.82
R.” in MIT’s Technology Review. E12. Reprinted with permission. 1985. ´Òµ is dominated by the ﬁrst pole and its reﬂected pole. From Nobuyuki Yoshigahara (private communication). It is at Ñ ¼¼. MIT 59. Unknown. Submitted by Lawrence Kells. MIT 55. E13. All rights reserved. E10. 1992. 3. Einhorn and I. “Puzzle Corner.
. E5. “Puzzle Corner. E4. Reprinted with permission. p. J. giving ´½¼¼¼µ
¢ ½¼ ¼
(c) From observing the behavior of × Ò´ÒÝ½ ½ µ one can determine the smallest Ñ where the magnitude of ´Ñµ is less than the magnitude of ´Ñ · ½µ. edited by Allan Gottlieb. 8. private communication). Problem 7 from the 1980 Leningrad High School Olympiad. p. From Yoshiyuki Kotani (private communication). Submitted by S. Problem 3a. All rights reserved. E15. From David Singmaster (private communication). E2. Vol. From Nobuyuki Yoshigahara (private communication). M5. October 1992. From Nobuyuki Yoshigahara (private communication). 1990.” in MIT’s Technology Review. Modiﬁcation of Problem 1. 1987. Unknown. E3. p.” Pi Mu Epsilon Journal. Reprinted with permission. All rights reserved. E8. 302. E1. (a) and (b) original problems. “Puzzle Section.
¢ ½¼ ¾
½¾¾ ¢ ½¼ ¾
All problems written by the author unless otherwise indicated. “Puzzle Corner. No. I. From David Singmaster (attributed to Roger Penrose. M4. edited by Alan Gottlieb. M2. 10. From Dieter Gebhardt (private communication). Unknown. From Nobuyuki Yoshigahara (private communication).” Crux Mathematicorum. M3. E16. Problem 3. No. Unknown. E11. Thus ´½¼¼¼µ is very nearly ½ ½¼¼¼Ü½ × Ò´½¼¼¼Ý½ ½ µ. J. HESS
(b) For Ò ½¼. 9. E14. edited by Alan Gottlieb. October 1987. MIT 55. Problem 1. Unknown. p. E7. M6.
. 10. From Karl Scherer (private communication). 142. MIT 53. Problem 2. H6. (a) Unknown. 13. A special case of a problem by P. H5. edited by Allan Gottlieb. p. p. All rights reserved. From Robert T. p. 10. Submitted by Nobuyuki Yoshigahara and J. 1986. All rights reserved. No. 8. M20. Crux Mathematicorum. January 1989. 2. H1. Vol. H9. (b) original problem. All rights reserved. Pi Mu Epsilon Journal. 1994. No. MIT 47. Problem 1289. M15. Problem 1995. 1986. All rights reserved. Submitted by Cops of Ottawa. H2. edited by Allan Gottlieb. Vol. MIT 55. Problem 4. Problem 3. Reprinted with permission. M16. Purdy. Vol. First published as Problem 3. Submitted by David Singmaster. p. October 1988. Wainwright (private communication). p. Reprinted with permission of the Canadian Mathematical Society.” in MIT’s Technology Review. Vol. 1987. February–March 1997. 1987. p. 13. M19. Crux Mathematicorum. From Robert T. No. “Puzzle Corner. Reprinted with permission. p. 155.” in MIT’s Technology Review. 86. M18. Vol. 12. 10. edited by Allan Gottlieb. MIT 53. 180. p. and then as Department Problem 860. Crux Mathematicorum. p. (a) From Yoshiyuki Kotani (private communication). Submitted by Carl Sutter. 1989. Problem 1194. Problem 944.” in MIT’s Technology Review. Reprinted with permission of the Canadian Mathematical Society. 285. Wainwright (private communication). p. p. 10. H8. 20.” Pi Mu Epsilon Journal. All rights reserved. All rights reserved. H10. 290. Reprinted with permission. H11. No. Part (d): Problem 7. “Puzzle Corner. 1988. 5. 8. H3. Submitted by Sydney Kravitz. 526. H12. May–June 1994. Crux Mathematicorum. 1984. 7. Part (a): Problem 870. 12. All rights reserved. Vol. Submitted by Bob High. All rights reserved.PUZZLES FROM AROUND THE WORLD
M7. No.” MIT’s Technology Review. “Puzzle Section. Submitted by Jerzy Bednarczuk. No. Reprinted with permission. 1995. 9. Reprinted with permission of the Canadian Mathematical Society. All rights reserved. From Karl Scherer (private communication). 1997. M17. Crux Mathematicorum. Unknown. 1988. All rights reserved. All rights reserved. Submitted by Joe Shipman. From Robert T. edited by Alan Gottlieb. 3. Reprinted with permission. Akiyama. Reprinted with permission. Crux Mathematicorum. “Puzzle Corner. Wainwright (private communication). M14. Erdos and G. 1994. Reprinted with permission of the Canadian Mathematical Society. From Yoshiyuki Kotani (private communication). Vol. “Puzzle Corner. Part (b): Problem 1225. Reprinted with permission of the Canadian Mathematical Society. M11. Vol. 282. No.
Hess from Albert Latter (private communication). Submitted by Richard I. 12. p. From Robert T.
. H18.84 H14. H20. All rights H17. Crux Mathematicorum. 1986. 118. I. Part (a): Problem 1190. Reprinted with permission of the Canadian Mathematical Society. HESS
1987. 9. Unknown. H19. 13. Unknown. No. Vol. All rights reserved. Vol. 242. Crux Mathematicorum. 4. Reprinted with permission of the Canadian Mathematical Society. Wainwright (private communication). Problem 1231. No. p. reserved.
but the one that grabbed us the most was the Hexiamond. Remove pieces ½. ﬁve had reﬂexive symmetry. and then by twelve more.O’Beirne’s Hexiamond
Richard K. 3. The next solution in my collection is Figure 7 by Mike Guy (March 1960). and . When I began to write this article. 5. He showed us many remarkable puzzles. 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. 7. But it does appear in his column in the New Scientist. If we count reﬂections as different. and several copies had to be manufactured. One of these is the regular hexagon. and in January 1960 he found the solution shown in Figure 6. but I haven’t found one yet. 5. O’Beirne thought that the result would be more pleasing if the Hexagon were in the center. Rearrange pieces 1. Maybe the only other places where it has appeared in print are Berlekamp et al. and 8 from Figure 3 and replace them in a different way. ¿. Or try using 0. . It must have been in 1959 that O’Beirne noticed that. since everyone wanted to try it at once. 3. Martin  uses O’Beirne’s names for the shapes but does not distinguish between reﬂections so his problems only involve 12 shapes. while seven did not. 6. I was sure that I would ﬁnd references to the Hexiamond in Martin Gardner’s column. 1. It soon became necessary to devise a classiﬁcation scheme. No one went to bed for about 48 hours. since it was not easy to decide whether a solution was new or not.  and the not very accessible reference Guy . 6. ½.
. Nor is it mentioned in O’Beirne’s own book . Guy
Tom O’Beirne is not as celebrated a puzzler as he deserves to be. and in Figure 7. In the interim he had found solutions with the Hexagon in two other of its seven possible positions (Figures 4 and 5). It was soon after this that O’Beirne visited the Guy family in London. particularly on this side of the Atlantic. among the shapes that can be formed by adjoining six equilateral triangles. which can be surrounded by six more hexagons. giving a ﬁgure (Figure 2) having the same total area as the 19 shapes. then there are 19 shapes (Figure 1). We became adept at ﬁnding new solutions based on the old. and ¿. and in Figure 5.
with its handle on the left. . 7. the multiples of 4 have rotational symmetry through 180 degrees. We don’t know how O’Beirne decided whether to give a piece a number less than or greater than 10. which appears to be turning to the left. and the Sphinx. to the Signpost. Negative (bar sinister?) labels are given to the Bar. to the Yacht. Pieces 4 and 8 have only rotational symmetry and exist in enantiomorphous pairs. together with those that have rotational symmetry only.86
R. K. ¿.
O’Beirne had already suggested a numbering of the pieces from 1 to 19. as do pieces 1. to the Crook. Notice that 0 has the greatest symmetry. which have no symmetry. By subtracting 10 from the Hexagon. 3. and Lobster each have a single symmetry of reﬂection. exist in enantiomorphous pairs. Pieces with odd-numbered labels have no symmetry and. a parallelogram drawn in the opposite way from the usual textbook fashion. The Butterﬂy has the same symmetries as a rectangle: Its position is described by that of its body. The symmetries of the Hexagon are those of the dihedral group D6. The Chevron. GUY
Figure 1. and each is positioned by its apex. and the other even numbers have reﬂective symmetries. with its hook on the left. 5. but our mnemonics are as follows. . . ½. and subtracting each of the labels 11 to 19 from 20 we arrive at the labeling in Figure 1. to the Snake. Crown. sailing to the left. whose head is on the left. The nineteen Hexiamond pieces.
. . with its pointer on the left. . to the Hook. and 9.
Figure 2. and Lobster. 1959). giving the positions of the Chevron. but gives enough detail to enable comparisons
Figure 7. O’Beirne’s ﬁrst solution (November 30.
¿ Ã¾ Ð
Figure 3. Butterﬂy.
We classify solutions with a string of ﬁve labels. Crown. This doesn’t completely describe the solution. Hexiamond board.
Figure 6. Hexagon.
reﬂect the board if necessary to bring the Hexagon into the right half. 8. it is possible to place the Chevron in such positions. depending on how far it is from the center. and t are omitted.
Figure 9. reﬂect the board if necessary to bring the Butterﬂy into the right half. If all three are symmetrically placed. and are given only 0. (The exact ﬁgure is 89. Then append a subscript ¼ ½ Ü indicating its position on the clock. This will be a capital letter. We won’t count a solution as different if it is just a rotation or a reﬂection. together with an even subscript. The subscripts a. l. 6. The subscripts on C are about ½ ¾ more than their clock hour. If both Chevron and Hexagon are symmetrically placed. e.5%. of the Chevron from Figure 8.” the edge that bisects it. More than three-quarters of the positions found so far have the Chevron in position h. piece number 2. since placing the Butterﬂy there does not allow a solution to be completed. shown in Figure 10. K. F only even ones. 2 or 4 for a subscript. Coding the Hexagon. 2.) Next use Figure 9 to describe the position of the Hexagon. is that of its “body. 0. is unique and requires no subscript. or Ü the side of the board it is nearest to. reﬂect if
. This is indicated by a lowercase letter. the center. GUY
to be made quickly. piece number ¾. reﬂect the board left to right in order to bring it onto the right half. Note that the letters a and n are missing. if the Chevron is on the left-hand side of the board. is pointing upward. 4.
R. f and m are equidistant from opposite sides of the board. as in the ﬁrst solution in Figure 14. A to G. and G. A. Coding the Chevron.
The position of the Butterﬂy. s. B. so ﬁrst rotate the board so that the Chevron. Then. but it’s clear that they can’t occur in a solution. D. Read off the lowercase letter at the apex. If the Chevron is central (positions b to g in Figure 8). piece number 0. A and E have only odd subscripts.
Figure 11.O’BEIRNE’S HEXIAMOND
Figure 9. R are omitted. Coding the Butterﬂy. in Figure 11.
The Lobster. these are in the same positions as the capital letters used for the Crown. The position of the Crown. Coding the Crown. c. 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. piece number 6. Positions s. is located by a lowercase letter in Figure 12. 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. Positions S and T are symmetrical and carry even subscripts. Continued: Coding the Hexagon’s position.
necessary to make the apex of the Crown point toward the right half of the board. M. A. F. and r don’t lead to legal solutions. Positions C. Except for a. t. piece number . Q. as they don’t allow legal solutions. and u are symmetrical and carry only even subscripts.
. g. is given by the capital letter at its apex.
Remember that the code doesn’t specify the solution completely. is about 50. for example. Coloring arguments don’t seem to lead to much restriction. There are already more than 4200 in the collection. and 1 legal positions
. but perhaps some reader will be more perspicacious. which will be deposited in the Strens Collection in the Library at the University of Calgary. We have found 247 of these cases (no fewer than twenty-six bit the dust during the writing of this article). there are respectively 21. although some of these turned out to be so. It isn’t very easy to give a good upper bound. 39. h.
Now that you know how to classify solutions. based on how rarely duplicates appear. note that O’Beirne’s ﬁrst solution (Figure 3) is of type Ô Â . GUY
R. There are at least ten solutions of class Ô ¼ ¼ . There are 508 essentially different relative positions for the Chevron and Hexagon that have not been proved to be impossible.
Figure 11. With the Chevron in positions g. K. and o. Continued: Coding the Crown’s position.000. (Can you ﬁnd a larger class?) How many solutions are there? A wild guess. Continued: Coding the Butterﬂy’s position.
In 1966 Bert Buckley.
for the Hexagon. the second by the present writer. the discerning solver will soon wish to specialize. As we go to press. Coding the position of the Lobster.O’BEIRNE’S HEXIAMOND
Figure 12. suggested looking for solutions on a machine. I didn’t think he would be successful. Figure 13 shows two of Bert Buckley’s machinemade solutions. For example.
. the ﬁrst found by John Conway and Mike Guy in 1963. With such a plethora of solutions. and solutions are known in all of these cases. then a graduate student at the University of Calgary. how symmetrical a solution can you get? Figure 14 shows two solutions. 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. each with as many as 11 of the 19 pieces symmetrically placed. but after a few months of intermittent CPU time on an IBM 1620 he found half a dozen or so solutions from which it was possible to deduce another ﬁfty by hand.
Conway showed that you can’t have more than 15.92
R. We have always drawn the hexiamond with one of these directions vertical.
Figure 14. You can
. It seems unlikely that anyone can beat 11. at angles of 60 degrees to one another. but you may prefer to put one of them horizontal. found by John Conway in 1964. each with 11 symmetrically placed pieces. rearrange pieces ¾¾ or ¿¿ and note that the hexagon formed by ¼½½ has the symmetries of a rectangle. in Figure 14. Some solutions come in large groups. Solutions found by machine by Bert Buckley. rearrange pieces ¾¿¿.
Notice that there are two different kinds of axes of symmetry. or ½¿ . To deduce the others from the single diagram shown. Figure 15 shows eight solutions. For example. GUY
½ ÒÜ ¼ Ó¿
Figure 13. K. All edges of the pieces lie in one of three different directions. pieces placed symmetrically with respect to the two kinds of axis. after which you can swap ¼¾ with ¿ . respectively 13.
Figure 16 shows a solution displaying three keystones that can be permuted to give other solutions.O’BEIRNE’S HEXIAMOND
Figure 15. and so on . In May 1996 Marc
½ Ô¼ Ü ×¾
¼ Ç¼ Ð
The most symmetrical solution?
rotate ¼¾ ´¿ µ or ¼¾ ´¿ µ and produce two “keystones. keystones. Figure 17. here ¿ . Figure 18 has the Hexagon in position G and the other pieces forming three congruent sets. 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.” which can be swapped. shows that keystones ´ µ can occur internally and don’t have to ﬁt into a corner. found by Mike Guy. Readers will no doubt discover other curiosities and ﬁnd their own favorites.
has an inter-
. nal keystone. Figure 19 is one of at least eight examples that have six pieces meeting at a point.
there are no fewer than 40 relative positions of the Chevron and Hexagon that determine such a unique solution:
¼ Ñ Ø
¿ Ô Ô Ø ½ Ø
¼ Ô Õ ¼ Õ Ü Ö Ø ½ Ø Ø ¼ Ø
How long will it take you to ﬁnd them all without peeking at Marc’s database ? After writing a ﬁrst draft of this article. On the contrary. are
130 195 533 193
Figure 18.460 584
637 914 749 498 1238 and.
Six pieces meet at a point.447 F 5727 G 1914 I don’t think that this need take the fun out of one of the best two-dimensional puzzles ever invented. president of Kadon Enterprises and discovered that she markets a set of “iamonds”
. for those who prefer their puzzles to have just one answer. I wrote to Kate Jones.
R. classiﬁed according to the position of the Chevron.489 B 15. Õ
Paulhus wrote a program that used only a few days of computer time to ﬁnd all the solutions.717 C 6675 D 7549 E 11. K.518
A 75. Their numbers. by the Hexagon’s position
Ø Ù Total 264 1094 124.
Kate Jones’ Iamond Ring. Winning Ways for Your Mathematical Plays. discovered by Kate herself. pp. K. diamonds. Berlekamp. if you want to use them to make O’Beirne’s hexiamond puzzle. are given in the following table.
 E. R. 1982. so.
. H. J.O’BEIRNE’S HEXIAMOND
under the trademark “Iamond Ring. . doesn’t do justice Figure 20. 787–788.
Figure 20. Academic Press. you must get two sets. London. 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. Guy.” For this purpose the iamonds are not counted as different if they are reﬂections of one another. and R. The numbers of iamonds. triamonds. Conway. in which each x-iamond is labeled Ü ½ Ü to the very appealing colored pieces of the professionally produced puzzle.
R. 259 (61-11-02) pp. 9 (1960) pp. Assoc.  George E. Malayan Math. Nabla (Bull. K. Martin. 1991. Guy. pp. 144–153. 316–317. II pp. especially pp. Polyominoes: A Guide to Puzzles and Problems in Tiling. of America Spectrum Series. GUY
 Richard K. 1965. see Puzzles and Paradoxes No.  Thomas H. 168–170. 44: Pentominoes and hexiamonds. O’Beirne. 97–106. Math. A database for O’Beirne’s Hexiamond. 104–106 and 152–153. submitted. New Scientist. Some mathematical recreations I. Soc. Oxford University Press. Puzzles and Paradoxes.  Marc Paulhus.).
In 1742. The book was called Sei Shonagon Chie-no-ita (The ingenious pieces of Sei Shonagon).Japanese Tangram: The Sei Shonagon Pieces
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.
(a) Tangram pieces. 16 centimeters wide and 11 centimeters long. so I sent a report about the Japanese tangram to him. I received a letter from Kobon Fujimura. The introduction bears the pseudonym Ganreiken.
Sei Shonagon Chie-no-ita is a 32-page book. In fact. but nobody knows its real author. a court lady of the late 10th and early 11th centuries. The tangram came to Japan from China in the early 19th century.
(b) The Sei Shonagon pieces. I heard that Martin Gardner had been planning to write about tangrams. Sei Shonagon. was one of the most clever women in Japan. There are 42 patterns with answers. a little book about Japanese seven-piece puzzles was published. a famous puzzlist in Japan. and the Japanese edition of Qiqiaotu Hebi (The Collected Volume of Patterns of SevenPiece Puzzles) (1813) was published in 1839. but their
shapes are inaccurate. The puzzle was introduced to the world by Martin Gardner in his book Time Travel ½ , and he elaborated as follows. “Shigeo Takagi, a Tokyo magician, was kind enough to send me a photocopy of this rare book. Unlike the Chinese tans, the Shonagon pieces will form a square in two different ways. Can you ﬁnd the second pattern? The pieces also will make a square with a central square hole in the same orientation. With the Chinese tans it is not possible to put a square hole anywhere inside a large square.”
A square with a center hole
I know of two other books that are collections of patterns of the Sei Shonagon pieces. In about 1780, Takahiro Nakada wrote a manuscript entitled “Narabemono 110 (110 Patterns of an Arrangement Pattern),” and Edo Chie-kata (Ingenious Patterns in Edo) was published in 1837. In addition, I possess a sheet with wood-block printing on which we can see patterns of the Sei Shonagon pieces, but its author and date of publication are unknown.
Some patterns formed from the Sei Shonagon pieces
½1988, W.H. Freeman and Co., NY
How a Tangram Cat Happily Turns into the Pink Panther
Do you want to create, or better generate, a new two-dimensional puzzle? Nothing could be easier! Just take one of the many well-known puzzles of this type and submit it to a geometrical transformation. The result will be a new puzzle, and — if you do it the right way — a nice one. As an example, let us take the good old Tangram and try a very simple transformation, a linear extension (Figure 1).
Figure 1. Tangram submitted to a linear extension. The result is a set of pieces for a new puzzle, a Tangram derivative called Trigo-Tangram.
While the tangram pieces can be Ô arranged in an orthogonal grid, the pieces of our new puzzle, due to ¿ as extension factor, ﬁt into a grid of equilateral triangles, a trigonal grid. That’s why I gave it the name TrigoTangram. Puzzles generated by geometrical transformations I call derivatives ; our new puzzle is a Tangram derivative, one of an inﬁnite number. Depending on the complexity of the transformation, certain properties of the original puzzle are preserved in the derivative, others not. Tangram properties preserved in Trigo-Tangram are the linearity of the sides, the convexity of the pieces, and the side length and area ratios. Not preserved, for example,
Figure 2. Suitable problems for a puzzle derivative can be found by transforming problems of the original puzzle. For Trigo-Tangram, two problems can be generated by extending this original Tangram problem in two different directions. The transformation of the solution can be a solution for the derivative (right image) or not (upper image).
are the angles and the congruence of both the two big and the two small triangles. For our derivatives, we do not only need the pieces, we need problems as well. Problems? No problem at all! We just take some ﬁgures for the original puzzle and submit them to the same transformation as the puzzle itself. The images of the ﬁgures will be suitable problems for the derivative. In Figure 2 this is done for a Tangram problem and its solution. As you can see, the corresponding linear extension can be performed in two different directions, which gives us two different problems. And you can see another important fact: The transformation of the solution does not always render a solution for the derivative. Figure 3 shows some problems for Trigo-Tangram. So, take a piece of cardboard, copy and cut out the pieces, and enjoy this new puzzle. Here is another way to have fun with a puzzle and its derivative: Take the original Tangram, lay out a ﬁgure, and try to make a corresponding ﬁgure with the pieces of Trigo-Tangram. In many cases, you get funny results. Figure 4 shows how a Tangram bird stretches its neck and how a Tangram cat turns into the Pink Panther.
TANGRAM CAT HAPPILY TURNS INTO PINK PANTHER
Figure 3. Some problems for the Trigo-Tangram puzzle.
Transforming two-dimensional puzzles provides endless fascination. Just look what happens if you rotate the Tangram in Figure 1 by ¼ Æ and then perform the transformation. The square and parallelogram of the original are transformed into congruent rhombi, which makes the derivative less convenient for a puzzle. You obtain better results by applying linear extensions diagonally (Figure 5). Since in this case the congruence between the two large and two small Tangram triangles is preserved, the two puzzles
Figure 4. A Tangram bird stretches its neck, and a Tangram cat turns into the Pink Panther.
Figure 5. Transforming the Tangram by diagonal extensions leads to two derivatives which are more Tangram-like than the one shown in Figure 1.
generated are even more Tangram-like than the Trigo-Tangram. To create problems for these derivatives is left to the reader. One set of problems will serve for both of them. I hope I have stimulated your imagination about what can be done by transforming puzzles so you may start your own work. Take your favorite puzzle, and have fun generating individual derivatives for yourself and your friends!
Polly places precisely ﬁtted ﬂagstones around various plantings in her garden in order to suppress weeds. and her sarcasm rather put me on guard. the sides are divided in the ratio of 3 to 5. 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. His wife Polly is an avid gardener. but I should have known better. She presented me with the following problem. Frankly. She wrote back with profuse thanks for such an “elegant” solution.Polly’s Flagstones
I wish to report on some recent correspondence with my good friends the Gahns in Calcutta. is shown below. I was surprised that Polly would bother to write about such a trivial dissections problem. She and Paul have a bent for geometrical recreations. so they are always on the lookout for creative and original solutions to their various landscaping projects. a classic dissection of the square. She had one large stone that was perfectly square.
In order to meet Polly’s required dimensions. You may remember meeting Paul in The Puzzling World of Polyhedral Dissections. The simple solution to this problem which I then sent to her.
She immediately wrote back and suggested that it was a shame to cut up two beautiful large stones when one might do. 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. I came up with the scheme shown below. each one enclosed by a rectangular perimeter. The smaller and larger holes have dimensions that are irrational.104
S. how about a scheme whereby the stones could be rearranged to form any one of three different-sized rectangular openings. If you think that Polly was satisﬁed with this solution. pencil. which she proposed that we cut into four pieces. then you don’t know Polly very well. but after tinkering for a while with paper. and scissors. With the dimensions shown. COFFIN
Now she posed another problem. 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. It was then I realized that from the start she
. to allow for more ﬂexibility in landscaping. cut into four pieces. but this time to form a rectangular perimeter enclosing a rectangular planting space. Suppose we were to consider starting with a large rectangular stone. the medium-sized rectangular hole will have dimensions exactly one quarter those of the original stone. She decided that she preferred mostly rectangular rather than square planting spaces. This required a bit more reﬂection than the ﬁrst problem.
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. And for good measure.
are left for the reader to discover. For added recreation. so I wrote back and asked what made her so sure it was even possible. The same solutions are possible. Now what sort of scheme do you suppose Polly will come up with for that other stone. I decided to play into it. note that the pieces can also be arranged to form a solid parallelogram (two solutions). The actual arrangements of the pieces for the four different solutions. the square hole will be one foot square. and a different trapezoid (one solution). but for this example let it be one foot from A. An interesting variation is to let points A and F coincide. making one of the pieces triangular.POLLY’S FLAGSTONES
had just been setting me up. and use that distance from F to locate point G. by return mail arrived her solution. One of the rectangles has a pair of solutions — the others are unique. Draw EF. Locate the midpoint E of side BC. Sure enough. one square and three rectangular. Subtract length CE from length DF. The location of point F is arbitrary. but it does. except that one of the trapezoids becomes a triangle. the rectangular one?
. shown below.
For the purpose of this example. Bisect EG to locate H. let the original square stone ABCD be four feet on a side. and draw JHK perpendicular to EHF (most easily done by swinging arcs from E and G ). Don’t ask my why this works. a solid trapezoid (two solutions). With these dimensions.
As we loitered around the airport newsstand. Imperial Earth. Arthur C. based on Solomon Golomb’s presentation in 1953 to the Harvard Mathematics Club. It expired one month before the May 1957 issue. In late 1976. declares himself a “pentomino addict”. so I did not get to see the historic column introducing pentominoes (Figure 1) to a world audience. created an ever-widening ripple effect. focusing especially on the pentominoes. in his mostly autobiographical Ascent to Orbit.
That column was inspired by a 1954 article in the American Mathematical Monthly. a paperback
Figure 1. crediting Martin Gardner’s column as the source.Those Peripatetic Pentominoes
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. My favorite part of the magazine was the “Mathematical Games” column by Martin Gardner. Thwarted from including pentominoes in the movie 2001: A Space Odyssey. Golomb’s naming of the “polyomino” family of shapes and their popularization through Martin Gardner’s beloved column. In 1956 I received a gift subscription to Scientiﬁc American as an award for excellence in high school sciences. Clarke. Later editions of the book actually show a pentomino rectangle on the ﬂyleaf. The subscription expired after one year. and I was not able to renew it. a group of expatriates stationed in Iran took a weekend trip to Dubai. Clarke wrote pentominoes into his next science ﬁction book. as the game HAL and Bowman play (the ﬁlm shows them playing chess).
Well. I commissioned a local craftsman to make a set of pentominoes in inlaid ivory.
. and after some preliminary doubts we went for it.108
K. and the pieces turned out to be thick enough to be “solid” pentominoes. It was a thrill to ﬁnd mention of Martin Gardner in the back of the book. but idleness was not my style. and inevitably a dominotype game idea presented itself to me. and soon friends got involved. of course that was the way to go! A little research turned up the fact that “pentominoes” was
Figure 2. This magniﬁcent set invited play. A friend’s suggestion that we “make and sell” that game I had invented popped up just then. to be shared with friends. Super Quintillions. By fall of 1979 a wooden prototype was in hand. I bought the book and soon caught pentomino fever. Exploration of the pentominoes’ vast repertoire of tricks was a ﬁne way to spend expatriate time. JONES
rack with a copy of Imperial Earth caught my eye. Fast forward to December 1978. Back home I was in limbo. when the Iranian revolution precipitated the rapid evacuation of Americans. a specialty of the city of Shiraz. Playing ﬁrst with paper and then with cardboard was not enough. A longtime Clarke fan. having sold my graphics business and with no career plans for the future. My husband’s job took care of our needs.
The most important lesson was that one needed a product “line”. We had much to learn about marketing. And it was encouraging to us neophyte entrepreneurs when Games magazine reviewed Quintillions and included it on the “Games 100” list in 1980. and most of them sneak in some form of pentomino or polyomino entity among their other activities. And so Quintillions begat a large number of kindred puzzle sets. Martin Gardner. Fives quints quintillions! It was with great pride. Here (chronologically) are the many guises and offspring of the dozen shapes that entered the culture through the doorway Martin Gardner opened in 1957. and the product names are trademarks of Kadon. Leap: a ¢ grid whereupon polyomino shapes are plotted with black and white checkers pieces in a double-size.½ Super Quintillions: 17 non-planar pentacubes (plus one duplicate piece to help ﬁll the box). joy and reverence that we sent one of the ﬁrst sets to the inspiration of our enterprise. Inc. keeping the checkerboard pattern (Figure 3).. not just a single product.
these are products made by Kadon Enterprises. checkerboarded format.
Figure 3. so we’d have to think of another name. The puzzle challenge: Change any one into another in the minimum number of chess knight’s moves. 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).THOSE PERIPATETIC PENTOMINOES
a registered trademark of Solomon Golomb.
column. In another use.
Colormaze: Square tiles in six colors form double polyomino shapes with no duplicate color in any row.
Quintachex: the pentominoes plus a ¾ ¢ ¾ square checkerboarded on both sides (different on the two sides). What is the minimum number of moves to exchange the congruent groups of black and white knights?
Figure 4. The pieces can form duplicate and triplicate models of themselves. formed with checkers (Figure 4). with checkerboarded arrangements (Figure 5). 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). or diagonal.110
Void: a ¢ grid on which a “switching of the knights” puzzle is applied to pairs of polyomino shapes from domino to heptomino in size.
and double-size polyomino shapes (Figure 8). in a four-way symmetrical tray of 89 unit squares (Figure 7).
Sextillions: the hexomino shapes can form double through sextuple copies of themselves and various enlargements of the smaller polyominoes. convex. Sizecompatible with Poly-5. They can form various single.
Poly-5: two-dimensional polyominoes of orders 1 to 5. concave—on the four sides of a square.THOSE PERIPATETIC PENTOMINOES
Figure 6. Snowﬂake Super Square: the 24 tiles are all the permutations of three contours—straight.
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.112
Lemma: A matrix of multiple grids lends itself to polyomino packing with checkers in three colors.
Triangoes: orders 1 and 2 polytans permuted with two or more colors (square. including partitions into multiple shapes (Figure 10). triangle. JONES
Figure 8. The pieces can form diagonally doubled pentominoes and hexominoes with colormatching adjacency of tiles (Figure 9). parallelogram).
Figure 12. double checkerboard hexomino shapes deﬁned by pawns are transformed with knight’s moves (borrowed from the Leap set).
Gallop: On a ¢ ½¾ grid. 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).
Multimatch II: The 24 tri-color squares with vertex coloring (each tile is a ¾ ¢ ¾ of smaller squares) can form polyominoes both as shapes with colormatching and as color patches on top of the tiles (Figure 11).THOSE PERIPATETIC PENTOMINOES
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 six pieces comprising 20 balls can make most of the pentominoes in double size. Throw a Fit: Mulit-colored cubes form pairs of pentominoes with colormatching. The 20 distinct pieces form a ½¼ ¢ ½¼ rhombus (Figure 14). This is a limited-edition set. JONES
Tiny Tans: Four triangle-based pieces can make some pentomino shapes. The T and U puzzles are actually dissected pentominoes (Figure 13).
Figure 14. Perplexing Pyramid: A Len Gordon invention.114
A natural side-effect of his raising the consciousness of the public about the joys of combinatorial sets has been the proliferation of other sets. In retrospect. Game of Solomon. I cannot imagine any other career in which I would have found greater satisfaction. through articles in various publications. Octominoes: The 369 eight-cell shapes exist in limited-edition sets. with kind and encouraging words in correspondence. humor and special themes of these two games set them apart from all others. the inspiring and nurturing presence of Martin Gardner has been there. then. The Hexacube: The 166 planar and non-planar hexacube pieces plus 4 unit cubes form a ½¼ ¢ ½¼ ¢ ½¼ cube.THOSE PERIPATETIC PENTOMINOES
Heptominoes: These sets of the 108 seven-celled polyominoes are produced by popular demand. A vast territory just begging for exploration. We were honored that Martin selected us to produce his Game of Solomon and the Lewis Carroll Chess Wordgame. sized to smaller members of the family. Throughout the years of Kadon’s evolution. All these are ﬁnding homes in the Kadon menagerie. sized to Sextillions and Poly-5. The charm. and always helpful information. fulﬁllment and never-ending challenge than in the creation of
Figure 15. The above concepts arose and became possible because Martin Gardner fostered a love of mathematical recreations among his readers. and we are dedicated to their care and continuance. polytans. using Martin’s game rules. 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. this set is sized to Quintillions. in the reprints of his columns. polyiamonds.
. besides polyominoes: polyhexes.
. Lewis Carroll Chess Wordgame. Martin Gardner.116
K. the pentominoes will announce themselves in yet another manifestation. if there is a 90-degree angle or parallelogram to be found.
beautiful playthings for the mind. JONES
Figure 16. Thank you. And assuredly.
Robert E. Neale
Flexagons are paper structures that can be manipulated to bring different surfaces into view. The four-sided ones discussed here are “ﬂexed” by folding them in half along an axis, and then opening them up a different way to reveal a new face. Sometimes the opening is in a different direction (mountain or valley fold), sometimes along the other axis, and sometimes both. Although usually constructed from a strip of paper with attachment of the ends, ﬂexagons seem more elegant when no glue or paste is required. Interesting designs of the faces can be found when the material is paper, or cardboard, with a different color on each side (as in standard origami paper). I call these “self-designing.” What follows is the result of my explorations. Although the text refers only to a square, the base for construction can be any rectangle. There are ﬁve different starting bases (paper shapes; see Figures 1–5). There are three different ways of folding the bases. Warning: Not all the results are interesting. But some are, my favorites being three: the ﬁrst version of the cross plus-slit (for its puzzle difﬁculty); the ﬁrst version of the square slash-slit (for the minimalist base and puzzle subtlety); and the third version of the square cross-slit (for its self-design). Note that the text assumes you are using two-colored paper, referred to, for simplicity, as black and white.
Figure 1. Square window.
Figure 2. Cross plus-slit.
R. E. NEALE
Figure 3. Square plus-slit.
Figure 4. Square slash-slit.
Figure 5. Square cross-slit.
The ﬁrst ﬂexagon I saw that did not need glue was shown to me by Giuseppi Baggi years ago. It is a hexa-tetraﬂexagon made from a “window” — a square (of 16 squares) with the center (of 4 squares) removed (Figure 1). It has six faces that are easy to ﬁnd. ½ I have recently used this window model for a routine about the riddle of the chicken and egg: “What’s Missing Is What Comes First.” This involves decorating the paper with markers, so is a separate manuscript. It should be noted, however, that there are two ways of constructing this ﬂexagon, both of which are discussed below. Same Way Fold. As the dotted line indicates in Figure 6, valley fold the left edge to the center. The result is shown in Figure 7. As the dotted line indicates in Figure 7, valley fold the upper edge to the center. The result is
½Directions for constructing the hexa-tetraﬂexagon are found on pages 18–19
of Paul Jackson’s Flexagons, B.O.S. Booklet No. 11, England, 1978.
shown in Figure 8. As the dotted line indicates in Figure 8, valley fold the right edge to the center. The result is shown in Figure 9.
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.
R. E. NEALE
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.
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.
moves. The other two faces are solid black and solid white. The other cells come together to form two boxes open at the top. and. You will make two. exactly as indicated in Figure 16.
. This puzzle is interesting for two reasons: It is tricky to construct. and the second ﬂattens it again. number the base from one to six. (Note: The trick fold for constructing the ﬂexagon can be done in a slightly different way that has the four faces not continuous. The ﬁrst renders the base 3-D. The four faces are continuous and identical.122
R. very quick. involving changing the ﬂat ﬂexagon into a three-dimensional ring and back again. Orient them just as indicated. They indicate that you are to make the base 3-D by pushing two opposite corners down and away from you. the other two faces are quite difﬁcult to discover. The numbers in parentheses are on the back side of the base. One result was a puzzle made from a “cross” — a square (of 16 squares) with the four corner squares removed. they are turned over so you see the 5 cells. There is a 6 cell at the bottom of each box (see
Figure 16. So reach underneath and hold the two free corners of the 1 cells. This base is manipulated in a third way. Note the two arrows in Figure 16. E. the 3-D Tricky Way.) 3-D Tricky Fold In order to follow the instructions and understand the diagrams. and two slits in the center that intersect in the shape of a plus sign (see Figure 2). while four of the faces are easy to ﬁnd. When you pull them down and away from each other. one corner in each hand. NEALE
The window model inspired me to think of other ways to form a ﬂexagon without having to attach the ends to each other. being black and white checkerboard patterns.
Do not open the model as you do when ﬂexing. the left and right halves going back away from you.SELF-DESIGNING TETRAFLEXAGONS
Figure 17. Life and Other Mathematical Amusements. You can correct this easily by tucking the wrong cell out of sight. however. then 3. pull them away from each other. The 3-D boxes will collapse. pp. a 2 instead of a 1.¾ Flexing. Sometimes. Change your grip so you are holding the two inside corners of the 5 cells. Check both sides. the base forming a ﬂat square of four cells. Begin with face 1 on the top and face 4 on the bottom. you will ﬁnd that another number has appeared. Mountain fold the model in half on the vertical axis. 64–68. move the lower inside packet of squares (with 2 and 3 on the outside) to the left..
Figure 17). 1983. Rather. Collapse the tube in the opposite way.H. Now open the model into a tube — a cube open at both ends.) Now push down on these corners. Freeman & Co.
. Now the base should be ﬂattened into a compact model with four cells showing on a side. and at the same time. 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 four 4 cells on the other side. Once this happens. Faces 5 and 6 are found by the following procedure. This creates a new arrangement. replacing it with the proper one. Flex it in the usual way to show face 5. New York: W. The model is completed. or a 3 instead of a 4. The result is shown in Figure 18. You can ﬁnd the faces numbered 1 to 4 by the usual ﬂexing. and in Martin Gardner’s Wheels. (These are the corners opposite the ones you were holding. and the upper inside packet of squares (with 2 and 3 on the outside also) to the right. you should have four 1 cells facing you.
I discovered a way to make a ﬂexagon without removing any portion of the square. checkerboard. with a single slit.) It is made from a square with two slits intersecting the center in the shape of a plus sign (Figure 3).
About a year or so ago. but they are not continuous. with face 1 at the back.
and then face 1. Four faces can be shown.) It is pleasing that a ﬂexagon can be constructed from a square. These follow. Using the 3-D Tricky Fold gives the same results as the Alternating Way Fold. use the Alternating Way fold. The designs of the four faces are two solid black and two checkerboard. solid black. All faces are of the identical solid color.) Four faces are easily found.
(Repeat “slash-slit” out loud quickly at your peril. and repeat the moves just given. For best results. (The goal was inspired by my desire to make a ﬂexagon from paper money with minimal damage. both the third and fourth are derived from the ﬁrst face. as mentioned above.124
Figure 18. or rectangle. The Same Way Fold can be used also. I have worked out some variations on the theme of removing no paper. Two faces can be found only by forming the rings. and four faces can be shown: checkerboard. Flexing is continuous. Indeed. begin on face 4. but they are not continuous. The other two faces are solid white. (Alternating Way gets you nowhere. To ﬁnd face 6. but with poor results. E. solid white. Another version of this ﬂexagon can be made by using the Same Way Fold. More recently.
This does not appear to offer as much variation due to the repetition involved. triangle in each corner. the move can be done quite secretly in the course of doing the ﬁrst ﬂex. the other two creased the other way. 1. solid black. the ﬁrst face takes on a quite
. but pull gently as you do it. 1. Many versions of these are possible. All have six faces. and the other two easily. checkerboard. Use of paper money is recommended. and then move the two little ﬂaps inside the slit to a position outside it. arranged as in face 2. once you understand what is required. 5. The ﬂexing is continuous. solid black. Then ﬂexing is possible. The Same Way Fold can also be used. solid black. 4. The faces are these: solid white. to be seen only at the very end of the ﬂexing. solid black. and best of all. Being sure you have the right axis on which to fold the model in half. 2. but with different designs: eight triangles of alternating color. ﬂexing is not possible without an additional move. 2. then challenging the knowledgeable paperfolder to duplicate the action. Using the Same Way Fold. 6. However. 4. leaving a white square in the center. There are two versions. The result ﬂexes exactly the same as the one above. circling about the center. two white triangles. Or. two white triangles arranged as in face 1. the second allowing some presentational by-play. two black triangles in a different arrangement. arranged differently from the previous two. However. 3. solid white. you can create some mischief by folding. Use of the Same Way Fold shows four faces continuously. The faces are as follows: two black triangles opposite each other. and two of the sides are just a little tricky to ﬁnd. two black triangles opposite each other. Here are four of them. Using the Alternating Way Fold. two white triangles opposite each other. solid white. in the following manner: face 1. The model will show four faces. The designs are these: checkerboard. The obvious way is to open up each side of the slit. etc.
These models are made from squares with two intersecting slits making an X (Figure 5). you can tuck the ﬂaps properly as you make the valley folds. however (Figure 4). Obviously. ﬂexing and unfolding. and the little ﬂaps will be adjusted automatically and properly. a second solid white. same as the second face. with the inner ﬂaps arranged differently: two opposite ﬂaps creased one way. do the folding.SELF-DESIGNING TETRAFLEXAGONS
It must be a slash along the diagonal. ﬂexing is not continuous. two black (and six white) triangles opposite each other.
the design changes. triangles in each corner so the center forms a white square. and the other two are deceptively hard to ﬁnd. Use of the Alternating Way Fold creates a puzzle version. two triangles on each face. solid black. with one triangle in each corner of the model. four faces each with different arrangements of three triangles. For example. also. solid white. but written more than a year or two before that. combine a half-cross base with a slash-slit. checkerboard of four squares. when a half of a corner is removed. The designs are these: checkerboard square. one triangle (out of eight). NEALE
different appearance when ﬂexed to the back of the model. although no movement to a ring is required. Four faces are shown continually.
. these reveal seven different designs: solid color. This manuscript was submitted for possible publication after the Gathering for Gardner in January of 1993. instead of all or none of it. but differing in both color and arrangement. One of the items was taught at the Gathering and appeared in the Martin Gardner Collection. and use the Same Way Fold. (This is my favorite of these four because the designs are handsome and the two solid color faces are a little tricky to ﬁnd. I constructed one to reveal the maximum of variety in design. Additionally. The second and third showings of the face reorient parts of it. checkerboard square. The designs on one pair of faces are different. And so on. E. So although my multiversion shows only six faces. as the eight triangles are rearranged in an interesting way — four triangles of alternating color circle about the center.
Other ﬂexagons can be constructed from the X-slit approach. the methods given above can be combined to create still different design possibilities.) The Alternating Way Fold can be used with the inner ﬂaps arranged differently. For example.126
R. This uses the fact that all ﬂexagons show some faces more than once. For another example. four large triangles alternating in color.
except to describe their common design feature. thus:
Then bend the wire into some animated shape. 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. To ﬁll this need. which almost by deﬁniton do not require close tolerances and are just about the easiest to fabricate. such as in the very simple example below. Liberty Bell.
. This led to a couple of not very original or interesting string-and-bead puzzles (the Sleeper-Stoppers). so I will not waste space on them here.The Odyssey of the Figure Eight Puzzle
I became established in the puzzle business in 1971. Take a foot or two of ﬂexible wire and form a loop at both ends. The three rather unexciting puzzles that came out of this (Lamplighter. with wire passing through the loops in various ways. I also tried to come up with something in string and wire that could be licensed for manufacture. I may have been inspired in this by a neat little puzzle known as Loony Loop that was enjoying commercial success at the time. hand-crafting interlocking puzzles in fancy woods and selling them at craft shows as fast as I could turn them out. I turned my attention to topological puzzles. and Bottleneck) are described in the 1985 edition of Puzzle Craft.
For a slight variation of this. The puzzle editor made the surprising observation that it was topologically equivalent to the Double-Treble-Clef Puzzle made by Pentangle and therefore solvable. and I did not spend a lot of time trying to digest it. full of diagrams and such. decided to add the Figure Eight to the line of puzzles that he made and sold in his spare time. COFFIN
Finally. I published this simply as a curiosity in a 1974 newsletter (later reprinted in Puzzle Craft). but being a novice in the ﬁeld of topology. none of these ideas enjoyed any success.104
S. More recently. The proof appeared to be a rather complicated. and tried to remove it. they reported that topological puzzles in general were their least favorite of any puzzles I had produced. 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. (In a recent survey of my puzzle customers. To add to the confusion. I received a seven-page document from someone in Japan.
I soon became convinced that this was impossible. Alaska. knotted a cord around it. Royce Lowe of Juneau. he begged me for help. I had continued to
. during an idle moment one day I formed some wire into the ﬁgure-eight shape shown below. I was at a loss for any sort of formal proof. Over the intervening years. When some of his customers started asking for the solution.) The story does not end there. Alas. however. somewhat to my surprise. But a careful check showed that the two were not quite equivalent. Some readers misread that purposely vague write-up and assumed that it must have a solution. some of these puzzles can be quite bafﬂing to solve. purporting to prove that the puzzle was unsolvable. Unlike the simple example shown above. 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). which then left them utterly bafﬂed as to ﬁnding it. since the Pentangle puzzle was. Next it appeared in a 1976 issue of a British magazine on puzzles and games.
perhaps disguised in another form. and by that time I was rather tired of the whole thing. When it comes to puzzles. Or will it mischievously rise again.THE ODYSSEY OF THE FIGURE EIGHT PUZZLE
received numerous requests for the solution. 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. as topological puzzles so often do?
. it is often the simplest things that prove to have the greatest appeal. probably not even realized at the start.
will analyze the problem with logic and stratagem. a back injury from an auto accident and lots of encouragement from puzzle collectors brought the following and many other puzzle ideas to fruition. a mathematical problem. 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. mechanical puzzles are an open book. Wire lends itself very easily to topological problems because of its inherent nature to be readily
. My own interest in puzzles began in early childhood. In fact. with everything visible. then reason out the solution to include returning the parts to their original starting position. puzzles entertain. and the search for puzzling challenges will undoubtedly continue. and here is where the separation between the common puzzler and (to borrow a phrase from Nob Yoshigahara) a “certiﬁable puzzle crazy” lies. the solutions can elude even the sharpest and quickest minds of every discipline. those who like puzzles generally like to solve just about any problem. they were never quite enough of a challenge to satisfy my hunger. About twenty years after my introduction to those ﬁrst little wire teasers. Regardless of one’s ability to solve them. Be it a paradox. Somewhere in the back of my mind I knew I could come up with better puzzles than were currently and commercially available. Puzzles can go beyond an understanding of the problem and its solution.” on the other hand. Unlike magic or illusions with misdirection and hidden mechanisms. A “puzzle crazy. the search for answers drives many of us on. or a puzzle. all parts exposed ready for minute examination and scrutiny. magic. Although entertaining. Wire disentanglement puzzles are topological in nature and can vary widely in both difﬁculty and complexity of design. In spite of this.Metagrobolizers of Wire
Many people love the challenge of solving a good puzzle. 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. mystify and educate.
Thanks also to Tom Rodgers for his support of my work. It is my suspicion that the subconscious mind is constantly at work attempting to ﬁt pieces of countless puzzles together. This puzzle idea came to me as I was driving to San Francisco to sell my puzzles at Fisherman’s Wharf. IRBY
and easily formed into whatever permanent shapes may be necessary to present a concept.” and for presenting me with the opportunity to meet Martin Gardner. Many thanks must go to Martin Gardner as an inspiration to the millions enlightened by his myriad works. Often it is easier to come up with and develop a new puzzle idea than to give it a good and catchy name. sometimes it succeeds! If you devote your mind to something. Unfortunately. for asking me to participate in “Puzzles: Beyond the Borders of the Mind.
The Bermuda Triangle Puzzle
. I am frequently asked to explain the thought process involved in coming up with a new puzzle. I really can’t answer. either you become good at it or you are devoting your mind to the wrong thing.
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. I neither know or understand the process of any creativity than to say that it just happens. A couple of examples of ideas that have “popped” out of my mind at various times are explained and illustrated below.132
R. in 1971 or 1972.
Sometime later one of us stated that we probably would have nightmares about it that night. The triangle can be moved over the entirety of the larger conﬁguration. After making the prototype. as I sat playing with it. 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.
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. which is a triangular piece. The solutions to many puzzles can be elusive until the puzzle has been manipulated many times. the UFO being a ring with an abstract shape mated to it. Most of the large conﬁguration to which the triangle is attached is there simply to bewilder the would-be solver. As with most of my puzzle ideas. my wife joined me in critiquing my latest design. The Nightmare
. 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. We didn’t have any nightmares about it. The Bermuda Triangle rates about a medium level of difﬁculty. this one came to me fully formed and complete. although moderately difﬁcult for the average puzzler to solve initially. The puzzle is generally set up with the ring around the Bermuda Triangle.METAGROBOLIZERS OF WIRE
The object of the Bermuda Triangle is to save the UFO that is trapped in the puzzle. this one is relatively easy to remember once the solution has been seen. taking the UFO with it as it moves. but the name stuck.
A cord encircles the two inner loops of the puzzle. the ﬂexibility of the cord allows one to make mistakes not possible with rigid pieces. I rate the Nightmare an 8. There are two outer and two inner loops. quickly compound into a tangled mass of knots soon precluding any progression toward the solution.134
R. with the wire ends making small rings that are wrapped around the wire in such a manner as to eliminate any usable ends. The Nightmare puzzle is made from one continuous strand of wire.
. In addition to the difﬁculty in conceptualizing the convoluted shape of the puzzle. Any wrong moves. not promptly corrected. and the object is to remove the cord completely from 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. On a scale of 1 to 10. The difﬁculty may be increased by. after the cord is removed. adding a ring to the cord that will not pass through either of the small end rings then attempting to replace the cord.
In 1978. I was going to write a column on “Games” so I made a pilgrimage to Hastings-on-Hudson to visit the Master. I read Martin Gardner’s columns religiously. and corresponded with him occasionally from 1963 on. and at the top an hourglass ﬂoats. I assumed it had to do with some law that I had forgotten since high school. that I not only could have thought of it myself.
. There on a shelf was the infamous Floating Hourglass itself. Omni. It seems impossible that a transfer of sand from top to bottom of the hourglass would have any effect on its overall buoyancy. I had the pleasure of ﬁnally meeting the Master Explainer. was to be launched. An unusual toy is on sale at a Paris shop: a glass cylinder. I turned it over and the hourglass stayed at the bottom. The next month. It was so simple. I should have. Can you guess the simple modus operandi ? I gave the problem some thought but couldn’t come up with any good theory. as a college student. I could ﬁnally try out the curious toy I had read about so many years before. Then it rises slowly to the top. as a graduate student. in the months before a new science magazine. when the answer came. ﬁlled with water. and then as an editor of Psychology Today. If the cylinder is inverted a curious thing happens. The effect was like seeing a good magic trick or hearing a good joke. so absurdly obvious.Beautiful But Wrong: The Floating Hourglass Puzzle
One of the problems in Martin Gardner’s August 1966 “Mathematical Games” column was The Floating Hourglass. the hourglass remains at the bottom of the cylinder until a certain quantity of sand has ﬂowed into its lower compartment. just like Martin said it would. I was delighted.
For years I have itched to tell this story. the Super Egg) and poet/artist (Grooks). When Martin allowed me to look through his ﬁles. Hein told him about a toy he had seen in the Paris airport. the Danish sculptor/inventor (the Soma Cube. had visited Martin in early 1966. made by Ray Bathke of London. I immediately ordered some. MORRIS
At the 1991 Puzzle Collectors Party in Los Angeles.
The First Theories
Piet Hein. This book ﬁnally gives me the chance to tell the history of the Floating Hourglass Puzzle. I asked my clever readers to submit theories to explain it. a treasure trove of letters and drawings. I saw my second Floating Hourglass and knew I could ﬁnally write about it.136
S. The results appeared in the January 1993 issue. since I only published puzzles in Omni that I knew my readers could ﬁnd. Tim Rowett had brought one from England. He didn’t bring one back. I found a thick folder on the Hourglass. but no magazine article could possibly contain it. Gardner knew how it must work and wrote about it in his August and September columns without even having
. My September 1992 column introduced the Floating Hourglass 25 years after I ﬁrst heard about it. but his description was clear enough.
The hourglass rises not because there is little sand left in the top chamber. don’t weigh the package while they’re running. He drew a cartoon of himself on an elevator with Einstein.. not the amount of sand that has fallen or is left. He knew of Gardner’s “answer. Until you could see and touch one. In a shop there he saw a double-glass: two cylinders side by side. break it open or learn how it is made. but because the rate of falling sand has decreased. Hein tried the double-glass in the shop a few times. pondering an oversized hourglass. Hein had an agonizing experience in Milan.” Note that he wrote all this after the September Scientiﬁc American was out. For him this meant that the ﬁnal proof was not yet in. Martin’s theory relied on friction between the glass and the cylinder.. “That was all I wanted to know. Hein believed the impact of the sand grains hitting the bottom of the glass exerts a downward force. A “sunken” hourglass that stays in place at the top of a tube can’t be explained by sand grains falling in the opposite direction. He knew immediately that his impact theory was doomed.” Hein wrote Gardner on September 16. When turned over. but Pien wasn’t convinced. the glasses stay in place at ﬁrst and then one rises while the other sinks. or you’ll have to pay a higher postage. There it stands at the bottom and changes its weight!.
A Painful Paradigm Shift
Just a couple of days after writing the letter and drawing the cartoon. 1966. just enough to see that it worked. Hein issued a mock warning: When mailing hourglasses. He thought the inside of the cylinder he had seen was too smooth to offer much resistance. an ‘effective change in mass’: “The hourglass is heavier while the sand is falling.” he wrote “and all that was needed to make my intellectual headache come back much worse than
. Hein was obsessed with hourglasses. He designed an hourglass-powered perpetual motion machine and created a fantasy ocean full of bobbing hourglasses. This seems to solve the whole problem. Since the glasses change their weight whenever the sand falls inside. He felt there must be something more.” but he also knew that Martin had never examined an actual hourglass sample. the best criterion for a theory is its beauty.THE FLOATING HOURGLASS PUZZLE
seen one. all theories were valid contenders. something to do with the falling sand. a ﬂoating hourglass in one and a sunken hourglass in the other. “Imagine if the hourglass were opaque and you didn’t know it were an hourglass at all. In the absence of knowing the truth. Italy.What keeps the hourglass down is the falling of the sand. And Martin had already admitted that Piet’s impact theory was beautiful.
” He later wrote about the “Aha” experience. I found myself in possession of the solution. But it does not make it any less beautiful. not the sand. with the exception that the sand was running downward in both hourglasses. The falling sand is just for misdirection. 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. I must admit.” Then. “somewhat worried and depressed on behalf of my beautiful theory.” Hein was forced to make a paradigm shift. like a small ﬁsh on the center of a huge platter. It really hurt. the solution of the hourglass mystery. exactly half moon (D for dynamics) were left behind us and all was dark. My dear Watson. he was somber. “I should like to meet the person who invented this effect and designed it so as to hide the solution so elegantly for us.” he explained. When the cylinder is inverted. And on that tablet lay. The hourglass puzzle is caused by an hourglass effect. “When the Alps in the lite of the (superelliptic) setting sun and the growing.
. Only when enough of it has seeped down does the glass begin to rise.” he wrote. He knew that the sinking glass directly refuted his theory.“At an altitude of 8500 meters. There must be two liquids of different speciﬁc weights that don’t mix completely.138
S. not in the falling of the sand but in the ﬂowing of the liquid. “I couldn’t think of anything else but the principle. and are indistinguishable in color and transparency. but what a clever ruse it is. 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. I admit. but of thinking. Hein realized. “This is not a question of fumbling one’s way to a solution. as the plane soared over Mont Blanc.”
The Solution Is in the Solution
The key to the puzzle isn’t in the glass but around it.” When Piet Hein left Milan on the afternoon of September 21. the heavier liquid is on top pushing down. “I am glad you think my explanation is beautiful. MORRIS
the ﬁrst time.” He couldn’t bear to abandon his pet idea completely. there was a lull in my mind. being false makes it less right. it’s very simple. “So do I. but the relevant displacement is in the liquid. Hein had a sudden crystallizing insight . and he found it painful. He rationalized in jest: “My theory works. Hein was awed by the brilliance of it all.” he wrote on September 22. a tabula rasa.
“I am writing to put your mind at ease (on the impact theory posed by Altman). also from the U. On September 6th. It is possible that one version of the toy works on the principle you mention. I am sure that your explanation was correct. In short. and to suggest that you not publish a correction. Reid later adapted his letter for publication and his short article “Weight of an Hourglass” appeared in American
. all you have to do is shake your own hand. but did not bring one back with him.”
The Hourglass Letters
While Hein was having his epiphany in Europe. You are the inventor. Maryland: Another solution to the hourglass science teaser is that the momentum carried by the falling sand causes the hourglass plus to weigh more than its static weight by an amount Ô¾ sand . Gardner was beginning to wonder if his friction explanation told the whole story. 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. Naval Ordnance Laboratory at Silver Spring. and it didn’t last long. Naval Ordnance Lab. Albert Altman wrote from the U.S. A letter dated September 29 came from Walter P. where is the rate of the ﬂow of the sand. and the height through which the sand has fallen. Hopelessly confused? Martin Gardner?! That is something that surely doesn’t happen very often. perhaps depending on the temperature. 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 FLOATING HOURGLASS PUZZLE
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. One man wrote that he owned one of the Paris cylinders. the acceleration of gravity. or perhaps still another one.” Reid went on to show mathematically how the impact of sand hitting the glass’ bottom is exactly balanced by the loss of the sand’s mass while it is in free fall. but reported that his glass sometimes ﬂoated and sometimes sank. at this point I am hopelessly confused. Reid. Gardner was getting letters in response to the September issue.S. and the other on the principle I suggested. I have not yet seen the toy. Well. The principle is simply too clever to be true. having relied (unfortunately) on an account given to me by a friend who examined the toy in Paris.
” he wrote to Hein. He turned out to be a Czechoslovakian glassblower named Willy Dietermann. 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. In his lab. When Hein asked about the liquid.”
By this time. completely invisible. Gardner cites it in the hourglass puzzle reprinted in Mathematical Circus. Stong shone polarized light through the cylinder and found that the refractive index of the liquid did not change from top to bottom. April. the other down. He also reported it had a freezing point of eight degrees below zero. which eventually sends one hourglass up. and tracked down the maker. Mr. but which do not separate completely.L. He concluded the liquid inside was homogenous. and found a transparent cylinder into which it would ﬁt. Dietermann explained it was 90 to 95% water and 5 to 10% a combination of alcohol and formic acid. the only scientiﬁc writing on the subject. the original “Amateur Scientist” columnist of Scientiﬁc American. freed the glass from the wooden frame.
Hein and the Horse’s Mouth
All was settled for a few months. but then in the spring of 1967 the glass rose again. It worked just like the big version. The glasses do tip to the side when inverted.”
. the inventor staggered as if struck. “then I wrapped copper wire around the middle of the hourglass. MORRIS
Journal of Physics. “Red” Stong. 35(4). 1967. He decided he had to write a correction after all. but that is just a holding operation until the liquids have had time to change places. so there is no visible division between the two. A slight difference in refractive power is concealed by the curvature of the glass cylinder. Gardner had his own hourglass tube. He loaned it to C. Gardner wrote to Dietermann but received no reply. The main principle is the liquid hourglass effect.140
S. By snipping off the ends of the wire I was able to make its weight such that it slowly rises in the cylinder. He bought a cheap egg timer. This remains. and he conﬁrmed Hein’s two-liquid theory. Gardner was playing amateur scientist himself. “I ﬁlled the cylinder with water. Meanwhile. Hein wrote that he went back to the shop in Paris where he ﬁrst saw the glass. It was the ﬁrst time his secret had been guessed: two liquids of different density. to my knowledge.
” reasoned D.” I received over 1200 letters about the hourglass after publishing the puzzle in Omni. He wanted to expand his “perpendicular thinking. that the hourglass ﬂoates up with the pocket of warm liquid surrounding the glass’s neck. About 40% of those readers with incorrect answers cited heat as a factor in the hourglass’s behavior. but they were creative products of human thought. and Hein’s French was worse than his German. The Dane and the Czech had to speak German because Dieterman knew no Danish or English. Ontario. Let others measure a refractive index or a freezing point.” Perhaps Dieterman was leading him along. others. and deserved to be prized for that alone. “The top and bottom of the hourglass are so thin as to sag under the weight of the sand. Q. Some argued that this warms the surrounding liquid so the hourglass stays down until the liquid cools again.
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. or perhaps there was an honest misunderstanding about the use of different liquids. Many correspondents blamed the “impact” of falling sand for keeping the glass down. Hein wanted to think the problem through. beautiful explanations. G. Canada. ﬁrst on the bottom of the glass and later on a
. of Richmond Hill. Some even used mathematical formulas to show how much force a sand grain exerted. There certainly was plenty of room for misinterpretations. making it expand slightly and then rise. “When enough sand falls into the bottom chamber. We may never know what Hein got from the “horse’s mouth. sorting them into different piles. The other 60% broke down into about 15 different theories. As I read them. California. More than 50 readers thought that the hourglass was ﬂexible. the hourglass widens and wedges itself into the cylinder. increasing the hourglass’s volume. they said. Falling sand generates heat. of Los Altos Hills. it ‘bubbles’ the bottom end out. Some reasoned that when the sand presses down from the top. This proportion stays at about 40% with each new batch of mail. They may have been wrong. But most in this category thought the heat warms the air in the glass.” wrote B. and his German was better than his French.THE FLOATING HOURGLASS PUZZLE
Sanity was restored and the correction was never printed. Others decided that the hourglass is ﬂexible only at the ends. He wanted to search for alternate. I found the largest single category was always the “correct” theory.
of Tillamook. Many believed the air at the top of the hourglass lifts it to the top of the tube. Oklahoma. California. When the air is in the bottom half. and air in the bottom of the hourglass became compressed until “a jet of air shoots into the top half of the hourglass. the water below the glass can push up only on the circular end of the glass.” wrote T. I promised copies of the book Omni Games to the ﬁve “most interesting” entries. T. “It just takes a long time for the hourglass to get started. H. Therefore.142
S. supposed that sand blocked the hole. imparting just enough lift to overcome inertia and start the glass moving up. They concentrated on the air bubble that constantly rises.” explained D. of Coweta. wrote one reader. a greater surface area. Florida. but the calculations have to consider the amount of time each grain of sand is falling and weightless. Gagne of Eglin Air Force Base. Another line of reasoning put the emphasis on the liquid: “The solution is in the solution!” wrote H. then the denser liquid is at the top when the tube is turned over. ﬁrst in the hourglass and then in the tube. the hourglass begins to rise. W.” argued P. If the liquid is naturally cooler and denser at the bottom. N. When the air rises to the top half. Oregon.” perhaps because the liquid is very viscous. water can push up all around the inverted cone. as a system. while the weight of the falling sand is required to release the mechanism. Correct answers to this puzzle aren’t very interesting because they’re all virtually alike. It eventually seeps down below the hourglass and buoys it up.”
. is rising from the moment the column is inverted. A surprising number thought it was all an illusion. Movements within the system don’t alter the weight of the system. I awarded books for incorrect answers only:
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. North Carolina. The momentum of rising provides enough energy to engage the mechanism as it reaches the top of the cylinder. “It’s the same principle that causes a snow cone to pop out of its cup when you squeeze the bottom. of Glendale. “The hourglass. About 4% of those who wrote in thought that the shape of the hourglass affected its buoyancy. “When enough air reaches the top chamber and exerts its pressure there. the two tiny opposing forces exactly cancel each other out. MORRIS
mound of other sand grains. of Chapel Hill. The theory may be correct. A.” John J.
” The ﬁfth book. and Daniel E. Arizona.D.D. goes to Holzapfel.THE FLOATING HOURGLASS PUZZLE
“The hourglass is composed of a ﬂexible material such as Nalgene. Shoreham. Physics Teacher. If this is wrong. then my name is John Holzapfel and I teach chemistry. New York. Ph. Watson Research Center in Yorktown Heights. therefore.S. 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. Ph. the loss of friction enables it to rise.
. Cliff Oberg of Clarkdale. New York. The “correct” answer. as Martin Gardner wrote it in 1966: When the sand is at the top of the hourglass. The resulting friction against the side of the cylinder is sufﬁcient to keep it at the bottom of the cylinder. Dinger. a theory about the density gradient of the liquid was signed “Bob Saville. a high center of gravity tips the hourglass to one side.. “P.” He added. Shoreham-Wading River High School.” wrote Timothy T. The two share a prize for having the conﬁdence to submit their incorrect theory on company stationery: IBM’s Thomas J. Finally. After enough sand has ﬂowed down to make the hourglass ﬂoat upright. Edelstein.
. 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. rather amusing. With proper insight it can be solved in a few minutes. Three “ﬂats” of size ¢ ¢ .e. Three “longs” of size ¢ ¢ . Usually there is an “aha” reaction when students see that a cube can actually be constructed from the pieces. has exactly two solutions — each with a different set of three colors. For more advanced students it would be reasonable to ask in how many essentially different ways can the cube be constructed from the eight (uncolored) pieces. if colored as above.Cube Puzzles
A new. The problem then is to arrange the eight pieces into a cube so that opposite faces have the same color. 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
´ · µ¿
¿·¿ ¾ ·¿ ¾· ¿
Gardner calls such models of mathematical theorems “look-see” proofs. One “Biggie” of size ¢ ¢ (where ¾ ). Without this insight it typically takes several hours to arrive at a solution if one can be found at all. This puzzle. Now color each piece with six colors according to the scheme in Figure 1. In fact he has recommended that every teacher of algebra construct a set of eight pieces for classroom use. This will depend on just what is meant by the words “essentially
Therefore.½ distinct ways of constructing the cube. and 3 are any six distinctive colors. B. and 3. C. 2.
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. 2. 0. additionally. Solution Hints. If you cannot.146
J. B. That is.½½ . These hints are to be regarded as progressive. The combinatorial puzzle uses only six colors in its construction.
Call the six colors A. and C will turn out to be a solution set. proceed to the next hint. B. 1.½ ¼. A. FARRELL
Figure 1. after reading a hint. 0). There are 93 solutions in this case. each of the 48 faces of the pieces is colored with a unique color. but there still remain a great many cases that are nearly right. C. A. and the solver has the additional clue that a face is all of one color. then there are ´ ¿µ´ µ´¿ µ¾ . If. each
. 1. Notice that the eight pieces must form the eight corners of the completed cube — one piece for each corner. Trial and error is not a very fruitful way of trying to solve this puzzle. try again to solve the puzzle.
nor can C and 3. take any piece (Biggie is a good choice). 1–2–C. etc. we may choose another piece. Place Biggie as in the diagram so that A–B–C is a corner. say a ﬂat. that four pieces will have the counterclockwise order A–B–C and the other four the order A–C–B. This means that A and 1 cannot appear together in a solution. A–B–3. B. it may happen that on that ﬂat. using Biggie as a guide. Of the 20 possible sets of three colors (out of six) we are left with only eight possibilities: A–B–C. B–2. Of course. and C. and C are showing. A–2–C.CUBE PUZZLES
piece must have on it a corner with the three colors A. 1–B–C. When eliminating possibilities. B. We would then know to eliminate anything with A and 2. and C in some order. on any other piece. Likewise B and 2 cannot. These three ﬂats must cover all or part of the colors 1. 1–B–3. For instance. B. Since it is also true that. keep A opposite A. it is convenient to turn the three ﬂats to allowed colors. 2 opposes A. One or two more tests with other pieces lead to A–B–C (or 1–2–3) as a candidate for a solution. (It is a fact.
To determine the colors A. and use it to reduce further the possible solution colors. colors that oppose each other cannot appear in the same solution. and 1–2–3.)
Figure 2. This would force A and B to be together. and notice that the three pairs of colors A–1. and 3 of Biggie. 2. but not necessary for the solution. and C–3 oppose each other. A–2–3. Place all three ﬂats so that A. It is easy to visualize exactly where a particular ﬂat
and 3). 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. If only one solution is desired. main diagonal (upper left to lower right). ﬁnally. Then place the longs and.148
J. 2. Smally.
. That is. There is another solution to this puzzle using the colors 1– 2–3. and one to slip in when our enemies try the puzzle that is colored so that no solution is possible!
Figure 3 shows the schematic for a Magic Die. take any piece and interchange two of the colors A. FARRELL
must go by looking at its A–B–C corner. B. The solver will notice that the completed cube has a “fault-free” property. We like to have three Smallys prepared: one that yields two solutions. This will change the parity on one corner of the cube so that a solution is impossible using that set of three colors. or off-diagonal around all four lateral faces is always 42. The Magic Die has the amazing property that the sum of any row.
Figure 3. This will assure that only one solution is attainable with the colors A–B–C (the proof is left to the reader). column. and C (or two of 1.
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
You can construct a Magic Die puzzle by taking twenty-seven dice and gluing them together into the eight pieces of our combinatorial cube puzzle— being sure the dice conform to the layout shown in Figure 3. (Alternatively. it will be extremely difﬁcult to ﬁnd.)
. There will be only one solution to this puzzle (with magic constant 42). and.
A typical set is shown below in Figure 1. the nine color puzzle was ﬁrst introduced during 1973. but after reading this article it should not be
. Typical solutions are shown at the end of this paper. using the name “Nonahuebes. analyze it thoroughly.
Figure 1. combinatorics. sets.” included it among the puzzles I produced in New Zealand under the aegis of Pentacube Puzzles. and requires logical decision-making. This puzzle is of particular interest because it is of simple construction. The object of the puzzle is to assemble the pieces into a cube with all nine colors displayed on the six faces. Ltd. has an easily understood objective. parity. ordered pairs. is a manipulative puzzle. Altogether there are three cubes each of nine different colors. I was intrigued by the puzzle’s characteristics but did not.
According to Reference 1. A complete analysis requires the use of a computer. until recently. and isomorphisms. Included in its analysis are the concepts of transformations. and. as “Kolor Kraze. has numerous variations and solutions.” I learned of the puzzle in 1977 from Reference 2. backtracking.The Nine Color Puzzle
The nine color puzzle consists of a tricube. in Canada. and twelve different dicubes with each cube of a dicube a different color. where each number represents a color. with each cube a different color.
Since one person can assign the fourth and ﬁfth colors to orange and red and another person to red and orange. some with the tricube on the edge and some with the tricube through the center of the cube. It becomes necessary. none of the nine planes may contain two cubes of the same color.251. to separate these cases into disjoint sets of isomorphisms. The third observation. A proof is given in Section 2. the location of the third cube is determined. Since during the course of trying to solve this puzzle a conﬂict often occurs. when two cubes of the same color are in position. only thirty dicubes are required. The individual dicubes can be colored thirty-six different ways. in which the number of isomorphisms using the set of thirty-six dicubes is also determined. as determined in Section 4. and the most difﬁcult aspect to investigate was the fourth observation: the color combinations need not be as shown in Figure 1. How many solutions are there for each color combination? And how many color combinations are there? This paper addresses these last two questions. A proof is given in Section 3. Thus. The question then arises: How many solutions are there? Most intriguing about this puzzle. FARHI
a difﬁcult exercise for a moderately proﬁcient programmer to verify the results obtained. only 133. The ﬁrst observation noted is that the cube can be assembled with the tricube along the edge or through the center of the cube.677. Twelve of these can be selected in 1. is that there are numerous solutions.700 (36!/12!24!) different ways of which. The second observation is that for a cube to be assembled with the nine colors on each face. and only 10. the results are shown in Table 1.152
S. it becomes apparent that most of the 133.
. this rule then tells the experimenter to backtrack. but not in the center of a face.105 combinations are isomorphisms. then. The ﬁnal part of this analysis was to determine the number of puzzle solutions for each disjoint set of dicube combinations. The previous question now becomes more interesting. The method of sorting the 10.691 combinations into 148 disjoint sets is described in Section 6. Fortunately.691 combinations need to be sorted into disjoint sets. obtained after some experimentation. This was achieved through the use of a computer program.105 of these combinations meet the three-cubes-of-each-color requirement. as is shown in Section 5.
DZ[yz 1439 624 96 4 80 6 15 .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
.COeyz 360 180 8 0 16 16 0 . 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center .DZpez 720 312 198 12 83 10 16 .:Zppp 10 10 0 0 0 0 0 .
No.DOeyz 720 312 20 0 12 0 4 .:Z[yz 60 60 0 0 0 0 0 .THE NINE COLOR PUZZLE Table 1.
No. 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73
Isomorphisms Number of Solutions dicubes straight L shape Name 36 30 Edge Center Corner Edge Center /Obppp 720 84 38 2 104 46 20 /Oceyz 720 84 18 0 46 6 8 /Ocmyz 1439 168 34 3 27 12 2 /Ocnpz 1439 168 126 4 117 16 19 /Ocnyz 1439 168 40 1 36 11 11 /Ocnzp 1439 168 49 3 45 24 9 /Ocoxz 1439 168 112 5 14 3 2 /Ocozp 1439 168 18 1 35 28 9 /Oewez 720 84 32 2 38 48 4 /Oewnz 1439 168 52 3 20 15 1 /Oewpp 1439 168 17 6 38 12 4 /Oeyxp 720 84 36 0 12 6 4 9DEeyz 180 36 0 0 16 0 0 9DNeyz 30 6 0 0 0 0 0 9DOeyz 360 72 0 0 32 12 4 9DXeyz 720 108 10 0 0 0 0 9DYeyz 720 108 22 2 38 18 6 9DZ[yz 1439 216 69 3 32 17 2 9DZepz 1439 216 71 7 32 16 7 9DZpez 1439 216 55 6 52 13 5 9DZppp 720 108 54 12 72 18 20 9ODeyz 360 36 4 0 4 0 0 9OF[yz 360 36 76 0 72 8 4 9OFepz 720 72 82 0 45 0 6 9OFppp 180 18 208 0 134 8 0 9Ob[yz 720 72 54 4 41 12 2 9Obepz 1439 144 145 7 48 13 5 9Obmyz 1439 144 87 2 31 8 0 9Obnpz 720 72 2 0 18 4 0 9Obofz 720 72 68 4 28 2 0 9Obpez 720 72 22 2 25 4 0 9Obpfp 720 72 174 6 71 10 0 9Obppp 1439 144 30 3 26 5 4 9Odmpz 720 72 186 6 54 10 2 9Odmzp 360 36 16 0 6 12 0 9Odoyz 360 36 76 8 66 16 4 9Oewpp 360 36 40 0 16 4 0 DDDeyz 15 1 0 0 0 0 0
THE NINE COLOR PUZZLE
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
” Other possible locations are reﬂections or rotations of these and are not considered as separate puzzles solutions. There are only two possibilities. as mentioned in Section 2. There are two rather startling phenomena shown in Table 1. thus leading me to believe that they originated from the same source. It is not necessary that the tricube be straight. The tricube location ﬁlling spaces 1. Verifying this would be an interesting exercise for the reader. Stranger yet. there are two color combinations (Numbers 73 and 121) which. 14. these two cases have no puzzle solutions. no matter how the colors are swapped. Additional Comments
The Kolor Kraze puzzle shown in Reference 1 and the nine color puzzle shown in Reference 2 are isomorphic. map onto themselves. Each dicube when similarly checkered has one even and one odd number. Is this a coincidence?
2. where odd numbers represent one state and even numbers the other.
The three layers of the cube are checkered as shown in Figure 2. and. The number of solutions for these cases are included in Table 1. Since the cube has fourteen odd numbers and thirteen even numbers.
. 10. Permissible Locations of the Tricube as Determined by Parity Restrictions
Figure 2. the tricube must ﬁll two odd numbered spaces. While some color combinations within the set of thirty dicubes have up to 720 isomorphisms. an irregular tricube may have three locations. and 23 is referred to as a “center case.THE NINE COLOR PUZZLE
1. and 19 is referred to as an “edge case” while the tricube location in spaces 5.
then one color is selected from the third row and if color 3 has been selected twice. Since the corner colors are displayed on three faces. 11. 2.
4. and 3. there are three possible locations. the edge case ﬁlls spaces 2. and 11. then a third dicube with color 2 and any color from the second row is selected. FARHI
If an irregular tricube is used. The corner case ﬁlls spaces 1. and 14. and on a centerface such as shown by the letter “B. accounting for two faces. and 11. the center case ﬁlls spaces 5. The “No Two in the Same Plane” Rule
Figure 3. If color 3 has been selected once.” The remaining six corners must have cubes of different colors. cubes of the same color are never in the same plane. then color 3 is combined with two colors from the third row. Determination of the 133. the remaining two cubes must be located on an edge.105 Possible Dicube Color Combinations
The tricube colors are deﬁned as colors 1.
One cube is centrally located and thus not visible.” The remaining two colors must all be on edges.158
S. If none of the previous selections include color 3. 5. If color 2 is selected from the ﬁrst row.” In each case. The two dicubes with color 1 may use any two colors in the ﬁrst row of the following table. such as shown by the letter “C. The only way the remaining two cubes of that color can be displayed on all six faces is for them to be located on opposite corners as shown in Figure 3 by the letter “A.
3. 2. If color 2 is not selected from the ﬁrst row then color 2 is combined with any two colors from the second row for the third and fourth dicubes. then none are selected from the third row.
6). given the second numbers 2. 7). (4. 6). 5. 4. 2). 8).691 color combinations. 8. Color 1 2 Selection 2 3 4 5 3 4 5
This can be justiﬁed by deﬁning the colors associated with color 1 that are not 2 or 3 as colors 4 and 5. Using this algorithm. Identiﬁcation and Separation into Disjoint Sets
The color combinations in Figure 1 use dicubes with color (1. 3). 6. using the same algorithm as before. (4. Similarly if color 2 is not associated with colors 3. 8). (7. color 5 is combined with colors in the ﬁfth row such that there are three cubes with color 5. 5). 8. only 10. or 5. (3. 9. etc. 9. 3. (2. 9). (6. the two new colors are identiﬁed as colors 6 and 7.THE NINE COLOR PUZZLE
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. 7. (5. a computer program established 133. 4. 4). Identiﬁcation of the second color of these ordered pairs is sufﬁcient to deﬁne all the colors.
5. For example. 6.105 possible color combinations. (5. (7.
6. 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. 9). and 9 one can logically deduce the ﬁrst colors by the three
. (1. This results in requiring only thirty cubes and. 9) and (8.
cannot be interchanged with other colors. on the ends of the tricube. Colors 1 and 3. Color 2. 35. and 122. 6). can be swapped with each other. and then assigning a printable character to each. 6). (4. 3). Any permissible interchange of colors is an isomorphism. 8). (24).. 9). The use of this six-character identiﬁcation greatly simpliﬁed the process of sorting the 10. then additional criteria for recording solutions are recommended. and (99). Thus this color combination is identiﬁed as /:Z[yz. 68. and six at a time including a triplet of two at a time. (35). The remaining six colors may be interchanged in numerous ways: two at a time. (1. The head of the list then became . but not with any of the other colors. 58.160
S. the ﬁrst entry in Table 1 is . The ASCII code for 23 is not a printable character. 67. which is in the center of the tricube. and (8. 5). Subtracting twenty-three results in: 23. Identiﬁcation was further simpliﬁed by placing these numbers in pairs (see Note below). 9). four at a time. Note: The smallest possible number is 23 and the largest is 99.691 color combinations obtained by the algorithm described in Section 5 were listed in order according to their ASCII characters. (5. (98). If the solutions are to be catalogued. by adding twenty-three to each and using the ASCII character set codes (ASCII stands for American Standard Code for Information Interchange). . (2. and 99. 8). 9). pairs of three at a time and four at a time with two at a time. 7). has sixty isomorphisms. (4. Decoding a name is quite simple. The ﬁrst entry. This process then identiﬁed the 148 disjoint sets of isomorphisms. (4.:Z[yz. 91. 3). The 10.691 cases into 148 disjoint sets of isomorphisms. (6. (68). 90. These were removed from the list. its isomorphisms are determined and removed from the list. (7. ﬁve at a time including three at a time with two at a time. For example. (67). An inspection of the ASCII table will explain why it was decided to add 23. The process was then continued until the list was exhausted. The solutions are identiﬁed by a three-digit number indicating the color of the buried cube and the two “edge” colors. Not all of these mappings produce a new isomorphism. Shown in Figure 4 are an edge solution and a center solution for the color combination of Figure 1.:Zppp. three at a time.
. The ASCII codes for these characters are: 46. FARHI
cubes of each color requirement and the restriction that the second number be greater than the ﬁrst. (7. 121. (5. Different solutions often have the same three-digit identiﬁcation. including pairs of two at a time.:Z[yz. 98. It is then a simple process to recognize that this represents the dicubes colored (1. 2).e. i.
Often two dicubes may be swapped with two others having the same colors. 6) and (7. J. Paris.THE NINE COLOR PUZZLE
Figure 4. Abrams. and Botermans. hardcover. and Torbijn.
The edge solution is interesting in that new solutions are often obtainable by using two transformations.
. Typical solutions. Puzzle sets or puzzle solutions may be obtained by contacting the author by E-mail at sivy@ieee. 200 pp. For example. CEDIC. Using these transformations. 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. Whenever the tricube is on an edge and shares a plane with only three dicubes. New York. 7) and (6.. softcover. Distracts 4. in French. J. Jerry.. Stan Isaacs has suggested that graph analysis would be useful in determining the number of solutions.
 Slocum. the plane can be translated to the other side resulting in a new solution. 9) in the bottom layer may be swapped with dicubes (4. France. 1977. 176 pp.  Meeus. Harry N. Unfortunately. Creative Puzzles of the World. Polycubes. 1978. the reader should now be able to determine ﬁve more edge solutions. P.org. all 148 graphs have not been compared. Jack. dicubes (4.. 9) on the upper right.
Figure 2.” The name was chosen because there are two quite different puzzles involving two different blocks numbered 2(a) and 2(b). to “Twice. Italy.
There are nine square blocks (including two numbered 2).Twice: A Sliding Block Puzzle
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. Blocks A. and 2(b) have a single horizontal channel. The puzzle was invented by Dario Uri from Bologna. cannot get there at all. 7.
Figure 1. Blocks can either have a horizontal channel. a vertical channel.
Fixed into the base of the board are four pegs. Channels (or grooves) are cut into the bases of some blocks. one being used in each puzzle. allowing them to pass over the pegs in the corners (see Figure 2). in some cases. one in each corner (see Figure 1). but only eight are used in each puzzle. 163
. and was originally issued in 1989 with the name “Impossible!!” Subsequent additions and improvements made over the next couple of years led to a change of name. or no channel at all. and these blocks can only reach the corners from a horizontal
Written by Edward Hordern and reproduced with the kind permission of Dario Uri. both channels.
The delight of the ﬁrst puzzle is that it is very easy to move block A to just above the bottom right corner. So near. however. Blocks 3 and 6 have no channels and cannot go into any corner. and the second puzzle uses block 2(b).164
. It is quite an achievement just to solve them. a plan can be made as to which blocks have to be moved so as to allow block A to pass. they must only move in a cross-shaped area.
Figure 3. Both solutions are believed (but not proved) to be minimum-move solutions. For the expert. Once this has been mastered. This causes something of a trafﬁc jam. HORDERN
direction. can only move freely up and down the center of the puzzle.
Figure 3 shows the start position. Since they can’t go into the corners. The object of the puzzle in each case is to move block A to the bottom right corner. The real “problem” blocks are block 7 in the ﬁrst puzzle and blocks 7 and 2(b) in the second. as well as block A. which has to be overcome…. and yet so far! In puzzle 2 it is quite a task to move block A more than a square or two. The ﬁrst puzzle uses block 2(a). These blocks. The puzzles are both rated as difﬁcult. Blocks 4 and 2(a) have a single vertical channel and can only approach the corners from a vertical direction. or to get anywhere at all!
Hints for Solving
In both puzzles the “nuisance” blocks are 3 and 6. the shortest known solutions are 50 moves for the ﬁrst puzzle and a staggering 70 moves for the second. Blocks 1 and 5 have both channels (in the form of a cross) and can go anywhere. the second being the harder of the two. 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. only to get hopelessly stuck….
16). all interactions between the four pieces take place in one plane. with only three moves needed to remove the ﬁrst piece. because these can be so constructed that several moves are needed to separate the ﬁrst piece. The ﬁrst design I saw is Jeffrey Carter’s.” a puzzle of mine. The same movements. depicted in A. Bill Cutler’s “Bafﬂing Burr” and Philippe Dubois’ “Seven Up” were the ﬁrst attempts. which occur in many different designs.
. until the two smaller pieces are free. the two large pieces move into each other along a zigzag line. One result is the “Zigzag” planar burr. It takes ﬁve moves to separate the ﬁrst piece. Since about 1985 a “Most Moves Competition” for six-piece burrs has been running. Oskar van Deventer
Usually “burrs” are considered to be three-dimensional puzzles. who was inspired by “Dead Lock. Muroi wrote me that his idea was inspired by a design of Yun Yananose. True planar burrs are rarely found. so in some sense it might be considered a planar burr. The most interesting are those with internal voids. in backward order. in Japan.” which requires twelve moves! Recently. The idea of a two-dimensional burr immediately appealed to me. will separate large pieces.Planar Burrs
M. an ingenuously designed puzzle that he called “Four Sticks and a Box. However. but nevertheless requires twelve moves (!) to get the ﬁrst piece out of the box. it cannot be realized as a planar burr because in two dimensions each piece should be disconnected. K. Dewdney’s Scientiﬁc American column (January 1986. depicted in Figure 1.” The puzzle has no more than four movable pieces. Though Muroi’s puzzle is three-dimensional. To solve the puzzle. I am not completely aware of the state of the competition. I received from Tadao Muroi. The puzzle is not very difﬁcult to solve. Carter’s puzzle has four pieces. The most common are the six-piece burrs. and in April 1986 I made some attempts to ﬁnd a design of my own. p. but as far as I know Bruce Love is the record holder with his “Love’s Dozen. This puzzle has two congruent large pieces and two congruent smaller ones.
Some months ago.
M. particularly inspired by Muroi’s “Four Sticks” (so closing the circle of mutual inspiration). which I have called “Nine and One-Half Moves. I took up the challenge again and succeeded in ﬁnding a new design of a planar burr. Zigzag. Nine and One-Half Moves. V AN DEVENTER
Figure 1. The three pieces of the puzzle form a square with internal voids.” This is is a true two-dimensional burr of only three pieces and it needs no less than nine and one-half moves to separate one piece from the other two.
. we can glue one piece of the puzzle between two square plates as indicated in the top left corner of Figure 2. By using opaque plates the design is hidden as well. The ninth move is a slide-plus-rotate move.PLANAR BURRS
The pieces. so I count it as a move and a half. and the moves required to separate the pieces are depicted in Figure 2. A round hole can be used to hide a coin. In order to prevent us three-dimensional people from cheating.
The following are my favorites:
unnamed (44) (Singmaster–Klarner): Box: ¢ ½½ ¢ ¾½ Pieces: (44) ¾ ¢ ¿ ¢ unnamed (45) (de Bruijn): Box: ¢ ¢ Pieces: (45) ½ ¢ ½ ¢
. The puzzles are grouped according to whether there are holes in the assembled puzzle and whether the pieces are all the same or different. It is 3-dimensional box-packing puzzles where the box and all the pieces are rectangular solids. All Pieces the Same
“Aren’t these puzzles trivial?. but complete solutions are not given. date of design. but the “tricks.Block-Packing Jambalaya
My primary interest over the years has been burr puzzles. but there are some interesting problems. If known.” you ask. Well. No Holes. David Klarner gives a thorough discussion of this case in . The tricks to most of the puzzles are discussed here. but there is another small category of puzzles that is especially intriguing to me. the inventor of the puzzle. Presented here are 11 such “block-packing” puzzles. you are not far from being completely correct. the total number of pieces is in parentheses. I know of no other small group of puzzles that encompasses such a rich diversity of ideas.
7.” or unique features that the puzzles employ are many and varied. and manufacturer are given. The number of such puzzles that I am aware of is quite small. For each puzzle.
(13) ½ ¢ ¾ ¢
In the ﬁrst of these. and has many solutions. ¢ ¢ Box 3: ¢ ¢ ½¼ Pieces: all 18 pieces from ﬁrst two boxes
. The 18 pieces are all different. it is easy to see that the cubes must be placed on a main diagonal. (1) ½ ¢ ¾ ¢ ¾. function similarly to the ½ ¢ ¾ ¢ ¾ pieces in the ﬁrst puzzle. ¾ ¢
¢ . ½ ¢ ¾ ¢ ¿ ¢ . the three ½ ¢ ½ ¢ ¿ pieces must be used sparingly. See  or  for more information. The solver must determine exactly where the second set of pieces must be placed. (1) ¾ ¢ ¾ ¢ ¾. but it does make a nice set of puzzles. ¿ ¢ ¿ ¢ ¿. and the three boxes offer a wide range of difﬁculty. 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 rest of the pieces. By analyzing “checkerboard” colorings of the layers in the box. The pieces of the second type are smaller and easier to use. ¾¢¿¢
8. and then the rest is easy. The large box is moderately difﬁcult. ¾ ¢ ¢ . although not exactly alike. ¾ ¢ ¿ ¢ . The middle-size box is difﬁcult — there is only one solution. ¾ ¢ ¿ ¢ . No Holes. (6) ½ ¢ ¾ ¢ ¾ unnamed (18) (John Conway): Box: ¢ ¢ Pieces: (3) ½ ¢ ½ ¢ ¿. rigorous analysis of the puzzle can be done by hand in about 15 minutes. Pieces Mostly Different
Quadron (18) (Jost Hanny. but must be used efﬁciently to solve the puzzle.
unnamed (9) (Slothouber–Graatsma): Box: ¿ ¢ ¿ ¢ ¿ Pieces: (3) ½ ¢ ½ ¢ ½. ½ ¢ ¿ ¢ . not counting rotations and/or reﬂections. ½ ¢ ¿ ¢ ½¼. No Holes. In the second design. the three individual cubes are obviously easy to place.¿¢¿¢ . but they must not be wasted. ½ ¢ ¿ ¢ . A complete. Naef): Box 1: ¢ ¢ Pieces: ¾ ¢ ¿ ¢ ¿.
Quadron does not use any special tricks that I am aware of.170
B. The seven pieces can be placed in the box in ten different ways.¾¢ ¢ .¿¢¿¢ ¢ . The small box is very easy.
Box 2: ¢ ¢ ½¼ Pieces: ½ ¢ ¿ ¢ .
When running these programs on more powerful computers. I wrote such a program on my ﬁrst computer. ¾¼ ¢ ¼ ¢ ¾. the rate increases to 40 pieces/second. ¢ ½ .450. The total CPU time used was about 8500 hours. The program displayed the status of the box at any instant using color graphics. ¢ . ½ ¢ ¼ ¢ . not counting rotations and reﬂections.BLOCK-PACKING JAMBALAYA
Quadron also makes for a nice entrance into the realm of computer analysis of puzzles and the limitations of such programs. but the process cannot be completed. By the end of the runs. ¢ ½ .
. and ¢ ½¿. ¢ ½¿. ¢ ½ . ½ ¢ ¢ ¼. Of the 18 pieces. a Commodore 64. One or two individual layers can be constructed. The analysis was done on about 20 powerful IBM workstations. ½ ¢ ¾ ¢ ¾. There are four solutions to the puzzle.
Boxed Box (23) (Cutler. I painted pieces of an actual model to match the display. 10 are too wide to ﬁt into the box sideways and 4 are of width 5. This leaves 4 pieces that might be placed sideways. I did a complete analysis of the third box.480 solutions. about once a second a piece is added or removed from the box! Using compiled BASIC. ¢ . ½ ¢ ¿ ¢ . There are 3. which is no good for this purpose. but there are more efﬁcient algorithms that can be used for block-packing puzzles. ¾½ ¢
¢ ½ ¼. and they all have three of these four pieces placed sideways. it is tempting to solve the puzzle by constructing three layers of pieces. copied from a model in the collection of Abel Garcia): Box: ¢ ½ ¢ ¾ Pieces: all pieces are of thickness 2 units. The Commodore 64 is such a wonderously slow machine — when running the program in interpreter BASIC. Bill Cutler Puzzles): Box: ½ ¢ ½ ¢ ½ Pieces: ½¿ ¢ ½½¾ ¢ ½ ½. all very similar. the machines had constructed 2 1/2 trillion different partially ﬁlled boxes.
Since all the pieces are of the same thickness and the box depth equals three thicknesses. In early 1996. ½½ ¢ ½½ and two each of ¢ . The second box was analyzed in about a minute of mainframe computer time.
Parcel Post Puzzle (18) (designer unknown. 1978. ¢ ½¿. The programmer can use algorithms that are used for pentominoe problems. ¢ ¾½. ¢ ½¼. ½¼ ¢ ½½. ½ ¢ ½ ¢ ¾ ¢ . the widths and lengths are ¢ . or the equivalent of one year on one machine. ¢ ½½. The result was a fascinating demonstration of how a computer can be used to solve such a puzzle. the difference between the three boxes is stunning: The ﬁrst box can be completely analyzed in a small fraction of a second. The solution involves use of the following obvious trick (is that an oxymoron?): Some piece(s) are placed sideways in the box. ¾¢ .
¾ ¢ ¿¼ ¢ ¢ ½¿ . See . ¿ ¢ ¾ ¢ ¼. this puzzle still fascinates me. There are many other 23-piece solutions that are combinatorially different from the above design. 1981. The dimensions of the pieces can be modiﬁed. ¾ ¢ ¢ ½¾¿. Simpliﬁed (15) (Cutler.
10. Holes. ½ ¢ ½ ¢ ¾½. basic. Almost 15 years later. ¾ ¢ ¿ ¢ . where the smallest is greater than one-quarter of the sum. ½¼ ¢ ½ ¢ ¾ . I like the dimensions above because it tempts the solver to stack the pieces three deep in the middle dimension. Bill Cutler Puzzles) Box: ¼ ¢ ¾ ¢ ¾ Pieces: ¢ ½ ¢ ¾ . 1976) Box: ½ ¢ ½ ¢ ½ Pieces: (27) ¢ ¢
This sounds like a simple puzzle. ¢ ½¼ ¢ ½½
11. ¢ ¢ ½½. Hikimi Puzzland) Box: ½ ¢ ½ ¢ ½ Pieces: Two each of ¢ ¢ ½¼. ½¾ ¢ ½ ¢ ½ . ¿½ ¢ ¢ . There are 21 solutions. ¾¿ ¢ ½ ¢ ¿. ¢ ¾¼ ¢ ¾¼. ¢ ½¼ ¢ ½½. ½¼ ¢ ½ ¢ ¿½. ¿ ¢ ½½¼ ¢ ½¿ . ¢ ¼. ¢
¢ . ½ ¢ ½ ¢ ¾½
The original design of Cutler’s Dilemma had 23 pieces and was constructed from the above. ¿ ¢ ¾ ¢ ½¾ . CUTLER
¾¾ ¢ ½¼ ¢ ½¿½. The box is a cube with side equal to the sum. They can be any three different numbers. ¿¿ ¢ ¿ ¢ ¿ ¢ ½¾½. ¢ ¢
The dimensions of the pieces are all different numbers. The smallest number for which this can be done is 23. The net result was a puzzle that is extremely difﬁcult.
. The pieces ﬁt into the box with no extra space. The extra space makes available a whole new realm of possibilities. See  or  for more information. but it is not. none having any symmetry or pattern. ½½ ¢ ½ ¢ ¾ . ½½ ¢ ½¾ ¢ ¾. Holes. ¢ . ½¼ ¢ ½½ ¢ ¾. I will not say anything more about this design except that the trick involved is different from any of those used by the other designs in this paper. ½½ ¢ ½ ¢ ¿¼. ½½ ¢ ½ ¢ ½ . ½¼ ¢ ½¾ ¢ ¾ . ½¼ ¢ ½ ¢ ¾ . 1986.172
B. Pieces Different
Hoffman Junior (8) (NOB Yoshigahara. ½½ ¢ ½½ ¢ ¾ . Pieces the Same or Similar
Hoffman’s Blocks (27) (Dean Hoffman. version by cutting some of the pieces into two or three smaller pieces.
edited by D. This should keep him busy for quite some time!
 W.  D. 20–23.” J. 1981). It is a valuable weapon in every puzzle collector’s arsenal. this is not another oxymoron. including the one in your hand.
.¿ ¢¾ ¢ . being sure that the unﬁlled space is concealed at the bottom and is stable. 104–111. Pack all the pieces. Klarner. 212– 225. Recreational Math. into the box. 122–127.
The Melting Block is more of a paradox than a puzzle. 12(2).½ . pp. Spring 1973. “Brick-Packing Puzzles.  M. except one. This ﬁts into the box with a little room all around.” J. Miscellaneous
Melting Block (8-9) (Tom O’Beirne) Box: ¢ ¢ ½¿¿ Pieces: ½ ¢ ¾ ¢ . By the way. but rather a tautology to the 99% of the world that would never even have started to read this article. but seems to the casual observer to ﬁll up the box completely. Place the box on your puzzle shelf with the remaining piece hidden behind the box. Recreational Math. I won’t say which one (it should be easy to ﬁgure out). You are now prepared for your next encounter with a boring puzzle-nut. (No. and then dump the pieces onto the ﬂoor. (This second construction is a little more difﬁcult. February 1976.¿
¢ ¢ ¢ ¢
. Hoffman. 112–117.” in The Mathematical Gardner. Cutler. When the ninth piece is added to the group. 6(2). Gardner. Mathematical Games column in Scientiﬁc American. pp. Mathematical Games column in Scientiﬁc American. pp. the pieces can be rearranged to make a ¢ ¢ ½¿¿ rectangular solid. one of the puzzles listed above is impossible.. Gardner. readers. The eight pieces ﬁt together easily to form a rectangular block ¢ ¢ ½¿¾.BLOCK-PACKING JAMBALAYA
12. pp. being careful to keep the renegade piece hidden from view. Klarner (Wadsworth International.) This is a great puzzle to show to “non-puzzle people” and is one of my favorites.  D.¿ ¢ ¢ plus a second copy of ½ ¢ ¾ ¢
. “Packing Problems and Inequalities. pp. Show off the solved box to your victim. February 1979. “Subdividing a Box into Completely Incongruent Boxes.) Pick up the box and last piece with both hands. ½ ¢ ¾ ¢ ½ ¢ ¢ . 1979–80.  M.
Written on the occasion of the Puzzle Exhibition at the Atlanta International Museum of Art and Design and dedicated to Martin Gardner.174
. The disassembly and/or reassembly of the cage remains the primary function of the puzzle. Consider a three-dimensional (3-D) interlocking assembly in the form of a cage with a ball in the center. while examples are given for most groups.g. but most attempts so far have either been far too specialized in application or too general to provide the basis for a deﬁnitive classiﬁcation. etc.Classiﬁcation of Mechanical Puzzles and Physical Objects Related to Puzzles
James Dalgety and Edward Hordern
Background. related to such objects. go to http://puzzlemuseum.) Deﬁnitions. An interlocking puzzle should be classiﬁed according to its interior construction. a wooden cube. In cases where it seems possible to place a puzzle in more than
For updated information and illustrations. sphere. To provide a logical and easy-to-use classiﬁcation to enable nonexperts to ﬁnd single and related puzzles in a large collection of objects. some knowledge of the subject is required. “Mechanical Puzzles” is the descriptive term used for what are also known as “Chinese Puzzles. or teddy bear may all have similar Cartesian internal construction and so should all be classiﬁed as Interlocking– Cartesian). A mechanical puzzle is a physical object comprising one or more parts that fall within the above deﬁnition. 175
. and Jerry Slocum. A puzzle is a problem having one or more speciﬁc objectives. rather than its outward appearance (e. contrived for the principle purpose of exercising one’s ingenuity and/or patience. but particular thanks are due to Stanley Isaacs. Objective.” Several attempts have been made to classify mechanical puzzles. and patents. books. 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. Method. David Singmaster. barrel. (As presented here..com. Many people have provided a great deal of help. A puzzle should be classiﬁed by the problem that its designer intended the solver to encounter while attempting to solve it.
It is understood that specialist collectors will further subdivide the subclasses to suit their own specialized needs. crossword puzzles). however. For example. usually. HORDERN
one category. have signiﬁcant freedom of movement in relation to each other. Tanglement–Rigid & Semi-Rigid is awaiting a thorough study of the topology of wire puzzles. such as “INT-BOX. The linked parts. was contrived only to educate and amuse. although jigsaws with sloping cuts in fact have a relevant third dimension.g. if the route has been found.
.. A puzzle will be referred to as two-dimensional (2-D) if its third dimension is irrelevant (e. unless they are on or part of some physical object. The full abbreviations consist of three characters. while perhaps puzzling the infant. unlike the parts of an interlocking puzzle. Tanglement Puzzles (TNG) have parts that must be linked or unlinked. The pieces have L-shaped grooves.176
J. It would be classiﬁed as Route-ﬁnding because. thickness of paper or plywood or an operation involving a third dimension such as folding). It is based on a ¿ ¢ ¿ sliding block puzzle under a clear plastic top. DALGETY AND E. A few puzzles may have to be cross-referenced if it is absolutely necessary. and Route-ﬁnding. one category will be dominant. It will be noted that the deﬁnition of a puzzle excludes the infant’s “posting box. it also excludes the archer attempting to get a bull’s-eye. Most standard jigsaws are 2-D. so they must be classed as 3-D.” which. A good example of multiple-class puzzles is the “Mazy Ball Game” made in Taiwan in the 1990s. then the dexterity and sequential movement must also have been achieved. Thus the puzzle requires Dexterity. Also excluded are puzzles that only require paper and pencil (e. 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. and the blocks themselves must be slid around so that the ball can exit at the top left.
The fourteen main classes are as follows:
¯ ¯ ¯
Dexterity Puzzles (DEX) require the use of manual or other physical skills in their solution. the exercise of whose ingenuity is entirely incidental to the original warlike intent of the sport. plus up to four characters. Sequential movement. which may be ﬂexible. hyphen. Routeﬁnding Puzzles (RTF) require the solver to ﬁnd either any path or a speciﬁc path as deﬁned by certain rules.g..” These are the standard abbreviations for the classes that have been chosen for relative ease of memory and conformity with most computer databases. it must be classiﬁed in whichever is the most signiﬁcant category.
Measuring. close it. undo it. Pattern Puzzles (PAT) require the placing or arrangement of separate pieces of a similar nature to complete patterns according to deﬁned rules. a padlock and its hasp. Other Types of Mechanical Puzzles and Objects (OTH). Interlocking Puzzles (INT) interlock in three dimensions. Jugs and Vessels (JUG). The pattern may be color. They may clip together but do not interlock in 3-D. faces of a cube. or folding. The pattern required may be the matching of edges of squares. or otherwise get it to work.. Logic. pouring. Ambiguous Pictures and Puzzling Objects (AMB). i. Sequential Movement Puzzles (SEQ) can be solved only by moves that can be seen to be dependent on previously made moves. nor do they involve general assembly/disassembly of parts that interlock in 3-D. Vessels having a mechanical puzzle or trick in their construction that affects the ﬁlling. This group is for puzzle objects that do not easily fall into the above categories and cannot be categorized into sufﬁciently large groups to warrant their own major class.CLASSIFICATION OF PUZZLES
¯ ¯ ¯ ¯
¯ ¯ ¯ ¯
Opening Puzzles (OPN) are puzzles in which the principle object is to open it. Puzzles in which something appears impossible or ambiguous. shape. and the principle objective is to restore them to their unique original form. They are solved by hinging. Folding and Hinged Puzzles (FOL) have parts that are joined together and usually do not come apart. or a nut and bolt. Math. Some have a container and are posed as packing problems. Cutting. Many clip-together puzzles are “non-interlocking. Jigsaw Puzzles (JIG) are made from cut or stamped-out pieces from a single complete object. the differences must be sufﬁciently minor so as not to obscure the similarity of the pieces. They usually comprise a single object or associated parts such as a box with its lid. etc. Trick. Mystery.e. Also. remove something from it. The mechanism of the puzzle is not usually apparent. ﬂexing. provision is made for puzzles pending classiﬁcation. texture. Where the pattern is due to differences in shape. etc. Included in this group are Balancing.” Assembly Puzzles (Non-Interlocking) (ASS) require the arrangement of separate pieces to make speciﬁc shapes without regard to the sequence of that placing. or the pieces are mutually selfsustaining.
. one or more pieces hold the rest together. or drinking therefrom. and Theoretical Puzzles. magnetic poles.
while not strictly puzzles. etc.178
J. 3-D. triangular. This has resulted in an unwieldy list. Number of Pieces. need to be classiﬁed as part of the collection. etc. and any group/non-group moves. Linked: Pieces joined together by hinges. 2-D. or partly used in solution. 3-D. the type of structure. strings. Required for INT JIG ASS PAT RTF SEQ FOL. HORDERN
Ephemera (EPH). We are now working on deﬁning what should be included automatically in the subclasses where they are relevant. and the standard Stellated Rhombic Dodecahedron has a six-piece Diagonal Cartesian structure. where XXX is the main class and YYYY is the sub-class. which. Required for INT JIG ASS PAT.
Prior to 1998 the subclasses attempted to incorporate the number of dimensions. 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. Group Moves. 2-D on 3-D. Type of Piece Structure. We hope to reduce the number of sub-classes substantially in the near future. Whether needed. ribbons. Required for SEQ. Required for INT ASS PAT. Examples of puzzles in each class are given in the right-hand column.
Thus a standard six-piece burr uses square rods in a Cartesian structure. Number of Dimensions. 2-D. and 4-D. 2-D to 3-D. hexagonal. Puzzle Class Abbreviations (PZCODE) are standardized to a maximum of eight characters: XXX-YYYY. This category has been included because most puzzle collections include related ephemera.
. A detailed classiﬁcation with second-level classes is given in the separate table. DALGETY AND E. not needed.
Good. i. **required Class + Subclass **required Theme or advertisement (subject) Materials Dimensions A. Generalized Information
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
Generic name of puzzle or objective if not obvious. Fair. biggest piece. envelope.. Source Date Cost Insurance value
Information Speciﬁc to This Object
Information Speciﬁc to This Collection
. C. i. **A required for scale What the dimension refers to. electric. clockwork. In practice.e. box. puzzle may have a cracked glass top or be missing one piece. d = diameter.e. references.. designer Manufacturer or publisher’s name and country Type of manufacturing. solar Patent and markings Notes..e. i. Poor) Condition qualifying note. mass produced (over 5000 pieces made).. without paid curators.CLASSIFICATION OF PUZZLES
Guide to Making a Catalogue or Database for Puzzle Collections
Headings for cataloguing puzzle collections could include the following items.e. if known Number of set in collection A picture or photograph Location/cabinet/bin Acquisition Number Condition (Excellent. i. assembled puzzle If powered. but otherwise be in excellent condition. 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. craft made (commercial but small volume). it is probably advisable to be selective and limit the amount of information recorded. battery. B.
visiting places en route. “Pike’s Peak” Icosian Game. Most hedge mazes
.” puzzles using tops Pentangle “Roly-Poly” puzzles Ramps.” Engel’s “Black Box”
“Frying Pan.” “Theo der Turnier. Mercury manipulation Water-ﬁlled puzzles View by mirror Tomy’s “Pocketeers”
DEX-BALL DEX-SDRY DEX-LQOB DEX-INLQ DEX-MIRR DEX-MECH DEX-TOOL DEX-RTFL DEX-HIDD DEX-ELEC DEX-PINB DEX-OTHR RTFRTF-CHNG
RTF-STEP RTF-UNIC RTF-SHOR RTF-CPLX
Dexterity. 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.” colour mazes. “Le Pendu. plain balls into holes Dexterity with sundry objects and/or obstacles Liquid objects Dexterity in liquid Indirect viewing Mechanized Using tools and magnetic tools Route following dexterity Objects concealed from view Electrica and electronic dexterities Pinball-related dexterities Other dexterities. jumping beans. number totalling mazes. avoiding objects.” “Tandem Maze” (complex).” “Yankee.” Tomy’s “Crazy Maze. bridges. DALGETY AND E. “Bootlegger” (complex) Ring and hole mazes. HORDERN
Detailed Puzzle Classiﬁcation
Dexterity Puzzles Dexterity or other physical skills in their solution
Cup and Ball. Konigsburg Bridges “Worried Woodworm.180
ball in cube of cubelets
TNGTNG-RIGI TNG-R&F TNG-FLEX OPNOPN-BOX OPN-LOCK OPN-HID
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.” Chinese rings. pens. “Mayer’s Cube. “Tak-it-Apart. Soma. gears. poison rings Nuts and bolts. O’Beirne’s “Melting Block”
.” “Margot Cube. knives.” “Nine of Swords. “T” puzzle Polycubes.” keychain cars “Plato’s Plight”
Interlocking Puzzles Boxes that disassemble Cartesian (has parts along three mutually perpendicular axes) Interlocking polyhedral Other rigid shapes
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. “Hoffman” cube. Dalgety’s “Devil’s Halo” Leather tanglement puzzles Boxes. cast “ABC. purses Padlocks Chippendale tea chests. puzzle rings Hess wire puzzles. berrocals. Pentominoes. checkerboards.” “Hazelgrove Box” Strijbos aluminium burr box Burrs. cutlery.” Cutler’s burr in a glass Cofﬁn’s “Saturn” JWIP and keychain animals. Oscar’s “Dovetail.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. Oskar’s keys.
” “Even Steven. DALGETY AND E. Squashed Soma. Apple and Worms. and weaving 2-D patterns Patterns with 2-D parts on 3-D surface
JIG-PART JIG-SLOP JIG-LAYR JIG-2D3D JIG-BLOX PATPAT-2DPG
“Sculpture Puzzles” “Toyznet”
Queens on chess board.” Loyd’s Donkeys. Jensen’s “Tricky Laberint” Magic squares. Waddington’s “Black Box” Match puzzles.” Gannt’s “Chandy” Can include puzzles double-sided
JIGJIG-STD JIG-3D JIG-2DID
Jigsaw Puzzles Standard jigsaws 3-D jigsaws 2-D jigsaws with identical pieces for secondary objectives 2-D jigsaws which only partially cover the plane 2-D jigsaws with nonperpendicular/sloping cuts Multiple-layer 2-D jigsaws 2-D jigsaws with parts that can be made into 3-D objects Picture cubes/blocks Pattern Puzzles Pattern arrangements of points. overlapping. nine-piece ivory cube Pack the Plums. “Lapin. weaving puzzles “Dodeca”
.182 ASS-POLY ASS-SHAP 3-D Assembly Polyhedra and Spheres 3D Assembly Other Shapes
J. pegs. according to predetermined rules Patterns of sticks Pattern: arrangements of number 2-D matching edge patterns Arrangement of 2-D pieces on 2-D surface to make a pattern according to predetermined rules Stacking. number puzzles Heads and tails “Testa”
PAT-STIX PAT-NUM PAT-2DEG PAT-2D
Stacking transparent layers.” “Phoney Baloney. Josephus. or pieces. “Managon. BlackBox. HORDERN Ball pyramids.
” “Spin Out” “Why Knots. “Masterball.CLASSIFICATION OF PUZZLES PAT-3D 3-D pattern puzzles with separate parts
183 “Instant Insanity. Oscar’s “Solar System. Panel Nine.” “Flexicube” “Clinch Cube”
.” “Backspin. Fit Pz “Wolf. Tomy’s “Great Gears” Raba’s “Rotascope.” “Missing Link.” Laker Cube Bognar’s planets. “Dodecahedru” — rotating faces of Dodecahedron Psychic Pz.” Rubik’s “Clock” Rubik’s Cube.” Skor Mor “Instant Indecision. Rubik’s “Magic.” “Turntable Train”.” Waddington’s “Kolor Kraze.” Chinese balls in ball..” “Jugo Flower.” “Orbit. counter and peg moving puzzles 15s Pz.” strung cubes Flexagons.” “Kaos” Rolling eight cubes “The Brain. Tit-Bit’s “Teasers” “Inversions”
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-ROLL SEQ-MMEC FOLFOL-SING FOL-HGOP FOL-HGCL FOL-HSEP
“Tower of Babel. Tower of Hanoi.” Mobius strips Rubik’s “Snake.” “Hexadecimal. Sheep and Cabbage” Solitaire.
ﬁve square puzzle 19th century riddle prints Anagrams. “Jail Nixon” Strip polyhedra
Standard “block holes and suck” solution
“Jolly Jugs” Cadogan teapots. crosswords on bathroom tissue
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. or those that can only be represented on a computer
. Archimedes gold Cork for three holes. 12 Golf Balls. Chinese winepots Royale’s patent
Luminations “Columbus Egg” Jugs and liquids. rebus plates and prints. HORDERN Map folding. DALGETY AND E.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
hold to light advertisements Obscure non-rebus heiroglyphic prints Requiring red and green glasses for either 3-D or movement effects Framed strip prints showing different views from different directions
. OHOs.” “Find the 5th Pig” needing coloured overlays. weight illusions
Distortions Archimboldesque objects
Hidden image pictures (no manipulation required) Hidden image pictures (manipulation required)
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
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. soot on unglazed part of ashtray Landscape turned to make a portrait. Topsey Turveys. courtship/matrimony Optical illusions.” Penrose triangle “Vanishing Leprechauns. Oskar’s “Escher Puzzle.” Hooper’s paradox Anamorphic pictures Pictures and objects of one subject made from completely unrelated objects Devinettes (obscure outlines). impossible dovetails.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.” random dot stereograms “Naughty Butterﬂies. “Spot the Difference.
DALGETY AND E.186 EPH-OTHR XXX-XXX DEL Other puzzle-related ephemera Lost records Deleted record
For database integrity only Puzzle disposed of. no longer in collection
is certainly 50 (since Ü is then 50). if y is greater than x). Propositions 1 and 2 can’t both be true! Which of the two propositions do you actually believe? Most people seem to opt for Proposition 2. suppose Ý. Proposition ¾. I will now prove the following two obviously incompatible propositions. the excess of x over y. is greater than the excess of y over x. is 100. is equal to the excess of y over x. 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. and the excess of Ý over Ü. if Ü is greater than Ý.A Curious Paradox
Consider two positive integers Ü and Ý. if x is greater than y. if y is greater than x.
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 ¾ Ý¸ Ò Ø Ü ×× ½ Ýº Ì Ù× Ø Ü ×× Ó Ý ÓÚ Ý Ö Ü¸ Ý × Ö Ø Ö Ó Ý ÓÚ Ö Ü × Ø Ò Ý ½ ½ Ýº Ë Ò Ý ¾× Ö ¾ Ø Ö Ø Ò ½ Ý¸ Ø × ÔÖÓÚ × Ø Ø Ø Ü ×× Ó Ü Ø Ò Ü¸ × ¾ ¾ 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.. And isn’t 100 surely greater than 50?
. if Ý is greater than Ü. say. Now. The two amounts are really the same (i. The excess of x over y. Proposition ½. Then the excess of Ü over Ý. if x is greater than y. is certainly 100. one of which is twice as great as the other.e. We are not told whether it is Ü or Ý that is the greater of the two. But look.
Theorem 1. All rights reserved. no incomplete investigation is unbiased. No. no apathetic people are human beings. Therefore. Every unbiased investigation is an impartial investigation. Reprinted with permission. While his logic is impeccable. 191
. introduced by Dodgson. Proof. our next objective is to broaden the approach to encompass other models of mathematical proof. Charles Lutwidge Dodgson. Numerous additional examples. p.A Powerful Procedure for Proving Practical Propositions
Solomon W. 383.
Reprinted from Mathematics Magazine. Golomb
The eminent Oxford don. 5. All dragons are uncanny. For example. Theorem 2. The purpose of this note is to exploit this powerful proof methodology. the conclusion (that no Scotsmen are dragons) is not particularly surprising. Vol. Therefore. All incomplete investigations are biased. 67. Therefore. with special emphasis on obtaining conclusions having political or moral signiﬁcance. December 1994. Proof. All human beings are different. closely following Dodgson’s original mod-el. Apathetic people are not human beings. no Scotsmen are dragons. All apathetic people are indifferent. Every incomplete investigation is a partial investigation. demonstrated the applicability of formal mathematical reasoning to real-life situations with such incontrovertible rigor as evidenced in his syllogism: All Scotsmen are canny. could be adduced. to a broader range of human experience. However. nor does it shed much light on situations that we are likely to encounter on a daily basis.
W. Finding further theorems of this type. 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. If a government is arbitrary. Theorem 3. and extending the method to other models of mathematical proof. it is sufﬁcient to prove it for an arbitrary government. Proof. We exploit this to obtain the following important result. is left as an exercise for the reader.192
S. it is obviously unjust. And since this is true for an arbitrary government. it is true for all governments.
. All governments are unjust. To prove the assertion for all governments.
for example.Misﬁring Tasks
Years ago I went on a collecting trip. A devilishly difﬁcult kind of mathematical problem. except for the accumulated webs and dust. each can be accomplished for some hit-and-miss pattern of integer values until. can be done for all Ò Æ . What is a task parameterized in n that last misﬁres for n = 6?. 7. Martin Gardner gave me some help. not for example “prove the impossibility of ”. others scraped together a token from here and there. using all of its area. More precisely. visiting non-innumeretic friends asking for contributions of a very particular kind. things run smoothly. the task must not be designed in an ad hoc way with the desired answer built in — must not. after some last “misﬁring” value. There are mathematical tasks like that: Parameterized in n. divide it into n non-overlapping subsquares). 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. or for n = 9?. not from a generative construct such as “divide Ò things into groups of 17 and 5. I browsed. not as much as I expected. e. involve an equation with poles or zeros in just the right places. it often misﬁres a few times before running smoothly. 8.”
. When a cold engine starts. it must be positively stated. but the bottom of my basket ended up barely covered and.g. I thought. 6.. to be sporting.
and. I pilfered..e. One such task is to partition a square into n subsquares (i. might be to pose such a question in reverse. it must at least seem to have originated innocently. where 5 is “missing” from the sequence. so it remains. I pass it around once more for contributions. etc. This can in fact be performed for n = 4. before throwing it out.
its members do exhibit a special charm. KNOWLTON
For a mathematician or logician. 3. (Generally known. 6. the next have 8.. has such a severely constrained answer that it is almost impossibly difﬁcult. such a partition for 11 is 2. 7. well known. (Ron Graham is the only person I know who would ¾ ¿ think of such a problem. But “difﬁcult” may be the wrong word. and who did. edges. Divide a rectangle into n disjoint subrectangles without creating a composite rectangle except for the whole (Frank Sinden). for me at least. The tetrahedron has 6 edges. proved by Ron Graham. ask themselves “Why is 9 missing?”) Place n counters on an inﬁnite chessboard such that each pair exhibits different numbers of row-mates and/or different numbers of column-mates. and 77.194
K. I have answers only for N = 5. 11. even after developing a complete proof. Cut a cube into n subcubes (called the “Hadwiger” problem.g. of course. Zbinder who demonstrated a dissection into 54 subcubes). or 9. 47. my guess is that the matter lies somewhere on the non-crispness scale between agreement on “proof” and agreement on “elegance. But a misﬁre problem. these are attrociously imprecise speciﬁcations. tracked for years by Martin Gardner. 9.” Meager as the current collection is. ﬁnally clinched independently by Doris Rychener and A. and who then went on to prove that 77 is the largest integer that cannot be so partitioned. with a rather obvious proof: Partition a square into n subsquares. Design a polyhedron with Ò edges. The trouble is that a misﬁre problem is
.) Cut a square into Ò acute triangles with clean topology: no triangle’s vertex may lie along the side of another triangle. others quite esoteric:
´ µ ´ µ ´ µ ´ µ
Stated above. 12 . but the stated answer implicitly but erroneously suggests that the task can be performed for any Ò). what’s the question?” Usually such a setup is too wide open to be interesting. (Charles Cassidy and Graham Lord. 9. a sociological rather than mathematical question. One of Martin Gardner’s books contains essentially this problem as a wire-identiﬁcation task. e. Possible for 8. as I think I have deﬁned it. But here. conjectured by Ken Knowlton. (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. Partition n into distinct positive integers whose reciprocals sum to one. 10.)
What we have here is an inﬁnite number of problems of the form “Here’s the answer. 6 since 2 + 3 + 6 = 11 and ½ · ½ · ½ ½. Some of these are well known. 5.
a math problem with no implied search process whatever. for Ò ½¿. Instead of searching directly for a task last misﬁring. Or arguments as to why this is too mushily stated a challenge. I invite further contributions to the collection. except to review all the math and geometry that you already know. it’s much more likely that in the course of your mathematical ramblings you will someday bump into one. say.
Drawing de Bruijn Graphs
The simplest kind of de Bruijn graph.
Figure 1. two edges directed out and two edges directed in. just subtract ¾ Ò from it to bring the number back into the range from ¼ to ¾ Ò ½. in which the number of nodes is a power of two. Out from the node labeled Ü the two directed edges go to the nodes labeled ¾Ü and ¾Ü ·½. Ò 197
. When the number of nodes is ¾ Ò we can label them with the integers from ¼ to ¾ Ò ½ and notice that they correspond to all the numbers expressible with Ò bits (Ò binary digits). at each node. If ¾Ü or ¾Ü ·½ is bigger than ¾Ò ½.
Pictures up to Ò can be found in the classic “Shift Register Sequences” by Solomon W.
Figure 2. the numbers from ½ to ¾ Ò ¾ in order going south down the Greenwich Meridian. Edges out from ¼ to ¾ Ò ½ ½ on the meridian go east to the dateline. “Mathematical Games. See Scientiﬁc American. The earliest reference I know of to the m-pire (empire) problem is P. because Hal Fredricksen told me he had not seen it in any book. The scheme would be to put ¼ on the North pole. Vol. pp. 2.
½In 2-pire fashion each node of the graph is allowed to appear in two places in
. The picture for Ò suggests 2-pire fashion½ for drawing the de Bruijn graph on the sphere without any lines crossing each other. while edges out from ¾ Ò ½ to ¾Ò ½ go west. February 1980. No. TAYLOR
There is reason to think that the picture for Ò may ﬁll a gap in the existing literature. 242. J.” by Martin Gardner. Heawood’s 1891 paper in the Quarterly Journal of Mathematics.
. and down the international date line. available from Aegean Park Press. ¾Ò ½ on the South pole. 14–22.
Bigger de Bruijn graphs get really hairy.
A player who cannot make a legal move loses. The initial position of the game consists of a number of points called spots. regardless of the opponent’s strategy. Since draws are not possible. Shown below is a sample game of two-spot Sprouts. Furthermore. then a graduate student. and Daniel Sleator
Sprouts is a popular (at least in academic circles) two-person pen-and-paper game. either the ﬁrst player or the second player can always force a win. and John Horton Conway. a professor of mathematics. Guy Jacobson. A single existing spot may serve as both endpoints of a curve. Impartial games have variants where the condition of victory is inverted: The winner is the
Figure 1. and the last player to make a legal move wins. Each curve must be drawn on the paper without touching itself or any other curve or spot (except at end points). a spot may have a maximum of three parts of curves connecting to it. with the ﬁrst player winning. Which of the players has this winning strategy depends on the number of initial spots.Computer Analysis of Sprouts
David Applegate. Most people (including us) learned about this game from Martin Gardner’s “Mathematical Games” column in the July 1967 issue of Scientiﬁc American. 199
. adding a new spot on each curve drawn. Sprouts is an impartial game: The same set of moves is available to both players. It was invented in Cambridge in 1967 by Michael Patterson. Players alternate connecting the spots by drawing curves between them. A sample game of two-spot Sprouts.
G. grows. Many sprouts positions that occur during the search are the sum of two or more non-interacting games. Many seemingly different Sprouts positions are really equivalent. the complexity of the problem overwhelms human powers of analysis. Each additional spot adds between two and three turns to the length of the game. computers have come a long way since then.and spaceefﬁciency to solve the larger games. We used standard techniques to speed adversary search. Conway said that the analysis of seven-spot Sprouts would require a sophisticated computer program. our program is the only successful automated Sprouts searcher in existence. Our program makes use of these sum identities when evaluating normal Sprouts. Of course. Our representation strives to keep only enough information for move generation. These ideas are not nearly as useful in analyzing misère Sprouts. After many hours of late-night hacking and experimenting with bad ideas. AND D. and that the analysis of eight-spot Sprouts was far beyond the reach of present-day (1967!) computers. SLEATOR
player who cannot make a legal move. caching the results
. We have written a program that determines which player has a winning strategy in games of up to eleven spots. performed by Denis Mollison (to win a 10-shilling bet!). and in misère games of up to nine spots. This is the principal reason that we are able to extend the analysis of the normal game further than the misère game. such as cutting off the search as soon as the value is known. Games with small numbers of spots can be (and have been) completely solved by hand. The ﬁrst proof that the ﬁrst player loses in a six-spot game. JACOBSON. This is called the losing or misère version of the game. ran to 47 pages. n. Our program is successful for several reasons:
We developed a very terse representation for Sprouts positions. Sometimes it is possible to infer the value of the sum of two games given the values of the subgames. APPLEGATE. The combination of this low-information representation and hashing (whereby the results of previous searches are cached) proved to be extremely powerful. we managed to achieve sufﬁcient time.200
D. positions can become fantastically complicated as the number of spots. and also increases the number of moves available at each turn signiﬁcantly.
Our Sprouts Program
As far as we know. While Sprouts has very simple rules. but as the number of spots increases.
The data for misère Sprouts ﬁt a similar pattern.cs. and an asterisk indicates a new result obtained by our program.edu/~sleator
. our computational resources. or 5 modulo 6.cmu. The ﬁrst player has a winning strategy in n-spot Sprouts if and only if n is 3.COMPUTER ANALYSIS OF SPROUTS
of previous searches in a hash table. and our ingenuity. Sleator believes in the Sprouts Conjecture and the Misère Sprouts Conjecture.½
½For a paper describing these results in more detail. and searching the successors of a position in order from lowest degree to highest. The size of the hash table turned out to be a major limitation of the program.
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£
A “1” means the ﬁrst player to move has a winning strategy. 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. We are still left with the nagging problem of resolving a bet between two of the authors. The only remaining case required to resolve the bet is the 10-spot misère game. We discovered that saving only the losing positions reduced the space requirement by a large factor. The n-spot Sprouts positions evaluated so far fall into a remarkably simple pattern. characterized by the following conjecture: Sprouts conjecture. This problem seems to lie just beyond our program. The ﬁrst player has a winning strategy in n-spot misère Sprouts if and only if n is 0 or 1 modulo 5. a “2” means the second player has a winning strategy. Applegate doesn’t believe in these conjectures. go to
http://www. Misère sprouts conjecture. 4. and we devised and implemented two methods to save space without losing too much time efﬁciency. To reduce the space still further we used a data compression technique.
But the (stationary) right end. with human programmers as the vector. found by Hickerson (at U. Georgia Dear Gathering. like smallpox. new developments in the game are still spreading through computer networks. But if you interrupt it. by singlehandedly popularizing Conway’s game of Life during the 1970s. has to be Time 0: Note that it extends leftward with velocity ½ . the Life bug no longer poses a threat. clandestine Life programs spread like a virus. Davis) has period 5! Stretching between is a beam of darts that seems to move rightward at the speed of light. The toll in human productivity probably exceeded the loss in computer time.C. of which the weirdest. Back when only the large corporations could afford computers because they cost hundreds of dollars an hour. alien Life forms.
. With the great reduction in computation costs. with the left end (found by Hartmut Holzwart at the University of Stuttgart) repeating with period 4. as of late 1992. this is not the case.Strange New Life Forms: Update
The Gathering for Gardner Atlanta International Museum of Art and Design Atlanta. Martin Gardner. Sadly. leading to hundreds of unnatural. infecting some of the world’s best brains and machinery. it would be nice to report that. sabotaged the Free World’s computer industry beyond the wildest dreams of the KGB. the beam shortens in both directions at the speed of light. The biggest shock has been Dean Hickerson’s transformation of the computer from simulation vehicle to automated seeker and explorer. While the Life programs distributed with personal computers are harmless toys whose infective power is 100% cancelled by TV. and the solution of most of the questions initially posed by the game. and then both ends explode.
Hickerson’s Synopsis There have been developments in various areas by different people: Oscillator and spaceship by myself.” here is a synopsis of recent developments computer-mailed by Hickerson when the ineffable Creator of the Universe and its Laws himself asked the same question. and myself Oscillator and Spaceship Searches. David Bell. For a few months I ran it almost constantly. Bill Gosper. and I’ve run it occasionally since then. Paul Callahan. if you’re wondering “Whatever happened to Life?. Bill Gosper. and Hartmut searches Holzwart Construction of oscillators mostly by David Buckingham and Bob Wainwright Glider syntheses mostly by Buckingham and Mark Niemiec New glider guns by Buckingham. In 1989 I wrote a program to search for oscillators and spaceships.STRANGE NEW LIFE FORMS: UPDATE
So. and myself Large constructions mostly by Buckingham. and a period 4
. It found many oscillators of periods 3 and 4 (including some smaller than any previously known.
6. speed c/4 orthogonal spaceship. based on construction universality. but 5. it consists of a still life (“eater3”) and two spark producing oscillators. which implies that all sufﬁciently large periods are possible. David Bell wrote a similar program that runs on faster machines. of periods 5. Also. JHC!).) 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. and c/4. c/3. so we also get glider guns of those periods. and one period 5. one period 4. which reﬂects a symmetric pair of parallel gliders. one period 4. and 47 are the only ones we know with the right sparks. More recently. inﬁnitely many period 2. a few oscillators with periods 5 (including some with useful sparks) and 6. For example. The · ¾ Ò ¾ · ¾ Ò ¼ · ¼Ò ¾¿¼ · ¼Ò ¾ ¾ · ¿ Ò and · ¿ Ò use a device discovered by Wainwright. we now have nontrivial examples (i. Another amusing oscillator uses the glider crystallization that Bill mentioned in his centinal letter a few years ago. 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/2 orthogonal spaceships. speed c/3 orthogonal spaceships. 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. (These are used in some of the large patterns with unusual growth rates mentioned later. (Mine runs only on an Apple II. Due mostly to the work of Buckingham and Wainwright. in which a B heptomino is forced to turn a 90 Æ corner. Hartmut found another with speed 2c/5. speed c/4 diagonal spaceship (not the glider).) Periods ½¿ · Ò and ½ ¼ · Ò use a mechanism developed by Buckingham. A period 150 gun ﬁres toward a distant pair of pentadecathlons..) Construction of Oscillators. The ﬁrst glider to hit them is reﬂected
resembling your initials.e. The turn can take either 64. We
also have an argument. he and Hartmut Holzwart have been producing oodles of orthogonal spaceships with speeds c/2. I built a p155 oscillator in which 20 Bs travel around a track of length ¿½¼¼´ ¾¼ ¢ ·¾ ¢ µ generations. Some of the c/3s and c/4s have sparks at the back which can do various things to c/2 spaceships that catch up with them. 6. (No doubt you ﬁgured that out for yourself long ago.206
B.) Lately. the 73 gen turn can emit a glider. inﬁnitely many period 3. For the multiples of 8. 65. or 47. we put several copies of the B in the track to reduce the period to one of those mentioned. (Any greater nonmultiple of 4 would also work.
with a total of 36 AK47s. Using normal kickbacks. In addition to the long-familiar period 30 and period 46 guns. Glider Syntheses. forms a trafﬁc light. they don’t work with period 30 streams. When they’re all gone. there’s a period ½ ´ ¢ ¿ · ¢ µ gun that’s 119 by 119 and a period ½¿ ´ ¼ ¢ ¿ · ½ ¢ µ ¾ gun that’s 309 by 277. based on the period 100 centinal.STRANGE NEW LIFE FORMS: UPDATE
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. A honey farm starts to form but is modiﬁed by an eater and a block. it uses various mechanisms to interleave larger period streams. Fortunately. 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 largest by far that’s actually been built is Bill’s period 1100 gun (extensible to ¼¼ · ¾¼¼Ò). New Glider Guns. The other guns are all pretty big. and then starts forming another honey farm in a different location. I’ve found some eaterassisted kickbacks that give any even period. there’s now a fairly small period 44 gun that Buckingham built recently. centinal-based ¼¼ · ¾¼¼Ò. you (JHC) described how to thin a glider stream by kicking a glider back and forth between two streams. It’s fairly easy to get any period down
. of size 149 by 149.) It’s also possible to build a gun that produces a glider stream of any period ½ . Buckingham has also found extremely weird and elegant guns of periods 144 and 216. Here’s how to turn 2 MWSSs into a glider: In Winning Ways. emit gliders that can crash to form MWSSs. In the period 94 gun. so we can build a long row of them that is unstable at both ends. two such rows. Subsequent gliders grow a crystal upstream. The period 94 gun is 143 by 607. there are other ways to get any multiple of 30. it’s based on the “AK47” reaction discovered independently by Dave Buckingham and Rich Schroeppel. Rich has already mentioned this. based on a period 72 device discovered by Bob Wainwright. two gliders delete one pair of beehives. its bounding box is 13. The gun itself has a larger period. Buckingham mercifully outmoded it with a comparatively tiny. (Sadly.112. If you delete the trafﬁc light. It emits a glider. the resulting period must be a multiple of 8. which hit eaters. For example. an eater stops its growth. When the crystal reaches the guns. and it begins to decay. forming gliders that stabilize the ends of a row. 11 gliders add a pair of beehives to the crystal. the cycle repeats every ¢ ¾ generations. the process begins again.584 by 12.
300. . Here it’s shown deleting a glider. This will make it possible to produce glider streams of any period. its bounding box is 373 by 372. Ò · ) generations.” (Less than 14 is physically impossible. I have a working construction to build a P14 glider stream by inserting a glider between two gliders 28 gens apart! . For example. 150. 210. it escapes:
There are also combinations of two period 30 guns that give fairly small guns with periods 90. 120. We can also synthesize light-. this reaction leads to a period 30 LWSS gun: Large Constructions. (I built a pseudo-period 15 gun based on his reactions. based on the
. middle-. . GOSPER
to 18 this way (period 23 is especially simple).)
Returning to Hickerson’s account: Many of the large constructions mentioned later require a period 120 gun. The smallest known uses a period 8 oscillator (“blocker”) found by Wainwright to delete half the gliders from a period 60 stream. and 17. . . 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. so we have guns that produce these as well. and heavyweight spaceships from gliders. and Buckingham has found clever ways to get 15. more generally. 16. if the glider is delayed by 60 (or. and 360. 10.208
) But some of these properties can also be achieved without universality. Even inﬁnite growth is rare for small patterns. similar to the one you described. the reaction also sends back a glider (or two) which triggers the release of the next salvo. respectively. ﬁring gliders southwest and northeast. In addition. (It has one small difference: the “test for zero” is not a separate operation. In Winning Ways you (JHC) describe how to pull a block three units using two gliders and push it three units using 30. puffers. This is even true for guns. depending on which cells you’re looking at.” versions of which were built independently by Bill and myself. I’ve also built some aperiodic patterns with bounded populations. there are several
. Probably the second such pattern is the “exponential aperiodic. headed east and west.STRANGE NEW LIFE FORMS: UPDATE
small things we normally look at. Both of these patterns have populations tending to inﬁnity. although almost all large patterns grow to inﬁnity. so some of us have spent many hours putting together guns and puffers to produce various results. But it’s fairly easy to build a pattern for which this isn’t true: Take two glider puffers. (Figuring out how to achieve a particular behavior makes a pleasant puzzle. 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. and probably others. actually building the thing is mostly tedious. These use a glider salvo to push a block (or blinker).) The ﬁrst such pattern (other than glider guns) was Bill’s breeder. using a kickback reaction each time it hits one of the glider waves. Bill also built an “arithmetic aperiodic” in which the gap between occupations increases in an arithmetic progression. a signal is produced whenever a decrement operation reduces the value to 0. Add a glider that travels northwest and southeast.) Most of my large constructions are designed to achieve unusual population growth rates. Instead. whose population grows like Ø ¾. and breeders. it eventually becomes periodic. If you look at a ﬁnite region in a typical small pattern. Using this I built a sliding block memory register. Smaller breeders are now known. such as Ø ½ ¾ Ø¿ ¾ ÐÓ ´Øµ ÐÓ ´Øµ¾ Ø ÐÓ ´Øµ and Ø ÐÓ ´Øµ¾ . and linear growth with an irrational growth rate. I found a way to pull it one unit with two gliders and push it one unit with three gliders.
) Paul Callahan and I independently proved that arbitrarily large puffer periods are possible. (Basically this mimics the way that a ternary reaction can double the period of a glider gun. Give a glider puffer of period Æ . so I’ll just describe mine. 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. 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. providing us with natural syntheses (and probably estimates) of rare objects. this happens every ¾Æ generations. Following Dean’s update. entirely by crashing gliders together. two of them destroy it and the third escapes. which adds two period 60 input streams and produces a period 60 output stream. so they can’t produce another one. Thankfully. there’s already a MWSS in the way.)
—Dean Hickerson To clarify. Warm ones. building such a thing from standard glider logic is straightforward. the MWSS takes the place of the stable intermediary of the reaction. (Of course. The difﬁculty is comparable to stacking water balls in 1G.) Dave Buckingham and Mark Niemiec built a binary serial adder. he recorded the conditions that led to these (and many other) discoveries. but I haven’t built that. Buckingham’s “awe-inspiring glider syntheses” are the constructions of prescribed. The next time the gliders try to crash. often large and delicate.210
B. (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. whose populations are unbounded but do not tend to inﬁnity. Instead.
. His construction is more efﬁcient than mine. but harder to describe. sometimes even oscillating objects. GOSPER
“sawtooth” patterns. but they used some very clever ideas to do it more efﬁciently. months-long series of literally millions of random soup experiments.
STRANGE NEW LIFE FORMS: UPDATE
ºººººººººººººººººººººÓººººººººººººººº ºººººººººººººººººººººÓºÓººººººººººººº ºººººººººººººººººººººÓÓºººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººÓºººººººººººººººººººººººººººººº ººººÓºÓººººººººººººººººººººººÓºÓººººº ºººººÓÓººººººººººººººººººººººÓÓºººººº ººººººººººººººººººººººººººººººÓºººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººÓººººººººººººººººººÓººººººº ºººººººººººÓÓººººººººººººººººÓºÓººººº ººººººººººÓÓºººººººººººººººººÓÓºººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººÓÓÓºººººººººººº ººººººººººººººººººººººÓºººººººººººººº ºººººººººººÓºººººººººººÓººººººººººººº ºººººººººººÓÓºººººººººººººÓÓÓººººÓÓºº ººººººººººÓºÓºººººººººººººÓººººººÓºÓº ºººººººººººººººººººººººººººÓºººººÓººº ººººººººººººººººººººººººººººººººººººº ººººººººººººººººººººººººººººººººººººº ÓÓººººººººººººººººººººººººººººººººººº ºÓÓºººººººººººººººººººººººººººººººººº ÓºººººººÓÓººººººººººººººººººººººººÓÓº ºººººººººÓÓºººººººººººººººººººººººÓºÓ ººººººººÓºººººººººººººººººººººººººÓºº ººººººººººººººººººººººººººººººººººººº ºººººººººººººººººººººººººººÓÓºººººººº ºººººººººººººººººººººººººººÓºÓººººººº ºººººººººººººººººººººººººººÓººººººººº ººººººººººººººººººººººººººººººººººººº ºººººººººÓÓºººººººººººººººººººººººººº ººººººººÓºÓºººººººººººººººººººººººººº ººººººººººÓºººººººººººººººººººººººººº
Conway called such oscillators billiard tables, and along with the rest of us, never imagined they could be made by colliding gliders.
ººÓÓºÓÓººººº ººÓºººÓººººº ºººÓÓÓºººººº ºººººººººººº ºººººÓÓÓºººº ººººÓºººÓºÓÓ ºÓººÓºººÓºÓÓ ÓºÓºÓºÓºÓººº ºÓÓºÓºººÓººº ºººººÓÓÓºººº ºººººººººººº ºººººººÓÓÓºº ººººººÓºººÓº ººººººÓÓºÓÓº
Finally, after a trial run of the foregoing around the Life net, Professor Harold McIntosh (of the Instituto de Ciencias at Puebla) responded: “The humor in that proposed introduction conjures up images of untold taxpayer dollars (or at least hours of computer time) disappearing into a bottomless black hole. That raises the question, ‘Have other hours of computer time been spent more proﬁtably?’ (Nowadays, you can waste computer time all night long and nobody says anything.) “Martin Gardner is a skilled presenter of ideas, and Life was an excellent idea for him to have the opportunity to present. Not to mention that the time was ideal; if all those computer hours hadn’t been around to waste, on just that level of computer, perhaps the ideas wouldn’t have prospered so well. “In spite of Professor Conway’s conjecture that almost any sufﬁciently complicated automaton is universal (maybe we can get back to that after vacations) if only its devotees pay it sufﬁcient attention, nobody has yet
come up with another automaton with anything like the logical intricacy of Life. So there is really something there which is worth studying. “It might be worth mentioning the intellectual quality of Gardner’s presentation of Life; of all the games, puzzles and tricks that made their appearance in his columns over the years, did anything excite nearly as much curiosity? (Well, there were ﬂexagons.) “Nor should it be overlooked that there is a more serious mathematical theory of automata, which certainly owes something to the work which has been performed on Life. Nor that there are things about Life, and other automata, that can be foreseen by the use of the theories that were stimulated by all the playing around that was done (some of it, at least).” The Life you save may not ﬁt on your disk. —Bill Gosper
ØØÔ »»ÛÛÛºÑ Ò ×ÔÖ Ò º ÓÑ» Ð Ò »Ð
Many remarkable discoveries followed this writing. See, for example,
» Ò Üº ØÑÐ
STRANGE NEW LIFE FORMS: UPDATE
An early “stamp collection” by Dean Hickerson. These are representative oscillators of the (small) periods (indicated by the still-Life numerals) known by the end of 1992. As of the end of 1998, all periods have been found except 19, 23, 27, 31, 37, 38, 41, 43, 49, and 53.
M. Oskar van Deventer
In October 1983 I discovered hollow mazes. A. K. Dewdney presented the concept in Scientiﬁc American, September 1988. In this article I shall give a more detailed description of the subject. The ﬁrst part is about multiple silhouettes in general, the second part is about their application to hollow mazes.
Silhouettes are often used in, e.g., mechanical engineering (working drawings) and robotics (pattern recognition). One particular mathematical problem reads: “Which solid object has a circular, a triangular, and a square silhouette?” The solution is the well-known Wedge of Wallis (Figure 1).
Figure 1. Wedge of Wallis.
A silhouette deﬁnes a cylindrical region in space, perpendicular to the silhouette surface. With this meaning the word no longer refers to an object, but only to a region of a ﬂat surface (Figure 2). Two or more silhouettes (not necessarily perpendicular to each other) deﬁne unambiguously a region or object in space that is the cross-section of two or more cylinders. Dewdney introduced the term projective cast for an object deﬁned by a number of silhouettes. Not every object can be deﬁned as the projective
M. O. V AN DEVENTER
Figure 2. (a) Negative of a silhouette; (b) The cylindrical region in space deﬁned by it.
cast of a (limited) number of silhouettes: One cannot cast a hollow cube or a solid sphere. There are two interesting topological properties that apply to silhouettes, to projective casts, and to mazes and graphs in general. First, a graph may have either cycles or no cycles. Second, a graph may consist of either one part or multiple parts (Figure 3).
Figure 3. Topological properties of mazes and graphs: (a) multiple parts with cycles; (b) one part with cycles; (c) multiple parts, no cycles; (d) one part, no cycles.
Dewdney used the term viable for a silhouette or a maze that consists of one part and has no cycles. Even when silhouettes are viable, their projective cast can still have cycles (Figure 4) or multiple parts (Figure 5). There is no simple connection between the properties of the silhouettes and the properties of their projective cast. The basic problem is the analysis
The resemblance between multiple silhouettes and hollow mazes is evident. The two control mazes on opposite sides of the box are identical. since a spar cannot jump from one part to another. There can be no cycles since the center of a cycle would fall out. Three viable silhouettes yielding a projective cast with two parts. In this way. Three viable silhouettes yielding a projective cast with one cycle. Each spar passes from one side of the box to the other.HOLLOW MAZES
Figure 4. sliding along the slots of the control maze on each side. A cursor consisting of three mutually perpendicular spars registers one’s position in the hollow maze.
A hollow maze is a rectangular box with six sides (Figure 6). and how can a projective cast be given certain properties by its silhouettes? I do not know whether there is a general answer. a single three-dimensional maze is produced from three pairs of two-dimensional mazes.
and synthesis of silhouettes and projective casts: How can the properties of a projective cast be found without constructing it from its silhouettes. Each side is a two-dimensional control maze: a surface into which slots have been cut. Control mazes cannot consist of multiple parts.
Figure 5. Control mazes are always viable because of their construction.
but I lack the mathematical tools. Count the links in the hollow mazes of Figure 7. For example..) Cut and Connect.
Now we can take another look at the topological properties of the projective cast. This method guarantees that the eventual hollow maze is viable. viable or not:
Trial and Error. their paths have zero width). assuming that it cannot have cycles: We count the number of grid points of the projective cast. a square or cubic grid is used for control mazes and their projective casts. There are several ways to construct a hollow maze. the total number of links is found after taking all cross-sections. the control mazes are line mazes (i. V AN DEVENTER
Figure 6. Helmut Honig. The dotted cross-section of Figure 7(a) has six links through it. (Working at random. because the control mazes are line mazes.) It can easily be tested whether or not a projective cast is viable. A simple hollow maze. The number of links can be counted from the control mazes without constructing the projective cast. It seems that the projective cast cannot have cycles anymore. the control mazes are viable.216
M. a ¿ ¢ ¿ ¢ ¿ projective cast has 27 grid points. (I’m sure that there must be a simple proof of this.
I have developed several restrictions for the investigation of hollow mazes:
¯ ¯ ¯ ¯
There are three (pairs of) perpendicular control mazes.e. a German computer science
. Cut the projective cast several times and link the pieces again (Figure 8). O. Then there must be exactly 26 “links” between these points for a viable cast.
Regular Control Mazes.HOLLOW MAZES
Figure 7. Test of the viability of a projective cast. (a) ¾· ¾ ½. (b) ½ · · · · ¾ ·
· · · ·¾
½. See Figure 10.
Figure 8. showed me that not all viable hollow mazes can be constructed by this method. The cut and connect method for constructing hollow mazes. viable! not viable.
student. Figure 9 is a counterexample. How should identical control mazes of this kind be placed to yield a viable hollow maze?
M. Replace each point by a hollow maze (see Figure 11).
Recursion. Regular control mazes.
There is still much research to be done on multiple silhouettes and hollow mazes. Construction of a hollow maze by recursion. V AN DEVENTER
Figure 11. A viable hollow maze that cannot be constructed by the “cut and connect” method. O.
Figure 10. and perhaps some of the problems mentioned above will be solved in the future. I keep receiving letters from readers of Scientiﬁc American.
The third is the problem of Selling Different Amounts at the Same Price Yielding the Same. which contained statements such as “Mary is twice as old as Ann was when Mary was as old as Ann is now. Sometimes the difﬁculty is in setting up the equations. Ü · Ü or Ü or Ü . In this case. compiled around A. See Tropfke  for this and related problems. I discuss three examples of such diophantine recreations. the equations and their solution are relatively straightforward. but there is an additional diophantine requirement that the data and/or the answers should be integral. where the problem was in understanding the phrasing of the question. for which I give an algorithm for ﬁnding all solutions and a new simple formula forthe number of solutions. leading to one equation in one unknown. In the late nineteenth century. done to illustrate the ideas. 500 (although the date of this is uncertain. but the solution requires some ingenuity. The simplest are just problems of the ‘aha’ or ‘heap’ type.e. such as exploiting the symmetry of the problem. In a third class of problems.” (See Loyd )
A Simple Age Problem
Age problems have been popular since at least The Greek Anthology . for which I have just found a reasonably simple condition for integer data to produce an integer solution. with the solution then being easy.D. it is not always straightforward to determine when the problem has solutions or to ﬁnd them all or to determine the number of solutions. The second is the Ass and Mule Problem. complicated versions like “How Old Is Ann?” appeared. The ﬁrst is a simple age problem.. and the material goes back several centuries). Other times the equations are straightforward. In The Greek Anthology.Some Diophantine Recreations
There are a number of recreational problems that lead to equations to be solved. the problem of Diophantus’ age gives us
Ü Ü Ü · · · · ½¾ ¾
. A person. albeit less poetic. but with the same numbers. As eight is to sixteen. what were their ages on their wedding day? We can state the general form of this problem as follows. reportedly from The American Tutor’s Assistant. We man and wife have been.
is times as
Thus the problem in The American Tutor’s Assistant and Bonnycastle is Problem (3. from what I’ve said. after years. Now we give a diophantine aspect to our problem by asking the question: If . 2) does not have an integral solution. What were our ages when we wed? A more typical. 1824 .220
D. . but after they had lived together 15 years. version. Problem ´ old as . appears at about the same time in Bonnycastle. 2). when are and integers? For example. more comprehensible. My age was to that of my bride As three times three to three But now when ten and half ten years. Now tell. The solution of these is easily
assuming . et al. he was only twice as old. the problem is quoted in Bunt. 15. at the time he was married. was 3 times as old as his wife. It is easily seen that the problem leads to the equations
where and found to be denote the ages of
´ · µ
. and are integers.
times as old as
. 3. Since is integral if is.½ When ﬁrst the marriage knot was ty’d Between my wife and me. which is the earliest I have found. . The type that I want to examine is illustrated by the following.
µ. Her age to mine exactly bears. SINGMASTER
Between these are a number of types of problem. 1791. we can give a pretty simple answer: is integral if and only if
½I haven’t actually seen the book. I pray.
did you ever consider when I would be half as old as Helen?” Now it was my turn to be worried. The next day. soon you’ll be the same age!” Jessica seemed a bit worried. and now you are only twice her age. The mule says to the ass: “If you give me one of your sacks. I saw her doing a lot of scribbling. four years ago. We pick arbitrary integers and . Nonetheless. and I teased Jessica by saying “Seven years ago.” The ass responds: “If you give me one of your sacks. with ½. 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.” said Jessica. then let be any multiple of ´ µ . you were 9 times as old as Helen. I will have as many as you. “Dad. If you are not careful. you were 5 times her age.” How many sacks did they each have? The general version of the problem for two individuals can be denoted ´ µ for the situation where the ﬁrst says: “If I had from you.” “Don’t be so positive. as she stomped off to school. one can still get unexpected results.SOME DIOPHANTINE RECREATIONS
If we let denoted as
be the greatest common divisor (GCD) of ´ ½µ. she said to me.
This gives us the kind of condition that allows us to construct all solutions. you were 3 times her age.
Ass and Mule Problems
The classical ass and mule problem has the two animals carrying sacks. and I began muttering — “That can’t be. although one can consider permitting some or all of the values to be rational numbers. I’d have
. I will have twice as many as you. you’re always older than Helen. then (3) holds if and only if divides
½.Our neighbour Helen is just 8. and went off muttering. as in the following problem: My daughter Jessica is 16 and very conscious of her age. that’s just the limit! By the way. This pretty much settles the problem. six years ago.
and then determines which values of and give integral solutions. though. Here I present a reasonably simple criterion that allows one to generate all the integer problems with integer solutions. which is similar to. which I would not have predicted.222
D. Consider ´ · ½ · ½µ.
. Luckily. but somewhat more complex than. I ﬁrst noted that Ü · Ý has a symmetric expression.” and the second responds: “And if I had from you. This also divides ´ · ½µ´ · ½µ ´ · ½µ ´ · ½µ ½. This still doesn’t give us a very simple or satisfactory criterion. so we only need to check one of the second terms in (2). assuming ½. ´ · ½µ´ · µ ´ · µ´ · ½ · ½µ. Parameswaran for the letter that inspired this investigation. I said I knew of no way to decide this that was simpler than actually seeing if the solutions were integers. but it is traditional for the parameters and the solutions. that Ü is an integer if and only if Ý is an integer. One can see from (2). where ´ · ½ · ½µ denotes the GCD of · ½ and · ½. The values of Ü and Ý are integers if and only if ½ divides both ´ · ½µ´ · µ and ´ · ½µ´ · µ. unfortunately. say Ü Ý." Many versions of this problem occur. and I thought that the integrality of this was equivalent to integrality of Ü and Ý. I am indebted to S. The criterion allows us to pick arbitrary and . this false trail led me to the following simple result. I am also intrigued to see that any and any can be used. The general problem ´ µ leads to the system of equations
The solutions are readily computed to be
times you. the age problem. I’d have times you. assuming ½. I ﬁnd it particularly striking that and only enter via the sum · . to be integers. Now the last statement of the previous paragraph can be divided by to give us that Ü and Ý are integers if and only if
´ · ½µ´ · µ ½
´ · ½µ´ · µ ½
´ · ½ · ½µ
This seems to be as simple a criterion for integrality as one could expect. which is if and only if ½ divides GCD ´ · ½µ´ · µ. Ü and Ý are integers if and only if the second terms in (2) are integers. and it is obvious from (1). however. this doesn’t hold. Hence a natural question is to consider the diophantine question Which integer problems have integer solutions? In the early 1990s . The general solution of the problem gives somewhat complicated expressions for Ü and Ý. Looking for a more symmetric solution. as before.
but by analogy with the argument above. it is direct to show that Ü and Ý are integral if and only if
¬Æ «¾ ´¬ · « · µ
«´ · µ
Dr. Æ «. It is then quite feasible to consider rational and . or even the same. We easily see that ´ · ½ · ½µ. this is not really essential. loses all number-theoretic interest. but we see from (6) that Ü is an integer if Ý is an integer. though the converse may fail. so Parameswaran’s version has integral solutions if and only if
µ · ´ · ½µ´
´ · ½ · ½µ
From this.SOME DIOPHANTINE RECREATIONS
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 see that Ý is integral if and only if divides ´ · ½µ´ µ. Setting ¬ «. one can easily generate all examples. Thus he considered the problem where equations (1) are replaced by
´ µ ´ µ
· ´ · ½µ´
This is easily solved to obtain
The expression for Ü is a bit complicated. possibly even the same answers. Are there problems where the answers are integral. but doing so doesn’t seem to make the result signiﬁcantly clearer. One can deal with rational Ü Ý by scaling the problem so that one has an integer problem. Now consider ´ · ½µ. so you don’t know whether the statements are made by the same or different animals. and I leave it for the reader to discover them. so it seems most interesting to assume that we have integral and and we want integral Ü and Ý. From the ﬁrst equation. Thinking about Parameswaran’s version led me to wonder if there was any case where both versions gave integral answers. in either case? Perhaps surprisingly. That is. we get a fairly direct way to check whether Ý is integral. there are problems where the answers are indeed the same. Parameswaran interpreted the problem as though the second statement was also made by the ﬁrst animal. the admission of irrational values.
. where GCD ´ ¬ Æ µ ½. suppose you can hear the animals but can’t tell which one is speaking. however. Although the problem traditionally has all integral parameters.
“If I had half of what you have. the Ass and Mule problem may not be in any Chinese source. 2. 5) and (2. depending on whether ‘you’ refers to all the others or just the next in cyclic sequence. and the problem is of type II. 2. The earliest three-person version that I know of appears in India. 1. including some with ﬁve persons.. 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.” He gives (30. and Western sources. That is. 850. Tropfke  discusses the problem and cites some Chinese. He proposes “to ﬁnd two numbers such that each after receiving from the other may bear to the remainder a given ratio.D. 2. By about the ninth century. not only in Europe but also in India and the Arab world. Arabic. and extended versions where a person says. so that. e. I’d have 5 times you plus 1 more. Heiberg’s edition  gives the problem in Greek and Latin verses. 10. The earliest four-person version I have seen is Arabic. Below I give a summary of versions that I have noted up through Fibonacci. the problem is of type I. SINGMASTER
History of the Ass and Mule Problem. (10. the problem was a standard. 3. e. in al-Karkhi (aka al-Karaghi) from about 1010 . and it has remained a standard problem ever since. Two versions. Indian. Byzantine. Diophantus studied the problem in general. The other Chinese reference is a work of about A. The earliest known version of this problem has an ass and a mule with parameters (1.. 4).g.” This is a different type of problem and belongs in Tropfke’s previous section. surprisingly. 2. who typically gives many versions. about A. 485 that has only been translated into Russian and may well be similar to the earlier Chinese example. The problem is readily generalised to more than two participants.” This includes all the relevant sources cited by Tropfke except Abu al-Wafa (only available in Arabic) and “Abraham. but then there are two forms. The examples cited in the Chiu Chang Suan Shu  state.C. in the work of Mahavira . the second equation of (1) would be replaced by This is our problem ´ µ. 50. “If you give me 7. appear in The Greek Anthology .D. 3) as an example.” who is elsewhere listed as probably being
. an inconsistent example. I’d have 50.g. Alcuin gives an unusual variant of the problem in that he assumes the second person starts from the situation after the transfer mentioned by the ﬁrst person takes place. 1) and is attributed to Euclid. He also covers similar problems with three and four persons. from about 300 B.224
300 B. 2. 2) I-(9. 1. 1. but this form is less clearly expressed and has inﬁnitely many solutions. 2. 4. 7) I-(7. and to the second thirty apples. 11. 5. 10. 22. 11. 3. 3. 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. 510 9Ø century c. 7) II-(1. 11. 2. 1. 850 c. 1) (30. 6) I-(7. 6) (1. 23.5) I-(25. noting that this system is inconsistent Fibonacci. 510 c. 2. with the problem statement giving the 6 as a 7 Fibonacci. 3. 9. 2. 5. and to the youngest ten apples. 8. 5) (100.C. 5. 6) I-(7. 250 c. 9. 10. 2. 9. 9. so I will ﬁrst describe the later form which is ﬁrst found in Fibonacci . 2. 3. 4. certain apples. 5) (2. 5. there seems to be no common name for it. The problem is treated in one form by Mahavira . 50 in the ﬁrst printed English riddle collection “The Demaundes Joyous. c. done two ways
Selling Different Amounts at the Same Price Yielding the Same
Although this problem is quite old. 3. 10. 3. 3) (10. and all these three sold in like many for a penny.” printed by Wynken de Worde in 1511 . 4) (2. Now how many sold each of them for a penny?
. And he took to the eldest daughter ﬁfty apples.SOME DIOPHANTINE RECREATIONS
from the early fourteenth century and hence after Fibonacci. 850 c. 8. 5. 11. hence the above title. 4. 9. 11.
(1. Sridhara . although the material is believed to be based on older Arabic sources. 7) I-(7. 4. which is rather a mouthful. 3. 4. to sell. 2. 10. 5. 5) c. and Bhaskara . 11. A man had three daughters of three ages. 1. which daughters he delivered. 0. 2. 4. 10) (7. 5. 6. 6) I-(7. 3. 50. with the problem statement giving the 6 as a 7 Fibonacci Fibonacci. and brought home in like such money. I quote a version that appears as No.
or lowering prices late in the day in an attempt to sell their remainders. but we both conjecture that he was counting the positive solutions. 25. This may happen by the sellers going to different markets. for which he often gives fractional answers. 50). distinctly against the spirit of the problem.226
‘In like many’ means ‘the same number’ and ‘in like much money’ means ‘the same sum of money. or raising them late in the day when few items are left and new buyers arrive.’ Anyone meeting this problem for the ﬁrst time soon realises there must be some trick to it. it can take a fair amount of time to work out a solution. if the prices are also given. and it is not immediately obvious how to ﬁnd or count all the solutions. which is essentially the same as the ﬁrst edition of 1694. In 1874. of which 36 are positive. Tartaglia  justiﬁes the different prices by having large and small pearls. (12. which would produce the positive solutions with the smaller price equal to 1. Fibonacci only gives two-person versions: (10. or sometimes fruit.. and he gives ﬁve solutions for his ﬁrst version — but there are 55. 30). 30. (12.
. ﬁnd solutions where the amounts received are in some ratio. 33). Neither Glaisher nor I have seen the relevant work of De Lagny. I have now arrived at a reasonably simple approach that turns
¾I have not seen these editions. Let us denote the problem with numbers ½ ¾ by ´ ½ ¾ µ. he added several more versions as well. The extensively revised and enlarged edition of 1725 (or 1723) outlines a fairly general method and correctly says there are ten solutions. He gives a fairly general rule for generating solutions . considers (10. which normally are all of the same value. Although the objects are usually eggs. or simply changing prices after lunch. During the 1980s. 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. of which there are indeed six. Labosne revised the material. Fibonacci goes on to consider solutions with certain properties — e. I tried several times to ﬁnd an easy way to generate the solutions. is there a solution. 40). 30. 32). so that the above is (10. Fibonacci is one of the few to give more than a single solution. According to Glaisher .g. 30) and gives two solutions. This is historically by far the most common version. The question of ﬁnding all solutions for a three-person problem seems to have ﬁrst been tackled in the 1600s. contrary to De Lagny’s statement that there are six solutions. Even with this trick exposed. ﬁnd all solutions with one price ﬁxed. dropping Bachet’s rule and replacing it with some rather vague algebra. the 1612 and 1624 editions of Bachet  ¾ give a fairly general rule for this case and apply it to (20. such as the ﬁrst receives twice the second. The 1708 English edition of Ozanam .
We have ¾Ò · ¿ unknowns connected by these ¾Ò equations. Then we have the following:
where is the constant amount received. whenever Glaisher gives the number of solutions. Ayyangar  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. Because of the symmetry between and . Indeed. Glaisher’s paper contains most of the ideas involved. 50).SOME DIOPHANTINE RECREATIONS
out to be essentially a generalisation and extension of Ozanam’s method. the sell. which he starts considering on the ﬁrst page. that ´ µ ´ µ ½. we get ´ ½ ¾ µ is a rational number. but I prefer to deal with the original values. i. subject to some conditions. We thus expect a three-dimensional solution set.. Ayyangar’s method is quite similar to my method. By (4) below. i. Let be the number of items that the Ø person sells at the higher price . Normally the objects being sold are indivisible.e. Ø person has items to Consider the problem ´ ½ ¾ Ò µ. and we assume Ò ½ to avoid triviality. but he is so verbose and gives so many pages of tabular examples. Hence ¾ · ¾ . he displays or indicates all of them. so we want the numbers to be integers. we can assume and have assumed . Most previous methods yielded solutions involving some choices. He also assumes that the are an arithmetic progression. 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. 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. special cases..e. but did not thoroughly check that all solutions are obtained. This is equivalent to scaling the prices in equations (2) and reduces the dimensionality of the
. (One can reverse and in order to make the s increase. and let be the number of items that the Ø person sells at the lower price . This makes ½ ¾ ¡¡¡ Ò and ½ ¾ ¡ ¡ ¡ Ò. No one else has noticed that the number of solutions in the Western version can be readily computed. 30. We can reorder them so that ½ ¾ ¡ ¡ ¡ Ò . so I later sketch his method in my notation and ﬁll in a gap. this doesn’t change the problem. and we can assume that and are integers with no common factor. but the amounts sold at the higher price are the smaller numbers and hence are easier to deal with.) From ½ · ½ ´ ¾ ½ µ.
the yield is speciﬁed. which is the condition (5)
for all the equations (4) to have integral solutions. divides . Glaisher considers ¿.228
D. In some versions of the problem. This tells us that . we have «¾ ¾ . are often given with fractional values. i. «¾ .. Hence ¾ . ´«¾ · ¾ µ «¾ ½ · ¾ «¾. which restricts the solutions to a subset of those for the case ¾½ ½. and we note that ½ «¾ ½ . in particular ½ ¬ for some integer ¬ — e. SINGMASTER
solution set to two. From just before (6). the most common solution of the (10. 50) problem has ¿ ½ — and the usual solutions have each person selling as many groups of ¬ as possible. This is also equivalent to scaling the prices. it follows that (4) has an integral solution for ½ if and only if divides ½ and this holds for
each if and only if
½ leaves us Ò ½ equations:
It is now convenient to adopt the following:
Thus (5) can be rewritten as for some . 30..
. this implies that ¾ divides . so the previous equation becomes ´ µ« ¾ ¾ . and the solution set is again twodimensional. these prices are considered the same as ¾½ ½. In the present approach. This is quite reasonable. as solutions that differ only in price scale are really the same. so that
« is a multiple of Now we will take « ¾ as the ﬁrst basic parameter of our solution. and they are considered to give the same sets of solutions. we must have ¾ dividing «¾. unless one imposes some further condition such as being the reciprocal of an integer. and (4) gives us
divides . From the case ¾ of (4). From (6). giving us
´¿µ ´ µ ´
Subtracting each of these from the case
µ´ ½ µ ´ ½ µ Since ´ µ ½ implies ´ µ ½. Let ´«¾ ¾ µ. ½ with the condition that be integral. so «¾ ¾ . and I claim that any such «¾ generates an integer solution of our problem. Now use equation (1) to eliminate from (2). Since ´«¾ ¾ µ ½. we have ´ µ« ¾ ¾ . however.e. We also get ´«¾ · ¾ µ . where ´ µ ´«¾ ¾ µ ½. Solutions in the literature.g.
´ µ ´½¼µ
Now the largest value «Ò can have is
In our example. 8. Then ¾ ½ . we get no solutions. 6. . ½ ¾ ¿. we have ´ µ´ ½ Òµ Ò . ¿ ¾¼. From the case Ò ½ ½ ½ of (4).
indicating no solutions. Recalling that «¾ is a multiple of ¾
½ . For «¾ . 9. so we have ¾ ½ «¾ Ò
. and we choose «¾ as a multiple of ¾ ¿. ½ ¿ . . and there are solutions with « . and «¾ gives a negative value. From the case (4). since «¾. ½ ¾ . we have ´ ½ µ ¿. 8. We get solutions in the above section for each « ¾ that is a multiple of ¾ and in the interval ½ «¾ ½ . we ﬁnd ½ «¾ . 7. Counting the Solutions.)
Hence. . For «¾ ¿. so there are just six positive solutions. We can now determine the possible values of « ¾. namely (10. 25. 9. there are
Ò . ¾.e. and there are solutions with « ¾ . We can now let ½ vary as the second basic parameter of our solution and we can let ½ ½¼. . For «¾ . consider the problem discussed in Ozanam. we can let ½ until Ò becomes zero. i. . Thus Ozanam is correct. hence ¾ ¿ ¾. ½. hence ¾ ¿ ½. 4. For such an «¾. set ¾
«¾ Ò ¾
. 30).e.SOME DIOPHANTINE RECREATIONS
As an example. We can now let ½ ½¼.. for a given
«¾Ò ½ · ½ solutions ¾ In the above example. we have ´¿ ½ µ ¿. ½ ¿ . so that Ò ¼ when
(This expression is an integer. i. but four of these solutions have either ½ ¼ or ¿ ¼. giving us three solutions. «¾ gives ½¼ ¢ ¾¼ ½ · ½ ¿. «¾ ¿ gives ½¼ ¿ ¢ ¾¼ ½ · ½ solutions. giving us seven solutions. 5..
Looking again at our example. in our example. Hence we normally get ¾Ã situations with a zero value.. SINGMASTER
Then (10) becomes
The upper bound may not be an integer. i. unless it happens that ½ «¾Ò ¾ . i. we have ¾Ã ½ solutions with zero values. and then Ã ½ ¾ Æ ¾ ½ Æ· ´ ½ ¾µ¾ ¾ . we have Ã
´ ½ Ò · ½µ ½ Æ
Ã ´ ½ · ½µ
Ò Ã ´Ã · ½µ ¾
¾ and ¾ ¢ ½½ ¾¼ ¢ ¾ ¢ ¿ ¾ ¢
Most of the earlier writers excluded zero values and only considered positive solutions.e. then ¾ we get Ã ´ ½ ½µ ¾ Æ ´ ½ ½µ Æ· ´ ½ ¾µ ½ . i. we have Ã ½ Æ ¿ Æ· ½. when the upper bound for given in (13) is an exact integer. ½ «¾Ò ¾ by (7). when two zero values occur in the same solution. perhaps as intended by De Lagny. Most of the examples in the literature have the in arithmetic progression (AP). This led me to ask when a problem admits only one solution or only one positive
½ Ò ½
Ã ´ ½ · ½µ
´Ò ½µÃ ´Ã · ½µ ¾
. We then have Ã ½ Æ ½´ ½ · ½µ ¾ Æ· ´ ½ ¾µ´ ½ ½µ ¾. A little thought shows that then is the common difference of the progression. . All but one of these have ½ even. These occur when ½ ¼. Tartaglia’s problem 139 is the one with two sizes of pearls and is (10.. so we let
where [ ] is the greatest integer function. so we can write ½ ·´ ½µ . We can count the number Æ · of positive solutions by ﬁnding the number of solutions with a zero value.e. A few cases occur that are not APs (see below). and when Ò ¼. 90). 20.e.
Again. The majority of examples are APs with Ò ¿.
Substituting (11) into (8) and summing gives us the number of solutions. we ﬁnd that this occurs when ½ Ò . ½ ½ . which is trivially an AP. Ò Ò ½:
A number of early cases have Ò ¾. Rearranging. Subtracting ¾Ã or ¾Ã ½ from Æ yields Æ· . A few longer APs occur — I have seen examples with Ò .230
D.. Then we have Ò ´Ò ½µ . In that case. If ½ is odd. From (15). Æ· ½¼ ¾ ¢ ¾ .
this turns out to be if and only if ½ Ò · ½. given as ´Æ Æ· µ. For an AP. we have Ã ¾ and Æ . we are speciﬁcally given the greater price . for their capitals respectively. and resold [a part] so: and disposed of the remainder at one for ﬁve panas. What was [the rate of] their purchase? and what was [that of] their sale? This requires some explanation. Combining this with (16) shows that this implies Ò ¿. 110) has Æ ½.SOME DIOPHANTINE RECREATIONS
solution. the Indian version is somewhat different and gives inﬁnitely many solutions. It is assumed that you can buy items per pana. so the above gives all cases where Æ · ½ and Æ ¿.e. To illustrate. Here the numbers are not the numbers of objects. eight.e. For example. ½ Ò ¾.. along with the ﬁrst known (to me) dates and sources. hence ½ Ò ½ Ò Ò ½ or ´ · ½µ . The Indian Version As mentioned earlier. which gives us . Ayyangar’s translation of the problem
. Æ· ½ implies that the unique positive solution occurs with ½ and has ½ ½ ½ and Ò ½. and the number of solutions given in the source. as in Tartaglia’s example. consider the following problem from Bhaskara. so the the numbers of items will be and our equation (1) becomes
´ · ½µ ½ ¾
Further. We will get Æ ¿ only if Ã ½. In general. . but when Ò ¾. while when Ò ¿. (10. and it is understood that the lesser price corresponds to selling ¬ items per pana. Three traders. used as an example in Ayyangar. i. Example instanced by ancient authors: a stanza and a half. ½ ¬ . i. Tropfke  and Glaisher  cite a number of sources that I have not seen. Below I tabulate all the different versions that I have noted. and thus became equally rich.. Æ ½ implies that the only solution occurs with ½ and has ½ and ¼ . where ¬ is an integer. we ﬁnd Ã ¿ and Æ . but the capitals (in panas) of each trader. bought leaves of betle [or fruit] at an uniform rate. having six. Ò ½ ½ 20. this is if and only if Ò ½ . Glaisher  also studied this problem and was intrigued by this observation. Initially I thought that Æ · ½ would imply that Æ ¿. Similarly. A little manipulation shows that this works if and only if
For an AP. and a hundred. the numbers of solutions. noting the analogous cases with 10 and 11 people.
1 given. 40. 36) (25. 1) (7. 901) (10. Bhaskara makes the following comment: “This. 30) (18. 1 given. 48. 60) . 6 given. 1 given. 36) (30. 603. (20. 21) (225. 20. 100) (45.’ Thus (2) becomes
We now have ¾Ò · ¿ unknowns and ¾Ò equations. I must point out that some Indian versions give and/or as fractions! The Indian authors do not get very general solutions of these problems. 1 given. 33) (10. 68. none given. 56. 32) (12. none given. 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. 82) (17. Sridhara ﬁnds a one-parameter solution set. 55) (78. 40) (18. 1 given. though it is not as easy to see how to do this as in the Western version. 196) (45. 119. 1) (100. so we again have a three-dimensional solution set. 1 given. 90) Date 1202? 1202? 1202? c. 32. 140) . 29. 40. 80) (10. (20. 50) (30. one with fractions. 1 given. 1 given. 60) (117. 454. 20. 1 given with fractions.4) (70. 33. 170) (305. 50) (11.232
Version (10. 11. At this point. 16)
D. 35) (11. 60) (10. 4 given. in some way to reduce the solution set to two dimensions. 30. 55) (55. 1 given. 20) (10. 1 given.552. 1 given. 40) (10. 26) (20. 30. . 37) (27. 16) (16.9) (100. 30) (12. 36) (27. 81) (10. 2 given. all given. 1 given. 33) (20.6) # of Solutions (55. . 81) (55. 752. 15) (31. 40. 20) (64.400) (25.
makes this more speciﬁc by saying they ‘sold a part in lots and disposed of the remainder at one for ﬁve panas. 1300 1300s 1489 1489 1489 c. 30) (8. 1 given. 17. 36) (78. SINGMASTER
Solutions in Source 5 given. 1 given. 50) (10. .
. 25. 30. 12. 55) (16. 136) (3. 1 given. One can scale the set . 49) (3.
In the like suppositions. as has brought out a result in an unrestricted case as in a restricted one. we must have Ò ´ ¼µ ¬ ½
. and ´ µ can be a proper divisor of Ö. disappoints. but in order for «½ Ò ½ . Write Ö×. but since his notation is different.SOME DIOPHANTINE RECREATIONS
which is instanced by ancient writers as an example of a solution resting on unconﬁrmed ground.e. so our equations become
· ¬¬ ·¬
´ ¼µ ´ ¼µ
Ayyangar gives a reasonable solution process for this problem. Then is relatively prime to if and only if divides ×. then subtracting to eliminate and using the same abbreviations as before. Then ´ ¼¼¼ µ is equivalent to ¬ ½ ¼ ´mod µ and this is solvable if and only if ´ µ ½. we get
¬ ¬ ½
´ ¬ ½µ «
and the condition for all the equations to have integral solutions is
If we set
´ ¼¼ µ
´ ¬ ½µ
We need to determine which and ¬ will make ´ ¼¼ µ hold. (Ayyangar notes only that ´ µ must divide . which is if and only if Ö divides ´ µ. and such a supposition introduced. which is if and only if Ö divides . Eliminating ¬ . ´ ¼¼¼µ ´ ¬ ½µ ¼ ´mod µ Let ´ µ. I will sketch the solution in the present notation. i.” Amen! For ease of expression.) Both and ¬ can take on an inﬁnite range of values. where Ö comprises all the prime powers in whose primes also occur in .. owing to restriction. pointing out an improvement. the answer must by the intelligent be elicited by the exercise of ingenuity. This gives us a rather peculiar condition. we let ¬ ¬ . has been by some means reduced to equation. when the operation.
“A classical Indian puzzle-problem. In: T. 1884. 3292. 12.”  Ayyangar. Problemes plaisants & delectables qui se font par les nombres. 3rd ed. Soc.. John Murray. An Introduction to Algebra.. See No.. Wynken de Worde. there are only a ﬁnite number of solutions. and Other Places of Public Education. p.. H. Bijanganita.(475). Analysis similar to that done before shows that there are then ½ · ´ ¬ ½ Ò µ ¬ solutions. annotated by David Singmaster and John Hadley. Facsimile with transcription and commentary by John Wardroper. 19.. Johnson. pp.900 and of 3294.
 Alcuin. with Notes and Observations. The ﬁrst nine editions appeared “without any material alterations.” c. 15. 1782. In: Henry Thomas Colebrooke. Math. 150 B. 1910. Gordon Fraser Gallery. 4th ed. Braunschweig. Lyon. trans.” which “may be considered as a concise abridgment” of his two-volume Treatise on Algebra. reprinted by Blanchard. aka Bhâskara.so and ¬ ½½¼. designed for the Use of Schools. 178–186. 54. Nunn et al. pp. 1879.. 1817. By the way. SINGMASTER
For given and ¬ satisfying ´ ¼¼¼ µ and ´ ¼µ. pp. Problems to Sharpen the Young. A. Lyon. Paris.. 106–115.. 1511. Wiesbaden. Gaz. Prob. 242–244. The Historical Roots of Elementary Mathematics. so . L. 170. Vieweg. 1624. 1150. 1805. Translated by John Hadley. 16: “De duobus hominibus boves ducentibus — Two men leading oxen. London. Lucas N.  Bonnycastle. 1964. 1976. Indian Math.. London. Gauthier-Villars. Labosne. Claude-Gaspar. J. 33. 5th ed. reprinted by Dover. I examined the 7 Ø edition. 2nd ed. P. 214–223.  Bachet.. 24. 1923–24.  Diophantos. 5th ed. nos. Arithmetica. J. 76. pp. 1959.  Bhaskara. Diophantos of Alexandria.  The Demaundes Joyous. Rigaud. 15. pp. 86–88. London. 1973. 18. 1874. A. et al. pp. Translated into German by K. Vogel. he produced a 10 Ø ed. 122–126. Algebra. pp. 6. including Sandig. P. 24. 134–136. and the 13Ø edition. Nos. John. 1612..) Chap. which may be the same as the 1815 edition. Prob. Neun Bucher arithmetischer Technik.  Chiu Chang Suan Ching. Rigaud. 2nd ed. Mar 1992. (There have been several reprints. Bhaskara solves his example by taking ¾. London. PrenticeHall..C. Bîjaganita. Book I. 1968.234
D. 4392. London. Krishnaswami. giving total numbers of 3295. pp. 3200 and 16. 1824.  Bunt. 21. Revised by A.” In 1815. with Arithmetic and Mensuration from the Sanscrit of Brahmegupta and Bhascara. 1st ed.470 . 1813. Heath. “Nine Chapters on the Mathematical Art. “an entire revision of the work. 13. Prob. v. Oxford University Press. 10. Paris. 102– 126.
p. Paris. De Gruyter. 1959. L. Glaisher. Liber Abbaci. 6 of the facsimile. Rome. C. Sam. This is the oldest riddle collection printed in England. London. I. Reprinted by Georg Olms Verlag. p. Nicolo. Pinnacle or Corwin. aka al-Karagi. (1202). 1923–24. 76–77. Teubner. Opera. Metrodorus (compiler).. 1–131. 1976. General Trattato di Numeri et Misure. edited by Grandin. Abo^ u Beqr Mohammed Ben Alhacen.D. Tartaglia. V. pp. English version: Recreations Mathematical and Physical. Mahavira. 850. Ragacarya. aka Srîdharâcârya. 1725.” c. 1696. London.” A. VIII. Bonwick. 4 Ø edition. Curtio Troiano. Leipzig. Vol. Ozanam. Loyd. 90. aka Leonardo Pisano. Edited by J. prob. no. Paton. 28. 256r–256v. Alkarkh.” L’Imprimerie Imperiale. edited and published by B. Cambridge. 53 and 346. Heiberg and H. Sridhara. Translated by M. J. Loeb Classical Library.
. Paris. Prob. ex. prob. Johannes. pp. pp. vol. 1694 (not seen). pp. Vol. book 16. 1912. Berlin. 1010. 50.SOME DIOPHANTINE RECREATIONS
1971. and Helmuth Gericke. Lamb Publishing. Pâtîganita. reprinted 1976. 2 Ò edition. “Ganita-sâra-sangraha. In: Scritti di Leonardo Pisano. Shukla.000 Puzzles. Sect. Jombert. L. W. Government Press. Boncompagni. 3. Amsterdam. A. pp. pp. Venice. Paris. Fibonacci. Untitled manuscript called “Alfakhrî. R.. Part 1. Madras. pp. The Greek Anthology. 900. New edition. et al. 286–287. Euclid. 298–302.” Messenger of Mathematics. Geschichte der Elementarmathematik. 53. 1853. prob. pp. Transcribed and translated by K. pp. 44–49 and 94.. four vols. 1708. Sam Loyd’s Cyclopedia of 5. Jacques.. Vol. Edited by Sam Loyd II. S. Karin Reich. 1556. Tropfke. Tricks and Conundrums. Hildesheim. Menge. Paris. 1914. Edited into French by Franz Woepcke as “Extrait du Fakhrî. Revised by Kurt Vogel. 60-62. 1980. 5. 1982. Harvard University Press. 24. Arabe de la Bibliotheque Imperiale. Reprint. 68–70. 1857. c. 1916–18. 1228. 5. 1916. MS 952. Mass. 1: Arithmetik und Algebra. 24. 201–210. surviving in a single example in the Cambridge University Library. pp. Lucknow. Lucknow University. aka Mahâvîrâ(cârya). 136–139. 26–27 of the transcription. R. Supp. 79–80. prob. “On certain puzzle-questions occurring in early arithmetical writings and the general partition problems with which they are connected. Recreations Mathematiques et Physiques. and Heinemann. Translated by W.
or a winning path for black [1. and Black’s goal is to put black stones in a set of hexagons that connect the left and right sides of the rhombus. White’s goal is to put white stones in a set of hexagons that connect the top and bottom of the rhombus. for in any Hex board ﬁlled with white and black stones there must be either a winning path for white.
.Who Wins Misère Hex?
Jeffrey Lagarias and Daniel Sleator
Hex is an elegant and fun game that was ﬁrst popularized by Martin Gardner . 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. Assume. (This fact is equivalent to a version of the Brouwer ﬁxed point theorem. 3]. The game was invented by Piet Hein in 1942 and was rediscovered by John Nash at Princeton in 1948. there must be a winning strategy for either the ﬁrst or the second player. A hexagon may contain at most one stone. First we observe that the game cannot end in a draw. for the sake of contradiction. as shown by Gale .
A game of 7 ¢ 7 Hex after three moves.) Since the game is ﬁnite. Two players alternate placing white and black stones onto the hexagons of an Æ ¢ Æ rhombus-shaped board.
Proof. our result shows that in optimal play the loser can force the entire board to be ﬁlled before the game ends. As in the case of Hex. the ﬁrst player can win by moving in an acute corner. Let È be the player who has a winning strategy. LAGARIAS AND D. 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. either the second player or the ﬁrst player has a winning strategy. the proof of the existance of a winning strategy does not shed any light on what that strategy is. the winning strategy will work. Gardner attributes the discovery to Robert Winder. was later developed by Yamasaki . 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. Furthermore. the player simply chooses another empty cell in which to play. and Black wins if there is a white chain from top to bottom. and now imagines that this one is empty. An abstract theory of “Division games. In addition to showing who wins. It sufﬁces to prove the second assertion. For any winning strategy Ä for È
. If the strategy requires the ﬁrst player to move in this non-empty cell. SLEATOR
that the second player has a winning strategy. who never published his proof. and let É be the other player. A small step was made in that direction by Ron Evans  who showed that for even Æ .” which includes Hex and Misère Hex as special cases. The ﬁrst player can make an arbitrary ﬁrst move. and an explicit winning strategy for the ﬁrst player is not known. the losing player has a strategy that guarantees that every cell of the board must be played before the game ends. Here we present an elementary proof showing who wins Misère Hex. it is not the case that the second player can always win at Misère Hex. then follow the winning strategy (reﬂected) for a second player (imagining that the hexagon containing the ﬁrst move is empty). on even boards the ﬁrst player can win. Misère Hex has also been called Reverse Hex and Rex. Since the extra stone can only help the ﬁrst player. Because the game cannot end in a draw.238
J. Theorem: The ﬁrst player has a winning strategy for Misère Hex on an Æ ¢ Æ board when Æ is even. This contradicts our assumption that the second player has a winning strategy. and on odd boards the second player can win. Contrary to one’s intuition. In fact. This fact is mentioned in Gardner’s July 1957 column on Hex. the winner depends on the parity of Æ . and the second player has a winning strategy when Æ is odd. and the ﬁrst player wins. The purpose of this note is to analyze a variant of Hex that we call Misère Hex.
when it is É’s turn to move there must be at least three empty cells in the imaginary game. Player É applies the following strategy. (Note that Ñ´Äµ Ñ´Ä Ê µ. Suppose the position is modiﬁed by ﬁlling in any subset of the empty cells with É’s stones. The position is still a win for É. erases the circle. When this happens she has also won the real game. We must show that Ñ´Äµ ¼. so there are at least two empty cells in the real game. suppose that É is the second player. and draws a circle around the cell containing this move. This relationship will be maintained throughout the game. She makes an arbitrary ﬁrst move. such a position contains a È -path. Now. We shall make use of the following monotonicity property of the game. Player É begins by encircling Ô¼. Because Ñ´ÄÊ µ ½. when it is È ’s turn to move there must be at least two empty cells in the imaginary game. The state of this game is exactly like that of the real game. By deﬁnition. because of the monotonicity property. except that in the imaginary game the encircled cell is empty. that cell contains a É-stone. Let Ô ¼ be È ’s ﬁrst move. playing out ÄÊ in an imaginary game. Eventually É will win the imaginary game because Ä Ê is a winning strategy. in that we will describe a new strategy for É in which (in effect) É makes an extra move and then plays the reﬂected version ÄÊ of È ’s hypothetical winning strategy Ä.WHO WINS Misère HEX?
deﬁne Ñ´Äµ 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. The contradiction will be to show that under this assumption É has a winning strategy. she plays instead into another empty cell (chosen arbitrarily). and there must be at least one empty cell in the real game. From now on she applies strategy Ä Ê in what we shall call the imaginary game. Suppose that É is the ﬁrst player. Consider a terminal position of a game that is a win for É. This contradicts our assumption that È has a winning strategy.) The proof is complicated. 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. Therefore it is possible for È to move. We are now ready to prove the theorem. and draws a new circle around the move just played. while in the real game. We shall argue by contradiction. The proof splits into two cases depending on whether É is the ﬁrst player or the second. The basic idea resembles Nash’s proof that the ﬁrst player has a winning strategy for Hex. When the strategy Ä Ê requires É to play in the encircled cell.) Similarly. and further modiﬁed by changing any subset of É’s stones into È ’s stones. because none of these changes would interfere with the È -path. supposing that Ñ´Äµ ½. (We’ll see below that È will not have won the real game. however. Thus the real game can continue.
818–827. and transfers the circle from its current location to the new cell. Simon and Schuster.  M. 73–83. pp. Player É eventually wins the imaginary game. Math. The fact that Ñ´Ä Ê µ ½ ensures that both players can continue to move. and the monotonicity property ensures that the corresponding position in the real game is also a win for É. Excursions into Mathematics.240
J. A winning Opening in Reverse Hex.
. The Game of Hex and the Brouwer Fixed-Point Theorem. Kyoto Univ. and J. M. 1979. pp. pp. Bleicher.
 A.  R. The American Mathematical Monthly. then É arbitrarily chooses a different empty cell in which to move. then by erasing the stone in the encircled cell. New York. The imaginary game is obtained from the real game by ﬁrst changing Ô¼ from È ’s stone to É’s stone. Sci. Theory of Division Games. Yamasaki. It is easy to see that the relationship between the real game and the imaginary game is maintained. Gardner. 14. as follows. The position in the real game is obtained from the position in the imaginary game by putting É’s stone in the encircled cell. 1969. pp. Evans. LAGARIAS AND D. The position in the imaginary game is a winning position for É. Crow. and changing the contents of Ô¼ from a É-stone to a È -stone.  D.. New York.  Y. Gale. J. Publ. If the strategy ÄÊ requires a move into the encircled cell. This contradicts our assumption that È has a winning strategy. pp 189–192. Summer 1974. Res. 1959. Beck. Recreational Mathematics. 327–339. 1978. 86 (10). Inst. SLEATOR
The imaginary game and the real game differ in up to two places. 7(3). 337–358. Worth. The Scientiﬁc American Book of Mathematical Puzzles and Diversions.
Ý´ · Á µ
½Due to the author’s son. Then the vertices of can be labeled with 0’s and 1’s so that every neighborhood is odd. 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. of square . The linear algebra is.An Update on Odd Neighbors and Odd Neighborhoods
Proof: ½ Consider the matrix equation ´ · Á µÜ ½ where 1 is a column of 1’s. of course. Later. be a graph with no loops or multiple edges. if the problem is generalized to all graphs the proof is quite elementary.
½ ½ ¼ ½
½ ½ ½ ¼
¼ ½ ½ ½
½ ¼ ½ ½
. is the set of squares sharing an edge with square and does include square . since we are interested in only odd/even properties.
The ﬁrst solution of Problem 1 was said to be quite difﬁcult. cellular automata ideas were used. while a graduate student. However. Let be the Theorem 1. Suppose Ý is a ½ ¢ Ò vector of 0’s and 1’s. Æ is odd if there are an odd number of squares in Æ with labels “1". Figure 1 shows a 4 ¢ 4 example with a desired labeling. Let adjacency matrix of . the integers modulo 2. Bryan. Æ . A neighborhood. over ¾. We need to show that this equation has a solution for every adjacency matrix .
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 . to all graphs and prove a theorem. so Ý Ý Ì ¼. Theorem 2. It is quite possible that some of the buyers are odd. as before. for every graph . so ´ · Á µÜ ½ for every adjacency matrix .242
L. Is there an Ò so that the developer can sell all the lots in the Ñ ¢ Ò rectangle and satisfy the politically correct criteria that all neighborhoods have an odd number of odd neighbors? When this paper was originally written. E. with all neighborhoods in ¼ odd. and some are not odd. If not all neighborhoods of are odd. the answer to the Developer’s Dilemma was unknown.
Figure 2. We will generalize. But this time the generalization does not yield a solution to the Developer’s Dilemma. Now for some new questions (at least at the time that this paper was submitted). 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. Let be a graph with vertices labeled with 0’s and 1’s in any way. ÝÝÌ ¼ and rank´ · Á µ rank ´ · Á ½µ. Ñ
But is symmetric. we assume all lots in the ﬁrst strip of Ñ lots have been sold.
The Developer’s Dilemma
Suppose a real estate developer has an inﬁnite (at least sufﬁciently long) strip of land Ñ lots wide. Finally. then can be embedded in a graph ¼. a proof that the Dilemma can be solved will also be presented. the vertices of can be labeled with 0’s and 1’s so that every neighborhood in is odd. Therefore.
Clearly all neighborhoods ÆÚ with Ú Ú ¾ ¼ are now odd in ¼. are also added. Perhaps the Ñ ¢ ½ rectangle using the starter Ú ½ has the property that all neighborhoods are odd. Let ¼ be the graph formed from by adding a new vertex. We call Ú ½ .
Proof. Let be the number of vertices labeled 1 and adjacent to vertex ¾ .
Ü½ Ü¾ Ü¿ Ü Ü
If not. Û. Edges from Û to vertices Ú ¾ .
½ ¼ ¼ ½ ¼ ¼ ½
Figure 3. any 0–1 Ñ-tuple.)
. all neighborhoods are odd. and we describe an algorithm that can be used to label the next column so that all Æ ½ are odd for Ú½ and unit square in column 1. Æ is even in ´µ is odd and Æ is odd in ´µ is even. The Developer’s Dilemma can be solved. (See Figure 3(b).ODD NEIGHBORS AND ODD NEIGHBORHOODS
It is rather interesting that only one vertex must be added to to prove Theorem 2. Let the Ñ ¢ ½ strip of unit squares be horizontal. 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. Hence. a “starter" for the Ñ ¢ Ò rectangle we seek. So Æ Û is odd ¾ ¼. We need only to show Æ Û is odd in ¼. there are an even number of vertices ¾ labeled 1 and Æ is even in . But. These vertices now are all connected to Û with label 1. Proof. is even labeled ½ since the sum counts the edges from vertices labeled 1 exactly twice. with ÆÚ even. Figure 3 shows that this phenomenon can happen. Label Û with a 1. Theorem 3.
otherwise Ü ½. and we generate a second Ú Ú ·½ somewhere to the right of the ﬁrst Ú Ú ·½ (see Figure 4). Figure 5 shows an example where Ú ½ col´¼¼½½¼µ. Now that every neighborhood for Ú½ 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. · is odd.. E. and each new column will be dependent on just the preceding two. to show that Ò
¾¿ is a nice exercise for the reader. · · is odd.
· is odd. we can repeat the process for column 2. We will have a solution to our problem if the Ò · ½ column has all entries 0. Let Ú Ú ·½ be the ﬁrst pair that repeats. i. Contradiction! Hence there is a column of 0’s and the proof is complete. and that label can be assigned so that the neighborhood is odd. let Ü¿ ¼. SHADER
will now have all but one entry labeled. let Ü ¼.e. Each neighborhood
If If If If If
This process can be repeated as many times as we like. For Ñ . either we get a column of 0’s or there must be an adjacent pair that repeats in our process. Ò is independent of Ú.e. i.
Ú½ ¡ ¡ ¡ Ú Ú ·½ ¡ ¡ ¡ Ú Ú ·½
Perhaps the next question might be For Ñ ﬁxed. otherwise Ü½ ½. We can generate new columns moving from left to right or from right to left.
Using the second occurrence of Ú of the ﬁrst occurrence of Ú Ú ·½ .
¼ ¼ ½ ½ ¼
½ ¼ ½ ½ ¼
¼ ½ ¼ ¼ ¼
½ ¼ ½ ¼ ½
¼ ¼ ¼ ½ ¼
¼ ½ ½ ¼ ½
¼ ½ ½ ¼ ¼
¼ ¼ ¼ ¼ ¼
Figure 5. otherwise Ü¿ ½. let Ü¾ ¼. is there an Ò such that for every starter Ú the Ñ ¢ Ò rectangle has all neighborhoods odd?.
Ú ·½ . · · is odd. otherwise Ü ½. otherwise Ü¾ ½..244
L. Æ Ò is odd for square Ú in column Ò Ò. and Ò . let Ü½ ¼. · · is odd. let Ü ¼. Of course. Since there are only a ﬁnite number of adjacent pairs of the Ñ-tuple. Ñ is not limited to the value 5.
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 these are suitably placed and oriented. half of the potential reﬂections would not be used. 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. If the ray would not return along the entry paths. It is obvious that the strategy of
. passing once again by entrance mirror E.
Figure 1. Maximum number of reﬂections for two and three point mirrors (E = entrance mirror.Point Mirror Reﬂection
M. 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. Oskar van Deventer
It is well known that a ray of light can reﬂect many times between two ordinary line mirrors.
going to every other mirror. V AN DEVENTER
Figure 2. which is an odd number. so remains valid. However.
using a perpendicularly reﬂecting mirror P yields the maximum number of reﬂections. since the light paths are used in two directions. and what should the positions and the orientations of the mirrors be? A well-known geometrical property may help us: When the Æ mirrors are put on the corners of a regular Æ -gon. from the other mirrors. at the other Æ ¾ mirrors (other than the E and P mirrors) there can’t be Æ ½ reﬂections. O.
Æ¾ Æ · ½
However. At mirror P there can be at most Æ ½ reﬂections. At each mirror a maximum of Æ ½ rays can be reﬂected.
Mirrors on the Corners of a Regular Polygon
Can the theoretical maxima Ê ½ and Ê¾ be achieved for any Æ . Only at mirror P can there be an odd number of reﬂections. So the upper limit for an even number of mirrors is Æ · ´Æ ½µ · ´Æ ¾µ´Æ ¾µ or
Æ ¾ ¾Æ · ¿ Ê½
For odd Æ the parities are okay if E and P are the same mirror. Therefore. At mirror E there can be one additional reﬂection for the entering ray. Æ ½ from the other mirrors and one from the entering ray. At mirror E there can be at most Æ reﬂections. the theoretical maximum number of reﬂections is Æ ´Æ ½µ · ½. for even Æ there is a parity complication. see Figure 2. but only Æ ¾ to make it an even number.246
M. At all other mirrors there must be an even number of reﬂections. Number of reﬂections for different mirror types. the angle between all the neighboring pairs of light paths is 180 Æ Æ . as shown for the regular pentagon in Figure 3.
starting at mirror E. Omit one or more light paths (lines of the Æ -gon grid). Add the starting ray at an angle of ½ ¼ Æ Æ with the Æ -gon. Light paths in a regular pentagon. The ray of light. Rotation of a mirror. Non-Prime Number of Mirrors. after passing mirror E again.
Figure 4. For prime Æ the maximum number of reﬂections can always be achieved.
Using such an Æ -gon grid. makes hops of one mirror. If the number of mirrors is non-prime. Step 2. two-dimensional space is sufﬁcient to ﬁnd solutions for each Æ . say Ñ ¢ ¼Æ Æ . For Ñ ¼ we have the normal orientation.
. A systematical trialand-error search for the maximum number of reﬂections may be performed using the following strategy: Step 1.POINT MIRROR REFLECTION
Figure 3. Ñ ½ yields a rotation of one step clockwise. We take Ñ ½ for E (= P) and Ñ ¼ for all other mirrors. Since Æ is prime. see Figure 5. the investigation of various cases is less straightforward. Only discrete orientations of the mirror are allowed and are multiples of 90 Æ Æ . and so on. the ray will pass every mirror once at each tour from E to E. it makes hops of two mirrors. 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. This is illustrated in Figure 4.
Prime Number of Mirrors. Draw the Æ -gon and all possible light paths (the Æ -gon grid). and so on. Prevent any obvious loops. since a ray can neither enter nor leave such a loop.
however. Reﬂection diagrams for prime numbers of mirrors. (a) Hexagon grid plus entering ray. there are no loops and all of the remaining light paths are used. Searching strategy for Æ = 6. Indeed. The grid consists of 15 lines. reject the candidate solution. At least two light paths have to be omitted for Step 2. shown separately in (c) and (d). In this case. V AN DEVENTER
Figure 5. we do not count 2 ¢ 13 + 1 = 27 reﬂections but only 19. (b) A candidate having 19 reﬂections and a loop. If this is true. If not. Figure 6 illustrates the procedure for the case of six mirrors. Count the number of reﬂections by following the entering ray. decreased by 2 to account for the loss of the two vertical lines on the left and right sides. and so on. Calculate whether the number of reﬂections is equal to two times the number of remaining light paths plus one. Repeat Steps 2 and 3 for all candidates missing one light path.248
Figure 6. then for all candidates missing two light paths.
. The candidate has 13 light paths. a choice of a set of 2 points out of 6. until one or more solutions are found. there is a loop and the candidate has to be rejected. O.
If no light paths are omitted. nor a proof that a solution can always be put on our Æ -grid.
entrance mirror E is positioned in corner number one (
½). 8.POINT MIRROR REFLECTION N
4 6 6 8 9 10 12 14 15 16
11 27 27 51 69 83 123 171 199 227
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
N = number of mirrors. There will always be an Æ Ò loop. except for E. but might be possible. A general solution of the problem has not been found. see Figures 7. R = number of reﬂections. ¯ For Æ there are two solutions. Data for composite Æ up to 16. since light paths have to be omitted to account for all Æ Ò loops.
c = corner number.
A problem was deﬁned regarding the maximum number of reﬂections of a ray of light between Æ point mirrors. For Æ ¾ to Æ ½ . For Æ composite and odd the theoretical maximum of Ê ½ can never be achieved. since only the Æ -gon grid seems to have a geometry that suits the problem.
. which has Ñ ½. 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 position of the perpendicular reﬂecting mirror P is marked with an asterisk*.
m = mirror orientation number. It is likely that this is true. Actual solutions for Æ ½ or Æ ¾¼ have not been found. For all other Æ up to 16 there is only one solution. ¯ For even Æ this sharper maximum is given by Ê¾. d = number of deleted paths. A theoretical maximum was found. the value Ê¾ can be achieved. and 9. in which Ò stands for any divisor of Æ . Ñ must be 0 for all mirrors.
M. and many questions remain open:
¯ ¯ ¯ ¯ ¯
Is there a better search strategy for composite Æ ? What are the properties of solutions for Æ ½ and higher? Are there more Æ with multiple solutions? Can a sharper theoretical maximum for odd and composite expressed in a closed form? Are there solutions that use the back side of the mirrors?
. solutions were found by a systematic trial and error search using a regular Æ -gon grid. No general strategy was found. Deleted paths for composite Æ up to 16. V AN DEVENTER
Figure 8. For composite Æ up to 16.
which can be reached for prime Æ . O.
POINT MIRROR REFLECTION
Figure 9. Reﬂection diagrams for composite Æ up to 9. lasers or burglar alarms?
¯ ¯ ¯
Are there solutions that can’t be put on an Æ -gon grid or that require three-dimensional space? Is there a general theory to the problem? Are there applications to. for example.
In particular. For example. etc.How Random Are 3x + 1 Function Iterates?
Jeffrey C. which predict the behavior of “extreme” trajectories of ¿Ü · ½ iterates. in a trajectory starting from an integer Ò should be of size Ò ¾´½·Ó´½µµ as Ò and trajectories should be of length at most ½ ÐÓ Ò before reaching ½. The predictions of these models are consistent with empirical data for the ¿Ü · ½ function. This paper surveys results concerning more sophisticated probability models.
. deﬁned by
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. Huge numbers of computer cycles have been expended studying it. deﬁned by
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 . it has now been veriﬁed for all Ò ¾ ¼¾ ¢ ½¼½ . where Ì ´¾µ ´Òµ Ì ´Ì ´Òµµ. Introduction
The well-known ¿Ü · ½ problem concerns the behavior under iteration of the ¿Ü · ½ function Ì . The ¿Ü · ½ Conjecture combines simplicity of statement with apparent intractability. Simple probabilistic models appear to model the average behavior of the initial iterates of a “random” positive integer Ò. Lagarias
Abstract. see Oliveira e Silva . The ¿Ü · ½ problem concerns the behavior under iteration of the ¿Ü · ½
function Ì . the largest integer occurring .
We let Ø´Òµ denote the largest value reached in the trajectory of Ò. and Wirsching .e. The total stopping time ½ ´Òµ of Ò is the minimal number of iterates needed to reach ½. LAGARIAS
The ¿Ü·½ problem appeared in Martin Gardner’s “Mathematical Games” column in June 1972 .e. For example. Furthermore the resulting pattern is periodic in Ò with period ¾ . C. Ò Ó ´½¿º¾½µ Ì ´ µ´Òµ ½ ½ ´Òµ Ñ Ò
where ½ ´Òµ ½ if no iterate equals ½. if one takes the uniform distribution on ½ ¾ . who studied problems similar to it in the 1930s. i. Finally we let Ö´Òµ count the fraction of odd iterates in a trajectory up to the point that ½ is reached. and who has stated that he circulated the problem in the early 1950s . 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. see . It is usually credited to Lothar Collatz. i. each possible binary pattern occurs exactly once in ½ Ò ¾ .e.. To describe iterates of the ¿Ü·½ function we introduce some terminology. Prior to that it circulated for many years by word of mouth. ). It was also independently discovered by B. Müller .254
J. Surveys of the known results on the ¿Ü · ½ problem can be found in Lagarias . The ﬁrst mathematical papers on it appeared around 1976 (. so that
×ÙÔ Ì ´ µ ´Òµ
where Ø´Òµ ½ if the sequence of iterates is unbounded.
The ¿Ü · ½ Conjecture is made plausible by the observation that ¿Ü · ½ iterates behave “randomly” in some average sense.. This basic property was independently discovered by Everett  and Terras .. 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. then the distribution of the -vector
½ ½ ´Òµ
Ì ´ µ ´Òµ
½ ´ÑÓ ¾µ for ¼
Ò Ì ´Òµ
Ì ´ ½µ´Òµ ´ÑÓ ¾µ
is exactly uniform. and now more than one hundred papers have been written about the problem. i.
. Thwaites in 1952.
the number of steps ½ ´Òµ for Ò to iterate to 1 should be nor ¡ ½ Ô mally distributed with mean ½ ¾ ÐÓ ¿ ÐÓ Ò and variance ½ ÐÓ Ò. which may be enough for a pseudorandom bit to be extracted. . The pseudorandom character of ¿Ü ·½ function iterates can be viewed as the source of the difﬁculty of obtaining a rigorous proof of the ¿Ü · ½ Conjecture. Garner  and Korec ) they also seem to possess some residual structurelessness. There is excellent numerical agreement of this model’s prediction with ¿Ü · ½ function data. et al. one expects the iterates to decrease in size and eventually become periodic. These mathematical models predict that the expected size of a “random” Ò after steps should be about ´ ¿ µ ¾Ò.3X + 1 FUNCTION ITERATES
One may summarize this by saying that the parity of successive iterates of Ì initially behaves like independent coin ﬂips. Even though ¿Ü · ½ function iterates possess quite a bit of structure (cf. On the positive side. cf. Several heuristic probabilistic models describing ¿Ü ·½ iterates are based on this idea. and a branching random walk model for backward iterates of Ì . Since a number Ò is mul¿ tiplied by ½ ¾ if it is even and approximately ¾ if it is odd. for an explicit constant ½. Since this is less than ½. This model has the drawback that
. . 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. . The repeated random walk model results of  show almost perfect agreement between empirical data for ¿Ü · ½ function iterates for Ò up to ½¼ ½½. . . Cloney. Lagarias  and Luby . Here we survey some recent results on stochastic models.  proposed using the ¿Ü · ½ function as a pseudorandom number generator. In doing so we are in the atypical situation of modeling a purely deterministic process with a probabilistic model. see . which concern extreme behaviors of ¿Ü · ½ iterates. The simplest of them assumes that this independent coin-ﬂip behavior persists until ½ is reached under iteration. obtained in joint work with Alan Weiss and David Applegate. one expects that on average it changes multiplicatively by their geometric mean. 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. so that the average number of steps to reach 1 should be
½ ½ ÐÓ ÐÓ Ò ¾ ¿
¾½¾ ÐÓ Ò
Furthermore. Lagarias and Weiss  studied two different stochastic models for the behavior of ¿Ü · ½ function iterates. These models consist of a repeated random walk model for forward iterates by Ì . see .
J. More recently Applegate and Lagarias ( and ) studied properties of the ensemble of all ¿Ü · ½ trees. That is. see . the successive iterates ÐÓ Ì ´ µ´Òµ are modeled
. It remains a challenge to exploit such regularities to obtain new rigorous results on the ¿Ü · ½ problem. More generally. indeed actual ¿Ü ·½ trajectories coalesce to form a tree structure. They compared empirical data with predictions made from stochastic models. These models make a prediction for extreme values of ½ ´Òµ that is shown to coincide with that made by the repeated random walk model. They 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. C. LAGARIAS
it does not model the fact that ¿Ü·½ trajectories are not independent. Repeated Random Walk Model
Lagarias and Weiss  studied a repeated random walk model for forward iterates by Ì . Lagarias and Weiss introduced a family of branching random walk models to describe the ensemble of such trees. In this model. 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
see Applegate and Lagarias  and Wirsching .
14. consisting of lower bounds for the number of integers Ò below ½. This leads to two conjectures stated in Section 4. The ¿Ü · ½ Conjecture remains unsolved and is viewed as intractable. which are deﬁned in Section 3. 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. 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. Various authors have obtained rigorous results in the direction of the ¿Ü ·½ Conjecture. for these probabilistic models the deviation from independence exhibited by ¿Ü ·½ trajectories does not affect the asymptotic maximal length of extremal trajectories.
For the extremal trajectories for Ø£ ´Òµ in (14. see Figure 1.3X + 1 FUNCTION ITERATES
¿ by starting at ÐÓ Ò and taking independent steps of size ÐÓ ¾ or ÐÓ ¾ ½ £ ´Òµ is the step number with probability ¾ each. the limiting graph of these trajectories has a different appearance. £ ´Òµ is given by (13. For each initial value and the expected size of ½ Ò an entirely new random walk is used. This random variable is ﬁnite with probability one. The random variable ½ at which 0 is crossed. For the extremal trajectories for £ ´Òµ in (14. this explains the name “repeated £ ´Òµ? Large deviation random walk model. The length of such trajectories approaches the limiting value ½ ´Òµ
ÐÓ Ì ´ µ ´Òµ ÐÓ Ò ÐÓ Ò
. Ø£ ´Òµ ´½ º µ ¾ Ð Ñ ×ÙÔ Ò ½ ÐÓ Ò Finally. so that Ø£ ´Òµ ÐÓ Ò.24)." What is the extremal size of ½ theory shows that with probability one
The constant ÊÏ is the unique solution with tional equation
£ ´Òµ Ð Ñ ×ÙÔ ½ Ò ½ ÐÓ Ò
ÊÏ ³ ½
½ ´½ ¾ ÐÓ ¿ µ of the func-
´ µ ×ÙÔ ¾
Next.1) it is a straight line segment with slope ¼ ¼¾¿½ starting from ´¼ ½µ and ending at ´ ¼µ. we consider the random variable
¾ ÐÓ ½ ¾ · ¾ ¿
½ ½ ´Òµ
Ì ´ µ ´Òµ
Ì ´ ½µ´Òµ with ½
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. Large deviation theory also asserts that the graph of the logarithmically scaled trajectories
´½ º µ
approaches a characteristic shape. Large deviation theory shows that.4). let Ø£ ´Òµ equal the maximal value taken during the random walk. with probability 1.
the ﬁrst with slope ¼ ½¿½ for ½ ÐÓ Ò.5) approximates the union of two line segments. and the second with slope ¼ ½ ¿ for ÐÓ Ò ¾½ ÐÓ Ò. solid for ½¼). and ending at ´ ¾µ. starting at ´ ¾µ and ending at ´¾½ ¼µ. and the graph (14.
Figure 2. C. see Figure 2.
maximizing ´Òµ in ½¼
½¼ ·½ (dotted
ÐÓ Ò. starting from ´¼ ½µ.
Ø´Òµµ ÐÓ Ò in ½¼
. solid for
½¼). Scaled trajectories of for ½ . Scaled trajectories of Ò maximizing ´ÐÓ ½¼ ·½ (dotted for ½ .258
6026 0. ½
½¼.1) with numerical data up to ½¼ ½½ is given in Table 1. where ´Òµ ÐÓ Ò .5967 0.527 5 837.5857 0.180.24 16.815 10 13.328. Vyssotsky  has found much larger numbers with high values of ´Òµ.456. The agreement of the data up to ½¼ ½½ with the repeated random walk model is quite good.6020 0. More recently V.13 26.77 31. ½ ´Òµ It gives the longest trajectory.91 32.527 has ½ ´Òµ ½¾ ½ and ´Òµ ¿ ¾ ½ . 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
. In Figure 2 we plot the graphs of these extremal trajectories on a double logarithmic scale.5741 0. Section 5] shows that in ½¼ ½½ trials one should only expect to ﬁnd a largest ´Òµ ¿¼.266.91 19.194.511 7 63.68
0.64 32.48 18.769. and where Ö´Òµ gives the Ì ´ µ´Òµµ in this trajectory. and that
Ò ¿ ¸ ¸ ½¸ ¼¸ ¸½ ¼¸ ¸ ¸¾ ¸ ¼¸¼ has ½ ´Òµ ¾ and ´Òµ ¿ ¾ . fraction of odd entries in the iterates ´Ò Ì ´Òµ The trajectories of these values of Ò are plotted in Figure 1.5818 0. The deviations from the stochastic model are well within the range that would be expected from the size of the expected standard deviation for such models. He found that Ò 12. Maximal value of ´Òµ in intervals ½¼
½¼ ·½.527 Random walks model
½ ´Òµ ´Òµ
70 108 165 214 329 429 592 593 706 755 21.4).254 9 4.728.38 41.6020 0.890. The large deviations extremal trajectory is indicated by a dotted line. 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. an analysis in [15.799 6 8.6030 0.6091
Table 2.3X + 1 FUNCTION ITERATES
1 27 2 703 3 6.884.171 4 52.94 31.400. and the large deviations extremal trajectory is indicated by a dotted line. The agreement with the stochastic model is quite good. Extensions of the data can be found in Leavens and Vermeulen .
A comparison of (14.127 8 127.5927 0.5841 0.68 24.
¾Ò ½ ©
½.86 7.02 21. respectively. (See Table 3 in Section 4.75 5.817. and the number Ê´ µ of distinct edge-labeled ¿Ü · ½ trees of depth seems to grow ½ ¿. In iterating the multivalued function Ì ½.528.645
21.32 2.83 3.903 2.48 12.559 Random walks model Table 3. The branching structure of a ¿Ü · ½ tree of depth is completely determined by its root node Ò ´ ÑÓ ¿ ·½µ.1).671 5 704.909 1. ¾Ò ¿ The restriction to nodes Ò ¼´ÑÓ ¿µ is made because nodes Ò ¼´ÑÓ ¿µ never branch.65 7.831 9 8.897 2.819 1.
15.06 4. The set of inverse iterates associated to any Ò has a tree structure. Figure 3 shows pruned ¿Ü · ½ trees of depth starting from the root nodes Ò and Ò ¾¼.099 1.05 19.86 13.560 1. LAGARIAS
1 27 2 703 3 9. In this case.566.049. or 7 ´ÑÓ
ÐÓ Ø´Òµ for ½¼ ÐÓ Ò
½¼ ·½.65 5. where the branching at a node in the tree is determined by the node label ´ ÑÓ µ. Branching Random Walk Model
Lagarias and Weiss  modeled backward iteration of the ¿Ü · ½ function by a branching random walk.14 11.84 19.) empirically like . In what follows we use the term 3x+1 tree to mean pruned ¿Ü · ½ tree. In fact. there are duplications. see  for more explanation.362.260
J.790 1. these trees have the minimal and maximal number of leaves possible for depth . C. 4. using (15.59 23. ½
½¼.46 2.511 10 77.000
Ò 13.675 7 80.03 19.86 5.788 1.511 6 6. Largest value of
ÐÓ Ø´Òµ ÐÓ Ò
Ñ Ü ´Òµ
. hence there are at most ¾ ¡ ¿ possible distinct ¿Ü · ½ trees of depth .20 6. In  these trees were called pruned 3x+1 trees because all nodes corresponding to Ò ¼´ÑÓ ¿µ have been removed.804. 5.663 4 77.24 16.10 20. it proves convenient to restrict the domain of Ì to integers Ò ¼´ÑÓ ¿µ.976 1. so do not have any signiﬁcant effect on the size of the tree.391 8 319.631. if Ò ¾ or 8 ´ÑÓ µ.791 1.
Theorem 3. and the value assigned to that vertex by the rules above is exactly Ù . where the values of the inverse image nodes of a node Ò depending on whether the tree has one or of value Ò are ¾Ò or ¾Ò ¾¿ two branches. so that an edge from Ò to ¾Ò is labeled ¼ and an edge from Ò to ¾¿Ò is labeled ½. È ÊÏ . where ÐÓ Ò£ ´ µ is the smallest value taken among all leaf nodes of depth of a tree. Theorem 4.1). and the root node is assigned the value 1. where ¬ is the unique positive number satis£ fying ´¬ µ ¼.3X + 1 FUNCTION ITERATES
128 64 32 16 8 4 40 20 10 5 13 224 112 56 28 14 7 74 37
starting from root nodes Ò
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. Pruned ¿Ü · ½ trees of depth Ò . i.e. We assign labels ¼ or ½ to the edges. it was necessary to determine the distribution of the number of leaves in a tree in the branching process. The random walk at a given leaf node ends at a position Ù on the real line.1].
[15. satisﬁes (with probability one)
ÐÓ Ò£ ´ µ È³ ½ ÐÑ ½ ÐÓ This constant È ¬ ½ .. In  it was shown that the expected number of leaves of a tree of depth is ´ ¿ µ . Lagarias and Weiss [15. In order to prove the result (15.4] show that the expected size of ÐÓ Ò£ ´ µ.2).
. cf. where the random walk starts at the origin on the real axis at the root node and takes a step ÐÓ ¾ along an edge labeled ¼ and a step ÐÓ ¾ ¿ along an edge labeled ½. The branching random walk arises from associating a walk to each path in the tree from a root node to a leaf node at depth . and where ¾ £ ´ µ ×ÙÔ ´½ º¿µ ÐÓ ´¾ · ½ ¿ ¿ ¼ This constant is provably the same constant as (14.
they proposed that the normalized extreme quantities remain bounded. they studied the normalized extreme values
Æ ´ µ
Æ ·´ µ
Table 3 gives data up to ¿¼. Let Æ ´ µ and Æ · ´ µ denote the minimal and maximal number of leaves. LAGARIAS
and as ½ the leaf probability distribution is of the form Ï ´ ¿ µ . from a given root node .) The random walk arises from assigning the labels ¼ and ½ to the edges of the resulting tree. 7.2]. (See Athreya and Ney  for a general discussion of such processes. Distribution of 3x + 1 Trees
Applegate and Lagarias ( and ) studied properties of the ensemble of all ¿Ü · ½ trees. Here the pruned 3x+1 tree Ì ´ µ is the tree of inverse iterates of the ¿Ü · ½ function grown backward to depth . where Ï is a random variable satisfying
´½ º µ
Û ´Üµ Ü
´Üµ is strictly positive on ´¼
½µ. 8 pictured in Figure 4. see [15. 5.
16. respectively. Theorem 3. 2. The branching process is a multi-type Galton–Watson process with six types labeled 1. C. Applegate and Lagarias  empirically studied the extremal distribution of the number of leaves in a ¿Ü ·½ tree of depth as the root node is varied. Based on this data.262
J. with all nodes that have a label congruent to ¼´ÑÓ ¿µ removed. 4. This can be formalized as Conjecture C . Since the number of leaves is expected to grow like ´ ¿ µ . where ¼ represents the entering vertex being even and ½ represents the entering vertex being odd. Ê´ µ counts the number of distinct edge-labeled ¿Ü · ½ trees of depth . In this table. Let
ÐÑ Ò ½ ¼
´ µ Ò ½
Ð Ñ ×ÙÔ · ´ µ ½
Then these quantities satisfy the inequalities
Applegate and Lagarias  compared this conjecture with predictions derived from one of the branching random walk models of . the corresponding random walk
4). Transitions of the branching process . 1/3 each) 2. Let Æ ´ µ and Æ · ´ µ denote the expected value of the smallest number of leaves. choosing the root node uniformly ´ÑÓ µ.
¾ along an takes a step ÐÓ ¾ along an edge labeled ¼ and takes a step ÐÓ ¿ edge labeled ½. so that each vertex of the tree corresponds to a random walk ending at some position Ù on the real line. and ¢ for a ﬁxed constant ¢ ½. choosing Ê´ µ Ê´ µ ¿ ·½ so ½ ¢ ¿. 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. The parent (bottom) always ¾Ò (edge label 0). The paper  presents more evidence in favor of Conjecture
½These facts are derived using the probability density for Ï given in (15. 5 or 8
Figure 4.3X + 1 FUNCTION ITERATES
(prob. The number of leaves in actual ¿Ü · ½ trees empirically exhibits less variability than is predicted by this branching random walk model. This value represents the initial value of the ¿Ü · ½ iteration starting from that vertex of the tree.) We consider as random variables the smallest and largest number of leaves that occur among this set of trees. (For actual ¿Ü · ½ trees. 1/3 each) 2 8 5 4 1. and we consider the value of that vertex to be Ù . The analogues of normalized extreme values in the models are
´½ º¿µ ´ µ ¿
Æ ´ µ
Æ ·´ µ
In  it is proved½ that and
Ð Ñ ×ÙÔ · ´ µ ÐÑ Ò ½ ´ µ
These results for the branching random walk model do not agree with Conjecture C above. We make Ê´ µ independent draws of a tree of depth generated by this process. respectively largest number of leaves. 4 or 7 1 7
688 1.243 0.475 0.99 13.166 2.131 2.264
J.557 1.352 0.751 1.481 0.746 1.240 0.33 1.489 0.490 0.307 0.394 0.83 99.748
0.488 0.507 0.249 0.490 0.47 996.728 1.869 1.342 0.190 2.802 1.534 0.780 1.259 0.633 0.
.492 0.054 2.562 0.753 1.488 0.482 0.491
1.265 0.792 1.791 1.32 17.03 177.271 0.507 0.38 236.09 56.57 42.486 0.292 2.405 2.758 1.762 1.390 2.289 0.750 0.748 1.758 1.50 315.481 0.232 2.742 1.500 1.344 2.83 1771.738 1.778 1.562 0.16 4.67
0.633 0.270 2.534 0.310 2.750 1.026 2.68 31.534 0.273 0.
C and formulates additional conjectures concerning the average number of leaves in ¿Ü · ½ trees that have a ﬁxed node Ò´ÑÓ ¿ ·½µ.802 1.252 0.81 4199.782 1.898 1.255 2.750 0.747 1.12 74.75 5599.465 0.211 2.375 2.738 1.911 1.419 2.238
1.60 747.501 0. Normalized extreme values for ¿Ü · ½ trees and for the branching process.49 9.489 0.422 0.76 23.481 0.375 0.36 3149.486 0.923 1.802 1.669 1.741 1.490 0.746 1.475 0.401 0.422 0.305 0.491 0.451 0.255 0.21 5.281 0.62 1328.300 0. LAGARIAS
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
4 8 14 24 42 76 138 254 470 876 1638 3070 5766
1 1 1 2 2 3 4 5 6 9 11 16 20 27 36 48 64 87
2 3 4 6 8 10 14 18 24 32 42 55 74
¿Ü · ½ Function Trees
1.774 1. C.77 2362.688 1.500 1.37 3.784 1.302 0.986 2.34 420.77 133.118 2.782 1.327 2.62 7.78 2.45 560.475 0.268 0.409 0.360 2.898 1.076 2.481 0.246 0.433
10850 20436 38550 72806 137670 260612 493824 936690 1778360 3379372 6427190 12232928 23300652 44414366 84713872 161686324 308780220
100 134 178 237 311 413 548 736 988 1314 1744 2309 3084 4130 5500 7336 9788
114 154 206 274 363 484 649 868 1159 1549 2052 2747
Korec. Ò¸ ´¾Ò ·  C. Comp. Luby. Discrete Math. On the origin of the ´¿Ò · ½µ Problem (in Chinese). Maximum Excursion and Stopping Time Record Holders for the ¿Ü · ½ Problem: Computational Results. Applegate and J. Athreya and P. (June). Pomerance. Math. pp. The ¿Ü · ½ Problem. Applegate and J. D.. pp..  H. Goles.  J. Natural Science Edition.  J. New York.  G. C. pp. Muller. C. TreeSearch Method. Pseudorandom Number Generators in Cryptography and Number Theory.  L. 1972 Number Theory Conference. Computers Math. Leavens and M. 68. 42–45.  Martin Gardner. 427–438. 49–52. 55 (1985). pp. Univ. and G. J.  J.  L. on Stochastic Processes. Hamburg 12 (1991). pp. Mag. Tartani. 42. C.
. Proc. Density Bounds for the ¿Ü · ½ Problem II. of QuFu Normal University.  T. Math. H. Weiss. pp. Lagarias. Ges. of Colorado. Colorado. R. Mathematical Games. 114–118. Lagarias. A. pp. Math. 1996. B. 3. Appl. Garner. Annals Appl. Vichniac.. Math. Conway. Lagarias. Collatz. J. Density Bounds for the ¿Ü · ½ Problem I. Experimental Math. The ¿Ü · ½ Problem: a Quasi-Cellular Automaton. Das ¿Ò · ½ Problem. C. Generalized Parcal Triangles. Krasikov Inequalities. C. 1972. Boulder. pp. Muriel. Princeton University Press: Princeton.  T. R. 64 (1995). Imitation of an Iteration. pp. 44 (1994). Cloney.  D. 58 (1985).I.  I. Rawsthorne. 24. pp. pp. 1990. 411–426. No. 3rd Intl. The ¿Ü · ½ Problem and Its Generalizations. 25 (1977). Slovaca. 349–360. 57–64. On the Distribution of ¿Ü · ½ Trees. Vermeulen. Lagarias. Applegate and J. Math. Lagarias and A. E. (1995). 1990.  D. pp. E. Everett. Oliveira e Silva. Mitteilungen der Math. Vol. 1972. pp. ed. No. in: Cryptology and Computational Number Theory. pp. Pseudorandomness and Cryptographic Applications.  M. pp. and R. Math. 12. pp.3X + 1 FUNCTION ITERATES
 D. Physics and Geometry-Locarno. and Cellular Automata. E. Y. Comp. Unpredictable Iterations. 3–23. Scientiﬁc American 226 (1972). Appl. Proc. (1999). Monthly 82 (1985). ¿Ü · ½ Search Programs. On Heights in the Collatz ¿Ò · ½ Problem. Feix. Springer-Verlag. Proc. 172–176. 101–117. Conf. 79–99. 9–11. Merlini.  K. Prob. 4. (1976). Advances in Math. 229–261.  J. 371– 384.. Branching Processes. Comp. Symp. pp. Iteration of the Number Theoretic Function ´¾Òµ ½µ ¿Ò · ¾. 115–143. 85–89. The ¿Ü · ½ Problem: Two Stochastic Models. C. Ney. Complex Systems 1 (1987).  M. Providence. Math. 64 (1995). AMS. Lagarias. The ´¿Ü · ½µ ¾ Problem: A Statistical Approach. NJ. 2 (1992). Amer. C. C.  D. 11 (1992). 231–251.
72–76. 1681.  G. The Dynamical System Generated by the ¿Ò · ½ Function. Lecture Notes in Math. New York. J.  B. 35–41.266
J. No. private communication. pp. Appl.  V. 21 (1985). Math. Thwaites. Vyssotsky. pp. Math. Springer-Verlag. LAGARIAS
 R. Acta Arithmetica 30 (1976). The Collatz Problem. Wirsching. Bull. A Stopping Time Problem on the Positive Integers. C. My Conjecture. 241–252. pp. Intelligencer 7 (1985). Wagon. Terras.  S.
. 1998. Inst.