You are on page 1of 57
Solving Mathematical Problems A Personal Perspective Terence Tao Department of Mathematics, UCLA, Los Angeles, CA 90025 OXFORD Contents Profoceto the frst edition Prefoce to the second edtion ‘Strategies in problem solving ‘Examples in number theory ‘Examples in algebra and analysis Evclidean geometry Analytic gecretry Sundty examples References Index 01 Preface to the first edition roca an ancient Greck philseper, sid: “Thashrfore,bmatemaicesereiod you ftv ems of ‘hese ges let heron dca she aks tbe nd a ibe teinedery be being ole ines esse abces ‘Shino end ignrace whch areca by th ior {jt ike mathematics because iis fn ‘athenanca problems, oe ule, are impoctant to eel mathemati (ike rolving rea he problems), jst a fables, stores and anecdotes ane Jinporant othe young in understanding elif. Mathersatealproblers fe snited” matheticy where ekgant schon has aleady been end (by smseene ele, of coum), the questcn sipped ofall super ffoousnes and posed in an interesting an (hope) though owing ‘way Inher sere vo peorpetng frp ching a geod nth ‘ematical prcblern cain vo a “ideand seek couse in golyespecng: Yeu ave gen nugget to Bind, and ya know what i 1k ie, tha ‘ur there scmewere,da t tte hatd to reach. hei is unearthing tithin your capabilities, snd yor have conveniently been ven the at ‘suiprene i ta get ce ay be biden iy cumin pee bu it Srl repute ingest eater than digging 2 reach nts book I shall sve sled. problems from various levels and branches of mates. Seared problems (*} indicae an adiioeal evel kfc, cer becae sone higher mathersbes or acme clever tine are equi doubetared questions (°") ae sit ut vo a greater degre. Scene problems have adonal exeries atthe ed that ean be telved na sinilar manper of involve » anlar piece of rathertics ‘While volving thee pelle, wil ry 0 demonsarexome rch of the trade chen protlensolvng. Two ofthe main weapons—expeienee ad Tnoweye—ere ot eay to pt into a book they have ro Be aquired over tine. Bot hee ace rany singles tice thatoke leseimetoleaen, There ae ways of ooking ota pb that make i easier w bad 2 feasible aac plan. There are systemacc ways of educing 2 problem imo successively ‘mplrsub-prblen. But, onthe other hand, saving the peoblers is nt ‘everything. To rear ro the rok nugget aalegy ip-miang the nigh tourhood with ulldces chaser than dong careful survey, a bit of cology, anda small amount of dissing. A solution should be relatively shor, understandable, and bopoflly have a touch of segance, should also be fan wo discover. Trutsfonming a ie, hor lite geometry question jixos avening monster ofan equation by textbook coorinste peony docs not have the same tate of victory a 4 wo lve vec elton ‘As an example of elegance, hee te sandal vel in Puelidean seomety cio angle ae coo) “This neat i oe ie oul baked by cooerate omen. Tr 10 do 60 for « few minutes (hours?) then look ar this solution: " oor. Call she tangle ABC. Now let P be the imersection ofthe serpeodcube bisetoes of AB and AC. Because P'&-on the AM bisvton APT = PBI. Because P is on the AC bisector, JAP| = IPC]. Combining the wo HPT = [PG Bar his means that btm be othe BC tsi, ences he Bectoes ate eancurent (Incidentally, Piste cnc ef ABC o The following reduced diagram shows why [API [Pf Ps on the Alt perpendicuse bet: congruent ranges wil pullit of ey. = ES, ‘This kindof solution—and the sarge way that obviows facts mesh form a natso-obvious faces past of the henry of mothers 1 hope hae yeu too will appreciate his ea. Acknowledgements Thanks to Peres O'Halloran, Ver Telit, an! Lenny Np for eee O,and® > dy wecanalg asumed » 0 without os of generality + (som of anges in a ingle) a + 6+» = 180°: + (ie re) 6 —d)/sina = Bain = (b+ ins = (coring le) = bd? + (6-44)? 20d + deans = (re formula) ¢ = (1/2006 — bsiny ~ (1/2) — db + sing = CO ta snes + fieron’s formula) B = os = b 4 ds — bye — by where 6 = (=~ O-+b+ b+ a)/2 isthe seipesieter, + (ingle inequaiy) b +d < b+ A. Many of tev facts may prove to be wees or distracting. But we ean use some judgement separate the valable at som te unhelpful ones. The ‘equalities at likely berore vel han the inequalities, ine our objec Int and dats ceae in the for of equals. Aad Heron's formula looks ‘xpecaly promising, becuse the semiperimeter simplifies tos = 36/2. Se ‘we can highlight ‘Heron's forma being ily to be seul "We can ofcourse abo draws picture. Tiss often quite hep for geo- tty suction though ths ate ite does net seem add much: Modify the problem slighty. There are many ways to vary 3 prebler| nro one which may be ease to del with: (6) Consider» rpecal ate of the problem, such a exer of eeneste (by Salve a simplified version afte pear, (© Formulate a canjeemse which wuld imply dhe problem, and ty prove that Rest (@) Derive some consequence of the problem, and ey to prove tht fst (©) Reformolae the problem (eg ake the contrapostive, prove by cm readin, ory Some substation). (0 Examine solutions of sila problems. {e) Generale the problem. ‘This i wat when you cannoreven pet stared ona proble, because sl fag fora simpler elated problem somes rea the way to go 00 the main robles, Sina, considering exeme cases and slving the peob- [em ith adetionalaeumprionscan sloshed lighten the general Soliton. ‘Ba be warned hat pec ese ry ther ature, pial and wore leg ‘ant rechnigue could conceivably apply te them and ye have absolutely no Ul in elving the general eae. Tie tends to appr when the special ‘ase foo spel Start with modest sxeumpeios By, brea then yu [re etching as only as ponte vo the spi of the probe, Siegen peter g In Problers 1.1, we can try a special caue such as d = 0. in ths ‘ase we need to find the side length of an eilateral tangle of een Fn this cass i is a standard mater to compute the anewes, wich is b= 208/308 Ths indicates thatthe general anwar sould ako involve square coots ot fourth soot, bur dose not utente sugerst how Wo po about the problem. Conieraion af sma problems draws ite a wel, ‘ecep one pets further evidence that 3 gung-ho algebraic attack whats ected, ‘Modify the problem sigufcandy. tn thie more agsesive type of sate, we perlorm majr mexiieatiors to x peblem such a6 fen lng data, swaping the data wi the abject, or Depating the objective (eae tying disprove a sateen ater than pene) Bascal, we try topush the problem untlit beaks, and then yoo ntl where he beak own occured this idenifes wha the key components ofthe dat are, 26 ‘well a where the mais dificly willis These execs can also help to tng an instinctive fel of wha steps ae likely 0 work, an whi ‘nes a key a In regaed vo our parsculc question, one could replete angle with a quadrateral, itl, te, Neth lp chev: the problem just ete moce ‘complicated. But on theotberhend,onecanseetharenedces not telly need tangle inthe question, bt ut she dnenstons of their. We dona realy need to know the postion of the rangle. So here i ter coer ton that we should concentotecn the idee and srg ve abs) nd nor on coordinate geometry cia appeact, We cold omic some objective; for example insted of working out all the sides and angles we could worl: ox ju the ies, fe example Boxe te can ntice tha by the cosine and sve rls, dhe angen of she erage willbe desermined anyway. So ies ony neccesary to we forte sis, But we know thar the sides have lengths Bd By and b+ d, 20 we cay red {oid wha bef the pron, ‘We can alko omit wore dat, ike te anithmetcdiffercved, but then swe seem to have several pele oltions, and no eng draco slve ‘he probiem. Similars, oruting the area # will ot lave enough dan to lich a solution. (Sometimes one can potelly exit ata for iste, by only specifying thar che area lager ce smaller than sore threshold ‘a; bur this geting compleated. Stk with the simple options fst) [Reversal ofthe pra swapping data with object leat some ince ‘tng ideas though. Suppose you had a riage with artheeete diffrence 4, and you wanted to shrink ie (or when) ttl the ate becomes ‘One could imagine our tangle arin and defoering, we preseing ‘he arithmati ference of He ses. Slay, ne could conser all gles with fixed are, and mld the wane ino one withthe sides in the correc hte progision. Thee vtec work inthe long ba will solve tis question by anothee approach. Da vot fret hou seamen eens ‘bata qution can be soled in more than one way and no parila wy ‘an really be judged the abslute baw. ‘Prove rails about our question. Data istheret be used, 0 ne sho ick up the dita and play with fe Caneprduce more meaning data? ‘Als, proving smal elt could be beneficial later on when ying rore the tain ret find the answer However sal he esl do nfo {4 it could have Bearing ater on. Heiden, ves you someting do you are sick, Ta ‘Fralate.° poblem ike the rangle question, this tacit no | elu. Burone cant. For example, out tactical goal to sce For TS ‘epends cn the ee parameters d and t. In ether words, bs telly 8 far tore ~ Hd,0),(ifthienoestion looks out of place 3 geomet gestion, ‘then thats only because geemsery tend to ipore the Fanctionsl depend ence f object For example Heron's ocnala gives am expliiform forthe are Ain tes ofthe side 2 and in ther words expresecsthe fe on A(@,4,¢.) Now we can prove some mirereulteatext thi fanction 4d), sch as Bt) = et) (ecause an anitmete progression as sn equivalene arithmetic progression wich inverted acdinese deren), fr bee) = bat) (isis done by lasing te riage haces Dia bids) by B.We cull even ry iffereniaeb with respect to dor «For this parole problem, these ae low ut to ptt sme no. tmalaations, for instaneeseting t= Lord = I, and also prone sway to heck the ial answer. However, in this problan tse wes rin ut 0 fly lve mie advantages and re will oto then hee Simplify, exploit data, and reach tactical gels. Now we have set ap rotten and have + fow equations, we shoul ect lok at eating tur eactical goals thar we Baye enablsbed. In sitpe problem, hee ae tsly standard ways lf dmg his: (For example, alebricspiicsnen Is usualy discussed thoroughly in high-school eel wextbocks.) General, this ports the lengst and mest difSculpare ofthe problem: however once can vod peting ln oneremerbers hewn theo che daa a how they can beusd, and ost imporsandy the objective rs also a good ‘deatonotappl any gen technique or method blindly, bt eo sink akead snd see where one could hope such techiguem ake ont thie can allow fe wo save enormous amet of tne by eliminating unprofitable ice Sons of guy before sinking lv af efor in them, and onveraly to five the most promising dietons priory To Problem 11, we ar already concentrating on Hero's foal, We «an ethic o attain eu facia goal of waving for b Afr aly we have alveady need dha che ine and coin rues ca detertine oy once bis ‘now. As farther evidence that thins pongo be a ep foreard nor that "Herons formula invlues d and tin eeence, aes all ur data (we have aleeadyincorporatdthe fact abouhe sides bengin eric progression Sheegesngrtemtig 7 into eu notation), Anyway, Become formula in tert becomes 2G e999) O29 42) | a e eb? — 4a) ra ‘Now we veto solve for b. The rg handsideiss ply in bes loge and as contas) ania facta pada in Now undo on its own This means cha we lave genalaed oo fora pushed out ck emreksnnunterteoy ‘00 much, Lerus reel ourselves in apsn, Something eae could be preserved when we switch gts. Le us tae» look at rome examples—let us tke 4036 agai, siace we have aedy gor seme experience wath thi uber. “The dig switching pees are 4065, 4096, 4609, 4690, 4906 4960 6049, 6094, 6408, 6490, 6904, 6940, 046, 90¢4.9406,9460, 9604, 964, What do they havein common? They have the seme st figs. That ll very well and god, bor the we of dig isnot «very exeful mothemcticl ‘ebjet (nt many theorems and ook use his comer), However, the sm of tags 62 more conventional weapon. And, wel if two mariets heve the sue set of digits, then they have to have the sare cigs. So we lave anche: tra of information: dg siching pcterves the digit sum Combining this with eu previous iota we have ne replacmers question [ktherea power of2 such at there another power of wih hese sume of eit andthe se digt-sim athe ns power of 2? ‘Aes, i this question is es, the exginal qustion i tue, Now this {question is bit easir vo cope with thar the ori, because ‘ornbet of Ags” and “age sum are strand runberheuy salt ‘With shi new conerpein mi lets Took ate eit sire ofthe powecs| ‘of, sings oar ew question imvelvesthem. Well we have Power of Digs 1 26 65,536 25 2 sa wo 1 4 not rete 19 8 204814 524288 29 7 4096 198K S76 3 5 §192 20 10 163882 32768 26 rea this we note that + Thedigt-sums tend ro be quite small Fo ntsc, the digi sum of 217 ina mere 14, This i setaly a small eof had hack, conse sal a ‘bes ac ore likely to match than ate big number. (LF 10 people each randomly pick one wo-dige aunber there ex sncahle (2.596) chen ‘ofa match, bu if hey each pick 10 dig numbers, then vere ly 21 os one in milion chcnce of msteh: something about a Jxsy 2 the hanes of winning the latter.) But the smalzass ofthe names alo aids in picking out pater oo pethap ici not ll bad ecw + Some cigit-sums mare: for example, 16 nd 1026, Butt scr thar the Aigceus soul climbs higher anyway you would expec theta 100. Aig pores of ? wonld hive a higher dig than 4 10-dit one Bot ko remember that we are confining ours to powers of 2 with the ‘ae nber oF cgi, so dis ides wll mo be of much help ‘The upcor ofthese observations i is cgitaums have an esiy ape ‘cable wsercopicsructure (owl increasing ith sf Fee i ay peobuble though noe preven) thatthe dig sum of 2 is appronimacely (45 logo 2) 1.489 fer larg) hut leusy cvoscope ruc The gis just fcsuate too much, We mentioned earth sto digs wes unwieldy now itscems tht digits snot aah ithe. Is there acer ‘eduticn of the problem tha wil eve us with scmething we can realy work wth? Hmm. We mentioned exer that “digicam wnt a ‘conventional ‘weapon in mathemati, Takealockathe preceding gestion fer iatn. But the only real way digit sums ean be succesfully “ramen? is by ‘considering the digte-sum module 8. One ay real that a ranber sequel tits dgitsum medal foe exarle, 3297 = 3 x10" 42 10849 10847 » 10 (mo 9) Soar Peo RTT ed 3424947 (ned 9) because 10 sequal to ted. ‘So new our new medifed suction is flows: scherea power of 2 such thar there isanother power of? withthe cane number of digs and he ste digit-aim modulo 9 asthe int power fz: "Now we ean us the fact that 2 number sequal oti umn melo 9 corephate this question nani: sthrea power of 2 such that theres anether power of with the ve number of dg and these emande ed 9) 5th power OF? [Note tha che pesky notions of curansig dis’ set of digs’, and “sum of dit’ have been cempletely eliminated, which leds promi We itorpesnanteteary Nowler usmodifytheahove bec digitsums of powers of andsce what swe ee (ood 3) wal aot) Power of? Remainder Poyerof? Reminder Power of 2 Remainder to 264 65,536 7 2 2 sot aslo 5 a4 107 2624s 1 88 208 5 Soe2n8 2 6 7 40961 1048 576.4 Bos iz 2 @ 1 163844 m2 327688 ‘What we have to proveisthat no two power of 2 have the sane reminder (aed 9) andthe sme unter edges Well looking ede rable hae ae several powers o 2 wih the ame reminder 1, 64,4096, and 262144 for ‘ample. Bur none of thse four have the same nomber of dit, Ine’, poner of 2 withthe same remainder (od 8) sem to bese separared that thar is ne hope of them having the same number of gs ac, the powers of 2 with dhe same rerainer sets oe quite rally spaced sed onecan quickly sc tht the remainders (rod 9 acpeat theses vet) Sixsteps. This conjecture ca be eaily pened by mele anhmetc a wae 27 = 64 =2" (mee! 9 because 64 = 10d 9). ‘This ceslt means that the reminders of the powers of 2 will repeat ‘themselves ene, Like a epeating decals 12,4, 8755 42,4875, 15248,7,5.oc- Ths i tem mears that two powers of 2 with the sme Alig sum (med 9) mse at eae x step pats Bat en te powers? canoe posibly ave the same mumber of digs, because one wed be 6 tines biper than the ote, eas. Sochis means tat there ate no powers tf 2 with the sroe numberof cis ané the sme digit sum (red 9, We have now proved ot modified question, 20 we can work hackwacd ut wwe reach our odnal question, and wit out the ful aaswers OOF. Suppose two powers of 2 a sted by dgitawitching, This nears tat they have the suve aunbe of digits, and lio have the same {git sum (med 9). Bu the dig sums (mod 9) are peo with 2 ptt ‘of 6, with no repetitions within any given period 0 the two powers are atleast si eps part. Ba then ie imposible fr them to have the sare ‘umber of digs, 2 conan, o 22 cognate “This prolem was simpled repeatedly uotl the more unusable end unfriendly pares ofthe problem were exchanged with more atarah scl ‘and co-operative notions. This smplifeaten can be ait of hi-and mi alfa ther isalway the danger af oversinplifeacion ot mivkmplifeation (plying nc aed hrvng). Buen his question, almost anything was ‘beter than playing around with ipic-suitching, so splifiestion cold not do much more harm. There is a chance thst tanewering an smgling ‘may land you into a wild goose chav, barf you ae relly ick anyway, anything & worth sry. 22. Diophantine equations A Diophentine quatcn ian algebraic equation theclasic one iba? +2 = 2) wth the constraint hall vaibles ar integers Thetsualebjecneieto findallsolutions tthe equation. Generally there emo than one selon, ren if everything ieegral. These equations can be slved slgebaialy, bur one also can use the mumbertbecretiel methws of intr division, ‘modular either, and iepral arertstion, Here one Pronten 23 (Australian Mathematics Campton 1987, p. 19 Find all regers w ch tha the equation Ifa + Ib — n/a + 6) is satsed lor some nena ine ilies fan Bit a O20) | “This ems like a standard Dioghantineequition so we wou! probly begin by multiphing carte dnorinators to ge and thon (ot bt = nab e [Now what? We could eiminae the m sod sy that abycas bt (csi heir symbol | hat we used in Problem 2.1) ory to cen enttte om the fact that ab a unre These venues ate good but they do nocstem te werk on thi reblem. The elticnships ofthe lend righ ses of 2) are mot tong enough, Onesie i agua, he eter product 2 .eamerintcy ‘Gne thing to hep in mind when problem-solving is co be prepared to bandn temporal one inercing bur folless-approach and ep & ‘more promising one. One could ty algebra to attack the problem, then reapply aumber theory ler if algebra failed vo work Exped 2) and cellesing ers we can get +O —mab+bt=0, snot ih ute gained ¥fe-nalor oaS[o-maferai—, “Thslooks very esr, butacully we can tum ths messiness to our advan age We know chat 6 and are ltegers bu thre i gue soot in ‘the formula. Now ths can enly work i the eon aide he agar roo, (r= 2984, 53 pefect square Bur thir means tha 4 ee dana gure © | sare. This is ery restive. Bacau the gaps benwee the squares tt higher than # afer th at few square, oly ee to tet low mabe cfm eases ouhae (x= 27 has tobe & and hence misetet Dor Now ‘we cam werk each cae separately, Sang cer an example of each o a rool that nosuch example exis (Case I 1 = 0. Feeding this boc it, say, (2) we pe (2+ BF = 0, and thse 6-0. But ds peat of in Our erga gator re now have 30/0, which illegal Hence cae bed Case 2:04. Resin, 2) gives ws (4 b? = ab, whic upon collecing terms gives of ~ 2ab-+ bP 20. Factoring this we ge (2 DP ‘oust equal b. This ota coneradicticn, tu an examples 4 = B ‘works when pot into the cignal equation (2 So car answer was » = 4, but i wae obtained by the rather inelegart ‘method ofthe quadkati formule. Using cium, ba fete duces square rot term wich phe thatthe wre the square rece rust be a pefect square, ocanenlly comes in useful _iophantre problems can ec exteely dfcale when ene ofthe vari bles appears in the eqponert the meet notorious ef thee fe Formats las theorem, whch ests tha there are no ratoral number scion to 2+" =c" withn > 2 Fortunately, there ate other pcbles invlving exponents which are ese to hand Prowtens 2.4 (Tayoe 1989, 7). Find all whiions of 27 = a? where wand are ites. 22 cegoen tan This hind of question selly ads trial and errr to fn the sight tack. With Diophartine equations, the mot elementary methods ate modular antes and faceration. Modular aime rarer the eat a, ‘lon o asuiaie modulus, sowetimesconstae (eg 0d 7), oF (med 18) ‘rsometimes variable (eg (id pq. acrorzation alters the pte inte the feu (cer (Ector) = (seething ee), whee the rp ad side «ouldbea constant (the best posible esl) pine, aaguare or sxnething tls that has 3 limited choice of factor. For example in Problem 23, beth rmethos were considered canyon, bax dtcarded in favour ofa alzbeic anpecach, which sectually a factriatenechnagueindxgie(emeabat weevemunll ge (r—2)? —4= (guar "Now its best to ry eleenary vce fire, 6 may save lt of ashing about in ccles lter- One may have abn hee methods and ‘wed te analy the approximate uation xayPiTaxe which ean git ito some serious umber theory invohing topics such ax ‘ontnued factions Pelsequaton, ard econ relations ear be dove, but we wl ook for he elegant (i, lax) way out ‘Obtaining wef fctrtaation is teat to impr, excep when | ‘seven, Then we gota dlfereee of to sqvaes (ital fctonaion in Proghanen equations) Whe so at Pee 2 HIM whee m= nf2. Then we ean say that ~ 29 and +2", big facts of Zymust be 7,1, 1,007: and futher breakup im vss sor shows thst ‘here arene seetons if we ase is even ae that abe uel she facorcation method cartel uit does noe el us where the ctl Seluens ae and how many of them there ae (Although we do now kw hae mast beet) “The roodulraridhmetc approach nent. The stroegy sto se them lusto get id of one or more of the ters Fer example, we coud wee the equation medulo sto ebtain 2°47 =0(nedx) ‘or roaybe module 7,10 84 Bax (wed. Unforunately,chse methods do noc work wel a all. Bue before we gve ‘ep, ther is one more mods to try. We ted eliranating the and the ‘2? ern; cam we eliminate the 2" tem ite? Yes, by chine, ry sed 2. Then we pet 047 == od 25 when > 0, amd 1472 (ood 2) hes = 0. This nor roo bel ae we hose alos eliminated he role of 0 «completes Bur isl doesnot work, ec tern the nico «ould ke or 1,30 we have nor seal exe any posses, Torre the vals of x we ne co choose 8 dilfernt mole, With sie of shovuhimeo ses the vals un the righe-fun salen now tet ley made 4 yea of 2 Ii ote word, we hase OFS =e poet whens on 24322 (me 4) when & Fes=x mals whoo “i esnise a? mut he O (ned) oF (od), pos (3) hited, Thismeans sam anly beO wr 1A quckshetk then shows tha aly ‘on work, andy must be Sr The main ie, when waving Diophaeine equations of the for “nd allseltions’ is liming all bu fst runs of possiblities Th another reason why the tnd 7) a ne! wou ne wenks fe hey id, hey would hare clrinaced all de cases ule the ned 4) apptosely which elunated al bur benef [Exencist 2.2. Find the largest postive ingen such that 3+ 100 visible by m+ 10. (Hint use (ried + 10). Get the by ning the face tt = 10 med 4103), Prone 25 (Hasta 190,973), Prove dae for ayo | nepaieineger nthe nunbee "+ 2" 34 4 scsi hy Sand nly f nis ao die by 4 This poder books a bic uating at fst equaions Mk the abone vay remind oncof Fermat's ls dheorer, whichis norris for sirslabily, Bur our question mach ier We wiht a th 's ori noe iil by 5. Unless a tet factorization eden, we ill have to use the mls approach. (Tats show tha [42° 3p = 0 ted $) for wot diisie by 3, and P42" + 384 4 FO cd $y others) Besse we ate usa such sll numbers, we can evakste some of the volves of 42°43" £4" ined Sbenually The Fest wa odo hiss 0 work ut fed 51,2" ma! 8h, nd Shand 4 tno $l Fetore an tied [Naw itis abveus that some pia evant tn fat 2% 3% ane all pesnde with period 4. To prove eh contre, we waa alle ‘nth the dfition of period “Take 3 forezanple- Saying thai pera with peri juste thee 372 5 ima 9, Bur this easy 0 prone 8 ve Pct = 3" (omnl 5) because 81 = 1 ned 5) earlowanterteer Silay we can prove 1%, 2°, and 4¥ are pesoic with pesca 4 ‘This means thot 1" 42" 43°44" is peciodic wih period 4 Tis tn tun implies that we nly need fo prove our ution foe ® ~ 0, 43,3, came periciciy wil tke care f all the vaber eases cfm Bit we have alesdy shown the question co be tr in thew eases ie the above table). So we ate done. (By the way. there i ¢ owe eventary meth available we assme tat mt de simply pat up sod cancel vecms) ‘Whenever eying to prote equations inseving parame (n his ase ‘perc is always handy, sone no longo nats ces al ses ofthe paramere o verify the equation Ciecking one pera eg © 1,2,and 3) wil be ufo, [Now we turn to reer problem coneeming suns of powers Protea 2 (Sey ok 962 pH 6) Lat fe os | ‘wuters with odd. Prove tha the smn 184 28 "<9 is die byl p2 bond | “Tis quetion,by the way, ina andard exercise Beroulhpobroral (or sore astute applicarions of che Remeinder Theorem), at mvescoing penton of mathematis that hs rany applications or without the sed hammer of encull polyoma ore Riemann ¢ fancon) we wil at fave use pli old ruber theory. Fast of lve know chat 12-4 0.ean abo be writen inthe fon + 1)2. We form stall we ue? The forme ie more sesthet, ba Bt ‘eles ina dvs quetion. (es ahesyeesir if the vor ss0 prodocy rather than astm) might have bee wel here was sure ice facteriaticn of 14 28+--+ 9 which inwelved 1-42 Fb ‘here not at leat, noe an oslo one) If there was Sore Way fo ele alsbly by 1424-4 wo diviry by 1424-4 (v4 1) then Induction might bea way coo, bu tha doesnot sem likely ether Soe wily the mr 1/2 formation insta ‘So, using mou arithmetic (ubich i she mest flexcble way to prove ‘that one mimber divides ancther our cctv isto show ther, Tp octet Oma net 1/2. {eta ignore forthe mesmenc hein the 4 19/2 Then we ae tying te prove saneting of the form actor 1)» ctor 2H(expresion). he ro feces ate copie then en objectives egivalent o proving both of (actor Hlexpresicn) and Caco 2 expen) ‘eparately. This should be snpler o proves exer to prove visibility ifthe divisors are sale, But thee I an acying in the way To eal ‘ith cat we wl st breakup ito cases depen en hehe is ever ‘or od" The eases are quite sar and | wl oly do the se when ns fren foths case we cam wre = D0 soo staring a messy ‘erm in dhe following equations—Ite howskecpng hig like hs belp ‘sation run smooth). Replacing ll hem by Zr we have pve Home Barty 8 (od man +1), Ibu since mand 2-1 are copie, this i eguivalent ro proving VD am 0 trod mt and tatty, + mb =0 (mod mi Lec ustacke she (pod 2041) pare ft ee ite sina Problem 2.5 but isabiccase,besuse we know tht odd Using the rcs 2n 2 inequialen go 1,2 Lsequivalent 0-2, ands en, so oor expression TERRE Gal becomes TEED ham omit boon C2 CE ed 2B ‘We have done sis co that we ean do seme nice cancling kis oso (CEs equal to. Therefore (apt meh. The upshot of thin that ‘he above sum canbe pairwise cancelled: 2 and (2) wl cancel 3 and (C3) will cane, et este 0 2+ 1), desire. “padeganyangneteven cogent Somes 8 2 omegechtan "Now we have to do the (ed) part thas, we have to show Ta or DEF Ot Ort DEE + Om— D+ am = 0 med m9. aa we are working modulo, sosome ofthe above ten can besimplifid sreand 2oate both equivalent tO (wed mann +1 sequent I, ‘m2 is equivalent 2, ed 0 on So the above surraton samples TEE Ron DEE OLE On DE 0 (red ur several ermsappeae tic, recombining (and itching the 0) we get 20423 4 6 — DE eed "Now we can alnest do the sume thi fo the (2 +1 cas, cxcept there is smal ich when mi even Hi Or ca efrrdate the above expresion ae 2a 2 bm DO on DDI +2 4h ened. an doshesame procedure of cnelion aber. atin ieven sm = 2pysay) tees aide, which es ne cael vith athe ‘ether words ths ase he expen docs no caller O immediatly, Dur iaten! cance 2! (av0d 2p). ut tis of course, sequal 0 0, Regards of whether Ms edo even, weve proved thar 1+ 24 3 pst ie dive by nd 1/2 If Exercise 24, Complete the i ofthe ao prob whine ‘sat haps hw [Now let us turn to speci type of sus of power? problem, nay sums of recprocals ‘Protos 2.7 (Shiarsky et al. 1962, p17). Le p bea prime vamber ‘rete than 3. Show that the numerster ofthe (edie) faction IAF B44 19-0 is cvisie by #2. For example, when pi the faction is 141/24 1/3-41/4.= 25/12, andthe mimerstor ie choy divisible by ‘This question is 2 "rove hat question, neta Find sor ‘Show thee exist? question, soi should nee be compet posible: However, we have to prove something abeut a nurnentcr of «reduced fection nt seething exly dale wth! This numerator wl ed tobe transformed ing something moce standard, Uke an algehrsicexpesion, wo that we ca ‘manip i beter Also he gues Coss nx nrc ivaility by prime, it neds civil by dhe square of «pie. Thi ie sigicaty Ferd. We weld like to somehow ede de prtlem to mere prime ivi te make the predien meresebable- So by locking at the shape of the question, we Have the felling ‘objetvs to ke in in: (0) Expees the numerator as 2 mathematical expresion, so that we ean ampli (6) Aim 0 reece the prem from a -2-divisbliyproblan co seth simpler, reshape ap dvsibilypeeblem, Lets tackle (fist Fast ofall wecon pets numerator cay, but the reduced murat neesonly. By adn up de rsctions under scm How docs thas belp us poone tht he reduced tumerator i also visible by p*2 Wel, what isthe reduced ruerster? lei he origin ‘umerator ser seme cancelation with the denominator. Con ceelling etroy the property of pvt? Yn, fa muliple ofp cancelled Burmhiples of p cannot be cancel, betas the dnomintoriscopine top (pisprim, and (pT ean be expresses prov of rambers ess than p). Aha This meses that we ely neat wo prove that he vy locking, eardonnntster ‘pumerator abo is dive by p*. This is ete thon the aber name cause now we have. equatam ese Bx BHP D HLS Hep) FLD 3x32) =O (mod ph, (desi, we have switched over eo modular abet, which i sully the ‘bt way tshow that one nunber divides nother. Hetwever fhe quceon involves mere than one dvksblty for example, somehinginvlting ll “visors of certain number, caer technique arg semctines ener) “Although we have gor sn equation row irs 4 mess Our pen ak so simply it. What we have now on the lefhand se a indie sm of indfte products (Idee just ean tat there ate det dot Oe fhe ‘xpresion) However, we can repcesen theif produts mate rey, Exch infinite prot Resicaly the msnsers fmt to pl waliped ‘ogether, exept fr ene mune, 3), eich is Ferween and pT «can be expresed more compactly a (pI) legate te divide by 4 modulo p* because i copra top, So no out ebjetv inte peove oe =o way Over, z We fcrcze histo get nft edad oo[betede « > (Remember thot we ate destig with molar avthmesi, so dha 4 umber ke 1/2 vill be equalene to a iter. Fo comple, 12-6) = 3 (ad 9) "Now lok at what we ive: something ofthe form (factor) « 09) ro ic wre no for the modular atime, then we cen uly coy tha ‘one ofthe facies 0. With macula arthete, we can sy nes he one ‘hing, but we have tobe careful Lucy the tfc, p-—1) cope ‘0 9" because (p~1) is eopeine op) so we cn dive tout The spe ofthis hat (6) is equivalent etidy Ttaty (Sox tha his looks very similar to cur cignal question the only iter ‘ec being that weareconsderng the entre ston, ex ja the uate ‘of, Burne eanor just jmp fem one frm to anctber without are The shove complications were never) "Now we have reduced dhe question to proving rather benignoking ‘modular archer equation, Bt where to go on fom here? Pehupe oe ‘example will ip. Let us exke the sane example asthe ene given the ‘question: mame, p= 5. We have shabetevenseire sioiay =0(med25) 5 Gesied. But why oe ths wor? The numbers 1,13, 17. and 19 seem tobe random, but magically add up vo the right amount. Pepa ik fake, Leeuw uy p= 7. wtyraa pepe pegs be bata 25433437 610541 amd pepe pepe ded 14254 33437610541 ned 4) =O trmod 491 ‘Ths asthe same ‘fukines about i How docs this wrk? tr noe eat how werything manages to care! ot mala Pelape, herp Jing ebective (bin rin, ve can prove it rf iy shat es Otome ps o He orhingee; it wil ie somthing do, (Besides, if we can oe sole this (mod p) problem, thre no way that we wil be able te solve te (exud problem.) eur eat ube the smplr problem (7) is eh ese to work out For cexample, when pi 5, have $342 44 end 5 ned 5), hile when ps7 we have statet eld tises 243 ¢6tecd A1424 34445 46 (00d 2 xed 7. 20 Rearsinanteteny "Now we have pattern eering the reciprocals 9, /25..41/~1) {vod 9) seem to cover althe ides 1,2, y(>~ 1) id phexatly ones, For example, inthe above equation widh'p = 7, che numbers 1 +4 8 243.4 6 earcmgeto form 1+2+344+546 which 60. Tochedea lengthier example mod 11 yes Lys 1 PhEt Ot pa T 4644434942484 7454 10 mod I) HTN 24344454647 4849-4 10 (ed 1) ‘This tate, showing thatthe reciprocal mers canbe errangd in this rderty fashion, works neatly for (ep), bt dare no gence eesly 1 (mod p"), stad of foundering around eying to fa sun block te 2 round hole although canbe dove if you push hard enough its beter to find Hock thats more round. So what we have to do tow ie End smother proof ofthe fac that f+ 1+ | toon (fp 1) = O md plone ‘that rnerles at eas arly, to he (nod p?) case [Now it tie to use experience with thse ste of prablens. For ample, if we ae fresh from soleng Problem 2, we rw at syne ‘mec, oF at-syrmetry canbe expkited,especallym wed sraantes, In the problem of prong (7) we can make the sutn mute ane oyncc by replacing p~ L with “1, ~ 2 with 2 ad go or wget Bhyh ppt gt ghee ‘Aod now we can pair off ar cancel easly (thet smo “idl erm’ that oes noe pico 8p. an ed pme). Can we dete same in (ned pi)? The answers sort of. When ve sesed the problem (med we aed ‘off Wand 1p ~ D, 172 nd 1/(p~2), advo forth. When we ty he same poring in (med, what we ger not i hi (2 “Woon b oe ‘Now this eis locks ikea complication rather thana simplification. But we have gained very important factor of p onthe righthand se, Now, Inscad of having o prove tt expression) = 0 (ned p?) ve now hae to prone something ike (0 x expresion) = 0 (ned?) which is equivalent proving someting of che form Coxpesion) = 0 (mee, In other words, we are now reduced 0 « (mod ph suesion instead of 1 ined py quatien. Now we have sched ebjectve (B) pve above edu the question co ha of smaller mahi, whic well wot the Slight ineease m complex. ‘Andis quickly seen hat the apparencineressein expression empexiy| fs jut lusonar, ae the (med! p) can get dof ler mone ter than (ne pcan. Now, weeny have to shorter poten tredeg tt ety to unet ps Tx@=0 Fx@=H PH DARw HHA usp ~ 1 is euivalen 0 ~1 (med p), p~2 neguieles ro 2 (nod ph and vo forth, the equation reduces 10 + etiam Pimms cexcquvalently w= one dekehet -twnip. ‘This equation s oe too ad, expt chat thesis on te lean side ‘ends in an obscure spor (at 1/(P~ 1/2), rather than the mere nara Ute ~ 1, for example. But we can “double us, making we of the fat TO , 22 Pemtinnaareeay eo =e ont ayus rate roa -|pt 2 torts Tir eaF a * =f gta] oon ® Soprving that (1/1214 +1 ~1)/27 neat 0 (nod p) woudbe {0 forall posaveiegersm Show tha f(r} fo ll poste integers | ‘This equation looks insulin ro prove what we want. Aer ll, ow fF) 41 10) [Now le us see what we ean deduce, The standard method of desing with these equitions is by substituting peter vale ince the vibes oo et ts stare ih 122/41 “This doesnot ell ws mck about /@) o FH) ast lone, bt he +1 0 the righthand side hms thatthe /(2) cant be oo sll nf 8f ape nto the posi iepers,F/(1)) mane bea sa 30 (2) i at ese 2 "Now, we have ro show tha 2) is actully 220 we may been the gh {tack (Always try to use tactics tat ge you ase tothe abet, we allavalable dite approaches have been eau’, Only then you should think abou going sideways, ex—occsenally—baviwards) So, can we show tha /() sat last 3? Well, we can ry (10) again tog (Q) = ff) + 1. By using the seme aegnen as above we ct say that (6) sat lease 2. Bu ean we say something stronger? Earee we sid fF) was atlas. Peshaps (2) 6 least 2. Indeed, since we “secret know tha fo should eventually equal nme knew tha (GD) {82h we canoe use thar fac ye since we cc sctualy se Wht We are ying 0 peone) Wh this ine oF eho oe can apply (10) ye op 10) 22) +t 2/(A)— 0-414 123. Here we plugued f(2)— ozo she of one enn. This works because swe aleady know thar f(2) 1 isa eat | ‘evi seems we ean deduce tht f(M > Because we use the feta £42) wa at ease 0 prove tht /(3) v8 a ene 3 he eral pool reeks ‘Lindon ‘The inductions usa ie wcky thovgh, Consider the ext ese, show ing that (4) 2 4. From (10) wesw that /1 > FFG) 1 We sloey ove that f(3) > 3, s0 we woul ike de ot (3) 3, node thar weean conclude /()) +1 2 4.To do thet, we wou let have in Ine fect fhe forn ‘n> 3, then fn) 2 3 The eases way to do that {so put that kind of foc ino the indcton we are ying to prove More preci, we wil show Laseca 3.1, fle) > forall a> PROOF. Weiner on n + Base case = 11 This is cbvious: we ace given that fom is «postive inter ees fn) isa est 4 Induction case: Assume dat the lens wer for mand we wall ty prove fm) = +1 forall m2 "41, Well for any me > m+ 1, se scope stants Yee can use (10) 0 oben fon) = fn ~ 1)) + 1. Now for 1) > beace fle = 1) = by induction hypothe. We ca po further sie flon~ 1) = ny then bythe induction hypothess again [f= 1)) = ‘Therefore, m) = flor D)-+ 1 2 1, andthe indoction Kypothes is proved, a we specialize Lerma 3.1 the cas 1, e obtain cur stgcae flo) 20 fora postive aspen, ay [Now hat? Wells wihall anctional equation quetons once we havea now result we should jus play around with cand try to Fecombane ith Previous resus. Our ony previous result i (10) £0-we con put cue new ‘uation int (10). The sly useful eu we pt M4 02 MFO 412 F041 hich flrs once we replace by fi) i 1.10 other word, fon > Fe This very usefl formulas thi means tot fis amines hnceat (wt ebious fom (10), i 2) Ths eae Ut fe) Fn if a aly sme> nThis eons cur engl uation Toe > fifo could be reformulated a ma 1> fom, ‘this, wih (01), proves whet we waned ‘Prosi 32 (Austaian Mathematics Competition 1984 p. 7h Suppose fs 3 function om the pore ttpers which thes ieee ‘vals withthe fllowingpropernex on | (8) Flomn) = fond foe all postive integers rm ad x | | (0 flo > f60) fe > m Fi (1983) (ith seasons ofcourse [Now we have o find out a parla valve of f. The bust wy isto ey to evaluate all off, nor jt /1983) 1983 i just the year ofthe gucion “anyoy.) This f course, asumtog tere ie only one elution off. But implicit inthe question isthe fact tat there cnly ene posse vale of (1983) (here there woul be note than one awed ad bec of the ordinariness of 1983 we might easoably conjure tat thee nly ‘one olution 0 ‘So, what ae he properties of f2 We know that (2) = 2. Repeated pplication of () ek (4) = (AQ) = 4, F8) = (HF) = 8, Indeed, sn ey induction shows that (29) = 2" foe alm. So f(s) = when is « power of 2. Perhaps fs) forall x Plugging tis bck no (2), (5). and (6) shows thar this work: (3) = = ir one elton of a (0, sae) Soi we think thar eh is ex ene solution off, thm tht ene has tobe it So we might wan to prove the more genera ene ucstont The only function from the pesitive ineggers wo the integers satislying, (a) rai Grindancn Tose So we have to prove tht iff sais (b), and (3), then f) = 1, {O) = 2, (6) = 3, and soon, Lt us fis y vo prove fC) = 1 (oh until equations we should ry small examples fs get el ofthe ‘uestin). Well, by (2) we know shor /(1) = 2) ad we know /2) 2, of) ses an 2 And by (, wee (vith = Hand 10) =fnfer, andthe 2-40, “This means f1) muse 1a deed ‘We now have (1) ~ 1 and /(2) = 2. What about FC? (0) 30 bel and (b) ely gies (3) inves of ober numbers ike (6) or [ON wich [aks of ace mich tp (elds 12) <{0) <1 Ina 2 i 2 and (4) is 4,50 22) <4, ‘Bur the only integer berween 2 and 4 is 3. Hence (3) mst be 5, corp mates “This ives us a che: /Q) was 3 only Because c was an inceger (see how this sil to dhe previous question, f(r +1) > /(1)2). Without this restrtien,/(3) could have heen 2-1pc¢ 35, or whatever, Let us sce we ‘an vse this clue more often ‘We aac know (6) = let us ery to work our (3). Using the hope of doing what wedi tof) we eet 10) <1) <1. [Now f(y i 4. Bur what about /(6)2 Never feat: 6s 2 ees 3,80 £06) ~ foye) 16. Therefore, (5) Berween 3 a6, and st be 5 Tinscrs ote ging wes wehie row werkatowal hein ff) upmon= Because we sem to be elyng on past resus to attain the new one, the general poco smells heavily on induction. And because wea a st sino eos el Buse revs we pohly ed Loe 32. (9) = forall Poor. We ase strong indction Fire we check the base ene dees 141) = 1 Yes, we have aleay shown this. Now ssn thar > 2, fe shar / 0) =) Fora saller thon. We wart to shove fn) = Leoking at a Fw examples we will oe ace thar weave 1 ine case oven and el (Case 1 i even ln this ese we can vite = Det or some nee is ess than m, so by stone induction hypethess fl) =n. Hence flor) = f(r) = fifo) = 2n = 8 desired (Cac 2: eis ed. Here we wre = 20+ 1 By fe) fm) Fmd < {(2n-2).By song induction (2) ~ 2vand (m1) = m4 beets both 24-1 and De are salle than me. Now by (8) f+ 2) = FIP D Bont I) = 2a 2,40 our neq beers 2 1) burch doesnt say rach abou st ad ges excep tha they are of the sane sg. However, have nae pec fst: fp st ny old sare ps one: He quae qe of celine of ier factors. Can we ue these troy our avantage? ‘Wall the nicest factor one ca have 0, becuse ha wes the cece tepresion 0 (cially, there ae abn excisions here ving 0 ftor I the las thing ene wants fo have, bee one say win cancel shit factor} &~ a) 0 when, wel ina, 9 hee we have an ies: plugin ferx, Weget Mahe, Geting ack wo pla) and ge) this esl ear that Paglad 1. ‘what does this mean? Very lit, uns oe remembers that pan g have Incegescoefcints and choc the uate eters eo. The upshot ofthis thar pt) and g(a) ate both eteer. So we have te integers mul i eanponger sane te 1. This can only oscur whence ings are ether both 1, o¢ both Insbortand, Pla) = qa) = #1 fee alls =0,4...4m ‘One should ite cata wth de & notation here; we hace ht fey) ‘nd aay), fr instance, ze egul to each ether but pr) and pln) cod have the sme sign or the opposite sgn, fr al we ow rg nom We have found, more’ or lees, the wale of play, -sptn) and an, 9, 0 each polyesial epepae! by n po, But pelt ‘ails only have s many depres of feedom as thew degree Now ff, so the dere of p pls the dre fy eq he degre off, which s 2 ‘This means that one af the polyno sayy has ¢ deseo a mest In summary, weave pelyomil with deiee at mest but rere fo lie on m gives points. Hepeilly ths ean be exploited to @ conmraditon, hic is what we are seach or ‘What do we know abour a ply that has depree at met 2 Well ‘thas at most rots. Do we know anything sbout the sot of Wel ssa factoraffs0the roots pare ako ets off. What are the rors el > “There are none! (Well, rove oo the real Hine at Kast) f i always ese (i foe, alway at eas 1), an bee can ave no rot. The eats ‘ur thar p can have no roots, What dos ie mean when algeria hat ‘bo roots? encanta ever cesses O, tat i reve chain vi In jther word, 7 i thr alvays pesiive oe alvaye native This nes ut fro ease, bu we ca save» lit of woth by wiser that Se ase implies the oh. Indeed f we have ee fatoiation 2) porn, ne automadkaly have another factezation 4) = (Pts) glo). s0 bs lays repative, wecan alas lp the fcorvation aed end up with new factonvaton where p ie alas postive ‘So without ks of geerality we wl rae p 0 be always postive. We already know tha pa) either #1 or 1, anemone new ti postive 8 well, 50 pla) has to be +1 fr all Al ga) forced tobe egal te Pla) 50 al) 8 ako 1 foal. Nos what ‘ell, pos) and at are forced io take on the vlc of +1 atest nies, ‘This cam be phrased in wen of ors, as follows: tx) —1 and gf) 1 fave at lest rots ut ps) —U hase of ott Bese Px) itself hos depres of at most». This meane thatthe only ny pls) can Ina mrooes if pte) —1 has ese exactly, This in arm moe hat pe) ‘sof degree mand hence 4G) fake of diet m. ‘To sum up what we know so fr: We have sssuined that fa) = psd). and g are both postive igor polyacials of depres mad po) 446) = As or aernaively pla) —1 = gla) ~1 = 0, fr all Now we ‘know the roots ofp) — 1: they ate che a. They ate the en rots of P(e) ~ 1 Frcame pls) ~ Lean only hase ros at nt. Ths rears het 22 mime 2 POs) Lis of he form 1) 1 = He ae). a) and sinialy gf) =i ofthe fora, 48) — Ln xa) 03). en, inte rand sare seme constants. Toi tt more about rand, ember that p ad gate iter polynias. Th leading coeficent ofa) — 1 ‘and the leading cette of g(x) ~ Lie Ths menos that rand shave (be incgars "Now we appl thes formula for lx) and qe it ou ial errla Fox) = pines) ad pet oa) FT = OO 0 M9). DD XGbe ape —ay)...be-a) #0 (ana ‘This equation compares two expiiy defied polyuiats. Tbee hing, te do now tro compre coefioeses Comparing the 2 ceefint we get 1 — rand bcsine rad te Inger this means tat ether rs = +1 cr T Ler fre suppose thatr = s = 1. Our plynoml equation noi beep" tea? + (ee ape a) a9 4D Menara). ean) +0, Upon expanding and canceling, this becemes Dx ayvie—a).. be a9) =0 which is idiculous i most old forall). The ase ard we are done. SS eanaindtcadaeae “Brees. Let 7G) be palyomial wth int! octet ad eet ib be integers. Show that /(2) = /() can only equal 1 when 3b ate msec, Hine: far fe) F064 Euclidean geometry | Archimedes wl bern hen Aechyssoger, besuse Inegngc eae macerated ot (GE Fad, "A Nehemaicns Apa ueldean geometry was the fat branch of mathematics to be eat in anything lie he moder fashion (with pestle dations, teerers, ‘uso fori)sand eves now geometry is conducted in avery lps, hy ‘nt fashion, There are severel basic ess whieh canbe utd oyna ‘cally atack and rsohe questions abourpromsncsl objet and ideas This des cane talent extremes with coordinate geomet, which tasfoems pins, lines, wang, aed eles inc a quate mess of coordinates ‘rly converting remem into age Bu the tre ben of peomety isin how a nes-cbvioesleeking fer an be shoven tobe sera ree bythe repent sppicasen of obvious fact Take, for exarpte, Tales heoren (Eaca T31 ‘Toros 4.1 (Thales there). ‘The ale subuended by a danser ie 2 ‘ight angle. fn her word, athe dam Blow, we ave ZAP = 0? Poor. If we dea the ine segrent OP, we have dived Tuto «wo issdestiarges (because [OP] = [OA and (OPL ‘We we [ABI to denote te kength ofthe Line segrent AB). Now wing the fact that nonce angles have equal base angie snd that he sum of the gles in angle 180, we have ZAPB = £APO+ ZOPB = ZPAO + PD = wo" = 2APB, PAB LBA, snd hence APB tbe eight angle, a Geomery is fillings ie this re yousaa check by drawing pice toreand measuring anglesand ergs but ae not immediately elvis ke ‘he theorem thatthe points ofthe four sige of queeteal lays make up a parallelogram. These fact—they have 2 certain smn about them. Prostin 4.1 (Australian Mathematics Competition 1987, p. 12), | ABC arrange that ini cine Thang Bers ofA, 2, | rns the cite at D, Ey F, respectively. Show that AD ie pepenica toEF. J ‘The fist sep, of couse, to dea pcre snd abel wha we er | have taken the iberty of labeling the incense (the inersecson of all she bisesoe, and kay to be impetar as wello ee eerie 2M of AD and EF (which is where we want w eablsh cur riheange). “Ths, we ean write our ebjecve now 26am equation: we want ro show thot ZAME = 90 Thisisafeasilclooking probleme the diagram in cany to dry he con clusion fs quite evident from the iguze. With sucha problem, a dee Approach would probeby wore quite necy We need to compute an angle at M. At frst sight he pone Mi exer ‘emremarkabe, But fer ling in some dota we see that we ales have 2 ‘wea of eer angle mainly doe coll he ange bec, and anges and ces around the place. chaps jus by finding enough anges we can Aeteine ZAM. After all thse ae beaps of theorems ust waiting to be tse the sum of angles av a tangle add up eo 180" the arglesubended byachord onan areisabray the tr; he angle bscors ate concer "We rcedseane angles to are with With che ‘an’ trangle being ABC, sd with all he angle gets and cies sr! ff revolving about this ange it migh be best wo star with the anges = ZBAC, f = ZAPC, Y= 2BCA {itis tadisonal to we Greek lees 0 denote ange. OF couse, we havea + B+ y ~ 180°. Wea fil ia kr of ether angles, tos forenample, CAD = a2 (esha ifyou daw asec of the en sad fin de anges youre) Then, we an make use ofthe achat he agen ofthe rangle add up to 1807, and werk ou sone of inne ange. Fox ‘example iste incest of ABC fhe icrecion of AD, DE and CP, ‘hen Wwe cn easy sy that ZATC = 190° /2~ yy conser the tangle AIC. In fet we ean ge viel ol evar agles—exexp those tM, which ae the ones we really ant So, we est omichone wpe ‘er angle st Yn erm of ange that ae cored! to AL Wel this cssly dong; we coul! ey, fr example, that ZUM, which we wane Be 90° cam be writen a5 IME = 180° — ZF AIF = 180" ~ ZAtF~ CFE. “This is progeess because the two angles ZAIE and ZGEE ate much more ely worked out Indeed we have ZAP = 180° = ZAC = 02 +y and because ual chords sbted eqs angles we ave A PE = ZCRE = 0/2. Hence IME - 1807 — 0/2 —p/2—y/2 = 180" 180"72 = 90", a desea mbes eg ea ce tyaigh a cieweee emma Pieter etary Sacuesunerrmnee Sucachnimecerereeenere Sonate =e Paani nc n,n nde MC de cer | etre 8 meu AG Bg ee ot Ce ef ‘These bisectors meet at O. Suppose that [OD| = ]OE|. Prove thaeeicher | Eerie scarey =| We should fest daw a pictre. Kis le ticky Because one has to sezhowe make OD anal OF the sane lengthy but we ean chet bt by snaking ABC issels or ZBAC = 60" fice we hn thats what supposed to happen stiyway). This ges roo pone confgeatons « LLNb ‘We have only ove pice of data JOD| = 1OF}, and we wish to prove & strange ooking results choice of to properties about ovr tance Bor both properties are anglereatd (ienees tangles ave eal foc angles, and angle bisector are clealy angle related), 50 one eat sath ‘his bean angle problem at is, yay) ‘Once we have decided to atack the problem witha, rerin to fepbrase the given data JOD) = [OF| in ere of angles, The oie way ‘sony th bcese ODF i innclen, CODE ~ COED. Tat ay ok prc, bi is ite hard ro equa the anges COFD and. ZODE any other ongk. In partictar, we wood We thove anges to heer of angles = ZBAC, B= ZARC, y = ZACB bec we want to pone that cer # = 60° or a= y. (Also, ABCis thera’ ange al ter pots sem fom that angle. eis logical eeence rns ll quanti sald bein terms ef the nin triangle) Ba ewe are other ways to tame sider into angie. Lvs look t OD and OE, We want wo rate these lens ro our angles ‘eBay There are several waystoelatesdesand angles: bsitrgnnomety, ‘Siler eiangls, sscls and equilateral tingle sine ae cenne rey to mension few. Bai trigonometric anges and cick, and ‘we do nce have many of ther, We have fe sini tang either and ‘we have already considered the cee eagle approach. The cesine rele ‘sually complicates other tan simplifies, and jar erentes mere unre lena. This only lvves the sine rule asa eal akeratve. After aly ic seats sds co aig quite de. Will 0 ue the sine role, we need a trange oto, preferably those ‘that contain OD and OF and have a let of ages tha we already know. Leoking tthe dingram, and rang cut anges we can guts thas he ‘wiagles AOD, CCD, AOE, BOE could be well. The angles AOE and ‘AOD havea common sid, which should meke the problem single, se we Should ery these tangles it alaya ty toleokforcomnesons Knowing thar evo quantities ae equal would noe elp soles you onc tha it some way) Since we ate only looking a fou of he sx pois (A.D,F,O) ‘we could draw aredoed diagram ro deal wth ther afer all why should fone have to deal with eles eater) "We om that ZEAO = ZDAO = [2 and bythe ft thatthe sm of the angles ina eames 180" we ca week out ZAEO = 180" ay B47 ad sry we hive ZADO = 180" —@ — 8/2 = y 1 BD ‘One can ao filma couple more args connecting 4, D, FO. ae eve ‘lly ur enced dispar locks srmething ike ti tated al ow tp for elon Now we cn ze the sin re. To get woukableexpesion fr |OD} and OE (ich wat why we wanted the sne rl in the fst place) mc canony 1D! __ 10 wat Fino) ~ aly + TDN ~ Ste /2+ APD HOB 1A Sia) Sat 72D) ™ Sala +77 ‘Now tat we have an equation, our given data (JOD = [OFF may reve into somthing wef. Tv length [OA appears in bo of de above ets equations, 0 we shoul perhaps pu (OD) and JOE terme of TOA Wepre and 10 [ODI = |OF] ifand only nty + 1/2) = ip + y/2). (Actually, Ssccnehing ily could happen sch sina) ~0 It os tt take lng ‘ee thar hese svaion only east mentee, degeerare chen and ee freak cases are easly eat with separately, Tsay erember to watch ‘nr fr these things) We have now converted an equality abo ees nto an equakty about anges. More fnporanty, hee cole re relevant wo nur ebjeive rhc involves the anges 0, 6,y) vo we mast be heading ithe night direction ‘The quewion snow ales commltely alge ‘Anyway, ost are equal. This can mean wo thing: PPR By veer 180° (8 + 2 ‘Weare ping close and closes to our objecting the snes ate go and we have ako forthe ist te yxten a tatement that mele aoe Now ‘is mot hard ee that the fs case ade to f= y, while heed eae Beby = 120", and hence o = 6. And we have managels connec ‘nou to stumble onto or objective. ‘And thre we bave it. Sometimes, we can just lap on our given data and hannerng tino an equation resembling eur objocive fete coe, iavlving and chenaphing singleton tower ve wnt W's ce deo fered space Yorks ellwhatkc eee singleton clingy cae ars he protien, Usause hen we can have a ls how to appa ‘reneblen br gata snpliping sd waning ur sin ig ‘took rand mor ive aoe When the cece tease tre may have orate th itive toe we kom ich den Dupe nest prblon deste Prowtin 4.3 (Australian MatheraticsCempetion 1987, p13). () Les ABFE be a rectangle and D be the mtesecion of te dingonals AF and BE. 8 straight Hie trough E res the extended line A at G andthe exended line FB at © s that DC] = [DG Show that |ABIFC) = IFCI/GAl= IGAIAAE. ‘With geometry problems, onc can either work forward (messing sides snd angles spsteancally) or Bacar (toming the ead ees nt sre thing eqoivalee bur sper to wore with) Sanply drawing figure aa _pesing conclusion ishepfl comets, but au gut ard aw {nth cae. How do you forceitv thatIDC| = IDGP Abtof rislandertor {and taking a peck athe cenchsion [ABY|RC| = |FCVIGAI = ICAVIAE crencaly allows onto dear a daca pet Lecus try the forward approach, Hackantalash cnndinatepromesry is ‘ve long and boring way thts prone ro abysialcomplicaons nd huge aoe Lae st ong ai ange a ks lke a tempting location to place the oxigan and corinne axes). Vector romety alo ner suited to wel wih equations ike [DC] = [DG] eat ‘he vector versio are stil usely neater dan the coordinate countepa)

You might also like