Introduction to GEOMETRY second edition H. S. M. COXETER, F. R. S. JOHN WILEY & SONS, INC. New York + London = Sydney + TorontoPreface am grateful to the readers ofthe fst edition who have made suggestions for improvement. Apart fom some minor corrections, he pncipal changes are a follows. ‘The equation connecting the curvatures of four mutually tangent circles, now known athe Descartes Circe Theorem (p12) s proved along the ines aggested by Mr Beveroft on pp. 91-96 of "The Lady's and Geatleman’s Diary forthe year of our Lord TH2, being the second after Bisset, de- signed princpaly for the amusement and instruction of Students in Mathe tates: comprising many wef and entertaining particular, interesting to Alt persons engaged in that delightful pusuit™ ror similarity in the plane, new treatment (pp. 73-76) was suggested by ACL Steger when he wasa sophomore a the Univers of Toronto. For Sinlarity im space, diferent teatinent(p. 103) was suggested by Professor Maria Wonenburger. Anew exerese on p90 attoduces the useful concept Of inverse dite. Anoter has been inserted on p. 127 to exhibit R Krasnodgbki's drawings of symmetrical loxedromes ‘ages 205-208 have Been revised soa o clarify the treatment of efinies (hich preserve collinearity) and equines (which preserve are). The now material includes some challenging exercises For the discovery of fie ‘srometses(p. 287), credit has been given to von Staudt, who satiated Fano by 36 years ‘Page 395 records the completion, in 1968, by G. Ringel and J. W. ‘Youngs, of projet begua by Heawood in 1890. The rests that we now now, for every hind of surface except the sphete (or plane), the minimal umber of colors that wil sfice for coloring every map on the surface "Answers are now given for practically all he excises a separate booklet Js no longer needed. One of the pretest answers (p-483) was kindly sup ple by Professor P. Suz of Budapest HSM, CoxeterPreface to the first edition For the las thirty oF forty year, most Americans have somehow ls it terest in geometry. The present book constitues an attempt to revitalize this sly neglected subject “The four parts correspond roughly (0 the four years of college work However, most of Part Il can be read Before Part {and most of Part 1V before Par IL The st cloven chapters (hat is, Parts 1 and 10) wll pro= ides course for students who have some knowiedge of Euclid and cl Ientary analytic geometry but have not yet made up thir minds 9 spe- size in mathematics, o for enterprising high schol teachers who wish to See what is happening just beyond their usual corniulum. Part IIL deals ‘ith the foundations of geomety, including projective geometry and hyper= otic non-Eucideen geomet. Part IV snlroduces diferentil geometry Combinatorial topology, and four-dimensional Fuclidean geomet. Tn spite of the lage mumber of ross references, cach of the twenty-t¥0 chapters is reasonably sel-contined; many of them canbe omitted on fist reading without spoiling one'senjayment ofthe rest For instance, Chapters 13.6.8, 1B, and 17 would make a good short course, There are relevant exetises at the end of almost every secon; the hardest of them are pro- ‘ied with hints for their solution, (Answers to some ofthe exercises are {Bven athe end ofthe book, Answer to many of the femning exercises fe provided in & separate booklet, available fom the publisher upon re Quest) The unifying tread that runs through the whole work is the ide ‘ta roup of transformations or, ina single word, symm ‘The camomary emphasis on analytic geometry & Ikely to give students the impression that geometry is merely a part of algebra or of analysis, Te is refreshing to observe that there are some important instances (such 38 the Argand diagram desribed in Chapter 9) in which geometric ides are feded as essential fools in the development of these other branches of Inathematis, The scope of geometry was spectacularly broadened by Klein In his Erlanger Program (Erangea program) of 1872, which stressed the fet that besides pane and sold Euclidean geometry here are many othe frometrics equally worthy of attention, Forinsance, many of Fuels own Proposition belong tothe wider feld of affine gometry, which i valid notonly in onnary space but al in Minkowskissacetine, 3 sueceflly Solty Etna apes theory of reat eomety aac nt eninge analy, and conmolgy but ao in Vineet ad cutiann teh hosed wth the theory of Soup an ttn (eerste geomet help the desig of expen Bere iey Tat sero topology (Chapter) haber Meclape u wey hatin sande on some end of ner sacle Speman ta tet the stangen program and Ey stage hn Ge ede ape of fms anaed problem Ua of Susttnemer ey ple apcan Be coloed wih or oir The gail pew oto Shur of cre dled at simmer nt vue fo chat ches and tere Sivas Obisbome: Lamenbre Nova Seuss Ann arbor, Mihi Stanford Caio and Feder Niw'prnsmch ong wih svete publi aren en tothe ems Sip Mathematica New York Cy hy ination ofthe ate Profenor SS Cisbure The mot poplar ofthese epee ces wan he aon th oem ston a psa wich inemboied in Chapter "Apr the geeal emphans on the sea of transformation nd On she deat fspnting some tine i vch sun envionment sf te space and abt fue The Cie novels ees flows» mpl eum oh orhaeter )the ue of dominoes to ast so the sence space groups of twordneninal eywaiogaphy (68a onsrvton rt itarlan pont af laive secon 628) 8 dnp fiom ofthe penal ce pesvingtrarsfrimon (6° and the pa Simrty Tey an taplanatn® of plans 611 encode euiment of Soheters plem 25) an sconomed scm of axons inate geomet {8 Eatjan “absolute treatment of ration eoupe (issy an dementaytentment ofthe norospere @ 168) and ofthe tlm tna sada vm (183), the secon of pest err Soncemning he ap ofthe monkey side (19. an appicton of eo dine par courdnas to tie foundations of hypertee igononety {Fond ite closincan ot eular maps om the phere, projet lane than and Kim bute (235 andthe soggvion of enol Honey. comb 225) Tes nce thks tM, W. ALDAabi, 3.1 Burckbarh Werner Fen che M. Rely ot Scr sna A Shek for ena reading ars ‘haps alo 1H G. Fore, Mastin Caner, and Sno ho Reipin prootendng to 8 H-Gouid 1 Litfewend and) Som for Fermin to quote cevtin pana fom ther pled works aa to BIC Esher Ritossa the Roy Sosy ot Canada permis ‘rcp teas HSM. Coxeter Part | 1 RANGES Euclid Primitive concepts and axiozns Pons ainoran “The medians and the centroid ‘The incrle and the ceumcicle ‘The Euler ine and the orthocenter The ninespoint eile “Two extremum problems Morteys theorem 2 REGULAR POLYGONS 2 aa 3 ba 2s 26 2 2k Cysowomy ‘Xml tection omer Symmetry Groups The product of two reflections The kaleidoscope Star polygons 2 ISOMETRY INTHE EUCUDEAN PLANE ut 32 33 35 Direct and opposite isometries Translation Gide election Reflections an haturns ‘Samay of reels on fometries Contents 2% 2% 2 » 0 Fn u 36 ” » o 626 3 Hietmsew's theorem Patterns on a stp 4 TWODIMENSIONAL CRYSTALLOGRAPHY 4 a3 44 45 46 ar Lattices and thee Dirichlet regions The symmetry group of the general lattice The at of M,C. Escher Six patterns of bricks ‘The crystallographic restrition Regular tessllations Sylvester’ problem of collinear pots 5 SIMILARITY INTHE EUCUDEAN PLANE Bl 52 3 Sa 55 56 Dilatation Centers of smite “The nine-poin center The invariant point of a siniaity Direct similarity ‘Opposite similarity © CIRCLES AND Serene “ 62 83 6a 6s 66 87 6s 6 Inversion in a circle Orthogonal circles Inversion of ines and crles “The inversive plane Couxal circles ‘The eice of Apollonius Cirle: preserving transformations Inversion ina sphere The elliptic plane 7 ISOMETRY AND SIMILARITY IN EUCLIDEAN SPACE m a wa 15 16 Direst and opposite Bometis ‘The central inversion Rotation and translation ‘The product of three reflections Twist Diltive rotation Sphere:preserving vansformations ror tot Part It 8 COORDINATES a a2 3 83 a3 86 87 ae Cartesian coordinates Polar coordinates The circle Conies “Tangent, are length and area Hyperbole functions ‘The equiangular spa “Three dimensions 9 COMPLEX NunBeRs 94 92 93 94 9s 96 97 Rational numbers Real numbers ‘The Argand diagram Modulus and rmplitade “The formula ev" + I= 0 Roots af equations ‘Conformal tansformations 10 THE AVE PLATONEC SOUDS ton 102 103, toa tos Pyramids, prisms and antirisms Drawings and models Euler forma Rashi and angles Reciprocal polshedra 11 THE GOLDEN SECTION AND PHYLLOTAXIS ut 12 "3 ua ns Extreme and mean ratio De dina proportione The golden spiral ‘The Fibonaces numbers Phyiloass wr M0 13 U3 us, 4 135, a7 135 as 0 1s var 13 ae as ue ue 10) 152 155, 157 10 12. tet 16s 169Port ill 12 ORDERED GEOMETRY rt 2 123 1a ns 2g ‘The extacton of two dstint ometres fom Euclid Intermesiacy Sylvester's problem of collinear pots Planes and hyperplanes Continuity Pali 12. AFFINE GEOMETRY Bt 32 43 ne BS Be BT Bs Bo The axiom of pralelism and the “Desargues" axon Dilatations Affiaities Equation ‘Two-dimensional latices Vectors and centroids Baryeentic coordinates ‘Ain space ‘Threeimensional ties 14 FROJECTIVE GEOMETRY 4a 142 43 a 1s ae 147 43 19, Axioms fo the general projective plane Projetive ordinates Dessrguee' theorem ‘Quadsangular and harmonic ses Projects Collineations and corelations The conc rojetve space uchdean space 15 ABSOLUTE GEOMETRY 1s free 1S sa 135, Congruence Paras Isometry Finite groups of rotations Finite groups of someries Geomotical crystallography us im 181 13 186 1 191 1 193 19 20 20s 22 216 2 bs 20 24 ae 28 22 2 28 256 261 269 203 2s 268 2 mm 187 The poybedal kaleidoscope 58 Discrete groups generate by inversions 16 HYPERBOUC GEOMETRY 161 The Evcidesn and hyperbole axioms of parallelism 162. The question of consistency 163 The angle of parallelism The finiteness of wanes ‘Area and angular defect ‘Gree, Roroeyces and equidistant curves Poincare’s“halplane” mode! ‘The horosphere nd the Euclidean plane Port IV 17 DIFFERENTIAL GEOMETRY OF CURVES 174 Vectors in Euctidean space 12 Veoto fetions and their dvivatives 173. Curvature, evolutes, snd volutes 17a The catenary 105 The tacinx Tinted curves ‘The circular helix ‘The general helix “The coacho-spiral 18 THE TENSOR NOTATION 181 Dual huss 182. The ndamentl tensor 183. Reciprocal lies 184 The critical atic of sphere 183. General coordinates IRG The alternating symbol 19. DIFFERENTIAL GEOMETRY OF SURFACES 191 The we of two parameters on a surface 192. Directions om a surtace 193 Normal eurvature x9 2a 207 281 ass Et bos 296 302 M8, 32 3 aT 39 zr 2 as 26 ne ms a 2 as 37 3a ae §Wa 19s 196 197 198 Principal curvatures Principal directions and ines of curvature Untbties Duin’ theorem and Lio’ theorem Dupin’ indiatie 20 GEODESICS 201 302 203 204 20s ans 207 Theoremaegregum ‘The diferential equations for geodesics The integral eurvatue of a geodesic angle “The Enler Poincar characteristic ‘Surfaces of constant curvatre ‘The angle of parallelism The pseudosphere 2), TOPOLOGY OF SURFACES q 1 1 1 i Orientable surfaces [Nonorientable surfaces Regular maps ‘The fourcolor problem ‘The siccolor theorem {Asufelent namer of color foe any surface Surfaces that ace the Fell number of colors 22 FOURDINENSIONAL GEOMETRY ma nr 23 ns ns rasues, REFERENCES ‘The simples foursimensional ures Araceztary condition for the existence of {pg 7) Constructions for regular polytopes (Close packing of equal spheres A statsticalRoneyeomb ANSWERS TO EXERCISES 32 356 39 361 363 366 308 BA ay us 316 an ed 280 32 3S ae or 33 9 396 aw 30 Pit 05 au aa as as a9 Plates 1 The group pa, generated by two parallel lide reesons ” 11 The group em, ponerse by aroetion and pull ie ‘testo 0 ML Awise model the reglar Ibe, (5, 3,3) os IV Chsepacke cls inthe Eulen plane 4‘Mathematics possesses not ony tth, bt supreme beauty ae beouty cold and austere, like thot of seuptre, without appeal te ony port of our weaker notre. sublimely pure, end copable ofa stem perfection {uch oF only the great ort can show seraan RUSSEL 1672-1970)Part |1 Triangles In this chapter we seview some ofthe wellknown propositions of ele mentary geometry, suessing the role of symmetry. We feer to Eucli’s Propositions by is ova numbers, which have Been sed throughout the world for more than two thousand years Since the time of F. Commandino (1509-1575), who translated the works of Archimedes, Apollonius, and Pappus, many other theorems in the same spint have been discovered. Soch reaults were studied in great deta during the nineteenth century. As the present tendency is to abandon them in favor of other branches of mathematics, we shall be content to mention afew tha scm particular imerestng eucup Cocks wr il ng fe il th eta he eet ey Sie one Hao 181 90° About 300 mc. Bucid of Alexandeia wrote a treatise in thirteen books called the Elements. Of the author (sometimes egreably confused with the tater philosopher, Euclid of Megara) we know very ile Proclus (10. 485 a.) sud that he “pu together the Element, collecting many of Eu- thet — See ‘of points has a limit. ee lid “prinape of seperation” tad in proving 1, aes he ges sion whether figure canbe moved without changing incr sacue ‘This pingple i aowaday replaced by » forthe ext assumption soch asthe ion othe nih of angle with tall” igure 20 26, IfABC isa iriangle wih D onthe side BC extended while D's noo dt ore Cand 8 ante Ct ndifBC 2 BC. Ch = Ca AB = A'B’, BD = BYD’, then AD = A'D’. a = igure 10 ‘Tun anomcane wed etn he ton soars es mens to more complicated figures sich st angle, sa that we ca $y pre Gisely what we can bythe flation ° om LABC = 2 ABC. ‘Then we no longer need the questionable principle of supe Brood tt principle of superposition in or- ro iiangles have two sides equal two sides respectively, and have the ans coined bythe eu ies oa, he il a hae th hd ides ‘egal, and thle remaining angles equal respective im fac, they vl Be con. ‘wont tangles Cha Se shi wa it in i epamtsacn1.3. PONS ASINORUM ina: road a rove Sy ing wp th nes Toral Shoo sure ope nosy pepe! sit A. bode 2168) [onion 3, 28 15. The angles at the Bove of an tsosceles rangle ae eu "The name pons asinorum for this famous theorem probably arose fom. tne brigelte appearance of Eusi's figure (withthe construction lines re- {oid in his rather complicated proof) and from the notion thal anyone un Ste to cross ths ridge must be an ass Fortunately, afar simpler proof ‘tas supplied by Pappu of Alexandria about 240 a. (Figure 130) A A Le ARC bean sons wine Wi AB gu AC. Les cons thi spi oito anges anduieia ba way. Since 48-~ AC and AC AB be 9 Ime Talc tee ww see 40 4B” Aw he age BAC cea oe gc Cid rs ne fame. Thfors he concponding pats oh ses ABC 4c areca Inpaiir “The pedagogical ite of comparing the ones tangle ABC wih {ult toometines avoided by jong the apex 4 to, the mips ofthe ise BC. The meian 4D may be regarded abo mrorteetngB to ReSoningy ne sy that an noel angle x Symmetial by reflection, SSN RS Baer! tymmein. (OF course the Weaized miter wed i {lometry hat no thickness ix slvered om both side, so tht it not only Fetects B into © bu ls ests C into.) Any figure, however irregular its shape may be, vss symmetrical i tre when we place it next toa mirror and waive the distinction between ‘object and image. Such blateral symmetry is characterise ofthe external Shape of most animals Given any point P on cther side ofa geometrical misror, we ean con- struct its reflected image ” by drawing the perpendicular from P ta the itor and extending this perpendicular lin ton equal distance om the tothe side, so thatthe miror pespendiculay bisets the line segment PP ‘Working inthe plane (Figure 138) witha line AB Tor miro, we draw (80 ‘ices with centers 4, Band radi AP, BP, The two point of intersection ofthese circles are Pand its image 2 ewe 138 We shall ind that many geometrical proofs are shortened and made more vivid bythe use of elections. But we must remember tht this procedure fs merely a shortcuts every such argument could have Beea avoided by means ofa circumlocution involving congruent wiangles. For instance, the bove construction is valid because the angles ABP, ABP” are congruent, Pons asnorum bas many useful consequences, such asthe following five: IML, fa damevr of cic ie achord wih dosnt pas throuph he cee, ‘is pependar oo prendre EDN facile ange atthe cnte elt angle othe cicumfrnc, when the ray forming the anger mec the our in th sae ane IML21. na cele achnd bend pul angles tan two pits on he same ne of to ars decrmined byte chord ei Fgue Ie. POU = 2 PPO) 1IL22 "The ppote anes of an garage nbd nie ae tte qual (opens IIL32.fachord of cre be dr frm the pon of comic ofa tangent the angle ay a eran ect a ct pa of en ore ofthe chord. a Fp ‘We shall also have occasion 1 ute two famine theorems on similar angle:LX 1g ine be own pr! 0 ne ie of riage i wl athe cer suas ropanncey a fw sf he tangle be a opr thee Inge pm of ton if be parle romain se Vuk. iferepondng ange of te rangle ar egal ten coeypondng ie are potions ‘Combining tis last result with IH21 and 32, we deduce two signiisnt propetie of secant ofa cil Figure 13) Ree ML3S, fina cic sgh Hee eah ote, the rectangle cond by the eget of hone sal the tang ane the meen of ei GP's oP = 00 x OO HL fom a point ude acc a scan and angen Be down the rectangle contained Belle canted he atte the cl wil B ut the sare ihe tpn Gs. OP OP = OT Book VF also contains an important property of are: Lis. Simr wanes reo ae anther nh squared rat of ee arene sides oe, ABC ed 4 esa tangle tel eas age ithe alo AP a8 “This result yields the following easy proot fr the theorem of Pythagoras [see Heath 1, p33; 2, pp 210, 232, 269} 141. Ina righ angled rangle, he gure onthe hypotenuse sum of In the triangle 48, right-angled at C, draw CF perpendicular to the hypotenuse 1A, asin Figure 3d. Then We have thre similar rightangled tuianges ABC, ACF, CBF, with hypotenvses AB, AC, CB. By VL, the reas sfy AB 4 a iS oor co 3B a Evidenly, ABC = ACF + CBF. Therefore AB = AC? + CB? ewe 13 Using etangula Cartesian corinne, show thatthe sefeton in he yas (= ojreveres thesia of What happens when wert in thle =) 2 Dede 4 om 1L 36 (applied tothe ie with ene and rads 40) 3. Inside asguare ABDE, takes point Co that CDE isan ones angle with snaer 15° at Dand &Whatkindof tounge ABC? “4.” Prov the Erde Mord theorem: If Os ay pit inside wane AAC and 21g Rae of poet am O upon the apes ts B,C 04 + 08 + OC > OP + 00 + OR) (dtnc* Let Ps and Pa the et othe perpendicular fom Rand Q upon BC. De line analogous points Q, aed Qu Ry and fy on he other sien Using te many ofthe amples PRP, and OBR, express PP i tras of RP, OR and OB Aer sb ‘isting such eresons at 04 + 08 + OC > OAD? + PRYYRO + OBO + 00. PR POCR RR OR, collet he erm ining OF, 00, OR respectively) "Under what cicomstanses can the sgt > it ES 4 Be placed by =? 6 ne nation of Ex (04 x 08 x OC > (00 + ORKOR + OPKOP + 00) (A. Oppenheim, American Mathomancl Monthy, 68 96D, 9.230. See abo L 3 Mordel, Masomara Gate 46 1969, pp. 212215) 7 Prove the Stine-Lehmos theorem: Au) weighing two eal itera ange oto each measured from ere tothe oppose ie) wees (in Ha ‘eagle as vo ferent ole the smaller ange hse longer trl set) esta Bac Me Tinh ed 9.99: 1B1.4 THE MEDIANS AND THE CENTROID ‘lune wnyese Geet mates permenant” ae a (trays dra este nung dma om hn 7 1 (e221 ‘ewe 140 “Thelinejoininga vertex of rangle to the midpoint ofthe opposite side iscalled a median Tet wo ofthe three medians, say BB" and CC, meetin G Figure 14). LetL and Sf bethe midpointsof GB and GC. By Euclid Vand 4 (whieh ‘were quoted on page 8). both C’B' and LA are parale o BC and half as Tong. Therefore B°C°LM iva parallelogram. Since the diagonal ofa paral: Islogram hiset eachother, we hve BG=GL=18, CG=GM=MC. “Thus the two medians BB, CC’ rsect cach other at G._ In other word this point G, which could have been defined asa point of wsection of one tedian sale a point of trgection of another and siiany of the third We have thus proved fby the method of Cour 1, p58) the following LAY The vce medians of any riongle al pass through one point This common point G ofthe thre medians is caled the centroid ofthe wiangle. Archimedes (287-212 nc) obtained itas the enter of gravity of 2 telangolar plate of uniform density 1 Ayn hinge lai te cam ote meds of enge esbeven pnd pw p ete sm ‘ofthe sides. [Court 1, pp. 60-61.) = 1.5 THE INCIRCLE AND THE CIRCUMCIRELE bt Bacon 180-90) [as Os Dons Sey) redo octid IL tls us hat circles ymmetial by refeston in any diame- ter (whereas an elipse i merely symmetrial shout two special damstr the major and minor axes) Tefalows thatthe angle between two iter. Secting tangents bcted bythe ameter through ther common pont rod ames pce int a oe Sh at mesBy considering the loc of poiats equidistant fom pais of sides of r= age ABC. we sce that the inleral and extemal biseiors of the three fngles of the triangle meet by thes in four point I, JI a8 0 Figure isa. These points are the centers ofthe Tour circles that can be drawn 0 touch the thre lines BC, CA, AB. Oae of them, the center being inside the tiangle.s the center of the inscribed ciel ox miele (Euclid IV.) "The other three are the exceters Io Jy 1 the centers of the three escbed {Srses or evetles (Court 2, pp. 72-88}, The ead ofthe icicle and exci ‘les ate the tnradiue rand the exradi ro in describing a triangle ABC, iis cosomary to call the sides a= 8G cA, = AB, the semiperimeter rake sb +o the angles 4, BG and the area & Since A+ B+ C= 190", we have 1st BIC = 90" + 44 result which we shall find useful in § 1.9. ‘Since FBC is triangle with base a and height its thee such tangles we deduce daKetd tore Similarly A = YO + € ~ om = (6 ~ ore Thos 182 Baar (= of, = 6 Dn = 6 ~ Ore ‘From the well-known formula cos 4 = (88 + ct ~ o8)/2be, wefind also nat =F = ct 4 El 4 Dea + LAY, sind whence besin a [et = cag aoe 4 2 + Das [er brof-a bd + lab + Kar b— oP = le ae — Xs oh “This semarkable expression, which we shall use in § 184 is ateibuted to Heron of Alexandra about 0 4.), but it was really discovered by Archi imedes._ (See BL van der Wactden, Science Awakening, Oxford University Dress: New York, 196, pp. 228,77) Combining Heron's formla with 152, we odin se) Another consequence ofthe symmetry ofa irl is that he pespendicular bistbion of the tire sides of tclangle all pass through the circumcenter O, 1s ays Xs =) 2 (8)? = =D 7 at aoe re 1.56 Which sth center ofthe circumscribed eice o eicumcirele (Euclid VS), This isthe only ctcle that can be drawn through the thee vertices 4, B,C: tsradus Ris called the cireumradus ofthe tangle. Since the “angle at the center" 2 BOC (Figute 1.8), is double the angle 4, the congruent right- ingle triangles OBA", OCA’ each have an angle at O, whence ase nd ~ eB anc Draw AD perpendicular to BC, and let AK be the diameter through 4 ofthe etcumeiee as in Figue 1Se By Euclid I1L2, te ight-anged ti “AKC ate sila; therefore 5G —BXs— A +s ~~ a) +59 — ae — So ays— Bsa) =AGetntnn, Hence the five radiate connected by the formula 136 aR Let us now consider fur cits Ey, EEE tngent to one albert sit dsint pnt tach cre fy ha end deed ste repeal of is radius with sutabie sgn attached, namely, ial the contacts me ex ternal a inthe cue of the hight les fn Tighe Ld), she bend te ll ibcaiertar fe cues @eseniads washers (es pa thea or taPeers heavy circles the bend ofthis largest icles taken to be negative: and a ine Counts as eee of bend 0.” In any cae, the sui of al fur bends i positive, Tn leter of November 163 to Princess Elisabeth of Bohemia, René Descartes developed a formbla relating the rai of four mutually tangent ficken the “bend” notation itis 187 Mette eet twain teteter “This Descartes circle here was rediscovered in 1842 by an English ame ‘cr, Philip Beecroft, who observed thatthe four etles Ey determine another Se of for irles Hy mutually tangent atthe same sb points: Hy through | the tree points of comact of Ex, Ey Ey and so on. Let denote the bend OFM ifthe centers of En, Eq, Ey form a triangle ABC, Hy is either the 18 Inthe later (Figure 1.5) In ether case, we se from 1.531 that fen ton Hae ¢ +de 2) seas tnd of ourse we can permute the sub- Similarly nens 4 19m + m2 seripts 2.3.4, Hence Gay seit bee + Bete bo Dea = Se + Ss Since this expression involves and m symmetically, itis alo equal to (nye ths tet otuemt mem tm > Also, since (vte tea betetaatstebay nat ett eta nt of + dae (roms nom + mon) + (ms +) + One +) — (ne +) + IE = 2iyone + none + nan) + 20 = Dod + me + 19 a teta—nam™ ‘Adding four such equations ater squaring each sid, we deduce 2? whence zn, Doe? = Bel + Bat aie “Thus 1.57 has been peoved In 1936 this the oem was rediscovered aguin by Sir Frederick Soddy, who Inud received a Nobel prize in 1921 for his discovery of iotopes. He ex: pressed the theorem inthe form ofa poem, The Kiss Preise’, of which the middle verse runs as flows: Four cies tthe ising coe, The sale see tenes, “Th end jut he vee of ‘The tans fo he etre ‘Thou thew mtg ft Elid dam eve fn ug ods oe ‘Sock (tsb rsh wpe inSc so bn Send straight hie ‘The sum of the squares of al four bends Inkl te tuareof tr am 1. Find the locus of the image ofa aed point Py refeton in avaiable ne Ahrouph another aed pot 0. 12 For the pnera ange AUC eed he deter 2. The lene of he tangents rom the verte 4 to these ad 10 the hee excels ate epectvely 44 Thecicumcenter ofan obnse-anglt wan is oui the tae SS Wheres the cicumoenter of might anged ing? & Let U FW be the pins on the espostne ies BC, CA, AB of wingle 7 Atangeisghtanped Pun oly it + +20 = The bends of Beers ight os say 9. or ny fou umberssasing E+ + m4 here sa “Beer coe ee ee ee em ge RE gy m+ wh Oe (We Expt tenn 8 aor ancons fH Ae) 10, tthe ee, externally tagcat o ome aot, Rave cents feming sanpe AC, hey ae al tangent 1 ter eles (or pony a cle an) soe bens ae 11._Givena point Pon the citcumcie of an the fst of the perpenseulrs fiom Poth esd sl icons sgh ie (Tas ne commonly eae te So Solin of Pit reaps to he rane although wae metoned by W. Wale, tty pears ater Sion’ death ohowon Ys. 18) 12. Givenaviangle48C ad pons Pin plane Ost maton side noon the i= omc sc eth rd ange ede eto be perpen ‘sing the same F) and 45 BCs from 4aB2Cs. Then Ay BC del similar to AC. [Coy tip 23) (ina 2 Phas Pac. — 2 Peabo = CP Bsty) Thine har eee eteed by BM Sioa fom the tid deed range a ang fo he th ‘erwed won fan erg, Cloweae Mathomatea! Matty 490980), #52 260. 1.6 THE EULER LINE AND THE ORTHOCENTER Atgh he Gres wre int enn poy bt ln ‘seo oe fais of tana eee eae oe ae Mycsltenseryeher on cra eam From now on, we shall ave various occasions to mention the name of Euler (1707-1783), Swiss who spent most of his fein Russia, making i= portant contributions to all branches of mathematics. Some of hi simplest Uiscoveres are of such 4 nature that one ean Well imagine the ghost of Euclid Saying, "Why on earth dia’ I think of that™ re 1.80 the ciscumcenter O and centroid G of triangle coincide, each median is perpendiculac to the sive that it bisects, andthe tangle is "isosceles three ‘nayn that is equlateral Hence, if tangle AIC isnot equilateral, ts ucumcenter and ceatoid ie on a unique line OG. On his so-called Euler line, considera point H such that OH = 306, thats, GH = 20G (Figure 16a), Since also GA = 24°G, te later half of Euclid V2 tells us that is parallel to 4°0, which isthe perpendicular bisector of BC. ‘Thus A is Petpendiculart BC. Similarly BH is perpendicular to C4, and CH t0.AB. “The line through a vertex perpendicular to the opposite sides called an tirade. ‘The above ematks ef Cour 2, p. 101] show that The three attudes of any triangle apse through one point on he Euler line. ‘This common point HM ofthe thee altitudes is called the othacenter o he rian 1. Through ech vertex of nen ingle ABC daw ane paral to the oposite side, The peer bec o tess the tang rmed sagen aa ative prov tat th testes of AC areconcarent, 1 The ertocenrofan obtuse ange angle bes eee wane1. Wheel the othoener fright angled tangle? 4 Any triangle having ro equ altos isosceles ‘5 Conseut an onses tangle 4C with base BC given the median aioe BE (ne The centouitwo-thid the ay Tom B10 BCH, Pre ‘sexta “The aiade AD of any tingle 48C olen 1. ind the perpen dstnce rom he enti to these BC {tebe Euler line pases hough a vere, he tangle ether ihtanled or ese (oF bth > te Euler lines pr thie BC theangles Band Coats an Bun C= 1.7 THE NINEPOINT CIRCLE ‘hi cl the ly xing ae appa i ny con shen gana Dave Poe (910-) (Fedo 9.1) ‘Theft fh tudes hati he pint Din Fite 1.4 rm taco vung or sped tangle) of ABC. The cee of the ire aml scaled ne pt eedetor “Feuerbach ce) othe Sete aeebxese conten nthe fet ofthe tes aude SOESBo sc ther sgaean pot Tat PV Themudoins ofthe tree sds he mpi of he ines nn he oracentrto here Varese thejoe ofthe tread a fe on cae Proof{Coxeter 2,929). Leta’ B,C! A’, BY, C'be the midpoints of BC, CA, AB, HA, HB, HC, andlet D, E, Fhe the feet ofthe altitudes, asin Fig- tre 17a. By Fuchd VI2 and 4 again, both CB and B°C” are pralel to BC while both BC” and CB" are parallel t AH. Since V's perpendicular {© BC, it follows that BICC" in a rectangle. Similarly CA'C'A™ ia rectangle, Hence 4’4", BB", CC* are three diameters of acitcle. Since these diameters subtend ight angle at D, EF respectively, the same cise passes through these points to, 1F four point in plane ae joined in pairs by sx distinct ins, they are called the rerices of a complete quadrangle, andthe lines are is 5x sider “Two sides are said tobe opposite i they have no common vertex. Any point intersection of two oppoite sides called a agonal pot, There may be 'smany as three such points (2 Figure 1.78), Ta triangle 4BCis aot ight-angld, its vertices and orthocenter form a special kindof quadrangle whose opposite sides are perpendicular. In this {erminology the concurrence of the three alitdes canbe expressed a ole lows 1.72 Iftwo pairs of opposite sides ofa complete quadrangle are pars of perpendicular lines, he remaining side are ikeite perpendicular ‘Such a quadrangle ABCH is called an othocntri quadrangle, Is six sides BC, CA, AB, WA, HB, HC rth sides and altitudes ofthe triangle AC, and its diagonal points D. E, ‘ate the feet of the alttudes. Among the four vertices of he quadrangle, ‘our notation seems to sive a special oe tothe verter Cleaty, however, 1.73. Each yertex ofan orthocentric quadrangles the orthocenter ofthe ‘riage formed by the remaining thre vertices “The four trianles just one of which is acute-angles) onthe tangle and consequently the same nine-poiat ice Mis proved in books on ane geometry (wich as Coxeter , 8.7] thatthe midpoins ofthe x sides of any complete quadrangle andthe thre diagonal pointsal icon a cose. ‘The above remarks show that, when the quadrangle ISorthoventi, this “aine-point conic” reduces ta isle have the samexenests 1 OF te nin points decribed in 1.7, how many coin when the rangle (ohvoeten (equate? 2 The fet of thealitades decompose te ninepint iret he tes. Ifthe tangles sean, the ean st ofthe ne point ar dtbuted among ihe three eva fllows Onc as conta stone thes oi pote conta (Wo, a8 thethnd contains ee “On the are 4’ of he sine point oil tke the pot ones of the way fioma'toD. Take punts, Zsa on here ECE. Then XYZ sam eG lata angle Mt The inceoter and th encentr any ingle form an orhosentcquadanae (cass op. 24), Inthe notation of 15, the Ellie offs 10. {© Theour ingles that our in an otocente gandragle bave equal circum ra 1.8 TWO EXTREMUM PROBLEMS ast hv ec of mae ot ‘Cayce mtematr ban nm .t Hrd Bop 28) There wi nal cl we can ine ovo Hon tosamacer(872- | [Boer opts Wo ‘We shal describe the problems of Fagaano and Fermat in considerable detail because ofthe interesting methods used in solving them, The ist was roposed in 1775 by JF Toschi di Fagnano, who solved it by means of Siterential calcu. ‘The method given here was discovered by L. Feet hile he was a student [Rademacher and Toeplitz 1, pp. 30-32) FAGNANO'S PROBLEM. Ina glen acuteangled langle ABC, ineribea triangle OVW whose perimeter small as posible ‘Consider rt an arbitrary tangle UPI with U on BC. V on CA, Won AB. Let U, U" Bethe images of U by relection in CA, AB, respectively. Then ove vw 4s WRU Ey which ia path from U'to Usually a broken line with anges at and W Such put from to U" is minimal when itis straight a in Figure I.8a Hence, among all inscribed wiangles wth a given vertex Uon BC, the one with smallest perimeter occurs when F and W lion the straight line LU" Inthis way we obtain a definite thangle UVW Tor each choie of U on BC. ‘The problem will be solved when we have chosen Uso aslo minimize U'U", which sequal w the perimeter. ere te Since AU" and AU" are congruent nd clmages of AU by reflection in AC and 4B, they Luar =u. ‘Thus AU'U' isan isosooes triangle whose angle at i independent ofthe choice of U. The base OU" ix minimal when the equal sides are min ‘mal thats, when AU ie minima. In other words, AU isthe shortest dis {ance from the given point Ato the given line BC. Since the hypotenuse of ‘right-angled triangle is longer than ether cathetus, the desired locaton of| ‘Uissuch that AUis perpendicular to BC. Thus AU's theaitude fom 4 Thischoice of U yields a unique wiangle UVW whose perimeters smaller than that of anyother inseribed triangle. Since we could equally well have begun with Mor C instead of 4, we see that BY and CW are the alitides ftom B and C._ Hence ‘The triangle of minimal perimeter inscribed nam acut-anged tingle ABC ithe otic angle of ABC. ‘The same method can be used to prove the analogous result for spherical wiangles [Steiner 2, p48, No.7. The other problem, proposed by Pierre Ferma (1601-1665) likewise seeks omiaimize thesum oftheee distances The solution given here id oJ E Hotmann* FERMAT'S PROBLEM. Ina given acute-angled triangle ABC, locate a point ‘Pwhos distances from A,B, Chae the smallest posable sum. ‘Consider first an arbitrary point P inside the wiangle. Joinitw 4, B.C and rotate the inner triangle APB through 60" about B to blain CP, 50 that ABC and POP are equilateral angles asin Figure 18h Then AP + BP + CP =CP + PP + PC. + Een Lang ie Mamas, Zech fr mata ond ma shen Us (99, pp 333ae 186 ‘which a path fom C’ to C, usually a broken line with angles at P and P. eek path Goining C to Cby a sequence of thre segments is minimal ‘when its stain which ese (CBPC = 180" — LBP’ = 120" and LAPB = LPB = 180" = PPB = 120" "Thus the dested point P. for which AP + BP + CP is minimal, i the point from which each of he sides BC, C4, 4B subtends an angle of 120" ‘Tun ermat point is most simply constructed asthe second intersection of Aheline CC> and the cicle AAC (hati, the eircumece ofthe equilateral tangle ABC p Tras been pointed out for example by Pade 1, pp. 11-12] that thet angle 4BC need not be assumed tobe acute-angled. The above solution is ‘ald whenever there is no ale greater than 120 Tastead ofthe equilateral rangle ABC’ on 4B, we could just as well have nan cqulateraltiangle BCA’ on BC, oF CAB’ on CA, asin Figure 1.86 ‘Thos te thee lines 4, BB, CCl pass rough the Fermat point P, and Tiptavf them provide an allerative constuction fori, Moreover, the Tine segments 440 BB CC areal equal to AP + BP + CP. Hence evulaeral angles BCA’, CAB, ABC'are drawn ourwards on the sides of ertangte ABC the line segments AA’, BEF, CC are equal, concurrent hd clined at 1 oe another. 1. tage 18a UV and YW ma cau ange wh CA. Deduct se to cnet ape hence inate, Inter not eb angus ala tbl alt U iin the ecto UY, Wl found he age {UH net, that oa topped ton) “es {How dvs Fagnao’s poem clope when we appli angle ABC Invbich ee ape A robe? ne . 2 Thecus fete gate waglesn Figwe L8a ps hrough ‘sand ei ceates or fourth eqs tingle” “Tree bes, atthe vrces of an array tag are dil rows he topo aubieTrogh cache stead pied wha neight ang om blow the tate” above the thee hesdrare sled togetir an hen lesb the thee ‘is ar alleun bere wl he kot ome re? Four villages aria a he verze of sun of sd one mile, Tei santana ibn rn bt Bye eh neal toanake y+ mls fond Mow do they prose [Coun and bias tp 392] : — “Sure Fermat's problem for angle AC with A > 120%, a8 fora conve utiangle ABCD. “two pons P,P: asides tangle ANC ares stated ha ¢ CBP = Lenn Uacr © Cher s ePCh then CBPP = CPC 1. tour squares re placed exerally (rater 8 the ou sides of ay araelopom ha ees ante vers f sot ager. [Yap pp. 96-7 3. Leu, 1,2 bethecenters of oqars led evry on tees BC, CA. 4B ofa iege ARC. Thon te epment A ie congrent nd pepe 1012 (lo Wotkwdeoth, Wal ise : 1. Let ZX, U, Vb theceten of ares lace exerally on he sides AB, BC (cp, DA stay spl uaange or "qh ABCD. Then te sgment 20 Ge eo est ar) Sopa nt eer 827 1.9 MORLEY'S THEOREM ry te prc n mame oy ng ender {.€ Taimonh 19190 Tikka top 23) (One ofthe most surprising theorems in elementary geomet was discov red sbout 1899 by F. Morley (whose som Christopher wrote novels sch a Thunder on the Le). He mentioned it to his felends, who spread it over * Cutty. 5-17 Sess Mates 94 9.299 dott sh ham ai sscd Napalm and Fae tne sig genomthe world in the form of mathematical gossip. At lat after ten years, {rigonomerical proof by M. Satyanarayana and an elementary proof by MT: Naraniengnt were publsed." ere MORLEY'S THEOREM. The thre pdms of intersection of the adjacent ti secon of the angles of any triangle form an eullateral tangle. In other words, any triangle ABC yields an equilateral triangle POR if the angles A.B, Care tsected by AQ and AR, BR and BP, CP and CO, sin Figure 19a, (Much rouble experienced if we try a dtet approxch, but the difficulties dissppear if we work backwards, beginning with an equi Iateral tangle and building up a general wiangle which is afterwards iden tied with the given tangle ABC) ‘On the respective sides OR, RP, PQ ofa given equilateral triangle POR, cree toscles angles POR, Q'RP, RPO whose base angles a satisfy ‘the equation and inequalities ae R HY =I", a 60", BOO, y<60" tet ee Paty aa te HD) Sn: Morey, Armein Jour of Mathematics, 81 (92, 68-472, H.D. Grom Thamar ena Mo 88 (801238 nd el tre ee SoA Manteca) pp 25054. Asma pal ms devel bop cenit pe Extend the sides ofthe isosceles triangles blow thes bases unl hey meet gain in points 4, B,C. Since « + f+ 7 + 60” = 180", we can imme: iste iter the messurement of some other angles, as marked in Figure 19a. For instance, the tiangle AR must have an angle 60" —w at is vertex 4 since its angles at Qand R area + Band y + 0 ‘Referring to ISL, We ae that one way to characterize the incenterFof triangle ABC isto describe ta lying onthe bisector of the angle 4 at such ft distance that euie 90" Applying tis principle to the point Pin the wiangle PBC, we observe that the line PP (whichis 4 median of both the equilateral wiangle POR and the isosceles triangle P'OR) bisects the angle at P Also the half angle at Piss" aand LAPC = 180" — 9 = 90" + 0" — m, Hence Pisthe incenter ofthe triangle PBC. Likewise Q isthe inecnte of| QCA, and Rot RAB. Therefore ll the three small angler at Care equal; Iikewise at and at B. In other words, the angles of the Wiangle ABC are treeted "The three small angles at Aare each J4 ©. Thus 0° — a; similarly at B and 34, B= 6" 4B y= —4C ‘By choosing these values forthe base angles of our isosceles triangles, We ean ensure thatthe above procedure yields a tangle 4C that is similar to any given triangle “This completes the prof. 1 The thre nee PP, 00 RAT (Figure 19) are oncerent In oer word, te ise of 4B. C met ngain to form aot: angle POR eich is espe Ie withthe eultral rane POR (a geocrl POR snot equates) 2 Wht sales of «87 wl rte the tnpe ABC () elt) ght angie ol? Sketch te gure uch ce 3. Let Py and Ps om CA and 4) be the ngs P by reer in CP an BP ‘Tae he our points P,Q. Ry ae evenly spaced along act hough A the spec ease when The tingle ABC is equate these four pots ocr among the ‘ericesf eulreaseagon(0-gon)in wich 4 nthe vertex opposite tothe nde QR.2 Regular Polygons We begin this chapter by discussing (without proofs) the possibilty of ‘consructing certain regular polygons with the Instruments allowed by uci. We chen consider all het polygons egurdles of the question of onstructbliy, fom the standpoint of symmetry. Finally, we extend the ‘Concept of a reglar polygon $0 ast include star polygons. 2.1 cycoromy (Om tot tal An wah ond reugh eden 3 hop. 1 act's ostats imply resticion onthe nstrments hat allowed for making consocon™ namely the ettion fo rue (or saiahedee) ‘Rucompunce Hesomtced an gules ange (1), asauae 8) Tega pemagon 10.2 regular betagon (119), and renal 1-08 AneT5." Toe nsmber ses may be dbl sain snd asin epeated nue bscedons tip ntura to ask which oe eur pygns ea be SShtvted wh Eu mtrmentes Thi question way completely a= Steve bp Ons (1777-1855) ate ape of mnstce ie Sith 2 pp. 201 50a} Gaus found that euler, sy canbe 0 constructed if the odd me factor nar sine Ferma pie Ramat “The only known primes ofthis kind are Radsied HaBsias fetelai, Fae 1 = 257, Fy = W041 = 65597. ~ ‘To inscribe a regular pentagon in a given circle, simpler constructions ‘than Bvclds were given by Prolemy and Richmond. The former has been repeated in many textbooks. The later is as follows Figure 2a. “Toinserbe a regular pentagon PyPPsP4Ps ina cicle wih center O: raw the radius OB perpendicular to OPy; join Py to D. the midpoint of OB ‘sect the angle ODPp to obtain Nyon OPp: and draw NPs perpendicolar 10 OPy to obtain P; on the circle. ‘Then P,P} isa side ofthe desired pea: tagon- Richmond also gave a simple construction forthe (17) PaPs Pie (Figure 210). Join Py 10 J, one quarter of the way from O to B. On the Giameter through Py take EF 30 that 2 OJE is one quater of OJPy and ZHJE‘ 4S", Let the citcle on FP) a8 diameter eat OB in K, and It the cle with ceaer E and radios EK cut OPy in Ny (Between O and Fy) and 1Ne._ Draw perpendiculars to OPs a these two points, to cut the ongiaal izcein Py and Ps. Then the ate PsP (and likewise P,P) i y of the ci- ccumference, (The proof involves repeated application ofthe paciple that the rots ofthe equation 2 + 2x em 2¢ — I= Dare tan C and cot C) Richelot and Schwendeawein consiructed the regular 257-g0n in 1832. 4. Hermes spent ten years onthe regular 65537-g0n and deposited the man Script in a large box inthe University of Gottingen, where it may sil be ound "The next number ofthe form Fy = 2°* + 1s Fy = 4294967297. Pere tat incorrectly assumed it tobe prime. OT. Bennet gave the following ‘eat proof f that iis composite [Hardy and Wright 1, p. 1a} the number a S428 iviing both 54-2842 and 54-22 te =I divides their ference which is Fs SH W.Ridmons, Que Jounal Mahe 3689) 7.29627. HE, aden, tna Sas aon UNF " esoracdoy > Rasnnompty, Masai ee 4158 9310“The question naturally arises whether Fy may be prime fr some greater valueof Iie now known that this ean happen only i F divides 3-2 “Pl Using this criterion, electronic computing machines have shown that Friscomposte or 5 k Wb. Thrsfore Hermes’ constuction isthe is ‘ofits Kind that ill ever be undertaken! 1. Vest the cocci of Richman constuction for {8} Figut 210. 2. Assuming Richmond’ contraction fr {17}, bow wold ou Insribe [S1) i the te sil? ANGLE TRISECTION 1, on al (801725) (tot 9321 ‘Gauss was almost certainly aware ofthe act that his eylotomic condition is necessary as well a5 slit, but he does not seem to have said so ex pleity,“The musing step ss supplied by Wantel who proved that if The odd prime factors of are nor distinct Fermat primes, (n) cannot be ‘constructed with ruler and compasses. For instane, singe 7 is no a Ferma prime, Bul nsruments will nt sue for the eps heptagon 7}: and ince ie factors of are not distinc, he same is tue for the enneagon (9) "The problem of trsecing an arbitrary angle with ruler and compasses exercised the ingenuity of professional and amateur mathematicians for two hovsand years [Ball 1, pp. 333-335}. Is, of course, easy to tise cer~ {ain particular angles, such asa right angle. But any construction fort Seeting an azbitrary angle could be applied to an angle of 60", and then we ould draw a regular enneagon, In view of Wante!'s theorem, we may sty that i hay been known singe E37 that the clasieal tection problem ea never he solved “This is probably the reason why Morley’ Theorem (1.9) was not ds- ‘covered til the twentieth century: people felt uacasy about mentioning the trsectors of an angle. However, although the tiseetors cannot be con: SMracted by means ofthe ruler and compasses, they can be found in other ‘rays [Cundy and Rollet 1, pp. 208-211) Even i tese more versatile in- Siraments had never been discovered, the theorem would sill be meanig- Fal. Most mathematicians are wiling to accept the existence of things that they have not enable fo construct. For instance it was proved in 1909, that the Fermat numbers and Fy are composite, but ther smallest prime Factor sil main to be computed. “The number 2+ 1s sompoie whenever ai nat power of 2 aon ae en eee eee ee cates a vit’ pp. 366-372 2.3 ISOMETRY On way ttn tera foc, pale by th New senda, uh ten oon, Cor apn af psn you move eB em ne he tpt en rg sn oe oe aly or imogiton i 80 corto ray set monn Wy 185-1958) (We Yep We shall find it convenient o use the word transformation in the special sense of @ one-to-one correspondence P > P among al the points in the plane (or in space), that i, rule for associating palrs of points, ‘with the understanding that each pair has a frst member P and a second member P and that every pot occurs asthe fst member of just one pait fd also asthe second member of just one pai. Temas happen that the ‘members of «pair coincide, titi, that coincides with Pin this ase P ie alled an invariant point (or "double pont) of the transformation. Tn particular, an Gomer (or “congruent transformation,” of "eongn cence") isa transformation which preserves length, so that, if (2, P) and (0. Q') are two pais of corresponding points, we have FQ = FO": PO and PQ" ae congruent segments. For instance, ovation of he plane aboot Por about line through P perpendicular tothe plane) san Sometry hav ing Pas an invariant point, but a franslavon (or “parallel displacement") has no invariant point’ every points moved "A reflection i the special kind of iometty in which the invariant points consist of all the points ona line (or plane) called the mir. ‘A sll simpler Kind of wansformation (5 simple that it may at st seem to trivial o Be worth mentioning) isthe Ide, which eaves evry point Unchanged. "The result of applsing several ansformations successively is called ther product.” he product of two transformations the seni), cach sealed the mers of the other, and ther productin the reverse order Isaguia the identi. 2.31 Ifan(sometry has more than one inaran point it must Be ether the deny ora reflection “To prove this let and 8 be two invariant points, and P any point aot on the line A (Figure 1.36). The corresponding pint satisying ap Br, rust eon the circle with enter 4 and radius 4P, and on thecitle with een-‘er Band radius BP. Since P isnot on AB, these ctcles do not touch each ‘other but intersect in two points, one of which is. Hence is either P itself or the image of P by reflection in 4B. 24 SYMMETRY "ye retina ht av ame thy eset Wiliam 7971827) ‘When we sy that a igure is “symmetrical” we mean that we can apply certain isometries, ealed »ymmetry operation, which leave the whole Eure unchanged while permatingits parts For example the capital eters Fand ‘A Figure 24a) have bilateral symmetry, the mirror Being horizontal for the former, vertical forthe later. The letter (Figure 240) symmetrical by halfzuro rotation through 180" resston in a pono "ental ie version), whieh may be regarded as the result of reflecting horizontally and then vertically, or vie versa. The swastika (Figure 2) i symmetrical by rotation through aay number of right angle. AN ere 20 re 240 In counting the symmetry operations ofa figure, itis usual include the ‘identity any figure has this vial symmetry. Thus the swastika admit four distinct symmetry operations: rotations through 1, 2.3, or 4 right anges, ‘The ass the idemity. ‘The fst and thed are averses ofeach other since their products the Kent ‘This use ofthe word “prodvet” suggests an algebraic symbols in which te tansformations are denoted by eapital letters while denotes the iden iy. (instead ofl, some authors write E). Thus if Sis the counterclockwise _goater-cun, the four symmetry operations ofthe swastika are SS SiS and St ‘Since the smallest power ofS that is equal othe identity ithe fourth power aA we say that Sif period 6 Similarly S, being aalf-turn is of period 2 [see Coxeter 1p. 39). The only transformation of period Vis the entity. A teansaton is aperiodic tht i, thas no period), but itis convenintl sid to be of iaaite period ‘Some igres admit both reflections and rotations as symmetry operations. “The letter H (Figore 24d) has a horizontal miror (ike E) and a vertical miro lke A), as walla center of rotational symmetry (ike N) where the {womitrors intersect. Thue it has four symmetry operations the identity the horizontal reBection Ry, the Neral reBection Re, and the hal-urn Nr 240 ere 24d 1, Every itomeiry of prod 2 ler ebstin or aha te Bachna Teepe (oa hal agar, rears of) Crean co diate i) pla coordinates, (Take he nga tobe the center of oaion)| . 2.5 onours Sy, a wie aw 60 mo dee ming ‘ev dr, bay and precio eran Wr op 5 A set of tansformatons[Biekboll and MacLane 1, pp. 118-118] s sid to form a group if contains the inverse ofeach and the product of any (wo (Gaclading the product of one wit elf or withitsiavetst). The number of Aistnet transformations called the ordera the group. (Tis may be ether ‘nite or infinite) Clearly the symmetry operations of any figure form 3 soup. Thisis called the symmeiry group ofthe figure. Inthe extreme case When the Sgute is completely iregolar (ike the numeral 6) its symmetry roup is of order one, consisting ofthe entity alone“The symmetry group ofthe leer E or A (Figure 24) i he so-called edra group of order 3, generated bya single reflection and denoted by Ds (The name easily remembered asthe Grek origin of the word “dibedal™ {salimost equivalent tothe Latin origin of "bilateral") The symmetry group fhe letter N (Figure 248s ikewise of orde 2, but inthis case the genera~ forisa haleturn and we speak of the cele group, Cz. The two groups Dy tind Cy ave abtraclly identical or somo: they ae diferent geometrical resetations ofthe single abstract group of order 2, defined by the re 281 Ret ‘or R = Re [Coxeter and Moser 1 pI ‘The symmetry group of the swastika is Cy, the cyclic group of order 4, ‘generated bythe quaret-tura Sand abstractly defined by the relation S# = 1 ‘That oftheleterH (Figure 24d) is Da, the dinedalproupof order 4, genet- sted bythe two reflections Ry, Ry and absracly defined by the relations RRs [Although Cz and Ds bth have order , they are noisomorphie: they have diferent stuctore, diferent "multiplication tables.” To sex this tsuices {observe that C, contains two operations of period , whereas al the opera~ tions in Ds (except the identity ae of peed 2 the generators obviowshy snd thie product also since (RSF “This last remark illustrates what we mean by sayin that 2.92 is an ab siract defntion for Dy, nately that every tre relation concerning the genera~ tors Ry, Re ean algebraic consequence ofthese simple relations, An alter ative sbsract definition forthe same group is 33 RE= 1 RES, R= from which we ean easly deduce RyRs = RsRs “Te general eylc group Cy of order mas the abstract definition 258 sel Its single generator S of period ni conveniently tepreseted bya rotation through 360°7n. Then Sis rotation trough times this angle, and the ‘operations in Cy ae given bythe values of from Ito nor fom 040.0 = 1 Te particular, Cy occurs in nature asthe symmetey group ofthe periwinkle flower 282 Risk Re= RRs RiRRAR: = RRARER, = RARER = RR = Rye = 1 ages rlton through angle bout he gia 2 asformaton of (par ccd ent tf the cuon race a 246 THE PRODUCT OF TWO REFLECTIONS ‘oni oe dot Moet) {any group of transformations, the associative aw (RSiF = RST) [sautomatically satisfied, bu the commutative law RS =k doesnot necessarily hold, and care must be taken in inverting a produ example for (Rs = 89R, ‘not RS, (This becomes clear when we think of Rand Sas the operations ‘of putting on our socks and shoes, respectively) ewe 148 ‘The prodvct of elections in two intersecting lines (or planes) i a roa tion through tice the angle between them. Is Fach if 4,8, Dy -- are evenly spaced on a citcle with center 0, lt Ry and Re be the reflections in ‘OB and OC (Figure 262). Then Ry rect te triangle O48 into OCR, which erected by Reto OCD; thus RyRy isthe rotation through 2 AOC ‘or ZBOD, which is twice BOC. Since arotation i completely determined by its center and its ange, Ry sequal to the product of reflections in any two lines through O making the sie angle as OB and OC. (The elections in Od and OB are actually RyRaRy and Ry, whose product i RyRGR,™ RRs) In particule the hll-trn about O isthe prot of reflections ia ‘any two perpendicular lines through 0. Since Rife iss counterclockwise rotation, ReR is the corresponding clockwise rotation: in fet,‘Thiss the same as R,Reif the two mirrors are aright angles in which case RiReis a haletara aad (RRS) 1. “The produto quarter tras a he same sens) about Cand she alr shout teen of asquare having BCTor ie 2 LecACPO ane ARS be squares on te ies ACand BA of ange ABC. WE and Crema vel wl 4 vai ec, PS pans tough ted pot 2.7 THE KALEIDOSCOPE zis special case ofthe general dihedral group Dy, whichis, form > 2. thesymmetry group ofthe regular -gon, ()._(See Figure 2Tafor the cases ‘n= 3,4,5) Thisisevidealy a group of order 2n, consisting ofn rotations (Ghrovgh the m efectively distinet multiples of 360"/n) and m reflections ‘When mio exch ofthe m mirror joins vertex tothe midpoint ofthe o- poste side; when nis even, 4n mirrors join pais of opposite Vertces and tn Fiset pairs of opposite sides [sce ButhofT and MacLape 1 pp. 117-118, 135, ere 276 ‘The m rotations are just the operations of the eyeic group Cj. Thus the operations of Dy include all the operations of Cyn tecnieal language, Ca, ia mubgroup of Dy. The rotation trough 360", which generates thes group, may be described us the product S = RyRyof elections in two adj ‘ent mirrors uch as O and OC in Figure 27a) which are inclined at 180° Tet Ry, Ry, Ry denote the elections in their natural onder of a rangement, ‘Then R: Ries, being the product of reflections in two mirrors Inlined at K times 180", i rotation through K times 360°/n RiRaes = SA St, and them selections may be expressed as Ru RIS. RAS RSH nother words, Dyis generated by Ry and. By substituting RyRe for S, we Ths Ree a ‘ce thatthe same group equally wall generated by Ry and Re, which satis the relations an RY Ray = | (The fs ewo relations come fom 21 and the third fom 2.54) These lations ean be shown to sifce for an abstract definition (see Coxeter and Mose 1, pp. 6 36 ‘A practcl way to make 2 model of Ds isto join two ordinary mierors by @ hinge and stand them on the lines OB, OC of Figure 2.7250 that they are n- ‘ined at 180"/n. Any object placed between the mirrors yields 2 visible tmmages (including the objects). Ifthe objects Your right hand, maf the {mages wil ook Ike a et hand, stating the principe that since a elec tion everses sense, the product of any even numberof elections preserves sense, andthe product of any odd numberof reflections reverses sense. “The fist publied account of ths nstrement sceme to have been by [Athanasius Kitcher in 1646, The name kaleidoscope (rom eahor, beautiful ‘aon, form; and oxozns, to 368) was coined by Sie David Brewster, who ‘wrote a treatise on its theory and history. He complained (Brewster 1, p, TAT} that Kircher allowed the angle between the two mirors tobe any Su multiple of 360" instead of restiting it to submultiples of 180" “The case when n = 2s of course, familia. Standing between two per- pendicular mirors (as at a corner ofa room), you see your image ia each nd also the image ofthe image, whichis the way other people see ou. Having decided to use the symbol Dy for the dihedral group generated bby reflections in two planes making a “ihedal” angle of 180"/m, we naturally stretch the notation s0 a8 toallow the extreme value m = 1. Thus ‘Dyis the group of order 2 generated by a single reflection, that, the sym- imetty group ofthe letter Eor A, whereus the tsomorphic group Ca generated bya halt, isthe symmetry group ofthe letter N. acct tp he ai ne ah ie <1 Res inter i th was om te tnd of eter lsh OF oun he Feet ‘bovpmuhstur have sens en and butte dso pein ee ‘Beta Lonard sasion tra wed oto tn moder ae ean iingin Washo se smey woop Dean te Baal Tele ear Cag a sowfats wun Dy tutcrasioely ony Dy [Rep pp. 29-280) you ci an apple he ny oper ete te en ate ess op D Enemy the ep aby stag usin al ou die he we Eheim ech whch ecb on erode ft ak 1. Desi the metry gros of (aca tangle, (an ncle nee(© apanbat, (@) spartan (©) atoms, (1 arcana, 2 Useimvrsesand the tine lant rove algebra he "caneaiaion re "S_Show how the sl deting rations oD namely 2.71 win iced ager manipltio rom heigl laos Rea RAR, = RR 4. They group Cals aseroep of fad oa enum mia vis of ninpaicua stn tpime, the oly obproup of Cae Cy sland Cy aay bese 2.8 STAR POLYGONS Instead of deriving the dihedral group Dy from the regular polygon {n). wwe could have derived the polygon from the group: the vertices of the poly fon are just then images ofa point Pp (ihe Cof Figure 270) on one ofthe {feo mirrors ofthe Kalesdoscope, In fac, there 0 need to use the whole roup Da its subgroup Cy will suffce, The vertex Py of the polygon PoP... Pa-vcanbe derived fro the initial vertex Py bya rotation through Fetimes 360°/m "More generally, rotations about a fixed point O through angles 0, 20, 3 ‘transform any point Po (distinct from O) ito other points Py, Pa, Pa, fon the cicle with center O and radias OP). In general these points become increasingly dense onthe circle: bul ithe angle # commensura bie with aright angle, only faite numberof them wil be distin. In par ticular f= 360° /n where mis a positive Integer greater than 2, then there willbe points Pe whose sucessve joins PoP PP are the sides of an ondinary regular gon. [tus now extend ths notion by allowing to he any rational mumber greater than 2 say the fraction p/d (where p and dae coprime). Ac ingly. we define a (generalized) regular polygon (). where m = pd. ls p ‘etices are derived from P by repeated rotations though 360°/n, and its ‘sides (enclosing the center ines) are PoP PaPays soy Ppa Since a ray coming out from the center without passing through a vertex Wil crose dof thep siden this denominator ds called the densiy of the polygon [Coxeter pp. 33-94). When d = Iye0 that w= p we have the PuaPs Pi ordinary regular pon, (p}- When d > 1, the sides cross one another, but the rossing pointes are not counted as vertices, Since d may be any pos tive integer relatively prime to p and less than 4p, there isa regular poly~ ‘gon (x) foreach rational aumber n > 2. In fact, itis occasionally desi Able to include also the digon (2), although its tw sides coincide ‘When p= 8, we have the pestagos (5} of density and the pentagram {3} ofeensty 2, which was used a a special symbol by the Babslonians and by the Pythagoreaas._ Similarly, the cropram (4) and the decagram ('P] have density 3, while the dadecapram {{@) has density $ (Figure 28). These particular polygoas have names as well as symbols because they occu? as Faces of interesting polyhedra and tessellations* OF it & Polygoas for which d > 1 ate known as star polygons. They are fre quently used in decoration, The earliest mathematical discussion of them ‘Was by Thomas Bradwardine (1290-1349), who became archbishop of Can terbty forthe last month of fis life. They were also studied by he great Germain scientist Kepler (1371-1630) [see Coxeter 1, p. 114] twas the Swise mathematician L. Sehlaf (1814-1895) who first wed « numerical symbol sich as (pd) This notation i juste by the oeeurrence off tnulas that hold for (n) equally well whether m be an integer ora fraction. For instance, any sie of (») forms with the center O an soseees triangle OPP, (Figure 28H) whose angle at Os 2n/m (As we are iniroducing trigonometrical ideas, its natural to use radian measure and write 2 in- tad of 360",) The bate of this osceles tangle, being aside of the poly ‘gon, is conveniently denoted by 21. The other sides ofthe triangle are equal {o the crcumeadive R of the polygon. The altitude or median from 0 is the inradiusrof the polygon. Hence 281 R Lose, r= Loot, fm = p/, the area ofthe polygon is natrally defined to be the sum of the areas of the p isosceles triangles, namely 11.8. ones LS. Long gina. Mer, Uo payed, Po se rat of he aa Say, A BN (3) 6p. 01; a L owe nan 202 lr = pP cot. ‘when d = 1 thsi simply pot x/p in other exses our detition of area har the effect that every part of the interior is counted a number of times {gual to the “local density” of that part for example, the pentagonal re {Bom ia the mide ofthe pentagram [3] is counted (wie. “The angle PoPsPs between tho adjacent sides of (n}. Being the sum of the base angles ofthe ioseles triangle, sth supplement of 2e/n, namely 283 “The line segment joining the midpoins of two adjacent sides is elle the vertex figure ot (n). is length clearly 208 cost [Coxeter 1, pp 16, 94, xenessts 1. trtesides of polygon ind ina rear al equal he polygons el, 2 Ava potyon seed ino cic haan ol amber vers, and all it ov ace gual the plygon el. (Mave Ries) 1 ind te ans of the polgoas Gh Gh 4, Find se rai and verter sre of he pobgons (0D. Give poli eondinates or he Kh vertex Ps of polygon {a} of eieurais 1 wtb cemer athe oleae Py ob 10 caleand te sane amount oa fa 3 Isometry in the Euclidean plane Having made some wse of reflections, rotations, and translations. we nat- ually ask why a rotation or translation can be achieved 3s continuous Aisplacement (ot “motion”) while a reflection cannot. It is ao reasonable to ask whether there is any other kindof isometry that resembles a rete: tion inthis respect. After answering these questions in terms of "sens." ‘te shall use the information to prove a remarkable theorem (36) and to describe the seven posible ways to repeat a patern on an endless strip en, 3.1. DIRECT AND OPPOSITE ISOMETRIES (osm hop. By several applications of Axiom 1.26, ican be proved that any point P inthe plane of to congruent triangles ABC, 4°2°C” determines a core sponding point P'such that AP = 'P, BP = BP, CP-= CP. Likewise another point Q yields and PQ = P'Q" Hence 2.11 Any neo congruent triangle are relaed by @unigue tiometry $1.3, we saw that Pappuns proof of Pons asnorum involved the com parison oftwo coincident angles ABC, ACB. We see intuitively that this {sca dstinction of smee: if one i counterclockwise the ater is clockwise "opologica” property ofthe Fucldean plane that this distinction ‘cam be extended from coincident triangles to distne! Wianges: any two “dected” tangles, ABC and 4°B°C, either agree or disagree in sense. (For a deeper investigation of his intuitive dea see Veblen and Young (2, Pp. 61-62] oF Denk and Hofmann (1, p.56}) ITABC and 4'B'C are congruent, the jometry that rates them is aid te be director opposite according a it prestrves or reverses sense that i,according as ABC and A°B'C’agice or disagree. tis easily seen that this property of te lometty is independent of the chosen tangle ABC: ifthe Same hometrycelates DEF to D'E'F, where DEP agrees with ABC, thea ko DEP" agrees with 4’C. Clearly direct and opposite isometries ‘Combine like postive and negative numbers (eg the product of two appo- Site isometries dee) Since a election is opposite, a rotation (which is the product of two rections) is ditee In p isditect. Some authors call diect and opporite isometries “displacements and 1 ‘eral of "proper and improper congruences.” “Theorem 231 can be extended a8 fellows: ewe 3.12 Two given congruent line segments or pont prs) AB, A°B are ve lated by jst two tomers one ect and one oposite “To prove this, take any point C outside the line AB, and construct C50 thatthe tangle 4’B°C ss congrueat to ABC. ‘The two posible postions OFC {marked C, C” in Figre 3a) provide the two isometries. Since ether an be derived fom the other by reflecting in 4°", one ofthe isometries is ‘irct and the other opposite For a complete dscson we need the following theorem [Bachmann 1, pl 3.19. Every isometry of the plane isthe produc of ot mas three reflections. ‘If there sa invariant point, “three” con be replaced by "0." ‘We prove this in four stages, using 3.11. Trivially if he triangles ABC, -A'B'C coincide, the isometry isthe identity (which Is the product ofa re ‘ection with tsif) Ifa coincides with a”, and B with BY while Cand C” tate distinct, the triangles are related by the reflection in 4B. The case when ‘nly 4 coincides sith 4” can be reduced to one ofthe previous eases by e- flecting ABC in m, the perpendicular bisector of BB" (se Figure 3.15). Fie rally, the general case can be reduced to one ofthe fist thre cases by re- fecting ABC in the perpendicular bisector of AA’ [Coxeter 1, p. 35) ‘Since a reflection reverses sense, an isometry is diret oF opposite accord- ing ast the product of an even of odd aurber of reflections. ‘Since the identity ithe product of two reflections (namely of any ref: tion with tel, we may say simply tha any isometry isthe product of ro forties reflections, according ast diect or opposite. In particular, 3.14 Any isometry with an invariant point so rtation ova reflection a conding a is dre! or oppose 1. Name wo dst smetrien 2. Name coe oppoite omer. Is tere any ote ind? 13 AB and 0 age ated by a otaion, how can the center ation Be a8 structed The pependclsbcto of 44 nd BB are ot necessary di “E_The product frelecton in the ines through pia the eeection i a che ne ough the wae point [Bachmana 12.2 TRANSLATION “The particular isometries so far considered, namely teections (which are ‘opposite and rotations (which are died), have each at lest one invariant point, A famine isometry that leaves no point invariant sa ransasion Ibachmann 1, p. 7h which may be described asthe product of halF-turns bout two distinct points 0, O° (Figure 320). The fist halfwwrn trans forms an arbitrary point Piato P and the second transforms ths into P, withthe nl rest that PP" sx parallel to OO" and ovice as Tong. Thus the Tength and direction of PP" are constant. independent ofthe position of P. Since a translation is completely determined by is length and direction, the prodt of half-tarns about O and O's the same asthe product of hall= turns about Q and Q provided QQ" is equal and parallel to 00" (Thismeans that 00°Q'@ is a parallelogram, possibly collapsing to form four Collineae point, asin Figure 32a) Thus fora given Wanslation, the cen- fer of one ofthe two hal-turns may be arbitrary assigned BM The produc of wo translations isa translation For, we may arrange the centers so thatthe fst translation i the prod uct af haléturns about Ov and Os, while the second isthe product of hall- turns about Oy and Os. When they ae combined, the two halturns about ‘Oveancel, and we are lft with the product of halfturas about Oy and Os Similsty, sf m and m’ (Figure 3.26 ae the lines through O and O” per- pendicular to O07 the hal-trns abovt 0 and 0” are the products of re- Festions in me and’00', 00" and m'. When they are combined, the (wo ‘eflections in OO canes, and we ate left with the product of refetios in mand nf Hence 3.22. The product of reflections im two parallel mirrors isa translation through tice the tance between the miror. Ifa translation T takes P to P" and @ to Q the segment QQ is equal sand parallel to PP thetefore POQ"P” isa parallelogram. Silay, i an- bother translation U takes to Q it also takes P to 0"; therefore Tu = UT. (ln detail iQ is PY, is PE. But U takes PP co PT Therefore PP land P coincide, forall postions of P) Inter words, 3.28. Translations are commutative “The product ofa haléturn Handa translation Tis another hal-turn; for wwe can expres he translation asthe product of two hal-trns one of which IH say T= HEY, and then we have HT = WH! = HY 3.24 The product of a halfturn and aianslaion ia halfturn. WT isthe product of asus about 0 and 0, whats be product of half {When anion i expected as the product of wo eetons 10 what ex Ue Bbsjerdatrsnpangentcies eared ory 2. What she produc of roations trogh pote angles («and =) abou mo stint pins 'S Every product of thes bau s haltur [Bachna Ye .) 7. Expres the trasation though distance alo the ats 8 transoematon ‘of Caren coordinates IN )) = O18 te equation for ere whl the ei: 3.3. GUIDE REFLECTION ‘We are now familiar with three kinds ofsometry reflection, and translation. Another Kind isthe gid reflection (or simply" ‘which is the product of the feection in lie a and a Wansation along the Same line, Petre this line as a straight path through snow: then, conse: tive footprints are related by a lide. Suth an isometry i determined By is ‘ais @and the extent of the component translation. Since a refetan i op- posite whereas a tanalation i dret, their product is opposite, Thus a IB reflection san opposite isometry having no invariant point [Coxeter Yy P36) Ta gli election G transforms an aritary point P incoP* (Figure 33a), P and Pate equidistant Irom the axis 2 0n opposite sides. Hence The mdpoln ofthe Une segment PP les on the axl for all positions of P ere 338 Let R; and T denote the component reflection and translation. They «evidently commute, 50 that G=RT= TR, ‘We have seen (Figute 320) tha the taslation T may be expresied asthe product of two half-turns or of two parallel refetions. denying the ine ‘in Figure 3.30 with the line OO’ in Figure 3%, let RR’ denote the fe= ection in mm ‘Then the product ofthe two halfturns H=RR=RR W=RR=RRis T= HW = RRARIR = RR, and the glide eelection is G = RoT = RRR’ = HR’ TR, = RRR; = RH. “Thos a glide reflection may be expressed asthe product of thre reflections (wo perpendicular tothe thi), or ofa al-turn and a refistion, or of Feflection and a halturn. Conversely, the product of any hal-tura and “any reflection (r vice versa) sa glide reflection, provided the center ofthe haleturn doesnot lie on the mitfor. [Bachmann 1, p. 6] "We st in .13 that any direct sometty inthe plane is the product of to reflections that is 2 translation or a rotation according a the two mirrors ‘te parallel or intersecting; also that any opposite isometry with an invari fant point is ceflection. To complete the catalog of isometries, the only Femaining possiblity isan opposite isometry with ao invariant point. If Such an isometry $ transform an arbitrary point 4 into comsier the hallturn H that interchanges these two pois. The product HS, being an ‘Opposite isometry which leaves the point 4” invariant, can only be arelles- tion R. Hence the given opposite iometty isthe glide reflection 8 = HR = HR Every opposite lsometry with no invariant poles aside rection. In other words, 31. Every product of thre reflections is eter a single reflection or a de reection In particular, the product RT of aay reflection and any translation is a slide felection, degenerating to « pure eoflection when the mirror for Ris Perpendicular to the direction ofthe tansation (in which ease the refec- ions Rand RT may be sed a the two parallel reflections whose product [s'T)_ But sine a pven glide reflection G ha a definite axis (the locus of mnidpoins of segments PP), its decomposition into arefection and a trans ition along she mirror i unique (ani its Jecomposiion into a reflection find halfturn, where we may either take the mirror to be any line per: pendiculr tothe axis or equivalently take the ceatr of the half-turn tobe ny point onthe as. 11 Bis the midpoint of AC, what ids of omer wil aso (ABintoCB. GH) Bint BC 2. Every det iometey the product oft eens Every oposielsomeny i he rot of refecton and all 3 Desert the produc ofthe esto 8 00 and the asuen aout 0. eee eee ee eee ee ce eancnnnnne ne cecpencieular. | Every poi of hee pide eesions 2 refection of ge reece & The product of te elections rete THER RyRy ae thre efstions (RR amsation [Rademacher and oop 1.9) 1B Des the tansformation Gantt. sy the watemen that iran the eves) ‘34 REFLECTIONS AND HALF-TURNS “Thomsent has developed 8 very beauiul theory in which geometrical properties of points 0, Oy, Oz, --- and nes m my.» (undetsiond 19 be EI Gistine) are expressed av relations among the corresponding hal-turns HL Hoste. -and reflections R, Ry, Re,----- The reader can soon convince Fisel that the following pais of statements are logically equivalent RR, = RR + mand mare perpendicular. HR =RH <> Olesonm RiRsRy = RRR, > mma my are either concurrent or parallel TWH SHH, + Oisthemidpointof 0.05 HR = RH; + misthe perpendicular bisector of 010s Tere te tains 4) Haat = 1s) RA 3.5 SUMMARY OF RESULTS ON ISOMETRIES {Bean 2, Chen. Some readers may have become confused with the abundance of techni cal terms, many of which are famaliar words to which unusually precise Meanings have been attached. Accordingly, let we repent some of the t= tions, stressing both their analogies and their diferences +, Thames, The aes fees ey ya Rupee, Mamata Go ee eeeInall the contents that concern us here, a ransformation sa one-to-one comespondenee ofthe whole plane (or space) with itself. An omer 18 a Spesal Lind of ansformation, namely, the kind that preserves length. A sPesmeiny operation belongs to «given igure rather tan to the whole plane: Risin comely that transforms the gure into sll, in the plane a ret (sense-preseviag) isometry being the product of two reesions, a totation ofa translation according as it does or does tot ave an invariant pot, hats, according asthe ¢wo mirrors are inter- Tedding or paral. the latter case the length of the translation is twice {He danse between the mirrors: in the former, the angle ofthe rotation {twice the angle between the merors. In particular, the product of rec tions i two perpendicular mor isa half-ur, thats, a rotation through Teordgkt angles Moreover, the product of two half-turns i translation. aeppone (eee reversing) isometry being the product of thee rete: tion fe gener age refecton: the prodvct ofa eection and ransa- ‘dom Inthe special case when the translation is he ie ‘rough zsre distance, the glide reflection reduces toa single reflection, staal has whole line of invariant points, namely. al the points om the ‘To sum up 12.51 Any direct lomeiry iether a vanlaion or a rotation. Any oppo ice tamer ts either avefton ora glide reflection 1. 1S isan opposite omen, Sis tanto 2 tee Reyer are the eesti, (RGRGRGRG! isa earsatin along test ior (in Snce RoRsRe aa ReRaRy at de rections ir guae fe com ‘mutate, by 323; hos. (RRA RSPAGROF © (RERARYARIRLRF, shat Re ad RRGRGRGRS) ae commutative Bachan Ys 11) 3.6 HIFLMSLEV'S THEOREM vary hh dye of areca cone minty Mord (249 581 ‘Wesaysia 2.12 that to congruent line segments 48, 48a related by jushine oma: one det s8d one opposite oth isometries ave the sue enter on ener pnt collinear with 4 and B, thats every point 02 [ibladnscstragh tee 4 (or nstaoce, the midpoint of 4B is transformed seine Siipetat of 42". The opposte isometry Hs a reetion or glide ‘epee ene miror or ais ontn ll the midpoint of segs oi nins all the mic Pi te amincide the i diet some i hasan and thy a evince Coxeter 3, p. 267) HIEUMSLEV'S THEOROA. When ll she points P om one line are related by ‘an omer tall the pois P’ om another, the midpoint of the Segments PP ‘re dine and collinear or ele they al coincide. ewe 30 In particular, i 4, B, Care on one line and 4’ B,C’ on another, with 31 AB=AB, BC= BC (Figute 3.6), then the midpoins of 44°, BB, CC’ ate either collinear of oiedent (JT. Hjemsey, 1873-1950) 3.7 PATTIRNS ON A STR OO Any kind of isometry may be used to relate two equal circles. Fer in stance, the point Pon the fist cicle of Figure 37a is transformed into P* fn the second circle bya tansaton, into P* by areection, into P" by & ‘eum an ino Py ghd rfctn. "(rons have ben ted Jndiate what happens tothe positive sense of rotation ound the Bist ci tle) ‘These four lometries have one important property in common’ they leave invariant (as a whole) one infinite straight line, namely, the ine Sorina the centers ofthe wo ies: nthe orth case hi he on t-‘We have seen (Figure 3.2) thatthe product of rfetions in owo pallet smrrors mm isa translation, This may be regarded as the imiting ase {ar otaion whose center very far away; Tor the two parallel mirors are the limiting case of two mirrors intersecting a avery sll angle. Accord Ingty the infnite group generated by a single translation is denoted by Ci tind ihe infinite group generated by two parallel reflections is denoted by ‘De Abstavly. C. ithe “ree group with one generator” ITs the gem ‘ting translation, the group consists ofthe transitions TAT TT re 370 Similarly, D., generated by the refetions R. Rin parallel mirrors m,n (agore 370) consists of the febections and translations RRR RR, RL RL RR RRR, [Woseter 1, p. 76 i abstract detsiton is simply R= REAL “This group can be observed when we sit ina barbers chair between two ‘Daal mero (ee the New Yorker, Feb. 23, 1987, p39. where somehow The efection RR'RR'R yields a demon) “x Gitterent geometrical representation for the same abstract group De ig blaine by interpreting the generators Rand R’as hl-turs, There is thon intermediate representation in which one of them i rflection and the others hal-torn; but inthis ease their product is no longer a transla tion bits glide relection. ‘Contiaung inthis manner, We coud soon obtain the complete st of the sever infinite “one-dimensional” symmetry groups: the seven essentially {Tivunet ways to repeat a pattern on aatip or ribbon [Spetser Ve pp. 81-82); __Tplcotpanes ______enertore__Absrect Grown @LEEL " atstaon a lL Tee 1 pe fection = VV 2 eects (NNN 2 halter >. ov AWA Tecteston o and 1 half-turn) 9, D DDD into and 1 reflection Cx DL ae Farce Bix nih mo min ar both vee onin he mil Vel ing tbat tess ths Vino oct is nts os {i alo the pce btncen te two minor, ys the whol pater, ino an i) thee ws host mir an hem fhe ast Colum indateiet prod Conser prt. ral these groupe Sse (andt.stse some eon a henang the enrages oe nian eo psc bree tas Sisy speaking thet seven groupe are aot“Ldimesions” ba dines hatin ty se dienna mney romp nag sn re toni uh ne a wad ee Ont winner eoup pected by oe rat a Degenerate by two reletions Gn pot meron) —— 1. Tend he symmeuy groups of he toning pater: Behe! aba vdpabdpe 2 Which are he symmetry groups oa) cid 8) ane carve?4 Two-dimensional crystallography Mathematical crystallography provides one ofthe most important appl cations of elementary geomety to physis. The three-dimensional theory is ‘complicated, bat ite analog in wo dimensions is easy to Vsualize without being trivial Patterns covering the plane arse naturally as an extension ofthe snip patterns considered in $3.7. However, inspite ofthe Zestic~ tion to two dimensions, a compete account ofthe enumeration of iaiaite symmetry groups is beyond the seope of this book LATTICES AND THEIR IRICHLET REGIONS. {Boson 2 hp. 2) TE lg asa ise goo pt pe y to need Sune eae ate in ete ewe Ate translations X, Y._ Since the inverse ofa translation translation, and the prodvet of two translations isa tansation (3.21), this group consis entirely Df translations. Since XY = YX, these wansatons ate simply X°Y* forall fmegers ny. Abstract, this ithe “direct prot” C. % C., which has yx [Coxeter and Moser 1p. A0, Any object, suchas the numeral 6 in Figure ‘he. is transformed by the group pl into an inte array of such objects, forming a pattern. Conversely, pl isthe complete symmetry proup of the pattern, provided the object has nointrnsc symmetry. IT the object sin- ile pont, the paler an aray of points called x two-dimensional lati, tnhich may be pictured a he plan ofan inte orchard. Exch Itc point ‘naturally associated withthe symbol for the wanslation by which itis derived from the orignal point I (Figure 4.10). re 40 Aayone sanding in an orchard observes the alignment of tees in rows in many directions. This exhibits characteristic property of ltice: the line joining any two ofthe points contains infinitely many of the, evenly spaced, that is “one-dimensional lattice” In fack. the line joining the poiats 1 and X°Y* contains also the pots xoreyond = oxeeyeiy where dis the greatest common divisor of x and y, and m suns over all the integers. Tn particular, the powers of X alle on one line, the powers of Yon another nd ines parallel o these through the remaining lie points form 3 tesllation of congrisntparallelograme ing the plane Without itersties (Figure 4 1c). (We use the lerm fesellaton for any arrangement of polygons Biting together so as to cover the whole plane without over- lapping) aaa ga LU A gypical parallelogram is formed by the four points 1, X, XY. Y. The translation T= Xe¥" transforms this parallelogram into another one hav {ng the pot T (instead of 1 at its “AES” cornet. There is thus a one-to fone correspondence between the cols or tes of the tssellation and the transformations in the group, withthe property that each tansformation ‘kes ay point inside the original ello 2 point similar sitoate i the new cel Forthis reason, the pial parallelogram is called fndamental region. “The shape ofthe fundamental repon it far from unig Any parflelo- ‘gram wil serve, provided it has four lattice points for its vertices but no ‘thers on is boundary or inside [Hasdy and Wright Vy p. 28) Thisis the sometrcal counterpart of the algebra statement thatthe group generated By. iv equally well generated by X°Y", X°YS, provided ad — be “1 “Tocxpress the old generators in terms ofthe new, weobsorve that COYIMOEY = XH, ORYH OYE = Ye: But there is no need for the fundameatal region tobe a parallelogram at all; for example, we may replace each pir of opposite sides by a palr of con agrucnt curves asin Figure 4. Every possible fundamental region, whether we choose a parallelogram ‘any other shape, has the same aeu asthe typical parallelogram of Figure 4:1e. Por, inside a suficienly large circle, the numberof lattice points is ‘sual to the numberof replica of any fundamental region (with an signi fant erordueto mutated regions a the circumference): thus every possible Shape has for its area the same fraction of the area ofthe large circle” Its An interesting fet that any convex fundameatal region forthe translation {sr0up isa centrally symmetrical polygon (namely, parallelogram or «cen tally symmetrical hexagon) “Among the various possible parallelograms, we can select a standard or reduced parallelogram by taking the generator Y to be the shortest trans. lation (or one ofthe shortest in the group, and X to be an equal or test shortest translation in another direction, Ifthe angle Between X and ¥ then happens to be obtuse, we reverse the diection of Y. Thus, among all the parallelogram that can serve as «fundamental region, the reduced pa Telogram has the shortest possible sides. The translations along thee sides are naturally called educed generators. By joining the vertices X, Y of the reduced parallelogram, and the oor onding par of vertices of each replica, we obtain a tssellaton of com ferent ciangles whose vertices are lattice points and whose angles are ‘onobruse. Each latuce point belongs to sx of the triangles Tor example, the triangles surrounding the point | join io pars of adjacent points in the cjele XY CY v4, XY (Figure 4.16). By joining the circumoentersof these six triangles, ve obtain ‘he Dirichfe region (or “Voronoi polygon") ofthe late: a polygon whose Irior consists of all the points nthe plane which are nearer ov particu "Cw ed hie at asa sCevinang Mier and Cae Yosen P3938 TAM NSbh Comtonamay Mase 13.00, 9Iatie point (such athe point 1) then to any other Itticepoint*_ Such regions, each surrounding a latice pont, evidenly Ht together to fll the ‘whole plane; infact, the Dirichlet region ia particular kind of fundamental rexlon. “The latices symmetrical by the hal-turn about the point | (or ay other late point). For this half-tuen interchanges the pair of lattice points XeVi, XV (in technical language, the group p¥ hasan automorphism of period 2 which replaces X and Y by ther inverse) Hence the Dirichlet ‘exon symmetrical bya half ture Tes precise shape depends onthe relative Tengths ofthe generating translations X,Y and the angle between thers. If ‘his angle i aright angle, the Dirichlet region isa rectangle (ora square) since the ccumecnter of a righv-angled tangle i the midpoint of the hypotenuse. Inall other cases tea hexagon (ot necessarily a regular hexa> {0 but sine its cenrally synmetial, ts pais ef opposite sides are equ fan parle, ‘Varying the latce by letting the angle between the translations X and Y increase gradually to 90", we soe that tWo opposite sides ofthe hexagon Shrink til they become single vertices, and then the remaining four sides form a rectangle (or square) ‘Reflections in the four of six sides of the Divichlteegion transform the ental lattice poin I into fur or sx ther latice points which we naturally cll the neighbor of the pois Anya potest Diet pon perpen ie "Sketch the various ypesoflatice hat can are i X and are subject othe fel tong econ they ey have the same length andthe angle betwee them my be SA" ors" india the Diet eon in each cave and State whee the symbety rp fs ep Ca Dy, Dy oF Dr 4.2. THE SYMMETRY GROUP OF THE GEVERAL LATTICE aing186 182) (a ap] ‘Any given latices easily seen to be symmetrical by th halF-turn about the midpoint ofthe segment joining any two ltice points [Hilbert and Cohn- Vossen 1p. 73), Such midpoiate form a latice of ner mesh, whose gen tating fanslations are halas long as X and Y (see the “open” points in Figure 420. FL Dinca ree ran anpvae Maton 40 (850.56. 21620 The “general” atce curs when the reduced generators deri length snd the angle between them ie neither 90° aor 60". In sucha case, the tans ions XY" and the above-mentioned half-turns are is only syiametry ‘peretions, In other words, the symmetry group ofthe general atice is ‘erived from pl by adding an extra transformation H, which isthe halF-turn [bout the point |. This group i denoted by »@ [Coxeter and Moser 1, pp. ‘41-4 Tie generated by the hal-tuen H andthe wanslations X,Y, in terms ‘ot which the haléturn that interchanges the points Land T = X°V'is HT. {Note that T itself is the product of H and HT.) The group is equally well [generated hy the three hall turns HX, H, HY, oF eedungantl) by thes three Sod heir product UX H+ HY = BXY, ‘which are half-tars about the four vertices of the parallelogram shown in Figure 422 Tis remarkable that any trang or any simple quadrangle (not neces= sail convex) will serve asa fundamental region fr p2.Half-urns about the midpoint of the tice oe four sides may be dented with HX, H, HY (gore 424) oF HX, H, HY, HY (Figure 420.rien generators forall‘The art offing «plane with a repeating pater reached its highest de velopment in threent-century Spain, where the Moors wsed all the seven teen groups ia ther intents decoration ofthe Alhambra [Jones 1}. Ther preference for abstract patterns was due to thir strict observance ofthe Second Commandment, The Datch artist M. C. Escher tee from such Sevuples, makes an ingenious application of these groups by using animal Shapes for their fundamental repions. For istance the symmetry group of his patter of knights on horseback (Pate 1 seem at Gis sight 1 be pl fencraed by a horizontal translation and vertical tanslation. But bY {goring the distinction between the dark and ight specimens we obtain the ‘more interesting group pg, which is generated by two parle glide rte tions, say Gand G”., We observe tht the vet eanslation can be ex pressed equally well as Go GIs emarkable thatthe single relation Gt=G? provides a complete abstract defisiton for this group [Coxeter and Moser opi). Clearly, the knight and hs steed (of ether color constitute a fun ‘amcatal repon for pg. ut we mst combine two such regions, one dark ‘and one light, n order to oblin a fundamental region lor Similarly, the symmetry group of Escher’ patter of beetles (Plat 1) sects at fis sight to be pm, generated by two vertical elections and sve Seal translation. Buton looking mare closely we see that there are both thark and Hist beets, and thatthe colors are again interchanged by glide feflections. The somplete symmetry group em, whose fundamental region isthe right or left half ofa ete of either color, is generated by any such ‘erica lie reflection slong witha ectical reflection. To obtain a funds Inettal feion forthe “smaller” group pm, we combine the right half of = beetle of either color withthe left hal of an adjacent beetle ofthe other solr. ‘A hole beetle (of either color) provides a fundamental region forthe group pt (with one of ts generating translations oblige) or equally well for 1. Locate he ae of mo lide ebcions nich seat pin Pes nd I tbe repented by usations a he ple cece AA SIX PATTERNS OF BRICKS Figure 4a shows how sx ofthe seventeen two-dimensional space groups arise asthe symmetry groups of familiar pateras of rectangles, which We may think of as bricks or tes. The generators ae indicated a follows: a A broken line denotes a miror, a “lens denotes a halfturn, a small square denotes quarter-tutn (ie, rotation through 90°), and a “half arrow” de note side reflection, ‘In each cise, «convenient fundamental region is indicated by shading. ‘Tnissepion into some extent arbitrary excepein the case of pm, Where is enizely bounded by mirror. ‘The procedore for analysing such a patter isas follows. We observ that tne symmetry group ofa single rick sD: (of order 4), which has subgroups (Grand Ds. It all the symmetry operations ofthe brick are also symmetry erations of the whole putern, 4 in emma and prom, the fundamental fegion sa quarier ofthe Brick, two ofthe generators ae the reflections tht fgenerate Dz, and any other generator transforms the orginal brick into & Feighboring brick. IP only the subgroup Cor D, belongs to the whole pat Wan Pr a4 o3 eeve vee me Fire ate tern (the way C; belongs to p2 or pg, snd Ds to pmg oF pg). the Fonda- ‘mental regions hala brick, and the generator are not quite so obvious, na toe ptr it undertod that rik” rectangle ia which onesies ciccas long avaother nec as an) bi relied the whole pater in be ‘imc wayavany other In osha nngasge he sume) woups amine on he Uke) Arche st the on rani pater of ike? 4.3. THE CRYSTALLOGRAPHIC RESTRICTION Grrr itn2) ‘A complete account of the enumeration ofthe seventeen two-dimensional space groups would occupy too mich space ULI seems worthwhile to give arlow' elegant proot* thatthe only possible eylie subgroups are CoC Cy and Cy. Inter words ‘The onl possible perods fora rotational ssmmery operation of latice are Lt P be any ceater of rotation of period a. The remaining symmetry +8, aon, Popa! Map 6 (90 p7 ‘operations ofthe lattice transform Pinto infinitely many other centers of rotation of the same peiod. Let be one ofthese other centers (Figure 445) a the least possible distance from P. A third center, P is derived from P by rotation through 2e/m about Q: and a fourth, Qs derived from {Q'ty rotation through 2e/n about P. Of course, the segments PO. OP, PO, arcall equal, It may happen that Pand Q' coincide: then n = 6. Tn all oher eases, since Q was chosea a he last posible distance from P, we imusthave PQ’ > PQ; therefore n < 4 fn = 4 POPO isusquare. Ifm = 5.PQ' is obviously shorter than PQ. If n> 6, PQ crosses PQ. but itis no longer nveesry to use Q" we alteady have P< PQ, which i flr Slendy absurd) 1. US and ate rotation rouph 20m about P and Q, whats TIST? itnclades two sch votatons avg th tame ped and eefore sino telaton Trt priod presto than tncade fo independent trations 446 REGULAR TESSELLATIONS (esi esate ss rth noch ma gear sie woos maemo Gtr 028) It was probably Kepler (1571-1630) who fist investigated the possible ‘ways of ling the plane with equal regular polygons. We shall nd it con: Veneat to use the Schlafsymbol (p,q) for the tesellation of regula -gons, «surrounding ech verte [Sehla Lp. 213} The cases 63). 44. BG) arillustrated in Figure 4.60, where in essh case the polygon drawn in heylines is the verte figure the g-gon whose verties ae the midpoint of he g fees ata vertex. (Since tesellations are somewhat analogous to polyhedra, ‘tismatural to use the word edges orth common sides of adjacent polygons. ‘nd face forthe polygons themselves) Fora formal defisiion, we may say tha a wsellatin is regular iit has regular faces anda regular verex figure a each vertex. “The tecllation (6,3) ofl used for ed floors in bathrooms. Tt can aso beeen in any bechive. (4,4) i familar inthe form of squared paper in terme of Cartesian coordinates, is vertices are just the points for which both cand y are integers (3,6) ithe dual of (6,3) in the following sense “The dual of (po) is the tessllation whose edges are the perpendicular bi sectors ofthe edges of {p,q} (see Figute 4.6). Thus the dual of {p, 4) is (Gp), and vice versa; the vertices of either ate the centers of the faces of the other. In particular, the dual of (4.4) ian equal (4, 4). ere 4 ‘The posible values ofp and g ar easily obtained by equating the angle cota p-gon, namely (1 ~ 2/p)y othe value i mist have if g such polssoas ome together ata vertex: “The thre posible ways of factorizing 4 namely 4 Te yield the three tessellatons already described, However, before declaring ‘that these aze the ony regular tessllatons, we should investigate the Face tonal solutions of our equation; for there might conceivably be regular “war” tesellation(p, 4) whose face (p) and verter figure (q) are revl polygons ofthe kind considered in $28. For instance, Figute 4.6 show. n pentagons placed together ata conmon vertex. although they over lap, me might expect tobe able to add farther pentagons 038 form 3s sellation (5, 3} (whose vertex gure isa dacagram), covering the plane @ umber of times. Buta fact this numbers inte as we shal see Consider the general regular tesellaton (p,q), wherep = n/d._Witcov cs the plane only nite number of times, there must bea minimum dis- lance between the centers of pars of faces. Let P, Q be two such centers inimum distance apart. Since they are centers of rotation of period sgument used in FS proves tat the oaly posible values of wae 3,4 6. Thusd = and these ae alo the only posite values ofp. Hence here ‘reno regular star testelations [Coxeter 112 is actually posible to cover a sphere tee times by using twelve "penta- sons" whose ides are ares of great eles [Coxeter ¥ . 11} ‘To had the symmeuy group ofa regula tesellaton, we teat its face the way we treated one ofthe bricks in $44. Clearly, the symmetry group of {p,q} is derived from the symmetry group D, of one face by adding the re- fection inside ofthat face. Thus ts genetated by elections in the sides ‘of triangle whose angles are o/p (at the center ofthe fae) (athe mid Point of an edge), and #/q (ata vertex) This triangle ie a fendament Felon, since it is ansformed into neighboring tnangles by the three gen~ trating reflections. Since each generator leaves invariant all the points on fone side, the fundamental region is unigue: it cannot be modified by add ton and subtraction the way Escher modified the fundamental regions of some other groups, “The network of such triangles, ling the plane, is cut out by all the lines ofsymmetry ofthe regular texellation The lines of symmetry include thelines ofthe edges of both (p and its dual (4p). tn the case of (6, 3} and (3,6) (Figure 4.6), there edge lines sien the ease ofthe two deal (4.4)'swenced ako the diagonals ofthe squares, In Figure 46d alternate regions have been shaded so as to exhibit both the complete symmetry {21005 pom, pdm and the “direct” subgroups p6, p (consisting of rote tions aad translations) which preserve the colors and the direction ofthe shading [Brewster 1, p. 94; Burnside 1, pp 41,417) Tnstead of deriving the network of tangles fom the regular tssllation, ‘we may sonversly derive the tsrellaton fom the network. or this pur pose, we pick out a point in the network where the angles are =/p, that where p shaded and p white angles come tether. These 2p ranges com bin to form a face of (7 9) exercises 1._ Ses te oral tion of reps” on page 62. (emp thatthe faces neal alk in tht the vertices all rounded ale) "2. Give genoa argue co pose tha the midis ofthe edges of 2 ear teseltion belong fot tice (mt Conse he gro 9 genre by hal trae Shout the ouch mip) "ick ut the midpoint ofthe ees of (63). Vet that hey blows to a tee Do they constitute the whl ace? “4 Draw pron of Istes whos yet srs ate p2, pm. em. em, rom 4.7 SYLVESTER'S PROBLEM OF COLLINEAR POINTS. ord (294 AAs we saw in § 4.1, latices a disret set of points having the property thatthe line joining any two of them coatains aot only these two but in ritely many. Figure 47a shows a finite “orchard in which nine points are Arranged inten rows of thre [Bail p 105 It was probably the invest- {ation of such configurations that led Ssvester* to propose his problem of is Prove tha mt possible t ran ay fn mabe of rel pin hat 9 ight line thwagh ery of them shal pas hough ar ons heya ei he ame ih ne Niger 476 ere 70 [Neither Syvester nor any of his contemporaries were able to think of satisfactory proot. The question was forgotten til 1933, when Karamata and Erdos revived it, and . Gallat (alas Granvald) finaly succeeded, ‘sing a eather complicated argument, Splveser's “negative” satement was repheased “positively” by Motkin ‘Wm piste rel plane ae mono sig , hn are ei righ e centating enc re of he po ‘The following proof, which somewhat resembles Barlow’s proof ofthe ‘rystallographic restriction (645), is due to L. M. Kelly. 11.3 Sper, Mamata urn So fom eal Times (19) 9 Guo is Sean Saba tes Maa Mendy 81 Th, Ho; tMe key 9 lo p 2° Mn, Momeni Aro Mona Soong st M Key nt Soar Coan i a“Thenpoints Py... Paatejoined by at most min ~ Ilines PaPs, PaPa, ete. Consider the pars P, PsP, consisting of a point and a joining line which are not incident. Since there are at most Hn — In ~ 2) such pas, there must be a leat one, say Py, PP for which the distance PyQ From the point tothe line i the smallest such distance that ocurs. “Then te line PsP, contains no other point ofthe set. For fit contained a: least two ofthe points Ps, Ps, Py woul lie on one side ofthe per- peadicularPy0 (or possibly one ofthe Ps woul coincide with Q). Let the point beso mame that these two are Ps Py, with Ps nearer to Q (or coi ident with Q) Then Pr PsP) (Figure 476) is another pair having a taller distance than P,Q, which i absurd ‘This completes the proof that there is always a line containing exactly two ofthe points Of cure, thee may be more than one such line; in fact, Kelly and Moser proved thatthe number of seb lines is atleast 3/7 1. "The above prot ye sine PoP cating nly these two ofthe PS. The pat Q actualy le eter Fad Pe "I point a aot all om on li, hey Bae at asm stint joins [Coxcte , path 2. Draw a confuration fn pints for which the lower Ht of 34/7 ordinary” 5 Similarity in the Euclidean plane In later chapters we shall see that Euclidean geometry i by no means the ‘only possible geometry other kinds are just as logical, almost as useful, and fn some respects simpler. According to the famous Erlangen program (Klein's inaugural address atthe Universit of Erlangen ia 1872), the ex terion that distinguishes one geometry from another isthe group of trans formations under which the propositions remain true_ Th the case of Eu- lideun geometry, we might at fst expect this o be tbe continuous group ofall isometries. But since the propositions remain valid when the scale ‘of measurement i altered, as ina photographic enlargement. the "principal ‘sroup” for Euclidean geometry [Klein 2p. 133) includes also “similarities” (rich may change distances although ofcourse they preserve angles). In the present chapter we classy such transformations of the Eucidean plane In particular, “diatations” wil be seen to play useful role inthe theory ofthe nin-point center of «triangle. Thess and other “direct sates fare ueated in the standard textbooks, but “opposite” similarities ($5.6) seem torhave been sadly neglected. 5.1. DILATATION ‘fn hore gon tg dee (Codan 3, Oo. Itisconvenient to extend the usual detaton of paral by declaring that ‘vo (infinite straight lines are parallel if hey have either 20 common point for two common points. (Inthe latter case they coincide) "This eonven- tion enables ws to assert that, without any exception,5.1 For each pnt A and liner thee i just one line rough A parallel Two figures are said to be homorhevie if they are simile and similarly placed, that si they are related by a dilatation (or "homothecy"), which say he defined as follows [Artin ¥, pS ’ dilatation ea transformation which peserves (or reverses direction: hati, st anaforms each Tne int parle ne. ere 5.12 Two given parallel ine negments AB, A'B are related by a unique blatation AB 48 For any pont P not on 4 is transformed into the pot Pin which the line through 4” paleo AP sess the line through B parallel to BP (Fi Ate Sa) and any point Q on 42 is ansformed into the point in which A meets the line trough P* parallel to PO. In other words dilatation i completly determined by its eect on any to given points [Coxeter 2, 8S ‘Cleary, the inverse of the dilatation AB —> 4°B' is the dilatation AB AAR. Also AB -» AB isthe identity, AB — BA is half-cur (about the midpoint of 48), and ABBA isa patallelogram, AB — 4°B is trans- lation For any dilatation which ot the identity, the two points 4 and B may. be'so chosen that is not an invariant point and 42 isnot an invariant Tine. Such a dilatation 4B — 4°B" (Figure 5.15) transforms any point P fon-4A” into point Pon the parallel ine through 4" whichis 4” sell A Similarly, it transforms any point Q on BB'iato «point Qn BA, IF Aa tnd BB" are not parallel, these two invariant ins intersect ina invariant point 0, Hence [5.13 Any dilation tht i nota tronlation has a invariant pointe This invariant point 0 is wuqu. For, dilatation that has two invariant points O; and Oy can only be the identity, which may reasonably be te {garded a$@ translation, namely a translation through distance 2r0 [Weyl 1p. 031 Clesely, any point Pi transformed into «point Pon OP. Let ws write or = yor, withthe convention tat the number is postive or negative according as P and Pare on the same sie of O or on opposite sides. With the elp of Some homothtic tangles (asin Figure 5.18), we sce that Aisa constant, that is independent of the positon of P. Moreover, any segment PO i teansformed into segment | A | mes at lone. and oppositely direst if A <0. We shall use the symbol Of) for te dilatation wih center O-and No A. (Court [2 p. 40 prefers (O.AP") In particular, Of1) i the identity and O{—1) i a alf-sun. Cleat, the only dilatations which are alsa isometries are half-uens and translations. Th the case of translation, sch a symbol as O() is no longer aval Fawe 56 1 wars he inverse te ata OW? 2 We rout 040) and Ox) Ooh where is 0? ac, 20m he ston 0) nem of) pr saris, b Cresco 4. Explain heaton af he ponsaraph Figure Se an estrument evened by ‘ring Serato 0 eh pure ling sep eden ered amen ares ovo yf Md es of arly the respeane ras 44 48 roman clear when se Shape oh prac. ‘am changed. "The fsrumeat prod at Whos spol pint abeand tacng point at (or vise vera), pth tet ace over the ins of fen figure te pene pin rams a horuthese copy. The postions oO an Pare buble on in especie rod war tallow varios choles fhe ati 04 {reas ofcoue, fe taen 0 keep Oaad Folie with?) ‘How could the petograph te moe 0 a8 yi a lataton 7) wih 5.2 CENTERS OF SIMILITUDE Nil 8.18) Niger 520 [A dilatation 00), transforming Cnt C, tansforms a crle with center (Cand radio 7 ino eile with center Cand radius [Aj Conversely, fs we sein Figure 2a, two ctl have distnet centers C, Cand ne ‘ual rai ee they are related by two dilatations, Ot /a) and Oy(—a'/a), Sohose eaters O and Oy divide the regment CC” externally and internally Inthe tatio asa’ [Court 2, p. ISA], These points O and Oy are called the enters of simitde of the two cies, "To construct them, we draw an fitrary diameter PCP, ofthe frst cicle anda parallel radius CP of the Second (with Pon the same side of CC as P; then O lies on PP, and Os on iP I wo circles ate concentric or equal, they ar sl related by two diata- tions, but there is only one center of silitude- In the ease of concentric ‘Stes this is because the two diataions have the same center. Inthe ease ‘equal eirles iti because one ofthe dilatations isa translation, which has no center (The other the haltura about Os, which is now the ms point of CC") i [A Vanephen and ©. R. Veldamp (American Mathematical Mot 7 (195 p, 178) found ha forthe tage sone Exc 10a 81 (pgs 10 the ete of smite ofthe two “Soy irae the nee tnd the Cerone po the pit of concuens of eines joing the vers he pots cit fe te pecunelyoppostc nish hin ses 1. Ifowoequal irs have ao common pit hey have tw parle commen an- {sand to ober common tangents vouph Oy (miway betmeen the cate) If they oh hey Rie ay the common gn fy ite they hate ony 2 Any common tangent of two une ciles pases trough 2 centr of smi tude. Seu the penton fhe ene indie and recod the nmber of eon mon tangents eth ve enenally diferent stucco fro such cles, (To of the bre ne shown in Figure 320), Given 1 elation), O40) wh Ad, describe the poston of the ‘uma pint Com which Bt have te same ft 5.3. THE NINE-POINT CENTER ‘Consider an arbitrary triangle ABC, with cccumcenter 0, centroid Gand lothocenter H. Let 4" B,C’ be the midpoints ofthe sides and A", (C" the midpoints ofthe segments Ht, HB, HC, ae in Figure a. Cleary, both the tangles 4°B'C, 4"B°C” ate homothetic to ABC, being derived fom ABC by the respective dilatations G(—), MC). The former provides new proof thatthe medians are concurrent and inset one another. Since G(~4) and (4) are the wo dilatations by which the ine-point ciecle cam be derived from the eircumeircle [Court 3, p. 10s) the points 6G Hare the centers of similtude of these to circles, and the Ear line {Gif contains the ceaters of both crses: not ony the cicumcenter 0,48 ‘we know already, but also the mine;pointcemer N. Since the values of for the dilatations are =, the aine-poit radius is half the circumradius, ‘and the centers of simiitode HG divide the segment ON externally and internally inthe ratio 2: 1 (Figure $34), Thus isthe midpoint of OM teal wange whoo verace ae (11h Cemee dy ‘2. I-ABCH isan orhocencie quadrangle (oe 1.72), the four Ener ines of thet5.4 THE INVARIANT POINT OF A SIMILARITY ‘So cone allie por of green ie Wn a ave om ry of oie me Siig sear eae he 4 cies 865-188 (Wore Poe 0.8 [A similarity (or “similarity transformation,” or “simiitude”) isa trans formation which takes each sepment 4B into a segment 4'B” whose length ‘egiven by a may ‘where jis a constant positive number (the same fr al segment) elle the ato of magnification (Clara's “catio of smulitude”).Tfllows that any rangle trasformed into a similar tangle and any angle into an equal (or opposite) angle. When «= Is the simian isan isometry. Other spe ‘al eies are the dilatations O(=p Ate amie ite the lc retin or “pal at” Figuce Say which the prot of «lta O49) an rotation about (OF Anexhet i the dla reflection (Figure 5.46), which is the product of a Sinton O09 andthe recon i ine trough O. We woul not obtain Sothing new tin sie da) we replaced this tation Ow) by OC Forge Ot) = Ot =i) Oph and =I) va ature product of (fe) and wut out aout Oi the sane a the produc of Op) sae eTouan tough eet 2 and since {1} othe product oft pe erica rections, the produ of Oa) and the reflection ia 8 Kne m Uhrough 0 isthe same as the product of O{—) an he eefestion in the Fine ‘through © perpendicular to m. In Tat, 4 lative tection has two pet pendicular invariant lines (its axe), which are the internal and external ie Sectors of 404 (and of 2 BOM). ‘Geary (ef 3.10, 3.41 Any ovo similar triemgles ABC, A'HC: are related bya unique sim larity ABC ~» A'B'C. which i director opposite according a the tense of A'BC agrees or duagees with that of ABC. Tn other wonds, a simiasity is completely determined by its effect on any ‘hee given non-colinear points, For instance, the two triangles CBP. ACF ‘sed in proving Pythagoras’ theorem (Figure ia) are related by a itive tation, the product ofthe dilatation RAC/CB) and a quartertum; and the two triangles 4BC, ACF (in the same figure) are related by 2 diltive reflection whose ates are the Beacons ofthe angle Here is another way of expressing the same ide Any to line segments AB, A°B are related by just two similares: one dlirect and one oppose For instance, the segment 4 can be completed to make a squate ABCD fon ether sie ofthe ine 48, and similarly there ate two ways to place & square 4'B'CD' on a'8. The similarity ABCD + A'8 CD" is director opposite according a the senses round these two squares agree or diagsee “Theorem 5.13 suggests the possibility that every similarity with «1 ‘may have an invariant pont, ta given similarity i nota dilatation there must be at least one line tr Formed into a nonparallel ine, Let aad.’ be erresponding segments ‘auch a par of lines, ad let the given sila’ (directo opposite) be ‘determined by similar, but not congruent, paralllograms “ABCD. ond ABCD {lor example, by squares, ts above)Let P, 0, R, Sdenote the poate of intersection ofthe pairs of correspon ing lines AB and 4°, BC and B’C, CD and CD, DA and DA’, 38 in Figure S4e. Suppose the given siilaity transforms P (on AB) into P (on AB) and R (on CD) into R (on CD). Let O be the common point ofthe lines PR: PRC (which cannot be paral, fr, ifthey were, PRRP” woul! be ‘parallelogram, and the segments PR, PR’ would be congruent, contra iding PR aPR), Sine the point pairs PP’ and RR’ lie on pallet lines 4'8" and CD, oP _ or. On = OF “Therefore, the similarity leaves 0 invariant. Moreover O is onl nvr fant point For, if similanty with « 7 1 ad two invariant points Os and (On the distance OO; would be left unchanged instead of being mulipies bye Hence 1542. Any slay that mot am omer has just one ivariant pois Moreover, gen sw similar paaltograms ABCD and 4'5'CD: we cn asthe method indict in pre Ado comme the cet of te simi Ini that elts thm. For hung seen that Os onthe ie PR, we cn apply he same rearonng ing BC and DA instead of 4B and CD toshow tutes onthe ine 05. This ma ierent line, fo, if 0, R, S wee all collinear ve would have PA PS. PA agg RC BQ _ BC, Fam ro = re 4 RD” RS” RD” raking both P and R invariant. Hence, 0 con be comsraced a the pon of acto ofthe lines PR and QS. How ap the eof contin be ed fora deen root of Theorem S22 5.5 DIRECT SIMILARITY Consider a given dret similarity whose ratio of magniscation wis no Since thee is am invariant pont, ay 0, this similarity may be expresed as the product ofthe dilatation Ofp) and a direct isometry leaving O invariant, ‘By Theorem 3.14, such an isometry is simply @ rotation about 0. Hence, S51 Any cist similarity thai not an isometry is dlatation ora dila- ‘oureses 1. Whatis the produ of wo dave rotations? 2 How can eo ies be wed ote the invariant ot fhe dst iaiy ta eats wo given ncngrint pment om sopra ne? [Casey yp 185) 5.6 OPPOSITE siMILARTY Consider a gven opposite similarity whose ratio of magnification isnot 1. Since there is an invariant point. say 0, this simlanity may be expressed the product of the dilatation OG and an opposite isometry leaving O favariant. By Theorem 3.14 sch an isomety is simply the relection sm line through 0. Hence ‘S81 Any oposite similar that is ot am isometry i a date refection¢ sastpat asec nis ern me Sern pe gor mt tn a ed stile meceeitn te age to ene Lag onprces ta teraiiece mcm sh ni Me Bien tnt a pla eat erm Circles and spheres ‘The present chapter shows how Euclidean geometry, in which lines and planes play a fundamental role, ean be extended to imersive geomet, in ‘which this role i taken over by efcles and spheres. We shall sce how the ‘obvious statement, that ines and planes ae circles and spheres of infinite radius can be replaced by the sophstieted statement tht fines and planes fate those cicles and spheres which pats through aa “ideal” point called “the point at infty.” In $69 we shall biey discuss a sill more uns seomery, called ellpi, whichis one ofthe celebrated "non-Euclidesn eomewics. 621 INVERSION IN A ciReLE Can ib a a he aaa of pr wae ely long fortone strona es inoue nl on plop Yo nh saectin wn heh rie eee bse cana ‘tere ronge ot ed sre by ore a col Shon how on obra god fr th ne wal hh ao LL sme 097 All the transformations so far discussed have been similarities, which transform straight ines oto stright ines and anges into egal angles. The ‘tansformation called iverion. which was invented by Le J Magnus in 1831, is new in one respect But Tamil im another: it traasforms somestraight lines into circles, but i til transforms anges into equal angles. Like the reflection and the halfturn, iis involutory (thats of period 2). Like the reflection it has infinitely many inediant points; these do note fon staight line but ona ciee, and the center of the cise i “singular” ithas no image! Given a fixed circle with center 0 and radius k, we define the imerse of, any point P (stint from 0) tbe the point Pon the ray OP whose dis {ace from O satis the equation oP x on It follows from this defaition that the inverse of Fis Pitsell, Moreover, every point outside the eile of inversion is transformed into a point in side, and every point inside (except the center 0) nto a pint outside. The ‘irl is invariant in the strict sense that every pointon sinvarian. Every line through O is invariant asa whole, but not point by pont, "To construct the inverse a «given point P (other than 0) inside the cit cle of inversion let T be one end ofthe chord through P perpendicular to ‘OP. asin Figure 6.10. Then the tangent at T meets OP (extended) in the ested point P. For, sine the rightangled triangles OPT, OTP” ate simi Tar, and OT =k op _k ©” OF To construct the inverse ofa given point P outside the cite of inver sion, lt Tbe one of the points of intersection of this etce with the cle fon OF a diameter Figure 6-2). Then the desied point isthe foot of| the perpendicular from T to OP OP = {the inverse of P can easly be constructed by the use of eon passes alone, without a ruler. To do 50 let the circle through O with cen fee P cut the ciel of inversion in Q and Q'. Then isthe second inter: section ofthe circles through O with centers Q and Q" (This is easily seen by considering the similar isosceles iriangles POO, OOP") "There is an interesting connection between inversion and dilation 6.11 The product of inversions in wo concenrc circles wih radi k and HE the dilatation O() where w= (RP. “To prove this, we observe that this product tansforms Pinto P”(on OP) where OPK OF = OF Kop =e andsheretore op _(x)t or (t) Using compasses lone, construe the vercs of egal hexapo8 2 Usingcompasses alone, oats pint Bso thatthe spent OB ica ong st pen semen Of ‘Using compass slo, construc the inverse of point stat tom the ea ter O ofthe cle osaveron.Deseribesprsedure for iaverig poi aia ‘4 Using compasses alos, bit» stem egment 5 Using compan alone tits given spent, Descie a procedre Ford ‘dig Senet it any ie umber of egal par "Note. The above poblnsRelons tthe Geometry of Compass, which was de veloped ndepensenty by G. Mohr it Desk (167%) tod L Mascon aly (G79, For conce vesion af the whale story see Pele [1 pp. 2325} or Cowra a Robbie top. 62 ORTHOGONAL ciRcLES Acie in happy ing be Thr ow te oy purpeior Onaga ey (180-) ‘Two circles are said to be orthogonal if they’ ut at right angles, hat if ‘hey intersect in two points either of which the radius ofeach is tan gent othe other Figure 624), ‘By Euclid 111.36 (Se p. 8) any circle through a pir of fverse points is invariant: the ciel of inversion decompores i ingo two arcs which invert into each other. Moreover, such a ice is orthogonal to fhe cite of in Version, and every circle orthogonal to the crete of inversion is invariant {nhs sense. Through a prof inverse points we cen draw 2 whole pene!cites (infsitsly mans), and they are all orthogonal wo the cele of in- Version. Hence 6.21. The inverse ofa given point P i the second intersection of any 10 cities through P orthogonal 0 the circle of ierston ere 60 ‘The above remarks provide a simple solution forthe problem of drawing through a given poit P, a circle (oe line) orthogonal to two given eices. Let Py, Ps be the inverses of Pin the two citles, Then the cree PPPs (or the line trough these thee points, I they happen to be collinear is forthogonal tothe two given ices IO and Care the centers of two orthogonal citles and 78 in Figure {62u, the citle on OC as dlameter passes through the points of intersection 7, U,. ery other point on this ele is inside one ofthe two orthogonal circles and outside the other. Ic follows that a and b are to perpen ‘clas ines trough O and C respectively, ether ¢ touches ad b touches ‘Sora cuts y and lies ouside & or acs outside y and B cus 6.3 INVERSION OF LINES AND CIRCLES ‘We have seen that lines through 0 invert into themselves. What about other lines?” Let 4 be the foot ofthe perpendicular from 0 toa line not through O. Let abe the inverse of 4, and Pthe inverse of any other point, Poon the line. (See Figure 6 3a where, for simplicity, the eile of inversion has notbeen drawn) “Since OP x OP = kt = 04 x 04, the tangles OAP, OPA’ ate similar, andthe line AP inverts int the circle ‘on 04" as diameter, whichis the locus of points” from which Ot” sub tends right angle ‘Thus any ine not through O inverts int a ec trough (Ovand vice vers. ere 638 Finally, what about a ctele not through 07 Let P be any point on such ‘cil, with center Cand let OP mect the circle again ia O. By Euclid HLS gin, de product p=0P x00 is independent ofthe position of Pon the circle. Following Jueob Steiner (1796-1863), we call this product the power of O with espect tothe ice kis postive when O is outside the eile, zero when O lies on the ctl, and We naturally regard it as being negative when O is inside (so that OP fad 09 are measured in opposite directions) Let the dilatation O(R*/p) transform the given circle and its radius CQ into another cil (or possibly the same) and ts parallel radius DP” (Figure 63%, of. Figure 5.20), so that or _ op _ 00 = Oc =p Since OP x 0 = p, we have, by mulipicaton, OF x OF = Ke “Thus Ps the inverse of P, and the circle with center Dis the desired in-vets ofthe given cicle with ceater C. (The point D is usually no the in vene of C) ‘We have thus proved that the inverse of acitele not through is another ciel ofthe same kind, or possibly the same circle again. The later pos bility ours in just two cases: (1) when the given ctl is orthogonal fo the circle of inversion, so thatp = K and the dilatation s the identity: (2) when the given circle i the circle of inversion itll, so that p = —s# and te il tion isa halon, "When pis positive (se the left half of Figure 3b, s0 that 0 is outside the citle with center C, this ciel is orthogonal to the circle with center O and radius yp; that is the former circle i invariant under inversion with ‘espect to the latter. Tn eet, we have expressed the given inversion a3 the product ofthis new jversion, which takes P to Q, and the dilatation ‘O(k/p), which takes Q 19 P.When p ie negative (asin the ight half of Figure 630), P and Q ae interchanged by an “antiversion:” the product ofan iverson with radius \/=p and a haltur [Forde 3, p20) ‘When discussing isometes and other similarities, we distinguished be tween diet and opposite transformations by observing thei elect on rie angle. Since we are concerned only with sense, the triangle could have beea Feplaced by its circumeircle. Such 2 distinction can stil be made fo aver~ sions (and products a inversions), which tanstorm cices nto eles. In- Stead ofa tiangl we we a cece: not an arbitrary cle Bota “small” circle ‘whose inverse is also "smal thai, aezcle not surtounding O. Referring {gain tothe left half of Figure 63, we observe that P and 0 describe the ‘le with center Cin opposite senses, whereas Q and P describe the two Circles in the same sense "Thus the inverse points P and P* proceed oppo- site, and “Inversion is a opposite ansformation [efolows thatthe product ofan even numberof inversions is direct. One instance is familiar: the product of inversions with espect to two concentric clrles i «dilatation awe 26 ews 64 1. Porany wounegul ile that do otter one ofthe wo cee of se 653) the center of acre mich inverts cer the rence ite the i “ter Fortwo nega intscting ce, ot enters fmt hve th prop ‘ry. Wha happenin the cat equal itesetng cle 1. Expsn the action of Pao cl Figure 3) a stument invented by ‘A Pease inet for he pupote of Satin the iver of any ven aT feemed by four egel rods, hinged a the corners fa homibas APBP and te egal (Gong) ols coanestig 0 opp cores, an B to 4 Sed pot. When 4 Dene point nterted at ans aca pn at P(r vice ers) un theater 6 Ital over the curve fate Ogre, the peel oi dams te lovers Age. Tn parialar lu seveth rod tn another pe are aduced 0 to Lep Pom a ‘Be pasing tough O he loss of Pl be a aight line Thi kage gies a0 ‘art eolton of he iniortant mechanical problem of converting cela nto ve ‘inar motion, [Lamb 2,314) 2 Bxlan the action of Har’ nee igure 63, an instrument invested 8) HL Hart in 74 forthe sae purponet eater’ cal Tt egies nly fou rode hinged he cores of ered pralelogam” ACD (with 48 = CD, BC = DA). {Tne ie colina pats OP, ron the especve re 4B, AD, BC. remain ca Teron ain parle oC and BD) wen he shape ofthe rose paabiograe Ihunged Ar Blo, the instrament pivoted x. [Lamb 3p. 315} 4 Wil respect oa cco ai et p be the per ofan oui ait 0. ‘Then hee mith center O andra fave yo acide as ip 64 THE INVERSIVE PLANE Sg (9.138 We have seen thatthe image of a given point P by efetion ina line (igure 138) the second intersection of ny two circles through P ortho ‘onal othe mirror, and thatthe inverse of Pin a ccleis the second intr Section of any two eitles through P orthogonal tothe circle of inversion. Because ofthis analogy, inversion is sometimes called “reflection in ct le" [Blaschke 1, p. 47}. and we extend the definition of a cic so at n= ‘lade a straight linea special or Nimiting”) case: acl of infinite radius ‘We can then say that any thee distinct points lie on a unique cirle, and that any ctl inverts into acre Inthe same spirit, we extend the Euclidean plane by inventing an “deal” ‘ein at ifinty 0, which is both a common point and the common center (fall suaight ins, egarded as citls of infinite radius. Two cces with ‘common point ether touch each other or interact again. This remains ‘Sbvious when one ofthe ices reduces to a stright ne.” When both of them are straight, the lines are either parallel, i which ease they touch at (07 or intersecting. in which ease O” fe ther second point of interection [bert and Coha-Vossen 1, p. 281)‘We ean now assert hat every point has an inverse. All he tines through (0, being “crles" onhogonal othe cil of inversion, meet agai in Othe inverse of 0. When the center O is 0" itself the “citele” of inversion is straight, andthe inversion reduces to reflection “The Ewcidean plane with O”added is called the mere (or “conformal") ‘lana Ic ives inversion it fl status asa “eansformation”(§23): 2 one {owne corespondence without exception, ‘Where to curves coss eachother, hei angle of intersection is matorally ” invert into all the eicles through O and Pan “nuerseting (or "elipic") pncl of coanal ices, inctaing te straight line Opp" asa degenerate case."The system of concentric eicles with centerconsiatng of circles orthogonal o thes Lins, inverts ito a noninerseting {or"nyperbolic) pencil of coaxal circles (drawn in broken lines). These ‘ireles all mise one another and ar all orthogonal to the intersecting pen- ‘al. One of them degenerates toa (vertical) He, whose iaverse isthe circle {with center?) pasing through 0. ‘As kind of limiting case when O and P coincide Figure 68, the ct- es that touch a fxed line at Oxed point O constitute a angen (or "pal boli”) pene! of coaxal circles, They tvert ia cee with center 0) into All the lines parallel to the Axed line. Orthogonal to these lines we have nother system of the same Kin, inverting into an orthogonal tngeat pen- ‘lof costal circles, Again each member of either peal i or ‘very member ofthe ater. ‘Any two given ctles belong to pencil of couxal cies of one ofthese ‘three type, consisting of al th circles orthogonal to Both of any two clrees orthoganal to both the gen crles. (More concisely, the eoaral eles con- fix of al the eirles orthogonal tall the ctcles orthogonal to the given tlrsles) Two circles that evt each other belong to an itersecting penell (and ean be inverted into intersecting ines); two circles that (Ouch each ‘ther belong toa tangent pencil (and can be iaverted imo parallel ines), fo czcles tht miss each other Belong to noninterseting pencil (by the remark atthe end of B64), Each pencil contains one stright line (a ctcle of infinite radius) called the rad axis (ofthe pen, r of any two of is members)" Foran inter- ‘ctingpencl, this ithe line joining the two points common to all the ei ‘les (OP forthe “unbroken” circle in Figure 65a) Fora tangent pencil, Line Gin, de Ee Rhine, 18) itis the common tangent. For a noninterseting pencil isthe line mid way between the two limiting pont (ot eles of 250 radius) which are the ‘common points ofthe orthogonal intersecting pencil. For each pencil there isan of centers, which s the radial exis ofthe orthogonal pencil, Hence 4.51 {ftangents can be drawn tthe circles ofa coaxal pencil from a point om the radical axis, al these agents hae the same length ‘The radical axis of two given circles may be defined asthe locus of pints of equal power (§ 63) with respect to the wo ices. This power eam be measured as the square ofa tangent except inthe case when the given it~ ‘les intersect in two points O, P, and we are considering a point on the Segment OP; then the power isthe nepative number 40 x 4P. Te follows that, for thre circles whore centers form a triangle, the three radical axes (ofthe eirles taken in pats) concur in a pot called the rad. fal center. which bas the sume power with fespec o all three eiteles. IP this powers positive, sts square roa s the length ofthe tangents to any of the Circles, and the radical center isthe center ofa circle (f this radius) ‘whichis orthogonal to all the given cicles. But if the power i negative, ho such orthogoaal cle exist ‘Tae possiblity of inverting any two noninterseting ices nto concentric circles (by taking O at ether ofthe limiting points) provides «remarkably Simple proof fr Ssinr’sporism:* we have two (nonconcenirc) eile, fone isi the other, and erles are drawn soccessvely touching them and fone another, asin Figure 6. it may happen thatthe ving of touching ct les loss, tha is, thatthe ast touches the rs. Steiner's statement is that, Irthis happens once it wll sways happen, whatever be the position of the frst circle of the ring, To prove thi we need ony invert the origial two tlie into concent cise, for which the satement ie obvious over: p 23) ee sho Cone, ted ing of phere Spa Maen 8 stp Hoe ag fe Bm‘ngs te remsning members in pis ad inverts ach member of he otogl 22 The to ming point of aoninteneting pens ae vers eah thera any member the peel 1. ftw oles hve oor four common tangent thera aijoins the mi pois ofthese como tangents Iwo ces have comma tanget (1 oe "4. When a noninterscting pen fcousal cls verted into a pen con cee cles what happens the ting pons? ‘In Stine’ pi, the poi of contact of zs cts nthe ng al ie con cieand th wil serve to tavern two ogi ces ito each thet. DO ‘becom of the res inthe ig be om ek? ‘6. Fore rangecomiered ip Exerche 10a § LS (gue 16, be ini cca vith the "wo other els" (Soy eto. 6 THE CIRCLE OF APOLLONIUS The analogy between reflection and inversion i reinforced bythe fllow= ing PROBLEM, To find the lacus of a point P whove distances from two fixed points A, A’ are in a constant ratio Vu 0 that AP = WAP. When 4 = 4, the lous is evidently the perpendicular bisector of 44, ‘thats the line that reflect 4 into-A” We sal ze that for athe values Dott isa cece that inverts 4 into 4. (Apollonius of Perga,c. 260-190 8.) ‘Assuming p 71, let P be any point lor which 4°P = mAP. Latte ternal und extemal Biector of Z4Pa’ meet 44a Ay dy (a in Fla tre 6460, where 4 = 9). Take E and Fon AP so that 4°E is parallel to AyP and AF is parallel to 4sP, thats, perpendiculat to AyP. Since FP PE, we have AP_AP | Ady AP _ap PE Pa? Wa,~ FP = Pa’ (The former result is Euclid VI.) Thus and 4; divide the segment 44” {aternally and externally in the ratio t= and thee location i ndepend- ent ofthe position of Since 2 4,P4yisa right angle, Plies on the cit fle with diameter Avda (Conversely iF4, and Az are defined by their property ofcividing AA" in theratio ps and Pis any point onthe cvcle with diameter yz, we have ar PEO AA ds ‘Thus FP = PE, and P, being the midpoint of FE, isthe ctcumeent the rightangled wrangle EFA. Thecefore PA’ = PE and 4p AP ot Pa PEA 5 {Cove , p15) nly thetic of Apollon AA vets into A, Fo, if 0 its center an kite ay, the tances @'= 40 aod @ = 40 sly aah dds dds atk Rae ae te whence or = We. 1. When vais wil 4 aod" emai edhe is of Apollon form non ‘nesting peal with nd 4 for lining pots 2 Given ine fand two pont 4.4” (Aa 0 sate pots Pon For which sheato 2/40 he sunimam oe sma. (Hin Conse he cle oe A. ‘hts centr on "The poem sd © N.S Mendesoh, and the hit o Riche iihum) rest AAA inte of 44 tn he notation of igure Sta (nich embod igre 6.6) the ce on Avdzand Byes damelts ets 0 pois 0 and 0, uch hat the inp O48 nO aes ate ergs O48 nd 04. OF He SC ma sete SP RB wer amT ce CDs coo ue Ths ta ods with 8 eon a :Lette moni stance Hewes two snintestng cic be define a he rata ot ofthe ao of th a (he larger othe sale) ft consentic Tis into nich te en cineca be averted. The, stenting peas Seow eles inner any this order the tee oversees say uaa) + (an) = (0000) {6 Two sven unequal ives are rested By itty many ave rotations and ty infil) many doe eflsons, The ot of ara poi Gn tes cas) Theil faving fr dametr the segment jsing the two cer of imide ofthe [rence (This lonsietonen st the ce of mide the ven ies) What he corespendiag rent for wo gies equa cles? 1. The ieverss,n wo given cls, of point thir ee of simone images ot each ober by eet nth rain ox of he two ees (Court, p. 13 16.7 CIRCLE-PRESERVING TRANSFORMATIONS ving observed tha inversion ra transformation ofthe whole aversive sane incuding te point at ni) ito self, taking icles into ile, we atuallyack what isthe most general transformation of his kind. We ds- {ings two cases steording asthe point at init i, ors aot ineariant. ih the former case, not only ae cece transformed into circles Bat slo toes ino ies Wih the ep of ach 2} Gee p.7) we deduce that ‘away of angles is preserved, and consequently the measurement of angles ‘Syrercrveds so that every tangle is anaformed into a snulr tangle, 208 {he transformation se similanty G3). Sane ins ether hand, the given transformation T takes an ordinary point nto the pot at infinity 0”, we consider the product nT, where ie the version inthe uit cicle with enter O, This product JT; leaving” invari nt isa similarity. Lat A? be it ratio of magnifeation, and Je the inver- sion inthe cicle with center © and radius K-Since, by 6.11, JJ, the station 04%) esa 1.Pean be xpresed a J, where’ ian TH3s, the product ofan iverson and an isomety, "To sum wp, 6.71 Every cle prrersing transformation ofthe inverse plane iether 4 similarity or the product ofan inversion and an somet i fllows that every ctl preserving tunsformatin isthe product of at ‘ost four inversions (provided we reparda reflection ae aspect king of i ‘ersion) [Ford 1 p26) Such a tansformation i called«horieraphy or am nuhomogeaphy ascording asthe numberof inversions is even or add, The product of two inversion (either of which could be jst a refecton i ealled {voter or parable or dave homography according as the two inverting ‘les ate tnterseting, tangent, of noninterecting (Le, according ae the ‘orthogonal penil of vara citls is nonintersecung, tangent, or inter= Sccting). A special eases we ave, tespecively, toaton, a tanslation, fand 2 dilatation. The most important kind of rotary homography is the Mobis inoturion, whic, being the inversve counterpart of a half tun, isthe producto inversions in two orthogonal circle (eg. the product, ‘ofthe inversion ia a cicle andthe refeeuon na dameler). Any product of Tour iaversons that cannot be reduced to 8 product of two is called Ioodromic homography [Ford 1, 9.20, Slonand'azisomety) Sand ne usigue. Thre uly valid xpresion Sia Seb he ometry presides te fverson Why docs his revi prodet nse the me S" Under hat cumstances wil we ave = J? 6.8 INVERSION IN A SPHERE By revolving Figures 612, 624, 632, 63b, and 6:4 about the line of centers (OP of O4 oF OC), we se that the whole theory of inversion ex- tends readily ffom circles inthe plane to spheres in space. Given a sphere ‘with center O and radius & we define the inverse of any point P (distinct From 0) tobe the point Pon the ay OP whose distance om O satistes OP x OF = ke Alternatively, isthe second intertecton of three spheres through Porthog- ae ee ere retiree ee ance ere eee Serevided we include, as a sphere of infinite radius, a plane, which the inverse tofa sphere through 0. Thus, inversion sa transformation of inversive (or ‘conformal space, which s derived from Euclidean space by postulating potnt at infin, which lies on al planes and lines. ‘Revolving the cicle of Apollonius (Figure 66a) about the lie 44", we ob tain the sphere of Apollonia, which ay be described as follows 6.81 Given two points A, A’ anda postive number met Ay and Aa divide Ad" internally and extern hth ratio 1: Then the sphere on yA a3 ‘ameter ste lcs ofa poim P whase tances from A and A aren this ratio 1. trasphre with enter Oinveis ino 44nd Bit the wlaghes O4 and 2 tnterms of@ = 04 and) = OB, we Basen the notation of ES. 1) 4. -The*eos rai” ay fue points reseed by any inversion: [cae 1p 100) “4 To ophees which ou ach ther at avert it parallel plans Leta fy be thee spheres all ouhing pe aot Le ey, 0, Bea quence of pees toucng one anther accel) ab ll souchng fy Theo aches that ne havea ogo spheres inerocked with he oil Ang tite ine lovecat sphere whose ene the point of cont of aa 8.) 6.9. THE ELLIPTIC PLANE ‘nome vncee wo, he [Doi moved ier ad od oe he st th plc [Aegean i ow ‘Dnt ovo romeo to ponh igi Be ord oy ‘rash etpaer and oberg acer b beg pape wes 98) {Theta of vi Ee Let She the foot ofthe perpendicular trom a point N co a plane o, a8 in Figure Se. A sphere (not dase) with eater N and radius NS invers the plane into the spheres’ on NS as diamcter [Johnson 1, p. 108). We have seen that spheres invert into spheres (or planes: therefor ices, bing inte Section of spheres, saver iato eces (or lne)_ In particular. lth circles ino invert int circles (great oe small) the sphere and al the lines ino invert into css through N, Each point Pines a corresponding point Pon, namely, the second intersection of heline NP with. Conversely, ‘ach pont P” on except N, corresponds tothe pint Pin which NP" meets fn The exception ean be removed by making o an inversive plane whose point at infinity sth inverse of ¥- “This inversion, which pute the points of the inversive plane into one-to ‘one correspondence with the pots of a sphere, i known a sterengraphic projection. eserves as one of the simplest ways to map the geographical Hobe ona plane, Since angles are preserve, small lands are mapped wth {he cornet shape though on various scales according o thei latitude. Another way is by promonic (or central) projection, in which the point from which we project isnot N but 0, the center ofthe sphete, asin Fig 1ure69b, Each point Pina yields line OP joiningito 0. Thisdiameter tests the sphere in two antipodal points P,P which are both mapped on the same point P. Each ine m ina yieldea plane Om, joining tto O. Ths ‘iametral plane meets the sphere i great cirele. Conversely, cach great sree ofthe sphere, excep the “equator” (whow pla is parallel 0 9) cor fesponds toa line in «This time the exception can be removed by adding fo the Euclidean plane oa ine at ify (representing the equator) with ll it point, called pone ar infin, which epresent pairs of antipoda points ‘onthe equator, Thus, al the line paral to a given ine contain the same point at infinity, but fines in diferent directions have ifevent point at ‘ny all ying om the same ine a inBnity. (This Hea is de to Kepler apd Desargues) “Wen the line at insity is treated jst ike any othe ine, the plane 0 ‘extended is called the projerve plane or, more pecisey, the rea projective plane (Coxeter 2) Two paella ines mest i's pot at iaiy, aad an Srdinary line sets dhe lie tinny in a point at inity. Hence6.91 Any wo lines ofthe projective plane meet na pin. Instead of taking section of al he lines and planes trough 0, we could ‘more symmetrically (hough more abstractly) declare tha, by delntion, the points and lines ofthe projetve plane are the ines and planes through 0. ‘Thestatement 691 ie no longer surprising, stmerely says tht any two planes through O meetin atine trough 0. Eguvaleatly we could desare that, by defition, the Hines ofthe projective plane are the great circles ona sphere, any two of which meetin a pair of [ntipodal points. Then the points ofthe projective plane aze the pais of Sntipodal points, abstractly identised, This aberact idenicaion was ‘avidly deseribed by H.G. Wel is his short sory, The Remarkable Case of Daviion’s Eye, (A sodden catstrophe distorted Davidson eld of vision to that he saw everything asi would have appeated from an exaetly antip= ‘dal position on the earth) ‘When the inversive planes derived from the sphere by sereograpic pr jection, distances are inevlably distorted, but the angle at which two circles Intersects preserved In ths eas, the invesive plane has apart metric: angles ate measured inthe usual way, but distances are never mentioned [Graustein 1, pp. 377,388,395) ‘On the other hand, gnomonic projection enables us, we wish to give the projective plane a comple mete. The distance between two points P and ‘Qima (Figure 6) i defined to be the angle POQ Gn radian measure). and ‘he angle between two lines m and nin os defined tobe the angle between the planes Omsand On. (This agres withthe customary measurement of dis- ‘ances and angles on a sphere, st wedi spherical trigonometry) We have thus obtained the lips lane or, more pessely, the real projective plane with an elliptic metre [Coxeter 3, Chapter VI; E.T. Bell 2, pp. 302-3115, Bachmann 1p. 21 ‘Since the pons ofthe elliptic plane ae in one-o-wo correspondence with ‘he points ofthe unit sphere, whose lotal are x4, it follows that the total area ofthe elptie plane according othe most natural definition of “area") Js 2e. Likewise, the total length of line (represented by a" title") is =.” The simpleton that result from wsing the eliptic plane in Stead ofthe sphere is well ustated bythe problem of computing the area of ‘Spherical langle 4BC, whore sides are are of three great ctcles. Figure 6 3e shows these great circles, st instereographic projection and thea in fgnomonic projection. The eliptic plane is decompased, by the three lines 58C, C4, 48, nto four wiangula regions. One of them ithe given wiangle ‘3 with angles 4, B,C; the other three are masked a, 8, yin Figure 6.9. (On the sphere, we have, ofcourse, nt only four regions but eight) The two The sane “lpia mola ss timp aye comecon id te oe abla ap, tony ater rece ai Anal cane ee an ie Spyies stan mp crt mpi Atlpst a eh Sitges dcr apr og tno lines C4, 42 decompose the plane into two lunes whose areas, being po: portional fo the supplementary angles 4 and = ~ 4, are exactly 24 and 2a 4). ‘The lune with angle 4 is made up ofthe two regions & and a Hence Similarly A + f= 28 and A + 7 = 2C, Adding these three equations and subtracting Ababpeyad wwe deduce Girard’ “spherical excess” formula 692 AaAdBH CH hich is equally valid for the sphere andthe elliptic plane. (A. Girard, In Yention nowele om algbre, Amsterdam, 1629) 1. “Twositle inthe eipiplane may have a anya four point of ntti, 2. Thearenofap-gon inthe elipic place qual othe exes of Ks angle vam vet the angle sum of pgm the Eehdean plane7 Isometry and similarity in Euclidean space “This chapter isthe three-dmensonsl counterpart of Chapters 3 and 5. In $75 we ind a prof independent of Euclid Fifth Postlate) forthe theorem (aisovered by Michel Chases in 183) that every motion is & ni In $7.6 we see that every similarity (exept the twist and the glide Fefetion, which are hometies) a three-dimensional dlaiverotarion, ‘Ment ometies are falar in everyday if When you walk straight for wand you are undergoing a (ansation. When you turn a comes, is a ot Ton: when you ascend spit saieas, a twist, The transformation that Interchanges ourself ané Your image in an ordinary mirror is a refection, dnd tin cosy to se how you could combine ths With a rotation or a trans Itton tonbuain rotatory reflection ora glide rection respectively 7.1. DIRECT AND OPPOSITE ISOMETRIES giro me rg ort pee rr corgi © fir bay bea ae moan “The axioms of congruence, a sample of which was given in 1.26, ean be extended ina amural manner fom plane geometry to solid geomet. In Space an nomen (Weyl "congruence" sl any transformation that pre- somes Mola that fine Segment PQ is eansformed into a congruent se ‘ment PQ". ‘The mos familiar examples ate the ovation about a given Line {hrovgh given angle and the ranlaion i given diction through a ven tlistance In the former case the exis of rotation has alls pots invariant, inthe lauer thet so ivariant pat, except when the distance is 2et0 50 ‘that the translation is the emit A reflections the special kind of som ‘ry which as a whole plane of invaciant point: the mirror. By a sim- ple argument involving three spheres instead of two circles, we ean easily prove the following analogue of Theorem 231 7A If an tomery has ree now-colinear nariant pols, it must be elther the idemity ova reection When two tetrahedra ABCP, ABCP’ are images ach other by reflection {in their common face, we may spard the “broken line” formed bythe tee ‘edges 4B, BC, CP oss Kind ofradimentary screw ad the image formed by ‘AB, BC, CP’ as an oppositely oriented stew fone is right-handed the other iplefthanded. A models easily made from to piees of sti wie, with Incangled bends at and C. In ths manne the idea of sense ca be eX tended from two dimensions to thee: we can say whether two given con frucat tetrahedra agree or disagrec in sense. Ia the former case we shal, find that either tetrahedron cane moved ike a sere ins ut) the posi- ‘on previously occupied by the other: such a motion is alled a wis. This distinction arises in analytic geometry when we make a eoosdinate twansformation. If O isthe ongin and X.Y, Z are at nit distances along the positive coordinate ates the sense ofthe teteahedson OXYZ determines ‘whether the system of axes right-handed or lefthanded. (A coordinate transformation determines an isometry transforming each point, y2) nto "he point that has the sume coordinates inthe new stem.) ‘Ince an iometry i determined by its effect ona te 7.2 Any owo congruent tetrahedra ABCD, A'B'CD! are related by @ nique tomery ABCD» A°R°CD, whch is dre! or opposite according as the sense of 4°B'CD agrets or dsagrees with that of ABCD. (Some authors, such as Weyl, say “proper or improper” instead of direct ‘oropposite”) “The slid analogue of Theorem 3.12 is easily seen tobe: 7.13. Twosiven congruent triangles are related by just wo isometries: one lire end one oppetit. [Asa counterpart for 3.13 we have [Coxeter 1, p36: 74 Eeryometry isthe product of at mos fur refetons, If here isa aria pont, "orca be replaced by “thre.” Since reflection reverses sense, an isomety is director opposite accord ig as isthe product of an even or odd numberof telctions: 2 or # ia the former ease 1 or3 inthe latter. In particular a dzect isomety with ron,‘an invariant point isthe product of just two reflections, and since the two Inisrors have common point they have common line. Hence AS Every diet isometry with am invariant pon 4 rovation. [Also, as Euler observed in 1776, 7S The product oftwo rotations about lines hough a point Os another suck rotation exercise ‘he prot of rotation trough aot wo lnteectng ines tt orm an apaew isa route tough 7.2 THE CENTRAL INVERSION (One ofthe most important opposite isometries sth cemral inversion (or refeton in poi”), which transforms each point Pinto the poi P Tor this the midpoint of PP" is fixed point O. This can e described asthe product of reflections in any three motealy perpendicular planes through O. ‘Faking these thee mirrors tobe the coordinate planes x = 0,) = ‘wesc thatthe central inversion inthe origin transforms each point (2) into (—,—5.~2). “The name “cental inversion.” though wel established in the literature of crystallography, perhaps unfortunate: we must be careful to distinguish it from inversion in sphere, For most purposes the ceatral inversion plays the same role in three d= mensions as the halturn i two. But we must remember that, since 3 ‘in odd number, the central inversion isan opposite sometry whereas the hatetorn is diet dn space, the name halfturn i naturally used forthe rotation through w about a line (or the “election in a line), which ssl iret [Lamb yp. 9 Whats the produ of falfurs abou ths uty perpendcaar ines ough 7.3 ROTATION AND TRANSLATION ‘The treatment of translation in §32 can be adapied to three dimensions by defining «translation asthe product of two cental inversions. Wesoon Seethat ether the fist center or the second aay be arbitrarily assigned, and thatthe two saversions maybe placed by two hal-trns about parallel wxes ‘by two reflection in parallel mirors ‘Thus the prodvct of two relectons is either a translation ora rotation ‘The later arises when the two mirrors intersect in a line, the axis ofthe ro {ati Ta particular, the product of rections in wo perpendicular mierors ‘sahaletur, ‘The product of reflections in two plans through a line being a rotation bout is the same asthe product of reflections in two other planes through ‘Imaking the same dinedral ange as the given planes (inthe same sense) ‘Similar, the product ef reflections in two parallel planes, being 4 tans tion, the same asthe product of reflections in tvo other planes parallel to the given planes and having the same distance apart. 7A WE PRODUCT OF THREE REFLECTIONS ‘The tice simplest kinds of isometry, namely rotation, translation and e- Acction, combine in commutative pats to form the mist (or “screw ds Placemeat”) glide ejecuon and rotatory reflection. A twists the product of 2 olation wi a tanslatin along the direction ofthe axis. A ghde recon isthe product of a reflection with «translation along the direetion of line Tying the miro, thats, the produet of elections a thee planes of which {eo are parallel while the thirds perpendicular to both. A rotatory rete toni the product ofa reflection with a rotation whose axis perpendicular to the miror. When this rotation i halltura, the rotatory rlletion re- ‘duces to central inversion “Any rotatory refetion can be analysed into a central inversion and ae sidual rotation. For i the rotation involved in the rotatory reflection is Fotaton through 8, we may regard it asthe product ofa half-turn anda ro {ation through # + =(or — x). Thusa rotatory reflection can just as well be called a rotatory inversion: the product ofa central inversion and a rot tion whose auis pases through the cents. "Aay opposite isometry T that has an invariant point Os ether a single reflection or the product of reflections in thre planes through 0. It prod ‘uct TT with the Cetra inversion in O, being # dec isometry with an in- ‘arian! pat, i simply a rotation S about aie through O. Hence the given ‘opposite isometry isthe rotatory inversion Si =k 7.AL Every opposite ometry with an imarian point ia rotatory inversion Since thre planes that have no common pot areal perpendiclaro one plane the refections in them as applied toa point ina) behave lik the ce- fection inthe Hines that are thei seeions by e. Thus we can make se of ‘Theorem 331 and conclude that72 Every opposite omer with no invariant pol i a lide reflection 1. Whats the product of eesti te plas thou? 2 LetaC and 477 be wo congrient iizgsin isin plans, Conse the panda toro 44 BCCI aes ent oncom ito tans te ltd y say iverson ih emer. in [Ryne le hy aoa, he he plans wold imenet it ie) 5. "vey oppose nome is expres othe rod of efecto ad al 7s Twist The only remaining possiblity i dies isometry with no invariant point Lats be any tect isometry (with or without an invariant poind,tansfor fag amaritary point 4 into“ Let Ry be the refetion that interchanges A nd Then the product RyS isan opposite isometry leaving 4” variant By 7a, this rotatory inversion or rotatory reflection RaRGRa, the prod fit of totation Rey and 4 reostion Ry, the mietor for Ry being perpen Sular the axis for ReRy._ Since this fotation may De expressed 2s the ‘roduc of two reflestons in various ways (73), we ean adjust the mirrors for Reand Re 50 as to make the Former perpendicular to the miro for Ry Since both these plans remain perpendicular t the mirror Tor Re, we nO hve 5 = RRR the product ofthe two rotations RyRe, RoR Both of which are half-twens [vehlen and Young 2, p. 318) 71 Every dec isometry is expressible as the produc of v0 half ration, which may be ex- Ir the isometry has an invariant points a cotation, y eae lines. When there is no invariant point, the axes ofthe two halF-tuns are either parallel, in which case the producti a tanslation, oF skew, like two ‘opposite edges ofa tetrahedron. Two skew lines always lie napa of paral Iciplanes, namely, the plane through each ine parallel othe oer Since hull-turn is the product of reflections in any two perpendicul planes through is aus the two half-turs RGR, RaR with skew anes are re- poctively equal to R's, RR’, where the mirrors for Ry and R's ate par ill whe the other two ate perpendicular to them (Figure 75a). Hence RARARSRy RARER ‘where the interchange ofthe mide esctions is posible since the altura RaR’s may be equally well expresed as R'3R', We have now fulled our purpose of expressing the general direct isometry as @ ws. the product, GF the rotation RyRy and ihe anslation RR’, along the axis of the ot tion. (This aais meets both the skew ines at viht angles, and therefore retsures the sbortest distance Between them) In other words, 7.82 Every placement is elter a votation ora translation ora wis. (For an alternative treatment se Thomson and Tait [, § 102}) RORRRs 1. What And of sme tasforms the poi 2 eto (Go ree ere + 0 WCreren went enna 2 The prot flu about vo she hs ight angles satis ame, the poco lta abou the line short tenes anda wml hu ‘wim thi sovet tance, (bln nd Young [p38] med tes hala) 7.8 DILATIVE ROTATION Icon be rove by emetic Eo say etn Cevoe [19.21] In Euclidean space, the definition of dlazation exactly the same asin the plane, In fact, 5. ean be applied, word for word, othe dimensions, ex opt that the special dlstation AB -» Bd or O(~1) ie ot a halleturn bat & central inversion (7.2). Likewise, §5.2.appies co spheres just aswell as to citles: Figure 5.22 may be regarded as plane section of two unequs spheres wit their centers C, Cand their centers of simlitude 0, Os. Two ‘ual spheres are related by a vanslation and by a ental inversion. ‘oowever, an important diference appears when we consider questions of sense, In the plane, every dilatation ir diect, but in space the disation (00) isrect or opposite according as Ais positive or negative: fr example,vee |sowey AND SARIY N EUCCEAN SPACE cena inversion O(— is oposite, 3 ws have se. oe eee lac two la urs ae elated y a lary. wesley bean nomet oa dttion. BY amatral ek TESA is tainly we ae tik late rotarion o mean he prod ter Goon abou ne av) and ination whose center Oe reese aae geoueh © perpen isinvaran. bing wansformed aa rane re eEemural cola ation” of #35. Inthe spa according on abut vu hls, there te infil any ote aaah planes through Any sash plas es formed acorn 3 dative eeston Fane tan te prod of «oan though aage @ "saan 8 whee is om he 5) The flloning aes oa IU opel ess mc are aie aleern Retecton |We observe that this table inclaes sil Kinds of isometry, both direst and op este except the fansltion, twist and glide rection (hich have no in- Pesan pot). Stl more surprisagls we shall ind that, wit these same Uhree exceptions, every sina) 5a dave rotation "Fhe role of similat triangles Is now taken every similar tetrahedra. Evidently 17.61 Two shen similar tetrahedra ABCD, A'B'CD’ are related by @ unigtesinitariy ABCD + A'B'C'D, which director opposite according as Uhesense of 4 BCD’ agree ov disagrees with ha of ABCD. ‘nother words a similasity is completely determined by itseffect on any four given non-coplanar points and we have the following generalization of ‘Theorem 7.13 7.42. Two given similar triangles ABC, A'B'C are eelated by just m0 Limiteriies; one direct and one opposite rs a stp towards proving that every similarity which s nt an isometry fs dilative rotation, et us frst prove 7.63. Every smilariy which not a isometry has just ome ivariant point Consider any gen simaiy 5, whose rato of magnication i x #1 Let ramform an array pon tot IC umd wih A weave the dened invariant pn ‘Irno let Qe the pa tht vides te sexe sme 44 in th rap. extra or teal aooring at Sector ppt! hat constr Qo that Qe" = QA. Let D dnote diet or oponite ison Qk=4~) Then SD, having rato of mapiesion iva ect bomety whch eaves A ivan By Theorem 712,30 ao. thsion about some line ough A. The plane though perpen to unformed nt ti y oth SD and D-and therefore 9 thet Product On thi sven plans Side womens snr ick by Theorem 342-hasanineviant pin Fauna pin ivunige fori tere were we dnt invert pints the segment fed Uy tan would bev! intend of bing maloped by Having found the invariant point (or center) 0, we can caey oat asim plifled version ofthe above procedure, with O for 4. Since 4” and Q both foincide with O, Sis the product ofa rotation about a line trough O and {he dilatation Of) that i to sy, Sis a dltive rotation 746A Every simlay either an lsorery ora dave rotation, In other words, every similarity is either a translation, a twist a gid re- fection, ofa dave rotation, provided we tepard the lst possibilty as in- cluding all he special eases tabulated in the mide of page 102 By Theorem 7.62, there ae to dative rotations one direct and one op posite, which wil transform a given triangle ABC iato similar (but not ongruen)tranle 4'B°C. The rai of magnification, p71, given by {he equation 4° = AB. Let Ay and Ay divide 44” Internally and ex tecnaly inthe ratio 1 Let By and BC, and Cz divide BBY, CC ia the Same manner. Consider the three spheres whose diameters are 4s. BBs, GiCz._ These are “spheres of Appllonis” (Theorem 6.81; for example, the fits the lous of poins whose distances fom A and 4” are inthe aie Tq. Any poit 0 for which 4’ = 04, Of = nOB, OC = noc ‘must leon all three spheres. We have already established the existence of| two such points, Hence the centers ofthe two dilative rotations may be ‘onstruted as the point of intersection of these three spheres. Jn $3.7 we used translation to generate a geometric representation ofthe {nite eye group C. (which is the fre group with one generator). We See now thatthe sae abstract group has amore interesting representation in ‘which the generator isa diative rotation, Some thirty elements ofthis group ‘an be seen in the Nouri shell (Thompson 2, p. 843, Figure 418}cis 1. Hows the point (2) ttsormed bythe gener aie rotation whos cen retand wus athe on and the 2a 1 Find the aad angi fo the dave tation x s)-otes meh 14 How docs the above clastaon of sais dal wid he hrimensiona slasne recon the prado eitaton OG) a the reese 8 plane thea en Gould Theorem 763 be proved i the manne fhe exci a hee oS (ome BP 17.7 SPHERE-PRESERVING TRANSFORMATIONS “The reasoning used in § 67 extends readily from two to thre dimensions, delding the following analog® of Theorem 6.7 AV Every sphere-presersing transformation of nversve spaces either simian or the product fan nsraon (ia sphere) and an some very aprepeserving ansforatin san Be ened a the roduc of ree vom and terion where + none azang, Pr rsa ii ofme Cai ens 2 Ga vit pe a Part Il8 Coordinates Inthe preceding chapters, few exercises on coordinates have been serted forthe sake of those readers who are already aquainted with analytic geometry. Other readers, having omited such exercises, are avaling en Hightenment at the presen sage In addition to the usual rectangular Car- tesian coordinates, we shall consider oblique and polar coordinates. (The polar equation for an elise is important because of suse i the theory of forbs) After a brief mention of special curves we shall give an outline of Newton's application of calculus to problems of are length and area. The Section on three-imensional space culminates i a surprising property of the dovghnetshaped fort 8.1 CARTESIAN COORDINATES. ‘Tough he end a hid inp, et he mead oF sip por it my ede apo eto Ine Get geome fci, Acne nd Alan rea gaa.9) (er ete 2) ‘Analytic geometry may be described asthe representation of the points in ndimensional space by ordered sets ofm (or more) numbers called co- ‘ordinates. For instance, any position on the earth can be species by its tude, longitude, and height above sea level ‘The one-dimensional cae is well illustrated by thermometer. There is certain point on the line associated withthe numbe: 0; the positive int {ers 1,23,» are evenly spaced in one direction away from 0, the nega tive integers ~1, ~2, —3, in the opposite duccton, andthe fractional ‘numbers ae interpolated in the natural manner, The displacement frm ‘one point» to another point» isthe postive or negative number x” — >.In the two-dimensional ease, the position of point in a plane may be specified by its distances from two fixed perpendicular ines the axes. This TBitoncar be waved back over wo thousand years to Archimedes of Sy ‘hae and Apollonius of Pere, or even to the ancient Egyptians: but it was fie developed systematically by two Frenchmen: Pitre Fermat (whose problem about a thangle we solved in § 18) and René Descartes (1896-1650), ThuhcirTormalaton the two distances were taken to be postive o sro, ‘The [important idea of allowing one o both toe negative was supplied by Sit Tune Newton (1642-1727), and st was G. W. Leibniz (1646-1716) who fist Ug them “cooedinates™ (The Germans write Koordinaren the French fverdonnéee) For some purposes itis just a easy t use oblique aes, as inthe second part of Figure Sle. Starting from the origin O, where the ax intersect Re rcach the general point (x9) By going a distance x along the x-axis OX Md then detance j slong line parallel to the yanis OY. ‘The x-axis is id wave the equation O Because every pot (x0 sais this equ ton sila =O isthe equation of the-aNs. On any other line through {he ongin, consideration of Romthetic angles shows that the ratio / 1s eaetanty thus any line through the origin (0,0) may be expressed as ax} by =O, “Tobia the equation for any other ine, we take a point (x,y) ont. In terms of ew eoordinates derived by translating the origin rom (0,0) tetra) the ine may be expressed as ax’ + by" = 0. Since x’ = x — x Sidy’ yy the same lines rms ofthe orignal eorsinates as) yD onsay, an axtbyte “Thus every line has lneer equation, and every linear equation determines ine Taparticla,teline that make atercepspand yon the axes °. ry pra fortis equation is linear andi satis by both (9, 0) and.) Two lines ofthe form 811 are pall if they have the ste rato” includ ing. atone port, b= 0 To both i which ese they ae parallel the rau) The pat of ntrestion of two nonparalll ines i obtained by fon the wo smallaneos equations for ad iB 7, the equation 81 _may be solved for y in the form y= {ox +6), More general, points whose cordinats satis an equation Foss) = ory ~ fs) ean be pated by giving convenient values othe heii x and calculating the cortesponding valves of the ordinate 9. This frocedize ik parclarly appropiate when (x) sa onetalued fuscton of re nother eases we may peter four pate equations, expressing 2nd as unctons of ingle variable (or parame) 1. For iastanee i Py Aenotes the point (9) ay line through Phas params equations aaa eam eke yond Me whore X and Y depend on the direction ofthe line ‘Sometnes, for the sake of symmetry, the single parameter i replaced by two parameter; and, slated by an auxiliary equation. For instance, she Eel poe 9) om the ine hough to en pois sand Ps L KS AG tite YEAH Me th ‘This pont viding the segment Pai he aif: 1,5 he cena (or “Cena of gravity”) of masses at Py andra Poston outside thea terval fiom Py here ta = 0) to Phere t= O)aecovered by allowing ta ot be negative wile stl ating ta ~ Ine may july By iling them else charges” estead oases” For problems involving te dstanee Between wo point or the ange Be- tween two line it often advisable to ue rectangular aes that hed {nce from the origin to)) the square otf 2 + andthe tance Aaistesquareraet Gan + Oy Multiplication of the expression. ables ust normalize the equation Writing = Oi the form fax + by + chy a suitable number en ofthe general ines that a2 +58 = 1 (eam + aly +O — y+ 20 fe £0 — 9h, where = ast + By +6 weresognz it she locus of points gbiitant from me . (ey = Zaps = 288) and 99in other words, the line /= 0 serves asa mitror which interchanges these two points by reflection. I follows tha the fot ofthe perpendicular from Pst (s(x = a yx — bp), and that the distance from Ps tothe lines, (provided at + 6° 1). In particular, the distance from the origin to! = 0 ‘The locus of points at unit distance from the origin isthe circle fay hich has the parametric equations 088, y=sind or, with = tang, e 2» tee OPTS Inter of ener Cartesian oot, the point(s) wil e tans ao oy) by tenant) (cn) bythediaton 05, C0) batman mee ai. tats of etapa Cartesian corde he pt (willbe asad (ey sense 4 lsaadnt empathy ay tp PP Phen lite i as sine 8.2 POLAR cooRDiNaTEs MIA Newman 7.) (Wore ante 4895 9.19) For problems involving direction from a xed origin (or “pole") 0, we ‘often find it convenieat to specify a point P by its polar coordinate (8), Whete rs the distance OP and @s the angle that the direction OP makes with a gven inital fine OX, which may be idemtied with the axis Of ro. ‘angular Cartesian coordinates. Ofcourse, the point (is the same as (0 dno) for any integer mIt is sometimes desirable to allow rto Be ‘egatve 80 that (8) the Same as (— Given the Cartesian equation for a curve, we can deduce the polar equa tion forthe same curve by substituting a x= reo8, ya rsint For instance, he wait cle +92 = I has the pola equation (eos + (sin whic reduces to (The positive value of ris sufceat if we allow # to take all valves from —= tororfrom0to 2x) This procedure helpful in elemeataryteigonomety tvhere students often experience some dificult in proving and remember ing) the trgonometrieal functions of obtuse and larger angles. Taking an angle XOP with OP = 1, we can simply defines cosine and sine t be the absciss and ordinate of P. Polar coordinates are particulary suitable for describing those isometries (635) and sinlarties (154) which have an invariant point for this point ‘may be used asthe origin. Thus the general point (8) willbe transformed (0-4) by a rotation through a (2640) byahaltior, (= 8) by reflection in the inital (2a 8) byelection inthe line 8 Gir) by the dilatation O40, (ir 8 +) byadilative rotation wth center (jr, 2a'—8) bya dave refestion with center O and axis® = a Likewise inversion in the cree 7 = k (se § 6.1) wil transform (8) into err, 8. “The Cartesian expressions forthe stme transformations can be deduced at once, For instance, the rotation through w about O transforms (8, )) inno (9 where, by 821 (cos cosa — sin Pine) = xcose~ ysina (cos@sina + sineosa)=xsina + yeose. sin 0-4 @) ‘In particular, a quarter-turm transforms (x, ») nto (~), x) andi follows that a necessary ad sulicient condition for two points (9) and (x,y) lien perpendicular dretions Irom the vigin is ona wi ty at Svch a transformation as ye has two distinct atpecs: an “active” oF ali aspect, in which each point53) is moved to & new postion (x), and a “passive” or alas aspect, “which the point previously named (x) is renamed (x,y). The latter spect is sometimes ured to simply the equation of given curve, For i Manes, he eure a phy + b= becomes a(x cosa —ysino)t + 2h(xeosa~ ysina) (xsina + y cosa) "y blxsina + cosa}! =I in whch the cefcent of i ao longer 2h but 2h(ost a — sin? a) — 2a — B)cos asin a = 2hcos 2a —(a ~ 6) sin 2a ‘Since this vanishes when tan 2n = 2A/(a ~ B), the equation is simplified by rotating the axes through the particular angle tan 2h = Harcian 2 ‘The area ofa triangle OF Ps, where P has pola coordinates (r,s taken to be postive i) < 8, negative if @, > @s. With this convention, the area is frresin(@s — 00) ‘or in Cartesian coordinates, ridin cos, — 0s in) eae fous — x09) +) “Toyind the area of any triangle Py PsPs, we choose new axes parallel to OX, OY and pasing through Pa Sie the new coordinates of (i= Tor 2) Modi Jah the area of any tnangle PLP of pe Iu follows that a necessary and suicient condition for Py, Pa, Ps 0 be col Tincar in tha this three sowed determinant should be zero. "The equation forthe ine PyPs aay be derived from this condition by writing (x, for Gs) 1. Uscawatsinown wigonomtric omit ohainan expression for the quae ‘tthe utance between te pons wove polar cordate ar 8) 8) 12 Obtain polar cordate or the mio | Obtains ple uation for the ine We ves) “4 Use £22 obtain the conn for wo lines cx + by +6 = Vand ae + By + = O1w be perpendicular, Deduce tame. (Hin: Allow wake eg 'S Uses ute rtsion of ates to imply the eatin of he curve 48 + thy + IES 83 THE ciRCLE ufo (8e ‘The circle with centr y’)and rads k, being the locus of points (x) distane & from (x9) is GF tony Rak Thus sar Bayt tert Yreno isa circle with center (=, —f) whenever & + f2 > 6 H(xn) les on the cic, the tangent at this poiat Py ax tyr teeta) tfotmte=0 or Greer Ore Py + Gn +p 9 =0. For this ine pases through Py and is perpendicular tothe diameter zee ye Mts ety “The circle 831 is orthogonal to another irle Mbyte des + Yy ten 0 if fora suitable P, the center ofeach lies on the tangent at Pt the ater ‘Adding (tO UF set pi te to the analogous telation with primed ead unprimed ltrs intrchanged, ‘we see that the orthogonality ofthe two crcles ilies ew +2"Conversely, any two circles that satisfy this flation are orthogonal. In pat- ticular, the cles Maye terte aye -e= 0, whose centers le the a= and y-axes respectively, are orthogonal. Keep Thee constant and allowing 0/10 tke Various values, we obtain two or {hogonal pencils of eoaxal tle, whose radical axes arex = Oand y= 0 seeetinay. Ife = 0. we have two orthogonal fangen! pene, each con- {iptng ofall the civles that touch one ofthe axes a the origin. We > 0, {he eces 832, for various values of g, form a nonintersecting pencil lung the two point etles (eter tee. hich are the limiting points (= y¢,0) ofthe pen The ctles 8.3, which passthrough these two points, frm the orthogonal nrersecing pen. vianerses 1 iave x» ato (Sir a) apy his inversion to he ie 1 ad he circ 3. Find he lacs of» point) wove stances rom (0) a aD) ain tertol #860, *LOotaie he Cartesian equation ofthe cas fa pin th pdt of hse di tance om (0.0) and) eas, Deda he pla eaton oft "Base of Seb! wh hema flaca Bel “tC wo xual cides in contact. rd he ous the verte finger woh th ast ine post cle (12, be second amend (615) (Aner ‘nemscte*) Acinic ofradus lle witout iting on te ouside of fed cite rads ‘ne The lcs of pot Bed on fe dreumferece of the oling ce ald an ‘piel (when nivon integer, sx enspedeieycoi), Obtain he parametric eit e=tvs = tn ¢ bcos heawte + th Ja lee Din tain De (the card n = 2 (he mpl w = 3, andin = 3 See Shc the cases Rotaoa sp. 68] {© Shing theorist cusp (6. obtain she psa easton F=2 (1 eos) 1). eda that cords hough the cup ae of 0 forthe cad (834 with © 2nd on rh wig on eile of fd ois rk were n >I. Find parametric equation forthe hypo eoid (when mineral ‘pe meuped bypyled nich rhe oes o's pit Heed onthe ccamerence of {Beroling cite, Sketch the caes n= 2 (vain spring) n= 3 be ato, nd n= be exo iat he praeterin the at wo cae, tain. oe » [amb 2 pp 297-031 “Stine covered tha all he Simon ns for ay given ange touch det ‘Te ofthe ies, aumely those paral the sider of Merle equate wangle (819) ae the “spel tangent he Jl shes wih the mine pnt sie Tei pins of omc ete vee ofthe euler tangle 172 desrbe EA 3 on pe 20 For dea ee Bakes [1p 30-39, ep. 37) cones Ineo the salght ny ii, ple anaes it whch troy tel c's foi te Gross ere opera Socewesge by athe ov pment parable Colyton by aa of arene oa. “Sen ann red woke ne Le Span 97- (Sar 2.22) “There are several different ways to define a conic (or “cone section”). One ofthe most straightforward i the following (ef. $6.6): A con isthetoces of «point P whose distance OP from xsd poiat Os «times its dis tance PK from a fixed line HX (Figure 84), where ei postive constant ‘Other definitions for & onic, proposed by Menucchmus about 40 8. were reconciled with tis one by Pappos of Alexandria (oust century AD) ‘or posily by Enid see Coolidge 1, pp. 9-13], The conic i called am elipeife < Ysa parabola ie = 1a hyperbola it > 1. (These names are due to Apollonius) “The point O and the lie HX are called a focus and the corresponding directrix. The avo «called the ecenricit, is usually enoted by e (but then, to avoid aay possible misunderstanding we should add "where ¢ need hotbe the base ofthe natural logarithms” Litlewood 1p.) The chord ‘LL’ through the focus, parallel to the dzetris is alld the fru rectum: {ts length noted by 21 so that Ds OL = ett In terms of polar coordinates wi dlvctrx, we have 1 = OP = PK = LH ~ eos) the initial ine OX perpendicular to the oa a1 ercose sothat aan Fat ecoss, Since tis equation is unchanged when we replace Bby — retrical by reflection n the ital line. When 0 =O, whend = 2, except when € = I Wee 1, 842 makes rSnite and positive for all values of; therefore the clips is cloed (oval) curve” Ie = Ir is sil nite wad positive except ‘when #'= 5; therefore the parabola i nt closed but extends to infinity in fone direction, Ife > I, ris postive or negative aecording as cos is reater ‘rles than ~1/; therefore the hyperbola consists of two Separate branches, tBven by the conic sym: Vl + saad U(U ~ a: therefore the comic meets the ntl ine twice Ha<0 1 In the later case the above definition makes «negative, but we ean reverse ity sign without altering the equation 88. A mote important remark is thatthe equation is sil unchanged when we revere the ign af x ory. ThiS favariance shows thatthe ellipse and hyperbola ae symmetrical by eec- tion in either ani, and therefore also by the hall-turn abou he origin thei symmetry group is Di the notation of $2. For ths reason the origin ‘sealed the center and the elipse and hyperbola are called cenral comes 4 “The geometrical significance of @ and is indicated in Figure 84. For the lips, 2a and 2b are the major anid mior axes or the hyperbol, they te the imserse and conjugate axes The two branches of the byperbola xe endl ae lie in two opposite angular regions formed by the two fines + BeBee «FH‘These lines are called the asymprotes of the hyperbola. Ifa = 6, they are perpendicular, and we havea rectangular (oe “equilateral hyperbola. The = I, then 843 reduces yeudi-0 or, by refetion in the ine x = J eas yeas “This i the stand ‘equation for the parabola (Figure 8.40 “The most convenient parametric equations ae: forthe ellipse 246 x= aos y= bsint forthe parabola ear xed, y= and forthe hyperbola cosh, = bain 1 eter z (These functions will be discussed in $8.6) exenaises 1. What hind of uve ha he polar equation radio yer 2 What bind ofcure has the Carin ution (See Ex Sate ead of 112) "The sm (or erence) ofthe kinases oa pont o an elise or hyperbola) ‘the escent ofa rectanplr hyperbola? 5 Given pont Band C he oes of he vets 4 fa tngle ABC whose Ele Une parte to BC (aim Ex 93 the en of HL an lips whove nr as [AC wheter ans ei the ltde ofthe equate wrngle om BC (Hn It: Catecerrandt 0. the csmeeter euisan fom and (0490) incall tna quad form, 1 i oe fine b> 0 at FP the hie sg forall vale of andy excep x =) = 0 Is sad tbe poe definite perfec squat: poe seme fa > 0,50 ha sei 2 pct sae: nl Tied Beso ha Fis postive for some values ots and cea footer “Thc cuuation FU represent amelie Fis poaiv dente ptf para ines ‘scrote he a though engl ‘8 Dessite «geome interpretation forthe parte sin A [Hit Cm parame bin wih coe 08) 9 Imhutceapect the hyperbola NA more satiactory represented bythe than bythe quan 8487 1. The Gl = inserts he conic 842 ot the imagen ‘Sketch this curve for various ales of, When « = (0 that the conics para shia 1M. Theiler = a nsrs the restau hyperol = 3 6 nt the em (ee Bx 3 athe end of 83), 8.5. TANGENT, ARC LENGTH, AND AREA tm now ante eng tar ew at l (renters Manin of Nn, al 2. ha.) “Thecurves with which we shal be concerned are reeuifable” that, there is a well-defined arc length + between any two points P and Q on tach acurve, Using the temporary notation Po = P. Py = Q, we subdivide the Gvemare POby n — I points Py, Pay.» Py-s and consider the last upper bound n(n) ofthe lengths ofthe broken Hines PPL Pa tot Pas for all posible subdivisions |, ak ‘The cure is often ucfullyregutded asthe locus of a “moving” point. [Any two points Pand Pon the curve ate joined bya line called scant IVP is fxed while P' moves the secant wsualy approaches Limiting pos tion which is called the fangen’ at P-- When we use rectangular Cartesian Coordinates we draw PPM parallel fo the vans asin Figure 8.54, and let N be the foot of the perpendicular from to P'M.. The tangent has an ingle of slope. which may be defined asthe init of £ NPP". (The Bgure an be modified in an Obvious manner when this angle is obtuse oF exe he) ‘Tocompute we consider the ight-angled langle PP'N whose ses fare the “increments” of x). (all tending 0 210) Ae = PN, ay = NP, As = PP Thus . y= tim PR = tim 2 cos} = lim pp = fim A= ae sing = tim NP = tim 2 mw NP tim Pw = 3” as Since PP? = PN? + NPY (or since cost + sin Y = 1) the element of sr length de is given by ost it = det 4 bs, and the arcengths from (, 1) 0 (ss.92) oF fom = f to = fi safof te Gits-[1e ait awe 288 When a curve is given by its polar equation, the direction ofthe tangent Pisdetermined ether bythe angle which this tangent makes with the “rads” OP or (equally well) by he angle enone which it makes with the intl lie OX (Figure 8.5. As before, lt Pand be two neighboring points o the curve, so tha the tangeat at Ps the limiting postion ofthe secant PZ. Draw PN perpendicular to OP? [Lamb 2.284), Then 2 cos § = tim ME emer lim NP Pe 3 tang = tim ae Since PP* = NP* + NP® (or since cost $+ sint ¢ = 1), the element Tength drs now given by ase de = de + an, of.Since the area ofthe thin triangle OPP" difers by a second-order in itesimal from that of aetclar sector of radius and ange 3#. which is {}P1A0 the are of any closed curve surrounding the origin just once is 5 uf ea Such a formula can be translated into tm of Cartesian coordinates by means ofthe rlations 821, which imply ase de sothatxdy — yds = reossdy — rsin Bd and cm afeas fownve “This mus, of couse, be interpreted 3s a i fea- where x andy ate given in terms of parameter and the integration is Over the values of that take us all round the curve oh {rcosd—rsin Bal, dy = desind + rosa, os + sin Bydd = a Pere be The same formula can be used 10 comput the area ofthe “sector” ob tained by joining the origin toa given arc (Figure 88 or d), In polar co ‘ordinates, ifthe are goes fiom 8 = 8, 10.0 = By the area is sf ew When we transfer this to Cartesian coordinates, we regard the boundary of the sector ara loved “curve” consisting ofthe ave and the two radi. Since ds the rai (along which» remains constant) make a zero contribution to the integral onthe right side of 857. Hence, ithe ar gos from f = 1,10 Frag the area ofthe sectors aso fle [of Courant 1, p. 273, 1. Theline x — (4 4) 4 2h = Ona scat of the parabola 847, meting it inte pons shove peas se and. Maing Ftd fy, dee the equation {xh gn ce i wn preter a) eo Be pa nya kes a 2. Theline ova + Paina = oo 8 is secant ofthe ep 846, meting in the pit deduce the gusto =f Mating 8 tendo for te tangent a the point hoe parametric Robson 1s p.274) Obtain al te eas for he hyper bah Dedue that if (bes onthe central onic a, he agen ths pus ‘A. Atte point onthe line 45, he orn sing perpendicular the ange. i ierenting pray wit eget to «and then limiting, Main he erelope a onde 3, pp. 3637; Lamb 2.380) Hin8.6 HYPERBOLIC FUNCTIONS The hyper sin and oie have prope in freee 1 Aheretenguar perl xa elgou ft of he she lcm with fro he cor an he former incon red hyperbole are ale cer futons FW, oto (186-1938) {Hobson pp. 32930) As avery simple application ofthe formula 88, consider the unit wea yee lor (Figure 850). Since and © x cons msine=—y and Ba the tea ofthe sector from f= Oo anyother values #) an i[orene which, ofcourse, we knew already. More interestingly (Figure 8S), if he Surv isthe rectangular hyperbola x? — y2 = 1 or x= enh ty sinh sothat B mcosnt =x y and Ya cone =s, the area ofthe sectors agin sf brd) « a fant ‘Comparing the above results, we se clay the analogy that relates the cireular and hyperbolic functions In Figures 8 ¢ and d. we have a sector “OP ofthe cil or rectangular hyperbola, respectively. In both eases O4 OP fad the parameter fis tice the area ofthe sector.” In the former, OM a eosrand PAT = aint, Inthelatier, OM = cosh rand PM = sin (8 Mr 22 indie arena the eestor a he ener pero 7 THE EQUIANGULAR SPIRAL Inte svg ira oh Non oi nal labia ecm aceauntesthin tego tigre absence oun Si Ae Tag 0-198) [Monon 75575 “Thecitcler = a may e regarded asthe locus of the transform ofthe point (a, 0) by a continuous rotation, which transforms each point (7,0) ito (e+ a) where varies continuously. Similarly, the ay (of half ine) 8 ste locus ofthe transform of (a.0) by a.contnious dilatation, which yields ((.0) for all positive values of 7 By judiciously combining these evo trans Formations, we obiaia a conimous dative rotation. Let denote the ratio of magnifation corresponding to rotation through 1 radian. Then pis the ‘io of magnification fr 2 radians, for 3 radians, a for = radia, for fadians. Ths theilative rotation tansforms the general point (¢.8) into iy @-4 0, whete varies continuously. The locus of the tans- Form of a0) is the equiangular spiral (or “logarithm spiral), whose para metic equations wa 6 may be combined into the single polar equation an raw ‘ese wad hh Tl anaes Tong ge ha ae heme” eget re of ye sven ie song by» Shae io “pecan = a a= seat to he itatioe Ob Stnhous fyven siace = ros spew ashen een te posion cor ant tegen cot = 1 = ope ht hin como a which oul hve een een fom fact hat saron preserve age In ern of is constant angle which is about 80° in Figure 8.7a, we can write : peee aces thos expressing the spiral in its elascal form Feae™* (gad ¢constan, Fa Ha cme, ‘0 that ~ 5084 constant, This shows tha the length ofthe are from Pentwrens (rs, and thatthe length from the origin (¢ = 0) 0 the general point (although {his involves intnitly many turns is, reece. ve geil aman oe by mea fe nation 2 Sek the inven fancied with reget oa cle whose emer Othe siat iva {2 Themage ofthe pil by he aac aout its pole lovers of eqatangla sal ace fseromes) ‘THREE DIMENSIONS rosa oa eben ram heen cine oc ‘Grothe Onmat mers ecm cn ee sae (ep 178) To setup a system of Cartesian coordinates in space we use three axial planes meeting by pairs in tree axes OX, OY, OZ._ Starting from the origin ‘0, we reach the general pot (x2) by going a distance along the x-axis (Of, then a distance pin the direction ofthe y-axis OY, and finally a distance ‘File diecion of the=-ais OZ, The thtee axl planes have the equations wn 0,y = 0,2 — 0, which cam be taken in pairs to determine the axes. For Instance, the z-axis, consisting ofall the points (0, 0,2), has the two equa tions t=) = 0. Any line through the rign (0, 0,0) has parametric equations oa weM, yah rae The mutual atios ofthe coeiiensX, Y, Z determine the direction ofthe Tine. By a translation to new origi we see that the paral line through Girne is mek yar tlh raat Ze ing f, we obtain the two equations kam yom in ¥ Tae hic have tobe interpreted by a special convention when XYZ = 0. The entroid of masses yal (2,323) 4 fs a (3p) 22) mith +f = 1 (ies fot, faye + tase te + fa. “The ovgin and (x,y, 2 ate opposite vertices ofthe poralllpped formes bythe tees pairs of pale planesere 886 Qxem yeQyeys FeO tsar asin Figure 884, (In the tongue-twsting word “parallelepiped” we stress the wllable “ep.” belonging to the Greek prefix ep. which oceurs also in Such words a “epithet” and “epiyelid”) For the rest ofthe present sec tion we shall tke the axes to be mutvally orthogonal so that this isa ec {anguler parallelepiped (or "box"). The three-Jimensional extension of Pythagoras theorem shows thatthe length ofthe diagonal (0.0.0) (rn20 is the square root of 8 4+ + 21%; Similarly, the distance between (Geopezpand (9727s the square rot of VERO VRE. 1 he parameter tn 8.81 adjusted so that ee Marans St measures the distance from the origin to the general point (x,y, 2) the Tine, ‘The coeicente X,Y, Z,sailying £84 are called the direction cosines fof te line. because they are the cosines ofthe anges which the line makes ‘wth the eoordinate axes. More predly, they are the direction cosines of Dre of the two rays into which the line is decomposed bythe origin: the op- posite fay has the direction cosines —X,—¥, 2. Two rays forming an Engle meet the uni sphere watetel in two points, say (X, ¥, 2), (7, PZ), whose coordinates are equal to the ‘irecon cosines of the rays. By "saling” the Isseces triangle which these points form withthe origin, we obtain the expresion AN EY 4 Zz" {or the cosine ofthe angle between the rays. In partiul, the nays (and therefore alto the lines) ae a sit angles sey t zz 1e follows that the plane through the origin perpendicular to the line 8.81 is Xx t Wy + Zs =0. By a translation, we deduce the parallel plane through (3) £9: Nes Yy4z nr Ney 4 Ys + Zen Sill assuming a rereaxry yey ze we deduce that the two supplementary angles between the planes Ax pW e2eaT Keb yy 4 Ze=7 ar the angles whose cosines are HON 4 YY 4 229, We see now that every plane has alnear equation, and every linea equa Son etme alae april, the ane ta maker nese, 97 “Two planes ofthe form 885 ae parallel if thy ile only in thee “constant” termé TTheline of intersection of two nonparallel planes an be reduced {othe standard form 8.83 by eliminating ft sand then. ‘An equation connecting x,y, (not necessarily Hines) usally represents a surfaces two such equations together representa curve, the intersection of two surfaces. In particular, an equation Ray Involving only x and y, represents lind, the locus of line that passes ‘through a variable point onthe curve Fx.) = 2 = Owhile maining paral lelto thes-aXs. Athomogencou equation Sesy2)=0 (hos et sie is merely nulipied by a power ofa when x,y,z re replaces ‘by me, we) represents a cone the loca ofa line tht joie the origin fo variable point on the curve Jey N=O ee Important instances are the quadri yliners 2,including the ordinary cylinder af reoluion (or “ight eccular cylinder I y= A andthe quadrie cones at tbe bet 0 Including the cone of revolution (ot “right circular cone”) x4 4 38 = «2% The equation 1 + 22 = 0, which is satsied only by (0,0, Oh may be regarded either a 4 peculiar kindof cone oF asa sphere of radius 25, The general spre, having center (xy, 2") and radivs ki, ofcourse (8 FOV HERE R SE ‘We observe that thie an equation ofthe seond degree in which the coe- Sentofy2 2" are allen while there are no ters in 228 3 “The sphere x! J! + #1 = A whose coater isthe origin, inverts the point &Y,Z)into Gat BY RE WE PeE! WEEE RSP ‘The plane through this inverse point, perpendicular tothe line 8.81, namely, Neg Wye Zea, iscalled the polar plane of (X, ¥.Z) with respect tothe sphere. 1F(X, ¥,2 Ties inthe sphere, the polar plane s simply the tangent pane | ‘The thre-dimensional analogues ofthe conics are the quai suriees oF i undies, hose plane sections are conics (or ocasionally pairs of lines, ‘Mhich may he regarded as degenerate conic). These surfaces, whose equ fom arc ofthe second degree, include not only the eliptic and hyperbole “xlindes the quedtic cone and the sphere, but also the efpoid atnte the hyperbooid of ne set eee] a atea [R.J.7. Bat 1, p, 149, Fig. 4] the hyperbola of wo sheets ef 8, @-B-a [Salmon 2, p. 80 Fig, 1-4) the elie paraboloid ed atm and the hyperbli paraboloid scr 2-Pan (R.J.7 Bell, p 150, Fig. 42, The nature of these surfaces can be oughly discerned by considering their sections by planes parallel othe ordinate Planes. Their names were invented by G. Monge in 1805 see Blaschke V, PI Important special cases are the quadris of revolution, formed by revolving ‘conte about one ofits aes. For iastance, the spcil ellipsoid obtained BY evolving an ellipse about its major or minor axs isa prolate spheroid or an ‘oblate spheroid, spectively For the investigation of rirfaces of revolution itis often convenient to use pineal coordinates (8,2). which the fst woof the three Cartesian coordinates are replaced by polar coordinates ra vEFR while = retains its usual meaning. To revolve a plane curve Rasa yao bout the axis, we simply seplace x by thus the surface of revolution is Ar =o, ‘on in Cartesian coordinates, FWVI™FP 2) = 0 For instance, revolving the hyperbola 8447 about its conjugate axis, we obtain the hyperboloid of revolution (of one sheet) 2 tay @ Ca Replacing 2# + y by (xcosa +ysin ay + (yeosa — xsina}, we may ‘expres his equation in the form (cosa + ysina? ~ (ar/bP = ~ (pcos — asin + at which shows that, fr each value of a, every point on the line xoosa + ysina = az/b, yooea — xsina lies oa the hyperbolaid. Allowing ato vary from 0 1o 2, we obtain a con- ‘inuous system of generators: Hines lying eaitely om the surface. Reflecting inthe (x, plane by reversing the sign of :, we obtain a second system of senerators on the same hyperbole. The plane a7 cosa + ysin a = az/b, ‘through the center, touches the asymptotic cone Bey Blylong the line aeona” avine = BY and iets the hyperbolid in two parallel ines: one in each of the planes yoorn—xsina = =a “Another intresting surface of revolution isthe ring-shaped rorus (ao et=h @>d, which i obtained by revolving a circle of radius # about an exterior in in its plane, distant «from the center. This surface evidently comtains two systems of cices: the "meridian," of radius, andthe “parallels" in planes Parallel fos = 0), whore radi vary between a — banda 4B. Teisless ‘bious thatthe tors contains ls two “oblique” systems of eces of radius f such that two circles of opposite systems meet twice while two distinct Circles of the same system do aot most a ll but ae oterlocked” In fact by expressing the equation ae (REF op eta in te form (et bye eA OF + Mat — Be = AL + 9 405 (wens + ysin a) + (co8a — xin) Gh ey a +B Ab cosa — Hsin OE = Abt cos a + sin a) — Aa — Be foreach value of, the torus contains the whole of the setion ofthe sphete Ate ye ot 4 DY cosa — xing) by the plane ea9 Mx cosa + y in a) = Yar=BE:. Since the sphere can be expresed as (c+ bsinay + (y — beos a) 428 = at and the plane passes through its center (—B sa a, Bos a, 0), the secon is 1 reat cle and its radvs is. Allowing ato vary fom 0 to 2x, we ob- Sera ac tn Sete marc usr, tmerton 38918, p19 ist Rain rw sant on i, ots om ae tain continuous system of such circles, and a second system by reversing the sgn of 2. “The plane 8.89 meets the torus in two circles, one in each system (with replaced by « + sin the second system), Since these two citles ae se {dons ofthe two spheres AOE YEE Hat OF = = By cosa — ein), their points oF intersection are the two “antipoda™ points AaB ose, ta Hsing, +t VFI) (5 ae ) (orth signs agreeing). Since each ofthese isa point of contac, 889 8 4 Dzangent pane [R3.T. Bel p. 267 Computing 889 with 827, ve se thatthe “oblique” czcles on the torus lic in the same planes (through the center) asthe pairs of parallel genera {ors ofthe hyperboloid of revelation aay (his remark is de to A.W, Tocker) ‘The plane trough the sen pins. = Lana Fearemeat fbi palo ae ith diecton umber XY Zy he sorepnd ine tow ofthe dteennan replaced by Kon Bo 22 tn ers of genera Cartesian corinne, the pin (3 (x =». 2) by the ental oveson 0 (Gein) hythentaton Oy) 0, (Eyrzte) bya teanaton log the os 2A tn terms of rectangular coordinates the pit (2 il be Uasfrmed ato ia) Seaton ke ns Celnebe) bya ge retestan (474 4 nua oat ri pl (2) wl be atm ino Gane Saiao) byinepmeal dlabve rotation 42.5.a $20 + Bow Bietel insite se “ ‘andthe two systems of gneaton onthe ent Nyperolid ofone sheet Two ge nto otapposte ime iniret (or secasonally are parle, but two disint Soho ofthe same spston askew Ths overeston apes as To the cos {Ge of gear om the Byperole paral 6 Seoce + Sane)? o[fome—Zana)* 9 Complex numbers “The extension ofthe Euclidean plane tothe invrsive plane (86.4) oF to te elite plane (669) isthe geomet counterpart of a familiar procedure in algebra: the extension ofthe concept of number. Beginaing with the rural numbers suchas 1 and 2, we proceed to the integers, then to the a- tional aumber, then (othe real numbers, then to the complex numbers (and if we ha! ie we could continue with hypercomplex numbers). Each Stage is motivated by our deste tobe able to solve a certain kind of equa tion, Real numbers mere understood remarkably wel by the ancient Greeks. Complex numbers were used rely, by R-Bombeli (in his Algebra, Bologna 1573} and especialy by Euler, many years before they could be treated rigor ‘sly; that was how the word “imaginary” acquired is teehncal meaning ‘Topot “the square root of minus ane” on a frm foundation, its convenient hough not esent)to use 4 geometic representation. Such an interpre tation was suggested by 1. Wallis (1685), formulated completly by C. Wessel (1790), rediscovered by J, R. Atgand (1806), and rediscovered again by Gauss The present discussion of umber is not intended tobe a formal develop iment bot rather to emphasize the rle of geometry in the working rules. For {moze complete treatment see Robison (1, pp. 73-84) 19-1 RATIONAL NUMBERS 2.4 5m (2107 ‘The fist umbers that we consider in arithmetic are the natural numbers, os iC Qn ol sin, Ain Mahone Moh 9 (Mea ren arc on Compe mbes by CC. MacDuer neErechpdefoeming «sequence that begins with Land never ends, The problem of solving such an equation as ee2et motivates the discovery of the integers, which include not only the natural fhumbers (or "postive integer”) ut sao Zero and the negalive integers ‘The sequence ofintezers =2 10,12, hich has neither beginning nor end is convenienly represented by points ‘venly spaced along an ifate straight line, which we may think of as the ‘ani of ordinary analytic geometry. In this representation, addition and Subtraction appear as iansarionr: the transformation x» + shits each point through a spaces tothe eight if as positive and through a spaces {othe left ia is negative; i other Words, the operation of adding as the {tansaton that transforms O into a “The problem ofslving sich an equation as motivates the discovery of the ravonal numbers = a/6, where ai a in {egerand sa postive integer, these include nt only the integers @ = a but alo fractions suchas 1/2 (or {) and ~4/3. The rational nambers can rot be written down successively i their natural order, because between fny two of them there is another, and consequently an infinity of ethers; for eeample, between a/b and e'd we ind (a 4 0)6 + a). The corre- sponding points are dense onthe -anis, and a st sight seem to cover it ‘completely, Multiplication and division appear as darts: the trans oration % rx ie the dlstation O(, where O isthe origin (which rep- events Zero) in other words, nultipiation by isthe dilatation wih cen- ter 0 that transforms [into p.-OF couse w may be ether postive (Figure 51a) or negative (Figure %10). In particular, multiplication by —1 is the halfsurn about O. (The point | isjoined to an arbitrary point on the y= ais) ‘We cam derive the rational number a/b from the integer a by applying. the dilatation O{1/). which transforms B ito I. (Figure 9. irrates the derivation ofthe rational aumbers 3/2 and 1/2) This construction shows clearly why we eannot allow the denominator b to be zero. ‘There ‘would be no harm in allowing Bto be negative, but we naturally ideatiy 4i(=8) with a/b. Th the same sprit we usualy write each faction in its west terms” 0 that the numerator and denominator have no common factor. Use the method of igre et aaa 9.2 REAL NUMBERS gin ih he Mot Tre replec “an en he er cet ‘tachaeAndion Dien, tect on ro {Bede che 9 ‘The problem of solving such an equation as a2 motivate the discovery ofthe rea! numbers, which include not only thee: tional numbers but seo the rational aumbers euch as /2 and), which ‘cannot be expressed as factions. (Actually, = cannot even be expresed at 4 root of an algebraic equation) Pythagoris' prot of the irstionaity of ‘V2 was considered by Hard (2, pp. 32-36] 1 be one of the mont ancient IMstanees of strate mathematics, as fresh und signieant a when it wat an a area ee cg ee ee 4cannot beslved in thissene [Infld 1], Nevertheless, the fundamental theorem ff algebra (which Gauss proved in 1799) asserts the existence of roots for values of neven when explicit expressions are aot available, (For 8 heat prool sce Birkhoff and MacLane [1 pp 101-103})_In fat, numerical Solutions cam be found, correct to any assgned numberof decimal places exencist ‘tude, 24 Fe lng aginst wall with he exes suport os ciel BOX figs es placa he tm he alli one rion ede gas ea iota pine ell docthe adr eh? (Hin Take Ts tobe the ego wrote Be above the top te bon Oban am easton whose elven 0 9.7 CONFORMAL TRANSFORMATIONS ‘Wess in $9.3, thatthe transformation verte (ovhich adds to the complex variable = the complex constant 8) 84 transla- ‘ion, whereas (which mulipties by th complex constant a) dilatve rotation about the point 0, including as special caves dilatation (when ais eal) and a rotation (hen ja] =D. Tefollows that a dilative rotation about the general pont Yocsae—9 or Fea s(—ae ence he general irc smarty. as deren 6 the general ear aah [Fond 1, p. 3]; and this js tramslation oa dilative rotation according as a= loreal, (Inthelatercase,c = b/(I~ a) Since the produc ofan opposite similarity aaa election is direc, any ven opposite similarity may be expressed ah the product of given rfl Son and sable direct sim. Using he retin i the an (49.4), we se that the general opposite similarity isthe “conjugate linear transformation ae Since the rato of magnification is again Ja this lide reflection (poss bly reducing to a pure reflection) i ja] = I, anda dative reflection thes ‘We sawn $94, thatthe tansformation unit eee |2| = 1 Similary, 1s the inversion inthe isthe inversion in the cree |2| = &, of radius k, 16 follows tha the i ‘version in the general ctle Ris? aa Re =a) oF on zBy671, any cele preserving transformation thats not a similarity ithe product of such an inversion and an isometry Feptg © faptg. where |p| = 1. To express this product, we replace the = on the right of Dol by pt + gor ps + 4. obtaining est respectively, where b and dae certain expressions involving A, p, 3, 4nd Hence “Boery circle preserving transformation, director oppo, smear frac tional transformation wath gy paatts oe eared aed (od #9, where may be taken to be O or 1 according asthe transformation is oF Ro, similriy. ‘Conversely, very linear factional ransfrmation 9.72 transforms crls into circles The easiest way fo see this s by direct substitution in the equation [eou| = kor (G-we-m =k forthe general circle. This i clearly transformed into an equation of the Same Lind. The following alternative procedure is suggested by a remark ‘ofN. S. Mendelsohn. * Ie = 0, the transformation is similarity, as we have seen. Ife 7 0, wwe could arrange to have € = I as Before; but we shall ind it more con ‘enient to we e different normalization, namely to multiply all fa b,c, ‘Ge nocessary) by auch a number as to make the revised coeficients satisfy fad ~ be = I [Ford 1 p. I, Thea we have, inthe notation of continued Pracions, he identity tb eget Dt dt aan top rerpegoaer hic, of course, continues to hold when we replace z by # on both sides, “Thus the homogzaphy Beth #0 at—bead + amen Mate! Mon, 8894.9 1 may be expres asthe product ofthe nine simpler transformations ous, which ae alternately translations 7 snd MBbius involutions of the special frm 2” = 1/2: the prodvet ofthe inversion 2” = I/F and the felleetion 2” = = (The numberof steps could he reduced fom nine to four by using the diaive rotations” = c=; butt is intresting to aberve that this more complicated transformation ite product of uanslations and “horizontal” Mobis involution) For the anomozraphy ve@th etd we proceed inthe same way with one further telection 2” = 7 Since all these ae eice-peeserving vansformations, the desired result fllows. “The more powerfl methods ofthe theory of functions of «complex vri= able enable us to prove [Ford 1, pp 3,13] that every angle-prserving rans formation of the whole inversve plane is ofthe form 9.72. This shows that fngle preserving Wansformation and ctcle-peserving ansformations are ‘synonymous 1. When fa = Vanda 1 she wansfarmaion 12 When [al Ihe transformation £ angle does ane make ith he as? f+ bisa dive etic, What10 The five Platonic solids We saw, in 446, thatthe Euclidean plane can be filled wth squares, four tach vertex. Ife try to St squares together with only three at each Ver tex. we find that the figure close as soon as we have used sx squares, and ‘we havea cube (4,3). Similarly, we ca fl the plane with equllatral ie Snges, six at each vertex, and ii interesting (0 see what happens if we Use thee, four, or five instead of si Another possibility i 1o use penta ‘gon, thee at each vertex im accordance with te symbol (5,3). ‘With the possible exception of spheres, such pophedra are the simplest solid figures. They provide an ety approach 1 the subject of topology as well as an interesung exercise in trigonometry. They caa be defined and eneralized in various ways [se, eg, Hilbert end Cohn Vossen 1, p. 290, PYRAMIDS, PRISMS, AND ANTIPRISMS rors ch cacy etl be ay ged by mc ‘hom una ks eae Pec on pene Pat Torwhon hr chy need, [A comes polygon (cha (n), whore nan integer) may be described a ante repion of plane “enclosed by ate numer of ies the Seme tha eineror les enirlyon one nie f each ine» Analogous a Comes palhedon i fnite region of space enclosed bys Ate number of pins (once Teach pine tnt yer plans Eipeygon that we cll face Any common side of two face is ‘Te mow famine plas ate promi and prime ‘We shal be con- cerned solely with "gh epslar” poramide whos faces consi fre smegon atm isosceles inangles and with “right regular” prisms whose faces consist of two regular m-gons connected by rectangles (4 that there ate two rectangles and one mgonateach vertex), The height of sucha prism an always be adjusted so thatthe rectangles become squares and then we hhave aa instance of a uniform polyhedron: all the faces ace regular poly- {g0ns and all the vertices are surrounded ake [Ball p. 138}. When n = 4 the prism isa cube, whieh i not merely wiform but regular: the faces are all alike, the edges ar ll alike, and the vertices ae all lke. (The phrase “al alike” can be made presse with the aid ofthe theory of groups. We ‘mean tha here sa symmetry operation that wil ransoem any face, edge bor vertex into any oer Tac, edge or vertex) “The height of an m-gonal pyramid can sometimes be adjust so thatthe isosceles wiangles become equlteal, Infact, this ean be dane when n-< 6, but six equilateral ranges fll at into plane instead of forming a sli angle. A wiangular pyramid is called errhedron. If thee, aad there fore al fou, faces are equilateral the tetrahedron Is regular ‘By slightly distorting an m-gonal prism we obain an mgonalantiprism or ‘prsmatoid,” of “prismoid"), whose Taces consist of two regular mons connected by 2n isosceles triangles. The height of such an aniprism con always Be adjusted so that the isosceles triangles become equilateral, and then We have a uniform polyhedron with three triangles and an m-gon a ‘each vertex. When n = 3, the antiprism isthe regular octahedron. When ‘n= 5, ecan combige it wth two pentagonal pyramids, one on each “base.” to form the regular iosohedron [Coxeter 1, p. S.A pai of iosahedral dice ofthe Ptolemaic dynasty can be sen i one ofthe Egyptian room of the British Museum in London ‘We have now constructed four of the five convex regular polyaedra, namely those regarded by Plato as symbolizing the four elements: earth, ‘ve, ai, and wate.” The disrepancy between four element and fve solids id not upset Plato's scheme. He described theft asa shape that en velop the whole universe. Later it became the quintesenc ofthe medic: ‘alalehemists A model ofthis regular dodecahedron ean be made by t- ting together two “bowls,” each consisting of pentagon surrounded hy five other pentagons. The two bows will stall ht together Because their foe edges form a skew decagon lke that formed bythe lateral edges of [pentagonal aniprism (with isosceles lateral faces). Steinhaus(2, pp. 161 Te described 2 very neat method for building wp such model. From a sheet of cardboard et out two nes like Figure 1012, one for each bow ‘Runa blunt knife along the five sides of the central pentagon soe 0 make them into hinged edges. Place one net crosswise on the other, with the Scored edges outward, and bind them by running an slstic band alternately above and below the comers of the double str, holding the mode Rat with ‘one hand. Removing the hand so sto allow the central pentagons to move away from each other, we ste the dodeeshedrom ring ss perfect mods! (Figare 1010, The most elementary properties ofthe five Platonic solids are collectedque 10.16 igre 108 fn Table I on p. 413, Each polyhedson is characterized by a Shi sym- bol (p. which meaas that thas p-gonal faces q at each vertex. The furmbrs of vertices, edges, and faces are denoted by V-E. and F. They an easly be counted in each ese, but ther sigieance wil become clearer Cihen we have expressed them as functions ofp and q. We shal ls ob- {iin an expression fr the dihedral angle, whigh isthe angle between the planes of two adjacent faces exaneists 1. Gine am abernutine deci ofthe cetahaion (a a ipa, 2. Dose sold having he vero ad tanga ss | Dasrte the flowing sections: ) ofa regular teabedron bythe ple mi wy stecn tot oponte eda i ft cube by Be late may Bests Emo ‘decneizes i) dadeabodron ty te plan may betwen sw oppo fs. wa Sa comaaeat rhombi, with angles 6” apa 12°, wil together tf 8 ‘antec Carta cube") From te (wo oppoite “ace” ects of ths ey eeu crabedr can be ctf in sch > way tat what remains an ote ‘eon Taher word wo etna and ae viahedon cut bee opt to Teak npebedrn.Deduss ta he teraedron andthe octatedron Base 6p eects Geral angles aod that intly many specimens of these tw sls Semel oether ol the mle Elson pace Ba 1p 17) 10.2 DRAWINGS AND MODELS Yo bl # in aed you akin oh sottepay on pc! tet nee apne nomic! ope (sie Ba 651 Leonardo da Vinci made skeletal models of polyhedra wing strips of ‘wood for their edges and leaving the faces tobe imagined [Pacoli 1}, When aw NV Bae SD ad awe 10.38‘such a model is sen in perspective from a position just ouside the center ‘fone face this face appears as large polygon wih ll the remaining faces Sing its interior. Such a drawing ofthe slid is called a Schlegel diagram {Hilbert and Cotn-Vossen 1, pp. 145-146 Figure 10.2a shows each of the Platonic slis in three aspects: am ordi ary perspective view, tne which can be folded to make a cardboard model, land a Schlegel diagram. Each can be checked by observing the natre of ‘ince and the arrangement of faces ata vertex 1. ‘Shaoh Selo gra for pntagoral ati, 2. Whatis the smal number or acute-angled tape ito which a gven ob swse-angled range ea be Suse? (FW Le) "What isthe sallest number of sate-nged tangles into which a square can we ects? (Marus Outer") 10.3 EULER'S FORMULA ‘igh ewe fr ln tvs yas oh Te (2 9 0) “The Sehlege! diagram fora polyhedron shows at a glance which vertices belong to which edges and faces. Each face appears as a egion bounded by ces, except the “initia” face, which encloses ll the others. To ensure a ‘onetovone correspondence between faces and regions we merely have 10 tsocite the initial face with the infinite exterior reson “Any polyhedron that aa be represented by a Schlegel diagram i said to ‘be simply connected or “Evleian,” because its numerical properties sail Euler’ formula voEsP [Bitbert and Cohn-Vossen 1, p. 290), which i valid not only fr the Schlegel diagram of such polyhedron, but for any connected “map” formed by # finite number of point and line segments decomposing plane into non fverlapping regions: the only restriction that tere must beat least one "A proof reembling Euler's may be expressed as follows. Any connected imap can be built up, edge by edge, fom the primitive map that consists of {slope isolated vertex. At each stage, the new edge ether joins an old Vertex to4 nem vertex, asin Figure 10.3, of joins two old vertices, a5 ia Figure 10.30. Tn the former case, Vand E are each increased by I while ‘Fis unchanged: in the later, Vis unchanged while E and F are each in- teased by [- In ether case, the combination V — E + Fis unchanged. ‘Attbe begining, when theres only one vertex and one region (namely, all theres ofthe plane), wehave VoBSFat-041=2, ‘This value 2 is maintsned throughout the whole constuction. Ths, Eu- ler formula holds Tor every plane map. In particular, i holds for every Schlegel diagram, and so Toe every simply connected polyhedron, (For an- ‘ther proof, due to von Staudt, see Rademacher and Toeplitz [1, pp. 75- tf ewe 1036 ewe 10.38 In the case of the regular polyhedron {p,q}, the numerical properties sai the further relations 10.31 QV = 2E a pF. Tm fact i we count the q edges at each ofthe V vertices, we have counted very edge twice: once from each end. A similar situation arises if we count the p sides of each ofthe F Tacs, since every edge belongs t two faces, ‘We now have enough information to deduce expressions for V, Fas functions ofp and g. Infact, VEAP. 2 a Teper Dict area aie art vise 1032 * 4 BaMa A iar Since these numbers must be positive, the possible values of p and q are rested by the inequality 2p-+ 24 ~ pq > O0r 10.33 (Kg -D<4 Thus p ~ 2and g ~ 2 are two posive integers whose product is ess than geToor 2eL or 12 of 3+1 of 103 “These five possibilities provide a simple proof of Euclid’ assertion [Rade- inacher and Toeplite 1 pp. 8487) "Thee are jst fie conve: regular polyhedva G3), 43. BM, 3% GS) “The inequality 10.3 snot merely a nesesary coniton forthe existence ‘of {pea} but aio a suficient condition for in § 101 we saw how to com Strut a solid corresponding to exch solution. The same inequality arses in a more elementary manner when we con suructa model ofthe polyhedron from isnet. Ata vertex me haved PEON ach contributing am ange 1-2). (2) Tn order to form a slid angle, these q face angles must make a total ess than 2 Thus 1-Decoe o(-Be AV and Zp > 4F) "han faces ave alia the es ae al ate ad he vers are al ale ttc taces tr reguay "Show by an eampe ta hi eal fr pba ot ai 10.4 RADI AND ANGLES A sold model of (p,q) c48 evidently be built ftom F p-gonal pyramids ‘of suitable altude, placed together at their common apex, which the cen- ter 0) ofthe polyhedron. This point Osis the common center of three perce the crcunsphere which passes through all the veries, the mud sphere which touches all the edges a thei midpoints and the insphere which toch all he faces at thee centers. The crcunradis 9 appears a8 Lt- tel edge of any one ofthe pyramids (Figure 10.4), the midradiar as the alitade of lateral ace aad the ivadus 2R asthe altitude ofthe whole Pyramid 4 Here 10.48 ‘Such a p-gonal pyramid has p planes of symmetry (or “mierors") which join its apex Os 10 the p lines of symmetry ofits base. By means of these P planes, the sod pyramid is disocted into 2p congruent (irregular) tetra- hha of'a very spetil kind. Let 0500305 (Figure 10.4) be such att hedron, so tht Oy ia vertex ofthe polyhedron, O; the midpoint of an edge (0.0m: Os the ceize ofa face, and Os the center of he whole solid. (The158 ‘THe FVE PLATONIC SOUS et has en drawn to sel for the case of the cube (4, 3), in which {0501 = 0,01 = 0x02) ‘Siace the plane 0,020 perpendlaly Discs the edge Oy, 0x prpenclar to both O10 and O.Ds. Since (040;01' the plan of face the inradivs 0,0; petpendicalar to both (s0s and 00g. Thun the thee lies 00s, 0x02, 020s ae mutual pe Dendiclar andthe tetrahedron is "quadrietangulae™ al fou faces ae Fighiangledtranles Sela caled sacha tetrahedron an orthascheme {Concer 139) Maay rltions involving the radi R= 000s R= 0:05, ak = 010s can be derived ftom the four right-angled triangle, in which p01 206020, ~ #/p. But the whale story cannot be tod tll we have found the angle 6 = £010.01 which i half the angle subtended atthe center by an edge [Coxeter 1, pp. 22. ‘Another signifiant angle is = £0,040, whose complement, Z 00:03, is half the dihedral angle ofthe polyhedron, nother word, the dihedral angle is * — 2. In seeking these angles its useful to define the vertex gue of (pg): the poligon formed by the midpoints ofthe q edges ata vertex Op This sin ‘cede plane polygon, sincetts vertices lie on the cele of intersection of #0 spheres: the midsphere (with center Oy and radius :R = 030) and the sphere with center 0, and radive/ = 050. Wesee fom 2.88 that the er- tex igure of (pq) i (g) oF side eos > Since its plane is perpendicular to OsOn,its center Qs the foot ofthe pr pendicalr from 0, to Os0y (Figure 104), and is circumradius i 204 = feos 8 By 2581 with oe n/p for), this ercumadivs is teas Bose se F = co Fin Hence 10m cos ¢ = os [oxete 1 p21 eet inning 0d nowy ose 9, AR = Foot wea (lend), ond otf In orderw eliminate g its convenient 1a introdce the temporary abbrevia- Ke a sn — cost = sit 7 eet Sothat sin p = Kesea/g. Then vos wR rosa cosy = 008 ‘Thin sna ws compte i angle ay «tain (<55)in® Cals) (Coemrtp thither Speers Sea 10.5 RECIPROCAL POLYHEDRA ‘The Platonic solid (p,q) has Peiprocal, which may be defined as the polyhedron enclosed by certain set of planes namely, the planes ofthe ‘eriex gute atthe V vertices of (p, 2). Cealy its edges bstct the edges DF (pg) at eght angles Among there E edger, thore which bisect the p sides ofa fae of (p,q) all pas through a vertex ofthe reciprocal, and those hich bset the qedges ata vertex (p,9) form a face ofthe ecprccal. Ths The reciprocal of (7, 4} (4.7),