VAN NOSTRAND REINHOLD MATHEMATICAL STUDIES Editors Paul R, Halmos, indiana University taee a Frederick W. Gehring, The University of Michigan DIFFERENTIAL GEOMETRY oul A Mtoe TURES Om BODLAN AGES Sma gncn-EDTERES OM ELITE BOORORY VALUE POLES Nel eks-NOTES ON OFFERENTIAL GEOMETRY LespaloNbiTOPOLOGY AND ORDER by ‘tee. Srbean-NOTES OF SPETEAL AEDT ‘ger (yeon-HOES OM LIE ove PhosUETORES OM G40QUTS THEDEN 1 Sr NST THEORY > NOEL J. HICKS Geog W. Maky-AECTURE ON THE THEORY OF FUNCTION FA eOMPLEKWARARLE ‘The Unioorsty of Michigon as Are-CECTHBES OW QUSIEONFRNAL MAPPINGS 1a SMMC OBITS IM LEER TaPOLOGT A fxs TRIS NTE HERY OF FONETINS OF NE COMPLEX VRE luna Gatese-SLETED ROBLENS OH EFTONL SETS yand yn espa athin-ELMERS OF APHORATION THEDRY 5. COMPLEX MOMFOLE WITHOUT POTETAL THEORY -,ens-ARANT MEAS OR ToPOLOIAL GROUPS: 1 Wisn-USTURES ON ITER RUINS fn Ge, LECTURE SERIES OFERNTAL EOUTON, 1, et. —LECTURE SERIES FFERERTL ERLAINS, Vo. 1 rl Wgon-METORES 0M MESHRE a MTERATON Sh FogsTH ERODE THEORY OF wan OES tot ohssk>-ORICMET PROBLEM, EXTREMAL EACH AMD PRIME CS ons Saneton- OTE OLE ALGEBRAS ma ol. Sapro-SMEOTHNG AX RPROCMTION OF FONE Aa ete seman SURVEY OF MINA SOFAES meee nye VAN NOSTRAND REINHOLD COMPANY op atean-AOTES OF KOMBLOIAL ALGER es laser EDMETHIAL COMBRATORL TOOLEY, ve, ies Wee! cascicaaayersuiet candle ae Lest lserCEMETECL COMBIMTORAL TOPLOEY, Vo. aan A itn —NTROUETION TO ERONG TERY et Aili MOTES ON OPEATOR TARE 1. Ge TRENDS WATERYHl eerie fat Noerhane Ramen Contany herons PORGet: iat! New York ‘Chcogo "Millrae Dllan ‘Vax Nowrean Ratno19 Conran orsign Ofces Landon "Toronto Melbourne Copyright © 198 by LITTON EDUCATIONAL PUBLISHING, INC. Al igh reserved. No pazt af this wor coneed hy the ‘ooyightbaroon may be repradaced or sed fn any {orm or by any meuns—eraphis,sactroni ar chanel including phstoopying, recon, taping x infornatin storage aad retrieval systems—wvithoat itn perio tthe publisher, Manafectared inthe United Slots of ‘America ablshed by Vay Nosreax Rnynote Courasy 480 Weat rd Street, New York, NY. 10001 Published simltancosty in Canada by D. Van Nostrand Company (Cenada), Ls, wosressce PREFACE ‘The following paragraph presents a very brief history of differen lel geometry and the notation used in these notes. Differential geometry is probably 8 old as ony mathematica dis cipline and certainly wos well leunched efter Newion and Letboite had laid the foundations of callus. Many eesults concerning ste feces in espace were obtsined by Gauss in the first half ofthe nine teenth centray, nd in 1854 Riemann laid the foundations form more betract approach. At the end ofthat century, LevisCivitae and Ricci developed the concept of parael translation ia the classical language of tonsors. This approsch received a tremendous impets ‘tom Einstein's work on relativity. Daring the eatly years of this century, E. Cartan Initiated research and methods that were indepen ent of paticular coordinate system Gavariant methods). Chevalley"s book °The Theory of Lie Groups” (1946) continued the clasification of concepts and actatio, and it hes had remarkable effect on the ‘current situation. The complete global synthesis of Cartn's epprosch Was achieved when Ehrosmans formelated connexion in tems of fiber tondle, These notes utilize an invariant local method formulated ty Koszul ‘The first thee chapters of this book provide = short course on clas- cel differential goomety and could be used at the junior level with ltl outside reading ia Lneer algebra and edvanced calculus. The fet six chapters can be used fora one-semester course in differen tial geometry at the seniorgreduate Level, Sich « course would cover fhe main topics ofclessical differential geometry (except fr the Ieterial in chapter 8) slog modem language and techniques, and it would prepare a student for farther study inthe books of Helgesoe, Lang, Stemberg, ot. (Soe lis in following paragraph). The entice ‘book can be covered in fll year couse. A selection of chapters ‘ould make up a “topics” course or a course on Riemannian geomet.For example, @coutse on manifolds end consexions could consist of ‘chepors 1, 4, 5, 7, ad sections 9.1, 9.3, and 9.4. The reader with little experience should move though the First three chapters fisly ‘icky. The problems are of several typos: (3) those that provide explicit computations to test the understanding of the theory, (0) thooe that requite the student to prove theorems similar to those inthe text, (©) those that lead the student trough supplementary material, some of which may be an integral pot of the exposition, and (E) those that lead the student to books or papers inthe literature. An introduction to bundle theory and the theory of Lie groupe is covered via problem material. Our hope i to give tho reader a solid understanding of the basic concepta and to stimelate hia to frther reading and thinking in differential goonety. ‘Besides the specific references found inthe notes, we would like to mention the fllowing general ferences: Point set topology Kelley; Hocking and Young: Pervin, Linear algebra: Halos; Jacob- son, Advanced ealeulue: Buck Keplan; Nickersoa, Seared, and Spencer, Classical differential geometry: Eisenbatt Hilbert eed CCohe-Vossen; Strlk. Contemporary differential eonety:.Auslonder ‘nd MacKenzie; Crittenden and Bishop; Guggenheimer; Helgason; Kobeyeshi and Nonirs; Lang: Normizu; Stenberg. History of difer: ental geometry: Straik; Veblen snd Whites We will use th following conventions: “iff fo “land only i"; 17 for *Q. B. DLP; Cartan? wil seer tothe third reference inthe bibliography under Catan, and when there is only one reference for an autor, we omit the superscript 1; Fp, ¥,, and 3 will all be used to indicate a sim is to be made, and int later two eases, we hope the emitted information (fange ot index of summation) is cleat fom the conte. [A this time T would Like to expcess my gratitude to formar teachers 1X, Schwid and V. J. Varineau for thele early encouragement, to Mise Margaret Bi. Genova snd Miss Gili D. Hodge for their help in Syn the manuscript, and to L. M. Dickens for his contribution tothe under standing ofthe lustratlons. Finally, Iam indebted to W, Ambrose and H, Samelson for sharing their insights via courses, notes, end N. J. HICKS, CONTENTS ger 1. MANIFOLDS ‘Be Jeeta ot map 9 2. HYPERSURFACES OF R" X3 Partie surteces (normal sa ey oo ‘Chopter 4. TENSORS AND FORNS ‘Chapter 5. CONNEXIONS Pe Ee, ‘Ghapter 6. RIEMANNIAN MANIFOLDS AND SUBMANIFOLDS 18(Chapter 7. OPERATORS ON FORMS AND INTEGRATION (Chapter GAUSS-BONNET THEORY AND RIGIDITY (Chapter 9. EXISTENCE THEORY Chapter 10, TOPICS IN RIEMANNIAN GEOMETRY 1044 Jet ede ng cone ints, Bubliogrphy Inder # 105, ws 1 176 181 1. Monifolds. 1g this chapter wo define the fundamental concepts which we deal thoughout these aotes. Specifically, the notions of manfols, {paction, and vector, and the conceot of ditferestisblity (emoothncss), est bo carefully digested form slid foundation, fhetion 14. Nanitotds porn Let R be the set of reat numbers. Foran int n> 0, fe Rb the pod space ofodeed ees of real name r=. Thus R= (Cyrwiagh 98). Fort = room ft be he | "Pptia eoontinate (it actions of RO hess Re BF {42jn018,) = ae An open set of RP wil be a Set which i ope i the | © landed ete opto idaced bythe standard mttie faction dom RE thas i= (ajay) od B= (yr) ae pots fn Rs then vd, 8)= (37.6, "81 ‘The concept of dilleretnbility i bose ultinstely on the definition fs derivative in elementary calculie. Let rte en ater > 0 Recall tom edvoced calls that «nap ftom an opens A CR* into is called C" on A if possesuen contiaous pil derivatives fn A ofall orders 1) n Cron A Aesch of ite lt fonctions f, =u, of te Cen A for lua thus foe pin RY Me) = Eph) in2 Notes on Ditlerential Geomotey We now define « manifold. Let W be » set, An ncoorinate pat on If is @ pair (@, U) consiating ofa subact U of W end 1to 1 map ‘oF U onto an open set in R. One mcvordinate pat (9, U) ix Clad to enother mecordisate psic (QV) if the maps «2 and Ges" are Cmaps wherever they ae defined (thus their domeina of definition mast he open). A.C" nesubatas on Mi 6 ecllection of mcootdinate airs (yy Uy) each of which ix C* seated to every other member of {he collection, andthe anion ofthe sets U, ix MA maximal clfection of Cr elated a-cordinate pairs in called s'C* meatlan. Ifa C* matlan contains & C’ mubatlas, we say the subatlas induces or generates the atlas. Finally, an m dimensional C* manifold ore C" rmanifold is 8 et W together with aC" niatae, When eM is castomurily cated a locally Euclidean space or topological manifold, and only when £0 4s M celled a differentiable or smooth manifold. An atlas on set M Js often called a differentiable sructire oc @ manifold simctare on Me [Notice tht one set may possess mote thin one diferetiablestucture (Gee exanple 4 below), however, a definition of “equivalent differ cntiable structures is necessary before the study of different atlases 00 ‘et becomes messing (see Munktes"). Exch meoordinate pai ( U7) on set M indaces set of reel valuod functions on U' defined by =, u, 048 for= Ineane ‘The tance tions tyre, ate called coondinats factions ot coardinale system and U is called the domsin of the coostinate system. | We list some examples “7. Let be R° with a C*meaabatas equ to the pair consisting of Go the identity map and U = R™ “2 Let I be any open set of R® and let a C'msubatles be (he identity PW. 43. Let W = GLC, R), the group of nea-tinglar Relineer transformer tons ofA into itgelf. "Then Mf con be mopped 1:1 onto en open set in Re? and thus e manifold stnicture can be defined on M via cxample 2 f(a) is a matrix representation of an elesent of M vith respect to tN usual bse of, then map Ce, int the » ur arg Maye Basra yr The image set of this map will be open since i isthe invesse mage of an open set by the deteminent esp, which is continuous (indeed i is Cas & map on R*), chon. 1 Manifolds 3 4. Lett be the Iedimeesonal C* manifold of example 1 nd ot MR withthe C! Laubatins (2, R), whore +i the entity sping on Re Then £1, since xin not C? the oii, V5. Let Fhe a C* real valued function on RO", with +> O ond ‘n> 0, and suppose the gradient of f does aot vanish on fheonstant set M = [pin R°"* f(p)~ 0). Then at each point {2M choose any patsl derivative off that doesn't vanish, ‘sey the 7 one, apply the implicit function theores obtain & reighboihood of p(eltive topology oni) hich projects in @ 1 way into th w, =O byperplane of R°", and thos define «| sbatioe which makes M'e C” nenifld Tris example covers mony Classical hypersurfaces in R, Including aphere, planes, and cylinders, V6, ‘The process in example Scan easily be generalized to obtain Cra ~ # manifolds from *coastent sete” of a C* map fs RO RE whose Jacobian matix Is of rank # onthe constant set 7. Let F bea unvelent map ftom an open set in R* into R™, with O contains a C™ stlas (see Mankzes'), There is en example of Kervalie which exhibits a C® atlas on a et which edits 90 at For further work onthe “equivalence” of differentiable structure Mulnoet 4, aunkres! **#, and Smae! The following list of nine problems ere recommended in oder to faniliasize oneself with tho notion of C™ map. In particule the pro lems are aimed at obtaining umbers 6 and 7 which ore often use Tre list (remember Ais open in My which is @ C™ nemanield), 1. The map & A+. is C™ on A if fi C™ at each point p in A, 2, Ife AN, Fis © on A, and U isan open set contained in 4, thea fy Is CU. Chop. 1 Manitotde 5 3. Let U, be a collection of open sets in M and let fy: Uy = te Con U, for each h. IFT isa function whose domi is the non ofall U, and iff, = f, forall then fis C™ on ite dowsta, 4. Ih ARM is C™ on ACRY and f: B~R is C* onthe open et BRM, then § of is C™ on ANT. If: AW is C™ 00 ACM and (6, U) is coordinate ait on W, then Fos™* ie C™ on GAN). 6 Let Pho @C™ ovmenifold, IfF: A=W Is C™ on ACM end & Bis C™ on the open set BCA, then ¢c1 ie C™ on ante, ‘The map A+ is C™ on A Cw it for every coordinate pair (, U) ina subatiag on W the functions t, «ae C= 08 ANEW), fori 3 n> band g: RY RP by day, Cm on Re IEA: REAR by Ren thon hse C™ on RE, 9. Let fand g be teal valued functions that are C™ on the subseta A and B of W, respectively. Show that + ¢ and ff ate C™ on ANB, where (C+ eX) = Kp) +(e) and eK = Mohd For the record, we can and so do define & Lie group. A Lie group © iss group G whose uaderiying set is also a C™ wenifold such that the group operstions are C>, ie. the map gi: GxG = G where dle, f= a"! is C™ (wee problem 18 and 20). (One last bit of notation, let C™(4, I) denote the set of C™ fanetions rapping en open set Ain @ manifold WM into a manifold Section 1.3. Vectors and vector folds ‘The definition ofa tangent vector generalizes the “directions! Aevvative® in RP. 1X, isan ordinary (advanced ealeules) vector ste point ain R® and Fis e C™ function ina neighbochod of m thea dtine Xf =X, + (V Ny, where Vis the gradient vector field off. From the properties of the “dot” pradact an the operator V It ftlows that,6 Motes on Ditlerontia! Geometry Kylal +O) ak fs Bie Xl (kg > Akt, ‘whore is a C™ function in @ neighborhood of mand « and axe rest numbers. Notice X is not normalized to be a unit vector. We generale izo now to define s tangent vector on @ manifold as an opertor on C™ functions which obeys the above rules, Let be 8 C* nenanifold. Let m be ia M and Jet C™(a) denote the set of real valued functions that are C™ on some neighborhood of A tangent vector atm i eal valued faction X on Mm) havo the folowing properties XE) XE Xe, KON = 8D 2) AU) = CxO + Heike, Where food & re in C™(m), and b is in R, The set C™(n) is almost a ting (hore is slight problm with dowtins), and thus a tangent vector {8 often called a derivation on C™im). ‘The tangent space fo M a m, denoted by Ny isthe set of all tane sont vectors atm. Tis vector space over the cea feld where (Ee Y= XP + VF and (BX)E = AGED for X,Y ia MW, fin C™lm), and 8 real nunber, Let typing bea coordinate system about m (een, m i i the domain ofthese coordinate functions). We define fr each i, a coordinate wees tor at m, denoted (9/25), by dosnt) (tmp bore x, «4, 0 and the diffretistion onthe right side is a8 uta oa RY. The vetfication of properties (1) and (2) above we leave to the reader. In a moment we show these coordinate vectors for 1 base for the tangent space at LEMMA. Lot yu, be a coordinate system about m with 4m) = 0 fr alli. Then for every anction fin C™(m) there existe n (hep. 1 Maniftde 7 factions Fyeely in Cm) with) (8/9e ud ad = a) +, in ancighborhood of m Note the equality in qeestion ix an equality bepwecn functions, and fim) representa a constant function with value fy tho sum is taken foe f= I, 2s ated i the fate this relevant ange is 10 be understood.) Lat Fa fog roo Let & be the coontinate map belonging te the snd we know F is defined in a ball aboat the exgin ia R™, Be [pin RY distance from origin to p< rh. For (yuna) in B we have, gee 0 04 FE, Op) — FO) + FQ) = = B Ploy 8p Qn NS + FO) F023, | aati ty nn P00) + 3,6 ayy where Xu, iat F (a8) fi C™ in B since (@F/Ou,)¢ C™. Let fF, 06 andthe lenma is ved. // THEOREM. Let M be 9 C™ mmanitold aad le xy, Bo @ 2 ndinate system about m io W. Then i Xin Wy X = SAR VA/A,) op 0d the coordinate vectors form a base for W, which thus bas dimesion rool, We first prove the stated representation, Take X in M, and in Cm), Ie xf) £0 for alli ety, =, ~ xm). Then apply the lemma to £ with'respect to the coordinate system Yon), and notice (Gl/ay on (0/2 Ym. Next we see it 8 constant mp then————_- ——_ —. A $_$_s—_— 8 Notes on Ditlerential Geometry Xo) = eXU) = XU © 1X) = 26x) Which implies ©X(1) = Oand X(e Ths KF Xin) + 3 9) = 5 ay (0 yn = 3X, ~ 266m) = 3 0/05 km) hich proves the roquired represeatation. If¥ = 3,0(0/ds,) 0 then 04 Y2,~a,, thas the coordinate vector ace independent end span Nye ‘A vector ficld on a set A is @ mapping thet sesigns to cach point Pin As vector X, in M,. A field X ie C™ on A if Ast open ad for ‘cach real valued fanction£ that is C™ on B, the function (X1)(p) = X,fis.C* on ANB. TEX and Y ase C™ vector fields on A thet racket {e's C™ vector field Lt, ¥] on A defined by [X, YI, ~ X,079 - 200 Uf and are C™ fonctions, it ix trivial that [XP PAE Se) ~ [My YF « Uy Vlg, and Uf, PKaf = aX, YI for ain R. To check the product property, consider 0%, Vtg) = X08») — rOXU)) 2 Mvg gy N~ Yinke + 6X0, = Weg + (KONE) + KAKA» gx = YKE (700K) ~ CroLAD ~ arxt AX, Ya + (XY ‘Thus LX, is a vector field and the prof of ite C™ nature we leeve 5 @ probien, For ater use, notice that [X, ¥1—~ (Ys X1 1%, X1=0, and the ‘racket is linear ia each slot with respect to aditon, i>, Dk, + Xp Vl Oty, Yl Eye YI However Of, ar) ~ {QF — AYOX © aX, ¥, andi i tha property that prevents the bracket ape hep. Monitonds 9 ing from being a tensor (problem 10). Problem 13 gives geometric {nterpretation of the bracket, and in section 9.1 theo ave applications ‘avolving integrability conditions. For example, if x yn, iz 4.00 condinate system then [9/9r, /éx,] =O for al und j(aiaee cron partial derivatives of C™ fanctions are equal), sa actually this com {ition on n independent vector felts is sufficient to imply the fields se coordinate vector fies (section 9.0; The bracket operation also satiate the following expression which 4 called the Jacobi identity, UO, 21) (2, (4, YI Uv, iz, atl <0 uuhese X,Y, and Z are C™ fields witha common domain, Section 14 Tho Jacobien of map Let and be C™ manifolds of dimensions n and k respectively We defined shove the concept of a C™ map ffm W into Mr Such a sap induces a linear transformation fom each tangent space N, ito the tangent space Wg, This linear map is called the Jacobian map ‘or the differential of and we denote it by fe (often i ix denoted a, but we reserve the symbol d for the exterior denvative operator. Let be in, and we define f4X as a vector at Ki) in Ny taking @ function "which ie C™ i « neighborhood of Km) and setting (a = Algo. Tis trivial to chock that 4X ie 8 vector at fin) and the wap 1 i Linear, By selecting coordinate systee xn, sbout m and another Yip about fm), we can determine & natra representation for fe whichis called the Jacobian max of fe with seapect to the chosen coordinate systems. Let X, 2/3, ¥,= 9/0 thus Xyeinkyy st ', form a base for My and we compute fe by computing its ection oa this base. Nanely, FX, = 3(06%,)y,¥, by the representation theotes above, hence the matrix in qdestion t= the matix (lok pr) = (hy, Vax) for cic mand sick ‘The implicit function theorem ard the inverse function theorem cen be applied and formulated inthis language, The former we postpone, since we do not really need it fr some tine (see problem 16) but the latter is both useful and instructive. First a definition. A difleomor= ‘hiom is @ mop f: M+ W that is 1-1 und onto with both and IC™,WO Notes on Ditferentia! Geometry ‘nd if such anf existe, then M i dileomorphic to N. THEOREM, (Inverse fonction) Let M and W be C™ nmanitolds sand let M +N be C™. I for m in My the Jocobion fx atm fea omorphism of Monto Ney, then there is neighborhood U of w and ‘neighborhood Vf Km) such that fin diffeomorphism fem U to ¥ (iets Lisa Local difeomorphise stout), We teave It tothe reader to choose s coordinate system on oth sides and apply the theorem from advanced caleulus to obtain the 16> Slt, Notice the C™ demand of f and f" inplies the theorem could be stated ss a necessary aswell ara suficient condition forthe exiat~ ‘ence of local inverse. If ene only demas continuity of the inverse, then the map x +x provides « homeomorphism of R onto R whose Jacobian is singular at the o [Now conser the behavior of the Jacobian with respect to com posite maps. Let ¢ be a C™ map of W into the © manifold Ly ‘Then at each min My Uf 0De = feof for if ix C™ function about ila) aod X in M, then (0 MAXI ~ XH of of) = UakKho€) = (UalleXD\A. In tents of coordinate systems, the above computation e Iibits the chain rule and a multiplicative bchaviorof Jacobian mates. Wren fs a diffeomorphism of W into N, and X and ¥ are C™ fclds on M, then faX and fa¥ ere C™ fields on W with LX, YI = Uf, fa¥L Section 1.5 Corves and intogel caves In these notes curves wil be viewed a8 special case of mappings, ‘hus we will deal with “parameterized curves” almost exclasively. A carve in W i9 @ C™ map ofom an open subset of R into W. Often we speak of «curve fom [a 8) ito M where (a, 8] «closed interval ‘of real numbers, and in this case itis assed the domain of o i sctually an open set in R containing Lo, ble Let oboe cue in M with domain U. For each tin U define the {tangent of o ot tt be the vector T(), of Tt, at lf whore TC esld/dt, and d/at denotes the usu dffereatistion operator of teal valued functions oa R. Thus if yuant, « coordinate system about 0, then 70) = ¥ (ate, «09/8/98 gene By ifferentoting the coe ‘cxdinate paranet et functions x, "0 oft) one determines the coefficients of T() with respect tothe coordinate vectors associated with the eo. chop. 1 Manifolds n cdincte system. Notice this 7() isthe usual “velocity” vector ‘sociated with « parameterized curve in R?. Having the idee of cuve and tangent vector we can give a geo- retric description ofthe Jacobien fe associated with the map f: M +. For X in W, choose any carve oon M with o(0) = m and T4(0) = X. ‘Then f ovis a carve on N with fo a0) = fm) end indeed fx Tl). Thus we Mfll in the vector by @ euve, map the curve to N, sand take the new tangent vector.” ‘This device ie very useful if one Inows geometically the behavior of certain curves; eg, let M = Ute, y, 2) fe RY at + y* I, Let S be the unit sphere in R, and let f M's Sby fy, y, 2) = (a, yy. The particular fjost defined is called the “aphere map" oF the "Gauss map” from W to, since it ‘easentially uses a unit normal vector field to M in its definition. Ite Jacobian should he trivial to compute at each point from the above temas We cary the idea of “Filling in a vector" to classical setting. Let X be a C™ vector field on the manifold M.A curve isan ine togeal curve of K if whenever lt) is inthe domain of X then 7,0) = Xgcpr_ Thus we say the curve o Nite” X, and suggest the physical fxample of the velocity vector field (which gives 1) of «steady uid flow and its streaelines (which give integral carver). The local ex- Isteace of integral curves is guaranteed by the theory of oninary 0 end a unigne curve o: (br, B+ #)-«¥ sich thar ob) =m nd 0 an intogeal curve of X. Prook Let x yn, be & coordinate system sbout m whose domain Vis contained in the domain of K, Let X = 3)@/dx)) define C™ eal valued fonctions f, on U. Then the condition that a curve-« be an ins tegral curve of X becomes the condition ax; 20) at nthe domain of o, or writing (improperly) as usual (0) =, oot) we have the system of first order o¥dinary differential ations Bo leynurtyh for Tom Apply an existence and uniqueness theoren from difer ential equation theory to obtain +> O end functions 2 thet defines 00 the specified range with the requted propertien./ Actually the theorem from differential equations gives mach more that the above conclusion fori includes the C™ dependence of olae ons as we vaty the initial parameter b andthe point m (see section 93), We rtur to this later when discussing the existence of geodesice 1nd the exponential map (sections 5.1 and 9.3). For global tions see Palais? or Leng Chap. 1 Manitotis B It is conveniont to define a troken C* curve o on an inteval {a 5) 10 bea continuous map o from a, 8) into M which ie C™ 08 cach of finite numberof subintervals [a 8) [6,8 lavolb,_.0 bl Section 1.6 Sabmaniolie A.C” kmanifolé Mf i #submonifld of 0 C™ mnonifod Wf for ver point pin ther i a coordiote neighbocood Bf with coe cntinate fenctions Fyn, such that he set U = [min U: Ts) = =m) ~ Oli a coordinate aelghtohoot ofp in Wf wth cooriate fanctins x, = Flynt, = Fylyr There coosinate rystens ate called special of adopted coordiatesycons, Notice iis not requied that W 1 Uso elles” of W wey op ‘roach other *lices™ of W in Wf (oe problem 17), end hence the topology on M may not be the elative topology. ‘The definition of submanifold inpliee W ise subset of and & 0 defo the cuve ot) a8 follows: go ¢ perameter units on X Sntegral curve thru m 10 p,» go tants on ¥ curve thea pt Py 0 units on (=X) eueve thes p, to psy go t units on (~ ¥) uive thea p, to lh IC 40) = aye) show T,(0)= Ut, Pge (Gint: ase the lems in ction 9.1 and patil Taylor series.) Let and Nv be manifolds with M connected and let f and 4 be (© maps of Minto N. Show f, =O iff a constant on) = lm) atone min M and Let Ibe in CUM, R) and define the cifforential of fd to be the Tinos map of My into R where (d)y(X,.)= Xf Show £,0%,)~ (4), (OIG/30 whore ¢ isthe identity Coordinate fincion on R” It is Because ofthis case that in « general ‘case the Jacoblen f, is often called the “dillrential off Prove the Inverse Function Theorem (p. 10), State and prove ‘2 version ofthe Implicit Function Theorem of advanced caleu- 1s in terms ofthe Jacobian map% ra 3B. 1, 20, Notes oe Dillecotial Geometry Prove the last sentence inthe thd paragraph of section 1.6 ‘Show tht the image ofa regular (o, 40) anivstent curve o ‘napping es open interval into @ manifold Wis @ one-dimonsion submanifold of W. Let be w unt constant vector field on 'R? with irationl slope. Let Tbe the set of equivalence classes on R? where (a, 8) (e.g) iff a c= mand (Bd) =m for integers mand n. Show T i a two-dimensional meaiold (ich i called the Hat torus) in @ natural wey. Show Xin duces « vector field on T such that the image of one integral stuctire or BU) which i calld the bond of oven over Me Por i GLin, R) lt Ry BW) = BUD by Ry) 634» Shap Satyr Seige) HD =O yh Slow Ry Is C=. Let sy: U = BM by sya) my (8/8 {a/Si,),) oe min, Show ng in Cand # By i he ‘deatity B. “The map ae ealled the cordinate section imap over U. Lat Gi Ur OLte, R)» Oy din #)- Re Spl) spimbks Show Sis dite onto its image. The map 3m ealed «stip maps 1, U) and, V) are coordinate pair on Wf define oy UNV = Glin, By syn) 6 Solid = ay(a\. Sow ayy in C= itis cll srecrral faction fot BON). Show (4) = a) hich ustiies the sane riht actin for Ry. es fned bin BD let he Chy > ‘UW by 14) Be. SoS fy tw = Call the et Pye) the (vertical fide over mn M. Show Fig en 927 subman- fold of BD ea, in fo of Glin, RSonto Peg 0) she), show "0 fy ins left taslation on Gla, BA vector es BU sich at nf) =O ie called a vertical vector. For in BOD, let E, (8) Uy) (6) define a vector £0) oe problem 21) Siow E, 2a global C™ vetieal vector eld on BQN. Compute IE, F,,)| 18 Notes on Dittrential Geometry 2. Hypersurfaces of R= In vory real sense, this chapter and the next ate to0 special, ie tnuch of the theory bolongs to an artitrry submanifold of e soca Riemannian” manifold. We specialize beceuse we can obtain many ofthe concepts aad results of classical diferatil geometry quickly and easily. In so doing, we hope to develop the “geometnc’ satis tion of the reader suticiently to make Inter generalizations and defini= Section 2.1 The etandard connexion en B®. Recall in section 1.3 we shifted the classical nation of « vector fiom a * directed line sepment” to an operator on fonctions, es if Xai « bf» chin a familiar vector on R? from advanced colevion thon we rowrite X = o(9/@) + 43/2) + cl@/Ae) go It Fis « rel, valued C™ function on R3, then X0 ix @ derivative off inthe direction x MAX Vina, gH, at Ven eases Notice tht X need aot be a unit vector. When 2, 5, ad are C™ fnctions on R* themselves (possibly constant functions), then X is "|. C™ field and X1 is @ C™ real valued function on R® DIG) = HF = ate) (4 619) # 69) + lp) f= 0) 0) + 0) 2) + (9) 2, Since bth ofthe representations of «vector field given above ere sevkward to wet, let us simply waite X~ (ab), tes ving op {rig the coeficien funetions (or constants) @ and of te debt ase Feld dt, 2, da on R, Me now define the dvivaive of a vector fold ¥ in w direction X, Let X bea vector at pn RY and Set P= nye) b0 Bel shout thus each », Isa C™ real valued fusction oo he onan of F which ‘includes The coveiat deivatve of Yin the decion Xs te PectoDy Y= UyyyronX,yq) a8 a vector at pI and are O° fields with the see domain, then Dy¥ in a C™ field with domain Aa Chap. 2 Hypecsietaces of R° » For exanple take R?, let X = (a,b,c) let ¥ = Gy? + 42,» ay 23) and then Duy 1%, 20m, X61 2, X60, 30°] + 2yb, a Bete, ay? 4 Deyb + Aoy = here 4), and © may be functions or constants. ‘Tho properties of D which we now list are one ofthe ssn analytic, tools ofthese notes. Let X and W be vectors at pin R®, fet ¥ and be C™ fields about p, and let Fhe a C™ real valued function about Then ) BW. 2)-Dyy Dz © By, W)-Dyy +Dyr O Dyyyx¥ ~My © Dyin. unr, + Moldy. ‘These follow directly from the definition of D. It i important to sotice Dy can be computed once ove knows ¥ slong curve o that {19 46. 40) =p and 40) = 119) ~ fy. Por let Voge G40 seo a hen Dg (2 (0) anc B8 OS since By he shan rae, Bio) = $ Bip Me =x, «Wr, and TO) = ky. Tras GeO § SLOG = 4, 9.) and 1). That is an RP ~ vector fot that is C™ on the carve o wth tangent 7; then DY is a wolsetined RU-vectr field that is C= ono ‘Using the operator D, we can define pale vector ilds along a curve and geodesics. Let o be a C™ curve (in R°) with tangent T and lot ¥ bo an R’-vector field that is C™ one, The field Y is parallel along oif DyY =O along 0. The curve vis u geodesic it DT = 0, shes fits tangent Tie parallel along. es trivial to see these are the oma concepts of prllel fies and seodesics Ja RY for Let of) = (4 (then9t) and Fog (y(n, OD Then DY = yy /dtonsdyy/dt) = Ol each {0 is 8 constant fane>20 Notes on Ditlerenta! Geomcty ion of 20 Y na Sensi vector ldo RO evalestd on The curve aie geodesio lf Dy? daytheudin ah eae Inoles a) =f +d alae fenton ota es obec tees of sig ne Rote tat te pranteetion of «cue apatite dle The ervrlrton of he fnon of covarit dllentiaton 4 coneson on any C™ mano is clous tea a the extn ofan operat D which stn he tee ar rio Usted 4B) and eaign a C> vector fle Pana wo ee sai a 6 fei Day on 4 Noto tae sap eee he omeslon ov «anol tthe case of sect Rien otde Soever thre eusts one coneson hh she one Stet ey, and nthe cose of Dis hie an tet, we now expan how Be na Hencfoth, dente teu! dt poet ole prota of vector Y eed tngot oR” by oF, 2, Thon Pe 2 thn, 2°= Bagge WY and 2 06 Co Noe veh ds sin hon = on dene Poles of tangent vectors at each point. Thus i ¥ and ave, ‘then X, Y is a real aumbor and <,> satisfies the following popertce (© oynmetsic) YS = <¥, X>, ©) Citinese) x 4 Fy Z> =, 2>4 <4, 2s say Vos 9k, Vator a in (©) ositive definite) ck, X> > 0 for sll X¢ 0, (© (©™)1E and ¥ are C™ teks with domain A then hep, 2 Mypersrteces of R* a KEY, = ty V2 be aC fanction on A When (c) Is replaced by (2) (oon-singuan) =O forall X implies ¥ = 0, then it» somiRemannisn or pond-Rinamien) maslold. ta tle can, the fnctionel ©,» called tele rd, he mac ‘toy, the Romanian mei, ote ifintasinal mare oa, Retce the mrd Soest the receding sane! fering ta ete fenton (istance function) i the opsogcal sence. In Chop 6 te Simeon ofthe concopt i lied, fi ase castmay Wo xara = Cenemanian san tobe Macedon however for an te local diferent! goonety i enced his is nelevent oe 1D ins C™ connexion ns com-Riesennlanmenflé hen D tsa Risnsmion ceoexion ii elie th tere pope () ti (ls Caper 6h sere of Rinanen nes li isl eal hues asin he ne solqusens os Rienenlanconsexion i ved, In section 23 one soon tat nea hyerataces in Btn > 3) rove cnenpen of enanisn asaflde with» Rleneni cmmesion ‘Section 2.2 The ephore mop en the Weingarten men. An (n= Ssubmanifold of an manifold is celled Aypersuaface, ‘Throughout this section let M be a hypertutace of RP, let D be the naturel connexion on A, and assume Wis @ unit noma vector field that is C™ on Me Thus = and Wye X = O for ll pin Mand X in W,,. Such an W always exists locally For any pin W and ony vector X in M,, define the linear map Li: MM, by O Uw)-d,n, ‘The vector LX) lies in M,, since O = X = 2 by property (6) fr B, The map L is linear by properties (2) snd (2). ‘The map L is called the Weingorten map, and in the case of RP i hn a geomet interpretation asthe Jacobian of the sphere map Gauss map) shich we now explain.22 Notes on Ditterential Geometey Let = (ayn) 40th 9, a ved C™ ton on tnd 3a)" = 1. Thos the ag of WS" densd gna) on 1) Ri C= map of Mino the etree fot" called ophore mop or Guten map IX eo in aa iting X (Go AO) and TH) 9), tos none AB jana, 2 0) and 7 ' rnin? oy ‘The map L is C™ on M inthe sense thet if X 4s Con the subset Af W thee LOX)» (Ray, X4,) is also C™ on A since exch a, io Con = Lexy WP awl ee EON ce ee neighborhood U of p and let U be the associated coordinate neighbor Ain), where gm Cel eed ng ye ee ‘in U. Let us assume the coordinate map ¢ maps U onte Cm 2, for p al 2, le) = Ou Jip) forall i. Thee if 74) 1818 Be Cyan) biel ge This mop hap, 2 fypersices of Re 2 (rich iC) ndocon C™ map Do U by om Bowe Late tine Z ~37°%E, oo 70") the eld Z is w C~ extension of 2 to D ‘Actual the shove procirs allows sto enend an Rad Z that a C7 on U toa Geld Zon. Having the existence of such extensions we prove » proposition. PROPOSITION. Lot Z and U be specie! ncighborhoods of pox shove and ict Z and Z be C™ fields on U and U, respectively. Then Zis an extension of Z (ue Z, » ig(Z,) for in Dill ZD)y,~ ZU) for all {1n CG, R)_ IR and'Y aro C™ extensions to of C™ fields X and Yon U, then (%, Flin @ C™ extension of Lt, Yl. Prock IZ, ~ ig(2,) for p in U, where i: M+ Ri the inclusion, thea for fin C20, Bye) = Bt UilZ, M~ Z, {C0 1) ~ Zl oh Conversely, if the tw extreme tert are eas then the second equality follows. For the est ofthe proposition, consider for pin U UW, Fp, 7,0 X70), YD) =X, 01Ulg) = FX y) = OTe thus OF, Fis an extension of (X, ¥ THEOREM, The Woingerten mop ie celfadoint Proof, Teke X and ¥ in M,, ined X and V in C™ leds on a special neighborhood U ofp, nd extend X and ¥ to C™ telds Fon as above. Then sod <0, = LIS Du, Yo, Be2A Notes on Ditterential Geometry since X, i, Poo Kh Y>= 05 ¥ <¥, Xo// ‘The fundomental forms on W cen ow be defined in temas of Land the lanes product. HEX and Y ate in My tha Id, ¥) - ok, Po, MUL, ¥) = SLO), Yo, MI, F) = , VK, P= LAR), Yop te od these forms are called the frst, second, tind, ot nndon ‘mental forms on Ml, Notice W is a Riemaasian manifold nity ronae fensor defined by the fist fundamental form. Since the Inner rodect {s symmetic and Lis self-adjoint, the fundanientel forms meat aymmetic bilinear functions on My W, forall p in M. These forms are C™ in the tense that i X od 'Yate'C™ fields with domes A then is g Creal valued fancton co A Je first thee fms have's dict interpretation geometrically since {L tepresents the Jacobian of the sphere map. ‘The algebraic invariants ofthe linear map L nt each point now doe fine the imbedded geometric invaviants ofthe submanifole M wt exer ‘int. Thus the deteminaat of L at pt the tea! corvrure (omen cuevatare) K(p) of W at p, the trace of Lat pis the moan carvoren, HG), ste. The eigenvalues of L are the principal curvatores cod the igenvectons of L are the directizns of curvature or wincipal eect, Since Li selfadjoint there are always (n ~ 1) independent ducers curvature, IL is multiple of the identity map on M,. thers ie sa umbilic point of M. If L = 0 at p we call pa flat poick of M, Nene zero vectors X and Yin W, are conjugate if ~0. Avectas Got sot) is asymptotic it itis selbeonjugnte, howell LX, he 0, ‘A curve In is line of curvature i ts tangent in «principal vec at each ofits points. The following facts come immediately from these definitions, An asymptotic direction X is a direction of curvatute iff Lk ah kc ‘onlugate to all vectors. Conjugete dieections always nist since if EX £0 then there exists ¥ whichis orthogonal to L. Ifthe second fundamental form is postive or ngstive define a0 sepmptoie {itections exit. IF aad ¥ ae two directions of curate belongs ing to unequal eigenvalues, then X is orthogonal to ¥. The pros ot this is standard algebra, ne, O- wk, = Bh, By Chop. 2 Mypeesurtaces of R° % so fy 2h; implies = 0. 1X and ¥ are non-zero independent vectors with LX ~ AY and LY = — BY, then the vectors X + ¥ and X= ¥ sre orthogonal asyaptouie dizections spanning the same rub- space as X and ¥. Fisally one notices that E mus satisfy its chore acteristic polynomial, which will also give « relation between the fundamental forms, ie, fm = 3, then L? = HL. + Klidentity) 0 snd Mm ~ HUI KI=0. - When X isa principal vector, the Weingarten map says Dy ~ kX, vwhete iso principal curvature, and this equality is classically called the formola of Rodrigus, ‘Another classical concept isthe Dupin indcatrx ot each p in M Uthich is the subset of M, consisting ofall vectors X such thet Gon eee etn 3 and et X ad ¥ be nt obogonal pcp econ in Mth LEAK nd LY “AY. WZ ak BY, onl, Doha Bi Thane ndestic ete cue (orca a Ws en a ta! shine 1. Conder te tee canes © 1K()> 0, then and Fhve the same sigs or K bk = et so suppose they are pstve The nest hen felipe dtoned by Kav = Ivan p san ips ©) A8KIp) <0, the band have opposite sgn, he adc isto ype, dp Dyers po @) AEKp) =, ay = 0,8 0 then B= ven tno steht lies pref the nest, née wp in (When k=O, pian ui nd fat pit) ‘There is © geomettc interpretation ofthe indicatlx as an spproxie ‘maton tothe intersection of the sutface with a plane whichis parallel and close to the tangent plane; for details see Struik (p84). Section 2.3. The Gauss equation, As in the last section, let M be a hypersutface of R®, tet be the natural comenion on R®, Jet be unit somal fied that is C™ on M, and fot LOX) = DW for X tangent to Ms Let U and U be spect coordinate neighborhoods ofa point pin M and R® respectively, let Z be a C™ extension to U af a C™ field Z on U as unale2% — Notes on Dillrential Geometey IY iso C™ ld about p in M, and X in, define Dy¥ by @ oy -Dyy-4x, om, This is the Gauss equation, First notice DY is in M, for = Dg, No + Deh, Yo= KY, WoO. ‘since <¥, > =O in neighborhood of ps Next notice Wf X, ¥ are C= ‘on B, then By¥ = Dz, and W 4 VULK, ZW + LX, DLP) = Dig yz +L, YD, 259, quoting tangent and normal pats to zero gives (©) D D,Z~D,Dg2~ Dig, p= . Prool. Using the Gauss curvature equation (8), RU, YOY, X= LY, YOLK, HS — ‘depends only onthe metric <, > and the connexion D, and itis com pletely independent of the nmmal W or the map L» ‘Thur the total curvature KG) = RU, YOY, X> isan “intrinsic” lavariant that is ine ‘epondeat of the “inbedding” (i.e, of snd L.). The theorem is ‘generalized in Chapter 6, Section 2.5. Examplos. See Figure 2.3 for eketches of (1), and (3). 1. Lett be an (2-1) dimensional hypoiplane in R*, N= (ayn) determine @ constant unit somal field on Me ‘Thea LO) =D = Kaye Xa,) = Ofor all ¥ at all points of M, be, L= Oon all of M. Thus W consists entirely of flat (bic) points, the total curvature K sod mean curvature (od all thers) are identically zero. All the fundanental forms, except the frst, are completely singular. Every vietor is ssyaptotic anda direction of eurvatre, and all priacipel curvar 2 Let M be ; the anit sphere about the origin in RY, and let 9 be the outer nomal on §, 1. i P= ayn) then Me) ‘Thus the sphere map 9 isthe identity map, ny also the identity map, and hence L(X) = X for all X. Thus K = 1 H=(n~1) 00'S. All the fundamental for are equal to the et| Notes on Differential Geometry first fundamental form, all points ese umbilic, and all principe curvatures are unity. Every vector is direction of curvature ‘and there ate no asymptotic directions, 3, Let i be the cylinder C= [(tynty) in ROE SIM)? = 1) with N= the “outer” nor f= (0, Onn0, 1) we have for X orthogonal to, ‘and tangent to C we have LK X, Hence Kw 0, Hf (a2), al principal curvatures ere ‘amity except ane which is zero, ete. Fig. 23 Pleces of Ksampler (1, (2, (9) 4. Next let be an open poce of srface of evolution about the 2, ats in R? (opuely M i obtained ty revelving a C= Plane curve bout sn ais inthe plane) Let be a plane cone taining the 2 aie and ako in WP (and et ua eoaider a ‘int m nt on he ai tft) Siace the norsal ties in P, the vector Dg = 1X) es P and tore, whooe Xs the uit tangent toa meridian carver Prom the semstsprceding the examples there is diction of ern {ure orthogonal © X, 2 the unit vector ¥ tangent tothe peril curves it direction of curvature, The vector ild DX ie 2270 rothognal to X and must i in the plane P, hese DyX 2 FyM, 90 Dg ~0, and we se the meridians ae peo desis. I the paralfel curve trough mi © geodesie, thon DY ss oxnal to W ood not ze, since these carves te nt frodesics in R?, But D,¥ is onhogonal to e,, the» diction, hence parle curve i «geodesic on Wi he normal slong chap. 2 Hypersutaces of R tangent to So L(Y) «= BX and 1 a a direction of eurvae 31 the parallel cuve is horizontal (Le, orthogonal tothe =axis) If mis a point oa the 2axis, then every direction X is tangent toa meridian aad hence is @ direction of curvature, 20 m is lambiic and KGm) 2 0. 5. Let us apply the anslysis of example 4 toa tons, ie, let Mt bo obtsined by rotating a circle Cin the x, e-plane about the ‘axis whove wo assure the ciscle does aol intersect the = taxis. Thon the movidans generated by C are geodesic, at is wn |e the minimum length parallel A ond maximum length parallel B. ‘Along 8, M bas positive curvature, along A the curvature is regative, aad the curvature is ze onthe extzeme top and bottom E end F where is constant. Indeed, ifr, i the radius 1) is the radius of A and ¢, the radius of B, then a= (1/26, of Cand32 Notes on Dillerentiol Geomotty ‘Those expressions canbe devived follows. Let Xb the lt tangent eld ow cielo aus toa he ii i R* (ove Figure 2.4) 20K) = eam (th sin WP) parscterine the cise oft X. Thon evaluating i enter real Wee Ao gives of) ~ (om, sin W/o, (AN ot) = WUAR, oxi he cet ties ona surface then ne see Dot U0) = U/9K. Now aly this tote ices 0 the toe 6. We discuss mle surfaces and developabie wif aces billy, A raed strlace 18 «two-dimensional ubneld Wo Roch that trough ech point pin W there pease a segnet of « ‘aight ine he generator though) hich teem Men the somal field pre fad in alg the generar, (Gs th tangent plane is constont slang penetra), teste raed satace ie dovelopabe surface, Notice we ely cosider the C~ case, although he above definition can be posrloe, Let bea rated surface, and let X be a C™ vit vector tangent to the generor at each point of M. The generators ate geodesics In R?, 80 DX =0, andhence, from the Gauss equation, DX O and ~ 0 Go generators are asymptotic lies). Let Y bea unit ‘ecto field orthogonal to X in the neighborhood ofa pat, then K~ ~ LX, Yo? =~ LX, Yo" £0 in this noighbore hood. Thus a ruled surace has now-positive cuvatire. Porn dete. opable surface, 0=D,N = LX so K = 0, A theorem do to Massey (ce Cheptct 3) states closed connected surface is developable if Ke. We study the neighborhood of a point p in a ruled surface M. Let A@) be the C™ curve though p which is parimeterized by are-lngth and is orthogonal to the generstors at each point. Let T be the tae feat tof (aay T ~¥ along f) and ft KO) =p. Then the map (es) > i) + oX(0) gives a coordinate xystea Irom « neighborhood of (0,0) in R? to a neighbortod of pin M. Fig, 25. Ruled Stace Chap. 2 Hypersurtacos of R* 3 Let be a local ent somal for this coordinate neightothood, The nit GeldsX, T, 1” give an orthonormal frame along f, and we next ‘obtain the Frenet formelas for this frame. On f we have 1a, 12 TB, WS 200-7, X= 2D,X, HS Amplies BX normal to X. Similarly, DM sora to W and Dp? sormal tT, Thus we define factions (0, MO, 0) by where a= = TT, 1>~ <1, DyA>=—, ete: Hol ing » constant, we eet a curve f(2)= (+ SKU) on M with tangent AaT4 2D ,X—(1—an)P + soit (Gote this T() and NGO are vectors at K() which re rigidly translated in R? to f,(0 to give ACO). ‘The tangent space along a generator is spanned by A and X (and A is orthogonal to), hence this tangent space is constant along a generator iff =O. The Function o/c? + 2°) Js called the distribution parameter and it ie independent ofthe pasti- ‘cular orthogonal trajectory #(ehich we show lates). Thos (3) it developable, (6) K = 0, (6) ¢ = 0, (@) LX ~0, (ote = — = ©), and (0) DX is tangent to, ate all equivalent for M closed snd ‘onected (asstning Massey's theorem). ‘Assuming i i closed (and ruled with ¢ 0), on each generator there exists «distinguished poiat called the central point, and these. points determine the curve of striction on the surface. Fixing two eneratos, say fort, <¢,, we compato the length J(=) ofan orthoge ‘onal trajectory botwoen these two generators by TOd=" (ea, aye3 Notes on Dillerentiat Geometry f= 20s + ates creer Let us find the value of » which minimizes (s), and we get (a) <0 it 2a Ma? +e%e 20 ore = ala? +8) at ty as tty. Hence the curve of stricto ie the curve Fy lola? ex 25 a function oft. This is prcisely the point on each generator here the taagent plane is nome! toD,.X() since D ,X is orthogonal to X we know, ond 0~=~ a's ate +07 again gives s=a/la? 64), Ava problem we leave the formula forthe curvature, KU 2) =~ e2/ Bon 4 ae? +0252), fd hones the central pont on each gontatr is also characterized 25 the oint where Kis a mexinan (if «misimua). At he conta point, == (a+ c#)/e%, which shows the distabution parameter, Mea! +c) depends only onthe gnertor, 1a Opives the cesta poat on w peticular generator, Len, wo take our orthogonal curve f fom this cent point, then Dok te somal to W at «= 0 and a= 0. Thus the dsttaonparcncter b= Veand Kl) =~ c4/b s ota)? =~ pip + a?) Along this fenerator A= Ts sc! whore T and V ae vectors athe eental pointy heace the norms! Ms) along the genetter is given by MO) = 5c 4 NAL oe) ESTs pIlp? + 229, ‘Thus if 3s the angle between the normal N(s) and the normal N at te central point, we have tan b= 4/p, iu, the tangent of i dinectly Propoctional fo the distance rom the central points ‘This is Chasles ‘theorem (1839). ‘This also shows the tangent plane tars evenly (chap. 2. Hypersurtaces of R® 8 through 180° along 2 generator (timing through 90° on either side of, the centzal point). For references, sce Stik (. 189) and Willmore G10, We point out we could have viewed the rled surface discussed shove as being generated by the curve (0) and the field X(0) alone the carve. To generate aufaces in this way X need aot be orthogonal to T. Indeed, incase D.zT £0, then wo generate « surface via (8) —1f0 + 910), for small «> O (or small « <0); which we call the tangential developable of the carve f, which is the ede of regres: sion of these two surfaces. It isa surface, since A= T+ aD 7 is lndependent of X = T (for # £0), and the tangent apace along & gene ‘erator will be determined by T and DyT for alls; hence the surtace developable. It is, of couse, nt a closed surface in general (Gee, Stik, p. 66). “i Fie. 26 ypototid of Reveation Section 2.6 Some applications, Let W be o hypersurface of RP with unit normal W = (a tthowe cach 2, i¢ a C™ function on M and Sa? 1, For eny tin Ry et Mm fp eg: pin Wh. Thus if p= (pyvtg) iin My then (9) = Pest, (5, + a,(PhenPy +1050) is ia My. The map fis called the nafural map of M ato Mand ff is univslent, then M, is © por lel hypersurface of M with nit normal se, Mg, = Mfr all In. Let L, be the Woingatten map ca,36 Notes on Dilerential Geowety THEOREM. Let fM-ol, we jast described Then fe Xin My 160 =X + ALD, LAbX) ~LUDhvandf preserves principal de ‘tions of curvature, umbilie, and the third fandamental form. Aso GK, GY 10%, YY 4 Dee, YY ee? ML, YY, where I Hare the first, second, and hind fundamental forms oa MIL Rie principal carvatre of M tm in dicction K, then 8/1 +8) is the corresponding principal caratare of, ot Fi) ia direction 1a rool, To compate fk tako a curve ot) = (b(Bhmb() with X~ ((0)an 80), and compete the tangent to fe Mek) = (3, Demag lh thon Fo) ~ Go) +a, a8 ite tangent at =O is indeed K+ (2). Alzo Mt) = A> ek) fem the detinition of Fand W,- Thus LIX) = Dg¥ = (at Dyaonah(O)— Dy aN = UGX) This Shows GK, 1) = Lif LyfQ¥> Lk, LY> INK, Ye [Now lot X be a unit vector atm in M with LX = BX, 60 (0X) ~ LX = AK and (4X =(L p08 K. TET =O, thon FX =O and LUX) = 1X0, s0k =O and 10, dus 14 rk 20 EM, n'a hypereurlace Hence L4(,X) ~ Q/(l + rig, which shows {preserves dvetions of curvatire and unbilics. Finally, one can verify the expression for by direct computation wing {4X = X + eLA.// Corliss. Inthe hypothesis ofthe sbove theoren let n = 3, end fet the total eurvatre ed mean carvature of M Cand M,) be denoted by K (and K,) and Wt (and). Then Ky 2RML 4 6 PR) and HOE 208) /b ls 2K), THEOREM. Let W bo a connected hyporsurtace in R® consisting catty of unbilcs. Thon W is either an open subset of «hyperplane ora sphere. IfW is closed, thn M ix a hyporplane or a sphere Prools Take pin M and X, in Wyy X_ £0, Tnbed X, in a C™ Seid X show pwd let ¥ be dy other C~ Geld aout pv, aad Chap. 2 Hypersurtaces of R 7 Y, independent. Let L~ bo the Weingarten map where fis a C™ seal valued faction on M and, 1 is the identity of cach tangent space. By the Codezzi-Mainardi equation (0), 0 = D(O)-D, (08) ~ hx, Y= Ur, 0¥, -7,9X,, sine Dx¥ Dr «LY. The ndepedeae of X, al Y,inglon =0." Since M is comected, { aust bes constant fanction on M Goten 1. ‘Suppose L = M, Ais constant on MI ~ 0, then L = 0.00, £0 is constant on Mf Dg = 0 forall Xn M,) ad A ust be an ope subset of a hyperplane. I. £0, then we may ascume #>Oby changing the sien of W i snocossary. Let r~— 1/k and let fs M+R by) the preceding theorem, fora in My, (8) RIX = 0. Thos fy = 0, and siace tis connected, (is constant smap. Let c= p~(1/8)%, for any p ia Me Then all points of W one 17 enits omc Th Hin on den abet a phe dat ot radius 1k Probleme 2B, Let f bein “UR, R). Let M be the graph off; hs M = UG, 9 a Min A Lat WP 9 ATE sd let WWE, -f, 1). Show X= (1, 0, fSand ¥ = (0,166, pan tall and i eit crm that nC oom, Let Eck, £2, F =X, Yo, and @=<¥, Yo. Show 1) = I, 6-1, PI + gE — bP) UY) MU, 6-1, PK 6, Bf, FIP) KW, ny He AU, 6.465,,8 21,7). LK, By byy = SLE, Ys by EX, LID, 6,5 ~ LK, LY sd oy38 Notes on Ditlerentiol Geometry Siow K = (deb IBC ~ = lt ©, EG — FAIA = OF, — B,, 6/288 Show L2 ~ HL -+ Kl" O wheres the icy 18 (Cotpare wi section 34) Comput the invariants fr the ihe Rico (a aed race) sich is the ise of the map ds R? — RP defined by gh ¥) = (eon, asin a for a> 0 Show « cutve on @ surface is line of curvature if the su face nomels along the curve forma developable surface oe) 3. Surfaces in R* ‘Throughout this chapter, M will onote a surface in R3; isa twonimeasional C™ submainfold of R2, Lat be a > uni nonmal field on M (Such en W always exists lolly). Let D and D be the astual consexions on B® end respectively. Let L(X) = Biv be the Wein for X tangent to.Me Let U be the set of unbie lies on and let V= MU. Let K and H denote the Gauss curvature (otal curvature) and mean curvature fantions on Mf respectively Let hand & be the principal curvature functions on W where Ap) > Kp) forall pin M, Thus KIp) = det L, = Mp)M6p) and HUG) trace L, = Mp) + Mo) for pin Me Section 31, Smoothness andthe neighborhood of « nan-umbilic point. ‘The frst theoms establishes the amothaees ofthe inveriants of W andthe local existence of orthonormal principal vectors on P ‘THEOREM 1. The sot of umblice U is closed in My a0 ite com ploment V is open ia Me The lanctons K snd Hare C™ cn The Tanctons b= (H+ y= AR)/2 and b= = YH? = RY? ace C? on M and C™ ca V. For any p nV thre ie 2 neighborhood A of p with ACV and an exthonormal C™ hase field of principal vectors Ae root. For any m i M, lot B be the domain of «local cooeinate| system. By applying the Gras-Schmidt process tothe coordinate vector elds on 2, we obtain an orthonormal C™ base field Z, W on B. Since L is C™, the vectors L(Z) = aZ = BY and L(W) = 62+ oW sre C™ on B, and heace the functions 8, and c are C™ on 2. Thus K= ac? and =a + cae C™ on B, and hence, on 7 Chap, 3 Sirlaces in B® » ‘The eigenvalues f end k must satiny the algebraic equation A? — HA + K=O associated with the characteristic equation of L. Hence ‘we get explicit global expressions fo band by the qundrstic forma, and thoy are clearly contisuous, since they ate the componite of cone tinuons functions. ‘The set U is precisely the sot where h = or 4K =0, so by continuity, U se closed and V Is open. Since H?— 4K > 0.08 V, the functions h and k are C= on Ve For pin ¥, let B, Z, and W be am in the first paragrophy with BV, Wo distinguish two cates: (1) it Hp) £0 and (2) if Hp) = 0, Incase (1), choose the neighborhood ACB auch thet 6 £0 on A and et ¥" = 82 + Gh a)W and X*~(a~ANZ 4 BW. Then X Y* axe C= orthogonal nonvanishing letde on A with LY" = AY" and LAY = BX" [Let X and ¥ be wit folds in diectins X' and Y', respectively. In case (2), suppose alp) > cp), choose 4 CB 20 a> con A, sd let Ya 0 BW and X= 82 4 (o~ BM, en// Jn the next theorem we derive basic expressions for stying the neighborhood of a noo-umbilic pont THEOREM 2. Let m bo @ nowumbilic point on M and fot X and ¥ be an cxthonxmal C™ base tld of principal vectors on tho eigh= orkood A of m with ACV and LX = kX, LY ~ AP on A Defining the C™ lanctions a and b oa A by = ORV B) and (xn 8), then DY = ak; DyK = BY, DX =-2¥, Dy¥ =-bk, Wy h< ak a7, ant Kat 2H Eh W- K2Kh — XR) » OY — 298 oe oA. Prook, Since =1, <¥, ¥>=1, and ok, ¥>=0.0n 4, Xo, Yo=2D,¥, Y> 20D ,¥ = aX for sone C™ fection 2, which ‘re compute below. Sina}, DY ~ BY for some A Also 0 — XK 1D = 2D 5H, XO and 0 XCK, Po = =-U18)~ (Ya) - 9? ~ 63, and the final ‘expression for K follows by inserting the formulas for and b and compating.// Covttry. I m is 2 nonunbilc eiest point of both gincipal curvatures, then KU) = (Ch ~ ¥2R)/0h 8) TEM bas no unbiies| Section 3.2. Serlaces of constant cureetre, Let bo closed connect surface in R? with constant Gases ‘curvature K. Then Mis sphere, a developsblesurfce, or doesa't ‘exist, aconting as K > 0, X= 0, of K'< 0 respectively. The cases when K > 0 (ue to Liebmane) end K <0 ue to Hilbert) were solved tis amazing thatthe case K ~ 0 (ae to Massey) was solved vntil 1962. of Licbaand follows from 2 ame de to Hilbert. LEMMA. 11 is a positive constant on M, thea h cannot have @ relative maxima (and kconaot havea relative minima) a ay eh tumble point. root. Suppose m ia ¥ snd m is a relative maximum for hand & relative minimum fork (since K- Bk = constant). With the notation of ‘Theorem 2, X"h <0 and Y7K >0 at Thus by the above corollary, (a) <0, whichis costadiction.// [A theorem of Bonnet, rove ia Chapter 10, shows the “com sssumption in the following theorem can be replaced by “lose ‘THEOREM 3. A compoct connected surlace in R3 of conatant positive Gauss cirvatire ia a sphere. Chap, 3 Sortaces in R? a ‘Proof At all plat, the pincipal curate h> VK, since b> ‘k= K. Since W is compact, h must have an absolate maxioun m ia 1M, and m aust be unblic by Hilbert's lems. Thas Mim) - Ka) ~ VR, and hence < YR on M. Thos h~ VK, sll points ace bil aust be a spbere.// ‘The preceding theorem can be paraphrased by saying “a sphere canst be bent.” Fora precise interpretation of this phrase, soe CChepter 8, where s generalization, th rigidity theorem for convex fodies, ix proved. ‘A proof ofTlilbert’s theorem stating that a closed connected su face with constant K <0 cannot exit in R? is in Willmore. Here ‘io, the compact case is easily disposed of by the fet cooary ofthe following theorem; indeed, no compact M exists with variable Kon. there i ‘THEOREM 4. On a compact surloce Kw)> 0. rook. Lat xp) ~ |p give the distance fom a point p in R® tothe ‘origin. Then ro is continuo function oa the compact surfece M ‘0 t takes on a maximum at poit m in M. By a rotation (orthogonal ‘ransformation) of R, we may assume m les onthe sans (or 1 -axi) Let W'be © wait nomal to Mon a neighborhood of with W.- (0,0. Let X be any nit principal vector at m with LOX) = Bw — ax. Lat ei) ~ (A, (0, be C™ curve on M with anit tangent vece to X a ¢= 0; thus X= (AO), €0), MO). Since m iz an solute maximum of 1 01 roi on My AMO) <0, Letting X be the tangent to ©, we have at m, DyX = U0), 4°), (0). Decomposing this veo ‘or into tangent and normal componeats, we get, bythe Gauss equation, “LX, IW ~ (0,0, 4) ~ (0, 9, B10), 20 4 -B*(0)> 0. Since all principal cuvateres are greater than zor atm K(x) > O// Notice the theotem is true for any compact hypersurface in R° with | tev modification of the prot. Corollary. There is no compact byperustace in RP tive Gauss curvature at all points, Corottay. There is no compact minal (Hl ~ 0) surface in R* Pro ICH ~0, thea kh and K =-8? < O.// point m with th now-posi-4 Notes of Dillerential Goometey Before considering the case K ~ 0, seca that a generator on a surface W i 8 steight lie in R? that les on I wih the normal to W constant along the line. A developable surface is ruled surfece withthe normal constant slong the ling lines in the surface. Ia Aovelopabiesurfece is closed, the it has « generator though each ‘THEOREM 5. Lot W bo a closed comected surface in R? with K-00 Thon wither M is» panes or Brough each point of ses unique genoeator od all generators are parallel in B2, Moreover, the mesn curvature is constant alo foneratrs, and hence ‘he toandry of te uabilie net ie a union of those generators root, Supposing Mix ot pan; thea the set is non-empty. Let A be a connected neighborhood in ¥ ax described in theorems ‘and 2. Since Hf does not vanish on ¥ snd A in connected we my sssune H=4>Owhile k= Oon A. Theoren 1 gives an ortonomal | pair of C™ fields X and ¥ on , with LX = Osnd LY HY on A. Since | Yk~O6on A, veering to theowen 2we have a =Ocon A 500,70.” nd DyX =Oon A. By the Geuss equation, DX - DyX ~ with G, CH, We next show G, cV. (nthe nighboohod A ofp by theorem 2, 0-228 200m x91), ae 8 Hence, if sis the are length on G, in the direction X with» ~ 0 at then (U/H) = co dand = 1/len +c for pots in G, “A If thete was an umblic point at! on G, then Hs!) = 0. Ai = Jaf {a's 1s umbiic], Ale") ~ est + d) £0, since i continuous, Hence there ate no unbilics en G,, G, CV, and to avo om impossible Singulesity in Hf at © ~-c/¢ follows Min constant on G “After extending X and ¥ along G, by fetting X be the unit tangent {0 G,, an overlapping neighborhood argument will show X and Y toe ‘san pencil vectors; hence L{X) - Oand Li¥) - HY on G,~ ‘Then Div = 104) =O implies in content on Gy, 0 G, is @ generate. chap. 3. Suxtaces in R? 8 the neighborhood A, siace His constant in the X diecticn, by theorem 2, D,X = 0, and so DyX = DyX ~ = Lik, LY> = Flim), Y> for all X, ¥ in Mfr all in MHence ~ 0 forall ¥ £0 LAX'= PX forall X. One always has L — H+ KI =O, where 8 the identity ep; hence HL = (Ks FY. W Hm) £0, then m i an umbitic and Rw) ~ HC@)/4> 0, Im is umbilic and Hm) ~ 0, then Kim) = -FUm) <0, ‘but at an umbilic Kim) = Am) > 0 abrays, Ths the umbiic set is exactly the se of m whore Ma) £0, and hence U is open snd closed. Since W is cooseeted, either W = U and W is @ phere (F> Ora plane) of M = V, HO, and K =P <0,MH Woten on Dittrontiat Geometry ‘The lactassortion ofthe theorem follows from a copllary to theorem 4.// Section 3.3. Paratelsuraces (normal maps). ‘Let us state standard hypothesis fr some theorens (end problems) on *prallel surfaces": M isa closed connected surface in R? with (C™ anit nomal Wy, ¢is non-zero veal aumber, and is the map M+R? dined by Mp) p+ el, (zee zection 2.6). 1 ‘THEOREM 7. With che standard hypothesis, if is strictly come formal, then Mis «sphere, plane, or has constant moon carvatire HBr with no milion, rook, From section 2.6, 1 X i Ma, then Since fis stietly conformal, there is on M with LK, Yo = Fm ok, Y>= I 6 ehX » ALAN, YO for all X,Y in My forall min Me Hence, 222 6 ek «(FY <0 ‘ed, as always, L? HL + KI =O, 0 OF 2/0L =e 0 PVE. IH) + 2/¢ £0, then m ie an mili, an, indeed, U ~ fn nM: Wo) £~ 2/0, For ifm wnbilic ead Ha) = 2/r~ 2, then k= -1/ K=We2, K—(L~ Pfc? F/e? 0, nd 30 Fl) ~0, which is Ine possible. Thus M ~U or M = ¥, andthe ony possiblities give the emclusion ofthe theoren.// THEOREM 8. With the standard hypothe ‘scond fundamental form, thon Mia plane. rool. From section 2.6, forall X and ¥ in M, if preserves the SDK, Po fy hy P= LK, Y 4 Ys ths =X, oL2¥> forall X and Y, and hence L? ~ 0. ‘Thus the principal curvatures are zero, L~ 0, and Miss plane.// Chap, 3 Surtaces in R 4s Similar results are given as problems. The following thoore is ie to Bonnet, and the examples inthe next section show the hypothe= ti is oot vochous. THEOREM 9. Lot M be surface of constant positive Gauss corvatee K with no umbilics. Lot 1, 1/VK and 1, WR dling poralfel sets Mand My, rospectively. Thon M, ant M, ae immer sions 6f M which have constant mean corvatire VK ond YR, respoc= tively. 1EM* ia a suclace with cantant moan caret H (ncn rere) nd noa-zero Ges curvature, letting + ~-1/M yields pall set thot isan immersion of M with constant positive Gauss crvatre 7% rook The poof is « corllary tothe formulas for, and K, in section 2.6, The special assumptions avoid trivial cast (sphere ‘or cylinder) and singularities. For the first pat, f is non-singular, since for principal vectors 1X = (14 AM and Eck — LER VR LO, since there are 20 umbilics, ‘Then Hy = GH» 2VR)/(+ H/VK)~ VR, and similarly, ff ~ VR. For the second part, fis non-ringular, ince 1 yk 1 —K/H <0 ly E~H, 20 the othor principal curvature is zero and K contrary tothe hypothesis. Thea K, = K/G—1+ K/H?) = H.// Section 2.4, Examples (sulaces of revelation). ethos for computations with “parameterized” sur faces af introduced inthis section. Let A be an open sein Rand let fs A+R? be dofined by the thee real valued ot fnction f 6 ad By 50 Gla, ¥)~(le, vu), By ¥)) 40 ¥) in A. Write T= Up fg hgh where f, = AY, Ty, = Uys yg Bogle. Notice T,~ $(@/2u) i the tangent tothe ucparamees curves on iA. Let 8 assune (7, x T,) £0, whove "x ie the croseproduct of vanced calculus; that is'an immersion of A into R®. Let = (T, x 7,8 ith W = (7, x T,1 7 008 A. ‘To compite the Weingaten map L, UT) =D, =, This, ocited with W, otice UT T= My Te SWOT TYE, xT) WM Teww 4 Noten on Dillerentia! Geometry aT yy Sinilay, with sbviws values offend je Incase T, and 1, ae orthogonal, UT) =~ Jae 0d similarly for T,; hence, Pl ee ete Fate wt ee Ln La Moa Ti OE Fy Tot, TS ‘A itle more computation Is necessary to determine the matrix for L in tems of T, and T, when they ae nt orthogoa Specializing farther, let be a positive (et least C2) fnetion, and foru> O1et ha, #1 = (a 008», win, Ka) ote a tsiface of revolution.” Applying the shove analysis, one sees directly that 7, and T, are principal vector and K = 7/4 (OVP where Pa) ~'t/du. To find satfaces of constant carvture fone must solve the differential equation 4" = oKll + (021 task that is left to reader via several problems, For note details end picture, see Serie Section 3.5, Lines of curvatur, In this section we place some results involving lies of curvature, Jno curves whose tangent vectors ar principal directions of ear Chap. 3 Sertaces in R2 a Definition. A triply orthogonal system of surfaces in neigh- bethoed U of R? ia lamily of surlaces euch thot trouth each point Of U there passes exectly toe members of the fanily whose nomale so mutwallyporpendicule. THEOREM 10 (Dupin) Invesectng seraces fom a tiply exthogoncl eystom interect along line of curvature. Prools Let Sy $, and S, be mutually oxthogooal families of sar faces with unit sormals W , eapectively. Let LX =D yl, 9 uous. ‘The Geld, ie a tangent fo the intertection of Sand Sy v0 one rust show M, ina principal direction on S, and S, er LN) = 2 fori=1, 2. This i equivalent to showing L,,) is ortnogonl fo 1, and 1, for i= 1, 2. To be specific, consider Ly). Since E40) 8 tangent t0 Sy LM, 8,> = While = By Ny Np aM, Dy Noe AN, LN > ~~, Wm since 1,18 eitadjoint. Thus by syamety, es one cyllely permutes the indices, LM yy Nab = LB yy My = +L May My> = LN Ne Hence ~ O// ‘Examples of tply orthogonal eystens are given bythe coordinate surfaces inretangslar coordinates, cylindsical coordinates, and sphovical coordinates. Ancther example i provided by 8 system of confocal quadrics, ie. ho surfaces 32tx,)#/(a,— A) = 1 with 4, , Jot fu) = f"/CL~ B45) ad tet (ov) lw co vw sin v, Me) Show that the sutace of revolution detemined by for 0 0 tet ta) ~ f°" "a — o80r, Show thatthe sutluce of evolution induced by has constant ‘urvatere ~8? and draw its pragh (actin). 29. Find a surface of constant positive cuvatee that isnot an open nabsct of sphere. Show «surface is minimal (H = 0) symptotie veetors at each point. 31, Let a) = cosh “a for a> 1, and show the surface of revole: | lon induced by fcatenotd) is minimal surface (H = 0), 30. fF there ate orthogonat | 4, Tensors and Forms ‘The material i the first thee chapters was based on & nino ‘mount of structure, ie, monifold, functions, and vector elds, more= over, there ws @ stung bias on hyperartaces in Euclidean space By ths time the reader should be a home with these concepts, end before discussing general contexions on manifolds, ithe convenient to define tensors and foms. They are thre, snd they are usefly At {ines inthe past, one notices e stag compulsion to seek ot end label tensors ad nauseum, and objects that were aot tensors were ‘eyed with suspicion. In a sense, this chapter i the “7th section of Chapter 1; i is just more stracare that « C™ manifld has eutomet= cally, and Chapter 7 contaues the theme. tis hoped by braking the Aefnitions up they becone more digestible Let Whe a C™ mrmaaifld throughout this chapter, and let m be a Point Jae Since the tangent space M,, at mis an mimensional vec tor space, the theory of linear slgbra can be spied to define tensors snd forms. A prcovariant teaser atm (for p> O) or & eco tense at m ‘areal valued pulinea (elias in each slo) function on M, Moved, (9 copes). ‘Than is 2oco tensor atm if50 Notes on Dillerential Geometry alt 4, 2) alt, 2) + al¥, 2) alk, ¥ 4 2) = alk, 1) + a 2) OK, Y) = alt oP) «ral, ¥) forall, ¥ and Z in M,, snd ein R. In similar way, one defines & Vevalued reo tonsceat'm, where ¥ i any vector apace over Ry ne deed, ¥ coud be Mitel Let be the dial space of My. Thus MS i the sat of ro sald I-cotensors at mothe tof linear anctionals from My ae | 1o R, and MS is endowed with ite natural vector space structure (eu, one adds functions by adding thelr value and multiplies by ‘constant in an obvious wy). Similarly, th set of poo tensors atm, denoted by 7°70), isa vector space over RA p-ontravrint ot ‘Prconta tensor at (or p> 0) ix areal valved pliner function 00 (US). the cross product of p copies of Wz, andthe artaral vector | secs fred by pot era mi dated by T-%U,),Deie | R, (The sets of po tensor and p-conta tensors on any } W ace denoted by TAG) and THM), ceapectively,) 3 Vvalued pconta tensors are defined analogously. Finally, | po and peonta tnsor af me a (p+ d-liees real valued fone” tion on WW) and the vector space ofthese tensors is de= noted by Fur) Ip and 9 are greater thn ze, elemenis of TTevt ae called mixed tensors. Notice that « vector atm is w Toone ta tensor atm. Siilaly, thre is « special name for Toco tensor at m for itis called form at m A tensotis symmetric iff ts value remains the saa for all posible | permutations of its arguments (thus only T° ot T®? tonsors ean be symmetic). A tensor i stew-symmettic ot altrating iff tx value after any permutation ofits arguments isthe product of ts value bee fore the permutation andthe siga of the pemtation. For exenple, Tet abe a co tensor at mand lot x be permeation ofthe set (1, 2,3), Then is symnetsic if aU, Xap Xy) = aX ys Nya Xyy)= ‘Uy Xyy ) for all pemtations wand all vectors Xin.” When 71s delined by the fiat equality inthe above line, a's alienating iHf a= C1)" whore (1) ie the sign ofthe peemtation m. Thea & form a m (fot p> 0) is an alternating p-co tensor atm andthe set (Chap. 4 Tensces and Forms st of porms at mis denoted by FPN... A Oform at mie sal number thus FW) = R for any vector space W over R.A porn ie ead tobe of degeoe p. “Tensor fields snd C™ tensor fields are sow defined in ¢ way thet { analogous to the definition ofa vector fel, once 8 vector was, delined. For exemple, poo tance field on a set U is @ ping that assigas to each m in U a po tensor at mA peo tensor field (2.00 9 a0t U is C™ HEU is open an forall sets of C™ vector fields Kye 00, the fctionLaOX yng) Km) = aX CX) 1s! C™ function on U. A C"ppfonm felt on an open set Ot Caled 2 dilleretial pf on U. ‘The tensor product of covesiaat tensors ie defined as follows: i cain TW) and in T®-AW), then @ © isthe clement in T®F eH) defined by Co BHK yng) = al ren k BX ree for all X, ia W. Notice that (a, + 2;) © ~(a, 2 f+ (a,® fy 2818, 1/8) (a @B,) + la fy), and (a) )= 2 Ale ©) for rin R. However, 2 ® Bé B®.a im genera, bt (2 fy = 20 (8 @)). Thus the tensor producti bilinear and associative but not symmetric. The tensor product of contravatiant tensors or sited tensors is defined analogously, but the detals are omitted since these oducts ar rarely used in this study, i a and ae forms of degree p and g, respectively, then the lexterie, wage, of Grossman product ax ix defined to be the (p+ 9 form a= = (0/p! @)3(-1)%e@ (2% where the sum is taken ove sl permutations # ofthe set (1, 2,...p + ql: In problem 35 there isan ex- pression for « ~f thet avoids division. Notice that @ ~#= ()P"B sey 2-18, + ,)~a ~B, +4 fly, where fi, are forms ofthe same degre, sad (a"P) =y = an(Bl~y) which is proved by using problem 35. ‘To continue the detinitens in terns ofthe abstest vector spece| W over the tensor algebra TO) over W and the Graseman agora (exterior algebra) FW) over W oro delined asthe wesk dseet sums TW) =, gT PW) ard FW) = 3. PAW). By a weak divest sum, 3M, of oles over en index set fe means the set of formal finite linoar coabiaations of elements m+ must, where each 49 Ms or more precisely, 3M, = fa D2 1G@)~0 for almost all$2 Notes on Ditlrontiot Geometry tet nite sano) of elements ant then one wate =m, + Src HID m fri heck tad M-Oee} flat ne Skee nd Jace for ie etl) The tant macs tan ad the extrin podact cont ete dtseely tT) ‘04 FO, respectively, thus mang hom agra over Re MU tan open a in the meld et P90) bth sto pcos al pc tenet eden Ur a let WO) ond FD) oe Seto wal opsniy On th er Rend et, be he go tal var intone Van th the Feat of vcctor fas Ten he sve dein cane xed te the F,-modules T*-*(X,) and F*CL,) for p, q 20, where T%X,) FAX} 7F Toe aa teem on te crn are teed uminate the relation between T(U) and T(l,). To accomplish thin let nto op sot Yn Wo mei tee ents 4 Obs Hetdon Vey at omc ets ener an {ht we indent cach post of ¥ ‘THEOREM (characterization of C™ tensor). IfU in famed ‘pen eet in My thea TSU) i inomorphic T?-*X,) in a natural ways zy be 2 C™ base Feld 09 Ua et Wy, formt on U (see probiem 22). Ie i efficient to illustrate the proof for 7°? where p> 0, since the other cases ate tmalogous. Consider in T%™(U,), end let Freel ‘an clement in TM) defined by fa, DK) elo, one, Manon, a), Oa, sla or, Mal, Cae, ao, Ca whore X, are C= fields on Us Then a = for if Xe ing, then the funtion 5 elements of T-MX,), AX nk) = a 1 Kaen 20 Oe, 3,9, 0,) since is maltilinesr over Fy and each w (X,) is efuction ia Fy. Bat ais an element of T°), and notice Lal rn Ka) de onde only the vectors (Xl) and nt nthe Fels Xs Xyr Thus the map a -» @ defines da somorpiem of TN, ) onto TU).// ‘One can “oughly” paraphrase the above theres by saying thats SF grmultilinearfenction on vector fields oa U ix acteally a smooth piecing together of Renultiinear fnetions on W, foreach m in Corottany. Let U be open ia Me Let abe a map that assigns to cach famed open set V CU an clement ay in TC) wit ay in TAT AW) forall open framed V and W contained ia U. Then there is @ unique tensor a in TP“) such tat aly ay foreach framed open V CU. Moreover, im in U aad XyoyX, re in My while then Aig = pg and one us of folds and forme st min onder to evaluate oth ofthe tensors oa the right. // If the reader will become familiar with tensors end computations Involving thes linearity via some of the problems, then the above theorem end corollary should become sore natal “To close this chapter we study the wap on tensors induced by @ (C™ map £2 MM, whore M is C*n-manifld and Mis =n" fold. Because the Jacobian, maps vectors on M nto vectors on 1, it induces a map 1 of covariant tensors (end forms) on Minto ‘covariant tensors (and forms) on Mt. IE isin T®%U") ~ Fy for open U" on M', then *(g)~ g of is @ C™ eal valued function i Fy,54 Notes on Differential Geometry ais @ peo tensor at f(a) ia 1X, ia My, by where 0 = PAU"). the peco ter at m defined oa X,, (0% 9) eae laX Ufa C= cn he open sen na * 0 open at (ie Seve porphin po hs ors om and Tove he ober acest te olewr | Teac C= ons on Usb ny C= vector ld nO, and e show ha) se C™fncon on J Take mine STeT2¥ toot syten seat awh bane VEO, tad ley, TOP Sotanae stn teat fa) wr doula eC Deine eactaosnon V tby an by HS dn) onde SY otis yO.) 9, oI, ccolg or 8 0 Pe aletpvay (ve pon 2. Thon co on (1re¥X) = 30,0, 09 24° cm face ot, and since the cight ade is tion on V,(ftaX.X) is C= on V, and hence fa is C™ on U. Finally, one checks that lay + a) = fe, « Pays Py, 8 7) = (Fy) Uy), and 4, ~B,) ~ (7B,) ~AB,), where a, ate tensors of the seme dogs, y, are aay coveriant tensors, and fi, ro alterating covariant tensors. Thus f: Fl") -» PUM) isa degree preserving exterior-ale fg2bea map of the > foms a" into the C™ forms on i. ‘There ae certain natural tensors on every manifold called universal tensors. ‘These are mixed tensors that let the arguments “work on each other." For example, lt be the 1, tensor Iw, X) = W(X) fot X 10, 1H, and w in Mz. Another isthe 2, 2teasor ECW) Way Xyy X3) = ww CE), ate. ‘The 1, tensors, T?"(H), over a vector space W have eastural Interpretation, for there is « tural isomorphism of 7!-"(W) withthe up, Hom, (V, F), of Linear transformation of W into itself. If B Is in TQ), then lt B be the linear map Biz) there Z, problem 36), 22. a 38. Show T80H) is Let ain FAW) and A in FW). 1X (Chop. 4 Teteors and Forms 55 37. Bley Zz, 2, i8 2 base of W with dual base w yuesi of # (bed Problems, In these protlens, W isan min real vector space and Mis @ C™ mnanifold, Let ey, bea base of We For i 1, 3, where 8, =O iti fj and 8, 1. bate of W4, endfor @ in, 8S sm Tet w fe) = Show Wye ik a Xe 0 ne fH and tw ts Bh a bose of MOvi(e))=5,,). Ihain 79%), how a= Eh sale jh wine a remind ty values 09 #6 Fly another base, Jt ale, 63) = 2, aly), aad Bye Sasyee SHOW by = Saray omoephic to W. Show TW) has dineasion (+-9im Show FFG) hes dimension () = (a/ptin~ py. fete 0, show © BO york yagh= SEN oly iy ni where the sum is over all shule permutations for pan Lei 1Sicicporpelsicicp+ a then, © 3. Show TGV) is isomorphic to Hom g(, W), the set ofall Re LUneds maps of W ato W, vie the above map B —», end show this map is independent ofthe base Z,. Show the unie versal tensor / in T'1(V) coresponds tothe Kentity map on ® UW ojynnjey i8 8 C™ bate fold on U in My and Wy set of duel forms on U, show exch w, ie C™ on bs sy isthe Let bein CMM, Rl. For pin W and Xin M,, Jot (NX = Xk Show (dD, is @ form at pe Ixy, i & coordinate system ith domsin Uy show deja, iv the dual base to 3/2x pany 9/6, and df= BP (/ds, kde, on U. Show a in C™ on I, Show alt + g)= al dg end tle) ~ tg + el for @ in C™W, Wy puny it also a coordinate system on U and w ~ Shaya,56 Noten on Ditlrental Geometty BpUlyy shaw 8 = S70 (Oey) Saye, a 35,f dy, ld express fo, te ee of yr in Ox show at oa = oat on PW, or, i base of Wp, show there i coon Fyne, about p wth (On), forall = 40, Leta? be the set ofall onered pairs (ms w) for m Syms ron, 08 AU) by =, 90 w = Selig whic, With these coordinate functions, M® becomes a! C™ 2neminifld called the cotangent nlc of M. The faamental form W on M ie dotined by W,_,,, - Show W - Spjda, on FO). Show a ix a nondegenerate 2fom; en, ACK, Y) O forall ¥ i ia classe 41, 112 tangent to BW) at 8 = egy) Het (XD be rel ‘unbers sach that X = 39,(X)e,. Show w, eo C= iene on Bi) (see problem 22), 5. Connexions ‘This chapter is a study of «general coanexion on 2 C™ manifold, the concepts belongag tothe connenion and the different ways of ‘efining the connexion. These comenions ate histovically cafled sifine or linear consexions on a manifold, ‘The genealization to ‘connexion in principal fiber bundles is sketched in section 55, bet these generalizations will nt be focused upon in these notes, Section 5.1, Invriaat viewpoint ‘The approach to consexions tht follows is dae to Koszul and ix found in Noniza? and the fist chapter of Helgsson. The definition ‘was motivated in section 2. Let M be C™ manifold. A. connexion, iainitesimol camnezio, coveviant ciferentation on M is en operator D that easigns to each pair of C* fields X and ¥, with domain 4, 5 field DY, with dow main A; ond if Z 6e Geld on Awhile fis @ C= real valued fame | These properties imply the vector (DY), chop. 5 Connexions 7 tion on A, then D setisfies the following four properties: Dg 6 2)-D,Y 6 DZ Doggy) (2) =DyZ 4 Dy @ Py ¥ OF © Dyer) ~ any + Dy. ta point m in My depends ‘only on X, and the values of ¥ on some curve that fits X_. Fen et Gynt, bea C™ bate field oboatm, let X, = Syaiim\e,y, and Y 127bj6, om the domain ofthe base Geld (intersected with domain of P). OM), - Dye), E26). + 03,260, o gh ‘Thus a (0), Bo), and Xb, determine D,¥ completly ifthe fields D. , er known (e0e secien 5.2). ‘The existence of many manifolds with comexions bes teenie Iustated by the natural induced congexions on hypersurfaces of R. Leto bea curve in with tangent field T. A C™ vector field ¥ on ois pallet along 0 MD, ~O.on a. The cuve ais «geodesic 8,1 =0on 0. Thus acute ix geodesic if its tangent feld Ia ‘ parallel field along the curve, The following wo theorems give the ‘existence of parallel fields and geodesics. The dowein ofan index a unless otherwise specited. THEOREM. Let o be a curve on (a, S]withtangont T. For cach vector ¥ in Moca there is unique C™ field Ye) ono such that Ye ¥ and th tietd Yo) is parafel aong o. The mening P,, Maggy Moy ¥t Pil) = PC is Knoae isomorphism which ix Colfed poratfel translation slong o from oe) to of) of suneation is almays 1,58 Notes on Ditterentia! Geometry Prook, Let ty.upk, be a coordinate system about ofs) with do Y, and fet X ynoX, be the askociated coordinete fields. We define > tuntion Tob U by Dy X,= SPX, Let omap the tosees ls 6,)iatoW. HE ¥(0 isa tel ono with donain (By) then dete fanctons (0 on this dona by 70) = 34/00E AD) ‘Lat gl 2 yoo on, byl 50 TU) ~ S00) ere)» Ug), HVC is parte song o, then 0 Dy = Sait, « Se HK Thus YC) parallel elongo itt ey oy OE Sa Gers, foe ke Tyo and for 30 fa, 6,1 The condition Ya)» ¥ defines » {nol values 2), and the theory of ordinary differential equsions then gives s unique set of C™ fonctions 2,(), atisfying the atone ‘ations on the whole domain (a 8,1, ice the equations ate inet, This defines the parallel field 7(). Fort in [a 4), the map P,,, is linear because ofthe line the equations (5) Te ¢is ny sumber in fb we obtain P,,, by covering the cone pact set ola, t) by finite number of coordinate neighborhoods aad Palle! translating though each neighbotood vie solutions ofthe systems (9.// THEOREM. Let m bein M, X iM. Then for any coal number 2 thero exists a real number ¢> 0 and a unique curve oy defined on (~r, bee] such thar 8) ~ m, T3(6) —X, and ov geodon ty of Proof, Using the notation ofthe above proof, we must find C™ functions (0 that satisfy the second-order differential aystom, Sey, © Shey pe es de, ae Gee 0 ad sith initial conditions: (6) xm) and X » REj00%,- The theory of withthe F> 0 aad the Vase. $ Conesions ordinary differential eguatons provides us functions f,(0.// ‘The existence and uniqueness theory of ordinary diferential eq. ations will actualy give us much more than the conclusien of tt stove theorem. In particular, if we let olt; my X, 5) be the eucve ‘ovided by the theorem, then the mapping o fa setualty G™ with respect to alts peraneters fm Hand Tre torsion tensor of # connexion D is a vector valued tensor ‘Tor that assigns to each pair of C™ vectors X and F, with dovaty ‘A, 8 C™ vecter fold Tork, ¥), with domain A, by Tort, ¥)- DY = D,x = 1%, v1 (One checks easily that ToxX, ¥)=—TonY, X), Toe + ¥, 2) = Tosh, 2) + Tor, 2) and TonhyP)o1 Tor, Vy, whore fin Fr snd Z in Ug. Thus the value of Tor (X, ¥) ata pont m depen fly on X, "and ¥,, and not on the fields X and ¥, by the theorem tthe end of Chapter 4. If more than one eoenenion enters tae Alscussion, we write Tory for the torsion of the connexion D. Ib Tory = 0, thea we say D is symmetric, ot torsan fen ‘8s far as we knom, there is n0 nice motivation for the word ‘tosson® to describe the above tensor. In particulny thos nothing to do with the “torsion of « space curve". ‘The following definition of curvature hes been motivated in sece tion 2.4, ‘The curvature tensor of @ connexion D is « linear tansforetion valued tensor that assigns to each pair of vectors X and P ate & linear tuansfomation RUX, Y) of My into tell. We define RGR, YYZ ‘by imbedding X, ¥, and Z in C™ fields about m and seting © RW, 192 = (D,Dy2 ~D,D,Z ~ Diy.» \2)y Again we check linearity over the rng of C™ functions coe aflicints on the Hight to doterine the tensor character of B. Here, RU, Vi == ROY, M2, oad iff C%, then60. Notes on Dilfeentiot Geometry Chop. Conteniona 6 RUG, =D DyZ— WIDE ~ DyDyZ ADE —Dtae} | 6 ga 5:2, Cot viewpoint mix, YY. For local problems conceming connexion, ont can transform the roperties of D to cortuin properios of differential forms, By using fiber bundles associated with a manifold, one can also study globel pblems vis differential forms. Ne develsp the Local viewpoint here. UL, VIZ) ~ DglWPOZ + 1/21 —Dyl(KAZ + 1042) (Uk, YNZ Let D be @ connexion on an mmsnifoldW, an fix D end W thoughe cout this section. Let U be a fixed opea set (pethepa & coordinate 10,D,Z~(vxNZ ——dawain) in My ad Let ey nny, bea fined base field of independent (©™ vectors on U. Let # yn be the C™ Informs on U which are the dul base £0 eyyupe, at each point of U. Define n? connenion Irlocms w,,0n U which are assoclated with D andthe base field by ~ hy, 2 = AYN 6 NDA « (AND: = (XOD,2— (ND 2 ~ 10,052. HU, VINE — py, py = ARK, Dye, = Bw Xe, ‘The linearity of RU, YOZ with respect to ation (in each ofits ‘The w,, are linear by property (2) ofthe conaenion D, and w,, re Vatlables) i trivial to check (0% since WX aC field on U, then Dye ise C™ field, so #10) = ‘The tonsor nature ofthe torsion ad curvature wll again be vere | Dee) Is > fenton, fied in section 5.3 with exhibition ofthe classical coordinate repie= ‘The torsion and curvatare tensors may also be expressed vi sentations ofthese tenors | differential forms associated withthe base fie. Define 2eforms The conept of *comexion-reservng” map follows naturally. | Sua Ron Dep : Let i and i" be C™ manifolds with connexions D and D, respectively AC nap WM" is conneson preserving f4D,¥)~D', wa¥'| gy TR, vy forall vectors X and fields ¥. Note the right side is wellefised since (is a welldetined fold on some curve that fits 4X, A C™ . rap f: MW -~W" is geodesic preserving iff 44 is grodeste ia Mt for | OD RO Ve, = F7.,Ry cech eosin Tia comnenompesercing nop is | geodesic preserving. here the proprties of an altomatlng tensor ate checked for T, and Rue the properties of T aod R THEOREM. Lot {he « itfeomrphis of M onto Mand tet D' | “the fora wy wy Ty aad ate related by the Cartan structerl bea connexion on’. Then ther is a unique connexion Don M | equations which are equivalent tthe defiition ofthe tnion aed lor which fis connexion preserving. crvatre tensors. We merely express everything in term ofthe base Prool, Take X in Mand tet ¥ be a field about m Since fis a fleki. Let X end ¥ be C™ fields on U. ‘Then, sitfeo,{,¥ is «field about fim). Define Dy¥ = 14D}, .f4¥) The tification thet D is @ connexion is left ae an erercisee IK every geodesic g(t) ean be extended so it is @ geodesic for sl ‘in R, then the connexion D is complete = BglSe (Me,)~ Dy (Em) (Xe,)~ Bw (0%, YDe, TX, Ye, =D,¥ ~ Dy ~ 1%, ¥)62 Notes on Dilterentsl Geometry = 30kw 7) — Fe 08) —w 0%, Ye, + 36v (Pw, 00 — 9,008, 0, Byuating components, TyC8, ¥) (Bey ow Ny YY = Aw fF) = Yor X= 4, VL Since the expression oa the left isa 2form, a0 i the expression on the Hight (Caken as a whole) and indeed, it ix the exterior dvivaive dv, of w, evaluated on X ond ¥. With this motivation we define the lexterive derivative operator d on fens and functions (forts) a8 follows. For @ C~ functlon f with domain Ay let dX) ~ X6; thas dis @ (Cm Ieform on A. Let w be any C™ [form with donsin A. ‘Then de Js aC™ Dom with domain 4, defined on C™ fies X, ¥ on A by (12) au(X, ¥) = Nwly) = Poe) — wl, t We leave it tothe reader t check that the right side is linea in each slot over the ring of C~ fnetions on, aod hence that nC Yn) is defined for mie A sdependent of the fields X and ¥. I¢Fis 3 C™ function on 4, then °F = af) =O, To see this, let X nd ¥ be C* flelde on Ai then, 240%, ¥) = Kalo”) ~ Van) — aK, V1 AVE HE 0, Y=. Also note that if xyyuopt, a coordinate system on A then dx dy i the dual bose (0.9/28yans/Dty since de (0/0K,) = On,/8,~ 8, (the Kronecker delta). Now we caa write the frst Cartan stractral equation, de, Bh w, yelp By @ computation involving the definition of ROK, Y), which is com pletely analogous to the above computation, one obtsaa the second Chop. 5 Connexions 8 Cartan stracteral equation, an ae, 7 Shen ay Rip ‘Thete equations provide an alternete prof ofthe tensor character of Tard R, since they show that T, and R,, ee fora Section 5.3, Let U be a coordinate neighborhood, and let X yr, be the coe ondnate base field associeted with the system xy.--r4,"08 U- ‘Then ww, = de andthe associated forms w,» T, and R,defive functions Phe Tyne 4 Rays respectively, by Coordinate viewpoint Tyee w yh) 80 = 3,Phdty Tye= Thy X) 20 Vo ST yale, @ diy Rian =Rihy %) $0 Ry = SyRyadny® dy, From the stmtural equstions, we have (oe, 0a, Tae EK p Ky) = PL, Thye since dx, = 0, aad Rigg = a + Me) AP Hye X,) = Kym) = Kyu hy) MEYTG, PIP GD) Thy _ Ty a Oey. FROG PPA ‘which are the classics! cooeis te components ofthese tensors.64 Motes on Ditfrential Geometry Section 5.4. Dillerence tensor of twa connexion ‘The reference for this kection is Ambrose, Singer end Palais, Let W be a C™ manifold, and Jet D aed De connexions on W, For fields X and ¥ we define the diference tensor BCX, ¥)-D,¥ ~D,¥. The linearity of B inthe first slot is tivlal from properties ofthe connene fons (namely, (2) and (3)). To check the scond slot let f be C= fom the domain of X and ¥; then BCX, 1) = (KAY « DP ~ (XOY ~ yy - WY, Y) ‘Let BOY, ¥) = Sik, ¥) + AUY, ¥)be the standard decomposition of 4 bilinear tensor ito symmetric and skew-symmetric pieces; en. SUX, ¥) = 1/28, ¥) « BUY, x9) snd AGL, ¥) = /2080%, 7) BUY, 2 Actually, we ean express A in tems of the torsion tensors T and 7 Of and, respectively, fr 24(K, ¥) By DY ~ByX 4D, XT, YL, PITY, YY, PITY, 1-714, 1 THEOREM. The following statomonts are oguivatent (@) The connections D and D have the same geodesics (©) BGR, X) 0 forall vectors X, © =o, BoA Proot. (a) implies (bs Take X at m in M and let ¢ be the geodeste with initial vector X, Extend X along @ by letting X be the tangeat to 4; then BU, X)=DyX ~ D,X = 0~ 0, since gis a geodesic for both __ (b) implies (9: Let he a geodesic for D with tengent field X; thea DX = BUY, X) D,X =0 on 6; hence d ie o geodesic fr B. () equivalent (e Since $ is aysmmetic it ie detemsined by its diagonal values SCX, X), and BUX, X) = iff SUX, X) = 0. (6) equivalent to @ For B='8 + A.// ' hap. 5 Connexions 6 THEOREM, The connexions D and D ore equsl iff thoy have the same geodesics andthe same torsion tensors, ‘Prook. That the first pat impli¢e the second is trivial. Conversely, ifthe geodesics are the same, thea S = 0, and ifthe torsion tensors fare equi, then 40; hence B = O and D ~D,// THEOREM, Given « connexion: D on i thee is unique com rrion D having the same geodesice ax D and 210 torsion. Poook, Let DY =B,¥ ~ (/2)7(, ¥. tes tivial to check that D satisfies the reuited properties 0 deine a connexion. Here 2B W/2)T 4 since a torsion tenner i skew-aymmetiy this §~ 0, s0D and D have the same geodesics. Moreover, T= T= 2A 0, 90 Dihas ze torsion. The uniqueness follows rom th preceding theoren// ‘Tht if we patton connexions into equivalence classes by place ing two conrexions. with the same geodesics inthe same class, tea ‘in each class there exists unique trsiontee (ze torsion) cone rnexion. Moreover given any comtexion D and any skew-synmettic vectorvalued 2covariant tensor 7 there exists @ connexion with torsion tenor T andthe sone geodesics ax D. From the above proot wwe have Ti, ¥)= 20 ,¥ ~ D,¥, which provides a geometic inter ‘tation of the torsion tensor of « connexion a measuring the if ference between covariant difereatiation i the given connexion and covariant diferentiation in the torstonfee connexion with the saz erodesics, Section 5.5. Bundle viewpoint In this section we define @ connexion on the bundle of bases over 8 manifold and sketch a proof of the equivalence of such definition with our previous viewpoints. This is the fourth (and last) viewpoint ‘we consider. The bundle viewpoiat provides a natral “jumping off” for generalizations to connexons inal kinds of bundles, etd much ofthe research in differential geometry at this tine uses these com cepts. For more detalls the reader is referred tothe book by Crttender ‘and Bishop o the book by Kobayashi end Noirs. ‘Throughout this section let M te a C= nomanifod, let BBM) be the bundle of bases over M (see problem 22), and let x; BM be the665 Notes on Ditlerontiat Geometry natual projection map. IfD is connexion on My then by inept ing ordinary differential equations (5) above), we can pale trans Tate the tangent space along carves in M. If b= ay eyrne,) ia in B and ois « curve in M with (0) =m, then by paaltel translation we define @ C™ carve O10 = (tt €,(0hnyeq()) 0 B, where 6(0) isthe parallel translate of e, ~ (0) along o'0 ot. Since m °3 =, we say 36 “lft of o%, oF 2 Mies over 2,” and since 2 reads of «pate sllel base, we say 2 isa *horizoatal” curve in 8 Thus a connexion D on W yields unique "horizontal lifts® of C™ curves in M. The bundle definition of « connexion gives an independent method for de- ‘alng “horizontal Lite” (of curves in MD withthe corect properties, Recall at each point} in B we defined the subspace of vertical vectors Vy = [Xin By: a{X)= 0). A connexion on 2 in 2 mapping H that assigns to each b in B subspace H, of B such that @ A, -v,-0 (hence His @ R40.) = Hy, for all g in GLO, R). (G) His C% ies for each b in B there is neighboshood U and 8 sot of independent C™ vector fields EyyyE, on Uthat ive @ base for Hy for every bin U. 18 rll, 36 an isomorphism of H, onto My) ‘iH, we say X isa horizontal vector. Property (1) implies in By thore Iso unique decomposition X= Xy +X, with Xj, in Hh and Xy in Vand property (3) implies if X is C™ then Xy saad Xy a1e C™ fields. IE is a C™ Field with domain U in My then there fea unique C* horizontal field F on @ = 9°40) with 2F,) ~ Xjgy or all Bin. Having the existence of “horizontal lifts” for vector felde, one can ‘horizontally life curves in natural way. Thus if o iw curve in with tangent 7 (non-vanishing), extend T to @ C™ field is neighbor hood U of a univeleat pert of, lift T to horizontal field T on and take intogral carves of Tto find hovizotal lifts ofc. The por allel translation so defined willbe independent of the base (the tart= ing point for ) by property 2); ie. 12 ix horizontal (es a horizontal tangent), then Ry 02 ix also hereon fle 3 Cometone @ ‘Thowe is dual viewpoint involving differents! forms, To moti vate ity let H be connexion as described above and notice st each b= (it un8,) in B we can define 8 unique horizontal field £8) snth 2(E (0) 6, by (The fields Ey.yE,, are global independent horizontal C™ fields on B. Together with the natural vertical fields Byrn ng We tet & global base field on B. Let Fyn, BaF be the dual Ioforas to this base (where Wy. ar the stocal Morne of problem 41). Then Xin By, Xy = 30,20) CE, dor Hone knows X yy then, of course, Xy 2X =X ye Tha ghee int, (or giving H) is equivalent to giving “vertical projections” teach point in B, Thus # set of connection Toforms (fri, j= Ann on B18 @ Set of Lforms such thet (1) Bly, forma dual base to E,, stall bin B, 2 (0B 2) 20,4, For all Xin By, (2) 2, are C™ forall end ‘That the definition f = connexion on B in terms of Ifo in terms of 4, 8 equivalent is left as a problem. ‘Notice thatthe #, ean be weed to define a Lie algebea (of GLir, RD) valued Isform # by BON) ~ 3p, 00K, where the X,, are the ‘canonical left iavarient fields on GL(x, (zee problem 21). Finally, we connect with the Cartan viewpoint. Let ye, D6 ‘base field on the open set U in W. Define a C™ map f: > by f(a) ~(; (pyle) q) f08 min Us Since wo ff the identity of Us ine call fa soction over U. Let w,, be the connexion foms defined in section 5.2, and let F,, be the global forms defined above, Then wy =F) of on. “Thus the Cartan structural equations (13) and (1) (end the torsion and curvature -forms) can be cated upto global equations on . Problems,/Let M end M! be C™ manifolds. 42, Let x, Lx and x, =y be the usual coordinates on Rs Define 1 connexion D on R? by letting T}, =O except for P =P, = Set up and solve the dilfereatial equations forthe geodesics thru eny point in R®. Find the particular geodesic g with (0) =6 2 4, 46. 0 Notes on Distercatist Geometry Q, 0 and 7,0) = (8/Ax) + (8/47). Ix D complete? Do the fcodesics emanating fom the origin pass thru sll points of the plane? Ifo andy aze geodesics with (0) -oA0), and T,,0) = btg(0) for b in R, show y(t) =o) for ll posible Investigate the connexion D' with all P), = O except P= Let D be @ connexion on W. Let a) be an sategal cuve of the C* field X lot €,(Dsonseg(®) be a parallel base alone 6, and lot ¥(0) = 3y (de) be 8 C™ fd slong a. Show (DPN) Bley /dte (0) along o. Show (Dg¥ XO) = in (1/01, g¥ A) — veofss ¢ 0. Let {be a connexion preserving C™ map of if into M. Show 14(Tor (X, YD) = Tor eX, LeP) an 1 (BU, Y92)~ RUE, Ay xE2) A manifold W is paraelizare if thee is @ connexion D on ‘Min which parallel ttasslation is independent of curves, and such a D is called a flat connexion- Show W ix parallclizable lf there isa global C™ base field on M. IfD i a flat cone nexion, show st curvatre tensor is zero (see problem 85) Let G bea Lie group. Define the felt invariant connexion D on G by asserting all vector Fields in the Lie algebra & are Poallet fields. Show D is flat, G ix pralelincble, and if X snd ¥ ace in Gy then Tor (Ky ¥)~ -[%, ¥} Show that each geodesic on G isthe left translate ofa one-parameter sue HFOUD 0; Lt 0) = Ego ft for allt. Show D i complete, Let D be consoxioa on M. Ferm in My let H, denote the set of linear maps of M,, ito itself, obtained by parallel Uweaslation of Wf, arcund broken C™ cutves starting and end ing tm. Show Hy iz a group. ICM is connected, show H, ‘is isomorphic to Hy form’ in The group His called the holonomy group at m, end if is connected, thn the holonomy ‘rcup of Mis the group H =H, fr aay min M._ Restiiting the closed curves fo he nul-bomotope, one obtains the r= stricted holonomy group HS. If D is fat, show #1, = 0. If A isthe unit sppere in R2"and Dis the Riemannien conmexion, show that H = SOQ, R), wheze $02, R) isthe special othog- Pleas tmnt nti ° onal grup, ot rotation group, consisting of orthogonal maps ‘with determinant one 48, (Continuing problem 13,) Let X = 8/81, and ¥ = a/0n, for & coordinate ByStem xy, on M abou a and show ig (1/01 = 1K8/04,)1= ROP, XYG/9e,)~ 3,8 (9/2%,), where Fis the identity map and P,., i poalle translation slong y from y(0) to (0). Because f this, one often says RCL, ¥) is “infnitesiogal patllel translation around on inf nitesinal parallelogram spanned by X and F." 6. Riemannian Monifolds and Submonifolds ‘The definition ofa Riemannian (end a semi-Riessnaien) manifold swat given in soction 2.1. A manifold on which ane has singled out 1 specific symmatie and positive definite (or non-singular) 2-coveri= ant tensor field, called the metric tensar, is & Riemannian (or semi- Riemannian) manifold. tn this chapter we generalize the theory of Chapters 2 und 3 in 2 natural way. Mach ofthe theory applies to somi-Rienannian manifolés and submanifolds, bt, in genet, we uese things only in Riesinaian terns, Section 6.1. Length and distance ‘The metic tensor allows us to define lengths, angles, and distances. Let be a Riemannian manifold with metric tensor <)>. Let X and Y be in My. Deliae the fength of X by |X| = Vo, X.. Define the angle 6 between X and ¥ (both nonzero) by letting = (XP eos 0 where 0< 0, and notice the Schwartz inequality |X, Po] & PII makes this possible, The length ofa curve is now defined by integrating the length of its tangent vector field. Let g te a C™ curve of [a,b] with tangont field T (or T, if necessary), The length of trom ato be, denoted | by lof, ie dened by © elg= f° vera Te70 Notes on Ditterentiat Geometry The integral existe, since the integrand is continuous. The length of 4 broken C™ curve is defined as the (Cait) sum ofthe lengths of its (C™ pieces. ‘The number |o[? is independent ofthe parameterizetion offs image set inthe following gence: lot g be a ©? map of oy dl foto a Bl wth end points mapping to end points assume g(c) ~ and (0) ~ BY then FPeralo, Tele) at = fT tain Tyleen> Hartote “Tog, Tyg tet ogglt)= €°OT AAC) by the chain re. Thos we can write i= lo where q = ola) and p= of). bassically, the metic tensor ie almost alway expressed by the notation “ds? g,,drjde-* This meens one is giving the inner pro- duct on coordinate dein U with cooeinate factions ty, in terms ofthe coordinate bases: fe, if X,~ 3/0), then fy, hy” Xy> is 8 C™ function on U. IY ~ 3y,X, and Z= 32, Xp then cP, 2 Ree eudiy TH, giving the metex of functions ¢,, on U deters | ilies the Inter product oa U. ‘The *ds” only makes sense when one ‘curve o which mape into U, for then let (= ol! is)? oot, Po xg, BED A (py act, To ~ 24, Are Me is connected, = proudo-metri is dofined on M by Odom inf [lof #2 broken C™ curve fom p toa) Teivially, dip, m)2 0, dlp, p)~ 0, and dp, m) = dim, ph The tane Inequality is left ae problem. THEOREM. The peeudomotric topology on M equats the mani= | {old topology. ‘rool. (Alter Seifert and Thelfall,p. 44), Let m be any point in Mand 1et Kyau, be a coordinate system about m with domain U- chop. 6 Riemannian Manifolds and Submits n For pin U let alg) = cl, p) defined shove, end let Mp) ~ (2x (9) ‘whore we assume 4,(m) 0. Choose a > O80 4 =[p: Xp) ra «(ph Hence, (1) alo) rp, But i curve with, oft) = ty/2tp where (6) =p then oj = [" |Fg(des Raia) Hence, (2) a6 s Ree ‘The inequalities (1) and (2) prove the theorgm.// Coroftary. A connected Riomsnnian manifold W ix Housdort if the preadonmetic dis w metic, In Chapter 10 we show that geodesics are the curves that locslly minimize are length ic, the length ofa small piece ofa geedesic in ‘Wis precisely the distance between the end points ofthe piece. Henceforth we assume all manifolds we mention ate Hausdorf. A Riemannian menifold is complete i i ix complete se « metric space, snes, every Cauehy sequence must converge Section 6.2. Riemannian connexion and curvature A Riemsnnien connoxion D on @ Riemannian manifold M is con noxion D such thet, @)DyY = DyX = 1K, YI and 2h, Y= DA, YX DAY for atl fe ‘theorem of (seri-) Riemannian X,Y, snd Z witha common domsin, The fundamental Molds is the following: ‘THEOREM, There existe 9 unique Rismsnnian conection on (comin) Rieaanaian manifold. Riemannian consexion D exists and is unique jeneas implies D must agree Pool, We show on every coordinate domain U. The uniT2 Notes on Ditferontial Geometry on overlapping domains; hence D exists and is unique on all of H Lat mtg be the coordinate fields on U et gy = 00 0, sand let (@"), be the j* enty of the inverse matrix of € =, (hich i non-singular). HE (3) and (4) hold, then OK ky Xp eX ay AP Hy Kp a Dy Ky KP slace [X,, X,]=O for all s. By section 5.2, giving DonUis | equivalent giving fonctions P, with Dy (X,)~ 87.,PyX, and demanding properties (1) trough (4) of section 5.1 are valid. Thus 6) implies 23,0 85, = Nit + Nybys~ Hay hence cs Hy, (Abn, A My © Pye wassey Be + Bee Bie, Ths isthe classical exprsioe for the Christotfe function T' in tem ofthe metric tensor. Use (6) define D on U. A direct check of (3) and (@) shows D is Riemannian, andthe explicit represeatation | (6) shows Dis unique.// | ‘The above theorem is special case of more general theores | (Grobien 70). For the rst of this section let be « (gems) Riemanaien rmenifold and let D be the Riemannian conaexion on i. The Ricmann-Chriattfel curvature tensor (of type 0, 4) is the 4covariant teator KU, Y, Z, W) =<, RZ, WW> tor X,Y, Z, ond Win Me | THEOREM, The following relations are tue: (@) RO, YE RZ, XY ROY, ZK =O | () RUF, 2, W= RU, MZ) | (©) KIX, ¥, 2, W)=-KUY, ¥, 2) (@) UY, ¥, 2, WT=RUZ, ®, XT “The relation (ss called the frst Biachi identity and it hold for any symmetric connexion. These relations are equivalent tothe “synmetriee” ofthe indices ofthe classical R.,,, functions rool. For (s), ust the Jacobi identity, property (2) above, and compute, For (©), use RUZ, W) = RW, Z). For (b), ate property (4) to shift D from one slot othe other. For (), notice (0) implies (a KX, YZ, 1) + RU MY, 2) 0 KUM, Z, My YY =O. By wntog Ge) thee more times, cyelicly permuting the rguneats ofthe fist tena ‘one step from one line to the next, adtng all four equstions, and cancelling via (8) and (e), one obtsins (8). For X and ¥ in W,, let DAWN) ok, Be, rk, A WAGE, Y) 40, tet RX, Y) KOH, ¥, x, YY/ALY Py end by direct computations, using the above properties of K, one can show Rey, ¥) RAY, X)~ Ree, a¥) Rw 6, ¥) for, s and f not zoro. Thus if AUK, ¥) £0 and ad be 40, then RU, Y) = RlaX + BY, CX + dV), and we define K(P), the Riemannian cameature of the 2ndimensional subspace P of N spanned by X and Y, by RIP)» /ALX, ¥). In section 24, we showed Kg) ~ Kl) is the Gauss curvature of surface M in R®. Inthe Riemannian case, [AUL, Y)} ithe area of the parallelogram spanned by X and Let £:W ~+4" be @ C™ map between Riemannian menifolds, If thre is C™ real valued positive function F on M such that for fain Wf and all X, Via Wy, yh f4¥> = Fla)y then fis a ‘conformal (or strictly conformal) nap aad F is called the salo func= fiom. CF exists but F > 0 only, then {18 weakly conformal. I E = then is en isometry. IE 48 an isometey aad a diffeomorphism, thee fis isometric end W is isometric to M. IC is constant, then fis omathoti. ‘At this point, we explicitly call the reader's sttention to problem 52, which is considered an integral part ofthe theory of Riemannian manifolds.Th Notes on Ditlerentiat Geometry Section 6.3. Curves in Riemannian monifods ‘This section parallels the standard teatnent of cuves in advanced ‘aleulas, Let W be # Riemannian sessfold with Riemenaisa connex- Leto bee C™ curve in with tangent field V = 0/4), imhich can legitemately be called the “velocity vector” of 2 since “length” is defined, Assuming V does not vanish oa the domain of define the nit tangent vector T() = VU0/|V0)» and define the Speed function «= (ded) = |P(, <9 Vl) ~ e!(T() for athe dor Inain of 2 Define the geodesic eurature vector Held of oto be the field DT, and its length, isthe geodesic curvarare of o. Notice thot D7? ond Fy, at a particular point on the curve, do ot depend on the parameterization ofthe “point set of the curve® bat only onthe ‘rieatation (choice of *dtection*) andthe existence of @ C™ parame terization with non-vanishing tangent at the point. ‘The curve ois a geodesic (Dy ~ 0) Mf V has constent length and (@)D,T ~Oor(0)k, =O. This follows since D,V = 'D ,('T) T+ (a!)'D 7 and s* >Ovwhile DT i onhogonal to, (— 1800-0, 1)=2D,T, Toh When f,(0> 0, defn the (first) normal to @ at ok) to be the unit vector H() such thet DT = ky) at then, 0= TEN, T>= + Wy, Dy To = 0, define the second normal to at of) tobe the unit vector (0) such that Dy, y#,T = 1,8 IF, > Oon an inter wal, ten the above process can be continued (9 define, and where 1,50, one gets Ny, ete, The vectors Ty Wy Hann ar Called Frenet vectors, and the equations that express the Dy, in terme ofthe Frenet vectors are called the Frenet formulae. ‘hen if s = 2manifolé and k, > 0, then the Frenet formulae bee come DT =K,W and DW, ~-b,7. In this ease i is possible to locally Choose W, slong a Independently of D;T (oe univalent pieces fo), and leting DT = 0, woald define k,, which could take on negative values (G00 problen 72). ton D. i, i defined on an itera, | chop. 6 Riomannian Manifolds and Submenitotds 7 Section 6.4. Submonifods, “The theory in sections 2,3 and 2.4 is now generalized, Thiowgh= cout thls section et the K-manifold I bea (nonsingular) submanifold ofthe (semi-) Riemannian manifold a the semicRiemannian case, the submanifold W is nonesngular if the metric tensor ix nonsingalae tihencesticted f0 If forall m in W (thus Mis @ semi-Riemannian ‘aenifold under the induced metic tensor). The induced metric tee tor on Wis called the fiat fandamontal form on W. Let D be the Riemannian connexion on ‘THEOREM, For C™ folds X and ¥ with domain A on M (and tangent 10M), tine Dg¥ and V(X, Yon A by decomposing Dz int its eng tangential and normal components respective thes, © Day D,r¥a, Then D isthe Riemannian connexion on M and V is @ symmotié vectorvalued 2-covariant C™ tensor called the second fundamental The decomposition equation (9) is called the Gauss eqbation. ‘Proof. We mill establish the C* nature ofthe decomposition, ‘The rest ofthe proof will only be outlined, fori i 8 simple exercise, Use the properties of D (since i i¢ a contexion) to establish the roperties of D (eaking it « connexion) andthe tensor character of V its multilinear). Zero torsion for D implies zen torsion for D, and V is symmetri (ore the proposition in section 2.2, which gen ‘ralizes trvielly), Since D satisfies condition (4) (section 6.2), D does too, Hence D is Riemannian, and bythe uniqueness theorem, Dis the Riemsnnian connexion on i ‘To show D and V are C~on A, choose pin A. Let T and U be special conan doaie about gin Hand eps, with UA, and lt Zyoy2, 998 Z, ~Z yyy ~ Zyl be the coordi>. rate vector lilds on sad U, respectively. Apply the GramSchaidt process t0 ZyjonZ, on U to obtain C™ (the Gram-Sehnidt process 1s algebraic) orthonomal fields Wy, on Ouch thst yyy tdve @ C™ orthonormal base off for'm in U, while MyeignMely five l-voctor fields that are C™ on U and form a base ofthe rtho- onal complement 0 My for min U. Let X= 344M, and ¥ = SH Wy deline C™ fafetions x, and yon U fori!" and let76 Notes on Ditterontisl Geometry on: Then By, ~ S28, define C~ factions Bf Day = 31X7,W, + 39,050, whore and j= Lynd and ¢ = pvt hae Dyk 3h. y) 6 B09 885, and POG P= SLM oP, | se C* on ui? By decomposing the curvature R into tangent sid nosmal pats, we obtain the Gauss curvatre equation (10), and the CodaeeiWainaed ‘equation (11), respectively. Let X, ¥, and Z be C™ feléstangest to W with a comson domain. Writing the decomposition ofa vector ¥ se W tans nor 0) tan Fx, V2 = RUN, VIZ « tan D,VUY, 2)—B, PUK 2) and (nor RY, V2 = VX, D,Z)— VU, D4Z) ~ MUX, YZ) ++ nor DVO", Z)~B,VX, 2) Since V is a tensor, Lis Voy ¥,.) e wellefined and indepen ent ofthe fields ¥ aad ¥ used 9 compte it i the Gauss equation, we define X and ¥ tobe conjugate vectors at m if VX Y) 0. A vector Xia Wis an asymptotic vector if V(K, X) =O, and in any case, detin the asympotie (or normal) curvatre of Xy ky, by hy WOR, X). AL, = 0, them mis fat point on fis curve ia M with C™ unit tangent T, then ¥(T, T) is the ormal curvature vector field of and ky» (VT, 7) i the normal uvatire of Chop. 6 Riemannian Manifolds and Submanifotds 7 ‘THEOREM Oteusnien). AI cures on M with the seme unit tar {gent Tat point have the same norml curvature at that point. Tt fa curve on M with C™ une tangont Ty then (Ey)? = ,)2 +k lites the geodesic curvatures Kandy of on Wand M with ite omel curvature Ky. Moreover, ky =i cos & determines the angle 1 between the normal, of on Hand the normal curvature vector VF, T)il is dfined. root. The fs sentence follows since Visa tensor. The sec nd sentence follows from the Gauss equation DT ~D,T + V(t, 7) sine the vectors onthe ght ae orthogonal. Fr the third sentence, iFB, 0, then ky ky =O aed aot defined, If, > and ky 0, then'V(T, T)=0, i, fe tangent toMy nd = 9/2 (anything). 4, £0, 10. be the uit normal a diction of VF, T) and ky VT, Ty Wo = Dy, No oF, cos du// ‘The theorem and corollary atthe end of action 2.3 can now be generalized by replacing R* by Section 6.5. Hypersurtaces In this section, lt M be hypersurface in the Riemannian mani- fold i ond let he a C~ unit normal on i. Define the Weingarten ‘map L(X) = DW for X in Ny (as in section 2.2). The Gauss equee don for Nnow becomes (2) By DY = Lx, Yow since = Dg¥, No = XY, Yo=e¥, Lid> and 00, ‘Thus VOX, ¥) = ~ 1, and lo bea kesubmanifold ofthe Riemsanian manifold H- Let ¢ be a geodesic of W that Tie in My et T be the unit tangent tog let X be a unit fold tangent to M which Chap. 6 Riemannian Manifolds and Submanifotde n fs pallet in W along and orthogonal to T, ad fet P be the subspace sponnot by X and T. Then R\P)> RiP) along t, and RP) = KIP) itt His paaltel along § ia rook, We prove the theoren for k= n= 1, lesving the other cases to problem $5. Let W be a C™ unit normal on «neighborhood of « point on f and tet L(2)~D gW, Here DgT ~0 80 DT ~ 0 and =0, By the previous theorem, RiP) =K(P)" BG). 1 oqniy doen LNs To - Daw Bek = Dg Oya Serene! ‘Thete is & basic “rigidity” theorem fr hypersuraces of R® which Js our next goal. This theotem is e uniqueness theorem, and there is @ corresponding existence theorem that is proved in Chapter 3 When m = 3, the theorem was fiat proved by O. Bonnet (1867). Intuitively, thiatheovem states if two hypersurfaces in R® are Jsoneteic and their nomal are “bending the same*, then by a *igid motion” one can superimpose the two manifolds. THEOREM, Let M and W* be connected hyporsurtaces in for n2 3. Lot W and. be C™ unit normal elds on W and W',respoo: tively. Let F bea difeomorphism of M onto M thot preserves the fist and second fundamental forms. Thon there isan isometry G of Re with F- Gly. rool. Duting this proof let us use "prines” to denote concepts belonging i which correspond to familiar concepts for My ke, Let LQ) = Dy for Xin Mand L')~ Dy" for ¥ in MSs. The bye pothesis tates if X and Z are in W,, then

~My D> and LIP MY, FyZ> = LN, 2 Combining these statements, LUPM) Fyz> = LK, > for all Z which implies L' oF, =F, oL, ‘Thus the hypothesis could be rephrased as demand thet F bean isometry of onto W" whose Jacobian commutes with the Weingarten maps. Since an isometry i=8) Wores on Ditlerential Geometry conenion preserving, F,(D ,2)~D'y, ,PyZ fr vectors X and fields Z teagent tH. 7 Tp ia M, we extend the Jacobian of F to be & linear up of (R*) ato (Rs where p= Fipl Let W be in (Ry then WW,» a. where Wis tangent to fy so define ‘ FA) = FW) ay Xi in, and W is « C™ Feld of Revectors on My thon FB¥)~D,, tM, ‘howe B is» natural covarentdiferentiaton on R®, ‘This follows BW =D,W, +B la) = PM, LK, Wom aw + atx, ad FAD gM) = Dep, SFM, PAK, FAN DW' s P,Xlao FAW" 4 lao FNL ~B AF, By lao) -B, Pe, ow let oj.ut, be the usual othonormal fields on Rad deine G* fanctons'8,, on M by Fale)p = 32.6, (0Ne,)u. The fnctons 4g ate O singe, M, nd M6 C= then 8h m wae bs crthogoal since Fis an isomety. Then for en tangent vecee te M at any point pi M, we now Dye, 0, since e, ae peralel else on Thus, 9 -F Dye.) ~B,, Pe) ~ Sl, De, + 8,.Dp, 20] = E88, ley (Chap. 6 Riemannian Manito ent Submanitolde a 95, -Oforal rand». Since X ond p were ebhity sn Wi Contec, the tncions 6, re constant on Wand tha the jcoien Sr in constant enthogonhtensfmatinsltve to the atl ee nese, oth Nort dle » nup@ off ont tell which i ranlaton fl lowed by a hog nap ty eng Ap) Fp foro pn Wand eguing (Gp), ~ (Py), This tonpleely determines td th Jcoion ofc dhe eu! ete Jaron af F tall yoins Since ie comecteds FC ay Section 6.6, Cartan viewpoint and coordinate viewpoint In this section let Be s hypersace of « Riemannian rmenifold . Let p be in fe T he «special coordinate neighborhood of pin AM with U the cerespoading seighbothod ofp int, UCU, Apply the Geve-Schmidt process tothe coordinate vector fields en to cain sm enthonomel bab fel eyney on O HA (ry) 9 base of W, frm in U and (0) sora to (ths 6 provider lol normal forthe neighborhood U}. Lat f 'U = Ube the inclusion mp. ‘Applying the esse of section 8.2, let Fynan, be te deat I forms sasocited with eons, and 1%, fF 1 55) be the coe ed withthe Riomansian conexion Den 0, s0 nexion forms associ (18) Bye, = 39,8 (key 06 j= An Let Fy and 8, = Ry for 15, jm, he Wy = Ry 5 Then if X is tangent to fat mim U, by the Gauss eguoe 19) Dye, = 355! 8,08, 2) Mee) =, 06, for} = Anum 1. Thus for 7 <1 ae the connesion forms for the induced Riemannian coanection D onl. Moreover,8 Notes on Ditforentiat Geometry AY LO) =Byo, = SIH (Xe, since L(X) in, 80 Wy, -Oon U. Also w, =O. U, since 6, ix normal toi. Eqlation (18) isthe Gause equation and equation 21) {2 the Weingarten egution Let, I, It be the first, second, and third fundamental fom, respectively. Then for and Vin, m in U, 10%, ¥) = 3970 aw (P) MEY) 0 Lk, Yo = BPW, (F) MICS, ¥) = + = wy (X) + W(X) for atl X tengent to i 1.e.w,,~—w,, for connexion forts belonging to an orthonormal base (and this again shows ©, ~ 0). ‘Thus we can ito H nd I ia toms of, if we wish Certain relations are imped by the Catan stractural equations. ‘The equation <7, ~ 0 (on N,) implies I is symnetric, The equation a”, si, ~ 0 (onl) implies I ix yr R,, when restricted syncs For) 0, and K <0; L., the Riemsanisa curvature K(P) of all plane sections is « constant For K=0, le. ~ R® withthe usual Riomsnnian metric, This is ssually called Euclidean space ot fat space. Foc K > 0, let M = [aia R°" 33a? ~ 1/K), jes W is the dim sphere of radius 1/yK about the origin in RO". Tis @ Riemannian ‘manifold via the induced metic from R°"!. This is called spherical space or Rismann space, Letting W he the unit outer normal on M, then L(¥) = YX fo al vectors tangent to ad sll points are unbiic. By equation (17) above, K(P) = ~K where X end F ae unit orthogonal vectors spanning P. Since M is compact itis come plete, An altemate proof that W has constant curvature i= provided by the group of orthogonal tranformations on R"*, which proves ‘isometries that will mep any point mand plane section P st m, nto sny other poiat mand plane section P. Since an isometry preserves the curvature, this Would show M has coastant Riemannian curvature but would got evaluate this constant. For K<0, let W= (a in RP: Sto? <-4/KL. Lat xyase, be the sual coordinate funetions on RP, ist, (a) = ay let Ky ite, for 1 Tata eine a metic on W bythe fanctions gy, ~ yf 84 /A? where A= 1+ (K/4)S283. Then W with this metic is celled Inyperbolic space, ot Poincare space, Thus M is obtained by » cone formal change of the usual metric tensor on an open ball in R®, and Wis simply comected, since itis contractible, (One proves i has constant negative Riemannian curveture K by @ Airect computstion which we outline. Let K, be the Riemennian ‘curvature ofthe plane section spanned by X¥ X, ot ny point in M Le Ry AM = SA Ey «Ey die eto, ‘Then Ky, ~ APR, and compute vin the classical formulae for RE terms of Pj, aad in terms off. These formulae show I, unless two indices fre equal and I’) =F, =P, ha (tie P= Rj/2A. Thea Ry, =(K/A) ~ R28) /442 ond84 Notes on Differential Geometry Also ty diect computation one shows Ri, ~ Ounless k= i ¢= orkaj,t=i. Thes letting e, ~ AX, gives an orthoromal base 0,, sv@y at each point of M. Let P be any plane section at m in M and et yf, be an oxthonoraal base of P whieh we extend to an oxthor aoenal base of M,. Then the bate , is related tothe base f, via an ‘orthogonal matix, end one uses this fact to show K(f, f,)~K. Thos M has constant negative curvature. "To show Wis complete, let K = 82, #02 sinh BB cosh #0, 0,018 1 one shows the curve &: geodesic defined for al + and parameterized by ae length. Such a geodesic is obtained on every ty emanating from the origin 0 by syeametry. Thos B,(0, 0 =B,a(0, 2 sith Br/B cosh By ul = Bald, 2 si B1/B cosh Bo Which iso compact se, 80M is complete. ee Bat «)~ [mi Aytnyp) Sn wheve dy i he distance function i 7). Note that the imipping & whon genetalized to all rays in Ro, exhibits explicity the txponetal nap of onto M (se section 9.3). For K>0, let =, ed let, = 5,,/A? define @ Riemannian nettc op M ag above. The above computations show IT has constant Rienaanan cuvatere and Wf i tevally simply connected. Bat M is not complete since (0, 29/R) = R® is aot compact. Thus we eve an example of «cotfomal change of meric which changes « complete Romanian naniold into « nom-complete Rieaaaaieh men foie Section 6.8. Existence “The objective of this action is to show a paracompact connected ausdorf C™ manifold edmits @ Riemanaian metric. This is aecon plished by constructing a “partition ofthe unit function.” ‘The fane= tion 67071 isthe pelncipal tol which is used to show there are ‘many C™ fanctions on a C™ manifold LEMMA 1. 1f band. are coal numbers with 0< b 0, We outline « sequence of operations ‘which loads to the desired funetions, and we ilustste (and number) the graphs ofthese intermediate fanctions in Fig. Gl. Translate 1450 its graph moves (© ~ 5)/2 units tothe left this is m0. (1). Reflect the graph of (1) abost the yoaxi to obtain (2). Multiply (1) and (2) to obtain (3). Integrate (3) to obtain (8). Matipy (4) by @ scale factor to obtain (). Translate the graph of (5) to the tight to ‘obtain the desired function f// LEMMA 2. 11 b and c are ret numbers with 0< b . root. Let Pip) ipl) where £48 obtained fom lena 1.// LLENMA 3. FM is a Hausdorff C~ manifold and mi M, then there is coordina neighborhood U of mand @ C™ tumction fs Mm R such that fp) > 0 for pin U and fp) ~ 0 foe p notin U, rook. Let V be any coordinate neighborhood of m with coordinate map di V—» RY such that on) isthe origin 0, Choose real numbers band c with O00 Ug and f= Oo M~ Uy The function F = Saf Js a welldefined non-vanishing C™ function on If lace the covering We} locally nite. Finally lot 6 = f6/F.// THEOREM, 1M ix a connected Havadrlf C™ manifold then exch ofthe following tree properties implies the other two:raph of gi yn a) 3) 4) 5) raph of f: a (e-d)/2 me Picola! (eel katte stl » (@) Mis parscompact, (0) Madnite © Riemonnian metric, (©) Mix aecond-countablo (completely separable) rook, We show (a) implies (b) and give references forthe other Implications whose profs ave putely topological "Assuaing (3), apply lemme 4to obtain locally inte cover U) swith the partition of unity gg] On each coordinate neighborhood Uy define a local Riemsanisn metric tensor <, >, by demanding the coordinate map be ao isometry. ‘Then the tensor 4c, >, i global (C™ tensor on I that vanishes outside U,, At any point a My for X and ¥ in M,, let is constant 50, Let R and R! be two Lineor map valued show-rymmetiie 2 covariant tensors whose cortesponding K and K" ratisly prop- erties (a) teu (2) on p. 124. Show K's AYME R= Sl. ffs ¢ C™ stetly conformal map, show f, has no kemel and preserves angles. If is 2 complex analytic map, show ; Dy fal > = INBIEK,, Xo, where fe CC. [52 Let Matt be a stsctly confornal map with Seale func- sion F. Show fs (Riemannian) connexion preserving if F is53. 54, 58. 56, 57. 61. Notes on Dittoentia! Geometry fis an Isometsy, show f preserves the curvature tensor andthe Riemsnian cur With the standard hypothesis of section 2.3, show if Fis cone rexion preserving, then iis a sphere, « plane, or @ right eit ‘ular ejliader (see Hicks?) Let ibe s hypersurface in RP, le bo a C* unit normal, Yet be in CMU, RD, and let: Wow R by Lo) =p» CW, Show tha (,)X —X' «QW + efL1X) for X tangent to Me 38 an isomety for €> 0, show that Wis fat. Generalize the first two theoroms in section 6.5 to the case of kesebmanifold that is framed in an manifold for 1 < R™ | orthogonal, show f, = fin a naturel woy. Te youyey I on cethonomal base feld with duel base w ‘on, and W has constant Rlensnnian curvature K show the sssotiated curvature forms R,, Ke, ~,- WAC bas constant Riemannian curvature K and one defines & ete om Wd by ty Yh Uys Fado By K+ Tp ¥ py does M x A have constant curvature? Pie ol EL » 82. IE Xyanys, a coondinates ona hypersurface U in RY, Let 58/0 y= ey Xi by 2 iy Rp y= Soa Show tata, 34°18, (Weingarten eeuation), Rhy 3,07 (Oy)Pon — By by) Gaussrcurvature equation), nad ay a me, Os, 30.0 oh) (CodassiManiandk equation. 63, IFAs a Havedosff C™ manifold, A isa compact subset of M, B is open in M, and ACA, show there is an fin CUM, R) with (A) =0, 40 B)= 1, and 0 © (UN < 1, 7. Operators on Forms and Integration ‘This chapter dovelops more structure on a manifold. To conserve space, the teatment is fsiny bunt and many compatationsl details sare omitted. In the frst four sections M is « C™ trifold and A Js on open set in Section 7.1. Exterior derivative, For p> Owe define the exterior diferentiation map ds FAA) —+ £FO*%(A) shore F(A) is the set of C™ plore on A. If fn FCA) snd X is « C™ feld on 4, then a(x) = Xf For p> 1, letting w be a = 1) fom on A and X pa X, be C fields on A, then dW gg) = I AY WO york) 6 EycfDN Oty Xb K wonky whore Rindicstes tht the fed is omitted aon atgument Notice thatthe defiition is consistent withthe pet definition dn section 5.2. One proves that dis in (A) by using the che acterization theorem in Chapter 4 We aie the argument, That dt i linea with respect tation is tval. ‘That de i alterate dng cn be shown by suitching tno arguments ad examining the90 Hotes on Ditterentiat Geometry terms that don't immediatly change signs (this ust be done care fully. That di is Hinear over the sing FAA) then need only be checked in one ele. ‘Proposition. ‘The operator d has the following properties: dw ov) = dw 4 dy, where w and © are in FCA), Q) dw =) = be) =¥) + CPW de), for w in POCA) adv any form on A. (Any operator with this property is called an anti= derivation.) @ #adodna rok, Property (1) ollows vil fo te defion of and sation ectonaa. Forte eer ve popertiea weft eta Soca! opeewtation dr Let sy bow corte system Oe open at U, aad let ¥,= 0/0, The’ U, 26 Ivor © may © vance by #3 edo here te ums over allindices sch tat 129-2 aha, 0, let (Cy) KyrXgas) Proposition. The operator C hes the following properties ® Cp? Cw sv) Cge + Cav, ® Cgyy Cx rly © Cyl =2)-(Cyw) ~2 6 -1)P Cyr, for fin FO), K and ¥ in TH%A), w and vin POCA), and 2 in FAA), rook. Properties (1) through (4) ze eivial. Property (5) follows by induction on p, and itis sullicient to prove it whon w is product ofp Lorne by the loeslrepreseattion of fors.// ‘The operator Cy can be defined on coveriant tensors aad mixed tensors in an obvious way (with only (2), (3), ané (8) valid ia general), sd one can let C be zero on pute conrsvatiant tensors. Properties {@) and (4) indicate Cis a tensor map (eh satiderivetion valued Ie form of degree ~ 1 on FLAY),92 Notes on Differential Geometry ‘Thove is another fom of *contaction® induced by the natural iden ton of tensors of type 1, 1 ad linear naps, Let W be an din St 1S) $0 define tt 10, ¥ ty iW, and si vector space over R. For > 0, #>0,1 THA) = T= 19300) by taking @ is To KyyoorX yy in and letting DD yoy Kyo, 3h where Z yoyZ, 18 a base OfF and 2m the dul base of W*. One checks easily that tris well-defined independeatly of the patie Jar base ured. HQ in T!¥H), let r1-"@ = tr, The above operator induces an operator trl: T!M(A) — T'™1e8='(A) for an open set A ia au Section 7.2 Let X be a C™ vector feld onthe open set A. An operator Lig, called the Lie derivative via X, which maps T’(A) into itself Aelined as follows: (a) i fin FA, Lt = Xb; (if ¥ in TAY, LY = (8, VU (if w in TAY (L gw ¥) = Moe) ~ wl lt, YDy and Dit in THY yoy 9 TOAD, 8d Piygh, bo TMA), then £8 is defined by solving for iti the equation Lie devivative. DL lO yo ae Topol Mm gw ea) + lL gw ywayen¥) Bir nv La We call Lg a complete derivation because ofthe property (@), and note all teas in (3) are well-defined by (2), (), and (©) except the 1.0 term (indeed, (©) it “defined” by (4). One shows L.y9 is ten Sorby checking th linearity over FMA). Proposition. The operator L has the following properties: (Ly preserves forms, byes 2b yw ely chap. 7 Operators on Forms and Integration 9 @ Ly ev- ty vaw@ by Lyle =P) = (ya) ~B +a ~LxB where w and ate tensors of the same type, ¥ is any tensor, and a td fare forms. rool, An exercise (for (4) use Lglla @ 8)"1=[L x\a® BI") ‘There is @ more geometric definition of the Lie derivative Ly on ‘covariant tensors which we now discuss, Suppose the vector field 1X is defined and C™ on all of M. For each = in M let £0 be the integral curve of X (zection 1.5) through m with (,(0) em we know 1, defined for in @ neighborhood of zero, bet euppose each, is etined forall t and R. Then foreach tin R we could define & map Fs MM by Fm) ~£(0, withthe propertion F, oP, = Fy, and FE Wc R moi by Fl, 1) = Fim) would be C™ (fom the fact that X was C~ and the C™ dependence of glutions of erdinary dferatial equations upon initial conditions). Bach F, would be a diffe, since (Yt Fond Fy i6 tho identity map. A'nop F withthe above ‘operties is called @ parameter group of diferstiable transforma ons of Mand X is called its infinitesimal generator In general f, 1s not defined for allt, but one does obtain a loos! Irparamote group of local transformations iw neighborhood ofeach 1 in; Le for each m in M there is acighborhood U of my real amber b> 0, and a map F: U x, 8) if such that (1) F is C=, @) fort (-b, B), FP: U— PU) i 9 dill, (3) fr ty 5; and 4's in @b, B), F, oF, = F.,y and (8) for fixed p in U, 10) F(p) is an integral curve of 4. For'nore details see Palsis! "8H ? gad Nomiza! ©.5. LEMMA I+ Lot ¥ bo 9 C* fics in = neighborhood of min M. We chooee U and Bin the proceding paragraph to be sufliciently small 0 the image of F is contained in the domain of Y.. Then U6 Wy = His (PY pny) Palt rool, See Nomizu)(p 8.9A Notes on Differential Geometry Assuming lemma I, which gives us another geometric iterpetse ‘on ofthe bracket, itis tial to show the following lenma LEMMA 2. Lot w be 0 C™ plc at m Then (yn, = lig UP pe), it where Cp peel) =H pay AY goon MY The following is usefl relation between d, Ly, and Cy. Proposition, IEX is a C™ field on A, then Ly «doCy + Cgod when applied to C~ forms 00 A. rool, We verify this equality on funetion (forms) and Hors. ‘This is sufficient to prove the proposition, since locally form ix sum of products of functions and I-forms, andthe operators which we ‘equate above are both derivations; hence thee value on aay for is determined ty the valves on fictions and forms. For fin FMA) dC lf) 4 Cyl) 0+ dik) = X= Lyk For w ia BMAD (Cg 5 Cyaw XY) ~ Yih) « deh, ¥) = Pui) + Rw(P) — Yu) — wilt, YS yk Section 7-4. General covariant detivative, Let D be a connexion on My and let X be & C™ field onthe open set A. An operator Dy, called the covarient derivative vie X, which ‘maps T*(A) ito itself, i defined by using the recipe for dining Ly- The definition of D, proceeds exactly a the definition for Ly excopt for (8), and if¥ in TMA), DY is piven by the connexion D (soe section 5. When Dy, is substituted for Lg inthe first proposition of the pre~ vious section, oe obtsins valid properties for Dy ‘An operator &, ealled the general covariant derivative operator, which maps 77*(A) into 77*°%(A) is induced by D. If Bis in T*(A), Hype, te in TMA), 20d FoF gy ate in TMA), then ‘Chop. 7 Operators on Forms and Integration 95 DAO pomp Kye Dy Oe rnatty VypeaY ghe That (AG) is a tensor is left as problem. If @and 3 are tensors of, the same type, then (+ 6) AD + a, but 48 not tensor (eve problen 64). WA Sispand 1<) <4, then outa iehtod on TA), [An opertor div, called the divergence, which maps TMA) into Tr), for > 0 and #2 0, is defined by div =e"! oA. We site div 0 = t(.A0), where we assume the trace is taken on the last ‘covariant slot andthe Last contravariant slot. A tensor ix conserva liveitdiv 0-0. ‘The Riemann-Christolfel curvature tensor of type 1,3 i the tensor Kin THA) defined by © Kew, X,¥, 2) = wiRUY, 298) forw in TOMA) and X, ¥, and Z in TA), The second Bisnchi entity isthe equation (ARM, X,Y, 2, WY CARY KW, Ys 2) CARY, 2, Wyo » which is valid {Dis symmetric, and i is proved by noting the ex- pression @ —DyRUY, zyx) ~ RUZ, LY, WK ~ RUF, Z¥D,%9, when written on thee lines, permuting W, YZ cyclically from line to line, and then adding the thee lines, yields 2010, ‘The Ricci tensor is the 2-covarant tensor ©) Riel, ¥) = Ge"e2RYCK, ¥) = eet AX, ¥)9% Notes on Dilferentit Geometry (and this isthe negative ofthe Yclarsical™ Rico! tensor). Notice (PORK, ¥)= R(X, Y), The Ricci curvatre of x vector Xia the rumber Ric, X) (end this agrees with the “classical” Rice curv ture). IED is symmetsic, the fist Bianchi identity implies 10) Rie (X,Y) = Rie (F, XY 4 RUG, 1D is Riemannian, then RU, Y) is skew-symmetric by (6) in sec tion 6.2, 20 Ric is aymmetii. Hence there exists a seladjoint Linear map R?, called the Ricci map, dened on each M, with Rie (X, ¥) = ; indeed a Re) = Sn ROY, 2,2, for an orthonormal base Zy..2,. By (I), RE ia C™. The scalar carvature S(m) at each m it W is"detied by Sim) ~t(R,., ‘ Geni.) Riemannian metic induces many operations called raise ing’ and “lowering” of indices which we nou explain, The non-singt- lar metric tensor induces » nonsingular Linear mop G of W., oato MS for each m, La, AFX in Mg, then G(XKY) ~ Wi cise Lejce 1, and Bis in T define OH! a GOW, way Ky = Ho Similarly, define Gils Te Sigeeland bgjce by taking the form inthe i covariant slot (ofthe new tensor); aply™ ing G,, and inserting i into the © contravatioat slt (ofthe old tens ‘Thus ‘G1 -G on Tt, andthe I,letensorR associated with R* is siven by B= G41 Ric (where Rew, X)= m(REX)). ILL is ja C™W, Rh the drodiont fold of fis the field pred I G,(d) and the Laplacien of Fis the function del = div(rad 1 (Gometimes the notation del f= Af is used) “The operators Gand Gi: commute with A when possible, ie, (13) MOH Glo Men THEI Sa Land cise. 7 Operators on Ferme and Incegration ” Ji) MoGEF= GH oA on PH pall [As an example ofthe use of these operations we prove that as) AS 24ivR, rhich fs used in general relativity, Let Z juny, be an orthonoteal base of M, ond Wyre, the dual base, The second Bianchi identity implies S,[AKUWy Zy Zp Zp H+ MKWy Zip % Zp Z,) + Klwy Zy Zp My ZN ‘The frst ter of the sum gives ( AS)(X), while the other two each give “(div BYR). For (ast) = Care 86, ea = FG, HANAN) = 3AM, 2p Zp Ze Nh (div BY) = rb, er t2K CH) 2-5, ARwy Zp My Zp Zp and (Akon 2p Zp KZ) = (AGM, 2p 2, X,Z,) (0™KNE, X, 2, 2p 2) = ASO RIL, 2, X, 2, 2) (AG™RNE, 2, Zp MZ) = Kew BI, by (©) and (') in section 6.2.9B Hotes on Dittersnist Geometry Section 7.5. Integration of forms end Stokes" theorem. (Ope integrates prfoens over prckains, of singular p-chains, which we now define, Let I =[a in Re: 0< a, <1) denote the nit praquare for p >0, and 1 [0 in Rl, A.C™ pecube'o on Mis an Mrvalued C= faction defined on an open aeiphborhood ofthe unit square [in RP. A real C™ pechain cis a finite formal linear combination of C~ ‘cubes with real cofficents, thus © = 1,0; + f,0y% 0, where in R end 0, ae C™ pecabes. The aet C(t, of ail tat C™ p= choins is on abelian group (actually an Renodule) where one defines slditon by adding the coefficients of correspoeding pcubes ‘There ate fancies ways of defining C,(W, R). Lat Q, be the set of C™ pubes on M, Then C,(, R) ix isomorphic to set of ell fance tions mapping Q, into R which are zero except on finite number of elements, and the addition and scalar multiplication structure ov this function space ix obvious. Similarly, oe coald define C,(W, 2), the set of intogeal C™ prchains or C™ p-chains over the intern, ‘Then Ci R)=R Cyt. Z). More generally one could define C™ pe chains over say fing A with en identity element, and then by using the tensor product obtain the Armadale of C™ p-chsins on M over aay Armodule. There are conesponding goups obtained from C’pechains for any integer > 0. These groups are fundamental objects ofthe ‘cubical singular homology theory for M and ave studied in algebraic topology, (see Bilenberg and Steeazed). Because of ou differential seometey bias, me restrict ourselves to real C™ pchains, and let ©, =, Bi. ‘The'support of a preube ois the eet ol = oll, the image of I” under a. The support of a prchainc isthe set le| =U] foro, in where we say 0, inf the coefficient of o, is noneter, iv. ‘topting the functional viewpoint o(,) 40 iff, inc ‘To define the Boundary map @: C,.C,_,, datine maps a! and af fom 19" into 1? for} = Laon by 5) aft, tga = fai fpentyag) howe «= Lor If ein Q,, define dy and call the (9 ~ D-cubes @ oa} and oo g° to-all of C, by demanding it be liner, i 1)"4oog! e008), faces of. We extend @ Ale, +0) de + Ay Chop. 7 Operators on Forms and Integration 9 and Are) = 186 for rin R. A straightforward computation shows 0? = For p >0, let abe @ C™ cube, let w be poem, and let, 1, be the naturel coordinate function on RP. Since ow is «poem ef @ neighborhood of 1°, we may define & C~ function f on 1? by OW =f, ~ diya diy. Then 0 Lye sown st where the intogral on the sight isthe standard Riemann integral of f over 1? developed in advanced cslculus. Ie = Sry, ie a pckai, then J w= Zfe,fw; thus for fixed, the Integral over wis an Re homomorphism of C, into R, Since ats linea, it is trivial thet Si 02) = fos Sn, For forms wand phan & For p~0, et be a fanction oa M and 0, the Oacube with J (0) = my then ff fim) ~ ¢34(0), and we extend the integral off over sy real Ocha tobe Linear (as extended above). Let CP = Homg(C, R), which isthe R-module ofall Relies ho= rmomorphisms of C, into R. The set CP is called the module of rea (© peochains off. The adjoint 8 ofthe boundary operator 3 is called the coboundary operator and is detined by 5 fle) = Mae) fora prcochain fand a (p + Yechain c, Thus 6: CF + CP"? and 37-0 We define the Stokes" map S: FAM) ~» CP which aps p-forns on M into C* peochains on M by [Siw)lle} = fm, fore in C,». ‘The fo lowing theorem shows the Slokes’ map commutes with the differenti coboundary operator, ie, Sed = eS. STOKES’ THEOREM, Let w be a C™ plorm and o be a “Ip + Dheube, then (fae fw root, Define C™ functions @ yyy, on JP! by oly = 6a cu, = Bima Bigy ys Thea es100 Hotes on Ditlerenial Geomety 3, pe dy), so ote) = Spt ay, tpt)? Be pay one SEEN Se Maynoengy He fg f= | HCA ME SEH decd yg] = EHDY bo, ca) =2, 009% \whete we use Fubini'stheonem and inograt fits with respect toi" coordinate to obtain the lest equality For the other side we must compute f {Der L Notice (ef), = la) = ds, 6)» hay 0, or according 25 i, respectively. Thus leew = (oj oa hty-dy, and plat? ot for 5 or END cue DY 6,00! 9,00 ICDS a, oa} =, 008 ‘hich proves the desied equality.// We reas that Stokes” theorem is sinply genetalized“Tundamese tel theorem of cleus." Lat fs Mit boa Go mon ete ee form on Mand oa peu in My then iti val to show fw {ft which is essentially the classical substitution rule that deals ‘ith the behavior of intograls with respect to mappings, The Stokes’ mep induces @ map at the cohomology level that yields 8 algebre-isosorphism of tho dfferntial cohomology groupe of & ‘manifold with the real singular cohomology groups. This fact is called (he de Rham theotes (see A. Weil, and problem 21), (Chap. 7 Operators on Forms end lategeation 101 Section 7.6. Integration in # Rismsnion menilotd. ‘Let M be @ Riemannian manifold, let o be @ C™ curve in M, and Jet be a real valued C™ function on the imege of 0, ine, let oo ‘be C. Consider a “piece” of o, which we assume to be parancter= ized by are length on the interval [a,b], and define 0) fur afe [Todt where al, 8] denotes the restriction of o to the interval (4, 8 Call the intogral just defined the integeal of {over o restricted to (a, bh tnd when the interval is understood, we write simply J itis e (© real valued function { defined on a broken C™ curve o, we define {10 be the sum of the intogras off over the finite number of C= sub-curves determining a Notice that by assuming o patanetrizéd ‘by ate Length we are intograting over oriented or directed curves Wo wish to integrate real valued C™ functions over other subsets of, ond in some cases over W itelf. This coalé be accomplished by using the Riemannian metic to define a measure on M, bat for ou purposes we need not be so general, First we define onenteble ‘sanifolds and then utilize the theory developed above for inteprte 1 forms over chains, ‘An n-imensional manifold W is orientable if there isa nonevanishe ing C™ mforn w on M. hea M is orientable and we have selected Ws, we say M is oriented (by w) end w ix an orientation of M. IF is ‘oricated ty w, then an ordered base ue, of te postvely oriented if wy = bweyan.a¥, where 5S 0 ad w, are the Tfores dual to 0), We say M is nonvoriontable when M is aot orientable. If te stented and e,...8, @ positively oriented base off, then one votfies easily that a base fu, of Mis positively oriented if and only if det (by) > 0 where 3,6, For example, Ris orientable, and we orient it by choosing t,o nd, where u, aro the natural coordinate fanetions. It is ‘topological result that any complete (or closed) hypersuifece in Ro is orentbie, Let and W' be oriented amanifolds, A noncinguler C™ map f Of Minto Mi orientation preserving ifs maps « positively tented02 Wotes on Dillerontial Geometry ‘bate onto 4 positively oriented base ‘Let be an orented Riemannian r-nenifold, For min M lt rt, be positively oriented rtkoaornal base of M, with dual bse wnt. Define the form ¥ by Vy = Wy~ne=Wye The form vis ¢ welidelined (independent of the particular bare) C™ r-form on called the volume elemant 'A major problem now confronts us: the problem of *riangulating™ ox cubulating” a manifold. Tale ie «theory for breaking the mani- fold into “nice pieces” over which one can intogfate functions. For this purpose we define fundamental chains. Let Int(A) denote the Interior of «eet 4 [etl be an oiented ©™ nomanifold. A fundamental n-chaiafx.M is a chain ©~ o,1..40, such tht: (1) each 0, is an mcube that is an orientation preserving dffeo onto its imager and (2) Int (lo) Int (ji) 18 empty for i 2 - Figuze 71 schematic diagram of «fundamental 2-chain (withthe inoges of the faces of the eanoaics 2ecube numbered). EM i on oriented Riemsnnisn aemanifld,c is an achein, and f is @ C* real valued function whose donein contains |, then define Om fief where v isthe volume element on Mt. Let a subset A of M be fandor ‘mental Af there existe fundamental n-chaine with [o| = A. Notice © fundamental set is compact. Chop. 7 Operators on Forms and Intogration 103 Proposition. If ¢ and + te two fundamental r-chains with fe = Irie Ar and fis C™ function whose domain contains A, then fte= fhe. Thus define f= ff Prool, (King Lee). Lat e~ 0, Atroughout this proof let 1 Zeand 12) sa IA, ~ [6] Mlyie et By, = (o)"MA,)) and C,, =)" 4A). ‘Then yj? 20, isa diffeo of B, onto C,, and {6 fo, (7,977 78 = fe, OF AMI = fe, ipo Hence, 66 = Ey6 f= 3 if, (0% - Bf, 1% Wis a compact oriented manifold then M is a fundamental set (this is han, see Cairas). Thus if is & compact exleated Reman Jan manifold and i= a C™ weal valved function on M, then ff is welle defined, To handle the non-compact case, define the suppor of a function {9 be the set S, that isthe closure ofthe set [p in Ms ip) £0). Since any compact set of W is contained in a fundamental set (a nontrivial mast), HM is oriented and Riemannian, is C™ Wwth compact support, and S, Cundameatal set 4, then fl = fF is well-defined (independent of A). ‘The arse, volume, ot measare (depending on the appropriate dix mension) of «fundamental set A isthe number {qf, where f= Lon M. Fora deeper study of integration theory on manifolds soe the book of Whitney? Problems, Let M be 4 C™ amanifold end lot U be an open subset of 64, IC and ¥ ate in 72-5), F and g in C™W, R, and w in TOD, show Lg = w(Nldl + fe) Lig = ALeg¥)~ aA Lyd = leggy and Alfe) = Aw + 9 8 af. Thus L and Aare 65, 16X is a C* vector field on Uy 2 U; Zynga base of Mg ith dual base W yu Myy SOW (BWV X), = Ey (Dy XD ‘Show that the divergence of « C™ field on R? agrees withthe67. 68. 6. 7. nh . Let be a oviented Riemannian 2-enifold, Notes on Ditfoentiat Geometry sdvanced caleulus definition, 1 Let A be in TIN), Let Z yaa, be a C™ base fied on U dnd ftw ys, Bethe dual base on U. Show De) = 3, (D,Z,, and SAD yw, Z,) + Aly Dy) = Let M be Riomannien, let XyyunXq ye T bo an orthonormal ‘base, and let P, be the plane section spanned by X, and T: Show Rie (7,7) = 315K). Prove formulas (5), (7), (13), snd (14) ID has zero torsion, show di(X, ¥) ~ Dywh¥) ~ (Dw) Wife Riemonnian and G(X, ¥) = ch ¥>, show thet a cone ‘nexion Dia metsc proserving iff AG - 0. Given abitary A tn TG) spd Bin TQM) with ACK, ¥, Z)~ AU, X, Z) end Blw, X, ¥)~-B(w, ¥, X) forall», , ¥, Z, show there exits ‘unique connexion D on M with AG- A and Ble, X, ¥) = Mer, YD. (Poincare temina) Show every closed pform on R® is exact for p > Os follows: for bin R let gy: R"—+ RO“ by EAtyrmnty) = rv Bh LEE ROPE BY ME vrtge gd = pestis fovstav-nfsa stg 1&UT ~ 3/8 pond frp > O, dling the Tinear mop K2 FPR") -» FP"AR®) by K(w) = fi6,)* Cy» F4w)ab, and show GK « Kd equals the identity map on FAR") i ois an oriented C* cuwve in M with uit tangent 7, Tot 7, W be an orthonormal oriented base slong o and define the signed geodesic curve tore of to be the C™ fonction 8 with DT ~ BW ono. Mf 2, Wis an orieatedorthonoal parellel base eld along o end T =icos 0 « (sin OW, show b = di/ds = TOon a. Ux, 7 8, tn oriented orthogonal coordinate system on U in My, let E = and O= c¥, Y>. IE, and by denote the gedesic curvature along the xcoorinate and y-coordinate curves, r= spectively, show b, = 0/28 yGX3E/3y1,b, = (1/2GVE) (@6/ae) and K = (EG), EV ~ Ab 6/30. Show the | ‘yreuve are geodesics (with 738 paranetes) iff G is constet. Chap. 8 Gause-Bonnet Thoory and Rigicity 105 73, UA is Riemannien, (fy U) is #coosdinate pair, x, = 4, 4 6 ,,= My Xp where X, ~ 0/aty d = dt, fis in CU, RD, ad A ies fndanental et with A CU, BOW f= fe Slo ers eee 74, Let M be s surface in R? with sphere map y. Por min WM let Ae) bo the aren of B¢m, 1), the ball about oof radius rand let (AK2) be the ares of (Bm, 0). Show Kia) = Han [4,(2)/A0] aor 0, 8. Govss-Bonnet Theory and Rigidity In this chapter, M will denote @ connected oviented Riemannian manifold. Section 8.1. Gauss-Bannet formula. In this section, let = 2, let A be a fundamental set in My and let © bo a fardamental 2chsin with |e) - A. The oriented curve y = 3¢ ‘a called the bounding curve of A. A vertex of ¢ is a plat in M that is the image of a vertex in [? wader @ -cube in c. A face of € is the suppott of @ 2cube in c. An ef of is the face ofa Heabe in do for some 2ucube a in c. A Boundary edge of isan odge that is in y {A comer point of is a vertex of ¢ belonging to exactly two boundary ges. Ate comer poiat p of y , let T,(p) (the “tangeat int) end T.(@) (the “tangent out”) be the unit tangents at p ofthe Leubes in yi defined by the ovietation, going into” and “out from™ p, respec tively. The exteior corner angle ap) is the angle such that cos alp) = Tiel Pele) and Ocace or nacD according az T, Ty i m positively oc ngetively oxented bese, If, =, then a= 0, and if Ty ~-T, thes a= =r (See Fig. 8.1). 2% ig. 8.1 Comer Ansios